Methods and apparatus for high-resolution spatial and temporal mapping of large-scale electrophysiological dynamics and transcriptomic profiles within intact brain tissue

Abstract

This invention relates to an in vitro method for mapping the spatiotemporal electrophysiological dynamics and spatial transcriptional profiles of cells in a functional neuronal cell ensemble. This in vitro method includes simultaneously recording and analyzing spatial data of molecular activity and electrical network activity at the level down to individual cells using high-density electrobiosensors, spatial transcriptomics, optical imaging, and advanced computational strategies. The invention also relates to: methods for identifying the composition of a functional neuronal ensemble; methods for monitoring the spatiotemporal electrophysiological dynamics and transcriptional profiles of cells in a functional neuronal cell ensemble; methods for monitoring the cellular composition of a functional neuronal ensemble; methods for identifying compounds that influence spatial electrophysiological and transcriptional dynamics; and / or methods for identifying the cellular composition of a functional neuronal cell ensemble based on detected dynamics; and devices thereof. This invention allows for a better understanding of disease mechanisms, the identification of therapeutic targets, and the development of new drugs and treatment regimens.

Description

[0001] This invention relates to an ex vivo method for mapping the spatiotemporal electrophysiological dynamics and spatial transcriptional profiles of cells in a functional neuronal cell ensemble, comprising the simultaneous recording and analysis of spatial data on molecular activity and electrical network activity at the individual cell level using high-density electrobiosensors, spatial transcriptomics, optical imaging, and advanced computational strategies. The invention also relates to methods, apparatus, and uses for identifying the composition of functional neuronal ensembles, monitoring the spatiotemporal electrophysiological dynamics and transcriptional profiles of cells in a functional neuronal cell ensemble, monitoring the cellular composition of a functional neuronal ensemble, or identifying compounds that influence spatial electrophysiological and transcriptional dynamics, and / or for identifying the cellular composition of a functional neuronal cell ensemble based on detected dynamics. This invention allows for a better understanding of disease mechanisms, the identification of therapeutic targets, and the development of new drugs and treatment regimens. Background Technology

[0002] The brain has evolved to be able to robustly and efficiently process complex information in order to maintain homeostasis, navigate in space, make decisions, and perform higher cognitive functions (1). A coherent understanding of the brain’s complexity, from the molecular to the systemic level, requires the integration of multimodal data from different spatiotemporal contexts.

[0003] At the heart of this ambitious goal lies the integration of neuronal electrophysiology and molecular phenotypes at network and cellular resolution, as these form the basis of physiological function as well as neurodevelopmental and neurodegenerative diseases (2, 3).

[0004] Methodological advancements, such as patch-seq, have enabled single-cell transcriptomic analysis and morphological reconstruction of individual neurons following electrophysiological recordings (4, 5). Despite their significant implications, patch-seq still suffers from very low throughput and difficulty in resolving neural networks ranging from moderate spatial backgrounds to large spatial extents.

[0005] Electro-seq combines flexible bioelectronics with in situ RNA sequencing to map electrical activity and gene expression, but it is only applicable to simplified neuron culture systems (6).

[0006] WO 2023 / 091970A1 discloses a computer-implemented method for determining cellular omics profiles using microscopic imaging data. This method combines microscopic imaging data of cells or cell populations; determines a target expression profile of a set of target genes from the microscopic imaging data, the target genes identifying cell types or cell states of interest; and determines a single-cell omics profile of the cell or cell population. The target expression profile can be a target spatial expression profile, the microscopic imaging data can be obtained using label-free microscopy methods, and the cells or cell populations can be live cells or live cell populations.

[0007] By profiling the expression patterns of thousands of neuronal genes while preserving spatial organization, high-throughput spatial resolution transcriptomics (SRT) provides unprecedented insights into the molecular diversity of brain regions of interest (7, 8). However, SRT has limited temporal resolution, providing only a single snapshot of the transcriptomic atlas. Spatiotemporal transcriptomics can only resolve transcripts using multiple samples from different time points (9, 10).

[0008] US 2022-0068438A1 relates to methods, systems, and computer products for aligning single-cell data with spatial data to generate spatial maps of cell types and gene expression at single-cell resolution. Further described is mapping to a universal coordinate frame.

[0009] WO 2021 / 168455A1 relates to the spatial detection of nucleic acids (such as genomic DNA or RNA transcripts) in cells included in a tissue sample. Methods for detecting and / or analyzing nucleic acids (such as chromatin or RNA transcripts) are disclosed to obtain spatial information about the location, distribution, or expression of genes in a tissue sample. Procedures for performing “spatial transcriptomics” or “spatial genomics” are disclosed, enabling a user to simultaneously determine expression patterns, or the location / distribution patterns of expressed genes, or gene or genomic loci present in a single cell, while retaining information related to the spatial location of the cell within the tissue structure.

[0010] Aqrawe Z et al. (published in: A simultaneous optical and electrical in-vitroneuronal recording system to evaluate microelectrode performance. PLoS One. Aug 20, 2020; 15(8):e0237709, doi:10.1371 / journal.pone.0237709.PMID:32817653; PMCID:PMC7440637) disclosed a high spatiotemporal resolution electrical recording system using a planar microelectrode array and simultaneous optical imaging, suitable for evaluating microelectrode performance using a proposed “performance factor” metric. Neuronal activity was recorded simultaneously using both electrical and optical techniques, and this activity was confirmed by inhibiting action potential discharges using tetrodotoxin.

[0011] Brain-on-a-chip technology, supported by high-density complementary metal-oxide-semiconductor (CMOS) biosensing microelectrode arrays (MEAs), now allows for non-invasive, multi-site, long-term, and label-free simultaneous measurement of extracellular activity, capturing both local field potentials and spike activity from thousands of neurons with single-cell precision without compromising cellular integrity (13 to 17).

[0012] Despite the improvements mentioned above, the following unmet needs remain: to provide new and effective methods for analyzing the spatial domain of complex neural network dynamics across different scales.

[0013] Therefore, the object of this invention is to provide a platform that can simultaneously capture and integrate molecular and functional information across multiple spatiotemporal scales within intact brain tissue samples. This will bridge the gap between molecular architecture and brain activity, providing insights into neuronal plasticity and connection dynamics at the causal level. Other objects and advantages of this invention will become apparent to those skilled in the art upon further study of this disclosure.

[0014] According to its first aspect, the above objective has been achieved according to the invention by a method for detecting spatiotemporal electrophysiological dynamics and transcriptomic profiles of a functional neuronal cell ensemble, the method comprising the steps of: providing a functional neuronal cell ensemble to be characterized; recording patterns of local field potentials (LFPs) of the neurons in the ensemble with high spatial and temporal resolution using a suitable electrode array; obtaining spatial localization patterns of the LFPs of substantially the entire ensemble or ensemble within an intact tissue using optical imaging, optionally cryopreserving substantially the entire ensemble or tissue; performing optical imaging and spatially resolved transcriptomic analysis of the cells in the ensemble using a second array; and processing the obtained data, the processing of the obtained data including aligning the spatial localization patterns of the LFPs with the spatially resolved transcriptomic analysis, and thereby detecting the spatiotemporal electrophysiological dynamics and spatial transcriptomic profiles of the cells in the functional neuronal cell ensemble.

[0015] The method according to the invention can be performed in vitro or in vivo.

[0016] Preferably, the method according to the invention comprises the collection of functional neuronal cells to be characterized selected from neuronal tissue, brain slices, neurocancer samples, interconnected brain cell cultures, cultured neuronal differentiated iPSCs, spinal cord, bioelectric tissues such as olfactory bulbs and cardiac tissue, and portions or combinations thereof, particularly from neuronal tissue derived from mammals with behavioral conditions or disorders, mammals with neurological conditions or disorders (such as neurodegenerative disorders or cancer), and / or mammals being treated for one of these conditions or disorders.

[0017] A further preferred embodiment of the method according to the invention is that processing the data and aligning the spatial localization pattern of the LFP with spatial resolution sequencing and transcriptome analysis includes overlaying an optical image with an electrode array layout and overlaying an optical image with a second array layout, and the alignment of the two overlays includes image resizing and rotation.

[0018] According to its second aspect, the above objective has been achieved according to the invention by a method for identifying the composition of a functional neuronal ensemble, the method comprising performing the method according to the invention, and further comprising the step of inferring the composition of the functional neuronal ensemble based on spatiotemporal electrophysiological dynamics and spatial transcriptional spectra of detected cells.

[0019] According to its third aspect, the above objective has been achieved according to the invention by a method for monitoring the spatiotemporal electrophysiological dynamics and spatial transcriptional profiles of functional neuronal cell assemblies, the method comprising: repeatedly performing the method according to the invention at different time points; and monitoring the spatial electrophysiological and transcriptional information of molecular neuronal cell assemblies based on comparisons of the spatiotemporal electrophysiological dynamics and transcriptional profiles of cells detected at different time points.

[0020] According to its fourth aspect, the above objective has been achieved according to the invention by a method for monitoring the cellular composition of a functional neuron set, the method comprising: repeatedly performing the method according to the invention at different time points; and monitoring the cellular composition of the functional neuron set based on a comparison of the cellular composition of the functional neuron set detected at different time points.

[0021] According to its fifth aspect, the above objective has been achieved according to the invention by a method for identifying compounds that affect the spatiotemporal electrophysiology and spatial transcriptional profile and / or composition of cells in a functional neuronal cell set, the method comprising performing the method according to the invention in the presence and absence of at least one candidate compound, wherein, in the presence and absence of at least one candidate compound, or compared with suitable controls, differences in the spatiotemporal electrophysiological dynamics and transcriptional dynamics and / or composition of cells in a functional neuronal cell set identify compounds that affect the spatiotemporal electrophysiological dynamics and transcriptional profile and / or composition of cells in a functional neuronal cell set.

