Introduction

Neural interfacing technologies aim at establishing seamless and stable communication between manmade devices and native neural tissues[1] to provide innovative insights into the nervous system[2] and thus contribute to the exploration of the mysterious and intricate neural network.[3] In the 1960s, there have been studies implanting electrodes to stimulate neural tissues or record neural activity in animal models[4,5] and to understand the organization and operation of neural circuits. In the current state, implantable macro- or micro-devices are widely used in neuroscience research and therapeutic treatment of neurological diseases including Parkinson’s disease, depression, and obsessive compulsive disorder. [6,7,8] Modern interfaces enable interrogation of neural network dynamics by electrophysiological recording, while direct stimulation of neural tissue offers powerful therapeutic interventions.

Neural interfaces have evolved from simple electrodes to sophisticated devices, yet a key challenge remains: capturing neural activity across large populations of neurons with high temporal resolution and single-cell precision.[9] Another challenge is to achieve seamless, long-term integration between devices and delicate neural tissue. Most conventional neural probes are made of rigid materials that drastically mismatch the softness of neural tissue,[10,11] which causes inflammation, glial scarring, and signal degradation over time. Indeed, inflammatory responses and progressive neural degeneration at the implant site of conventional electrodes deteriorate the signal-to-noise ratio (SNR) over weeks to months.[12,13,14] These factors impede stable long-term recordings and limit the longevity of brain implants in vivo.

Recent developments in bioelectronic interface designs are tackling these issues through innovations in materials and structural design (Scheme 1). Flexible and stretchable electronics, composed of tissue-mimetic polymers and conductive composites, have emerged as promising alternatives to rigid silicon and metal probes.[15] These soft systems minimize mechanical mismatch and maintain stable contact with neural tissue over extended periods of time, enabling chronic electrophysiological monitoring with minimal immune response. Meanwhile, the advent of scalable, high-density electrode arrays has significantly expanded recording capability.[16] Microelectrode platforms now contain hundreds to thousands of channels, allowing large-scale parallel recording of neural activity with cellular resolution.

Scheme 1
scheme 1

Bioelectronics for neural interfaces. Schematic illustration of the key directions in neural interface bioelectronics.

Full size image

As neuroscience research advances to human-specific models, brain organoids derived from human stem cells have gained attention as a complementary platform to animal models to study neural development and disease.[17] These 3D organoids recapitulate many aspects of human neural architecture and function, offering experimental access to human brain biology in vitro.[18,19] However, conventional electrophysiological tools, such as planar multielectrode arrays, have challenges in probing cells within the organoids without disrupting their cytoarchitecture. To overcome these challenges, 3D neural interfaces have been developed to enable chronic monitoring and modulation of neural activity at the single-cell level within intact 3D cultures. These platforms offer the potential to map synaptic connectivity and functional circuits with high spatiotemporal resolution.

Beyond advancements in recording scalability and mechanical compliance, recent developments have led to multifunctional neural interfaces that redefine the boundaries of traditional electrophysiology. By integrating electrical, optical, and chemical modalities into a single platform, these systems enable precise, real-time, and multimodal interrogation of neural circuits.[20,21] Such multifunctional architectures support closed-loop neuromodulation, targeted drug delivery, and optogenetic control.

Moreover, the advent of living bioelectronics marks a paradigm shift toward interfaces that are no longer passive receivers, but active actuators of the nervous systems. Engineered neuronal constructs, soft tissue-like scaffolds, and biohybrid systems are being harnessed to create regenerative, self-adaptive, and even synaptically functional interfaces. These living platforms offer improved biocompatibility and signal fidelity, holding promise for next-generation implants that can grow, remodel, and heal alongside host neural tissue.

In this perspective, we summarize recent breakthroughs on scalable neural recording platforms, soft and stretchable bioelectronics, organoid-integrated 3D interfaces, multifunctional neural probes, and living bioelectronic constructs. We highlight how these advances redefine the capabilities of neural interfacing, paving the way for a future in which engineered devices and the nervous system operate in seamless and adaptive coordination.

Scalable neural probes for large-scale neural recording

The brain is an intricate network of various cell types that work together to process information, control movement, and generate thoughts. To understand how these circuits function, researchers need to record the activity of neurons across interconnected brain regions while maintaining single-neuron resolution. One of the key challenges is to design tools that are capable of capturing neuronal activity at millisecond scale[22,23,24] from large populations of neurons. Traditional microelectrodes are often limited to recordings from tens of neurons.[25] While powerful for visualizing large neuronal populations simultaneously, optical imaging techniques face limitations in penetration depth, temporal resolution, and phototoxicity.[9] This underscores the need for scalable MEA technologies that can record from thousands of neurons in deep brain regions with millisecond precision.

