Three-dimensional microinstrumented neural network device

Abstract

The present disclosure provides a device for interfacing neurons cultured in three dimensions, comprising a flexible scaffold structure with a flexible insulating layer folded to form multiple levels. The scaffold has mesh portions with apertures allowing three-dimensional neural networks to extend through. The device includes electrically-conducting patterns encased within the insulating layer and exposed in selected regions to form electrodes for neural interaction. The insulating layer may comprise SU-8 polymer and conducting patterns may comprise gold. The device enables stable extracellular action potential recordings from multiple planes, allowing quantitative tracing of neural network connectivity maps and real-time tracking of pharmacological effects. It further enables systematic tuning of connectivity strengths between neurons via electrical stimulations, establishing methodologies for three-dimensional spatiotemporal monitoring and modulation of neural networks.

Description

[0001] PRIN- 104576

[0002] THREE-DIMENSIONAL MICROINSTRUMENTED NEURAL NETWORK DEVICE

[0003] CROSS-REFERENCE TO RELATED APPLICATIONS

[0004] This application claims priority to U.S. Application No. 63 / 716,375, filed November 5, 2024, and to U.S. Application No. 63 / 903,414, filed October 22, 2025, each of which is hereby incorporated by reference in its entirety.

[0005] STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT

[0006] This invention was made with government support under Grant No. 1R21EB035681-01 awarded by the National Institutes of Health. The government has certain rights in the invention.

[0007] TECHNICAL FIELD

[0008] The present disclosure relates to three-dimensional neural network devices, and more particularly to a three-dimensional microinstrumented neural network device with flexible electrodes for long-term mapping and modulation of cultured neural networks.

[0009] BACKGROUND

[0010] Three-dimensional neural cultures have emerged as powerful tools for studying brain function and neural network dynamics. These systems offer advantages over traditional two- dimensional cell cultures by more closely mimicking the complex three-dimensional architecture and connectivity patterns found in native brain tissue. In three-dimensional neural cultures, neurons can form connections in all spatial directions, creating network structures that better represent the physiological environment of the brain.

[0011] The study of neural networks in three-dimensional environments provides insights into neural development, synaptic plasticity, and disease mechanisms. Researchers can observe how neurons migrate, differentiate, and establish functional connections within a controlled three-dimensional matrix. These cultures allow for the investigation of neural circuit formation and the analysis of electrical activity patterns that emerge as networks mature over time.

[0012] Monitoring electrical activity in three-dimensional neural cultures presents technical challenges. Traditional planar electrode arrays can only access neurons at the surface of three- dimensional cultures, limiting the ability to record from cells distributed throughout the PRIN- 104576 volume. This constraint restricts the comprehensive analysis of network- wide electrical activity and connectivity patterns that develop within the three-dimensional structure.

[0013] Long-term stability of recording interfaces represents another consideration in three- dimensional neural culture studies. As neural networks develop and mature over weeks or months, maintaining consistent electrical contact between recording electrodes and neurons becomes challenging. Changes in culture morphology, cell migration, and tissue remodeling can affect the quality and reliability of electrical recordings over extended periods.

[0014] The ability to both record from and stimulate neurons within three-dimensional cultures would enable bidirectional studies of neural network function. Such capabilities could facilitate investigations of how electrical stimulation influences network development, synaptic strength, and connectivity patterns. Additionally, the combination of recording and stimulation could support studies of neural plasticity and the effects of pharmacological interventions on network activity.

[0015] Applications of three-dimensional neural culture systems extend to drug discovery, disease modeling, and biocomputing research. These platforms can serve as models for studying neurological disorders, screening therapeutic compounds, and developing brain- inspired computing systems. However, realizing these applications requires robust methods for interfacing with neurons distributed throughout three-dimensional culture volumes.

[0016] SUMMARY

[0017] This summary is provided to introduce a selection of concepts in a simplified form that are further described below in the detailed description. This summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter.

[0018] According to an aspect of the present disclosure, a device for interfacing neurons cultured in three dimensions is provided. The device comprises a flexible scaffold structure composed of a flexible insulating layer, the flexible scaffold structure being folded to form a main portion having a plurality of levels. The flexible scaffold structure has mesh portions defining apertures, each aperture configured to allow a portion of a three-dimensional neural network in a neural culture to extend through the aperture. The device includes multiple electrically-conducting patterns encased within the flexible insulating layer, wherein the multiple electrically-conducting patterns are exposed in selected regions to form a plurality of electrodes allowing the neural culture to interact with the plurality of electrodes. This configuration provides the technical advantage of enabling three-dimensional volumetric PRIN- 104576 access to neurons distributed throughout a neural culture, allowing for comprehensive monitoring and stimulation of neural networks that more closely mimic the natural brain architecture compared to traditional planar electrode arrays.

[0019] According to another aspect of the present disclosure, the flexible insulating layer comprises a biocompatible epoxy-based polymer. This material selection provides the technical advantage of ensuring biocompatibility for long-term neural culture applications while maintaining the mechanical flexibility needed for folding into three-dimensional structures without compromising electrical insulation properties.

[0020] According to another aspect of the present disclosure, each electrode has an effective diameter of about 10 pm to about 50 pm. This size range provides the technical advantage of optimizing the balance between spatial resolution for single-neuron recording and sufficient electrode surface area for stable electrical contact and low impedance measurements.

[0021] According to another aspect of the present disclosure, the plurality of levels comprises 3 to 10 levels. This range provides the technical advantage of achieving sufficient three- dimensional coverage of neural networks while maintaining manageable device complexity and fabrication feasibility.

[0022] According to another aspect of the present disclosure, adjacent levels are separated by a gap of about 10 pm to about 500 pm. This spacing provides the technical advantage of allowing neurons to grow and form connections between levels while preventing mechanical interference between adjacent scaffold layers and maintaining structural integrity of the folded device.

[0023] According to another aspect of the present disclosure, the multiple electrically- conducting patterns further define multiple wires, each wire having a width of about 5 pm to about 25 pm. This wire width range provides the technical advantage of minimizing electrical resistance while maintaining sufficient mechanical flexibility for folding operations and reducing the overall footprint of the interconnect network.

[0024] According to another aspect of the present disclosure, the device further comprises input-output pads electrically coupled to the multiple electrically-conducting patterns. This feature provides the technical advantage of enabling reliable electrical connections to external recording and stimulation equipment while providing sufficient contact area for stable longterm electrical interfaces.

[0025] According to another aspect of the present disclosure, the ratio of electrodes to inputoutput pads is 1. This one-to-one ratio provides the technical advantage of enabling PRIN- 104576 independent control and monitoring of each electrode, maximizing the spatial resolution and flexibility of neural network analysis.

[0026] According to another aspect of the present disclosure, the ratio of electrodes to inputoutput pads is greater than 1. This configuration provides the technical advantage of enabling multiplexed electrode arrangements that can reduce the number of external connections while still providing access to multiple recording sites.

[0027] According to another aspect of the present disclosure, the mesh portions define apertures configured to allow neurons to form three-dimensional connections throughout the neural culture. This design provides the technical advantage of promoting natural neural network formation in three dimensions while ensuring that the scaffold structure does not impede normal cellular processes and connectivity patterns.

[0028] According to another aspect of the present disclosure, the flexible scaffold structure is configured to interpenetrate with a three-dimensional neural network while maintaining the cytoarchitecture of the neural network. This capability provides the technical advantage of enabling intimate device-tissue integration without disrupting the natural organization and connectivity of the neural network, thereby preserving physiologically relevant network dynamics.

[0029] According to another aspect of the present disclosure, the device further comprises spacers between adjacent layers to maintain a predetermined gap. This feature provides the technical advantage of ensuring consistent layer spacing during and after the folding process, which maintains uniform access to neural networks across all levels and prevents layer collapse that could compromise device functionality.

[0030] According to another aspect of the present disclosure, the neural culture comprises only a single type of neuron. This configuration provides the technical advantage of enabling controlled studies of specific neuronal populations without the complexity introduced by multiple cell types, facilitating focused analysis of particular neural mechanisms or responses.

[0031] According to another aspect of the present disclosure, the neural culture comprises a plurality of types of neurons. This configuration provides the technical advantage of creating more physiologically relevant neural networks that better represent the cellular diversity found in natural brain tissue, enabling studies of complex neural interactions and network dynamics.

[0032] According to another aspect of the present disclosure, the device further comprises a substrate coupled to the flexible scaffold structure. This substrate provides the technical advantage of providing mechanical support and stability for the device while serving as a platform for external electrical connections and handling during fabrication and use. PRIN- 104576

[0033] According to another aspect of the present disclosure, the substrate comprises fused silica. This material selection provides the technical advantage of offering excellent optical transparency for microscopy applications, low electrical conductivity for signal isolation, and chemical inertness for long-term stability in biological environments.

[0034] According to another aspect of the present disclosure, the substrate is coupled to at least a portion of the flexible insulating layer on a first level of the flexible scaffold structure. This coupling arrangement provides the technical advantage of providing secure mechanical anchoring for the folded structure while allowing the remaining levels to maintain flexibility for optimal neural network integration.

[0035] According to another aspect of the present disclosure, the device further comprises one or more retainer walls coupled to the substrate and positioned to define a volume for containing the neural culture. This feature provides the technical advantage of creating a defined culture environment that contains the neural network and culture medium while preventing contamination and enabling controlled experimental conditions.

[0036] According to another aspect of the present disclosure, the device further comprises one or more retainer caps permanently or removably coupled to the retainer wall with a gap and positioned to define a volume for containing the neural culture. This feature provides the technical advantage of creating a defined culture environment that contains the neural network and culture medium while preventing contamination, allowing exchange of gases and enabling controlled experimental conditions.

[0037] According to another aspect of the present disclosure, the one or more retainer walls comprise a retainer ring. This circular configuration provides the technical advantage of creating a uniform culture environment without comers that could create stagnant regions or uneven cell distribution, promoting homogeneous neural network development.

[0038] According to another aspect of the present disclosure, a method of producing a device for interfacing neurons cultured in three dimensions is provided. The method comprises fabricating a planar polymer scaffold on a base substrate using an insulating polymer, depositing a plurality of conductive patterns on the insulating polymer, encapsulating the plurality of conductive patterns with the insulating polymer, removing insulation from the planar polymer scaffold at selected locations to form electrodes, partially releasing the planar polymer scaffold from the base substrate, and folding the planar polymer scaffold into a plurality of layers to form a three-dimensional flexible electrode system. This fabrication approach provides the technical advantage of enabling precise control over electrode placement PRIN- 104576 and interconnect routing while utilizing established microfabrication techniques for reproducible device production.

[0039] According to another aspect of the present disclosure, the insulating polymer comprises a biocompatible epoxy-based polymer. This material choice provides the technical advantage of ensuring compatibility with biological systems while providing the photolithographic processability needed for precise pattern definition and the mechanical properties required for folding operations.

[0040] According to another aspect of the present disclosure, depositing the plurality of conductive patterns comprises depositing gold patterns having a thickness of about 25 nm to about 200 nm. This thickness range provides the technical advantage of ensuring adequate electrical conductivity for low-noise recordings while maintaining sufficient mechanical flexibility for folding without pattern cracking or delamination.

[0041] According to another aspect of the present disclosure, depositing the gold patterns further comprises depositing a chromium adhesion layer having a thickness of about 3 nm to about 25 nm. This adhesion layer provides the technical advantage of ensuring strong bonding between the gold conductive patterns and the underlying insulating polymer, preventing delamination during folding and long-term use in aqueous biological environments.

[0042] According to another aspect of the present disclosure, partially releasing the planar polymer scaffold comprises etching a sacrificial layer positioned between the planar polymer scaffold and the base substrate. This release mechanism provides the technical advantage of enabling controlled partial release of the scaffold structure while maintaining attachment points needed for subsequent folding operations.

[0043] According to another aspect of the present disclosure, the sacrificial layer comprises nickel having a thickness of about 25 nm to about 200 nm. This material and thickness range provides the technical advantage of enabling selective etching without affecting other device materials while providing sufficient mechanical support during fabrication processes.

[0044] According to another aspect of the present disclosure, folding the planar polymer scaffold comprises creating spacers between adjacent layers to maintain a predetermined gap. This approach provides the technical advantage of ensuring consistent layer spacing and preventing layer collapse, which maintains uniform access to neural networks across all device levels.

[0045] According to another aspect of the present disclosure, the three-dimensional flexible electrode system comprises 3 to 10 layers with electrodes distributed across multiple vertical planes. This configuration provides the technical advantage of achieving comprehensive three- PRIN- 104576 dimensional coverage of neural networks while maintaining manageable device complexity and fabrication yield.

[0046] According to another aspect of the present disclosure, a method for detecting or stimulating neural network connectivity is provided. The method comprises providing a device comprising a flexible scaffold structure folded to form multiple levels with apertures allowing three-dimensional neural network growth and electrodes distributed across the levels, creating a suspension comprising neural cells and depositing the suspension in a three-dimensional volume defined by the device, allowing neurons to interweave through the apertures within the three-dimensional volume, and recording electrical signals from the neurons using the electrodes to analyze neural network connectivity and development. This methodology provides the technical advantage of enabling comprehensive analysis of three-dimensional neural network formation and function over extended time periods with single-neuron resolution.

[0047] According to another aspect of the present disclosure, the suspension comprises a solubilized basement membrane matrix at a concentration of about 2 mg / ml to about 15 mg / ml and neural cells at a density of about 500,000 cells per ml to about 2,000,000 cells per ml. These concentration ranges provide the technical advantage of creating an optimal environment for three-dimensional neural network formation while ensuring sufficient cell density for robust network connectivity without overcrowding that could impede normal cellular processes.

[0048] According to another aspect of the present disclosure, the method further comprises applying electrical stimulation to only a subset of the plurality of electrodes to modulate neural network connectivity. This selective stimulation capability provides the technical advantage of enabling targeted manipulation of specific neural pathways and connections, allowing for controlled studies of neural plasticity and network modification.

[0049] According to another aspect of the present disclosure, the electrical stimulation comprises biphasic pulses having a duration of about 0.1 millisecond to about 2 milliseconds and applied at a frequency of about 0.1 Hz to about 100 Hz. These stimulation parameters provide the technical advantage of promoting neural plasticity and connectivity changes while avoiding tissue damage or excessive depolarization that could disrupt normal neural function.

[0050] According to another aspect of the present disclosure, recording electrical signals comprises monitoring action potentials from multiple levels simultaneously over a period of time to track neural network development. This capability provides the technical advantage of PRIN- 104576 enabling comprehensive longitudinal studies of neural network maturation and connectivity evolution with spatial resolution across the three-dimensional culture volume.

[0051] According to another aspect of the present disclosure, the period of time is at least 3 months. This extended monitoring duration provides the technical advantage of capturing the full developmental trajectory of neural networks, including initial formation, maturation, and long-term stability phases that are relevant for understanding neural development and disease progression.

[0052] According to another aspect of the present disclosure, a method for drug discovery or development is provided. The method comprises providing a device comprising a flexible scaffold structure with electrodes distributed across multiple levels interfacing with a three- dimensional neural network, recording baseline electrical activity from the neural network using the electrodes, applying a test compound to the neural network, recording electrical activity from the neural network after compound application, and analyzing changes in electrical activity to evaluate pharmacological effects of the test compound. This approach provides the technical advantage of enabling comprehensive assessment of drug effects on three-dimensional neural networks that more closely represent natural brain tissue organization compared to traditional two-dimensional screening platforms.

[0053] According to another aspect of the present disclosure, the method further comprises a step of washing the test compound from the neural network and recording recovery electrical activity to assess reversibility of the pharmacological effects. This additional step provides the technical advantage of distinguishing between reversible and irreversible drug effects, which provides valuable information for drug safety assessment and mechanism of action studies.

[0054] According to another aspect of the present disclosure, the test compound comprises one of a neurotoxin or an antagonist of a neural receptor. This compound selection provides the technical advantage of enabling studies of specific neural pathways and receptor systems, facilitating targeted drug discovery efforts and mechanistic understanding of neural function.

[0055] According to another aspect of the present disclosure, analyzing changes in electrical activity comprises measuring changes in firing rate, burst frequency, and connectivity patterns between neurons across the multiple levels. This comprehensive analysis approach provides the technical advantage of capturing multiple dimensions of neural network function, enabling detection of subtle drug effects that might be missed by single-parameter measurements.

[0056] According to another aspect of the present disclosure, the three-dimensional neural network comprises only a single type of neuron. This configuration provides the technical advantage of enabling focused studies of drug effects on specific neuronal populations without PRIN- 104576 confounding effects from multiple cell types, facilitating clear interpretation of pharmacological mechanisms.

[0057] According to another aspect of the present disclosure, the three-dimensional neural network comprises a plurality of types of neurons. This configuration provides the technical advantage of creating more physiologically relevant drug screening platforms that better represent the cellular diversity and complex interactions found in natural neural tissue.

[0058] According to another aspect of the present disclosure, a method for biocomputing is provided. The method comprises providing a device comprising a flexible scaffold structure with electrodes distributed across multiple levels interfacing with a three-dimensional neural network, applying spatially patterned electrical stimulations to the neural network through one or more of the electrodes, recording neural responses from one or more electrodes, and processing the neural responses to perform classification of input patterns, wherein connectivity strengths between neurons are modulated through the electrical stimulations to train the neural network. This approach provides the technical advantage of creating biological computing systems that leverage the natural computational capabilities of neural networks while providing precise control over network training and input-output relationships.

[0059] According to another aspect of the present disclosure, the electrical stimulation comprises biphasic pulses having a duration of about 0.1 millisecond to about 2 milliseconds and applied at a frequency of about 0.1 Hz to about 1 Hz (can be much higher, ~100 Hz). These stimulation parameters provide the technical advantage of promoting synaptic plasticity and learning in the neural network while maintaining physiological stimulation levels that preserve network health and function.

[0060] According to another aspect of the present disclosure, the spatially patterned electrical stimulations are applied to different combinations of electrodes to create distinct input patterns for classification. This approach provides the technical advantage of enabling complex pattern recognition tasks by utilizing the spatial distribution of electrodes to create diverse input representations that can be learned and distinguished by the neural network.

[0061] According to another aspect of the present disclosure, the input patterns comprise simultaneous stimulation of at least two electrodes on multiple levels of the flexible scaffold structure. This multi-level stimulation strategy provides the technical advantage of utilizing the three-dimensional architecture of the device to create rich, spatially distributed input patterns that better exploit the computational capabilities of three-dimensional neural networks.

[0062] According to another aspect of the present disclosure, processing the neural responses comprises converting spike trains into binned arrays and applying logistic regression analysis PRIN- 104576 to classify the input patterns. This data processing approach provides the technical advantage of translating biological neural activity into digital representations suitable for machine learning analysis while maintaining the temporal and spatial information content of the neural responses.

[0063] The foregoing general description of the illustrative embodiments and the following detailed description thereof are merely exemplary aspects of the teachings of this disclosure and are not restrictive.

[0064] BRIEF DESCRIPTION OF FIGURES

[0065] Non-limiting and non-exhaustive examples are described with reference to the following figures.

[0066] FIG. 1 illustrates a perspective view of a three-dimensional microinstrumented neural network device, according to aspects of the present disclosure.

[0067] FIG. 2 illustrates a cross-sectional view of an embodiment of an insulated strand, including the insulating sublayers, a metal patterned layer, and exposed portions of a top insulating sublayer to form electrodes.