[0022] According to its sixth aspect, the above objective has been achieved according to the invention by means of an apparatus for performing the method according to the invention, specifically comprising: a collection of neuronal cells to be analyzed; an electrode array adapted to record patterns of local field potentials (LFPs) of the collection during optical imaging; a second array adapted to perform spatial resolution transcriptomic analysis of the cells in the collection during optical imaging; and a computer program product for processing the acquired data, the processing of the acquired data including aligning the spatial localization patterns of the LFPs with spatial resolution sequencing and transcriptomic analysis, the computer program product optionally having additional buffers and instructions for use.

[0023] According to its seventh aspect, the above-mentioned objective has been achieved according to the invention by using a device according to the invention for detecting spatiotemporal electrophysiological dynamics and transcriptional profiles of cells in a functional neuronal cell ensemble, for identifying the composition of a functional neuronal cell ensemble, for monitoring spatiotemporal electrophysiological dynamics and spatial transcriptional dynamics of cells in a functional neuronal cell ensemble, for monitoring the cellular composition of a functional neuronal cell ensemble, or for identifying compounds that affect spatiotemporal electrophysiological dynamics and transcriptional profiles and / or for identifying the cellular composition of a functional neuronal cell ensemble.

[0024] Brain function is performed through the combined action of large sets of neurons and gene networks that share basic organizational principles, and information processing across a wide range of spatial and temporal scales that evolve with experience and disease.

[0025] Therefore, decoding spatiotemporal electrophysiological and spatial transcriptomic information in the same tissue while preserving cell aggregation location can specify spatiotemporal transcriptomic networks, spatiotemporal transcript spectra, and / or spatiotemporal transcriptomic maps in real-time relationships with connectivity and other multimodal data. Based on potential discharge information, molecular dynamics can be quantified in pseudotime, spatially resolved cell type composition can be deconvolved, and electrophysiological network features can be predicted from the transcriptomic spectrum with high accuracy.

[0026] As mentioned above, according to its first aspect, the present invention provides a method for detecting spatiotemporal electrophysiological dynamics and spatial transcriptomic profiles of cells in a functional neuronal cell ensemble. The method includes providing one or more functional neuronal cell ensembles to be characterized. Patterns of local field potentials (LFPs) of the neurons in the ensemble are then recorded at high spatial and temporal resolution using a suitable electrode array, while simultaneously obtaining a spatial localization pattern of LFPs substantially throughout the ensemble using optical imaging. Optionally, substantially the entire ensemble is then cryopreserved and can be stored prior to further analysis. Subsequently, optical imaging and spatial resolution transcriptomic analysis of the cell ensemble are performed using a second suitable array. In the context of the invention, local field potentials (LFPs) recorded in brain slices are electrical signals measured from a group of neurons (ensembles) in a slice of brain tissue. These signals reflect the combined activity of these neurons, providing insights into how these neurons communicate and function together. Processing the acquired data includes aligning the spatial localization pattern of neural activity patterns recorded as LFP signals (“LFPs” herein) by the electrode array with spatial resolution sequencing and transcriptomic analysis, preferably using machine learning algorithms, and thus including the timing and topology of the ensemble activity in a high-resolution representation. Thus, the spatiotemporal electrophysiological dynamics and spatial transcriptional profiles of functional neuronal cell assemblies can be detected.

[0027] In a set of specific embodiments of the present invention, the inventors hereby provide a method referred to in a preferred embodiment as “MEA-seqX”, which combines on-chip brain technology, bioimaging technology, and spatial sequencing technology within a cross-scale computing framework. MEA-seqX includes i) simultaneously recording electrophysiological firing patterns from large cell assemblies in brain slices at high spatiotemporal resolution and imaging the entire circuit for spatial localization; ii) performing multiplexed spatial spectral analysis of cellular transcriptomics from the same cell assemblies in the same neural circuit; and iii) specifying the spatiotemporal transcriptome network based on real-time relationships between the spatiotemporal transcriptome network and connectivity and other multimodal data, quantifying molecular dynamics in pseudotime based on potential firing information, deconvolving the spatially resolved cell type composition, and / or using machine learning algorithms to predict electrophysiological network features from the transcriptome spectrum with high accuracy, and preserving time and topology in a high-resolution representation.

[0028] Generally, the method according to the invention can be used in many different analytical scenarios concerning functional neuronal assemblies. In the context of this invention, the term "functional neuronal assembly" or "neuronal assembly" or "cell assembly" should refer to a group of neurons that are activated in response to a specific type of stimulus or during a specific behavioral process and function together as a unit. This concept is fundamental to understanding how the brain encodes and processes information (see also, for example, Holtmaat, A., Caroni, P. Functional and structural underpinnings of neuronal assembly formation in learning, Nature Neuroscience 19, 1553–1562 (2016), https: / / doi.org / 10.1038 / nn.4418, and Umbach, G., Tan, R., Jacobs, J. et al. Flexibility of functional neuronal assemblies supports human memory, Nature Communications 13, 6162 (2022), https: / / doi.org / 10.1038 / s41467-022-33587-0). For the purposes of the method according to the invention, the neuronal cell ensemble needs to maintain its basic function in vitro, that is, it needs to exhibit co-firing activity that basically reflects its function in vivo.

[0029] The functional neuronal cell set to be characterized can be selected from neuronal tissue (set), brain slices, neurocancer samples, cultures of interconnected brain cells, cultured neuronal differentiated iPSCs, spinal cord, bioelectric tissues such as the olfactory bulb and cardiac tissue, and parts or combinations thereof. Particularly preferred are neuronal tissue (sets) derived from mammals with behavioral conditions or disorders, mammals with neurological conditions or disorders (such as neurodegenerative disorders or cancer), and / or mammals being treated for one of these conditions or disorders.

[0030] Then, patterns of local field potentials (LFPs) of neurons in the ensemble are recorded with high spatial and temporal resolution using a suitable electrode array. LFPs are transient electrical signals generated in nerves and other tissues by the sum and synchronized electrical activity of individual cells (e.g., neurons) within those nerves and tissues. LFPs are “extracellular” signals, generated by transient imbalances in ion concentrations in the extracellular space caused by cellular electrical activity. LFPs are “local” because they are recorded by electrodes placed near the electrogenic cells. Determining the local field potentials (LFPs) of neurons in the ensemble can be performed using any suitable technique known to those skilled in the art and / or described herein. An example is a large-scale array of 4,096 electrodes performing simultaneous recording, used to study spontaneous and induced network activity propagating in acute mouse cortical hippocampal brain slices with unprecedented spatial and temporal resolution (see, e.g., Hu X, Khanzada S, Klütsch D, Calegari F, Amin H. Implementation of biohybrid olfactory bulb on a high-density CMOS-chip to reveal large-scale spatiotemporal circuit information, Biosensors Bioelectronics, 2022 Feb 15; 198:113834, doi:10.1016 / j.bios.2021.113834, Epub 2021 Nov 24, PMID:34852985, and Emery BA, Hu X, Khanzada S, Kempermann G, Amin H. High-resolution CMOS-based biosensor for assessing hippocampal circuit dynamics in Experience-dependent plasticity, Biosensors Bioelectronics, October 1, 2023; 237:115471, doi:10.1016 / j.bios.2023.115471, Epub June 12, 2023, PMID:37379793).

[0031] While the implementation of MEA-seqX as described herein involves a CMOS-MEA (4096 electrodes) (13, 16) constructed based on active pixel sensor technology (54), the method is not limited to any particular technology and is also suitable for accommodating a range of high-density technologies, such as those provided by switch matrix technology (26,400 electrodes) (55). The applications of MEA-seqX extend beyond ex vivo applications to include a wide range of in vivo studies.

[0032] Preferably, the method according to the invention includes a mode for recording local field potentials (LFP) that utilizes a high-density CMOS-based biosensing MEA chip (CMOS-MEA).

[0033] In a preferred embodiment of the method according to the invention, the characterized functional neuronal cell assembly is approximately 300 μm thick. Neuronal tissue (aggregates) in the form of dorsal horizontal brain slices. In the context of this invention, neuronal tissue (including cell aggregates) of any thickness can be used for LFP analysis, which still retains functionality and thus allows for the recording of suitable data. An example is approximately 500... With 100 Between, preferably about 200 With 400 Between slices / tissue samples.

[0034] Preferred is the method according to the invention, wherein the evaluation of the recorded LFPs and the assignment of the recorded LFPs to specific characteristics and shapes of the waveforms from interconnected slices / tissues and thus network layers include grouping of electrodes such as those used in arrays and the use of principal component analysis (PCA) and K-means clustering algorithms.

[0035] While acquiring spatial localization patterns of LFPs substantially the entire ensemble using optical imaging, recording of patterns of local field potentials (LFPs) is performed. Imaging is used to provide localization and / or allocation relative to signals detected by, for example, the array used. Given the provision of accurate spatial and temporal dynamics of LFPs for the purpose of providing suitable, standardized, and reproducible data for further analysis, the method according to the invention is preferred, wherein the term "simultaneously" should mean a time span of less than 30 seconds, preferably less than 10 seconds, more preferably less than 1 second, and most preferably, recording of LFPs and optical imaging is performed simultaneously.

[0036] In the context of this invention, the term "substantially" should mean a region, surface, or number that allows for the detection, determination, and thus calculation or prediction of information or data with statistically high correlation to the entire region, surface, or number being analyzed. In one example according to the invention, optical imaging is used to analyze substantially the entire cell assembly to obtain the spatial localization pattern of the LFP. Therefore, the region being analyzed must be sufficient to obtain the spatial localization pattern of the LFP, which has a statistically high correlation to the entire cell assembly present, as in the formulation used. Statistically high correlation in... Preferably at More preferably (For example, ANOVA).

[0037] Preferably, the method according to the invention includes, for example, imaging of the entire cell assembly network using bright-field imaging (including imaging of the entire cell assembly network using a microscope that can be equipped with a high-resolution imaging device, such as a modular stereomicroscope).

