Multilayered microelectrode array for monitoring electrophysiological signals of 3d neural networks in cerebral organoid

Abstract

Microelectrode arrays (MEAs) are widely used platforms for monitoring neuronal activity in biological systems. The advent of brain organoid technology—three-dimensional (3D) neural tissue models derived from human stem cells—has opened unprecedented opportunities to study network connectivity and developmental processes of brain-like tissues in vitro. However, most commercial MEAs are restricted to planar, two-dimensional configurations, limiting recordings to a single surface and failing to capture the multilayered architecture of organoids. To overcome this limitation, we develop a multilayered MEA (MLMEA) capable of interfacing with multiple depths within brain organoids. The MLMEA features customizable interlayer spacing achieved through the stacking of mesh-type electrode layers separated by mechanically compliant PDMS spacers. This design allows simultaneous, spatially resolved electrophysiological recordings across layers along the vertical axis, enabling detailed characterization of interlayer connectivity and network dynamics. Noninvasive integration of the MLMEA with cerebral organoid sustained for over four weeks demonstrates the stable, increasing neural activities with layer-specific electrophysiological metrics. Our findings demonstrate that MLMEA can reliably capture heterogeneous, depth-dependent neuronal activity, establishing it as a powerful tool for advancing 3D organoid-based neuroscience research and high-content drug screening.

Introduction

Stem cell-derived brain organoids have emerged as powerful in vitro models for studying human organ development and function. These three-dimensional (3D) cultures recapitulate key features of brain tissue, including complex cellular architectures, diverse neurogenic processes, and functional neural networks. Consequently, they have become invaluable for investigating neurodevelopmental mechanisms and disease pathology^1,2. Their applications span multiple domains: modeling neurodevelopmental and neurodegenerative disorders, screening drugs and testing toxicity for personalized medicine, and even serving as biological platforms for neuromorphic computing and artificial intelligence hardware^3–5. Importantly, organoid systems provide a human-relevant alternative to animal models, addressing long-standing ethical concerns.

A variety of neurophysiological techniques have been used to probe organoid function⁶. Patch-clamp recordings^7,8 enable precise single-neuron measurements but typically require tissue slicing or dissociation, which can damage cells and disrupt local circuitry. Fluorescence-based techniques^9,10 such as calcium imaging and three-photon microscopy, offer high-resolution insights into neural dynamics but are limited by penetration depth and often require genetic modification or dye loading. Magnetic resonance imaging has also been explored¹¹; however, challenges remain, particularly in achieving sufficient spatial resolution and minimizing artifacts caused by the small size and complex architecture of the tissue.

Among these approaches, microelectrode arrays (MEAs) have become particularly valuable for organoid research^5,12,13. They offer high spatiotemporal resolution, relatively simple fabrication, and simultaneous recordings from multiple electrodes, thereby enabling population-level analyses. They also permit noninvasive, long-term monitoring combined with the ability to externally stimulate neurons¹⁴. However, conventional two-dimensional (2D) MEAs face notable limitations when applied to 3D organoid cultures. Their planar configuration confines recordings to the organoid’s surface, failing to capture neural activity within the 3D structure. As a result, many studies have reported slicing organoids to enable electrophysiological measurement of internal neural signals^15,16. In addition, the mechanical stiffness of these MEA interfaces may also interfere with organoid cytoarchitecture development¹⁷, potentially disrupting the essential spatial relationships for proper neural network formation.

To address these challenges, 3D MEA platforms have been developed that enable more physiologically relevant electrophysiological analyses¹⁸. Various 3D MEAs have been introduced, including implantable^19–22, foldable^23–25, and net-type MEAs^26–28. Implantable MEAs use needle-like electrodes of varying heights to detect signals from within the organoid interior. Foldable MEAs employ self-folding leaflets that envelop organoids of various sizes (400–1500 μm), increasing electrode coverage around their surface. Net-type or stretchable MEAs provide conformal, perforated electrodes that gently support organoids while improving nutrient and medium transport. Collectively, these systems have revealed complex, spatiotemporally resolved neural network dynamics that are difficult to measure with 2D systems. However, achieving a truly depth-resolved 3D neural interface remains challenging. Most current platforms are still based on single-layer electrode architectures, which limit volumetric coverage and preclude multilayer mapping of neural activity throughout the organoid. As a promising alternative, highly stretchable mesh electronics can be embedded within growing organoids to achieve seamless, conformal 3D tissue integration^29,30. However, the continuous structural deformation of the tissue inherently alters the spatial coordinates of the electrodes over time, making it challenging to maintain the physical distance between recording sites, particularly regarding the z-directional customizability required for depth-resolved mapping. Recent work on 3D-printed scaffold-type MEAs interfacing with a neuron-hydrogel mixture³¹ shows promise for enabling more sophisticated MEA-tissue interfaces and functional analyses in 3D space. However, 3D-printed MEAs typically require relatively thick and rigid materials for structural integrity, complicating integration with soft organoid tissues. Additionally, photoinitiators and chemical additives commonly used in 3D printing can be cytotoxic or elicit undesired biological responses, making them less suitable for long-term organoid culture^32,33.

In this work, we present a robust, reproducible, and precisely controlled fabrication and assembly process for a multilayered MEA (MLMEA) platform designed to overcome these challenges. The MLMEA incorporates multiple stacked electrode layers in a customizable 3D configuration, providing porous, flexible mesh-type electrodes that support organoid growth and allow nutrient exchange. The device also demonstrates excellent electrochemical properties, enabling high-fidelity, depth-resolved signal acquisition throughout intact organoid. As a proof of concept, we cultured cerebral organoids on the MLMEA and recorded neural activity across distinct layers over time, allowing investigation of functional connectivity and network dynamics during organoid maturation. This MLMEA represents a significant advancement for multilayered, high-resolution studies of neural circuit function in 3D brain organoid models.