[0068] FIG. 3 illustrates a cross-sectional view of another embodiment of an insulated strand.

[0069] FIG. 4 illustrates a side view of a device of FIG. 1 positioned on a substrate, according to aspects of the present disclosure.

[0070] FIG. 5 illustrates a folding sequence for assembling a three-dimensional scaffold structure.

[0071] FIG. 6 illustrates a cross-sectional view of a portion of the flexible scaffold structure of FIG. 5 after all folding is complete.

[0072] FIG. 7 illustrates an isometric view of an embodiment of the neural network device, according to aspects of the present disclosure.

[0073] FIG. 8 is a plot showing representative recording traces from Day 51 to Day 150. Averaged sorted single-unit APs of corresponding traces are shown on the right column.

[0074] FIGS. 9A-9D are plots showing analyses of burst activities over six months, including (9 A) the evolution of burst frequency, (9B) the change in burst duration, (9C) the average number of spikes contained in each burst, and (9D) the mean interspike interval within the bursts.

[0075] FIG. 10 is a plot showing sorted spike trains from 16 representative sensors distributed in four different vertical layers at 49 DIV. PRIN- 104576

[0076] FIGS. 11A-1 IB are plots showing representative recording traces across four different vertical layers — one representative channel from each layer — from Day 38 to Day 72. Sensor 223 ’s data on Day 72 is omitted due to a malfunction in the commercial readout setup.

[0077] FIG. 12 is a plot showing averaged sorted APs from the corresponding sensors and days from FIGS. 11A-1 IB.

[0078] FIG. 13. Is a plot showing example recording traces (top) and sorted spike trains (bottom) obtained simultaneously from sensors 133 in layer 1 and 424 in layer 4 at Day 49.

[0079] FIG. 14 shows representative trace and perievent raster plot of three sensors, 322 (layer 3, blue), 114 (layer 1) and 231 (layer 2) in response to electrical stimuli applied to electrode 231 (vertical line represents stimulation time). Each small vertical line in the top right and bottom plots represents a sorted spike event from the corresponding sensor (n=24 stimulation trials). There is a blank space in the 231 distribution because of stimulation artifacts.

[0080] FIG. 15 is a plot showing evolution of the weight between the neurons recorded by sensors 322 and 421 over multiple days in response to electrical stimulation applied to 231.

[0081] FIG. 16A is a graph showing changes in prediction accuracy with training using 2D and 3D distributed sensors.

[0082] FIGS. 16B-16C are graphs showing the evolving connectivity strength, including a connection getting stronger (16B) and weaker (16C), in response to stimulations; weights were normalized with respect to the corresponding pre-stimulation weights of each pair.

[0083] FIG. 17 is a schematic for spatial classification, where 3D-MIND was stimulated through spatially distributed stimulators and the observed activity was used to create an array X that was fed to a logistic regression model. The regression model predicted the spatial distribution of the stimulator.

[0084] FIG. 18 is an illustration of input X for a logistic regression model. Activities observed from different sensors in response to the stimulations were used to obtain spike trains which were then binned into 2 ms windows for a duration of 300 ms. A perievent histogram was created with the start of the pulse as the trigger event. The counts / bin of this histogram served as the activity XI for sensor I. Activities from sensors were then appended at the end of each other to create an array X that was used as the input to the logistic regression model. Each X acts as a data point for classification.

[0085] DETAILED DESCRIPTION

[0086] The following description sets forth exemplary aspects of the present disclosure. It should be recognized, however, that such description is not intended as a limitation on the PRIN- 104576 scope of the present disclosure. Rather, the description also encompasses combinations and modifications to those exemplary aspects described herein.

[0087] Referring to FIG. 1, a device for interfacing neurons cultured in three dimensions may comprise a flexible scaffold 100. The flexible scaffold 100 may be composed of a flexible insulating layer that provides structural support while maintaining biocompatibility with neural tissue. In some cases, the flexible scaffold 100 may be configured to interface with biological neural networks in a three-dimensional environment.

[0088] The flexible scaffold 100 may include a main portion 102 that forms the primary structural component of the device. The main portion 102 may be folded to create a plurality of levels that extend in different vertical planes. In some aspects, the main portion 102 may comprise a first level 106, a second level 108, and a third level 110, each positioned at different heights within the three-dimensional structure. The multiple levels may provide spatial distribution of sensing and stimulation capabilities throughout the volume occupied by neural tissue.

[0089] As shown in FIG. 1, the flexible scaffold 100 may include a mesh portion 112 that defines one or more apertures 114. Each aperture 114 may be configured to allow portions of a three-dimensional neural network 116 to extend through the structure. The apertures 114 may facilitate the interpenetration of the device with neural tissue, allowing neurons to form connections in multiple directions while maintaining contact with the device components.

[0090] The neural network 116 may comprise cultured neurons that grow and form connections in three dimensions. In some cases, the neural network 116 may extend through the apertures 114, creating a distributed network that spans multiple levels of the flexible scaffold 100. The three-dimensional configuration may allow the neural network 116 to develop connectivity patterns that more closely resemble those found in natural brain tissue.

[0091] With continued reference to FIG. 1, the device may include multiple electrically- conducting patterns that may be encased within the flexible insulating layer of the flexible scaffold 100. A flexible pattern 118 may represent one such electrically-conducting pattern that extends through the structure. The flexible pattern 118 may be configured to provide electrical connectivity between different components of the device while maintaining the mechanical flexibility of the overall structure.

[0092] The multiple electrically-conducting patterns may be exposed in selected regions to form a plurality of electrodes. An electrode 120 may be formed where the electrically - conducting pattern is exposed, allowing direct electrical contact with the neural network 116. The electrode 120 may be configured to record electrical signals from neurons or to deliver PRIN- 104576 electrical stimulation to neural tissue. In some aspects, multiple electrodes may be distributed across the different levels of the flexible scaffold 100, providing volumetric access to the three- dimensional neural network 116.

[0093] The device may further include an input / output pad 122 that may be electrically coupled to the flexible pattern 118. The input / output pad 122 may provide an interface for connecting the device to external electronic systems for signal recording or stimulation delivery. In some cases, the input / output pad 122 may be positioned outside the region containing the neural network 116 to facilitate connection to measurement or stimulation equipment while maintaining the integrity of the biological culture environment.

[0094] Referring to FIG. 2, the flexible scaffold structure may comprise multiple layers that provide both structural support and electrical functionality. The flexible scaffold structure may include a substrate 202 that serves as a base for supporting the other components of the device. In some cases, the substrate 202 may comprise materials such as fused silica or glass that provide dimensional stability during fabrication and use.

[0095] A bottom flexible insulating sublayer 204 may be positioned on the substrate 202. The bottom flexible insulating sublayer 204 may comprise a biocompatible epoxy-based polymer that provides mechanical flexibility while maintaining electrical insulation properties. In some aspects, the bottom flexible insulating sublayer 204 may be formed from SU-8 polymer, which exhibits suitable biocompatibility for interfacing with neural tissue.

[0096] SU-8 polymer is a negative photoresist material that may be used in micro fabrication processes. In some aspects, SU-8 comprises an epoxy-based polymer that exhibits biocompatibility properties suitable for biological applications. The polymer may be processed using photolithography techniques, where exposure to ultraviolet light causes cross-linking of the polymer chains, resulting in a stable, insoluble structure in the exposed regions.

[0097] In some cases, SU-8 may provide mechanical flexibility while maintaining structural integrity, making it suitable for applications requiring bendable or foldable components. The material may exhibit low water absorption and good chemical resistance, which can be beneficial in biological environments. SU-8 may be deposited in various thicknesses, ranging from micrometers to hundreds of micrometers, depending on the specific application requirements.

[0098] The polymer may be processed at relatively low temperatures, typically below 200°C, which can help preserve the integrity of underlying structures or materials during fabrication. In some aspects, SU-8 may serve as both a structural material and an insulating layer, providing PRIN- 104576 electrical isolation between conductive elements while maintaining the desired mechanical properties for the overall device structure

[0099] As shown in FIG. 2, a metal sublayer 206 may be disposed on the bottom flexible insulating sublayer 204. The metal sublayer 206 may form portions of the flexible pattern 118 and may define the electrode 120 and input / output pad 122. In some cases, the metal sublayer 206 may comprise gold or other conductive materials that provide stable electrical contact with biological systems.

[0100] A top insulating sublayer 208 may cover portions of the metal sublayer 206 while leaving selected regions exposed. The top insulating sublayer 208 may comprise the same biocompatible epoxy-based polymer as the bottom flexible insulating sublayer 204. In some aspects, the top insulating sublayer 208 may provide electrical isolation between different conductive elements while maintaining the overall flexibility of the structure.

[0101] With continued reference to FIG. 2, an exposed area 210 may be formed where the top insulating sublayer 208 has been removed to expose portions of the metal sublayer 206. The exposed area 210 may correspond to locations where the electrode 120 or input / output pad 122 requires direct electrical contact with the surrounding environment.

[0102] Referring to FIG. 3, the layered structure of the flexible scaffold structure may be configured to provide both mechanical and electrical properties suitable for neural interfacing applications. The substrate 202 may provide a stable foundation for the multilayer structure, while the bottom flexible insulating sublayer 204 and top insulating sublayer 208 may encapsulate the conductive elements of the flexible pattern 118.

[0103] The flexible insulating layer of the flexible scaffold structure may be processed using standard photolithography techniques that may achieve close to 100% fabrication yield. In some cases, the SU-8 polymer used for the flexible insulating layer may be processed with specific baking temperatures including 65°C and 95°C during different stages of the photolithography process. These processing temperatures may provide proper cross-linking of the polymer while maintaining the desired mechanical and electrical properties.

[0104] The SU-8 polymer may be treated with oxygen plasma exposure to create a hydrophilic surface. In some aspects, the hydrophilic surface treatment may enhance the biocompatibility of the flexible scaffold structure and may improve the interaction between the device and neural tissue. The oxygen plasma treatment may modify the surface chemistry of the SU-8 polymer without affecting the bulk mechanical properties of the material.

[0105] As further shown in FIG. 3, the flexible scaffold structure may be configured to be folded to form the main portion 102 having the plurality of levels. The layered composition PRIN- 104576 including the bottom flexible insulating sublayer 204 and top insulating sublayer 208 may provide the mechanical flexibility needed for the folding process while maintaining electrical connectivity through the flexible pattern 118. The biocompatible epoxy-based polymer composition may allow the structure to be folded into complex three-dimensional configurations without compromising the integrity of the electrical connections or the biocompatibility of the device.

[0106] With continued reference to FIG. 1, the mesh portion 112 of the flexible scaffold 100 may be configured with specific dimensional parameters that facilitate integration with the neural network 116. The mesh portion 112 may comprise scaffold ribbons that define the structural framework of the device. In some cases, the scaffold ribbons may have a width of approximately 25 micrometers, providing sufficient structural support while maintaining flexibility for the folding operations.

[0107] The mesh portion 112 may include longitudinal gaps between adjacent scaffold ribbons. These longitudinal gaps may have a width of approximately 50 micrometers, creating openings that allow portions of the neural network 116 to extend through the structure. The longitudinal gaps may be oriented in a first direction to provide pathways for neural growth and connectivity.

[0108] As further shown in FIG. 1, the mesh portion 112 may also include transverse gaps between scaffold ribbons. The transverse gaps may have a width of approximately 75 micrometers and may be oriented perpendicular to the longitudinal gaps. The combination of longitudinal and transverse gaps may create a grid- like pattern of apertures 114 throughout the mesh portion 112.

[0109] The apertures 114 defined by the mesh portion 112 may be configured to allow portions of the three-dimensional neural network 116 in a neural culture to extend through each aperture 114. The dimensional parameters of the apertures 114 may be selected to accommodate the growth patterns of neurons while providing adequate structural support for the flexible scaffold 100. In some aspects, the apertures 114 may allow neurons to form three-dimensional connections throughout the neural culture by providing pathways for neurite extension and synaptic formation.

[0110] The flexible scaffold structure may be configured to interpenetrate with the three- dimensional neural network 116 while maintaining the cytoarchitecture of the neural network 116. The interpenetration may occur as neurons grow through the apertures 114 and form connections that span multiple levels of the flexible scaffold 100. In some cases, the mesh PRIN- 104576 configuration may allow the neural network 116 to develop natural connectivity patterns without disruption from the device structure.

[0111] The cytoarchitecture of the neural network 116 may be preserved through the specific sizing and spacing of the apertures 114. The 50 micrometer longitudinal gaps and 75 micrometer transverse gaps may provide sufficient space for neuronal cell bodies and dendritic arbors to develop naturally. The mesh portion 112 may act as a supportive framework that guides neural growth while allowing the formation of complex three-dimensional network topologies that resemble those found in natural brain tissue.

[0112] Referring to FIG. 5, the flexible scaffold structure may be configured for folding to create the main portion 102 having the plurality of levels. An intermediate flexible scaffold structure 500 may represent the device in a partially assembled state during the folding process. The intermediate flexible scaffold structure 500 may include a bottom scaffold layer 502 that forms the foundation for the multi-level assembly.

[0113] The bottom scaffold layer 502 may be positioned on the substrate 202 and may incorporate portions of the flexible pattern 118 that extend through the structure. In some cases, the bottom scaffold layer 502 may serve as the first level 106 of the final assembled device. The bottom scaffold layer 502 may include the mesh portion 112 with apertures 114 configured to allow portions of the neural network 116 to extend through the structure.

[0114] As shown in FIG. 5, a spacer 506 may be positioned along the bottom scaffold layer 502 to define the separation distance between adjacent levels in the folded configuration. The spacer 506 may comprise the same biocompatible epoxy -based polymer as the flexible insulating layer, providing consistent material properties throughout the device structure. In some aspects, the spacer 506 may have a controlled thickness that determines the gap between adjacent levels when the structure is folded.

[0115] The intermediate flexible scaffold structure 500 may include a folding line 508 that defines the location where the structure may be bent to create the multi-level configuration. The folding line 508 may be positioned to allow portions of the flexible pattern 118 to extend continuously through the folded structure while maintaining electrical connectivity between different levels. In some cases, the folding line 508 may be oriented to facilitate the creation of parallel levels that are separated by predetermined gaps.

[0116] With continued reference to FIG. 5, a folding direction 510 may indicate the direction of movement during the assembly process. The folding direction 510 may be selected to create a stacked configuration where each level is positioned above the previous level with consistent spacing maintained by the spacers. The folding process may be performed while the structure PRIN- 104576 is submerged in deionized water to facilitate gentle manipulation of the flexible scaffold structure.

[0117] Additional scaffold layers 512 may be created through successive folding operations along multiple folding lines. The additional scaffold layers 512 may correspond to the second level 108 and third level 110 shown in FIG. 1. In some aspects, the additional scaffold layers 512 may be positioned to create a total of 3 to 10 levels in the final assembled device. The number of levels may be selected based on the desired volumetric coverage of the neural network 116 and the specific application requirements.

[0118] As further shown in FIG. 5, multiple spacers may be positioned between the additional scaffold layers 512 to maintain predetermined gaps throughout the structure. A spacer 514 may be positioned between a first pair of adjacent levels, while a spacer 516 may be positioned between a second pair of adjacent levels. A spacer 518 may be positioned between a third pair of adjacent levels. Each spacer may have a controlled thickness that defines the separation distance between the corresponding adjacent levels.

[0119] The spacers may have controlled thicknesses ranging from 25 micrometers to 250 micrometers. In some cases, the spacer thickness may be selected to achieve specific interlayer spacing requirements for the neural network 116. The controlled thickness of the spacers may allow the device to be configured with adjacent levels separated by a gap of about 10 micrometers to about 100 micrometers. The gap dimensions may be selected to accommodate the three-dimensional growth patterns of neurons while maintaining adequate structural support for the flexible scaffold structure.

[0120] Referring to FIG. 6, the spacer configuration may be illustrated in a cross-sectional view showing the relationship between the flexible pattern 118 and the multiple spacers. The spacer 506 may be positioned at a lower portion of the structure, while the spacer 514, spacer 516, and spacer 518 may be positioned at upper portions corresponding to different levels of the folded structure.

[0121] The spacers may maintain separation between adjacent sections of the flexible pattern 118, creating defined gaps between the layers of the structure. In some aspects, the spacers may prevent the layers from sagging or collapsing during use, ensuring that the apertures 114 remain open for neural network growth and that the electrodes 120 maintain proper positioning relative to the neural tissue.

[0122] With continued reference to FIG. 6, the total vertical dimension of the assembled device may be controlled by the number of levels and the thickness of the spacers. The device may be assembled into three-dimensional tissue-like scaffolds spanning a total vertical PRIN- 104576 dimension from approximately 100 micrometers to approximately 1 millimeter. The total vertical coverage may be determined by the sum of the spacer thicknesses and the thickness of the individual scaffold layers.

[0123] The folding process may be enabled through the use of a sacrificial layer that is selectively removed during fabrication. The device may use sacrificial nickel layer etching with TFB wet etchant to enable folding of the scaffold structure. The sacrificial nickel layer may be positioned between the flexible scaffold structure and the substrate 202 during fabrication. When the nickel layer is etched using TFB wet etchant, portions of the flexible scaffold structure may become free-standing, allowing the folding operations to be performed.

[0124] The TFB wet etchant may selectively remove the nickel while leaving the biocompatible epoxy -based polymer and the conductive elements of the flexible pattern 118 intact. In some cases, the etching process may create a free-standing end of the flexible scaffold structure that can be manipulated during the folding process while maintaining attachment to the substrate 202 at specific anchor points. The selective etching may allow precise control over which portions of the structure become free-standing and which portions remain attached to provide structural stability during and after the folding process.

[0125] Referring to FIG. 4, the device may further comprise a substrate 404 that may be coupled to the flexible scaffold 100. The substrate 404 may provide a stable foundation for supporting the flexible scaffold structure during fabrication, assembly, and operation. In some cases, the substrate 404 may serve as a mounting platform that maintains the structural integrity of the device while allowing the flexible scaffold 100 to be positioned in the desired three- dimensional configuration.

[0126] The substrate 404 may comprise fused silica, which may provide dimensional stability and biocompatibility suitable for neural culture applications. In some aspects, fused silica may exhibit low thermal expansion properties and chemical inertness that can be beneficial for maintaining device performance over extended periods. The fused silica substrate 404 may be processed using standard microfabrication techniques, allowing for precise dimensional control and surface preparation.

[0127] As shown in FIG. 4, the substrate 404 may be coupled to at least a portion of the flexible insulating layer on the first level 106 of the flexible scaffold structure. The coupling between the substrate 404 and the flexible scaffold 100 may provide mechanical anchoring while allowing other portions of the flexible scaffold structure to be folded into the multi-level configuration. In some cases, the coupling may be achieved through direct adhesion between the flexible insulating layer and the substrate surface. PRIN- 104576

[0128] The flexible scaffold 100 may include an upper surface 402 that may be positioned away from the substrate 404. The upper surface 402 may correspond to portions of the flexible scaffold structure that are not directly coupled to the substrate 404, allowing these portions to be manipulated during the folding process. In some aspects, the upper surface 402 may include portions of the mesh portion 112 and the apertures 114 that facilitate integration with the neural network 116.

[0129] With continued reference to FIG. 4, the device may further comprise one or more retainer walls 406 that may be coupled to the substrate 404. The retainer walls 406 may be positioned to define a volume for containing the neural culture. In some cases, the retainer walls 406 may be offset from the flexible scaffold 100 to create a containment area that surrounds the device while providing access for neural culture media and biological materials.