[0038] Since the cell assembly being analyzed needs to remain viable, particularly when studying the use of compounds that may affect cell properties and / or activity (see below), the method according to the invention is preferred, wherein the neuronal cell assembly is continuously perfused with a suitable perfusion solution during recording. Suitable media are known to those skilled in the art and are described herein.

[0039] In the next step, optical imaging and spatial resolution transcriptomic analysis of the cells in the ensemble are then performed using a second suitable array. This can be done directly after the above recording. However, the method according to the invention is preferred, wherein essentially the entire ensemble is cryopreserved and can be stored prior to further transcriptomic analysis. A rapid freezing method is required to preserve the ensemble state as the basis for LFP recordings. Preferably, the method of the invention includes cryopreservation, which includes rapid freezing, preferably on dry ice, and optionally also includes freezing at a suitable temperature (e.g., approximately...). Storage. Further preferred is the method according to the invention, wherein the cell collection is appropriately stained prior to transcriptomic analysis, for example, H&E as described herein. The appropriate staining methods are also known to those skilled in the art.

[0040] To further optimize cell number at each point and provide a clearer transcriptomic profile, the thickness of the neuronal ensemble can be further reduced before performing transcriptomic analysis. For this purpose, the slices can be further frozen horizontally to approximately 50°C before transcriptomic analysis. Up to 10 The thickness between, preferably about 25 Up to 15 The thickness between them is preferably about 18. The thickness.

[0041] As described above, the preferred method according to the invention includes optical imaging for transcriptome analysis, such as imaging of the entire cell assembly network using bright-field imaging (including imaging of the entire cell assembly network using a microscope that can be equipped with a high-resolution imaging device, such as a modular stereomicroscope).

[0042] Then, spatial (local) resolution transcriptomics analysis of the cells in the ensemble is performed using a second suitable array. Methods for performing spatial (local) resolution transcriptomics analysis on an array are known in the art (e.g., in Duan H, Cheng T, Cheng H, Spatially resolved transcriptomics: advances and applications, Hematology, Nov 4, 2022; 5(1): 1-14, doi: 10.1097 / BS9.0000000000000141, PMID: 36742187; PMCID: PMC9891446; or Wang X, Almet AA, Nie Q, The promising application of cell-cell interaction analysis in cancer from single-cell and spatial transcriptomics, Semin Cancer Biol, Oct 2023; 95: 42-51, doi: 10.1016 / j.semcancer.2023.07.001, Epub). On July 15, 2023, PMID: 37454878; PMCID: PMC10627116 (in Chinese) and as described in this article.

[0043] Preferably, the method according to the invention includes the generation of barcode-tagged cDNA and sequencing on a second array, such as quantitative sequencing, such as spatial resolution transcriptomics (SRT) sequencing.

[0044] Another important aspect of the method according to the invention involves the collection, processing, and use of the large amount of data generated in the steps above. The acquired data needs to be processed to align the spatial localization patterns of LFPs with spatially resolved sequencing and transcriptome analysis data. Here, the combined input expression matrix (V) captures collective information from the SRT and n-Ephys (LFP rate) datasets (see...). Figure 2Applications of unsupervised machine learning, such as sparsely constrained NMF, enable joint analysis of SRT and LFP (n-Ephys) data to reveal distinct subnetworks of genetic, spatial, and electrophysiological features. Coordinating data sources toward a common high-resolution spatiotemporal genomic functional dimension and representation is preferably achieved using computational pipelines in Python.

[0045] As disclosed herein, one of the primary objectives of this method is to provide the timing and topology of ensemble activity in a high-resolution representation. This allows for the detection of spatiotemporal electrophysiological dynamics and spatial transcriptional profiles of cells within ensembles of functional neurons. Graph-theoretic analysis in the preferred embodiments described herein reveals small-world topology of hub complexes with dense connections, indicating molecular functional specialization and increased global communication capabilities across scales (29, 32, 50). Furthermore, by combining DPT and CAT analyses, this method tracks the pseudo-temporal ordering of cells along developmental trajectories and the progression of large-scale neural activity within their spatial context. This provides unique insights into the coordinated spatiotemporal dynamics of the hippocampal circuit.

[0046] A preferred aspect of the method according to the invention is processing data and aligning the spatial localization pattern of the LFP with spatial resolution sequencing and transcriptome analysis, including overlaying an optical image with an electrode array layout and overlaying an optical image with a second array layout, and the alignment of the two overlays includes image resizing and rotation.

[0047] The unique feature of the method according to the invention lies in its unprecedented ability to simultaneously capture and integrate molecular and functional information across multiple spatiotemporal scales within intact brain tissue. While existing technologies offer valuable insights, the method according to the invention further pushes boundaries by combining the advantages of various technologies. The method of the invention offers a unique solution to the limitations of spatiotemporal resolution by utilizing the high-density capabilities of CMOS-MEA technology combined with spatial transcriptomics, optical imaging, and computational tools.

[0048] Compared with methods according to the prior art, the method according to the present invention offers several advantages:

[0049] - The method according to the invention allows for the generation of knowledge about the molecular structure and spectrum of complex and extensive neuronal / brain activity;

[0050] - The method according to the invention provides a comprehensive perspective from the molecular level to the functional network level, which was not possible with previous methods and therefore could only be studied independently and separately.

[0051] - The predictive power of machine learning algorithms allows for accurate prediction of network-wide electrophysiological characteristics based on spatial gene expression, revealing multi-scale causal relationships between specific gene expression and neural activity;

[0052] - As mentioned in this paper, this method can be adapted to provide additional applications beyond neural cell sets, such as in other bioelectric tissues like the olfactory bulb and cardiac tissue, to offer insights into a variety of biological contexts.

[0053] To demonstrate the power of the method of the present invention, the inventors applied it to a classic enrichment environment paradigm of experience-dependent plasticity, in which the experimental intervention was performed on different housing conditions of laboratory mice. Enriched mice (ENR) were housed in larger enclosures and larger groups of syngeneic animals compared to standard-fed animals (SD) ((16, 21)). This highly influential paradigm elicited structural and functional changes throughout the brain and hippocampus. The inventors were recently able to demonstrate that its effects on the scale of the event correlated with changes at the level of hippocampal circuits (16).

[0054] According to the present invention, the MEA-seqX implementation allows for the exploration of computational dynamics and connectosomes in large-scale hippocampal networks, connecting to underlying transcriptional dynamics. The inventors' method enables insights into the causal relationships between these two dynamics. The present invention allows for the identification of the molecular identity and dynamics of large-scale neural circuits, and thus allows for the identification and development of novel multimodal models and biomarkers with high functional effectiveness in both health and disease contexts. Such biomarkers can then be used to develop diagnostic and screening tools, particularly, but not limited to, precision medicine.

[0055] The method according to the invention also allows for cell type identification within neuronal cell ensembles and highlights network-wide spectral fingerprints and heterogeneity of neuronal cell types within these ensembles (illustrated here in the hippocampal circuit). Furthermore, the platform's predictive capabilities using machine learning algorithms allow for accurate prediction and forecasting of network-wide electrophysiological characteristics based on spatial gene expression data, demonstrating multi-scale causal relationships between specific gene expression and neural activity, providing a deeper understanding of neural dynamics. This also advances research in machine learning and artificial intelligence.

[0056] In another aspect of the invention, the method according to the invention provides for identifying the composition of a functional neuronal ensemble. The method includes performing the method according to the invention as disclosed herein, and further includes the step of inferring the composition of the functional neuronal ensemble based on spatiotemporal electrophysiological dynamics and transcriptomic profiles of detected cells.

[0057] A further preferred embodiment of the method according to the invention includes: performing atlas enhancement or mapping on the cells in the set, identifying cell clusters and / or cell types such as functional clusters in the set, identifying different nerve cell types in the set, and / or identifying the electrical or spectral properties of the cells in the set.

[0058] This allows inventors to determine cell types and local tissue composition based on deconvolutioned gene expression patterns in order to construct multiscale spatial maps of neuronal heterogeneity and their firing characteristics within the same neurons (e.g., hippocampal cortex).

[0059] As an example, analysis conducted in the CA3 region revealed an increased proportion of spatially localizing marker genes Nosl and Inhba associated with pyramidal cell type (see also...). Figure 5 The initial application of CARD yielded a broad group classification of hippocampal cell types. This diverse group included astrocytes, endothelial cells, ependymal cells, macrophages, microglia, neurogenic cells, neurons, oligodendrocytes, and NG2 cells. Prior to any filtering, the inventors identified 85 distinct cell types, highlighting the experimental effectiveness of their slicing technique. After removing low-count cell types, the inventors retained a robust group of 76 cell types. Notably, when these cell types were exposed to two different transcriptomic inputs from SD and ENR, the inventors observed a consistent distribution of prominent cell types across both transcriptomic groups. Figure 5 (a). This result highlights the robustness of the inventor's method and the reproducibility of the inventor's findings.

[0060] Similarly, in another aspect of the invention, the method according to the invention provides for monitoring the cellular composition of a functional neuron cell set, comprising repeatedly performing the method according to the invention at different time points and monitoring the cellular composition of the functional neuron cell set based on a comparison of the cellular composition of the functional neuron cell set detected at the different time points.

[0061] In another aspect of the invention, the method according to the invention provides for monitoring the spatiotemporal electrophysiological dynamics and transcriptional profiles of functional neuronal cell assemblies, comprising repeatedly performing the method according to the invention at different time points, and monitoring the spatial electrophysiological and transcriptional information of molecular neuronal cell assemblies based on a comparison of the spatiotemporal electrophysiological dynamics and transcriptional profiles of cells detected at different time points.