Results

Design concept and key features of the MLMEA

Figure 1 illustrates the foundational design and operating principle of the MLMEA platform, developed for noninvasive electrophysiological monitoring of 3D brain organoids. The flexible, mesh-type MEA provides a supportive scaffold that promotes organoid growth in three dimensions, while micropores (150 μm × 150 μm) allow efficient nutrient and oxygen diffusion (Fig. 1a). Additional design specifications, fabrication procedures, and experimental protocols are provided in the Materials and methods section (Fig. S1–S3).

Fig. 1. Design concept and key features of the multilayered microelectrode array (MLMEA) platform.

a Schematic of a cerebral organoid cultured on the MLMEA. After placement on the upper MEA layer, the organoid grows in 3D space, supported by the porous structure of the MEA. b Exploded view of the MLMEA showing its modular assembly. c Optical image of the fabricated MLMEA after organoid placement. d Optical image highlighting organoid placement and mesh-patterned signal recording sites. The inset (top right) shows an optical image of the separated MEA layers, emphasizing the interlayer spacing

Upon placement, facilitated by the poly-L-ornithine and laminin coating, the organoid adheres to the mesh surface and extends neurites along the interconnects (Fig. S4). This process reflects the intrinsic migratory behavior of neural progenitor-rich tissue, which responds to topographical cues such as aligned microstructures^34,35. The 150 μm × 150 μm micropores provide ample space (significantly larger than individual neuronal somas of 10–20 μm³⁶) for cellular intercalation, while volumetric expansion and gravity-induced settling progressively engage multiple MEA layers³⁷. Consequently, this expansion-mediated colonization enables continuous, noninvasive recording through seamless integration with the porous multilayer mesh.

The exploded view highlights the modular assembly, which consists of two mesh MEAs, PDMS spacers that define the interlayer displacement (ILD), culture wells, and a glass substrate (Fig. 1b). Additional layers can be incorporated by stacking extra spacers and mesh MEAs during assembly. Figure 1c shows an optical image of the completed MLMEA device following cerebral organoid placement. The assembled MLMEA was maintained in a petri dish inside an incubator, and neural activity was monitored by connecting amplifiers to the external connectors on the device. Figure 1d provides a closer optical image of the organoid adhesion area and the mesh-patterned signal recording sites, with the inset showing the physical separation between the two MEA layers. For the proof-of-concept demonstration, each MEA layer was designed with 32 recording electrodes (30 μm diameter, 200 μm inter-electrode distance), 50 μm-wide interconnects, 150 μm × 150 μm pores, and ground and reference electrodes placed near the culture well (2 mm² area). A supporting bridge structure (600–700 μm width) was incorporated to mechanically stabilize the film and precisely maintain the interlayer spacing. To allow fluid exchange and minimize bubble formation between layers, large millimeter-scale openings (>1 mm × 1 mm) were included along the periphery of the bridge structures. Each MEA film is fabricated with a thickness of 10 μm—sufficiently thin to conform to the tissue yet mechanically robust for repeated handling and cleaning. Routine medium exchange via gentle pipetting during organoid culture on the MLMEA did not compromise electrode functionality or interlayer adhesion.

Customizability of the MLMEA

Figure 2 shows the customizable characteristics of the MLMEA platform. The ILD, a parameter controlling the vertical resolution, is precisely tuned using PDMS spacers of variable thickness (150–600 μm), as visualized by z-stack fluorescence imaging and optical images (Fig. 2a). Although minor increments (<15 μm) in ILD compared to the spacer thickness result from the applied PDMS adhesives during the integration process (Fig. S5), the differences were consistent and can be compensated by reducing the spacer thickness. Although the minimum ILD is highly dependent on the production of uniform PDMS spacers, the higher precision can also be achieved through adhesive-free bonding or surface-modification processes^38,39.

Fig. 2. Design customizability of the MLMEA platform.

a Schematic of interlayer displacement control using spacer components in a double-layer MEA. Representative z-stack fluorescence and optical images confirm precise control over vertical spacing. b Optical and fluorescence images demonstrating micron-level alignment accuracy between MEA layers. c Optical images of an MLMEA with four stacked layers and its integration with the printed circuit board, highlighting scalability. d Two bent MEAs with aligned structures, demonstrating compatibility with curved geometries. e Double-layer MEAs designed for multiple organoid placements, enabling high-throughput applications. Rhodamine B was used as a fluorescent dye to visualize the MEA structure

Optical and fluorescence images verify that the alignment error between stacked MEA layers is less than 10 μm, with consistent reproducibility across multiple devices (Fig. 2b). The design also supports multilayer stacking beyond two layers, as demonstrated by a four-layer MEA integrated with a printed circuit board (PCB) and connectors (Fig. 2c). This modular stacking capability allows for expanded monitoring depth, enabling finer analysis of layered dynamics. The platform further accommodates curved and folded MEA geometries (Fig. 2d). This specific configuration is presented as a geometric demonstration to highlight the mechanical flexibility of the polyimide platform, enabling future applications such as simulating cortical folding, examining axonal projection pathways across nonplanar surfaces, and modeling realistic tissue interfaces for fusing region-specific organoids or assembloids. While the electrochemical stability of flexible MEAs is well-established^23,25, deploying certain architecture for active biological recordings additionally requires rigorous functional validation and geometric optimization in further studies. The availability of multiple placement sites on a double-layer MEA (Fig. 2e) facilitates parallel electrophysiological recording of multiple organoids, supporting high-throughput screening experiments. Although this work focuses on double-layer configurations, each MEA layer is customizable in mesh pattern, electrode number, and geometric layout, allowing the MLMEA platform to be adapted to a wide range of experimental needs. A comparison with the key features of various 3D MEAs reported in the literature is presented in Table S1.