Two classical microelectrode designs have shaped the evolution of scalable neural probes: the Utah array [Fig. 1(a)] and the Michigan probe [Fig. 1(b)]. The Utah array features a grid of rigid, vertically penetrating silicon needles. Each silicon needle is equipped with a single recording site at the tip, which enables parallel recordings across large cortical areas.[12] However, it sacrifices depth coverage per electrode shank. In contrast, the Michigan probe[26] features silicon shanks with linearly distributed microelectrode sites, which allows laminar sampling at different depths of the brain with limited lateral coverage.

Figure 1
figure 1

Reproduced from Ref. 12, Licensed under CC BY-NC-SA 3.0. (b) Michigan-style neural probes with multiple recording sites linearly distributed along silicon shanks. Reproduced with permission from Ref. 26. Copyright 2004, Elsevier. (c) Microwire-CMOS hybrid system for massively parallel neural recording. Left: schematic of flexible microwire bundles connected to a CMOS readout chip via compression contacts. Right: scanning electron microscope (SEM) images of microwire array arranged in a dense 3D configuration. Reproduced from Ref. 27, Licensed under CC BY 4.0. (d) Neuropixels 1.0 with detachable headstage and large footprint. Reproduced with permission from Ref. 11. Copyright 2017, Springer Nature. (e) 3D stacking of Michigan-style multi-shank probes. (i) Schematic of individual probes with electrode distribution. (ii) Vertical stacking of multiple probes to form volumetric arrays. (iii) Optical image showing a high-density multi-shank configuration with 1,024 electrodes in a volume of only 0.6 mm3. Reproduced with permission from Ref. 28. https://pubs.acs.org/doi/10.1021/acs.nanolett.6b02673. Copyright 2016, American Chemical Society. (f) CMOS-based high-density electrode arrays for intracellular and synaptic activity mapping. (i) 32 × 32 pixel CMOS nanoelectrode array (CNEA) with subcellular resolution. Reproduced with permission from Ref. 16. Copyright 2017, Springer Nature. (ii) 64 × 64 pixel CMOS nanoelectrode interface (CNEI) integrated with a packaged chip. Reproduced with permission from Ref. 29. Copyright 2020, Springer Nature. (iii) Platinum-black microhole electrode array. Reproduced with permission from Ref. 30. Copyright 2025, Springer Nature.

Recent advances in scalable neural probe architectures. (a) Utah array: 3D silicon-based penetrating microelectrode arrays.

Full size image

One prominent example of Utah-style probe development is the microwire–complementary metal-oxide–semiconductor (CMOS) hybrid system. In this approach, dense bundles of polymer-insulated microwires are directly integrated into CMOS readout chips. For example, Schaefer et al. demonstrated a configuration that interfaces thousands of 5–25 μm microwires with a CMOS pixel array [Fig. 1(c)], enabling massively parallel extracellular recordings in vivo.[27] This method mitigated the bottleneck of one-to-one wire connections and allowed recordings from hundreds of neurons in mice.

Neuropixels probes, a representative advancement from the Michigan-style architecture, have been widely adopted in neuroscience research [Fig. 1(d)]. Harris et al. introduced Neuropixels 1.0 which integrated 960 densely packed electrodes along a 10-mm-long silicon shank and enabled dynamic selection of 384 active recording sites.[11] This platform enabled simultaneous tracking of neuronal activity across multiple brain regions with single-spike resolution. Moving forward, Neuropixels 2.0 offers a smaller form factor (approximately one-third the size of the original headstage in Neuropixel 1.0) along with both single- and quad-shank configurations.[31] Each shank contained 1,280 recording sites (total 5,120 in the quad version) while maintaining 384 simultaneously addressable channels. Compact electrode spacing provided great spatial sampling, and integrated motion correction algorithms supported stable, long-term tracking of neuronal activity in freely behaving small animals.

To integrate the lateral coverage of Utah arrays with vertical coverage of Michigan-style probes, researchers have developed 3D silicon needle arrays [Fig. 1(e)].[28] These devices integrate multiple Michigan-style shanks into a monolithic 3D architecture, enabling simultaneous recordings from 3D volumes. By incorporating over 1,000 electrodes within a small volume (~ 0.6 mm3), these arrays enabled high-density, multilayered electrophysiological recording, offering great spatiotemporal resolution and large-scale neural data acquisition.