[0130] The retainer walls 406 may comprise a retainer ring configuration that may encircle the flexible scaffold 100. The retainer ring may be fabricated from acrylic material, which may provide chemical resistance and dimensional stability suitable for biological culture applications. In some aspects, the acrylic retainer ring may be machined or molded to achieve the desired dimensions and surface finish for containing neural culture media.

[0131] As further shown in FIG. 4, the retainer walls 406 may define an inner volume of space 408 that may be configured to contain the neural culture. The inner volume of space 408 may encompass the flexible scaffold 100 and may provide a controlled environment for the growth and development of the neural network 116. In some cases, the inner volume of space 408 may be filled with culture media and biological materials that support neural cell viability and network formation.

[0132] The retainer walls 406 may be attached to the substrate 404 using PDMS (polydimethylsiloxane) adhesive. The PDMS attachment may provide a biocompatible seal between the retainer walls 406 and the substrate 404, preventing leakage of culture media while maintaining the integrity of the biological environment within the inner volume of space 408. In some aspects, the PDMS adhesive may be cured at room temperature or elevated temperatures to achieve the desired bond strength and sealing properties.

[0133] The retainer ring configuration may create wells for media that facilitate the maintenance of neural cultures over extended periods. The wells may be designed with specific volume capacities that accommodate the media requirements for the neural network 116 while providing adequate space for the three-dimensional growth of neural tissue through the apertures 114 of the flexible scaffold 100. In some cases, the well configuration may allow for PRIN- 104576 media exchange and monitoring of culture conditions without disrupting the neural network development.

[0134] The device may include one or more retainer caps 410 which may be permanently or removably coupled to the retainer wall. The caps may be configured to provide a gap to allow the exchange of gases and enabling controlled experimental conditions. The caps may cover the retainer wall(s), defining the volume for containing the neural culture. In some cases, the gap may include radial and / or axial gaps between the retainer cap and the retainer wall in order to provide a path for gases to flow (e.g., if the cap is not a tight-fitting or sealed cap, there will be gaps through which gases can flow). In some cases, the gap may include a channel 412 defined on an outer surface of at least a portion of the retainer cap 410 to allow gases to flow in or out through the volume of space. In some cases, the gap may include at a hole extending through the cap. The hole may be operably coupled to a valve.

[0135] The positioning of the retainer walls 406 relative to the flexible scaffold 100 may be selected to optimize the interaction between the device and the neural culture. The retainer walls 406 may be positioned at a sufficient distance from the flexible scaffold 100 to allow unrestricted growth of the neural network 116 while maintaining containment of the culture media within the inner volume of space 408. The spacing may also facilitate access to the input / output pad 122 for electrical connections to external measurement or stimulation equipment.

[0136] The neural culture used with the device may comprise various types of neuronal cells that can be cultured in three-dimensional environments. In some cases, the neural culture may comprise only a single type of neuron, providing a homogeneous cellular environment for studying specific neuronal behaviors and connectivity patterns. A single-type neural culture may allow for controlled investigation of particular neuronal characteristics without the complexity introduced by multiple cell types.

[0137] In other aspects, the neural culture may comprise a plurality of types of neurons, creating a heterogeneous cellular environment that more closely resembles the complexity found in natural brain tissue. A multi-type neural culture may include different neuronal subtypes that exhibit distinct electrophysiological properties, morphologies, and connectivity patterns. The plurality of types of neurons may interact through various synaptic connections, creating complex network dynamics that can be monitored and modulated using the device.

[0138] The neural culture may be created using rat hippocampal tissues dissociated from 17- day pregnant Sprague Dawley rats. The hippocampal tissues may be extracted from embryonic rats at this developmental stage, which may provide neurons that are suitable for three- PRIN- 104576 dimensional culture applications. In some cases, the 17-day embryonic stage may correspond to a developmental period when hippocampal neurons exhibit optimal viability and growth characteristics for in vitro culture systems.

[0139] The dissociation process may involve enzymatic treatment of the hippocampal tissues to separate individual neurons from the tissue matrix. The dissociated neurons may then be prepared for three-dimensional culture by combining with appropriate matrix materials and culture media. In some aspects, the dissociated rat hippocampal neurons may include various neuronal subtypes naturally present in hippocampal tissue, including pyramidal neurons and interneurons.

[0140] The neural culture suspension may comprise Matrigel. As used herein, the term "Matrigel" may refer to a gelatinous protein mixture that forms a three-dimensional extracellular matrix. Matrigel may comprise a solubilized basement membrane preparation extracted from mouse sarcoma cells, containing various extracellular matrix proteins including laminin, collagen IV, heparan sulfate proteoglycans, and entactin. In some aspects, Matrigel may provide a biologically active matrix that supports cell adhesion, differentiation, and growth in three-dimensional culture systems. The material may remain liquid at low temperatures and may gel when warmed to physiological temperatures, allowing for the encapsulation of cells within a three-dimensional network that mimics the natural extracellular environment found in tissues.

[0141] The Matrigel may be at a concentration of, e.g., 1 mg / mL to 1 g / mL, such as 1 mg / mL to 100 mg / mL, such as 1 mg / mL to 10 mg / mL. The Matrigel concentration may provide an appropriate matrix density for supporting three-dimensional neuronal growth while maintaining sufficient porosity for nutrient diffusion and waste removal. In some cases, a 1 - 10 mg / mL concentration, such as of Matrigel, may preferably be used to a create a gel matrix that mimics the mechanical properties of natural brain tissue, providing structural support for neuronal processes and synaptic formation.

[0142] The neural culture suspension may further comprise a neuronal cell suspension at approximately 750,000 cells per ml. This cell density may provide adequate neuronal populations for network formation while avoiding overcrowding that could limit three- dimensional growth patterns. In some aspects, the 750,000 cells per ml density may allow neurons to establish connections with neighboring cells while maintaining sufficient spacing for the development of complex dendritic and axonal arbors.

[0143] The combination of Matrigel and neuronal cells may be used to create a suspension that can be deposited within the inner volume of space defined by the retainer walls. The PRIN- 104576 neuronal cells may be present at any appropriate concentration, such as at 50,000 cells / mL to 5,000,000 cells / mL, such as 500,000 cells / mL to 1,000,000 cells / mL. The suspension may be allowed to gel at physiological temperatures, encapsulating the neurons within the three- dimensional matrix while allowing the neurons to extend processes through the apertures of the flexible scaffold structure.

[0144] As used herein, the term "neuronal cell" may refer to a specialized cell type that forms the basic functional unit of the nervous system. A neuronal cell may be characterized by its ability to generate and transmit electrical signals through changes in membrane potential. The cell may comprise a cell body containing the nucleus and most organelles, dendrites that receive signals from other cells, and an axon that transmits signals to other neurons or target tissues. Neuronal cells may exhibit excitability through voltage-gated ion channels that allow rapid changes in membrane potential, enabling the propagation of action potentials. In some aspects, neuronal cells may form synaptic connections with other neurons, allowing for communication through chemical or electrical signaling. The cells may be derived from various sources including primary tissue dissociation, stem cell differentiation, or established cell lines, and may retain their characteristic electrophysiological properties when cultured in appropriate conditions.

[0145] The neural culture preparation may involve pre-treatment steps to enhance neuronal adhesion and growth. For example, the device surfaces may be treated with at least one amino acid polymer, such as an amino acid homopolymer. Advantageously, the device surfaces may be treated with at least poly-D-lysine. The device surfaces may be treated with an amino acid polymer at a concentration of 100 pg / mL to 1 mg / mL, such as 250 pg / mL to 1000 pg / mL, to promote neuronal attachment. In some cases, a second coating treatment may be used, which may include a glycoprotein, such as laminin, at concentrations of, e.g., 1 pg / mL to 500 pg / mL, such as 2 pg / mL to 100 pg / mL, such as 5 pg / mL to 20 pg / mL, to be applied over the first coat (and preferably over an amino acid homopolymer coating) to further enhance the biocompatibility of the device surfaces and support neuronal process extension.

[0146] The neural culture media may comprise neurobasal media along with one or more supplements. Such supplements may include, e.g., B-27, penicillin-streptomycin-glutamine, and nerve growth factor (NGF) 2.5S. These media components may provide the nutritional requirements for long-term neuronal culture while supporting the development and maintenance of synaptic connections. In some aspects, the supplemented neurobasal media may be optimized for three-dimensional neuronal culture applications, providing appropriate osmolarity and pH conditions for neuronal viability. PRIN- 104576

[0147] The neural culture may be maintained under controlled environmental conditions. Such conditions may include temperatures of 30°C - 40°C, and preferably about 37°C, and with a 1 %-l 0%, and preferably a 5%, CO2 atmosphere. These conditions may support optimal neuronal metabolism and growth while maintaining the pH stability of the culture media. In some cases, the controlled environment may be maintained using standard cell culture incubators that provide consistent temperature and gas composition throughout the culture period.

[0148] Referring to FIG. 7, the assembled device may provide functional interaction between the flexible scaffold structure with multiple levels, electrodes distributed across the levels, and apertures allowing neural network interpenetration. The three-dimensional configuration may enable comprehensive recording and stimulation of neurons within the neural network through volumetric access to the biological tissue. The isometric view shows the integrated system where the folded scaffold structure creates multiple planes of electrode interfaces that may interact with neurons distributed throughout the three-dimensional culture volume.

[0149] The device may enable stable extracellular action potential recordings for extended periods during neural culture development. In some cases, the device may maintain stable recordings for at least 90 days, 120 days, 150 days, or 180 days in vitro with no decay of signal amplitude. The long-term stability may result from the biocompatible interface between the flexible scaffold structure and the neural tissue, where the interpenetrating configuration allows electrodes to maintain consistent contact with neurons as the culture develops and matures.

[0150] The electrodes distributed across multiple levels may provide simultaneous monitoring of neural activity from different spatial locations within the three-dimensional neural network. The stable recording capability may allow for continuous tracking of individual neuron activity over months of culture development. In some aspects, the signal amplitude stability may indicate that the device maintains consistent electrical coupling with neural tissue without degradation of the electrode-tissue interface over time.

[0151] The device may track burst characteristics as the neural culture matures over extended periods. Burst activity may represent coordinated firing patterns among groups of neurons that indicate network-level connectivity and synchronization. The device may monitor burst frequency, which may change as neural networks develop more sophisticated connectivity patterns. In some cases, burst frequency may increase as neurons establish more synaptic connections and develop coordinated activity patterns.

[0152] Burst duration may be tracked to assess the temporal characteristics of network activity. The device may detect changes in burst duration as neural cultures mature, with duration PRIN- 104576 potentially decreasing as networks develop more efficient communication pathways. The number of spikes within each burst may provide information about the level of neuronal participation in coordinated activity events. As neural networks mature, the number of spikes per burst may change to reflect evolving connectivity patterns and synaptic strength distributions.

[0153] The device may monitor interspike interval characteristics within bursts to assess the temporal precision of neuronal firing patterns. Interspike intervals may change as neural networks develop more refined timing relationships between connected neurons. In some aspects, the tracking of these burst characteristics over months may provide insights into the developmental progression of three-dimensional neural networks and the establishment of functional connectivity patterns.

[0154] The device may monitor pharmacological responses to specific drugs with different mechanisms of action. For example, one or more neural receptor agonists or antagonists, such as bicuculline (which acts as a GABA receptor antagonist) or CNQX (which blocks AMPA and kainate receptors), may be applied to the neural culture to assess the device's capability to detect changes in inhibitory signaling or changes in excitatory synaptic transmission. The device may record increased neuronal firing rates following application of the agonist or antagonist, demonstrating the ability to detect pharmacological modulation of neural network activity. The device may record alterations in network activity patterns following CNQX application, indicating sensitivity to drugs that affect synaptic communication between neurons. The multi-level electrode configuration may allow simultaneous monitoring of drug effects across different spatial regions of the three-dimensional neural network.

[0155] Similarly, pore-blocking compounds, such as tetrodotoxin (which blocks sodium channels), may be applied to assess the device's sensitivity to changes in action potential generation. The device may detect immediate suppression of neuronal firing activities following application of the compound, indicating the capability to monitor drug effects that directly affect neuronal excitability. In some cases, the device may track the recovery of neuronal activity following compound removal, demonstrating the ability to monitor reversible pharmacological effects.

[0156] The device may apply electrical stimulation using biphasic pulses for chronic stimulation of neural networks. The stimulation parameters may include, e.g., 0.1 millisecond - 10 millisecond pulse duration, preferably 1 millisecond - 5 millisecond pulse duration. The charge duration may be selected to be sufficiently low (such as around 1 millisecond) so as to provide sufficient charge delivery for neuronal activation while minimizing tissue damage. A PRIN- 104576

[0157] 0.05 Hz - 1 Hz, such as around 0.2 Hz, repetition rate may allow for chronic stimulation protocols that can modulate synaptic plasticity over extended periods.

[0158] The biphasic pulse configuration may provide charge-balanced stimulation that minimizes electrochemical reactions at the electrode-tissue interface. In some aspects, the pulse duration may be selected to activate voltage-gated sodium channels in neurons while avoiding excessive charge injection that could damage neural tissue. The repetition rate may provide intermittent stimulation that allows for plasticity mechanisms to operate between stimulation events.

[0159] The device may use cross-correlogram analysis with Gaussian fitting to quantify connectivity strength between neurons. The cross-correlogram analysis may assess the temporal relationships between spike trains recorded from different electrodes, providing information about functional connectivity between neurons. Gaussian fitting may be applied to the cross-correlogram data to extract quantitative measures of connection strength.

[0160] The connectivity quantification may be expressed as a ratio of curve heights to standard deviations from the Gaussian fitting analysis. This ratio may provide a normalized measure of connectivity strength that accounts for both the magnitude of the correlation and the variability in the data. In some cases, higher ratios may indicate stronger functional connections between neuron pairs, while lower ratios may indicate weaker or absent connections.

[0161] The device may track connectivity evolution between specific neuron pairs over time in response to electrical stimulation. The tracking capability may reveal bidirectional plasticity effects, showing both strengthening and weakening of connections depending on the stimulation parameters and the initial state of the neural network. Some neuron pairs may exhibit increased connectivity strength following chronic stimulation, indicating long-term potentiation-like effects.

[0162] Other neuron pairs may show decreased connectivity strength following stimulation, indicating long-term depression- like effects or homeostatic plasticity mechanisms. The bidirectional plasticity tracking may demonstrate the device's capability to monitor complex adaptive changes in neural network connectivity patterns. In some aspects, the plasticity effects may be reversible, with connectivity strengths returning toward baseline levels when stimulation is discontinued.

[0163] The device may function as a reservoir neural network for classification tasks. The three-dimensional neural network may serve as a biological computing substrate that can be trained to recognize and classify input patterns. The device may apply spatially patterned PRIN- 104576 electrical stimulations to different combinations of electrodes to create distinct input patterns for the neural network to learn.

[0164] The training accuracies improve substantially (e.g., at least 20%) after only a few (e.g., 3-5) epochs of training, demonstrating the learning capability of the biological neural network interfaced with the device. The improvement in classification accuracy may indicate that the neural network develops enhanced discrimination capabilities through repeated exposure to the training patterns. In some cases, the learning process may involve changes in synaptic strengths between neurons that optimize the network's response to specific input patterns.

[0165] The device may demonstrate superior performance compared to two-dimensional counterparts in neural network interfacing applications. The three-dimensional electrode configuration may provide faster learning rates compared to planar electrode arrays that only interface with neurons at a single surface level. The volumetric access to the neural network may allow for more comprehensive monitoring and stimulation of neural activity patterns.

[0166] Higher classification accuracy may be achieved with the three-dimensional device interface compared to two-dimensional systems. The superior performance may result from the ability to access neurons distributed throughout the volume of the neural culture, rather than being limited to surface interactions. In some aspects, the three-dimensional interface may provide more diverse input pathways and monitoring points that enhance the computational capabilities of the biological neural network.

[0167] The multi-level electrode distribution may allow for more complex stimulation patterns that can engage different subpopulations of neurons simultaneously. The three-dimensional configuration may also provide better representation of the natural connectivity patterns found in brain tissue, leading to more effective learning and classification performance. The superior performance characteristics may demonstrate the advantages of three-dimensional neural interfaces for applications in biocomputing and neural network research.

[0168] A method of producing a device for interfacing neurons cultured in three dimensions may comprise several fabrication steps that create a three-dimensional flexible electrode system from planar starting materials. The method may enable the creation of devices that provide volumetric access to neural networks through folded scaffold structures with distributed electrodes.

[0169] The method may comprise fabricating a planar polymer scaffold on a base substrate using an insulating polymer. The planar polymer scaffold may serve as the foundation structure that will later be folded to create the three-dimensional configuration. In some cases, the base substrate may comprise fused silica or glass materials that provide dimensional stability during PRIN- 104576 the fabrication process. The insulating polymer may comprise a biocompatible epoxy-based polymer such as SU-8 that can be processed using standard photolithography techniques.

[0170] The fabrication of the planar polymer scaffold may involve spin-coating the insulating polymer onto the base substrate to achieve controlled thickness. The insulating polymer may be deposited in multiple layers with different thicknesses depending on the specific structural requirements. In some aspects, a bottom layer of the insulating polymer may be deposited with a thickness of approximately 2 micrometers to provide the base structural layer for the scaffold.

[0171] The method may further comprise depositing a plurality of conductive patterns on the insulating polymer. The conductive patterns may define the electrical pathways that will form electrodes, interconnecting wires, and input / output pads in the final device. In some cases, the conductive patterns may comprise gold deposited to a thickness of approximately 100 nanometers to provide stable electrical conductivity suitable for biological applications.

[0172] The deposition of the plurality of conductive patterns may involve photolithographic patterning followed by metal deposition techniques. A lift-off resist process may be used to define the pattern geometry before metal deposition. In some aspects, a chromium adhesion layer may be deposited prior to gold deposition to enhance the adhesion between the conductive patterns and the underlying insulating polymer. The chromium adhesion layer may have a thickness of approximately 10 nanometers.

[0173] The method may comprise encapsulating the plurality of conductive patterns with the insulating polymer. The encapsulation step may involve depositing additional layers of the insulating polymer over the conductive patterns to provide electrical isolation and mechanical protection. In some cases, the encapsulation may be achieved through spin-coating and photolithographic patterning of the insulating polymer to selectively cover the conductive patterns while leaving specific regions exposed.

[0174] The encapsulation process may create a multilayer structure where the conductive patterns are sandwiched between layers of the insulating polymer. The top encapsulation layer may be patterned to expose only the regions where electrical contact is desired, such as electrode sites and input / output pad areas. In some aspects, the encapsulation may provide mechanical flexibility while maintaining electrical isolation between different conductive elements.

[0175] The method may further comprise removing insulation from the planar polymer scaffold at selected locations to form electrodes. The removal of insulation may be accomplished through photolithographic patterning and development of the insulating polymer. In some cases, the selected locations may correspond to sites where direct electrical PRIN- 104576 contact with neural tissue is desired. The electrode formation may create exposed areas of the conductive patterns that can interface with the biological environment.

[0176] The electrode formation process may involve precise control of the exposed area dimensions to optimize the electrical interface with neural tissue. The electrodes may be formed with diameters ranging from approximately 10 micrometers to 50 micrometers. In some aspects, the electrode size may be selected to provide adequate signal-to-noise ratio for neural recording while minimizing tissue damage during implantation or culture integration.