Characterization of the MLMEA

To reduce electrode impedance and enhance signal fidelity, all recording electrodes were electroplated with platinum black (PtB), as confirmed by impedance spectroscopy comparing bare gold electrodes with PtB-coated ones (Fig. S6). PtB deposition dramatically decreased the average impedance magnitude at 1 kHz from 1.26 MΩ (bare gold) to 18.1 kΩ (PtB plated, 120 s), thereby reducing thermal noise and improving the sensitivity for detecting weak neuronal signals. Phase-angle analysis revealed that PtB electrodes exhibited predominantly capacitive behavior, indicating improved electrode–electrolyte coupling and stable performance across physiologically relevant frequency bands. These electrochemical properties are consistent with previous reports on PtB-modified MEAs, which demonstrate enhanced signal fidelity and durability under long-term culture and repeated handling¹⁴.

Finite element method simulations were performed to evaluate the mechanical stability of the MLMEA under pressure applied at the recording site. Although a pressure of 1 kPa was used for simulation convenience, the actual mechanical load exerted by cerebral organoids under culture conditions—considering their mass (≈0.7 mg for a ≈ 1 mm organoid)—is significantly lower, typically only a few hundred pascals or less⁴⁰. Simulation results indicated that PDMS spacers between MEA layers effectively decreased the maximum principal strain and stress on the polyimide films and minimized z-directional deflection (Fig. S7). Specifically, the inclusion of spacers reduced the maximum principal strain from 0.195% to 0.174% and provided mechanical buffering that preserved electrode positioning and overall structural integrity during organoid growth and routine handling. Given that actual culture pressures are considerably lower than simulated loads, the mechanical impact on both the MLMEA and the organoid interface is expected to be even less pronounced, supporting the platform’s ability to provide gentle, compliant, and sustained interfacing with fragile 3D neural tissues. These results validate the design strategy for enhancing MLMEA durability, maintaining stable recording interfaces, and minimizing risks of mechanical failure or tissue damage during long-term electrophysiological experiments. Also, the mechanical buffering provided by the spacer ensures the stability of the interlayer spacing and minimizes the potential for electrode movement or sagging, which is essential for longitudinal tracking of the neural network.

Characterization of cerebral organoids

Figure 3 summarizes the cerebral organoid culture workflow and developmental progression. Figure 3a outlines the experimental timeline, beginning with human-induced pluripotent stem cell (hiPSC) culture, embryoid body (EB) formation, and neural induction, followed by a four-week maturation phase before placement onto the MLMEA for longitudinal electrophysiological recordings. Figure 3b shows immunofluorescence images of whole cerebral organoids after four weeks of culture, immediately before MEA placement, illustrating the spatial organization of neuronal populations at different maturation stages. CTIP2-positive neurons, characteristic of deep-layer pyramidal projection neurons, were broadly distributed, whereas SOX2-positive neural progenitors remained localized in distinct zones, supporting continued neurogenesis. In parallel, MAP2-positive neurons displayed well-developed dendritic arborization, indicating the establishment of a mature cytoarchitecture and functional network potential. Figure 3c provides high-resolution confocal images of these markers, emphasizing the interaction between proliferative progenitor zones and differentiated neuronal populations. The abundance of CTIP2-positive neurons is thought to support robust spontaneous activity and signaling across multiple depths, enabling reliable multilayer electrophysiological recordings on the MLMEA. Figure 3d highlights representative morphological changes at 7, 14, and 28 d after organoid placement on the double-layer MEA, capturing the shift from superficial surface growth to progressive infiltration into deeper mesh layers. Calcein-AM and DAPI staining confirmed preserved cell viability and intact nuclear morphology during this process, indicating that vertical expansion represents active tissue growth rather than cell loss or degeneration. Also, it substantiates the platform’s biocompatibility and minimized mechanical impact of the MLMEA on the cell condition while integration process. The vertical expansion enhances contact with electrodes across multiple layers, which is crucial for capturing neuronal activity throughout the organoid volume and reconstructing multilayer network dynamics.

Fig. 3. Characterization of the cerebral organoids.

a Schematic outlining the complete culture and measurement process from hiPSC maintenance to electrophysiological monitoring. b Representative immunofluorescence confocal image of a whole organoid showing the spatial distribution of neuronal markers (scale bar, 1 mm). c High-magnification confocal image of a specific region highlighting cellular organization and marker expression (scale bar, 100 μm). d Fluorescence images of organoids on the MEA at 7, 14, and 28 d in culture, illustrating progressive morphological changes and infiltration into the mesh layers

Spontaneous neural activity of organoids on the MLMEA

Figure 4 presents electrophysiological recordings of spontaneous neuronal activity in cerebral organoids cultured on the MLMEA platform over a four-week period. As shown in Fig. 4a, the organoids maintained a typical spherical morphology with predominant adherence to the upper MEA layer (first layer) at day 35 (7 d on MEA). By day 56 (28 d on MEA), the organoids exhibited significant volumetric thickening and clear infiltration into the lower mesh layers, indicating progressive tissue integration. These morphological changes were closely paralleled by the electrophysiological findings shown in Fig. 4b. Representative raw voltage traces from electrodes in both the first and second layers showed a gradual increase in spike frequency and amplitude with time in culture. Raster plots revealed that initially sparse and asynchronous firing patterns at 7 d on MEA transitioned into denser, more temporally synchronized bursting activity by 28 d on MEA. Consistently, mean firing rates increased across both layers, with a particularly pronounced rise in the lower layer (second layer), suggesting that network maturation and recruitment of deeper regions occur alongside morphological expansion. The detailed signal processing workflow and spike detection parameters are provided in the Experimental Section.