While neural probes Like Neuropixels and 3D shank devices have advanced large-scale neural recording, their signals remain limited to extracellular action potentials, which provide indirect measures of subthreshold activity and synaptic inputs. In contrast, intracellular recording offers direct access to membrane potentials, allowing the resolution of subthreshold voltage fluctuations, postsynaptic potentials, and even the integration of synaptic inputs that do not lead to spiking. To realize intracellular recording, Ham et al. combined CMOS-integrated circuits with vertical nanoelectrodes (~ 2 µm diameter, ~ 3 µm height, 20 µm pitch), enabling parallel intracellular recording.[16] With 1,024 recording pixels integrated on a CMOS chip, this platform could support recording and stimulation of membrane potentials in single cells and across the entire network. Recording electrodes were designed to be high-density sites that minimize impedance and maximize signal-to-noise ratio. Stimulation electrodes were fabricated with larger geometries and functionalized surfaces to provide higher charge injection capacity and stable current delivery. The working principle of stimulation electrodes relies on charge transfer at the electrode–electrolyte interface, which depolarizes the neuronal membrane and evokes action potentials. To further advance massive parallel intracellular neuronal recording, researchers integrated 4,096 platinum-black (PtB) electrodes into the CMOS chip.[29] Compared to conventional metals, PtB electrodes provide a high surface area and reduced impedance which enhances charge injection capacity and enables both stable recording and effective stimulation. The roughened PtB surface enhanced the charge storage capacity, thereby supporting safe and efficient stimulation. With more recording and stimulation sites, this array could detect synaptic signals and map 304 synaptic connections from 1,728 neurons. Moreover, these nanoneedle systems have been further transformed into microhole arrays[30] (~ 2 µm diameter holes, ~ 3 µm depth, 20 µm pitch) to further advance stability and scalability. The microhole structure enhanced the mechanical anchoring between the electrode and the cell membrane, allowing neurons to seal over electrodes without puncture. With 4,096 PtB electrodes and well-nested microhole design, the platform could allow neurons to seal over electrodes without membrane puncture and thus could capture more than ~ 70,000 synaptic connections among ~ 2,000 neurons. This array was capable of classifying synaptic types and re-accessing the same neuron. This series of development of CMOS-based approaches allows for high-resolution subthreshold voltage recordings and functional synaptic mapping, bridging the gap between traditional patch-clamp techniques and large-scale extracellular electrophysiology [Fig. 1(f)].

Advanced flexible bioelectronics for in vivo neural interfacing

Traditional rigid metal electrodes can provide high spatial resolution recording, but mechanical mismatches between rigid electrodes with Young’s modulus in the GPa range[32] and soft biological tissues in the kPa range[33] can cause tissue inflammation and degradation of recording fidelity over time. The advent of flexible bioelectronics presents a promising alternative, since these innovative platforms can provide mechanical compliance, long-term stability, and functional adaptability, making them great candidates for next-generation neural interfaces.

To develop flexible bioelectronics, there are two main strategies [Fig. 2(a)]. One is to synthesize intrinsically soft materials with low Young’s modulus.[34] The other focuses on reducing the dimensions of rigid materials to decrease the overall bending stiffness of devices.[35] Conductive polymers (CPs), which combine metal-like electrical conductivity with polymer-like mechanical flexibility,[36] have attracted significant interest for the development of implantable scaffolds.[15] Poly(3,4-ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSS) is a suitable candidate for bioelectronic applications[37] given its low Young’s modulus and good conductivity. Bao et al. pioneered a hydrogel-based elastic microelectronic platform for low-voltage neuromodulation,[38] marking a significant shift from traditional rigid interfaces to mechanically compliant bioelectronics [Fig. 2(b)]. By combining conductive polymer with elastic fluorinated photoresist (perfluoropolyether-dimethacrylate (PFPE-DMA)), this system significantly reduces interfacial impedance, thus enabling efficient charge transfer for localized nerve stimulation. The impedance reduction is primarily achieved by the porous PEDOT:PSS hydrogel structure, which allows deep ion penetration and enlarges the effective electrochemical surface area, thereby lowering interfacial impedance compared with flat metal electrodes. The PFPE-DMA photoresist ensures mechanical compliance and stable anchoring of the elastic conductive hydrogel, while functioning as a stretchable insulating encapsulation layer that preserves neural interface integrity through preventing leakage currents and restricting stimulation to the designated electrode window. This design enables localized and ultralow-voltage stimulation (~ 50 mV), improving spatial resolution and minimizing off-target activation. In vivo studies demonstrate its efficacy in peripheral nerve modulation with minimal power consumption, reducing risks associated with overheating and overstimulation. This work highlights the potential of conductive polymer-based hydrogels (CPHs) as active neuromodulation platforms for next-generation bioelectronic implants. In another research, the incorporation of PEDOT:PSS conductive network within a hydrophilic polyurethane scaffold equipped the hydrogel system with remarkable mechanical toughness (~ 3,300 J/m2), stretchability (> 400%), and electrical conductivity (> 11 S/cm)[39] [Fig. 2(c)]. Based on similar material strategies, Bao et al. embedded a laser-induced porous graphene network into a polystyrene-block-poly(ethylene-ran-butylene)-block-polystyrene (SEBS) elastomer matrix to develop a stretchable neurotransmitter sensor called NeuroString[40] [Fig. 2(d)]. To enhance electrochemical performance, iron oxide nanoparticles were incorporated into the graphene–elastomer composite. These nanoparticles reduce interfacial impedance by increasing effective surface area and providing additional redox-active sites, which improve charge-transfer kinetics during neurotransmitter oxidation. NeuroString was capable of detecting dopamine and serotonin fluctuations with high sensitivity, even in constantly moving biological environments like the gut-brain axis.