[0177] The method may comprise partially releasing the planar polymer scaffold from the base substrate. The partial release may be achieved through selective removal of a sacrificial layer that was deposited between the planar polymer scaffold and the base substrate during the initial fabrication steps. In some cases, the sacrificial layer may comprise nickel that can be selectively etched using chemical etchants without affecting the insulating polymer or conductive patterns.

[0178] The partial release process may create free-standing regions of the planar polymer scaffold while maintaining attachment points that provide structural stability during subsequent folding operations. The selective etching may be controlled to release only specific portions of the scaffold, allowing other regions to remain anchored to the base substrate. In some aspects, the partial release may enable the folding operations while maintaining the integrity of the electrical connections and the overall device structure.

[0179] The method may comprise folding the planar polymer scaffold into a plurality of layers to form a three-dimensional flexible electrode system. The folding process may be performed while the scaffold is submerged in deionized water to facilitate gentle manipulation of the flexible structure. In some cases, the folding may be accomplished using pipette-generated water flow to guide the movement of the free-standing portions of the scaffold.

[0180] The folding operations may create multiple levels that are separated by predetermined gaps maintained through spacer structures. The spacers may be fabricated as part of the planar polymer scaffold and may have controlled thicknesses that define the interlayer spacing in the folded configuration. In some aspects, the folding process may create 3 to 10 layers with electrodes distributed across multiple vertical planes.

[0181] The three-dimensional flexible electrode system created through the folding process may provide volumetric access to neural networks cultured within the device structure. The folded configuration may allow electrodes to be positioned at different heights within the neural culture, enabling simultaneous recording and stimulation of neurons distributed throughout the three-dimensional tissue volume. In some cases, the flexible electrode system PRIN- 104576 may maintain electrical connectivity between all levels while providing mechanical flexibility suitable for biological applications.

[0182] The method may include additional processing steps to enhance the biocompatibility and functionality of the three-dimensional flexible electrode system. The completed device may be treated with oxygen plasma to create hydrophilic surfaces that improve interaction with biological materials. In some aspects, the device may be sterilized using ethanol treatment before integration with neural cultures.

[0183] The fabrication process may involve specific material selection and processing parameters that enable the creation of biocompatible three-dimensional neural interfaces. The method of producing a device for interfacing neurons cultured in three dimensions may utilize carefully selected materials and controlled processing conditions to achieve the desired structural and electrical properties.

[0184] The insulating polymer used in the method may comprise a biocompatible epoxy-based polymer that provides both mechanical flexibility and electrical insulation properties suitable for neural interfacing applications. Biocompatible epoxy-based polymers may exhibit low cytotoxicity and may support long-term contact with neural tissue without adverse biological effects. In some cases, the biocompatible epoxy-based polymer may comprise SU-8, which has been demonstrated to be suitable for biological applications due to its chemical stability and biocompatibility characteristics.

[0185] The biocompatible epoxy-based polymer may be processed using photolithographic techniques that allow precise control of structural dimensions and surface properties. The polymer may be spin-coated onto the base substrate at controlled speeds to achieve uniform thickness distribution across the substrate surface. In some aspects, the spin-coating process may be performed at speeds ranging from 1000 to 5000 revolutions per minute to achieve the desired thickness uniformity.

[0186] The processing of the biocompatible epoxy-based polymer may involve specific temperature profiles during soft-baking and post-exposure baking steps. The soft-baking may be performed at temperatures of 65°C and 95°C for controlled durations to remove solvents and prepare the polymer for photolithographic exposure. The post-exposure baking may be performed at similar temperatures to complete the cross-linking reactions that create the final polymer structure.

[0187] The method may comprise depositing the plurality of conductive patterns, where depositing the plurality of conductive patterns comprises depositing gold patterns having a thickness of about 25 nm to about 200 nm. The gold pattern thickness may be selected to PRIN- 104576 provide adequate electrical conductivity while maintaining mechanical flexibility of the overall structure. In some cases, gold patterns with thicknesses in the range of 25 nm to 200 nm may provide sufficient conductivity for neural signal recording and stimulation applications while allowing the structure to be folded without fracturing the conductive elements.

[0188] The deposition of gold patterns may be accomplished using thermal evaporation or sputtering techniques that provide controlled thickness uniformity across the substrate. The deposition rate may be controlled to achieve the desired thickness while maintaining good adhesion to the underlying polymer surface. In some aspects, the gold deposition may be performed at rates of 0.1 to 2 nanometers per second to achieve uniform coverage and minimize stress in the deposited film.

[0189] The gold patterns may be defined using lift-off photolithography processes that create precise geometric features for electrodes, interconnecting wires, and input / output pads. The lift-off process may involve the use of sacrificial photoresist layers that are removed after metal deposition to create the desired pattern geometry. In some cases, the lift-off process may be performed using solvents such as acetone or specialized lift-off solutions that selectively remove the sacrificial resist while leaving the gold patterns intact.

[0190] The method may further comprise depositing a chromium adhesion layer, where depositing the gold patterns further comprises depositing a chromium adhesion layer having a thickness of about 3 nm to about 25 nm. The chromium adhesion layer may be deposited prior to gold deposition to enhance the adhesion between the gold patterns and the biocompatible epoxy-based polymer substrate. In some aspects, the chromium adhesion layer thickness may be selected to provide adequate adhesion promotion while minimizing any potential effects on the electrical properties of the gold patterns.

[0191] The chromium adhesion layer may be deposited using the same deposition techniques as the gold patterns, such as thermal evaporation or sputtering. The deposition of the chromium adhesion layer may be performed immediately before gold deposition to minimize oxidation of the chromium surface that could affect adhesion properties. In some cases, the chromium deposition may be performed in a high-vacuum environment to achieve clean interfaces between the chromium, gold, and polymer layers.

[0192] The thickness of the chromium adhesion layer may be controlled to provide optimal adhesion without significantly affecting the overall electrical performance of the conductive patterns. Chromium layers with thicknesses of 3 nm to 25 nm may provide sufficient adhesion promotion while maintaining the desired electrical characteristics of the gold patterns. In some PRIN- 104576 aspects, thinner chromium layers may be preferred to minimize any potential diffusion of chromium into the gold layer during subsequent processing steps.

[0193] The encapsulation of the plurality of conductive patterns with the insulating polymer may involve depositing additional layers of the biocompatible epoxy-based polymer over the gold and chromium layers. The encapsulation process may use the same polymer material as the initial scaffold fabrication to ensure consistent material properties throughout the device structure. In some cases, the encapsulation layer may be deposited using spin-coating techniques similar to those used for the initial polymer layers.

[0194] The encapsulation layer thickness may be controlled to provide adequate electrical isolation between different conductive elements while maintaining the mechanical flexibility needed for the folding operations. The encapsulation process may involve photolithographic patterning to selectively expose regions where electrical contact is desired, such as electrode sites and input / output pad areas. In some aspects, the patterning of the encapsulation layer may be performed using the same photolithographic techniques used for the initial scaffold patterning.

[0195] The processing parameters for the encapsulation layer may include similar temperature profiles as used for the initial polymer layers, with soft-baking and post-exposure baking steps performed at controlled temperatures and durations. The encapsulation layer may be hard- baked at elevated temperatures to achieve full cross-linking and optimize the mechanical properties of the final structure. In some cases, the hard-baking may be performed at temperatures ranging from 185°C to 205°C for durations of 30 minutes to 2 hours.

[0196] The combination of the biocompatible epoxy-based polymer, gold patterns with controlled thickness, and chromium adhesion layers may create a multilayer structure that provides both the electrical functionality and mechanical properties needed for three- dimensional neural interfacing applications. The material selection and processing parameters may be optimized to achieve long-term stability in biological environments while maintaining the flexibility needed for folding operations and integration with neural cultures.

[0197] The method may comprise removing insulation from the planar polymer scaffold at selected locations to form electrodes through controlled photolithographic processing. The removal process may involve selective patterning of the encapsulation layer to expose predetermined regions of the conductive patterns where direct electrical contact with neural tissue is desired. In some cases, the insulation removal may be accomplished using photolithographic exposure and development techniques that precisely define the electrode geometry and dimensions. PRIN- 104576

[0198] The electrode formation process may utilize the same photolithographic chemistry and processing parameters as used for the initial scaffold patterning steps. The biocompatible epoxy-based polymer encapsulation layer may be selectively removed using appropriate developer solutions that dissolve the exposed polymer regions while leaving the unexposed areas intact. In some aspects, the development process may be controlled through timing and temperature parameters to achieve precise control over the electrode dimensions and surface quality.

[0199] The selected locations for insulation removal may correspond to sites distributed across the planar polymer scaffold that will become electrodes at different levels when the structure is folded into the three-dimensional configuration. The electrode sites may be positioned to provide optimal coverage of the neural culture volume after the folding operations are completed. In some cases, the electrode positioning may be designed to create uniform spatial distribution across multiple vertical planes in the final three-dimensional structure.

[0200] The method may comprise partially releasing the planar polymer scaffold, where partially releasing the planar polymer scaffold comprises etching a sacrificial layer positioned between the planar polymer scaffold and the base substrate. The sacrificial layer may serve as a temporary support structure during the fabrication process that can be selectively removed to enable the folding operations. In some aspects, the sacrificial layer may be patterned to define which regions of the planar polymer scaffold become free-standing and which regions remain anchored to the base substrate.

[0201] The etching process may utilize chemical etchants that selectively remove the sacrificial layer material without affecting the biocompatible epoxy-based polymer or the conductive patterns. The selective etching capability may allow precise control over the release pattern, creating free-standing regions that can be manipulated during folding while maintaining structural stability through retained anchor points. In some cases, the etching process may be performed using wet chemical etching techniques that provide controlled removal rates and high selectivity.

[0202] The sacrificial layer may comprise nickel having a thickness of about 25 nm to about 200 nm. The nickel sacrificial layer thickness may be selected to provide adequate support during the fabrication process while allowing complete removal during the etching step. In some aspects, nickel thicknesses in the range of 25 nm to 200 nm may provide sufficient mechanical support for the overlying polymer and metal layers during processing while being thin enough to allow rapid and uniform etching. PRIN- 104576

[0203] The nickel sacrificial layer may be deposited using physical vapor deposition techniques such as thermal evaporation or sputtering. The deposition process may be controlled to achieve uniform thickness distribution across the base substrate surface. In some cases, the nickel deposition may be performed at controlled rates to minimize stress in the deposited film and ensure good adhesion to the base substrate.

[0204] The nickel sacrificial layer may be patterned using photolithographic techniques to define the regions where the planar polymer scaffold will be released during the etching process. The patterning may create a selective release pattern that allows specific portions of the scaffold to become free-standing while maintaining attachment at designated anchor points. In some aspects, the patterning of the nickel layer may be performed using lift-off photolithography or etching processes that create the desired release geometry.

[0205] The etching of the nickel sacrificial layer may be accomplished using chemical etchants such as TFB (transene ferric chloride-based) etchant that provide high selectivity for nickel removal without affecting the polymer or gold materials. The etching process may be performed at controlled temperatures and concentrations to achieve uniform removal rates across the substrate. In some cases, the etching time may be controlled to ensure complete removal of the nickel layer while avoiding over-etching that could affect the underlying base substrate.

[0206] The method may comprise folding the planar polymer scaffold, where folding the planar polymer scaffold comprises creating spacers between adjacent layers to maintain a predetermined gap. The spacers may be fabricated as integral components of the planar polymer scaffold during the initial photolithographic processing steps. In some aspects, the spacers may comprise the same biocompatible epoxy-based polymer as the main scaffold structure, providing consistent material properties throughout the folded device.

[0207] The spacers may be formed through controlled deposition and patterning of additional polymer layers that create raised features with predetermined heights. The spacer height may determine the gap distance between adjacent layers in the folded configuration. In some cases, the spacers may be positioned at strategic locations along the scaffold structure to provide uniform support and prevent layer collapse during and after the folding operations.

[0208] The predetermined gap maintained by the spacers may be selected to accommodate the three-dimensional growth patterns of neural networks while providing adequate structural support for the folded scaffold. The gap dimensions may range from approximately 10 micrometers to 100 micrometers, depending on the specific application requirements and the characteristics of the neural culture being interfaced. In some aspects, the predetermined gap PRIN- 104576 may be optimized to allow unrestricted neural process extension while maintaining proper electrode positioning relative to the neural tissue.

[0209] The spacer creation process may involve depositing thicker layers of the biocompatible epoxy-based polymer at designated locations during the scaffold fabrication. The spacer thickness may be controlled through multiple deposition and patterning steps that build up the desired height incrementally. In some cases, the spacers may be formed using the same photolithographic processing techniques as the main scaffold structure, ensuring compatibility with the overall fabrication process.

[0210] The folding process enabled by the spacers may create a three-dimensional structure where each layer is separated by the predetermined gap distance. The spacers may prevent the layers from sagging or collapsing under their own weight or external forces during handling and use. In some aspects, the spacers may be positioned to maintain structural integrity while allowing the apertures in each layer to remain open for neural network interpenetration.

[0211] The method may result in a three-dimensional flexible electrode system comprising 3 to 10 layers with electrodes distributed across multiple vertical planes. The number of layers may be determined by the number of folding operations performed on the planar polymer scaffold. In some cases, the layer count may be selected based on the desired volumetric coverage of the neural culture and the specific monitoring or stimulation requirements of the application.

[0212] The three-dimensional flexible electrode system may provide electrodes positioned at different heights within the neural culture volume, enabling simultaneous access to neurons distributed throughout the three-dimensional tissue structure. The multiple vertical planes may allow comprehensive monitoring of neural activity patterns that span the entire culture volume rather than being limited to surface interactions. In some aspects, the electrode distribution across multiple vertical planes may provide enhanced spatial resolution for neural network mapping and modulation applications.

[0213] The electrodes distributed across the multiple vertical planes may maintain electrical connectivity through the conductive patterns that extend continuously through the folded structure. The folding process may be designed to preserve the integrity of the electrical connections while creating the desired three-dimensional electrode distribution. In some cases, the conductive patterns may be routed through the folding regions using serpentine or curved geometries that accommodate the mechanical deformation without fracturing.

[0214] The three-dimensional flexible electrode system may provide superior performance compared to planar electrode arrays that only interface with neural tissue at a single surface PRIN- 104576 level. The multiple layer configuration may enable more comprehensive neural network interfacing capabilities, including enhanced signal recording quality and more effective stimulation delivery. In some aspects, the three-dimensional electrode distribution may allow for more sophisticated neural network analysis and modulation protocols that take advantage of the volumetric access to the biological tissue.

[0215] The layer spacing in the three-dimensional flexible electrode system may be controlled through the spacer dimensions to optimize the interaction with specific types of neural cultures. Different neural culture types may have different optimal layer spacing requirements based on their growth characteristics and connectivity patterns. In some cases, the spacer dimensions may be customized during the fabrication process to create device variants optimized for different neural culture applications or research objectives.

[0216] A method for detecting or stimulating neural network connectivity may provide comprehensive analysis of three-dimensional neural networks through volumetric electrical interfacing. The method may enable simultaneous monitoring and modulation of neural activity patterns distributed throughout three-dimensional tissue volumes, providing capabilities that extend beyond conventional planar interfacing approaches.

[0217] The method for detecting or stimulating neural network connectivity may comprise providing a device comprising a flexible scaffold structure folded to form multiple levels with apertures allowing three-dimensional neural network growth and electrodes distributed across the levels. The device may serve as a platform that enables comprehensive electrical interfacing with neural networks cultured in three-dimensional configurations. In some cases, the flexible scaffold structure may provide mechanical support for the neural culture while maintaining biocompatible interfaces that support long-term neural viability and development.

[0218] The multiple levels of the flexible scaffold structure may be positioned at predetermined vertical separations that accommodate the natural growth patterns of three- dimensional neural networks. The apertures distributed throughout the scaffold structure may allow neural processes to extend freely between different levels, enabling the formation of complex connectivity patterns that span the entire culture volume. In some aspects, the electrodes distributed across the levels may provide electrical access points at different heights within the neural culture, enabling simultaneous monitoring of neural activity from spatially distributed locations.

[0219] The method may comprise creating a suspension comprising neural cells and depositing the suspension in a three-dimensional volume defined by the device. The suspension creation process may involve combining dissociated neural cells with appropriate matrix PRIN- 104576 materials that support three-dimensional culture growth. In some cases, the neural cells may be derived from primary tissue sources such as embryonic hippocampal tissue that provides neurons suitable for long-term culture applications.

[0220] The suspension may comprise neural cells at densities that promote network formation while avoiding overcrowding that could limit three-dimensional growth patterns. The neural cell density may be selected to achieve optimal connectivity development within the three- dimensional culture volume. In some aspects, the suspension may include matrix materials such as solubilized basement membrane preparations that provide structural support for neural growth and differentiation.

[0221] The deposition of the suspension in the three-dimensional volume defined by the device may involve controlled placement of the cell-matrix mixture within the containment area created by the device structure. The three-dimensional volume may be defined by retainer structures that maintain the suspension in contact with the flexible scaffold structure while providing adequate space for neural network development. In some cases, the deposition process may be performed under sterile conditions to maintain culture viability and prevent contamination.

[0222] The method may comprise allowing neurons to interweave through the apertures within the three-dimensional volume. The interweaving process may occur naturally as neurons extend processes and form connections during culture development. The apertures in the flexible scaffold structure may provide pathways that guide neural growth while allowing unrestricted formation of synaptic connections between neurons located at different levels within the culture.

[0223] The interweaving of neurons through the apertures may create a distributed network architecture where neural processes span multiple levels of the scaffold structure. The three- dimensional connectivity patterns that develop may more closely resemble the complex network topologies found in natural brain tissue compared to conventional two-dimensional culture systems. In some aspects, the interweaving process may be facilitated by the biocompatible surface properties of the scaffold structure that support neural adhesion and process extension.

[0224] The method may comprise recording electrical signals from the neurons using the electrodes to analyze neural network connectivity and development. The recording process may involve simultaneous monitoring of electrical activity from multiple electrodes distributed across different levels of the scaffold structure. In some cases, the electrical signals may include PRIN- 104576 action potentials, local field potentials, and other electrophysiological signals that provide information about neural network function and connectivity.

[0225] The analysis of neural network connectivity may involve processing the recorded electrical signals to extract information about functional relationships between neurons located at different spatial positions within the culture. Cross-correlation analysis techniques may be applied to identify temporal relationships between neural activity patterns that indicate synaptic connectivity or functional coupling. In some aspects, the connectivity analysis may reveal the development of network structures and the evolution of connectivity patterns over time.

[0226] The analysis of neural network development may involve tracking changes in electrical activity patterns as the neural culture matures over extended periods. The development analysis may include monitoring of firing rate changes, burst activity evolution, and connectivity strength modifications that occur during neural network maturation. In some cases, the development tracking may provide insights into the mechanisms underlying three-dimensional neural network formation and the establishment of functional connectivity patterns.

[0227] The method may enable detection of pharmacological effects on neural network connectivity through comparative analysis of electrical activity patterns before and after drug application. The detection capability may allow assessment of how various compounds affect neural network function and connectivity at the network level rather than individual cell responses. In some aspects, the method may provide a platform for drug screening applications that evaluate compound effects on three-dimensional neural network behavior.