Fig. 4. Neuronal activity monitoring of organoids cultured on MLMEA.

a Representative optical images of cerebral organoids cultured on MLMEA for 7 d (top; day 35 overall) and 28 d (bottom; day 56 overall) after placement, showing progressive morphological development. b Electrophysiological activity recorded from the first and second MEA layers at 7 and 28 d after placement. For each time point, raw voltage traces (top), raster plots of detected spikes across channels (middle; blue: first layer, red: second layer), and binned firing rates (5 s bins, bottom) are shown. c Active channel ratio on the first and second MEA layers over 7, 14, 21, and 28 d in culture. Box plots display mean ± SE, whiskers represent SD, and individual data points are overlaid. Red dots indicate the mean for each day. d Firing rates measured from both layers at each time point. Box plots display SE (boxes) and SD (whiskers), with overlaid individual data points. Red dots indicate the overall average values for both layers combined. e Burst firing rates from both layers over time, presented with SE (boxes), SD (whiskers), overlaid individual data points, and overall average values (red dots). f Average signal-to-noise ratio (SNR) of detected spikes on both layers is shown as bar charts with individual data points. Whiskers represent SD, and the dashed line indicates an SNR of 4.5, used as a quality threshold. Each data point represents a 3 min recording. Statistical significance was determined using t tests; n = 6–8 measurements per group

The quantitative analyses shown in Fig. 4c–f further characterize the developmental dynamics. The fraction of active recording channels steadily increased over the four-week period, demonstrating progressive recruitment of neurons across multiple MEA layers (Fig. 4c). Importantly, the recorded signals maintained stable signal-to-noise ratio (SNR) levels, confirming the robustness of the platform and the reliability of long-term recordings (Fig. 4f). Both firing rates and burst firing rates displayed characteristic fluctuations after 14 d on MEA (Fig. 4d, e), which are attributed to the intrinsic activity cycles of developing neural networks—alternating between phases of heightened synaptic connectivity with coordinated bursting and subsequent periods of network remodeling and synaptic pruning⁴¹. It is important to note that these temporary fluctuations, specifically the observed non-linear change or slight decrease in mean firing rate at 28 d, do not suggest an adverse effect from the MLMEA. These evolving activity patterns are consistent with known features of neuronal maturation and synaptic plasticity^41,42, rather than measurement variability, underscoring the ability of the MLMEA platform to sensitively track the temporal evolution of electrophysiological states in 3D organoids.

Additionally, the quantitative analyses serve to substantiate the platform’s reliability against two major concerns regarding 3D integration. First, if the MLMEA had caused significant cell damage or hindered maturation due to toxicity or mechanical stress, the sustained increase in both the mean firing rate and burst activity, alongside the stable recruitment of active channels, would not have been observed. Second, the stable SNR throughout the four-week period indicates robust mechanical integrity. If the flexible mesh electrodes were to undergo significant deformation or sagging under media flow or gravity, the resultant mechanical drift would reduce the SNR and cause unstable signal acquisition. The observed stability in functional metrics confirms the robust mechanical performance and biological compatibility of the MLMEA, validating its suitability for longitudinal monitoring of neuronal populations during organoid development.

Analysis of 3D functional connectivity and network dynamics

Figure 5 demonstrates the capability of the MLMEA platform for depth-resolved 3D functional mapping of cerebral organoid neural networks through multilayer electrophysiological recordings. The schematic in Fig. 5a depicts an organoid cultured on a double-layer MEA device with an ILD of 260 μm, allowing simultaneous monitoring of neuronal activity across multiple depths within the tissue. The corresponding 3D functional connectivity maps revealed progressive development of neuronal networks during the culture period, characterized by increasing synchrony and strengthened interlayer communication (Fig. 5b). These observations are consistent with the maturation and integration of neuronal circuits within the organoids. To quantify synchrony, the spike-time tiling coefficient was applied, measuring temporal changes in spike timing and reflecting network plasticity and adaptive modulation. Layered synchronization of this type is important for recapitulating brain-like network dynamics in vitro, highlighting the MLMEA’s ability to perform depth-resolved electrophysiological measurements.

Fig. 5. Monitoring of neuronal network development in 3D space using the MLMEA.

a Schematic of the double-layered MEA with an optical image of the monitored organoid (inset, scale bar, 1 mm). b 3D functional connectivity map of an organoid cultured on the MEA. Connectivity was calculated using the spike-time tiling coefficient and visualized by line colors as indicated in the color bar. The size of the gray circles represents the relative mean firing rate at each electrode. c Schematic of the MLMEA with four layers and an optical image of the monitored organoid (inset, scale bar, 1 mm). d 3D functional connectivity map of an organoid cultured on the four-layered MEA, demonstrating enhanced spatial resolution. e Schematic of electrical stimulation-evoked response monitoring for network perturbation studies. f 3D functional connectivity map showing pre-stimulation spontaneous activity and stimulation-evoked activity. The yellow circle marks the stimulation site. g Correlation coefficient matrix (blue) and spike-time tiling coefficient matrix (orange) before and after current stimulation. Each grid cell indicates the synchronization intensity between pairs of the 61 detected spike templates, revealing enhanced network coordination following stimulation