Figure 2
figure 2

Reproduced with permission from Ref. 38. Copyright 2019, Springer Nature. (c) Formation of bicontinuous conductive polymer hydrogels (BC-CPHs). Interconnected domains of continuous electrical (blue) and mechanical (black) phases form BC-CPH network. Reproduced with permission from Ref. 39. Copyright 2023, Springer Nature. (d) NeuroString: a graphene-based stretchable neurotransmitter sensor. (i) Design overview shows the porous graphene structure embedded in SEBS elastomer and integrated with Fe3O4 nanoparticles. (ii) Optical image of NeuroString being implanted into the mouse brain. Reproduced with permission from Ref. 40. Copyright 2022, Springer Nature. (e) Multilayered, fluorinated elastomer-based neural probe design. (i) Schematic of the fabrication layout. (ii) Optical image of microfabricated stretchable neural probe with multiple stacked layers and aligned interconnects. Reproduced with permission from Ref. 41. Copyright 2024, Springer Nature. (f) NeuroWeb: ultrathin, skin-like brain surface neural interface. (i) Fabricated mesh with Au and SU-8 interconnects. (ii) Implanted NeuroWeb device conformally laminated onto a rodent cortex. Reproduced from Ref. 42, Licensed under CC BY 4.0. (g) NeuE for chronic single-cell interfacing. (i) Illustration of NeuE device with axon-mimicking morphology. (ii) Comparison of a single neuron and NeuE design. Reproduced with permission from Ref. 43. Copyright 2019, Springer Nature.

Flexible bioelectronics for in vivo neural interfacing. (a) Conceptual overview of flexible bioelectronic design created with BioRender.com. (b) Schematic of PEDOT:PSS-based micellar hydrogels and their responsive behavior.

Full size image

To address the challenge of long-term stability of hydrogels during their in vivo implantation,[44] Liu et al. developed perfluorinated dielectric elastomers[41] to ensure extended biofluid resistance and electrical stability. Perfluoropolyether-based probes demonstrated superior durability [Fig. 2(e)] and high-fidelity neural recordings over months of implantation. This work highlights the crucial role of materials innovation in improving longevity and precision in neural interfacing.

Another strategy to develop flexible neural interfaces focuses on their structural design to minimize the bending stiffness of rigid components without compromising their electrical performance.[34,35,38] Structural innovations, including mesh networks, serpentine designs, and ultrathin films, have been developed to integrate with the intricate architecture of brain tissue. Many of these designs leverage SU-8, an epoxy-based photoresist that provides high patterning fidelity and thermal stability. When patterned into submicrometer thicknesses and perforated structures, SU-8 can approximate tissue-like flexibility. For example, the NeuroWeb system [Fig. 2(f)] comprising ultrathin hexagonal boron nitride, graphene, and submicrometer-thick SU-8 demonstrated superior flexibility and adhesion to curved brain surfaces.[42] Mechanical testing revealed that NeuroWeb maintained low bending stiffness and resistance under deformation. Researchers also took inspiration from biological systems to develop biomimetic strategies for ultra-flexible neural probes. Lieber et al. developed neuron-like electronics (NeuE)[43] [Fig. 2(g)] by designing the electronics to match the structures of neuron soma and neurites at the subcellular scale. NeuE followed single-neuron activity from brain regions such as the hippocampus and cortex over 90 days. To enable minimally invasive probe delivery, Lieber et al. reported an ultra-flexible micro-endovascular probe based on the SU-8 mesh-like substrate.[45] This probe could be delivered to the brain through 100-μm vessels and realize neuro-monitoring through vessel walls.

Novel 3D neural interfaces for brain organoids

Animal models, particularly rodent models, have significantly contributed to neuroscience studies and greatly advanced our understanding of brain development and function.[17] Meanwhile, it is also important to recognize that certain species-specific differences between rodents and humans exist, which may limit the full extrapolation of findings. For example, human cortical progenitor cells typically undergo more proliferative cycles prior to neuronal differentiation compared to their rodent counterparts.[46] Additionally, notable distinctions have been observed in the gene expression patterns governing cortical development and function between human and mice.[47] Therefore, human-derived models, such as human brain organoids[18] and spheroids,[48] have emerged as valuable complementary platforms. These models recapitulate key aspects of human-specific brain development in 3D and, when used alongside traditional animal models, offer a comprehensive framework for investigating the evolution and complexity of the human brain.