[0062] Preferably, the method according to the invention is wherein the set of functional neuronal cells to be monitored and / or otherwise characterized is selected from neuronal tissue derived from mammals with behavioral conditions or disorders, mammals with neurological conditions or disorders (such as neurodegenerative disorders or cancer), and / or mammals being treated for one of these conditions or disorders.

[0063] In the foregoing aspects, this method monitors, analyzes, registers, and / or detects changes in functional neuronal cell assemblies over time. Repetitions can be performed once, twice, or multiple times. Because the structure of the neuronal cell assemblies is disrupted when the transcriptomic analysis steps of the method according to the invention are performed, repeat tests are performed using functionally relevant samples—e.g., adjacent brain slices, a second batch from neurocancer samples, a second sample of interconnected brain cell cultures, samples of cultured neuronal-differentiated iPSCs, adjacent spinal cord samples, and / or adjacent samples of bioelectrical tissues such as the olfactory bulb and cardiac tissue.

[0064] Monitoring may also include analyzing selected biomarkers during the transcriptome analysis step of the method according to the invention as described herein.

[0065] The monitoring according to the invention can also be performed in the context of identifying compounds that affect the spatiotemporal electrophysiology and transcriptional profile and / or composition of cells in a functional neuronal cell set, or as part of identifying compounds that affect the spatiotemporal electrophysiology and transcriptional profile and / or composition of cells in a functional neuronal cell set (see below), i.e., performed with or without the presence of the compound to be screened.

[0066] In another aspect of the invention, the method according to the invention provides for predicting the spatiotemporal electrophysiological dynamics and / or behavior of an ensemble of functional neurons. The method includes performing the method according to the invention and further includes providing spatially resolved transcriptomic data of the ensemble of neurons to be predicted. Then, as described herein, the (specific) electrophysiological dynamics and / or behavior of the ensemble can be predicted based on the data provided (see also...). Figure 6 ).

[0067] As an example of how to investigate whether the expression profiles of a single spatially resolved gene can predict electrophysiological features across the HC network, the inventors employed the gradient boosting (XGBoost) algorithm, known for its strong interpretability through the integration of multiple tree models (47). While previous methods have focused on gene properties related to electrophysiological and morphological diversity across cell types using low-resolution transcriptomics and electrophysiology (e.g., single-cell RNA sequencing and patch-clamp) (4, 5, 38), the method according to the invention assesses whether specific electrophysiological features can be predicted using spatially resolved transcriptomics data. The XGBoost model was trained using 70% of the spatial transcriptomics data points from the detected SRT points (i.e., 333 genes from the spatial background across HC tissues and six gene families as input) for each quantitative n-Ephys measurement. Three spatiotemporal n-Ephys metrics (LFP rate, amplitude, and time delay) were successfully predicted based on differential spatial gene expression. The relationship between cross-validation predictions and true values ​​was assessed using the Pearson correlation coefficient (r) of the SRT-spot-n-EPhys electrode. Figure 6 a and Figure 12 By implementing XGBoost on specific gene families that exhibit higher expression in ENR compared to the SD loop, the inventors observed significantly higher prediction accuracy on the ENR dataset. The XGBoost classifier achieved approximately 93% accuracy on the ENR dataset and approximately 70% accuracy on the SD dataset. Figure 6 (b). These results highlight multi-scale causal relationships between neural activity, plasticity, and the expression of different spatial genes within specific gene families, and align with predictions of network-level function regulated by prior experience (16). This method according to the invention elucidates key genes and cellular pathways shaping neuronal responses and overall brain function, while revealing regulatory mechanisms controlling neurodynamics, plasticity, and disease pathogenesis. The inventors’ integrative approach will significantly reduce experimental complexity. Specific genes used to predict effects on specific electrophysiological characteristics can then be targeted for manipulation to verify the functional role of specific genes (48) and can be used to screen for drugs that affect the role of specific genes in the prevention or treatment of diseases.

[0068] Notably, key genes essential for neuronal activity and function (such as Bdnf, Egrl, Homer 1, Npas4, Gria2, and Campk2a) showed increased expression levels within the DG transcriptome of the ENR, highlighting a rich transcriptional profile induced by environmental enrichment. Figure 8 ).

[0069] Then, another important aspect of the invention relates to a method for identifying compounds that affect the spatiotemporal electrophysiology and transcriptional profile and / or composition of cells in a functional neuronal cell set, the method comprising performing the method according to the invention in the presence and absence of at least one candidate compound, wherein, in the presence and absence of at least one candidate compound or compared with suitable controls, differences in the spatiotemporal electrophysiological dynamics and transcriptional profile and / or composition of cells in a functional neuronal cell set identify compounds that affect the spatiotemporal electrophysiological dynamics and transcriptional profile and / or composition of cells in a functional neuronal cell set.

[0070] Preferably, the method according to the invention refers to a therapeutic effect, such as modification of cellular composition, alteration of gene expression, changes in electrophysiological dynamics and properties, anticancer effects, neuroprotective effects, morphological effects, cognitive effects, and / or behavioral effects. As described herein, the corresponding changes can be identified and / or include methods known to those skilled in the art. Understanding the relationship between gene expression patterns and neural activity further allows for the early detection and prevention of neurological disorders. For example, identifying specific gene expression profiles associated with certain neural activities allows for the development of targeted interventions to alter or prevent these neural activities, thereby providing preventative approaches to neurological diseases.

[0071] In the context of this invention, candidate compounds may be selected from the group consisting of: chemical organic molecules, molecules selected from libraries of small organic molecules (molecular weight less than about 500 Da), molecules selected from combinatorial libraries, cell extracts (especially plant cell extracts), small molecule drugs, proteins, protein fragments, molecules selected from peptide libraries, antibodies or antigen-binding fragments thereof, or nucleic acids, DNA, RNA or siRNA.

[0072] These candidate molecules can also serve as a basis for screening improved compounds. Therefore, the method according to the invention is preferred, wherein, after identifying the compound and / or its changes / influences on the spatiotemporal electrophysiological dynamics and transcriptional profiles and / or composition of cells in a functional neuronal cell set, the method further includes a step of chemically modifying the compound. Generally, many methods for modifying the compounds of the invention are known to those skilled in the art and are disclosed in the literature.

[0073] Modifications to compounds are generally categorized into several types, such as a) amino acid mutations / changes to different amino acids, b) chemical modifications, for example, by adding additional chemical groups, c) changes in the size / length of the compound, and / or d) attaching additional groups to the molecule (including labeling groups, tags, linkers, or carriers such as chelating agents). In the next step, the modified compound is tested again in at least one of the tests described above, and whether the properties of the compound are improved compared to its unmodified state. In the context of this invention, "improvement" activities involve the desired increase or decrease in alterations / effects on the spatiotemporal electrophysiological dynamics and transcriptional profiles and / or composition of cells in a functional neuronal cell set.

[0074] Another aspect of the invention relates to a method for preparing a pharmaceutical composition, the method comprising performing a method according to the invention for identifying a compound and mixing the identified compound with at least one pharmaceutical carrier.

[0075] The term "pharmaceutical composition" refers to a formulation in a form that allows the biological activity of the active ingredient to be effective without containing any additional ingredients that would be unacceptably toxic to the subject to whom the composition will be administered. The pharmaceutical compositions of the present invention can be administered by various methods known in the art. As will be understood by those skilled in the art, the route and / or mode of administration will vary depending on the desired outcome. Pharmaceutically acceptable diluents include physiological saline and aqueous buffer solutions. "Pharmaceutically acceptable carrier" refers to a component of the pharmaceutical formulation that is non-toxic to the subject, other than the active ingredient. Pharmaceutically acceptable carriers include any and all physiologically compatible solvents, dispersion media, coating agents, antibacterial and antifungal agents, isotonic agents, and absorption delay agents, etc. Carriers can be suitable for intravenous, intramuscular, subcutaneous, parenteral, spinal, or epidermal administration (e.g., by injection or infusion).

[0076] The pharmaceutical compositions prepared according to the present invention may further comprise excipients such as preservatives, wetting agents, emulsifiers, and dispersants. The presence of microorganisms can be prevented by sterilization procedures and by including various antibacterial and antifungal agents, such as parabens, chlorobutanol, phenols, sorbic acid, etc. It may also be desirable to include isotonic agents such as sugars, sodium chloride, etc., in the composition. Additionally, prolonged absorption of the injectable pharmaceutical form can be achieved by including agents that delay absorption (such as aluminum monostearate and gelatin). The pharmaceutical compositions produced according to the present invention can be in liquid, dry, or semi-solid form, such as, for example, tablets, coated tablets, effervescent tablets, capsules, powders, granules, sugar-coated tablets, lozenges, pills, injections, drops, suppositories, emulsions, ointments, gels, tinctures, pastes, creams, wet dressings, mouthwashes, liquids, nasal preparations, inhalation mixtures, aerosols, mouthwashes, oral sprays, nasal sprays, or room sprays. Injectable compositions are preferred.

[0077] Another aspect of the invention relates to apparatus particularly suited for performing the method according to the invention as described herein, specifically comprising: a collection of neuronal cells to be analyzed; an electrode array suitable for recording patterns of local field potentials (LFPs) of the collection during optical imaging; a second array suitable for performing spatial resolution transcriptomic analysis of the cells in the collection while also performing optical imaging; and a computer program product for processing the acquired data, the computer program product including programming means for collecting and processing the data as generated, particularly for aligning spatial localization patterns of LFPs with spatial resolution sequencing and transcriptomic analysis and / or principal component analysis (PCA) and K-means clustering algorithms, the programming means optionally having additional buffers (see, for example, examples below) and instructions for use.

[0078] Another aspect of the invention relates to the use of a device according to the invention for detecting spatiotemporal electrophysiological dynamics and transcriptional profiles of cells in a functional neuronal cell ensemble, for identifying the composition of a functional neuronal cell ensemble, for monitoring the spatial electrophysiological dynamics and transcriptional profiles of cells in a functional neuronal cell ensemble, for monitoring the cellular composition of a functional neuronal cell ensemble, or for identifying compounds that affect spatiotemporal electrophysiological dynamics and transcriptional profiles and / or for identifying the cellular composition of a functional neuronal cell ensemble.