[0228] The method may enable stimulation of neural network connectivity through controlled electrical stimulation delivered via selected electrodes within the scaffold structure. The stimulation capability may allow modulation of synaptic strength and connectivity patterns through chronic stimulation protocols. In some cases, the stimulation may be applied in spatially patterned configurations that target specific regions or pathways within the three- dimensional neural network.

[0229] The combination of recording and stimulation capabilities may enable closed-loop protocols where stimulation parameters are adjusted based on real-time analysis of neural network activity patterns. The closed-loop approach may allow optimization of stimulation protocols for specific connectivity modulation objectives. In some aspects, the method may enable training of biological neural networks for computational applications through controlled stimulation and monitoring of network responses.

[0230] The method for detecting or stimulating neural network connectivity may involve specific suspension preparation parameters that optimize three-dimensional neural network PRIN- 104576 formation and development. The suspension preparation process may utilize controlled concentrations of matrix materials and neural cells that promote optimal network connectivity while maintaining cell viability throughout extended culture periods.

[0231] The suspension may comprise a solubilized basement membrane matrix at a concentration of about 2 mg / ml to about 15 mg / ml. The solubilized basement membrane matrix concentration may be selected to provide adequate structural support for three-dimensional neural growth while maintaining sufficient porosity for nutrient diffusion and metabolic waste removal. In some cases, concentrations within the 2 mg / ml to 15 mg / ml range may create a gel matrix that mimics the mechanical properties of natural brain tissue extracellular environment.

[0232] Lower concentrations within this range, such as 2 mg / ml to 5 mg / ml, may provide a softer matrix that allows greater freedom for neural process extension and migration. Higher concentrations, such as 10 mg / ml to 15 mg / ml, may provide increased structural support that can be beneficial for maintaining three-dimensional architecture over extended culture periods. In some aspects, the optimal concentration may be selected based on the specific neural cell types being cultured and the desired mechanical properties of the final three-dimensional network.

[0233] The solubilized basement membrane matrix may undergo gelation when warmed to physiological temperatures, transitioning from a liquid state during suspension preparation to a solid gel state that encapsulates the neural cells within a three-dimensional network. The gelation process may occur rapidly upon temperature elevation, allowing the neural cells to become distributed throughout the three-dimensional matrix before significant settling or aggregation can occur. In some cases, the gelation kinetics may be controlled through temperature management during the suspension deposition process.

[0234] The suspension may comprise neural cells at a density of about 500,000 cells per ml to about 2,000,000 cells per ml. The neural cell density may be selected to achieve optimal balance between adequate cell-to-cell contact for network formation and sufficient spacing to allow three-dimensional growth patterns to develop. In some aspects, cell densities within this range may provide adequate neuronal populations for complex network formation while avoiding overcrowding that could limit dendritic and axonal development.

[0235] Lower cell densities, such as 500,000 cells per ml to 1,000,000 cells per ml, may allow individual neurons to develop extensive dendritic and axonal arbors without spatial constraints from neighboring cells. Higher cell densities, such as 1,500,000 cells per ml to 2,000,000 cells per ml, may promote more rapid network formation through increased probability of cell-to- PRIN- 104576 cell contact and synaptic formation. In some cases, the optimal cell density may depend on the specific neural cell types and the desired network connectivity characteristics.

[0236] The combination of the solubilized basement membrane matrix concentration and neural cell density may be optimized to achieve specific network formation objectives. The matrix concentration may influence the mechanical environment experienced by the neural cells, while the cell density may determine the spatial distribution and connectivity potential of the developing network. In some aspects, the ratio between matrix concentration and cell density may be adjusted to achieve desired network architectures and connectivity patterns.

[0237] The suspension preparation process may involve combining the solubilized basement membrane matrix and neural cells under controlled temperature conditions to prevent premature gelation. The mixing process may be performed at temperatures below the gelation threshold, typically at 4°C to 10°C, to maintain the suspension in a liquid state during preparation and handling. In some cases, the suspension components may be pre-cooled before mixing to ensure adequate working time for deposition into the device.

[0238] The deposition of the suspension in the three-dimensional volume defined by the device may be performed using controlled pipetting techniques that ensure uniform distribution throughout the containment area. The deposition process may be completed rapidly to minimize the time between suspension preparation and gelation initiation. In some aspects, the deposition may be performed in multiple aliquots to ensure complete filling of the three- dimensional volume while avoiding air bubble formation.

[0239] Following deposition, the suspension may be allowed to gel at physiological temperatures, typically 37°C, creating a three-dimensional matrix that encapsulates the neural cells within the device structure. The gelation process may occur within minutes of temperature elevation, creating a stable three-dimensional environment that supports neural cell viability and growth. In some cases, the gelled matrix may provide a scaffold that guides neural process extension while allowing the formation of complex three-dimensional connectivity patterns.

[0240] The method may comprise allowing neurons to interweave through the apertures within the three-dimensional volume to form connections. The interweaving process may begin shortly after gelation as neurons extend processes and begin exploring their three-dimensional environment. The apertures in the device structure may provide pathways that facilitate neural process extension between different levels of the scaffold, enabling the formation of connections that span the entire culture volume.

[0241] The connection formation process may involve multiple stages of neural development including neurite outgrowth, pathfinding, and synaptogenesis. The three-dimensional matrix PRIN- 104576 environment may support each of these developmental stages by providing appropriate mechanical and biochemical cues. In some aspects, the interweaving of neural processes through the apertures may create network architectures that exhibit greater complexity and connectivity diversity compared to conventional two-dimensional culture systems.

[0242] The neural cells may establish both local connections with nearby neurons within the same matrix region and long-range connections with neurons located at different levels of the device structure. The apertures may serve as conduits that allow axonal projections to reach target neurons located at distant positions within the three-dimensional culture volume. In some cases, the interweaving process may result in the formation of layered connectivity patterns that resemble the laminar organization found in natural brain tissue.

[0243] The connection formation process may be influenced by the spatial distribution of neural cells within the matrix and the accessibility provided by the aperture configuration. The device structure may guide the development of specific connectivity patterns through the strategic placement of apertures and the dimensional characteristics of the scaffold levels. In some aspects, the interweaving process may be monitored through the electrical recording capabilities of the device, allowing real-time observation of network development and connectivity establishment.

[0244] The three-dimensional neural network that develops through the interweaving process may exhibit connectivity patterns and activity dynamics that more closely resemble those found in natural brain tissue compared to conventional culture systems. The volumetric distribution of connections may enable the emergence of complex network behaviors including synchronized activity patterns, propagating waves, and hierarchical connectivity structures. In some cases, the three-dimensional network architecture may support computational capabilities that can be harnessed for biocomputing applications.

[0245] The method for detecting or stimulating neural network connectivity may further comprise applying electrical stimulation to only a subset of the plurality of electrodes to modulate neural network connectivity. The selective stimulation approach may allow targeted modulation of specific neural pathways or regions within the three-dimensional neural network while leaving other areas unstimulated for comparative analysis. In some cases, the subset of electrodes may be selected based on their spatial distribution within the scaffold structure to achieve desired stimulation patterns that target particular network connectivity pathways.

[0246] The application of electrical stimulation to only a subset of the plurality of electrodes may enable spatially patterned stimulation protocols that can selectively influence connectivity development between specific neuronal populations. The unstimulated electrodes may serve as PRIN- 104576 control recording sites that allow monitoring of network responses to the targeted stimulation. In some aspects, the selective stimulation approach may allow investigation of how localized electrical inputs affect global network connectivity patterns and activity dynamics.

[0247] The subset selection may be based on the spatial relationship between electrodes and the desired connectivity modulation objectives. Electrodes positioned at different levels of the scaffold structure may be selected to create stimulation patterns that span multiple vertical planes within the neural culture. In some cases, the subset may include electrodes that are strategically positioned to influence specific neural pathways or connectivity motifs within the three-dimensional network architecture.

[0248] The electrical stimulation applied to the selected subset of electrodes may comprise biphasic pulses having a duration of about 0.1 millisecond to about 2 milliseconds. The biphasic pulse configuration may provide charge-balanced stimulation that minimizes electrochemical reactions at the electrode-tissue interface while delivering sufficient charge to activate neural tissue. In some aspects, the biphasic design may prevent accumulation of charge at the electrode surface that could lead to tissue damage or electrode degradation over extended stimulation periods.

[0249] The pulse duration range of 0.1 millisecond to 2 milliseconds may be selected to optimize neural activation while minimizing energy consumption and potential tissue damage. Shorter pulse durations within this range, such as 0.1 millisecond to 0.5 milliseconds, may provide efficient neural activation with minimal charge injection. Longer pulse durations, such as 1 millisecond to 2 milliseconds, may be used when greater charge delivery is needed to achieve reliable neural activation in specific network configurations.

[0250] The biphasic pulse parameters may be optimized for different neural cell types and network configurations within the three-dimensional culture. The pulse amplitude may be adjusted in combination with the duration to achieve threshold activation of target neurons while avoiding activation of non-target neural populations. In some cases, the biphasic pulse characteristics may be customized for different electrodes within the selected subset to account for variations in electrode-tissue coupling and local neural density.

[0251] The electrical stimulation may be applied at a frequency of about 0.1 Hz to about 1 Hz. The stimulation frequency range may be selected to promote synaptic plasticity mechanisms while allowing adequate recovery time between stimulation events. In some aspects, frequencies within the 0.1 Hz to 1 Hz range may be compatible with natural neural activity patterns and may avoid interference with ongoing spontaneous network activity. PRIN- 104576

[0252] Lower frequencies within this range, such as 0.1 Hz to 0.3 Hz, may provide intermittent stimulation that allows plasticity mechanisms to operate between stimulation events without overwhelming the natural network dynamics. Higher frequencies, such as 0.5 Hz to 1 Hz, may provide more frequent stimulation inputs that can drive stronger plasticity effects while remaining within physiologically relevant frequency ranges. In some cases, the optimal stimulation frequency may depend on the specific connectivity modulation objectives and the baseline activity characteristics of the neural network.

[0253] The combination of biphasic pulse duration and stimulation frequency may be selected to achieve specific connectivity modulation outcomes. The temporal spacing between stimulation pulses may allow neural networks to respond and adapt to each stimulation event before the next pulse is delivered. In some aspects, the stimulation parameters may be adjusted over time to maintain effectiveness as the neural network develops and connectivity patterns evolve.

[0254] The method may comprise recording electrical signals, where recording electrical signals comprises monitoring action potentials from multiple levels simultaneously over a period of time to track neural network development. The simultaneous monitoring capability may provide comprehensive assessment of neural network activity patterns across the entire three-dimensional culture volume. In some cases, the multi-level recording may reveal spatial patterns of neural activity that would not be detectable using conventional single-level recording approaches.

[0255] The monitoring of action potentials from multiple levels may involve simultaneous data acquisition from electrodes distributed across different vertical planes within the scaffold structure. The multi-level recording capability may allow detection of activity propagation patterns that span the three-dimensional network architecture. In some aspects, the simultaneous monitoring may reveal temporal relationships between neural activity at different levels that indicate functional connectivity pathways within the network.

[0256] The action potential monitoring may provide information about individual neuron firing patterns as well as population-level activity dynamics. The multi-level recording approach may allow discrimination between local activity patterns that occur within individual scaffold levels and global activity patterns that involve coordination between multiple levels. In some cases, the monitoring may detect the emergence of complex activity patterns such as traveling waves or synchronized bursting that span the entire three-dimensional network volume. PRIN- 104576

[0257] The tracking of neural network development through multi-level action potential monitoring may reveal the temporal evolution of connectivity patterns and activity dynamics as the culture matures. The development tracking may include assessment of firing rate changes, connectivity strength evolution, and the emergence of coordinated activity patterns over extended time periods. In some aspects, the multi-level monitoring may provide insights into the mechanisms underlying three-dimensional neural network formation and functional maturation.

[0258] The period of time for monitoring action potentials and tracking neural network development may be at least 3 months. The extended monitoring period may allow observation of long-term developmental processes that occur over weeks to months in neural culture systems. In some cases, the 3 -month minimum monitoring period may encompass multiple phases of neural network development including initial connectivity establishment, synaptic refinement, and functional maturation.

[0259] The long-term monitoring capability may enable detection of gradual changes in network connectivity and activity patterns that occur over extended time scales. The stability of the device-tissue interface over at least 3 months may allow continuous tracking of the same neural populations throughout the development process. In some aspects, the extended monitoring period may reveal developmental milestones and critical periods in three- dimensional neural network formation that are not apparent in shorter-term studies.

[0260] The combination of electrical stimulation applied to selected electrode subsets and long-term multi-level recording may enable comprehensive investigation of how targeted interventions affect neural network development over extended periods. The stimulation protocols may be applied chronically throughout the monitoring period to assess cumulative effects on connectivity development and network maturation. In some cases, the long-term approach may reveal delayed or progressive effects of electrical stimulation that emerge over weeks or months following the initiation of stimulation protocols.

[0261] The method may enable assessment of bidirectional plasticity effects through the combination of selective stimulation and comprehensive monitoring. Some neural connections may exhibit strengthening in response to the stimulation protocols, while others may show weakening or homeostatic adjustments. In some aspects, the long-term monitoring may reveal the dynamic nature of connectivity modulation and the temporal evolution of plasticity effects within three-dimensional neural networks.

[0262] The multi-level recording capability combined with selective stimulation may allow investigation of how local stimulation inputs affect global network properties and connectivity PRIN- 104576 patterns. The stimulation of specific electrode subsets may induce changes in network activity that propagate throughout the three-dimensional structure and influence connectivity development at distant locations. In some cases, the comprehensive monitoring approach may reveal emergent network properties that arise from the interaction between targeted stimulation and natural developmental processes.

[0263] A method for drug discovery or development may provide a comprehensive platform for evaluating pharmacological effects of test compounds on three-dimensional neural networks through controlled electrical monitoring and analysis. The method may enable assessment of compound effects on neural network function at the network level rather than individual cellular responses, providing insights into how drugs affect complex neural connectivity patterns and activity dynamics in three-dimensional tissue architectures.

[0264] The method for drug discovery or development may comprise providing a device comprising a flexible scaffold structure with electrodes distributed across multiple levels interfacing with a three-dimensional neural network. The device may serve as a biological testing platform that enables comprehensive electrical interfacing with neural networks cultured in three-dimensional configurations that more closely resemble natural brain tissue organization compared to conventional two-dimensional culture systems. In some cases, the flexible scaffold structure may provide mechanical support for the neural culture while maintaining biocompatible interfaces that support long-term neural viability throughout extended drug testing protocols.

[0265] The electrodes distributed across multiple levels may provide volumetric access to the three-dimensional neural network, enabling simultaneous monitoring of neural activity from spatially distributed locations throughout the culture volume. The multi-level electrode configuration may allow detection of drug effects that occur at different spatial scales within the neural network, from local cellular responses to global network-level changes in connectivity and activity patterns. In some aspects, the distributed electrode arrangement may provide comprehensive coverage of the neural network that enables detection of subtle pharmacological effects that might be missed using conventional single-level recording approaches.

[0266] The three-dimensional neural network interfaced with the device may comprise cultured neurons that have developed complex connectivity patterns spanning multiple levels of the scaffold structure. The neural network may exhibit activity dynamics and connectivity characteristics that resemble those found in natural brain tissue, providing a physiologically relevant testing environment for drug evaluation. In some cases, the three-dimensional neural PRIN- 104576 network may serve as a biological substrate that responds to pharmacological interventions in ways that reflect the complex interactions between drugs and neural tissue in natural brain environments.

[0267] The method may comprise recording baseline electrical activity from the neural network using the electrodes. The baseline recording process may involve monitoring neural activity patterns under control conditions before any test compound application, establishing reference measurements that serve as comparison standards for evaluating drug effects. In some aspects, the baseline electrical activity may include various types of neural signals such as action potentials, burst activity patterns, and network-level synchronization events that characterize the normal functional state of the three-dimensional neural network.

[0268] The baseline recording period may be selected to capture representative samples of neural network activity that account for natural variations in firing patterns and connectivity dynamics. The recording duration may be sufficient to establish stable baseline measurements while avoiding excessive data collection that could complicate subsequent analysis procedures. In some cases, the baseline recording may be performed over multiple time periods to assess the stability and reproducibility of the neural network activity patterns before compound application.

[0269] The electrodes distributed across multiple levels may enable simultaneous baseline recording from different spatial regions within the three-dimensional neural network. The multi-level recording capability may provide comprehensive characterization of baseline activity patterns that span the entire culture volume rather than being limited to surface-level measurements. In some aspects, the baseline recording may reveal spatial patterns of neural activity and connectivity that serve as reference points for evaluating how test compounds affect different regions or pathways within the network.

[0270] The method may comprise applying a test compound to the neural network. The test compound application process may involve controlled introduction of the compound into the culture environment at predetermined concentrations and timing protocols. In some cases, the test compound may be dissolved in appropriate solvents or culture media to achieve desired concentrations while maintaining compatibility with the neural culture environment and device materials.

[0271] The test compound application may be performed using controlled delivery methods that ensure uniform distribution throughout the three-dimensional culture volume. The delivery process may involve replacement of culture media with compound-containing solutions or direct addition of concentrated compound solutions to achieve target concentrations. In some PRIN- 104576 aspects, the application method may be selected to minimize disruption of the neural network while ensuring adequate compound exposure throughout the three-dimensional tissue volume.

[0272] The test compound may comprise various types of pharmacologically active substances including neurotransmitter receptor agonists or antagonists, ion channel modulators, enzyme inhibitors, or other bioactive molecules that may affect neural function. The compound selection may be based on specific research objectives or drug development targets that require evaluation in three-dimensional neural network environments. In some cases, the test compound may represent novel therapeutic candidates that require assessment of their effects on complex neural network behaviors and connectivity patterns.

[0273] The method may comprise recording electrical activity from the neural network after compound application. The post-application recording process may involve monitoring neural activity patterns following test compound exposure to detect changes in network function and connectivity. In some aspects, the post-application recording may be performed using the same electrode configuration and recording parameters as the baseline measurements to enable direct comparison of activity patterns before and after compound exposure.

[0274] The timing of post-application recording may be selected based on the expected pharmacokinetics and mechanism of action of the test compound. Some compounds may produce immediate effects that can be detected within minutes of application, while others may require longer exposure periods to produce detectable changes in neural network activity. In some cases, the post-application recording may be performed at multiple time points to capture both acute and chronic effects of compound exposure on neural network function.

[0275] The post-application recording may utilize the multi-level electrode distribution to monitor compound effects across different spatial regions within the three-dimensional neural network. The distributed recording capability may enable detection of spatially heterogeneous drug effects that affect different network regions or connectivity pathways to varying degrees. In some aspects, the multi-level recording may reveal how compound effects propagate through the three-dimensional network architecture and influence connectivity patterns between different levels of the scaffold structure.

[0276] The method may comprise analyzing changes in electrical activity to evaluate pharmacological effects of the test compound. The analysis process may involve quantitative comparison of electrical activity patterns recorded before and after compound application to identify statistically significant changes that can be attributed to pharmacological effects. In some cases, the analysis may utilize various signal processing and statistical techniques to PRIN- 104576 extract meaningful information about compound effects from the complex electrical activity data recorded from multiple electrodes.