Existing technologies for recording neural activity, such as patch clamp and planar MEA, often face challenges in long-term and high-resolution monitoring of brain organoids due to their limited compatibility with the 3D architecture of organoids. For example, patch-clamp techniques require slicing organoids in order to access interior cells, which causes irreversible tissue disruption.[49] Conventional planar MEAs are primarily designed for monolayer cultures, which limits their contact within 3D organoids [Fig. 3(a)].[19]

Figure 3
figure 3

Reproduced from Ref. 50, Licensed under CC BY 4.0. (c) Hydrogel-actuated e-Flower MEA system enabling mechanically adaptive interfacing. (i) Schematic of hydrated poly(acrylic acid) actuator causing polyimide arms to bend inward and wrap around brain spheroids. (ii) Structural layout of the full platform, including the main well, microfluidic inlets/outlets, and a six-arm platinum-polyimide MEA. Reproduced from Ref. 51, Licensed under CC BY 4.0. (d) Stretchable mesh nanoelectronics enabling seamless integration into brain organoids during development. (i) SU-8 scaffold with embedded microelectrodes. (ii) Schematic of interface strategy: mesh structure is introduced during early organoid formation and matures together with neural tissues, ensuring stable long-term recordings. Reproduced with permission from Ref. 52. Copyright 2022, Wiley–VCH. (e) KiriE for 3D organoid interfacing. Reproduced with permission from Ref. 53. Copyright 2024, Springer Nature. (f) Multifunctional 3D neural interface for cortical spheroids and assembloids. (i) Multilayer design: gold wiring and microdevices embedded between polyimide layers atop an elastomeric substrate. (ii) Fully assembled flower-like array with petals conforming to the organoid surface. (iii) Large-scale deployment of multiple units for parallel, high-throughput interfacing with minimal tissue disruption. Reproduced from Ref. 54, Licensed under CC BY 4.0.

Novel 3D neural interfaces for brain organoids. (a) Conventional electrophysiological methods for brain organoid studies: (i) Patch clamp and (ii) 2D MEAs (created with BioRender.com). (b) Self-folding shell MEA platform for conformal encapsulation of brain organoids. (i) A germanium sacrificial layer and patterned SU-8 bilayers are used to drive autonomous folding, while gold electrodes and PEDOT:PSS coating ensure biocompatibility and conductivity. (ii) A brain organoid is fully encapsulated by three folding electrode channels, achieving intimate surface contact and enhanced recording fidelity.

Full size image

To overcome this limitation, researchers have developed 3D neural interfaces that can flex, adapt, and even self-wrap around organoids, offering a new level of precision in neural monitoring and manipulation. To promote intimate organoid–device interface, Gracias et al. developed shell MEAs[50] where self-folding SU-8 polymer bilayers were employed to conformally wrap around brain organoids [Fig. 3(b)]. Due to their ability to autonomously encapsulate organoids, shell MEAs could optimize electrode positioning to maximize signal acquisition while minimizing external manipulation. This architecture enhanced spatial resolution and SNR, providing a high-fidelity platform for electrophysiological mapping. Another novel design was the development of e-Flower MEAs through a biologically responsive strategy[51] [Fig. 3(c)]. The e-Flower MEAs interface exploited hydrogel swelling process to autonomously conform to organoid surfaces upon hydration, enabling gentle yet robust contact. This stimuli-responsive behavior could precisely control the folding angle and curvature, thus realizing customization based on brain spheroid size. By incorporating dynamic adaptability, e-Flower MEAs demonstrated stable neural spike detection, extracellular field recordings, and chemically induced neuromodulation.

Long-term interfacing with growing brain organoids poses unique challenges, particularly while maintaining stable contact without disrupting their 3D architecture or development. To realize seamless integration with growing 3D organoids, stretchable mesh nanoelectronics were embedded into the 3D organoid via the natural organogenesis process[52] [Fig. 3(d)]. This platform accomplished real-time tracking of neural maturation and network formation over six months. Moreover, Cui and Pașca et al. introduced a kirigami-inspired interface design strategy, kirigami electronics (KiriE),[53] where mechanical engineering was employed to achieve seamless integration with brain organoids in suspension [Fig. 3(e)]. By leveraging mechanically optimized kirigami patterns, KiriE undergoes a two-dimensional (2D) to 3D transition in situ, enveloping organoids while maintaining non-disruptive contact. This structural adaptability facilitated stable long-term recordings for over 120 days, making it well suited for studies on neurodevelopment and disease modeling.

Moving beyond passive electrophysiological recording, multifunctional 3D neural interfaces could provide a holistic approach to neural interrogation by incorporating electrical, optical, and chemical stimulation capabilities. Rogers et al. integrated micro-LED arrays for optogenetic manipulation, microfluidic networks for targeted drug delivery, and multiplexed microelectrode arrays for synchronized neural activity mapping[54] [Fig. 3(f)]. With these modalities, this new type of multifunctional neural interfaces achieved precise spatial and temporal control of organoid function, facilitating the investigation of complex neural circuits and dynamic plasticity mechanisms.