Figure 5c shows the scalability of the platform with a four-layer MLMEA configuration featuring a reduced ILD of 160 μm. The associated functional connectivity map (Fig. 5d) confirms the platform’s ability to capture complex neural interactions spanning multiple layers with enhanced vertical resolution. Validation of tissue infiltration and electrode contact across multiple layers was supported by stable, robust multisite electrophysiological recordings. As described earlier, the mesh-type MEA supports 3D growth by enabling nutrient and oxygen diffusion through micropores in the mesh structure⁴³. As observed in Fig. 5b, d, the spatial shift of functional connectivity, along with the changed positions of the firing regions, reflects the natural, non-stationary developmental progression of the 3D network driven by neuronal maturation and synaptic plasticity. These spatiotemporal 3D mapping results show that the neural network is undergoing continuous, dynamic reorganization, highlighting the active pruning and re-routing of synaptic connections in 3D space.

Electrical stimulation experiments (Fig. 5e) further demonstrate the platform’s ability to probe network responsiveness. A cerebral organoid cultured for 31 d on MEA was subjected to single-pulse current stimulation (biphasic, 40 μA peak-to-peak amplitude, 1 ms duration). These parameters were selected from safe ranges validated in organoid MEA studies^3,44,45, followed by empirical screening (10–100 μA, 100 μs–1 ms) for each organoid to ensure effective activation. The MLMEA’s low impedance (18.1 kΩ at 1 kHz) ensured electrode stability without degradation, gas generation, or impedance drift. Figure 5f depicts the spatially resolved 3D functional connectivity maps showing pre-stimulation spontaneous activity and stimulation-evoked activity at the site indicated by the yellow circle. The stimulation evoked both localized and propagated responses, confirming the engagement of neural circuits. Temporal dynamics before and after stimulation were analyzed using correlation coefficient and spike-time tiling coefficient matrices (Fig. 5g). Both synchrony measures increased after the stimulation, indicating enhanced coordination and reorganization of the neural network in response to external inputs.

Discussion

In this study, we developed a 3D MEA platform designed for noninvasive, depth-resolved electrophysiological monitoring of human cerebral organoids. By integrating multiple flexible mesh-type electrode layers with spacers, the MLMEA overcomes the limitations of traditional planar and single-layer MEAs, enabling the simultaneous recording of neural activity at multiple depths within the organoid. The porous mesh architecture supports organoid growth by facilitating efficient nutrient and oxygen diffusion, thereby promoting robust tissue viability and integration. Electrical and morphological analyses confirmed progressive organoid maturation and vertical tissue infiltration into the MLMEA layers, which correlated with the expansion of electrophysiological activity across depths. Notably, the MLMEA platform captured complex neural dynamics characteristic of developing neural circuits. The neural networks showed a progressive transition from sparse, asynchronous activity at 7 days on MEA to synchronized bursting patterns after 21 days. This transition is reflected in increasing spike-time tiling coefficients across layers, which recapitulates the maturation trajectory observed in cerebral organoid development. 3D functional connectivity mapping further revealed evolving network motifs and interlayer synchrony during organoid development.

Our MLMEA platform addresses key challenges that persist in current 3D MEA designs (Table S1). Compared to penetrating needle-type MEAs, which achieve depth penetration but risk mechanical damage to fragile organoid tissues, the MLMEA employs a porous mesh architecture that minimizes invasiveness while maintaining structural support. Unlike foldable MEAs that wrap around organoid surfaces, our platform enables direct electrode contact across distinct layers with precisely controlled interlayer spacing (150–600 μm), offering superior depth resolution and customizability. The MLMEA also shows distinctive characteristics compared to net-type or stretchable MEAs by providing a stacked, modular design that facilitates parallel recordings from multiple depths simultaneously rather than single-layer surface monitoring. Critically, the mesh-type electrode geometry supports 3D tissue infiltration through its microporous structure (150 μm × 150 μm), promoting nutrient and oxygen diffusion without cytotoxic concerns. The selection of PtB electroplating was instrumental in achieving robust signal acquisition. By reducing electrode impedance from 1.26 MΩ to 18.1 kΩ at 1 kHz, PtB coating significantly enhanced SNR. This enhancement is critical for detecting weak neuronal signals in 3D recordings where electrode-tissue contact may be less intimate. PDMS spacers and polyimide electrodes provided mechanical compliance that reduced maximum principal strain from 0.195% to 0.174%, thereby minimizing mechanical perturbation of developing neural networks while maintaining precise interlayer spacing. This design philosophy—balancing electrochemical performance, mechanical compliance, and biological compatibility—distinguishes the MLMEA from alternative 3D approaches.

From a mechanobiological perspective, the physical constraints imposed by the polyimide mesh must be considered. While the ultra-thin, highly porous mesh architecture minimizes mechanical mismatch compared to solid substrates, it introduces physical boundaries absent in free-floating cultures. Immunostaining at 7 days on the MLMEA (Fig. S8 and Video S1) reveals preserved cytoarchitecture, SOX2+ progenitors, MAP2+ neurons, and Ki67+ proliferative cells comparable to controls, alongside progressive electrophysiological maturation. Nevertheless, these structural and functional readouts cannot exclude localized mechanically-induced stress or reactive gliosis at the tissue-mesh interface. Evaluating mechanosensitive gene expression, cytoskeletal remodeling, and glial stress profiles remains a key limitation and future direction.