[0277] The analysis of changes in electrical activity may include assessment of various neural network parameters such as firing rate modifications, burst activity changes, connectivity strength alterations, and synchronization pattern shifts. The multi-parameter analysis approach may provide comprehensive characterization of compound effects that encompasses both individual neuron responses and network-level changes in activity dynamics. In some aspects, the analysis may reveal compound effects that manifest at different temporal and spatial scales within the three-dimensional neural network.

[0278] The evaluation of pharmacological effects may involve comparison of observed changes with known effects of reference compounds or established pharmacological profiles. The evaluation process may include assessment of dose-response relationships, time-course characteristics, and reversibility of compound effects. In some cases, the evaluation may involve statistical analysis to determine the significance and magnitude of observed changes relative to baseline variability and control measurements.

[0279] The method for drug discovery or development may provide advantages over conventional drug testing approaches through the use of three-dimensional neural networks that more closely resemble natural brain tissue organization. The three-dimensional culture environment may enable detection of compound effects that are not apparent in two- dimensional culture systems or isolated cell preparations. In some aspects, the method may provide more physiologically relevant assessment of drug effects that better predict how compounds may affect neural function in natural brain environments.

[0280] The multi-level electrode configuration may enable detection of compound effects that occur at different spatial scales within the neural network, from local cellular responses to global network-level changes. The comprehensive monitoring capability may allow identification of subtle pharmacological effects that might be missed using conventional singleelectrode or surface-level recording approaches. In some cases, the method may enable detection of compound effects on network connectivity and activity patterns that represent important aspects of neural function relevant to therapeutic applications.

[0281] The method for drug discovery or development may further comprise a step of washing the test compound from the neural network and recording recovery electrical activity to assess reversibility of the pharmacological effects. The washing step may involve removal of the compound-containing culture media and replacement with fresh, compound-free media to eliminate the test compound from the neural culture environment. In some cases, the washing PRIN- 104576 process may be performed multiple times using sequential media exchanges to ensure complete removal of the test compound and any metabolites that may have accumulated during the exposure period.

[0282] The washing procedure may be designed to minimize disruption of the neural network while achieving effective compound removal. The media exchange process may be performed using gentle pipetting techniques that avoid mechanical disturbance of the three-dimensional neural culture structure. In some aspects, the washing may involve multiple rinse cycles with fresh culture media, with each cycle allowing sufficient time for compound diffusion and equilibration before the next media exchange.

[0283] The recording of recovery electrical activity may be performed following the washing procedure to monitor the neural network's return to baseline activity patterns. The recovery recording may utilize the same multi-level electrode configuration used for baseline and postapplication measurements to enable direct comparison of activity patterns across all phases of the drug testing protocol. In some cases, the recovery recording may be conducted over extended time periods to capture both immediate and delayed recovery responses as the compound effects dissipate.

[0284] The assessment of reversibility of the pharmacological effects may involve quantitative comparison of recovery electrical activity patterns with the original baseline measurements recorded before compound application. The reversibility assessment may determine whether the neural network activity returns to pre-compound levels or exhibits persistent changes that indicate irreversible effects. In some aspects, the reversibility evaluation may include statistical analysis to determine the degree of recovery and identify any residual effects that persist following compound removal.

[0285] The recovery monitoring may reveal different temporal patterns of effect reversibility depending on the mechanism of action and pharmacokinetic properties of the test compound. Some compounds may produce effects that reverse rapidly following removal, while others may exhibit prolonged effects that require extended recovery periods. In some cases, the recovery assessment may distinguish between reversible functional effects and irreversible structural or toxic effects that permanently alter neural network properties.

[0286] The test compound may comprise one of a neurotoxin or an antagonist of a neural receptor. Neurotoxins may represent compounds that specifically target neural tissue and may produce various effects on neural function including disruption of action potential generation, synaptic transmission, or cellular metabolism. In some aspects, neurotoxins may serve as research tools for investigating specific aspects of neural function or as potential therapeutic PRIN- 104576 agents that require careful evaluation of their effects on neural network activity and connectivity.

[0287] The neurotoxin may comprise compounds such as tetrodotoxin, which may block voltage-gated sodium channels and prevent action potential generation in neurons. The application of such neurotoxins may produce rapid and dramatic suppression of neural activity that can be readily detected through the multi-level electrode monitoring system. In some cases, neurotoxin testing may provide information about the compound's potency, selectivity, and reversibility that is relevant for both research applications and potential therapeutic development.

[0288] An antagonist of a neural receptor may comprise compounds that bind to specific neurotransmitter receptors and block their normal function without activating the receptor. The receptor antagonists may target various neurotransmitter systems including glutamate, GABA, acetylcholine, or other signaling pathways that are involved in neural network function. In some aspects, receptor antagonists may produce more subtle effects on neural activity compared to neurotoxins, requiring sensitive detection methods to assess their pharmacological impact.

[0289] The antagonist of a neural receptor may comprise compounds such as bicuculline, which may block GABA receptors and reduce inhibitory signaling within the neural network. The application of such antagonists may produce changes in neural activity patterns including increased firing rates, altered burst characteristics, or modified connectivity patterns between neurons. In some cases, receptor antagonist testing may provide information about the role of specific neurotransmitter systems in maintaining neural network function and connectivity.

[0290] The analyzing of changes in electrical activity may comprise measuring changes in firing rate, burst frequency, and / or connectivity patterns between neurons across the multiple levels. The firing rate analysis may involve quantification of action potential frequency from individual neurons or neuronal populations recorded from different electrodes within the multilevel scaffold structure. In some aspects, firing rate measurements may provide sensitive indicators of compound effects on neural excitability and overall network activity levels.

[0291] The measurement of changes in firing rate may be performed through spike detection and counting algorithms applied to the electrical signals recorded from each electrode. The firing rate analysis may include assessment of both spontaneous activity levels and evoked responses to determine how test compounds affect different aspects of neural excitability. In some cases, the firing rate measurements may be normalized to baseline levels to enable PRIN- 104576 comparison of compound effects across different experimental conditions and neural network preparations.

[0292] The burst frequency analysis may involve detection and quantification of coordinated firing events that involve multiple neurons firing in temporal proximity. The burst frequency measurements may provide information about network-level coordination and synchronization that reflects the functional connectivity between neurons distributed across the three- dimensional culture. In some aspects, changes in burst frequency may indicate compound effects on synaptic transmission or network excitability that affect the ability of neurons to engage in coordinated activity patterns.

[0293] The measurement of burst frequency may utilize specialized algorithms that identify periods of elevated firing activity that exceed baseline levels by predetermined thresholds. The burst detection may account for the temporal characteristics of coordinated firing events and may distinguish between different types of burst patterns that occur within the neural network. In some cases, the burst frequency analysis may include assessment of burst duration, interburst intervals, and the number of participating neurons to provide comprehensive characterization of network activity changes.

[0294] The connectivity patterns between neurons across the multiple levels may be analyzed through cross-correlation techniques that assess temporal relationships between neural activity recorded from different electrodes. The connectivity analysis may reveal functional relationships between neurons located at different spatial positions within the three- dimensional neural network. In some aspects, changes in connectivity patterns may indicate compound effects on synaptic transmission or network organization that affect the functional relationships between different neural populations.

[0295] The measurement of connectivity patterns may involve calculation of cross-correlation functions between spike trains recorded from pairs of electrodes distributed across the multiple levels of the scaffold structure. The connectivity analysis may identify both direct synaptic connections and indirect functional relationships that reflect network-level interactions between neurons. In some cases, the connectivity measurements may be quantified using metrics such as correlation strength, temporal delay, and directionality to provide detailed characterization of network connectivity changes.

[0296] The three-dimensional neural network may comprise only a single type of neuron. The single neuron type configuration may provide a simplified experimental system that eliminates variability associated with different cell types and allows focused investigation of compound effects on specific neuronal populations. In some aspects, the single neuron type approach may PRIN- 104576 enable more precise characterization of compound effects by removing confounding factors associated with cellular heterogeneity and complex cell-type interactions.

[0297] The single type of neuron may comprise specific neuronal subtypes such as pyramidal neurons, interneurons, or other well-characterized cell populations that exhibit defined electrophysiological properties and connectivity characteristics. The homogeneous cellular composition may allow standardized assessment of compound effects across different experimental conditions and may facilitate comparison of results between different test compounds. In some cases, the single neuron type configuration may be particularly suitable for mechanistic studies that investigate specific aspects of neuronal function or drug action.

[0298] The three-dimensional neural network may comprise a plurality of types of neurons. The multi-type neuronal configuration may provide a more complex and physiologically relevant testing environment that better reflects the cellular diversity found in natural brain tissue. In some aspects, the plurality of types of neurons may include different neuronal subtypes that exhibit distinct electrophysiological properties, morphologies, and connectivity patterns that contribute to network complexity and functional diversity.

[0299] The plurality of types of neurons may comprise combinations of excitatory and inhibitory neurons that create balanced network activity patterns resembling those found in natural neural circuits. The multi-type configuration may enable investigation of compound effects on different neuronal populations and may reveal selective effects that preferentially affect specific cell types. In some cases, the plurality of types of neurons may include various interneuron subtypes, projection neurons, and other specialized cell populations that contribute to network function and connectivity.

[0300] The multi-type neural network configuration may exhibit more complex activity dynamics and connectivity patterns compared to single-type systems, potentially providing more sensitive detection of compound effects that manifest through interactions between different neuronal populations. The cellular diversity may enable assessment of compound selectivity and may reveal effects that are not apparent in homogeneous cell populations. In some aspects, the plurality of types of neurons may provide a testing environment that better predicts how compounds may affect neural function in natural brain tissue where multiple cell types interact to produce complex network behaviors.

[0301] A method for biocomputing may utilize three-dimensional neural networks as biological computing substrates that can be trained to perform pattern recognition and classification tasks through controlled electrical stimulation and monitoring protocols. The biocomputing method may leverage the natural computational capabilities of neural networks PRIN- 104576 while providing precise control over network training through spatially distributed electrical interfaces that can modulate connectivity patterns and learning processes.

[0302] The method for biocomputing may comprise providing a device comprising a flexible scaffold structure with electrodes distributed across multiple levels interfacing with a three- dimensional neural network. The device may serve as a biological computing platform that enables comprehensive electrical interfacing with neural networks cultured in three- dimensional configurations that support complex connectivity patterns and computational behaviors. In some cases, the flexible scaffold structure may provide mechanical support for the neural culture while maintaining biocompatible interfaces that enable long-term training protocols and computational applications.

[0303] The electrodes distributed across multiple levels may provide volumetric access to the three-dimensional neural network, enabling simultaneous stimulation and recording from spatially distributed locations throughout the culture volume. The multi-level electrode configuration may allow targeted activation of specific neural pathways or regions within the network while monitoring responses from other locations to assess network-wide computational responses. In some aspects, the distributed electrode arrangement may provide comprehensive coverage of the neural network that enables sophisticated training protocols and pattern recognition capabilities.

[0304] The three-dimensional neural network interfaced with the device may comprise cultured neurons that have developed complex connectivity patterns spanning multiple levels of the scaffold structure. The neural network may exhibit computational properties that can be harnessed for pattern recognition, classification, and other information processing tasks through appropriate training protocols. In some cases, the three-dimensional neural network may serve as a biological processor that can learn to discriminate between different input patterns and produce specific output responses based on training experience.

[0305] The method may comprise applying spatially patterned electrical stimulations to the neural network through one or more of the electrodes. The spatially patterned stimulation approach may involve selective activation of specific electrode combinations to create distinct input patterns that the neural network can learn to recognize and classify. In some aspects, the spatial patterning may utilize different combinations of electrodes distributed across multiple levels to create complex three-dimensional stimulation patterns that engage different neural populations and connectivity pathways.

[0306] The spatially patterned electrical stimulations may be designed to represent specific input categories or classes that the neural network should learn to distinguish. Each spatial PRIN- 104576 pattern may correspond to a different input class, with the electrode combinations and stimulation parameters selected to create distinguishable neural activation patterns. In some cases, the spatial patterning may involve simultaneous stimulation of multiple electrodes at different levels to create volumetric activation patterns that span the three-dimensional network architecture.

[0307] The application of spatially patterned stimulations may be performed using controlled timing protocols that allow the neural network to process each input pattern before the next pattern is presented. The temporal spacing between stimulation patterns may be selected to accommodate the natural response dynamics of the neural network while providing adequate training frequency to promote learning. In some aspects, the stimulation timing may be optimized to maximize the effectiveness of the training protocol while avoiding interference between consecutive input presentations.

[0308] The method may comprise recording neural responses from one or more electrodes. The neural response recording may involve monitoring electrical activity patterns that occur following the application of spatially patterned stimulations to assess how the neural network processes and responds to different input patterns. In some cases, the response recording may utilize electrodes that are distinct from those used for stimulation delivery, allowing simultaneous stimulation and monitoring of network activity.

[0309] The recording of neural responses may capture various types of electrical signals including action potentials, local field potentials, and population-level activity patterns that reflect the computational processing performed by the neural network. The response patterns may evolve over time as the network learns to discriminate between different input patterns through repeated training exposure. In some aspects, the neural response recording may reveal the development of specific activity signatures that correspond to different input classes as the network undergoes training.

[0310] The neural responses may be recorded from multiple electrodes distributed across different levels of the scaffold structure to capture the spatial distribution of network activity following stimulation. The multi-electrode recording approach may provide comprehensive assessment of how input patterns propagate through the three-dimensional network architecture and influence activity at different spatial locations. In some cases, the distributed recording may reveal the emergence of spatially organized response patterns that reflect the development of computational representations within the neural network.

[0311] The method may comprise processing the neural responses to perform classification of input patterns. The processing of neural responses may involve analysis of the recorded PRIN- 104576 electrical activity patterns to extract features that can be used to identify and classify the different input patterns presented to the neural network. In some aspects, the response processing may utilize machine learning algorithms or statistical analysis techniques to identify patterns in the neural activity that correspond to specific input classes.

[0312] The classification of input patterns may be performed by analyzing the temporal and spatial characteristics of neural responses recorded following stimulation with different spatially patterned inputs. The classification process may involve training computational algorithms to recognize the neural response signatures associated with each input pattern class. In some cases, the classification accuracy may serve as a measure of the neural network's learning progress and computational performance.

[0313] The processing of neural responses may involve conversion of spike trains and other electrical signals into feature vectors that can be analyzed using pattern recognition algorithms. The feature extraction process may include temporal binning of neural activity, calculation of firing rate patterns, or other signal processing techniques that capture relevant aspects of the neural responses. In some aspects, the processed neural responses may be used to train external classification algorithms that can predict input patterns based on the biological neural network's responses.

[0314] The method may involve connectivity strengths between neurons being modulated through the electrical stimulations to train the neural network. The modulation of connectivity strengths may occur through synaptic plasticity mechanisms that are activated by the repeated application of spatially patterned stimulations during the training process. In some cases, the electrical stimulations may induce long-term potentiation or depression of synaptic connections that alter the functional connectivity patterns within the neural network.

[0315] The training of the neural network through connectivity modulation may involve repeated presentation of spatially patterned stimulations that reinforce specific neural pathways and weaken others based on the desired computational outcomes. The connectivity changes may accumulate over multiple training sessions, gradually shaping the network's response characteristics to improve classification performance. In some aspects, the connectivity modulation may create specialized neural circuits that are optimized for recognizing and responding to specific input pattern classes.

[0316] The modulation of connectivity strengths may be monitored through analysis of crosscorrelation patterns between neural activity recorded from different electrodes before, during, and after training protocols. The connectivity analysis may reveal how the training process alters the functional relationships between neurons distributed across the three-dimensional PRIN- 104576 network architecture. In some cases, the connectivity modulation may result in the formation of new functional pathways or the strengthening of existing connections that support improved computational performance.

[0317] The training process may involve iterative application of spatially patterned stimulations combined with assessment of classification performance to optimize the neural network's computational capabilities. The training protocols may be adjusted based on the network's learning progress to maximize classification accuracy and minimize training time. In some aspects, the training may involve multiple epochs of pattern presentation with performance evaluation after each epoch to track learning progress and determine optimal training parameters.

[0318] The biocomputing method may enable the creation of biological neural networks that can perform pattern recognition tasks with accuracy levels that improve substantially through training. The trained neural networks may exhibit classification capabilities that exceed those of untrained networks, demonstrating the effectiveness of the connectivity modulation approach. In some cases, the biocomputing method may achieve classification accuracies that are comparable to or superior to conventional artificial neural network approaches while utilizing the natural computational properties of biological neural tissue.

[0319] The method for biocomputing may provide advantages over conventional artificial neural networks through the utilization of natural neural computational mechanisms and three- dimensional network architectures that more closely resemble brain tissue organization. The biological neural networks may exhibit computational properties such as parallel processing, adaptive connectivity, and noise tolerance that are inherent to natural neural systems. In some aspects, the biocomputing approach may enable the development of hybrid biological-artificial computing systems that combine the advantages of both biological and artificial neural network technologies.

[0320] The electrical stimulation parameters may be precisely controlled to optimize neural network training and pattern recognition capabilities. The electrical stimulation may comprise biphasic pulses having a duration of about 0.1 millisecond to about 2 milliseconds. The biphasic pulse configuration may provide charge-balanced stimulation that prevents accumulation of electrochemical products at the electrode-tissue interface while delivering sufficient charge to reliably activate neural tissue. In some cases, the pulse duration may be selected within this range to optimize neural activation efficiency while minimizing potential tissue damage from excessive charge injection. PRIN- 104576

[0321] The biphasic pulse duration may be customized based on the specific neural populations being targeted and the desired activation characteristics. Shorter pulse durations within the range, such as 0.1 millisecond to 0.5 milliseconds, may provide efficient neural activation with minimal energy consumption and reduced risk of electrochemical side effects. Longer pulse durations, such as 1 millisecond to 2 milliseconds, may be utilized when greater charge delivery is needed to achieve reliable activation of neurons with higher activation thresholds or when stimulating through thicker tissue layers.

[0322] The electrical stimulation may be applied at a frequency of about 0.1 Hz to about 1 Hz. The stimulation frequency range may be selected to promote synaptic plasticity mechanisms while allowing adequate recovery time between stimulation events for the neural network to process each input pattern. In some aspects, frequencies within this range may be compatible with natural neural activity patterns and may avoid interference with ongoing spontaneous network dynamics that could disrupt the learning process.

[0323] Lower frequencies within this range, such as 0.1 Hz to 0.3 Hz, may provide intermittent stimulation that allows plasticity mechanisms to operate between stimulation events without overwhelming the natural network dynamics. The extended interstimulus intervals may enable the neural network to fully process each input pattern and consolidate connectivity changes before the next pattern is presented. Higher frequencies, such as 0.5 Hz to 1 Hz, may provide more frequent training inputs that can accelerate the learning process while remaining within physiologically relevant frequency ranges that do not disrupt normal neural function.

[0324] The combination of biphasic pulse duration and stimulation frequency may be optimized to achieve specific learning objectives and classification performance targets. The temporal parameters may be adjusted based on the complexity of the input patterns being trained and the desired speed of learning acquisition. In some cases, the stimulation parameters may be modified during the training process to maintain optimal learning conditions as the neural network develops and connectivity patterns evolve.

[0325] The spatially patterned electrical stimulations may be applied to different combinations of electrodes to create distinct input patterns for classification. The electrode combination approach may enable the creation of multiple unique input patterns that the neural network can learn to discriminate and classify. In some aspects, each distinct electrode combination may represent a different input class or category that requires recognition by the biological neural network during the training and testing phases.