Advancement of multifunctional neural interfaces

Multifunctional neural interfaces represent a transformative advancement in neuroengineering through integrating multiple modalities for simultaneous neural recording and modulation. Compared to conventional neural probes designed primarily for electrophysiological recording, multifunctional neural interfaces enable a more comprehensive characterization and manipulation of neural circuits. By combining electrical, chemical, and optical functionalities within a single device, these platforms can be applied to facilitate precise and real-time analysis of neuronal behavior in response to controlled perturbations.

Electrical stimulation is one of the most common neural modulation modalities for fundamental neuroscience and clinical practice. Recent advances in electrical stimulation have focused on increasing their charge injection capacity (CIC) while minimizing tissue damage. Various nanomaterial coatings, including iridium oxide,[55] platinum,[56] and conducting polymers such as PEDOT:PSS,[57] have been employed to increase the effective surface area and enhance CIC [Fig. 4(a)]. Structural engineering, such as carbon nanotube arrays,[58] porous platinum nanorods,[59] and graphene fiber-based stimulating electrodes,[60] and various quantum dot-based nanostructures including carboxyl-functionalized graphene quantum dots (QDs),[61] near-infrared responsive QDs[62,63] and AgBiS2 QDs,[64] as well as RuO2 supercapacitor-integrated optoelectronic biointerfaces,[65] have also been demonstrated to improve stimulation efficiency through enlarging the electrochemical surface area and enhancing charge transfer [Fig. 4(b)]. Moreover, the integration of recording and stimulation components into a single probe allows closed-loop neural interfacing with real-time feedback.

Figure 4
figure 4

Reproduced with permission from Ref. 55. Copyright 2016, American Chemical Society. Reproduced with permission from Ref. 56. Copyright 2020, Wiley–VCH. (b) Nanostructuring techniques for high-efficiency bioelectronic interfaces. From left to right: carbon nanotube (CNT) array electrodes; platinum nanorods; graphene fiber electrodes. Reproduced with permission from Ref. 58. Copyright 2006, American Chemical Society. Reproduced with permission from Ref. 59. Copyright 2019, American Chemical Society. Reproduced from Ref. 60, Licensed under CC BY 4.0. (c) Microfluidic chemotrode integrating fluidic and electrical modalities. The 3-inlet split-Herringbone Mixer microfluidic chip allows combinatorial chemical delivery, while the shank and underlying electrodes enable simultaneous recording and stimulation. Reproduced with permission from Ref. 66. Copyright 2015, Royal Society of Chemistry. (d) Wireless optofluidic neural interface. Schematic of a multifunctional microfluidic probe integrating micro-LEDs (μ-ILEDs), Joule-heated expandable reservoirs, Cu membranes, and fluidic channels. Reproduced with permission from Ref. 20. Copyright 2015, Elsevier. (e) Aerosol jet printing of bioactive nanomaterials for enhanced tissue-device integration. Reproduced with permission from Ref. 21. Copyright 2015, Wiley–VCH. (f) Bioelectronic neural pixel enabling simultaneous chemical sensing and electrophysiology. Reproduced with permission from Ref. 67. Copyright 2016, National Academy of Sciences. (g) Miniaturized multimodal neural implant. Left: schematic of a microelectrode array incorporating electrical pins, optical fibers, and fluidic channels in a single housing for multimodal stimulation/recording. Right: photograph of a rodent implanted with the system. Reproduced with permission from Ref. 68. Copyright 2012, Springer Nature. (h) Hydrogel-based neural probe with low mechanical mismatch. Left: photograph of an injectable hydrogel fiber emitting blue light. Right: immunofluorescence image confirming the reduced presence of GFAP+ astrocytes. Reproduced with permission from Ref. 69. Copyright 2018, Wiley–VCH. (i) Transparent graphene-based flexible neural interface. Left: A Parylene C–Au/Cr–graphene sandwich structure. Right: image of the fully assembled flexible device conformally mounted onto a curved substrate. Reproduced from Ref. 70, Licensed under CC BY 4.0.

Advancement of multifunctional neural interfaces. (a) Nanocoating strategies to enhance neural electrode performance. SEM images of IrO2 (left) and Pt (right) coatings.

Full size image

Neuromodulation-based therapies have been extensively applied in clinical settings to modulate neural activity in various neurological disorders. While effective in delivering compounds to specific sites, conventional infusion methods such as direct microinjection[71] do not simultaneously monitor neural responses. Therefore, a promising direction is to integrate microfluidic drug delivery systems with neural probes, thus allowing controlled neuromodulation while recording neural activity. Silicon-based probes equipped with microfluidic channels were demonstrated to realize simultaneous neural recording and localized drug infusion. Shin et al. incorporated a staggered herringbone mixer within the microfluidic platform to enable efficient mixing and delivery of multiple drugs into the brain[66] [Fig. 4(c)]. Moreover, wireless drug delivery systems were further developed to realize pharmacological modulation in freely moving animals, minimizing tethering constraints. For example, wireless optofluidic systems employing thermal expansion pumps enable on-demand drug release and real-time recording without external tubing[20] [Fig. 4(d)]. Additionally, drug-releasing coatings such as anti-inflammatory nanogels and conductive polymer-based drug reservoirs have been utilized to achieve sustained and electrically controlled release at targeted recording sites. For instance, Chen et al. developed a multifunctional neural probe by applying anti-inflammatory nanogel coatings to its surface[21] [Fig. 4(e)]. This nanogel-functionalized probe enabled stable and long-term neural recordings from the rat thalamic nuclei for more than four weeks. In another research, PEDOT:PSS-based recording electrodes were integrated with organic electronic ion pumps for multifunctional neural probes[67] [Fig. 4(f)].