While the MLMEA platform enables multilayer electrophysiological recordings with depth resolution, several limitations warrant acknowledgment. Precise 3D localization of individual electrodes within the organoid remains challenging, particularly after extended culture periods. Two factors contribute to this limitation. First, organoid growth patterns are inherently heterogeneous, resulting in uneven tissue distribution across electrode layers. Second, determining the precise 3D interface between organoid and electrodes requires complementary optical imaging. However, conventional imaging techniques such as confocal microscopy with z-stacking have practical limitations with restricted imaging depth and potential disruption to the stable culture environment. As a result, current 3D network analysis relies primarily on electrical signals without precise spatial context, meaning current connectivity analyses are functional and statistical rather than anatomically resolved. This lack of spatial information constrains our ability to correlate recorded neural activity with specific electrode locations within the organoid and derive detailed circuit-level mechanistic insights into organoid network architecture and connectivity.

Additionally, achieving identical positioning and uniform organoid coverage on the electrodes is a practical challenge. While initial manual positioning introduces minor spatial variations, the organoid’s autonomous migration and asymmetric expansion over long-term culture naturally lead to the off-centered coverage. However, the robust tissue-mesh integration ensures that these morphological variations do not compromise device reliability, yielding highly repeatable functional network maturation trends across multiple independent samples. In parallel, necrotic core formation is a recognized biological challenge, typically developing in spherical organoids exceeding 1.5 mm in diameter due to oxygen and nutrient diffusion limits⁴⁶. While optical imaging limits precluded precise quantitative histological assessment of core viability in organoids on the MLMEA, Fig. S9 demonstrates high viability (>90% live cells) throughout the MLMEA-integrated organoid, including the central regions (Video S2). Combined with the reliable detection of synchronized action potentials from inner layers at depths up to 640 μm, these results provide compelling functional evidence that the deep-layer neural populations remain viable and actively engaged in network dynamics.

Future optimizations are anticipated to address these limitations. These include the refinement of micropore size to enhance morphological guidance, the implementation of concave or bowl-shaped architectures to structurally constrain and center the organoids, and the exploration of alternative interfacing materials with improved biocompatibility. Furthermore, the integration of complementary imaging modalities combined with electrophysiological recordings will enable more depth-resolved spatial-functional characterization.

The modular and customizable design of the MLMEA enables diverse applications, from co-culture systems to assembloid models, and provides an advanced tool for dissecting the architecture and dynamics of human brain organoid neural networks. Beyond these applications, its multilayer configuration allows precise control over interlayer spacing and the integration of region-specific organoids into distinct layers, thereby creating unique opportunities to study inter-regional connectivity and emergent network dynamics. Future integration with optogenetic stimulation, chemical delivery, and automated high-throughput analysis will further expand its utility for neurodevelopmental research, disease modeling, and pharmacological screening. Overall, this work represents a significant advancement in bridging conventional 2D MEA systems with in vivo brain electrophysiology. The MLMEA offers a versatile platform to drive human brain organoid research toward previously inaccessible insights and translational applications.

Materials and methods

Fabrication of MLMEA platform

The design and fabrication of the MLMEA platform consisted of three primary components: mesh-type film MEAs, a PCB base, and the integration of PDMS (Sylgard 184, Dow Corning, USA) culture wells and spacers (Fig. S1). The mesh-type MEAs were fabricated using wafer-level microfabrication techniques. First, a 5 μm-thick polyimide layer (VTEC PI-1388) was spin-coated onto a silicon wafer and fully cured. A metal layer (10 nm chromium / 100 nm gold) was then deposited by electron-beam evaporation. Contact electrodes and traces were patterned via photolithography, followed by wet etching. Optionally, SU-8 6000.5 (Kayaku Advanced Materials, USA) epoxy doped with 0.01 wt% rhodamine B (Sigma-Aldrich, USA) was spin-coated, exposed, and cured to enable fluorescence visualization of the mesh.

To encapsulate the electrical traces, a second 5 μm-thick polyimide layer was spin-coated and cured over the first layer. A 20 nm titanium layer was deposited and patterned as a metal hard mask for reactive ion etching, which defined the mesh geometry and electrode contact openings. The polyimide was etched using oxygen plasma (power: 150 W, O₂ flow: 100 sccm). The titanium mask was then stripped using a buffered oxide etchant, and the films were thoroughly cleaned with acetone, isopropyl alcohol (IPA), and distilled water. During cleaning, the mesh-structured MEA films typically detached from the silicon wafer and floated freely in the solvent. Owing to their 10 μm thickness and millimeter-scale frame, the films are robust enough to be gently handled with fine-tip tweezers and transferred to temporary slide-glass substrates for drying. After drying, the contact electrodes were electroplated with PtB. For electroplating, the electrodes were immersed in a platinum solution (0.6 g H₂PtCl₆, 5 mg Pb(CH₃CO₂)₂, 20 mL H₂O) and subjected to a constant potential of −300 mV using a three-electrode setup (Autolab PGSTAT 302 N, Metrohm AG, Switzerland) with a platinum counter wire and an Ag/AgCl reference electrode.

The PCB base design was determined by the culture well dimensions, connector type, and the number of channels supported by the data acquisition system. To ensure ease of replacement and standardization, commercially available pin headers, flexible flat-cable connectors, and custom-printed PCB connectors compatible with Intan Technologies recording headstages were used. After soldering these components, the PCBs were cleaned with IPA and rinsed with distilled water to remove residual flux.