[0079] Then, another aspect of the invention relates to a method for preventing or treating neurological diseases or disorders and / or improving the cognitive function of a subject who requires such prevention, treatment or improvement, comprising administering to the subject an effective amount of a pharmaceutical composition prepared according to the invention.

[0080] Preferably, the method according to the invention is wherein the neurological disease or disorder in the subject is selected from the group comprising diseases or disorders that cause undesirable alterations / influences on the spatiotemporal electrophysiological dynamics and transcriptional profiles and / or composition of cells in a functional neuronal cell set, arrhythmias, hormonal disorders, Parkinson's disease, Alzheimer's disease, dementia, behavioral disorders, cognitive impairment, brain cancer, stroke, depression, affective disorders, comorbid depressive symptoms, substance addiction and abuse, obsessive-compulsive disorder, personality disorders (e.g., borderline), autism, Tourette syndrome, and anxiety disorders.

[0081] In the context of this invention, "subject" refers to a mammal, animal, or individual, such as a person or patient who has undergone prevention or treatment for a mental disorder or disease according to the present invention. In the context of this invention, the term "treatment" or "therapy" should mean an attempted remedy for a health problem (i.e., the subject's mental disorder or disease) as disclosed herein.

[0082] In the context of this invention, the term "about" should be meant to include a deviation of + / -10% of a given value, unless otherwise indicated.

[0083] The method according to the invention introduces a platform that provides an unprecedented ability to simultaneously capture and integrate molecular and functional information across multiple spatiotemporal scales within intact neural (e.g., brain) tissue. The method according to the invention offers a unique solution to limited spatiotemporal resolution by leveraging the high-density capabilities of CMOS-MEA technology combined with spatial transcriptomics, optical imaging, and computational tools to provide superior temporal resolution down to the individual cells within the spatial context of each cell. The invention fills a critical gap in understanding the possibility of understanding the molecular infrastructure supporting and resolving the integrity of large-scale neuronal interactions within physiological and experience-dependent plasticity paradigms (18, 49). Through the proposed comprehensive approach, the inventors identify spatially resolved causal regulation across molecular functional signatures, revealing the influence of environmental factors on coordinated neural activity and gene expression—a dubious but previously largely inaccessible link.

[0084] Graph theory analysis of the platform reveals small-world topology with tightly connected hub complexes, indicating molecular functional specialization and increased global communication capabilities across scales (29, 32, 50). Furthermore, by combining DPT and CAT analyses, the method according to the invention tracks the pseudo-temporal ordering of cells along developmental trajectories and the progression of large-scale neural activity within their spatial context, providing unique insights into the coordinated spatiotemporal dynamics of the hippocampal circuit. The method according to the invention also demonstrates the potential for cell type identification and highlights the heterogeneity of neural cell types and their network-wide spectral fingerprints within the hippocampal circuit. The predictive power of the platform using machine learning algorithms allows for accurate prediction of network-wide electrophysiological characteristics based on spatial gene expression, revealing multi-scale causal relationships between specific gene expression and neural activity, providing a deeper understanding of neural dynamics and opening new avenues for research in machine learning and artificial intelligence. Combining the method according to the invention with intelligent neural networks enhances the understanding of complex data, decision-making processes, and learning mechanisms (51, 52).

[0085] Furthermore, the method according to the invention holds promising potential in personalized medicine. The identification of spatiotemporal transcriptomic electrophysiological biomarkers may aid in the early diagnosis and treatment of neurological disorders, ultimately leading to precise personalized medicine (53).

[0086] Furthermore, in the context of other neurological and psychiatric disorders, the method according to the invention will deepen our understanding of disease mechanisms and therapeutic targets. The platform's potential to reveal molecular functional changes associated with various disease states paves the way for the development of novel therapies.

[0087] Further integration of additional modalities into the method according to the invention, such as proteomics and epigenomics data, will provide even more comprehensive insights into neural information processing (51). The platform's adaptability extends beyond neural tissue to bioelectrophysiological tissues such as the olfactory bulb and heart. This enables the study of gene expression and electrophysiological interactions in different contexts. Its application in heart tissue provides insights into cardiac function, thus contributing to advancements in cardiology.

[0088] The method according to the invention is also adaptable to a range of high-density technologies, such as those provided by switch matrix technology (26,400 electrodes) (55). Importantly, the use of the method according to the invention extends beyond ex vivo applications to include in vivo studies. Integration with state-of-the-art in vivo probes, such as Neuropixels (56), SiNAPS (57), or other emerging modalities, offers the potential to study functional neural dynamics and gene expression patterns in vivo, bridging the gap between laboratory findings and real-world biological contexts.

[0089] The platform according to the invention (which has the ability to simultaneously capture and integrate molecular and functional information across multiple spatiotemporal scales within intact brain tissue) has enormous potential applications in both research and clinical settings.

[0090] Its primary application lies in providing a deeper understanding of the principles of cross-scale computing in the brain, bridging the gap between molecular architecture and brain activity. This may be key in the following areas:

[0091] - Neural research: Understanding changes in neural plasticity, connectivity dynamics, and molecular function associated with aging, development, and various neural conditions.

[0092] - Drug development and testing: The platform can be used to simultaneously test the effects of new drugs or treatments on neural activity and transcriptome levels to provide insights into their efficacy and potential side effects.

[0093] - Diagnostic tools: With its predictive capabilities using machine learning algorithms, MEA-seqX can be developed into a diagnostic tool that can predict network-wide electrophysiological characteristics based on spatial gene expression.

[0094] - Precision Personalized Medicine: The platform holds the potential to customize diagnostic tools and treatments based on individual genetic neurophysiological characteristics.

[0095] - Early detection (prevention) of neurological disorders: MEA-seqX's ability to detect subtle changes in gene expression and neural activity may aid in the early diagnosis of neurological disorders. For example, certain gene expression patterns can indicate early stages of neurodegenerative disorders, such as Alzheimer's or Parkinson's, well before clinical symptoms appear.

[0096] - Disease progression monitoring: The method's ability to track changes in neural activity and gene expression over time makes it an excellent tool for monitoring the progression of neurological disorders. This may help adjust treatment plans as the disease progresses.

[0097] - Biomarker Discovery: MEA-seqX's comprehensive data analysis can help discover new biomarkers for neurological diseases. By identifying novel genes or neural activity patterns associated with specific conditions, it can expand the range of available diagnostic markers.

[0098] - Understanding disease mechanisms: Integrating genetic and neural data can provide insights into the underlying mechanisms of neurological disorders. This deeper understanding can not only aid in diagnosis but also contribute to the development of more effective treatments.

[0099] In the context of this invention, the methods provided go beyond single-cell approaches and focus on multi-scale network-level dynamics. This invention provides a more comprehensive and detailed understanding of hippocampal cortical cell types and their interactive, sequential electrophysiological properties across multiple scales.

[0100] This invention relates to the following items:

[0101] Project 1. An ex vivo method for detecting spatiotemporal electrophysiological dynamics and transcriptional profiles of a functional neuronal cell ensemble, comprising providing a functional neuronal cell ensemble to be characterized; recording patterns of local field potentials (LFPs) of the neurons in the ensemble with high spatial and temporal resolution using a suitable electrode array; obtaining spatial localization patterns of the LFPs substantially throughout the ensemble using optical imaging, optionally cryopreserving substantially the entire ensemble; performing optical imaging and spatially resolved transcriptomic analysis of the cells in the ensemble using a second array; and processing the acquired data, including aligning the spatial localization patterns of the LFPs with the spatially resolved transcriptomic analysis; and thereby detecting the spatiotemporal electrophysiological dynamics and spatial transcriptional profiles of the cells in the functional neuronal cell ensemble.

[0102] Project 2. According to the method of Project 1, wherein the functional neuronal cell set to be characterized is selected from neuronal tissue, brain slices, neurocancer samples, cultures of interconnected brain cells, cultured neuronal differentiated iPSCs, spinal cord, bioelectric tissues such as olfactory bulbs and cardiac tissue, and parts or combinations thereof, particularly derived from neuronal tissue of mammals with behavioral conditions or disorders, mammals with neurological conditions or disorders such as neurodegenerative disorders or cancer, and / or mammals being treated for one of these conditions or disorders.

[0103] Project 3. According to the method of Project 1 or 2, wherein the mode of recording local field potential (LFP) includes the use of a high-density CMOS-based biosensing MEA chip (CMOS-MEA).

[0104] Project 4. The method of any one of Projects 1 to 3, wherein optical imaging includes, for example, using bright-field imaging, including imaging the entire network of cell assemblies using a microscope such as a modular stereomicroscope.

[0105] Project 5. According to the method of any one of Projects 1 to 4, wherein the neuronal cell set is continuously perfused with a suitable perfusion solution during the recording period.

[0106] Project 6. According to the method of any one of Projects 1 to 5, wherein recording and optical imaging of the LFP are performed simultaneously.

[0107] Item 7. The method according to any one of Items 1 to 6, wherein cryopreservation includes rapid freezing, preferably freezing on dry ice, and optionally further includes... storage.

[0108] Project 8. According to the method of any one of Projects 1 to 7, wherein the cell collection is appropriately stained prior to transcriptome analysis.

[0109] Project 9. According to the method of any one of Projects 1 to 8, wherein transcriptomics includes generating barcoded cDNA and performing sequencing such as quantitative sequencing, such as spatial resolution transcriptomics (SRT) sequencing on a second array.

[0110] Project 10. According to the method of any one of Projects 1 to 9, wherein evaluating the recorded LFP and assigning the recorded LFP to specific features and shapes of waveforms from interconnect network layers includes grouping electrodes and using principal component analysis (PCA) and K-means clustering algorithms.