[0326] The different combinations of electrodes may be selected to create spatially diverse activation patterns that engage different neural populations and connectivity pathways within PRIN- 104576 the three-dimensional network architecture. The electrode combinations may be designed to maximize the distinguishability of the resulting neural activation patterns while ensuring that each combination produces reliable and reproducible responses. In some cases, the electrode combinations may be selected based on their spatial distribution to create input patterns that span different regions of the neural network and activate diverse neural circuits.

[0327] The creation of distinct input patterns through different electrode combinations may involve systematic variation of which electrodes are activated simultaneously during each stimulation event. The pattern creation process may utilize predetermined electrode selection algorithms that ensure adequate separation between different input classes while maintaining consistency within each class. In some aspects, the distinct input patterns may be designed to test the neural network's ability to discriminate between subtle variations in spatial activation patterns or to recognize more dramatic differences in stimulation locations.

[0328] The input patterns may comprise simultaneous stimulation of at least two electrodes on multiple levels of the flexible scaffold structure. The multi-electrode, multi-level stimulation approach may create complex three-dimensional activation patterns that engage neural populations distributed throughout the vertical extent of the neural culture. In some cases, the simultaneous stimulation of multiple electrodes may produce spatial summation effects that create unique activation signatures for each input pattern class.

[0329] The simultaneous stimulation of at least two electrodes may involve coordinated activation of electrode pairs or groups that are positioned at different spatial locations within the scaffold structure. The multi-electrode stimulation may create activation patterns that cannot be achieved through single-electrode stimulation, enabling the generation of more complex and distinguishable input patterns. In some aspects, the simultaneous stimulation approach may allow the creation of input patterns that mimic natural neural activation patterns where multiple inputs converge on neural circuits simultaneously.

[0330] The stimulation across multiple levels of the flexible scaffold structure may engage neural populations at different vertical positions within the three-dimensional culture, creating volumetric activation patterns that span the entire network architecture. The multi-level stimulation approach may activate neural pathways that connect different scaffold levels, enabling the creation of input patterns that test the network's ability to integrate information across spatial scales. In some cases, the multi-level stimulation may produce propagating activation patterns that travel through the three-dimensional network structure and create temporally extended response signatures. PRIN- 104576

[0331] The combination of simultaneous multi-electrode stimulation across multiple levels may enable the creation of highly complex input patterns that challenge the computational capabilities of the biological neural network. The three-dimensional stimulation patterns may engage diverse neural circuits and connectivity motifs that contribute to the network's pattern recognition capabilities. In some aspects, the multi-level, multi-electrode approach may create input patterns that more closely resemble the complex activation patterns that occur in natural brain tissue during sensory processing or cognitive tasks.

[0332] Processing the neural responses may comprise converting spike trains into binned arrays and applying logistic regression analysis to classify the input patterns. The spike train conversion process may involve temporal discretization of the recorded neural activity into fixed time intervals or bins that capture the essential temporal characteristics of the neural responses. In some cases, the binning process may aggregate spike counts within predetermined time windows to create feature vectors that represent the neural network's response to each input pattern.

[0333] The conversion of spike trains into binned arrays may utilize time bins with durations selected to capture relevant temporal dynamics of the neural responses while providing adequate temporal resolution for pattern discrimination. The bin duration may be optimized based on the characteristic time scales of neural responses and the temporal precision required for accurate classification. In some aspects, the binning process may involve multiple time scales to capture both rapid transient responses and slower sustained activity patterns that contribute to pattern recognition.

[0334] The binned arrays created from spike train conversion may serve as input features for machine learning algorithms that perform pattern classification based on the neural network's responses. The array format may enable efficient processing of the neural response data using standard computational tools and statistical analysis methods. In some cases, the binned arrays may be normalized or preprocessed to optimize their suitability for classification algorithms and improve pattern recognition performance.

[0335] The application of logistic regression analysis may provide a statistical framework for classifying the input patterns based on the neural response characteristics captured in the binned arrays. Logistic regression may offer advantages for biological neural network analysis through its ability to handle multi-dimensional feature spaces and provide probabilistic classification outputs. In some aspects, the logistic regression approach may enable quantitative assessment of classification confidence and identification of the neural response features that contribute most strongly to pattern discrimination. PRIN- 104576

[0336] The logistic regression analysis may involve training computational models that learn to associate specific neural response patterns with corresponding input pattern classes. The training process may utilize supervised learning approaches where the correct input pattern labels are provided during the model development phase. In some cases, the logistic regression models may be validated using cross-validation techniques that assess their ability to correctly classify novel input patterns not used during the training phase.

[0337] The combination of spike train binning and logistic regression analysis may provide a comprehensive framework for quantifying the computational performance of biological neural networks in pattern recognition tasks. The analysis approach may enable objective assessment of learning progress and classification accuracy throughout the training process. In some aspects, the processing methodology may facilitate comparison of biological neural network performance with conventional artificial neural network approaches and enable optimization of training protocols to maximize computational capabilities.

[0338] Example 1 - Fabrication, assembly, and characterization of 3D-MIND

[0339] A scaffold was designed with the following scaffold parameters: Width of each layer of scaffold 6.3 mm; length of each layer of scaffold 5 mml; scaffold ribbon width 25 pm; longitudinal gap between ribbons 50 pm; transverse gap between ribbon 75 pm; metal interconnect lines width 10 pm; and diameter of sensors 30 pm.

[0340] The main steps involved in the fabrication were as follows:

[0341] 1) Lift-off resist LOR3A (Kayaku Advanced Materials) and positive photoresist AZ 5214 (EMD Performance materials) were spin-coated on a 500 pm thick, 100 mm diameter fused silica wafer (University Wafer, Inc.) which was precleaned using acetone and isopropanol (IP A), dehydrated at 180 °C for 5 mins , and further cleaned using oxygen plasma. AZ 5214 was patterned and exposed using a laser writer (Heidelberg DWL 66+), developed in AZ 300 MIF Developer (EMD Performance materials), and a 100 nm thick sacrificial layer of nickel was deposited (Angstrom Nexdep). Lift-off was done in MR1165 (Dupont Electronics Materials) at 80°C for 1 hour.

[0342] 2) A 2 pm thick negative photoresist SU-8 2002 (Kayaku Advanced Materials) was spin-coated (bottom SU-8), followed by soft baking at 65°C and 95°C for 1 min each. The SU- 8 was patterned and exposed using a mask aligner (Suss MA6) and then post-exposure baked at 65°C and 95°C for 1 min each. The PR was developed using an SU- 8 developer solution (Kayaku Advanced Materials), rinsed with IP A, and dried with nitrogen followed by a hard bake at 185 °C for 1 hour. PRIN- 104576

[0343] 3) LOR3A was spin-coated on the sample followed by a layer of positive photoresist AZ 1505 (EMD Performance materials). The sample was then selectively exposed using a mask aligner (Suss MA6) and developed to create the pattern for metal sensors, interconnects and input / output pads, followed by a thermal deposition of 10 nm chromium and 100 nm gold using an Angstrom Covap. Liftoff was done in MR 1165 at 80°C for 6 hours, and the sample was then rinsed with DI water and IPA, and was dried with nitrogen.

[0344] 4) Step 2 was repeated to create another layer of SU-8 to passivate the metal lines, leaving the sensors (electrodes) and input / output pads exposed, the hard bake was done at 195 °C.

[0345] 5) 250 pm thick negative photoresist SU82075 (Kayaku Advanced Materials) was spin-coated (Spacer SU-8), followed by soft baking at 65°C and 95°C for 10 mins and 3.5 hours respectively. The PR was patterned and exposed using a mask aligner (Suss MA6) and then post exposure baked at 65°C and 95°C for 5 min and 16 min respectively. The PR was developed for 45 mins using an SU-8 developer solution, rinsed with IPA, and dried with nitrogen followed by a hardbake at 205°C for 1 hour.

[0346] 6) The sample was then exposed to an oxygen plasma (100 W, 1 min) to make the SU- 8 hydrophilic.

[0347] 7) The wafer was then coated with another protective layer of PR AZ 5214 and diced in the shape of the final MEA using an ADT Dicing saw.

[0348] Inspired by origami, the device was physically folded to assemble it from a 2D form after microfabrication into a 3D structure with precisely distributed electronic sensors. More specifically, after the photoresist in the fabricated device was removed using acetone and IPA, it was then put into a wet nickel etchant TFB (Transene) to remove the sacrificial nickel layer. The removed nickel allows one end of the scaffold to float while the other end stays attached to the substrate. The device was then rinsed in DI water and the scaffold was folded, similar to that shown in FIGS. 5 and 6. The main steps in folding are:

[0349] 1) The device was submerged in DI water and a pipette was used to create a gentle flow such that the free end folds at 180 degrees, the folded end was held in place by a thin-tipped pipette while the water was gently removed, and a finer alignment of the scaffold was done under a microscope.

[0350] 2) The water was allowed to evaporate, and the sides of the scaffold were glued using Kwik-Cast (World Precision Instrument) (Supplementary Fig. 3d). The glue was allowed to cure for 15 mins. PRIN- 104576

[0351] 3) Step 1 was repeated 2 more times to create a total of 4 layers, each layer spanning 5 mm x 6.5 mm.

[0352] 4) The entire device was then submerged in 70 percent ethanol for sterilization after which a custom-made retainer ring made of acrylic was glued (using PDMS Sylgard 184) to the center of the device to create a well for the media. A lid was also designed to cover the well to reduce evaporation while allowing exchange of gases.

[0353] Importantly, the SU-8 spacers prevent the layers from sagging by creating a controllable gap between each folded layer. After folding, the devices were assembled into 3D tissue-like scaffolds, spanning a total of about 100 pm to about 10 mm, such as about 100 pm to about 1 mm in the vertical dimension that was controlled by the spacers. In addition, the spacers also allowed alignment and positioning of the sensors across layers with precision comparable to the size of single neuron soma.

[0354] After folding, the electrical impedance and noise level of the sensors were characterized across a frequency range of 10 Hz to 100 kHz.

[0355] After a retainer ring was glued to the substrate (see, e.g., FIG. 7), it was filled with IX PBS (Phosphate Buffer Solution) to characterize the sensors. The device was inserted in the MEA2100-Mini-System (MultiChannelSystems) and the baseline noise from each channel was visually noted. Any sensor that had very high noise profile was considered to be potentially broken. A sinusoidal pulse of 1 kHz and 1 mV peak-to-peak amplitude (Agilent 33250A Function Generator) was applied to the PBS buffer using a platinum electrode. The response of the sensors to this signal was visually inspected and was used to further classify any faulty sensors. To characterize the noise profile of the sensors, unfiltered noise was recorded using the MEA2100-Mini-System and its Fourier transform was obtained. Electrochemical impedance spectroscopy (EIS) was performed using a Lock-in amplifier (Zurich instruments) to measure the impedance of the sensors (1 kHz sinusoidal voltage; 100-mv peak-to-peak amplitude). The EIS setup was further used to characterize the frequency response of the sensors by doing a frequency sweep from 1 Hz to 100 kHz. The stimulation response of the MEA2 100-MiniSystem was then characterized by measuring the voltage response of the sensor to an injected current or voltage. The MINI system was used to inject biphasic current pulses of amplitudes 5 pA, 10 pA, 20 pA, 50 pA, and 100 uA, each with pulse durations of 20 ps, 40 ps, 60 ps, 80 ps, 100 ps, 200 ps, 400 ps and 1 ms . The voltage response of the stimulation in the

[0356] PBS solution was recorded by an oscilloscope (Tektronix MDO4034-3) using a platinum electrode. Similarly, the response of the sensors to voltage stimulation was also PRIN- 104576 measured by applying biphasic voltage pulses of amplitudes 100 mV, 200 mV, 500 mV, 800 mV, and 1 V, each with pulse durations of 20 ps, 40 ps, 60 ps, 80 ps, lOOps, 200 ps, 400 ps, and 1 ms. After the sensor characterizations, PBS was removed, and the device was washed twice using cell grade water and finally sterilized with 70% ethanol.

[0357] Particularly relevant to AP recording is the performance at 1 kHz; impedance values of ~ 200 k and noise level of ~ 80 dB were obtained.

[0358] Example 2 - Stable long-term action potential recordings and responses to pharmacological stimulations with 3D-Mind

[0359] The assembled 3D devices were integrated with 3D cultured neural networks.

[0360] Pre-culture treatment of 3D-MIND and control planar MEAs. Poly-D-Lysine (Sigma Aldrich) was diluted in a Borate buffer ( pH 8.3 ) to get a 500 pg / ml solution. 1.5 ml of this Poly-D-Lysine solution was added to the culture wells of 3D-MIND and control planar MEAs that were precleaned based on company's recommendations. The devices were stored in the incubator for 24 hours, after which the solution was aspirated, and the wells were washed twice with cell grade water (Intermountain Life Sciences). A 10 pg / ml solution of Laminin (Coming) in HBSS ( Ca and Mg free; HyClone ) was then added to the culture wells and they were stored in the incubator for another 24 hours. The laminin was then aspirated, and the wells were washed twice with cell grade water before proceeding with cell culture.

[0361] 3D and 2D neural cultures. 17-day pregnant Sprague Dawley rats (Charles River) were euthanized using CO2, the embryos were extracted, and the head was removed and stored in HBSS (Cytiva) in an ice bath. Dissection was performed under a HEPA filter hood during which the embryo heads were stored in a petri dish over ice. The hippocampus was extracted from the heads in an HBSS filled petri dish and the hippocampal tissues were stored in Hibernate E (Gibco) media. To culture the neurons, the hippocampal tissues were washed twice with HBSS in a centrifuge (30sec, 800rpm ). After the second wash 900 pL HBSS and 100 pL IX Trypsin (MP Biomedicals) were added to the tissue and it was kept in a hot water bath (37°C) for 15 mins, gently shaking the tube halfway through. The trypsin was washed with HBSS twice in a centrifuge and after the final wash 2 ml of Neuronal media (Neurobasal media + B-27 ( 2% ) + PBS ( 1% ) + NGF 2.5S ( 0.1% ); all from Gibco) was added to the tissues. A flame polished Pasteur pipette was used to break the tissues into individual cells. The cell density was calculated using a Hemacytometer (Hausser Scientific) in a Trypan Blue dye (Sigma Aldrich). The cell solution was appropriately diluted to obtain a cell density of ~700 cells / pL and a portion of this cell solution was separated for 2-D seeding of cells. 50 pl of the PRIN- 104576

[0362] ~700 cells / .L cell solution was added to the base of the 3-D MIND and control MEAs to create a 2-D cell layer, the devices were then kept still inside the hood for 20 mins to allow for the neurons to adhere to the surface. The remaining cell solution was increased to a volume of 5 ml and then centrifuged, the media was aspirated, and fresh media was added based on the first cell hemacytometer reading to create a high-density cell solution. During this process, Matrigel (Coming) was thawed in an ice bath. Depending on the density of Matrigel supplied by the manufacturer, the high-density cell solution was added to the Matrigel and thoroughly mixed resulting in a final protein concentration of ~5 mg / ml and a cell density of ~750 cells / pL. After allowing 20 mins for 2D cells to adhere, the excess media was removed from the substrate. Fresh pre-warmed media was added to the control MEAs and they were moved to the incubator (37 °C). The Matrigel solution was then added to the folded scaffold and the device was moved to the incubator (37°C) for 45 mins for the Matrigel to cure. After 45mins, 1.5ml of neuronal media was added to the device. The device was then kept in a humidified CO2 (5%) incubator at 37°C. Half of the media was first changed after 24 hours, and subsequently half the media was changed every 2 days.

[0363] In order to characterize the interface between our devices and the 3D cultured neural networks, the SU-8 polymer was dyed with Rhodamine B, and the neurons were dyed with CellBrite Green and DAPI. Volumetric confocal fluorescence images highlighted three important features. First, neurons make connections in all 3D directions with extensive neurite growth. Second, the 3D-MIND with macroporous structures and multiple folded layers can interpenetrate with the 3D cultured neural networks. Third, neurites prefer growing along the SU8 scaffold: -75% of tracked axons followed the SU-8 scaffold lines on a plane that constituted -50% SU-8 surface, allowing for patterning and positioning of neurons. Importantly, the 3D-MIND did not disrupt the native neural networks as confirmed by the unchanged neuronal density with or without the device embedded.

[0364] With the 3D interpenetrated interface between the 3D-MIND and the 3D cultured neural networks, stable long-term tracking of the electrical activities from the neural cultures were demonstrated. Electrical recordings from 3D-MIND demonstrated that extracellular APs can be recorded for up to 191 days in vitro (DIV) with no decay of the signal amplitude. See, e.g., FIG. 8. This long-term electrical interface stability at single-neuron level laid down the foundation for subsequent BNN training via chronic electrical stimulation.

[0365] Moreover, robust burst activities were observed in the chronic recordings and tracked the evolution of key burst characteristics over time. The frequency and interspike interval of the burst gradually increased as the culture matured over months, while the duration of the PRIN- 104576 bursts and number of spikes within each burst decreased. See FIGS. 9A-9D. Importantly, the burst characteristics observed in 3D-MIND differ from those in 2D cultures grown on control planar MEAs. Specifically, in contrast to 2D neural cultures -where burst activities are relatively frequent with a unimodal distribution, bursts in 3D neural networks occur less frequently with bimodal distributions. These suggest two distinct types of bursts in 3D: shorter bursts with fewer spikes that decrease in strength rapidly and longer bursts that have a slower fall in spike amplitude. Additionally, compared to neurons in 2D cultures, neurons in 3D networks exhibit more non-burst activities that resemble the neuronal activities inside the brain. Crucially, the sensors showed stable signal-to-noise ratio (SNR) over the lifetime of the 3D cultures. Together, these results suggest that 3D neural cultures better recapitulate the fundamental energy and computational principles of the brains.

[0366] To test 3D-MIND as a platform for 3D chronic monitoring of pharmacological modulations and to confirm our recorded signals were physiologically relevant, 3D-MIND's responses to various drugs was investigated. The pre-drug firing activity was recorded from the devices that acted as the baseline before replacing the culture media with a drug and rerecording the firing activity. The drug was then washed twice with drug-free culture media, and the firing activity was recorded again. 3D-MIND's responses to three drugs were studied: Bicuculline-a gamma-aminobutyric acid type A(GABAA) receptor antagonist, Tetrodotoxin (TTX)-a sodium channel blocker, and Cyanquixaline (CNQX)-an AMPA-type glutamate receptor antagonist. First, GABAAreceptors are ligand-gated ion channels that have an inhibitory influence on target neurons, so a GABAAreceptor antagonist like Bicuculline increases the likelihood of a neuron firing.

[0367] As expected, addition of Bicuculline in 3D-MIND increased the firing rate of the neurons by ~ 7-fold in a few minutes that persisted after the Bicuculline was washed. The firing rate went back to its pre-drug rate 24 hours after the Bicuculline was removed. The increase in the firing rate is synonymous with lowering the threshold of the activation function of the nodes thus allowing us to pharmacologically modulate the training of the BNN. Second, opposite to Bicuculline, the addition of TTX immediately suppressed firing activities of all neurons, suggesting a rise in the threshold of the activation function. To ensure that the disappearance of neuronal firing activities was not caused by neuronal death, the TTX was removed 48 hours after the initial TTX addition. Immediately after the TTX was removed, the firing activities of neurons recovered with firing rates returning to pre-TTX level. Last, the PRIN- 104576 addition of CNQX temporarily silenced burst firings in the neurons. The burst firings recovered after the CNQX was removed.