PDMS culture wells and spacers (150–600 μm thickness) were fabricated using custom 3D-printed molds. Molds were printed via stereolithography (Form 3, Formlabs), followed by washing (Form Wash) and UV post-curing (Form Cure). Uncured PDMS mixed at a 10:1 ratio (base:curing agent) was poured into the molds and cured at 80 °C for 4 h. The cured PDMS wells and spacers were demolded, cleaned with IPA and distilled water to remove oil and dust residues, dried with nitrogen gas, and thermally treated in an oven for 2 h to ensure complete crosslinking and surface cleanliness.

The integration process was performed under optical microscopy to ensure precise alignment of the modular components. First, a glass substrate was bonded to the PCB using epoxy adhesive. Optionally, a glass MEA chip could be incorporated with an additional wire-bonding step. To achieve a consistent ILD, a minimal volume of uncured PDMS (10:1 ratio) was applied to the spacer surface as an adhesive layer. The first layer of PDMS spacers is placed on the glass substrate and cured at 80 °C for 1 h to maintain a gap between the MEA film and the glass substrate. After applying an additional adhesive layer on the bonded PDMS spacers, the MEA film was sequentially aligned, attached, and cured. PDMS spacers and MEA films are repeatedly bonded for each subsequent layer, ensuring that the total assembly stack maintained uniform flatness (Fig. S2). Finally, the culture well was bonded, and the MEAs were connected to the flexible flat cable connectors on the PCB base. To verify the sealing integrity of the assembled culture well prior to cell seeding, a leak testing procedure was performed. The culture well was filled with sterilized deionized water and observed for any leakage for over 30 min. If leakage was detected, additional PDMS adhesive was carefully applied to the compromised area, followed by a final curing step to ensure the long-term sealing stability required for cell culture.

Electrochemical characterization

Electrochemical impedance spectroscopy (EIS) of the fabricated MLMEA was performed using a potentiostat (Autolab PGSTAT 302 N, Metrohm AG, Switzerland). The culture well was filled with phosphate-buffered saline (PBS; Sigma-Aldrich, USA), and a three-electrode configuration was established with a Pt counter electrode and an Ag/AgCl reference electrode. EIS measurements were carried out over a frequency range of 1 Hz–20 kHz with an AC voltage amplitude of 10 mV.

Finite element method simulation

An analysis of the mechanical behavior of the MLMEA under pressure was performed using Abaqus. The MEA components were modeled as homogeneous and isotropic materials. Polyimide, PDMS, and gold were simulated as linear elastic materials with Young’s moduli (E) and Poisson’s ratios (ν) of E_PI = 2.5 GPa, ν_PI = 0.34, E_PDMS = 1.3 MPa, ν_PDMS = 0.45, E_gold = 78 GPa, and ν_gold = 0.44.

Stem cell maintenance and cerebral organoid generation

The hiPSC line WTC-11 (GM25256; Coriell Institute for Medical Research, USA) was maintained under feeder-free conditions. The use of hiPSCs was approved by the Institutional Review Board (IRB) of Korea Institute of Science and Technology (Permit Number: KIST-202305-BR-002). Culture plates were coated with Corning® Matrigel® hESC-Qualified Matrix, LDEV-free (5 mL; Cat. #354277; Corning, USA) diluted 1:25 in DMEM/F-12 (Cat. #11320033; Gibco, USA). Cells were cultured in mTeSR™1 medium (Cat. #85850; STEMCELL Technologies, Canada) at 37 °C in a humidified incubator with 5% CO₂, with daily medium changes. Colonies were passaged when reaching ≈500 µm in diameter and treated with 1.5 U mL⁻¹ Dispase II (Cat. #17105041; Gibco™, USA) for 10–30 min to detach and generate EBs, which were transferred to low-adhesion conditions for EB formation.

For neural induction, the EBs were cultured in STEMdiff™ Neural Induction Medium (NIM; cat. #05835; STEMCELL Technologies, Canada). On day 1 of neural induction, 3 μM SB431542 (TGF-β inhibitor; S4317; Sigma-Aldrich, USA) and 3 μM dorsomorphin (BMP type I receptor inhibitor; ab144821; Abcam, UK) were added to perform dual SMAD inhibition. This was followed by a 1 d treatment with 3 μM CHIR99021 (Wnt pathway activator; Cat. #SML1046; Sigma-Aldrich, USA) to promote early neural patterning. On day 5, the EBs were embedded in Corning Matrigel (Cat. #356230; Corning, USA) diluted 1:25 in NIM and plated onto Falcon™ tissue culture-treated dishes (Corning, USA) for adherent culture. After 2 d of adherent culture, the organoids were detached and transferred to a suspension culture on an orbital shaker at 75 rpm. Between days 7 and 10, organoids were cultured in a medium comprising NIM and DMEM/F-12 supplemented with 1% N2 (Cat. #17502048; Gibco™, USA) and 2% B27 supplement (Cat. #17504044; Gibco™, USA). Between days 10 and 14, the culture medium was replaced with a 1:1 mixture of this base medium and STEMdiff™ Cerebral Organoid Maturation Kit (Cat. #08571; STEMCELL Technologies, Canada). Organoids were cultured in suspension in a medium consisting of NIM mixed with DMEM/F-12 (Cat. #11320033; Gibco, USA) supplemented with 1% N2 (Cat. #17502048; Gibco™, USA), 2% B27 supplement (Cat. #17504044; Gibco™, USA), 1% non-essential amino acids (NEAA; Cat. #11140050; Gibco, USA), and 1% penicillin-streptomycin (P/S; Cat. #15140122; Gibco™, USA), and 0.1% β-mercaptoethanol (Cat. #21985023; Gibco, USA) for 3 d. The culture medium was then replaced with a 1:1 mixture of the same base medium and STEMdiff™ Cerebral Organoid Maturation Kit (Cat. #08571; STEMCELL Technologies, Canada) and maintained for 4 d.