[0111] Project 11. The method of any one of Projects 1 to 10, wherein processing the data and aligning the spatial localization pattern of the LFP with spatial resolution sequencing and transcriptome analysis includes overlaying an optical image with an electrode array layout and overlaying an optical image with a second array layout, and the alignment of the two overlays includes image resizing and rotation.

[0112] Item 12. A method for identifying the composition of a functional neuronal ensemble, comprising performing the method according to any one of Items 1 to 11, and further comprising the step of inferring the composition of the functional neuronal ensemble based on the spatiotemporal electrophysiological dynamics and transcriptomic profiles of the detected cells.

[0113] Project 13. According to the method of Project 12, wherein identification includes mapping cells in the set, identifying clusters of cells in the set such as functional clusters, identifying different nerve cell types in the set, and / or identifying spectral properties of cells in the set.

[0114] Project 14. A method for monitoring the spatiotemporal electrophysiological dynamics and transcriptional profiles of a functional neuronal cell ensemble, comprising repeatedly performing the method according to any one of Projects 1 to 11 at different time points, and monitoring the spatial electrophysiological and transcriptional information of the molecular neuronal cell ensemble based on a comparison of the spatiotemporal electrophysiological dynamics and transcriptional profiles of cells detected at different time points.

[0115] Item 15. A method for monitoring the cellular composition of a functional neuron cell set, comprising repeatedly performing the method according to Item 12 or 13 at different time points, and monitoring the cellular composition of the functional neuron cell set based on a comparison of the cellular composition of the functional neuron cell set detected at the different time points.

[0116] Project 16. According to the method of Project 14 or 15, wherein the set of functional neurons to be characterized is selected from the neuronal tissue of mammals with behavioral conditions or disorders, mammals with neurological conditions or disorders such as neurodegenerative disorders or cancer, and / or mammals being treated for one of these conditions or disorders.

[0117] Project 17. A method for identifying compounds that affect the spatiotemporal electrophysiology and transcriptional profiles and / or composition of cells in a functional neuronal cell ensemble, comprising performing the method according to any one of Projects 1 to 13 in the presence and absence of at least one candidate compound, wherein, in the presence and absence of at least one candidate compound or when compared with a suitable control, differences in the spatiotemporal electrophysiological dynamics and transcriptional dynamics and / or composition of cells in a functional neuronal cell ensemble identify compounds that affect the spatiotemporal electrophysiological dynamics and transcriptional profiles and / or composition of cells in a functional neuronal cell ensemble.

[0118] Project 18. Based on the method of Project 17, where the effects are therapeutic effects, such as modifications to cellular composition, alterations in gene expression, anticancer effects, neuroprotective effects, morphological effects, cognitive effects, and / or behavioral effects.

[0119] Item 19. An apparatus for performing the method according to any one of items 1 to 18, particularly comprising a collection of neuronal cells to be analyzed, an electrode array adapted to record patterns of local field potentials (LFPs) of the collection during optical imaging, a second array adapted to perform spatial resolution transcriptome analysis of the cells in the collection during optical imaging, and a computer program product for processing the acquired data, the processing of the acquired data including aligning the spatial localization patterns of the LFPs with spatial resolution sequencing and transcriptome analysis, the computer program product optionally having additional buffers and instructions for use.

[0120] Item 20. Use of an apparatus according to Item 19 for detecting spatiotemporal electrophysiological dynamics and transcriptional profiles of cells in a functional neuronal cell ensemble, for identifying the composition of a functional neuronal cell ensemble, for monitoring the spatial electrophysiological dynamics and transcriptional dynamics of cells in a functional neuronal cell ensemble, for monitoring the cellular composition of a functional neuronal cell ensemble, or for identifying compounds that affect spatiotemporal electrophysiological dynamics and transcriptional profiles and / or for identifying the cellular composition of a functional neuronal cell ensemble.

[0121] The invention will now be further described with reference to the accompanying drawings in the following examples, but is not limited thereto. For the purposes of this invention, all references cited are incorporated herein by reference in their entirety.

[0122] In the attached diagram,

[0123] Figure 1 An overview of the MEA seqX platform for cross-scale integration of brain dynamics is presented: a), e) The platform combines high-density CMOS-MEA, optical microscopy, and spatial sequencing technologies to examine the relationship between spatial transcriptomics and neural oscillation dynamics; b) A more detailed examination, illustrated by sensing electrodes coupled to SRT points, highlights the discernible difference in spacing size between the two platforms; c) High-density CMOS-MEA equipped with 4,096 sensing electrodes skillfully captures functional firing patterns (LFP) from all HC regions in spatial coordinates with high precision; d) Systematic mapping of gene expression profiles is achieved using a spatially barcoded SRT with 5,000 points, elucidating spatial nuances across HC regions; f) Python-based pipelines process the data to map transcriptional functional dynamics. This includes multifaceted spatial rescaling and alignment methods to establish direct correlations between neural activity and transcriptomic features; g) advanced analysis from multidimensional readings provides insights into multi-scale network features; h) topological maps capture complex networks of multimodal transcriptional functional connections; i) quantification of multi-scale neural dynamics with gene pseudotime and activity center trajectories; j) inference of cell type composition from spatial transcriptomic data associated with their firing characteristics obtained from n-Ephys; and k) automated machine learning algorithms to accurately predict electrophysiological features from transcriptomic profiles.

[0124] Figure 2This study integrates spatial gene expression patterns and network electrophysiological features. a) Comparative analysis of SRT expression patterns and functional n-Ephys features revealed enhanced gene expression patterns in the ENR corresponding to the DG and CA3 regions compared to SD. The statistical significance of the correlation between SRT and n-Ephys features was quantified using p-values ​​(padj) adjusted for Benjamin-Hochberg false discovery rate (FDR). b) Quantitative profiling of related genes based on network-wide LFP rates revealed an increase in related genes in the DG region of the ENR compared to SD, and a 2.5-fold and 0.6-fold increase in the CA3 and CA1 subregions, respectively. The significance of the correlation between SRT and LFP was measured using p-values ​​(padj) adjusted for Benjamin-Hochberg FDR. c) Unsupervised machine learning was applied via sparse-constrained NMF. The algorithm implements joint analysis of SRT and n-Ephys data to reveal different sub-networks of electrophysiological features, spatial locations, and genes. d) The combined input expression matrix (V) captures collective information from the SRT and n-Ephys (LFP rate) datasets. Each entry represents gene expression associated with network functional features at a specific spatial location. e) Spatial gene expression patterns are visualized through the foundation matrix (W). This matrix shows how gene expression patterns are distributed across spatial locations and associated with functional features. f) The coefficient matrix (H) represents the contribution of the foundation vectors to spatial location and n-Ephys features. The analysis reveals increased normalized gene expression and network eigenvalues ​​in the ENR network to highlight the spatially resolving components driving gene strength.

[0125] Figure 3Multiscale network topology analysis of spatial sequencing connectons and neural functional connectivity is presented. a) A plot of key multiscale graph measurements depicts characteristic hub nodes and rich club nodes defined based on their degree of interconnectivity in acute hippocampal cortical slices. Node degree corresponds to the number of attached links at a given node. b) Connectivity graphs of spatial IEG in the HC interconnection layer in ENR and SD. The network is visualized using Gephi to show the total connectivity from full-length sequencing SD data (nodes = 546, and links = 5,166) and ENR data (nodes = 877, and links = 50,974). c) Connectivity graphs of spatiotemporal functional neural activity in the HC interconnection layer under ENR and SD conditions. The network is visualized using Gephi to show 2% of the total connectivity in SD (nodes = 1,057, and links = 24,217) and large-scale ENR recordings (nodes = 2,003, and links = 36,312). The graph nodes in (b) and (c) are scaled according to degree intensity and colored according to HC module associations, indicated in the colored circle legend. Colored links identify intra- and inter-cluster connections. d) Percentage of quantified hub nodes and rich club nodes in different hippocampal transcriptome networks in SD and ENR. e) Percentage of quantified hub nodes and rich club nodes in different hippocampal functional networks in SD and ENR. f) Power-law distribution indicating the topology of scale-free transcriptome (SRT) functional (n-Ephys) networks with small-world properties in SD and ENR networks. The logarithmic plot of the cumulative connectivity distribution of the ENR (SRT and n-Ephys; blue) network shows a significantly heavier tail compared to the SD network (SRT and n-Ephys, gray), indicating the coexistence of low-order nodes with a few densely connected hubs, but higher than the SD network, which meets the cut Kolmogorov-Smirnov test. This is also supported at linear scales (inset) in all conditions, and their compliance with power laws is fitted by Pareto (). The Kolmogorov-Smirnov test is used to evaluate the goodness of fit of the log-normal function on the logarithmic curve (where the coefficients of determination of SRT, n-Ephys (SD), SRT, and n-Ephys (ENR) are R0, ... 2 The power-law distributions were fitted with values ​​of 0.95, 0.98, 0.96, and 0.97 respectively. Pareto's probability density function fits the linear curve with goodness of fit (where Rt = 0.95, 0.98, 0.96, and 0.97 respectively). 2 The values ​​of 0.96, 0.97, 0.95 and 0.97 respectively are fitted with power-law distributions.

[0126] Figure 4Analysis of multi-scale hippocampal dynamics is shown: a) Differential progression of DPT in IEG expression from SRT SD data illustrates the developmental trajectory of cells within hippocampal spatial regions; b) Same as in (a), but in ENR; c) Correspondence between DPT and CAT analyses in SD to infer colocalization, spatiotemporal alignment, and functional insights; d) Same as in (c), but in ENR; e) Quantification of differential DPT progression in interconnected hippocampal regions in SD and ENR. ANOVA test).