[0368] More specifically, 10 mM Bicuculline in DMSO (HY-N0219) was obtained from MedChemExpress. 15 pl of this solution was mixed with 1500 pl of pre-warmed culture media to make a working Bicuculline solution with concentration 100 pM. The cell culture was removed from the incubator and the culture media was aspirated. 1500 pl of the working bicuculline solution was added to the cell culture. Electrophysiology was done keeping the culture in the incubator. The bicuculline solution was then washed twice ( 5 min each time) using fresh pre-warmed media. Tetrodotoxin (TTX) was obtained from Cayman Chemicals (14963). 1 mg of TTX was diluted in 3.134 ml of Citrate Buffer solution (Sigma Aldrich; C2488-100ML) to make a 1 mM TTX solution. 1 ml of the resulting solution was further diluted in 9 ml of sterile water to make a 100 pM solution. 15 pl of this solution was added to 1485pl of pre-warmed culture media to make a final TTX solution with concentration IpM. The cell culture was removed from the incubator and the media was aspirated. The TTX solution was added to the culture and electrophysiology was performed. The TTX was washed twice ( 5 min each) with fresh pre-warmed media after 48 hours. CNQX was obtained from Cayman Chemicals (14618). 1.6-mg of CNQX was diluted in 1.380 ml of Dimethylsulfoxide (DMSO - Thermo Scientific, J66650.AK) to make a 5 mM CNQX solution. 50 pl of this solution was added to 450 pl of pre-warmed culture media to make a final working CNQX solution with concentration 500 pM. The cell culture was removed from the incubator and lOOpl of the working CNQX solution was added to the culture and electrophysiology was performed. The CNQX was washed twice ( 5 min each) with fresh pre-warmed media.

[0369] A MEA2100-Mini-System was used to record the electrical activity (at 20 kHz ) from the culture placed inside the incubator with Multi Channel Experimenter. A second order Butterworth bandpass filter ( 200 Hz to 3500 Hz ) was applied to the raw data. The cultures were recorded for 100 sec (a small file for ease of visualization) and 15 min (long recording for data collection and analysis). Different electrical stimulation parameters were tried to measure the response of the 3D-MIND. A biphasic 1 V peak-to-peak pulse of duration 1 ms was chosen for stimulation as it gave the best response. The stimulation pulse was administered every 5 sec (0.2 Hz) for a total of 30 pulses.

[0370] Example 3. Volumetric electrophysiology mapping from 3D neural networks over months. PRIN- 104576

[0371] Next, the 3D long-term mapping capability of 3D-MIND during the development of the 3D cultured neural networks was investigated. For the convenience of description, the sensors were number in an x-y-z format where x represents the layer (1 as the bottom layer and 4 as the topmost layer), and y and z represent the sensor row and column within the layer, respectively. First, electrical recordings from 16 sensors distributed into 4 layers, with 3 to 5 sensors selected from each layer, show clear single-unit activities from 3D-MIND, confirming the volumetric mapping capabilities of 3D-MIND. See FIG. 10.

[0372] Importantly, this 3D volumetric electrophysiology mapping is stable across all 4 layers over long-term, as evidenced by the selected (one from each layer) sensors' recording traces and their corresponding sorted spikes. See FIGS. 11A-11B. Second, the 3D volumetric electrophysiology mapping capability of 3D-MIND allowed us to track the neuronal connectivity as the 3D cultured neural network evolved. Specifically, a cross-correlogram constructed from the spike trains was used to infer the connectivity and information flow direction between pairs of neurons. See FIG. 12. To quantitively estimate the neuronal connectivity, Gaussian fitting was performed on the resulting cross-correlogram and used the metric obtained from the Gaussian fitting as an indicator of the connectivity strength / weight

[0373] Specifically, NeuroExplorer (Nex Technologies) was used to analyze the data. The filtered data generated from the recording setup was used to extract spikes using a threshold of 4 * (Median of the filtered signal). All other data processing such as spike sorting, burst analyses (Maxinterval method), cross-correlogram was also performed in NeuroExplorer. Second order Gaussian fit of cross correlograms was done in Matlab resulting in a curve of the form A ratio of the sum of the height of the curves and the sum of the standard deviation of the curves (al+a2) / (cl+c2) was used as the parameter to quantify the weight of the connection (connection strength).

[0374] The cross-correlogram of sensor 4-2-4 (424) with respect to sensor 1-3-3 (133) shows that the neuron recorded by 424 is downstream of the neuron recorded by 133 (see FIG. 13) with a weight of 0.57.

[0375] Last, this analysis pipeline was generalized to all the sensors to construct the 3D connectivity map of the 3D cultured neural networks. The 3D connectivity map was tracked over time and it was found that the connectivity map evolved over the course of time, as is expected in vivo. PRIN- 104576

[0376] Example 4 - Modulating network connectivity via chronic electrical stimulations and constructing BNNs

[0377] The capabilities of 3D-MIND were further expanded by incorporating electrical stimulations. As a proof-of-concept, first it was tested whether chronic electrical stimulations could modulate the 3D neural connectivity by altering neural connections and / or their weights. Neurons from a given layer were chronically stimulated and neuronal responses observed across the 3D volume. As an example, electrical pulse trains with 1 ms pulse duration and 0.2 Hz repetition rate were applied to electrode 231 (layer 2) and activities from all sensors were recorded. Representative recording traces from sensors 231 (layer 2), 114 (layer 1), and 322 (layer 3) were analyzed. See FIG. 14.

[0378] A perievent raster plot of these electronic sensors show that electrical stimulation of the neurons near sensor 231 can lead to an increased firing activity in neurons recorded by sensors 114 and 322, suggesting connections between these neurons. Applying crosscorrelogram analyses described above, the connectivity strength was tracked between neurons recorded by sensors 322 and 421 over multiple days while applying electrical stimulation via stimulator 231. Interestingly, it was found that the connectivity between these neurons strengthened over time, suggesting long-term plasticity akin to evolving weights between nodes in ANNs. See FIG. 15. The increased trend, however, fade away after the electrical stimulation was withdrawn, suggesting a dynamic plasticity in neural connectivity. Plasticity observed in 3D-MIND as a result of 0.2 Hz stimulation is consistent with previous reports.

[0379] The volumetric recording capability of 3DMIND was then employed to track the evolving neural network in response to external stimulations. As a demonstration, the connectivity between the three neurons near sensors 114, 322 and 421 were analyzed across days in response to chronic stimulation. Interestingly, it was observed the net directionality between 322 & 421 and 114 & 421 reversed along with changes in their weights.

[0380] Utilizing the plasticity of the neuronal connectivity and the long-term stability of the 3D device-3D neural network interface, 3D-MIND was employed to construct a reservoir neural network for classification. First, different spatially patterned biphasic electrical pulses were input from the volumetrically distributed stimulators of 3D-MIND. Next, we utilized the neuronal responses recorded from the sensors to fit a logistic regression model to classify the different input patterns. See, e.g., FIGS. 17 and 18.

[0381] For classification based neural network training, 4 stimulators (A, B, C and D) were chosen from different layers of the 3D-MIND to inject stimulation pulses. 2 of these 4 stimulators were simultaneously stimulated and acted as a single input, for a total of 4 different PRIN- 104576 input paterns (AB, BC, CD, and DA). This allowed overlap between different inputs showcasing the classification property of 3D-MIND for inputs with significant common traits. A 4.5-minute-long recording session was performed that consisted of 30 stimulation pulses which followed and preceded 1 min of no stimulation. The evoked neural response from 3D- MIND was recorded and converted into a one-dimensional array X to be fed into a logistical regression model. Specifically, the signal traces from t=0 to t=300 ms after each stimulation pulse from the most active sensors were converted into a spike train which were then binned into 2 ms time windows to form an array Xi. Such arrays ( Xi ) from each sensor were then appended to each other (retaining the same order in all training steps) to form a final array X=[Xi X2...Xn ]. 120 such arrays X ( 30 stimulation pulses from each of the 4 inputs) acted as the data set for training a logistic regression model for classification. The training of classification weights was done by the Scikitleaming package in Python with 75% samples used for training the model and the remaining used for testing the accuracy. The stimulation input paterns were given to 3 different neural cultures with a set of epochs every 12 hours for a total of 5 epochs. The training accuracies were obtained after each epoch.

[0382] The neural networks were trained over 2 days with an Epoch every 12 hours. In addition, in order to test our hypothesis that 3D device interfaces outperform 2D interfaces, a logistic regression model was trained in parallel with the outputs only from sensors sitting at the bottom layer of the same cultures. It was found that: first, representative confusion matrices for both the 2D and 3D models showed improved classification accuracy after training. In specific, the naive 3D-MIND started from a relatively low accuracy of classification at 58.3±3.3% (mean ±1 s.e.m), which was improved to 81.1±4.8% after 4 epochs of trainings. Second, it was observed that the 3D device interfaces outperformed the 2D counterparts at each intermediate epoch, with faster learning and higher classification accuracy at the end of the 4 epochs. See FIG. 16A. The learning of the 3D-MIND can be atributed to unsupervised learning through changes in network connectivity over the course of training . Last, to quantify the evolution of the 3D neural networks in response to training, the connectivity strength was tracked among neural populations and observed significant increase in the weight between some pairs of neurons and decrease in others. See FIGS. 16B and 16C.

[0383] These examples disclose 3D-MIND, an embodiment of the disclosed 3D flexible bioelectronics platform with precisely positioned sensors and stimulators creating a true 3D device-3D neural network architecture. It was demonstrated that 3D-MIND can be embedded inside and interpenetrate with 3D cultured neural networks without perturbing the morphology PRIN- 104576 or distribution of the neurons. Importantly, 3D-MIND enables stable tracking of the activity, connectivity, and development of the 3D cultured neural networks over, e.g., six months. In addition, 3D-MIND is capable of monitoring the 3D cultured neural networks' responses to pharmacological stimulations. Moreover, we demonstrated that 3D-MIND can be used to train the 3D cultured neural networks by modulating the connectivity between neurons with spatially patterned electrical stimulations over days. Furthermore, it was demonstrated that 3D-MIND can function as a tunable and fully observable reservoir neural network.

[0384] The disclosed platform has multiple advantages compared to existing methodologies. First, 3D-MIND with a tissue-mimicking scaffold structure, can form noninvasive 3D interpenetration with 3D neural network that retains the cytoarchitecture, network and computational characteristics of the brain. This enables stable long-term recording and stimulation of the 3D cultured neural networks. Second, the 3D locations of the embedded sensors and stimulators can be precisely distributed both laterally and axially, allowing controlled volumetric access to neurons within the cultured neural networks. This 3D device- 30 neural network interface is central to the performance of BNNs as evidenced by our comparative training studies between the disclosed platforms and sensors distributed in 2D. In addition, this 3D device-3D neural network interface allows for physiologically relevant understanding of brain's 3D network connectivity, potentially shedding light to the development of next-generation brain-inspired computing. Finally, the disclosed platforms are capable of long-term bi-directional monitoring and modulating neural connectivity, offering potentials to "co-train" artificial and natural neural networks.

[0385] With the ability to fully track 3D neural network, the disclosed platforms not only serves as a valuable toolkit for fundamental neuroscience in understanding the high efficiency and versatility behind brain's computational capability, but also, more importantly, offers a powerful platform for brain-inspired computing, allowing targeted training of the 3D biological neural network for the development of GPBIC. While the examples mainly utilized 0.2 Hz stimulation frequency, the disclosed platforms are capable of other stimulation parameters. Comparison of resulted connectivity evolutions in response to different frequency stimulations can potentially allow us to control training rates and give us hints about potential learning rules and mechanisms. With its 3D neural culture, the disclosed platforms potentially allow one to build on critical observations to understand the adaptive mechanisms of the brain by generalizing key findings in learning that were primarily based on 2D neural cultures or 2D recordings. PRIN- 104576

[0386] A number of implementations have been described. Nevertheless, it will be understood that various modifications may be made without departing from the spirit and scope of the disclosure. Accordingly, other implementations are within the scope of the following claims.

Claims

PRIN- 104576CLAIMS1. A device for interfacing neurons cultured in three dimensions, comprising: a flexible scaffold structure composed of a flexible insulating layer, the flexible scaffold structure being folded to form a main portion having a plurality of levels; the flexible scaffold structure having mesh portions defining apertures, each aperture configured to allow a portion of a three-dimensional neural network in a neural culture to extend through the aperture; and multiple electrically-conducting patterns encased within the flexible insulating layer, wherein the multiple electrically-conducting patterns are exposed in selected regions to form a plurality of electrodes allowing the neural culture to interact with the plurality of electrodes.

2. The device of claim 1, wherein the flexible insulating layer comprises a biocompatible epoxy -based polymer.

3. The device of claim 1 or 2, wherein each electrode has an effective diameter of about 10 pm to about 50 pm.

4. The device of any one of claims 1-3, wherein the plurality of levels comprises 3 to 10 levels.

5. The device of claim 5, wherein adjacent levels are separated by a gap of about 10 pm to about 500 pm.

6. The device of any one of claims 1-5, wherein the multiple electrically-conducting patterns further define multiple wires, each wire having a width of about 5 pm to about 25 pm.

7. The device of any one of claims 1-6, further comprising input-output pads electrically coupled to the multiple electrically-conducting patterns.

8. The device of claim 7, wherein the ratio of electrodes to input-output pads is 1.

9. The device of claim 7, wherein the ratio of electrodes to input-output pads is greater than 1.PRIN- 10457610. The device of any one of claims 1-9, wherein the mesh portions define apertures configured to allow neurons to form three-dimensional connections throughout the neural culture.

11. The device of any one of claims 1-10, wherein the flexible scaffold structure is configured to interpenetrate with a three-dimensional neural network while maintaining the cytoarchitecture of the neural network.

12. The device of any one of claims 1-11, further comprising spacers between adjacent layers to maintain a predetermined gap.

13. The method of any one of claims 1-12, wherein the neural culture comprises only a single type of neuron.

14. The method of any one of claims 1-12, wherein the neural culture comprises a plurality of types of neurons.

15. The device of any one of claims 1-14, further comprising a substrate coupled to the flexible scaffold structure.

16. The device of claim 15, wherein the substrate comprises fused silica.

17. The device of claim 15, wherein the substrate is coupled to at least a portion of the flexible insulating layer on a first level of the flexible scaffold structure.

18. The device of claim 15, further comprising one or more retainer walls coupled to the substrate and positioned to define a volume for containing the neural culture.

19. The device of claim 18, wherein the one or more retainer walls comprise a retainer ring.

20. The device of claim 18 or 19, further comprising a retainer lid coupled to the one or more retainer walls.

21. A method of producing a device for interfacing neurons cultured in three dimensions, comprising:PRIN- 104576 fabricating a planar polymer scaffold on a base substrate using an insulating polymer; depositing a plurality of conductive patterns on the insulating polymer; encapsulating the plurality of conductive patterns with the insulating polymer; removing insulation from the planar polymer scaffold at selected locations to form electrodes; partially releasing the planar polymer scaffold from the base substrate; and folding the planar polymer scaffold into a plurality of layers to form a three- dimensional flexible electrode system.

22. The method of claim 21 , wherein the insulating polymer comprises a biocompatible epoxybased polymer.

23. The method of claim 21 or 22, wherein depositing the plurality of conductive patterns comprises depositing gold patterns having a thickness of about 25 nm to about 200 nm.

24. The method of claim 23, wherein depositing the gold patterns further comprises depositing a chromium adhesion layer having a thickness of about 3 nm to about 25 nm.

25. The method of any one of claims 21-24, wherein partially releasing the planar polymer scaffold comprises etching a sacrificial layer positioned between the planar polymer scaffold and the base substrate.

26. The method of claim 25, wherein the sacrificial layer comprises nickel having a thickness of about 25 nm to about 200 nm.

27. The method of any one of claims 21-26, wherein folding the planar polymer scaffold comprises creating spacers between adjacent layers to maintain a predetermined gap.

28. The method of any one of claims 21-27, wherein the three-dimensional flexible electrode system comprises 3 to 10 layers with electrodes distributed across multiple vertical planes.

29. A method for detecting or stimulating neural network connectivity, comprising:PRIN- 104576 providing a device comprising a flexible scaffold structure folded to form multiple levels with apertures allowing three-dimensional neural network growth and electrodes distributed across the levels; creating a suspension comprising neural cells and depositing the suspension in a three- dimensional volume defined by the device; allowing neurons to interweave through the apertures within the three-dimensional volume; and recording electrical signals from the neurons using the electrodes to analyze neural network connectivity and development.

30. The method of claim 29, wherein the suspension comprises a solubilized basement membrane matrix at a concentration of about 2 mg / ml to about 15 mg / ml and neural cells at a density of about 500,000 cells per ml to about 2,000,000 cells per ml.

31. The method of claim 29 or 30, further comprising applying electrical stimulation to only a subset of the plurality of electrodes to modulate neural network connectivity.

32. The method of claim 31, wherein the electrical stimulation comprises biphasic pulses having a duration of about 0.1 millisecond to about 2 milliseconds and applied at a frequency of about 0.1 Hz to about 100 Hz.

33. The method of claims 27-32, wherein recording electrical signals comprises monitoring action potentials from multiple levels simultaneously over a period of time to track neural network development.

34. The method of claim 33, wherein the period of time is at least 3 months.

35. A method for drug discovery or development, comprising: providing a device comprising a flexible scaffold structure with electrodes distributed across multiple levels interfacing with a three-dimensional neural network; recording baseline electrical activity from the neural network using the electrodes; applying a test compound to the neural network; recording electrical activity from the neural network after compound application; andPRIN- 104576 analyzing changes in electrical activity to evaluate pharmacological effects of the test compound.

36. The method of claim 35, further comprising a step of washing the test compound from the neural network and recording recovery electrical activity to assess reversibility of the pharmacological effects.

37. The method of claim 35 or 36, wherein the test compound comprises one of a neurotoxin or an antagonist of a neural receptor.

38. The method of any one of claims 35-37, wherein analyzing changes in electrical activity comprises measuring changes in firing rate, burst frequency, and / or connectivity patterns between neurons across the multiple levels.

39. The method of any one of claims 35-38, wherein the three-dimensional neural network comprises only a single type of neuron.

40. The method of any one of claims 35-38, wherein the three-dimensional neural network comprises a plurality of types of neurons.

41. A method for biocomputing, comprising: providing a device comprising a flexible scaffold structure with electrodes distributed across multiple levels interfacing with a three-dimensional neural network; applying spatially patterned electrical stimulations to the neural network through one or more of the electrodes; recording neural responses from one or more electrodes; and processing the neural responses to perform classification of input patterns, wherein connectivity strengths between neurons are modulated through the electrical stimulations to train the neural network.

42. The method of claim 41, wherein the electrical stimulation comprises biphasic pulses having a duration of about 0.1 millisecond to about 2 milliseconds and applied at a frequency of about 0.1 Hz to about 100 HzPRIN- 10457643. The method of claim 41 or 42, wherein the spatially patterned electrical stimulations are applied to different combinations of electrodes to create distinct input patterns for classification.

44. The method of claim 43, wherein the input patterns comprise simultaneous stimulation of at least two electrodes on multiple levels of the flexible scaffold structure.

45. The method of any one of claims 41-44, wherein processing the neural responses comprises converting spike trains into binned arrays and applying logistic regression analysis to classify the input patterns.

Images