Optogenetics enables cell-type specific and temporally precise modulation of neural circuits. For example, emerging molecular photoswitches such as membrane-targeted push–pull azobenzenes[72] have been demonstrated to optically modulate membrane potential, offering a new strategy for precise and non-invasive neuromodulation. In addition, to integrate optical stimulation into electrophysiological recording device, researchers have developed multifunctional probes combining electrodes with transparent waveguide for light delivery. Silica optical fiber-based optogenetic systems[68,73] [Fig. 4(g)] have been used to achieve precise light delivery to deep brain regions. Moreover, miniaturized micro-LED arrays have been directly integrated into flexible probes,[74] allowing wireless and tetherless optogenetic manipulation in freely behaving animals. Soft and stretchable waveguides such as hydrogel-based optical fibers[69] [Fig. 4(h)] offer great mechanical compliance. In vivo studies demonstrated that these hydrogel fibers elicited significantly lower astrocytic activation compared to conventional silica fibers, as indicated by reduced GFAP expression around the implant site. Additionally, transparent electrode materials have been used to achieve simultaneous optical stimulation and electrical recording without generating optical artifacts. For example, transparent graphene-based electrocorticography arrays have enabled large-scale mapping of cortical activity under optogenetic control[70] [Fig. 4(i)]. Moreover, novel optoelectronic synaptic systems such as quantum dot–based artificial synapses[75] that directly convert light into neuromorphic photostimulation of neurons can realize simultaneous photodetection, memory, and neurostimulation functionalities in a compact architecture.

The rise of living bioelectronics

Conventional neural interfaces, typically fabricated using metals and semiconductors, inevitably generate foreign-body response (FBR) and glial scar formation.[76] The significant difference between the elastic modulus of these materials and the soft neural tissue typically results in micromotion-induced inflammation, neuronal loss, and degradation of recording and stimulation performance. To address these challenges, researchers have been increasingly focused on engineering interfaces with biological components. Living bioelectronics represents an emerging type of neural interfaces where biologically derived components are integrated into them or constitute the interface itself[77] [Fig. 5(a)]. Conventional bioelectronic devices typically employ synthetic conductors and coatings to interface with tissue, often resulting in static and non-adaptive interactions. In contrast, living bioelectronics leverage biologically active components—such as engineered cells or living scaffolds—to establish dynamic, functional, and even regenerative integration with host tissue.[78] This new strategy aims at addressing long-standing issues of chronic FBR, mechanical mismatch, and signal instability, meanwhile giving sights to cell-mediated modulation and therapeutic avenues.

Figure 5
figure 5

Reproduced from Ref. 77, Licensed under CC BY 4.0. (b) Step-by-step process for μTENN fabrication. Reproduced from Ref. 79, Licensed under CC BY 4.0. (c) Schematic of an optically modulated μTENN-based brain-machine interface. Reproduced with permission from Ref. 78. Copyright 2018, Wiley–VCH. (d) Restoration of the nigrostriatal pathway using dopaminergic μTENNs. (i) Dopaminergic aggregates are seeded into hydrogel microcolumns and cultured to induce axonal growth. (ii) Fluorescence imaging confirms directional axonal growth from a dopaminergic aggregate along the microcolumn, forming a functional tract between the substantia nigra and the striatum in rodent models. Reproduced from Ref. 80, Licensed under CC BY-NC-ND 4.0. (e) Vascular electronic scaffold (VasES) for cortical injury repair. Left: schematic illustration of VasES-guided neural regeneration. Right: fluorescence microscopy image showing integration of the VasES (red) with the regenerating cortical vasculature (magenta). Reproduced with permission from Ref. 81. Copyright 2023, Springer Nature. (f) TENGs for peripheral nerve regeneration. Top: overview of a TENG bridging a segmental nerve defect. Bottom: TENGs provide aligned axonal tracts and cell adhesion cues, facilitating host axons and Schwann cells regeneration via direct contact and secreted trophic factors. Reproduced from Ref. 82, Licensed under CC BY-NC 4.0.

Emerging living neural interfaces incorporating engineered biological components. (a) Schematic of a μTENN as a living neural interface. The construct consists of a neuronal aggregate seeded into a hydrogel microcolumn, from which long-distance axon tracts grow and extend toward host neural tissue.