[0127] Figure 5 Analysis of spatial cell types and their firing pattern fingerprints is presented: a) Regionally ordered heatmaps based on the CARD method depicted the proportions of 76 filtered cell types. Insertion lists and local cell type highlighting overlays on SD and ENR brain slices revealed the 20 cell types with the highest proportions. Comparative analysis of the hippocampal transcriptome between SD and ENR conditions revealed comparable distributions of prominent cell types; b) Integrating the CARD method into the MEA-seqX platform yielded spatially resolved composition of cell types accompanied by underlying oscillatory firing characteristics. This presentation included the top 20 cell types with the highest proportions, as determined by SRT readings, LFP rates, and oscillatory waveforms derived from n-Ephys readings across the entire hippocampal network; c) Focused examination of the DG region provided in-depth insights into the spatially resolved composition of cell types and their corresponding oscillatory firing characteristics; d) Similarly, regional assessment within the CA3 pyramidal cell (PC) network provided insights into the spatial composition of cell types and their corresponding oscillatory firing characteristics; e) The ENR transcriptome showed a higher proportion of spatially localized marker genes Cck and Penk associated with granule cell types in the DG. These proportions were significantly increased compared to the SD transcriptome. Similarly, the ENR showed a higher LFP rate and amplitude compared to SD. , ANOVA 0.05, ANOVA). f) Analysis in the CA3 region revealed an increased proportion of pyramidal cell type-related spatial positioning marker genes Nosl and Inhba in the ENR transcriptome compared to the SD transcriptome, which was correlated with higher LFP rates and amplitudes at the functional network scale. (ANOVA).

[0128] Figure 6Machine learning predictions based on multiscale transcriptional functional data are shown. a) The XGBoost algorithm is applied to predict network electrophysiological metrics (LFP rate, amplitude, and duration) based on transcriptomic data for each specified gene family in SD and ENR data. The predictions of n-Ephys metrics based on transcriptomic data are evaluated in SD and ENR using Pearson correlation coefficients (r). Significant differences between predicted SD and ENR values ​​across all gene families are indicated. (aNOVA). b) XGBoost performance is indicated by comparing the average accuracy values ​​from the final data output iterations across all gene families, all of which exhibit higher predictive accuracy for ENR data compared to SD data. The values ​​are calculated on the mean, and three standard deviations are determined within the chance threshold.

[0129] Figure 7 The following are presented: a) nFeature RNA and nCount RNA statistics and feature maps show the number of distinct genes and unique molecular identifiers (UMIs) in the transcriptomes of SD (n=2) and ENR (n=2). These distributions are part of quality control, highlighting consistent counts of unique genes and UMIs across all samples and conditions to effectively mitigate potential technical batch effects and experimental variability; b) condition-based UMAP analysis using both Harmony and Seurat procedures to differentiate transcriptomic profiles based on conditions. The resulting UMAPs reveal consistent organizational structures across the two conditions, demonstrating accurate cluster separation at high resolution; c) HC cluster segmentation based on spatial gene spectrum was performed on SD and ENR brain slices; d) selected clusters aligned with DG, CA3, CA1, and EC regions were localized within the UMAP representation. This low-dimensional depiction effectively demonstrates cluster separation based on gene expression profiles to capture tissue heterogeneity and progressive expression changes; e) classification of multi-level HC oscillation waveform shapes was established using PCA analysis and the K-means algorithm. This classification is based on features extracted from LFP events to reveal complex insights into the diversity of waveform patterns across different layers of the hippocampus.

[0130] Figure 8 Elevated transcriptional expression observed across the HC cluster in the ENR compared to the SD condition is shown in (a–f). Genes from six distinct functional families (including Immediate Early Genes (LEG), hippocampal neurogenesis, signaling pathways, receptors and channels, synaptic plasticity, and synaptic vesicles) were systematically classified. Notably, key genes crucial for neuronal activity and function (such as Bdnf, Egr1, Homer1, Npas4, Gria2, and Campk2a) showed increased expression levels within the DG transcriptome of the ENR, highlighting a rich transcriptional profile induced by environmental enrichment. ANOVA).

[0131] Figure 9 The analysis of gene expression interaction matrices within six gene families using mutual information reveals significant differences between SD and ENR conditions. Compared to SD, ENR shows enhanced statistical transcriptional connectivity and coordination activity, indicating a robust effect on HC subnetwork relationships. b) Cross-covariance of paired discharge electrodes was assessed using PCC to reveal complex connectivity patterns. The ENR connectivity matrix exhibits amplified local and global spatiotemporal interactions compared to SD, consistent with cross-correlation plot observations. These findings collectively reveal the subtle influence of ENR on dynamic transcriptional functional interactions within the hippocampal subnetwork.

[0132] Figure 10 The diagram shows the mutual information distance scores calculated for target gene families within SRT points in (a) to reveal differences in interactions within the HC subnetwork. The ENR transcriptome exhibits higher mutual information, indicating robust and coordinated activity and communication within gene expression patterns compared to SD. (aNOVA). (b) Quantize the correlation matrix using the PCC of LFP activities across the interconnect HC layer. ANOVA). As shown in (a), this corresponds to the increased spatial interaction in ENR compared to the SD network.

[0133] Figure 11 The application of sparsity-constrained NMF analysis in a) is shown, such as Figure 2 The results were validated in (c) to (f) and (b) to (d). Analysis quantified the characteristics of the decomposed gene sets, spatial locations, and connectivity to reveal simultaneous variations in network topology metrics associated with spatially resolved IEG expression and network connectivity.

[0134] Figure 12 This paper demonstrates the application of the XGBoost algorithm to predict network electrophysiological measures (LFP rate, amplitude, and duration) based on transcriptomic data from each specified gene family in the SD and ENR datasets. The predictions of n-Ephys measures based on transcriptomic data were evaluated in SD and ENR using the Pearson correlation coefficient (r). Significant differences between predicted SD and ENR values ​​across all gene families are indicated. (ANOVA).

[0135] Example

[0136] The platform of this invention has been tested using brain tissue from healthy mouse models. These mice were housed under different conditions, including standard and enriched environments, to investigate experience-dependent plasticity paradigms. Enriched environmental conditions are known to promote enhanced experience-dependent plasticity by increasing stimuli and differentiated social interactions.

[0137] The platform was employed to simultaneously capture and integrate molecular and functional information across multiple spatiotemporal scales within these intact brain tissues. This involved high-resolution spatiotemporal recording of the dynamics of functional firing patterns from large-scale cell ensembles, optical imaging for spatial localization, and high-resolution multipath analysis of cellular transcriptomics from the same ensembles in the hippocampal cortex.

[0138] The platform has successfully revealed the causal prediction potential of spatiotemporal transcriptomics at the cellular functional scale. This allows for the identification of upstream circuit computational processing and biophysical foundations, which is crucial for mapping the molecular identity of large-scale neural circuits.

[0139] The platform's development and testing have provided valuable insights into the molecular mechanisms underlying functional enhancements induced by varying environmental conditions. Its adaptability and capabilities hold immense potential for use in diagnostic tools and therapies for precise, personalized medicine based on individual genetic neuroelectrophysiological characteristics.

[0140] method

[0141] Animal and acute brain slice preparation

[0142] All experiments were conducted in accordance with applicable European and national regulations (Tierschutzgesetz) on 12-week-old C57BL / 6J mice (Charles River Laboratories, Germany) and approved by the local authorities (LandesdirektionSachsen; 25-5131 / 476 / 14). Female C57BL / 6J mice were obtained at 5 weeks of age and randomly assigned to two experimental groups—standard housing (SD) and enriched environment housing (ENR), as previously described (15). ENR-housed mice lived in specially designed cages containing reconfigurable toys, maze-like plastic tubes, tunnels, shells, and additional nesting materials. This ENR cage environment was shown to promote enhanced experience-dependent plasticity through increased stimulation and different social interactions. Mice were temporarily housed in their assigned environment for up to six weeks prior to the start of the experiment and remained there until the date of the experiment.

[0143] Acute brain sections were prepared according to the inventors' previous research (13, 16). Briefly, mice were anesthetized with isoflurane prior to decapitation. Before slicing, the brain was carefully removed from the skull and placed in a cooled, diluted sucrose solution. The brain was placed in a custom-made agarose container and fixed to a cutting plate. Dorsal horizontal sections (300) were prepared using a Leica Vibratome VT1200S (Leica Microsystems, Germany). Thickness). Slices at 0 Up to 2 The HC sections were then cut in a saturated aCSF solution containing 95% O2 and 5% CO2 (pH = 7.2 to 7.4), which was composed of 250 g sucrose, 10 g glucose, 1.25 g NaH2PO4, 24 g NaHCO3, 2.5 g KCl, 0.5 g ascorbic acid, 4 g MgCl2, 1.2 g MgSO4, and 0.5 g CaCl2 in mM. Next, the HC sections were cut at 34°C. The cells were incubated for 45 minutes and then allowed to recover at room temperature for at least 1 hour before being used for network electrophysiology (n-Ephys) recordings on a high-density neural chip. The perfusion solution used during recording contained 127 mM NaCl, 2.5 mM KCl, 1.25 mM NaH2PO4, 2.4 mM NaHCO3, 2.5 mM glucose, 1.25 mM MgSO4, and 2.5 mM CaCl2, and was aerated with 95% O2 and 5% CO2.

[0144] Extracellular n-Ephys recordings.