From day 14 onward, organoids were cultured in a 1:1 mixture of Neurobasal™-A Medium (Cat. #10888-022; Gibco™, USA) and STEMdiff™ Cerebral Organoid Maturation Kit (Cat. #08571; STEMCELL Technologies, Canada), supplemented with 1% N2, 2% B27, 1% NEAA, 1% P/S, and 0.1% β-mercaptoethanol. To promote neural maturation, neurotrophic factors including brain-derived neurotrophic factor (BDNF; R&D Systems, USA), glial cell line-derived neurotrophic factor (GDNF; R&D Systems, USA), neurotrophin-3 (NT-3; R&D Systems, USA), insulin-like growth factor 1 (IGF-1; R&D Systems, USA), and ascorbic acid (AA; Sigma-Aldrich, USA) were added at this stage. Suspension culture was maintained with regular medium replenishment until organoids reached the desired maturation stage.

Organoid placement on MLMEA

Cerebral organoids aged 3–4 weeks post-EB formation were used for MLMEA recordings. MLMEA chips were sterilized by immersion in IPA for 1 h, rinsed with distilled water, dried at 80 °C for 1 h, and UV-treated for 1 h. To enhance surface hydrophilicity, chips were treated with O₂ plasma and sequentially coated with 0.01% poly-L-ornithine (Cat. #P4957; Sigma-Aldrich, USA) and 10 µg mL⁻¹ laminin (Cat. #L2020; Sigma-Aldrich, USA).

Following chip preparation, cerebral organoids were gently transferred and carefully positioned onto the coated MLMEA chips to ensure optimal contact with the electrode sites. The organoids were allowed to stabilize on the chip surface for 24 h. To support viability and maturation, they were maintained in organoid differentiation medium with half-medium changes every 2 d. MEA signal recordings were performed directly in the culture medium, without media exchange immediately prior to measurement, thereby preserving physiological conditions during activity monitoring.

Electrophysiological monitoring of organoids

Neuronal activity was recorded using multichannel amplifiers (RHD 32-Channel Recording Headstages, Intan Technologies) and an electrophysiology recording system (RHD USB Interface Board, Intan Technologies), connecting the MLMEA platform to a computer. During monitoring, the MLMEA was placed inside an incubator housed within a custom-designed Faraday cage to minimize environmental noise and interference. Serial Peripheral Interface cables connecting the recording system were routed outside the incubator. Bandwidth and sampling frequency were set to 1–5000 Hz and 30 kHz, respectively.

Recorded signals were digitally filtered, analyzed, and curated using custom code modified from the Python-based open-source framework SpikeInterface, widely used in electrophysiological signal analysis⁴⁷. For spike sorting and clustering, the MountainSort5⁴⁸ algorithm was used with a detection threshold of 5.0 times the standard deviation. The overall electrophysiological setup and overall data analysis workflow are summarized in Fig. S3.

Immunostaining and fluorescence imaging

Organoids were fixed with 4% paraformaldehyde (PFA) for up to 1 h at room temperature (RT), followed by three washes with PBST (PBS containing 0.3% Triton X-100). Samples were then blocked in 6% bovine serum albumin (BSA; Cat. #160069; MP Biomedicals, USA) in PBST for 1 h at RT with gentle agitation. Primary antibodies diluted in blocking solution were applied and incubated for 1–2 d at RT with gentle shaking. After three PBST washes, samples were incubated for 2 h at RT with fluorophore-labeled secondary antibodies corresponding to the host species of the primaries. Following secondary staining, samples were washed thoroughly with PBST to minimize residual fluorescence. DAPI (1 μg/mL) was included in the final wash, followed by additional PBST rinses. Samples were then incubated overnight in a clearing solution and imaged using a confocal microscope (DMi8, Leica Microsystems GmbH, Germany).

For live-cell labeling, Calcein AM (Cat. #C34852; Invitrogen™, Thermo Fisher Scientific, USA) was added by replacing half of the culture medium with fresh differentiation medium containing Calcein AM at a final concentration of 1 μM. Organoids were incubated for up to 30 min at 37 °C, followed by immediate fixation with 4% PFA for 1 h at RT. After fixation, samples were washed with PBST, stained with 1 μg/mL DAPI (Cat. #D1306; Invitrogen™, Thermo Fisher Scientific, USA), and processed using the immunostaining and clearing protocol described above. Cleared samples were incubated overnight and imaged by confocal microscopy.

Scanning electron microscopy (SEM) imaging

Organoid morphology on mesh-type MEAs was evaluated using SEM (FEI Inspect F50, Thermo Fisher Scientific, USA). Samples were fixed for 2 h at RT in 2.5% glutaraldehyde in 0.1 M cacodylate buffer (pH 7.4), followed by two washes in cacodylate buffer and dehydration through a graded ethanol series. Samples were sputter-coated with a thin layer of platinum and imaged at an accelerating voltage of 1.0 kV.

Author contributions

N.K. and M.K. contributed equally to this work. N.K. and Y.J.L. conceived and designed the experiments. N.K., M.K., N.C., and Y.J.L. wrote the manuscript. N.K. designed, fabricated, and characterized the microelectrode array platform. M.K. and J.J. performed cell cultures. M.K. and H.K. conducted immunofluorescence staining and analyzed related data. N.K. performed electrophysiological recording and analyzed related data. H.N.K., N.C., and Y.J.L. supervised the project and provided resources. All authors discussed the results and commented on the manuscript.

Data availability

Data will be made available on request.

Conflict of interest

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41378-026-01328-8.