Full size image

A key representative of living bioelectronics is the micro-Tissue Engineered Neural Network (μTENN),[77,78,79] which consists of a neuronal aggregate seeded at one or both ends of a hydrogel microcolumn from which long-projecting axonal tracts extend [Fig. 5(b)]. These engineered axonal pathways act as biological “wires” capable of establishing functional synapses with the host neurons upon implantation. Unlike traditional metal-based electrodes which record or stimulate neural activity via charge injection, μTENNs leverage the intrinsic bioelectrical properties of axons and neurons to transmit and modulate signals. Significantly, using different neuronal subtypes during the fabrication process enables customization of living electrodes for diverse functional applications. For instance, μTENNs composed of dopaminergic neurons have been used to reconstruct the nigrostriatal pathway in Parkinson’s disease models, leading to restored dopaminergic signaling in the striatum.[83] Similarly, GABAergic or glutamatergic neuron-based constructs could be utilized for inhibitory or excitatory modulation, respectively, expanding the potential applications of living bioelectronics from disease treatment to brain-machine interface development.[78]

Living bioelectronics also exhibit remarkable potential as regenerative platforms for neural circuit repair. One of the most promising applications is the reconstruction of white matter tracts, which are frequently disrupted by traumatic brain injury, neurodegenerative diseases, and stroke [Fig. 5(c)]. Tissue-engineered axonal tracts, such as dopaminergic μTENNs, have successfully re-established nigrostriatal connectivity and restored dopamine release in rodent models of Parkinson’s disease [Fig. 5(d)].[80] In another research, Lieber et al. coated a vascular electronic scaffold (VasES)[81] with laminin, an extracellular matrix protein essential for promoting cell adhesion and migration. This biochemical modification enabled VasES to significantly enhance migration of neural progenitors from the subventricular zone into the damaged cortical region in the mouse cortical resection model [Fig. 5(e)].

Moreover, the living bioelectronic strategy can be applied beyond the central nervous system. These constructs have shown considerable promise in supporting the regeneration of the peripheral nervous system. In the context of peripheral nerve injury, tissue-engineered nerve grafts (TENGs) have been employed as living scaffolds to accelerate axonal regeneration and functional recovery[82] [Fig. 5(f)]. By applying controlled mechanical tension in specialized bioreactors, axons can be stretch-grown into aligned tracts extending several centimeters. When implanted to bridge segmental gaps in injured peripheral nerves, these constructs have been shown to effectively guide host axon outgrowth over long distances and restore function in challenging injury models. Notably, TENGs have demonstrated superior regenerative capacity compared to synthetic nerve guidance conduits and performed comparably to autografts—the current clinical gold standard (TableI).[82]

Table I Summary of Neural Interface Bioelectronics
Full size table

Perspective

Despite remarkable advances in neural interface technologies, challenges remain in achieving seamless, stable, and biologically integrated brain–machine communication. Among these, long-term biocompatibility is paramount. Immune responses, along with mechanical mismatches between neural probes and the soft, dynamic brain tissue, may affect long-term stability and pose challenges for clinical translation. Moreover, translating neurotechnologies from lab to clinic requires not only optimization of device performance in controlled settings, but also more rigorous validation of long-term safety, manufacturing scalability, and compatibility with surgical procedures and patient lifestyles. Bridging this gap necessitates interdisciplinary collaboration spanning materials science, synthetic biology, neuroscience, and clinical studies.

The emerging field of regenerative and living bioelectronics introduces a paradigm shift—embedding functional biological components such as stem cells, genetically engineered tissues, or synthetic protein scaffolds directly into devices. Biointerfaces may soon be capable of in situ synthesis of functional networks, enabling the dynamic remodeling of the neural–device interface.[84] This approach can not only improve biocompatibility but also create devices that evolve alongside host tissue.

Looking forward, the next generation of neural interfaces will likely embody a convergence of intelligence, adaptability, and biological integration.[85] One key direction is the development of closed-loop systems powered by real-time decoding of neural signals through artificial intelligence (AI). These AI-integrated interfaces can dynamically adjust stimulation patterns and data acquisition strategies in response to evolving physiological states, thereby offering more effective interventions for neurological disorders.

Complementing this vision is the concept of organoid intelligence (OI),[86] a strategy of biological computing that uses brain organoids as a living unit for information processing. These living systems can learn, adapt, and potentially interact with electronic networks in real time.[87] Though still in its infancy, OI offers an avenue for rethinking how biological substrates and engineered systems can co-process information and create hybrid neurobioelectronic platforms.

Altogether, the future of neural interfaces lies in the fusion of biology and electronics, not merely as components that coexist, but as systems that co-evolve, regenerate, and adapt. These advances can finally enable neurotechnologies to integrate into the brain not as passive sensors, but as active, living, learning extensions of the nervous system.