[0145] All electrical recordings are performed using a high-density CMOS-based biosensor MEA chip (CMOS-MEA) and an acquisition system customized for the current recording setup (3Brain AG, Switzerland). The CMOS-MEA consists of a 42-bit microcontroller. It consists of 4096 recording electrodes with a spacing of approximately 7mm. 2 The active sensing area. On-chip amplification loops allow for 0.1 kHz to 5 kHz bandpass filtering through a global gain of 60 dB sufficient to record both slow and fast oscillations (13). For extracellular recording, slides were moved and coupled onto the chip using a custom platinum harp placed over the tissue. To minimize experimental variation and maintain slide lifetime, a thermally stable perfusion system delivered oxygenated recording perfusion to the neural chip at a flow rate of 4.5 mL / min, and the temperature was maintained at 37°C throughout the experiments and recordings. The inventor used 100. Extracellular recordings at 14 kHz / electrode sampling frequency and 1 Hz recording frequency were collected from spontaneous network-wide activity via pharmacologically induced responses (16) (Sigma-Aldrich, Germany). All solutions were freshly prepared, and the pharmacological compound was dissolved in the recording perfusion fluid for use in experiments. A custom-designed modular stereomicroscope (Leica Microsystems, Germany) was incorporated into the system to simultaneously capture acute slide optical imaging and extracellular n-Ephys recordings of the entire HC circuit. During offline analysis, these images were used to maintain the spatial organization of brain slide tissue relative to the n-Ephys electrode layout.

[0146] Spatial analysis transcriptomics of hippocampal cortex sections

[0147] SRT Visium gene expression slices (10X Genomics, USA) were used to perform spatial resolution HC sequencing and transcriptome analysis. The SRT slices consisted of four distinct capture regions, each containing 55... 5,000 spatial barcode dots of 1000mm diameter were used to form a capture area of ​​approximately 6.5mm x 6.5mm, large enough to hold an entire mouse HC section. Immediately after n-Ephys recording, the sections were embedded in approximately 6.5mm x 6.5mm tissue-TEK Cryomold solution containing the optimal cutting temperature (OCT), frozen on dry ice, placed in WHEATON CryoELITE tissue vials, and frozen at -80°C. Slices were stored to maintain tissue viability until the date of the SRT experiment. Here, to optimize cell number for each point and provide a clear transcription profile, sections were horizontally frozen to 18°C ​​using a Thermo Fisher Cryostar NX70 (Thermo Fisher Scientific, USA). The tissue was mounted on SRT gene expression sections and fixed with methanol at -20°C. The procedure was performed for 30 minutes. Sections were stained with hematoxylin and eosin (H&E) and bright-field imaging was used to obtain morphological images of the sections. After imaging, the sections were enzymatically permeated on a thermal cycler for 22 minutes, and the resulting released mRNA bound to oligonucleotides containing thousands of spatial barcodes within each spot. To generate cDNA from the oligonucleotide-bound mRNA, an enzyme reverse transcription mixture (10X Genomics, USA) was applied, and the reaction proceeded at 53°C. Incubation was performed in a thermal cycler for 45 minutes. For cDNA second-strand synthesis, an enzymatic second-strand mixture (10X Genomics, USA) was applied and incubated at 65°C. Incubation was performed in a thermal cycler for 15 minutes. To denature the enzyme, alkaline elution buffer (EB) (Qiagen, Germany) at pH 8.7 was applied, and the final samples were stored in corresponding tubes containing Tris-HCl for each capture region. Finally, full-length cDNA with spatial barcodes was prepared by PCR amplification for library sequencing. To determine the optimal cycle number (Cq) via qPCR, a qPCR mixture from a KAPA SYBR FAST (Kapa Biosystems, USA) and 1 qPCR sample from each cDNA sample were used. Samples were added to clean qPCR plates. Following the incubation protocol, a Cq value of 15.7 was determined for the cDNA samples, corresponding to 16 amplification cycles. The amplification mixture (10X Genomics, USA) was added to the cDNA sample tubes, and the qPCR amplification protocol was completed based on the obtained Cq value. The samples were incubated overnight before proceeding with sequencing at 4°C. Store overnight. Library construction and sequencing were performed at the Dresden Concept Genome Center (DcGC) using a HiSeq 2000 next-generation sequencer (Illumina, Inc., USA).

[0148] Sequencing data was processed using the Space Ranger (10X Genomics, USA) pipeline to recreate the spatial arrangement, which aligns H&E-stained bright-field images with spatially barcoded gene expression data based on reference points in the slice capture region boundaries. The pipeline performs alignment, tissue detection, reference detection, and barcode / UMI counting.

[0149] Data Analysis

[0150] All basic and advanced algorithms used in this work were developed and implemented using custom-written Python scripts. All package extensions and plugins used have been referenced as required.

[0151] SRT quality control and gene expression normalization

[0152] Prior to data analysis, the single-cell analysis toolkit Seurat (58) and its extension STutility were used to exclude technology batch effects and experimental variations (https: / / ludvigla.github.io / STUtility website / ). These packages statistically quantify the number of unique genes (nFeature RNA) and the number of UMIs (nCount RNA) across all samples and conditions. To further characterize and discover shared hippocampal structures across the two conditions, another extension, Harmony, recomputed the UMAP embeddings and clustering to return an integrated low-dimensional representation of the data (59). Since approximately 5500 median genes were found for each point in each dataset, SRT points with fewer than 1000 unique genes in each dataset were filtered out from the analysis. Next, to reduce the total number of genes used in the analysis, mitochondrial and ribosomal protein-coding genes were filtered out from the analysis. Finally, to account for technology batch variations and detect highly variable genes, the overall gene expression for each SRT point was normalized by the total count of each gene across all SRT points, so that each point had the same count after normalization.

[0153] This is implemented using the scanpy.pp.normalize_total python package and is available on GitHub (https: / / github.com / theislab / scanpy) (60).

[0154] Oscillation mode detection and waveform classification

[0155] Prior to data analysis, oscillation patterns of LFP were detected in each record using a hard thresholding algorithm (16). Furthermore, the detected events were further processed and filtered using a low-pass fourth-order Butterworth filter (1 Hz to 100 Hz).

[0156] Finally, quantile thresholding was employed in a custom-written Python script to remove spurious discharge electrodes or non-physiological detection events (16). To characterize the different features and shapes of the recorded LFP oscillation waveforms and assign them to specific interconnected HC layers, the inventors implemented principal component analysis (PCA) and K-means clustering algorithms as described previously (13).

[0157] structural clusters

[0158] To characterize local and global hippocampal subnetwork behavior, functionally discharged n-Ephys electrodes were structurally associated with specific HC regions by overlaying light microscopy images of the hippocampus onto a CMOS-MEA layout. The electrodes were then grouped into clusters—DG, Hilus, CA1, CA3, EC, and PC (16)—based on structural markers on the HC slices. To characterize the transcriptional profiles within these six major regions, H&E-stained bright-field microscopy images of the SRT points were overlaid onto a layout using Loupe Browser (10X Genomics, USA), with the SRT points structurally associated with specific HC regions.

[0159] Multi-scale spatial alignment

[0160] To infer the correspondence between n-Ephys electrode-SRT point interfaces and their corresponding network-range functional electrical activity and transcriptome feature readouts using spatial localization, MEA-seqX implements a multi-scale spatial alignment process. To provide transcriptomic and electrophysiological profiles of the same cell set with a spatial background, automatic slice alignment is performed using image resizing and rotation. This alignment is based on optical imaging, the physical size of the n-Ephys electrode-SRT point interface, and associated hippocampal cortical structural inputs to place multi-scale data in the same dimension. First, MEA-seqX implements an automatic scaling algorithm based on the n-Ephys electrode-SRT point size to resize the light H&E-stained bright-field microscopy slice image from the SRT to the corresponding light microscopy hippocampal image from the n-Ephys. Importantly, n-Ephys electrode-SRT point matching is not one-to-one due to differences in technical resolution; instead, it is based on fractional matching of associated hippocampal cortical structural inputs. Therefore, each SRT point has an associated n-Ephys electrode with average electrophysiological features from the associated electrode. Next, slice spatial alignment and rotation are computed between the two slice images using the following procedure: i) Hippocampal cortical structure reference points {i, j} and {k, l} are assigned for each scale SRT and n-Ephys, respectively, where {i, k} is the midpoint in the top of the DG, and {j, l} is on the edge of the upper lobe of the DG; ii) SRT reference point i is aligned with n-Ephys reference point k to place the two scales in one dimension; iii) After the reference points {i, k} are aligned, the final alignment of reference points {j, l} is based on... and The difference between them. iv) Assuming the coordinates of the two arrays are known, use the distances between x, x', y, and y' to calculate... and The angle. v) To determine x is defined using the horizontal intersection between aligned reference points {i, k}, while y is defined using the vertical intersection between reference point j and the horizontal intersection. vi) To determine x' is defined using the horizontal intersection between the aligned reference points {i, k}, while y' is defined using the vertical intersection between the reference point l and the horizontal intersection. vii) The final rotation angle is defined as When applied, the reference point {i, k} is aligned as the final multi-scale reference point m, and the reference point {j, l} is aligned as the final multi-scale reference point n.

[0161] Average activity n-Ephys characteristics

[0162] To determine the correlation between spatial gene expression patterns and functional n-Ephys features, Spearman correlation was used to associate filtered genes with one of the network features (61), and the genes were ranked according to significance using p-values ​​adjusted for Benjamin-Hochberg false discovery rate (62). Functional network features of large-scale spatiotemporal LFP oscillations included LFP rate, amplitude, energy, delay, and positive and negative peak counts (13, 16).

[0163] Target gene list

[0164] A specific list of genes was developed based on the functional gene ontology. Genes related to immediate early gene families, signaling pathways, hippocampal function, and neurogenesis were compiled into six lists (24, 25).

[0165] Nonnegative matrix factorization (NMF)

[0166] The inventors implemented an unsupervised machine learning algorithm using sparse-constrained nonnegative matrix factorization to identify various spatiotemporal patterns emerging from SRT and n-Ephys networks (26). The NMF factorization was described and adapted according to the Scikit-learn 1.2.2 python package (sklearn.decomposition.NMF) (63). First, the input V matrix contains combined information from SRT (i.e., gene expression values ​​from IEG gene families) and n-Ephys (i.e., LFP rate or degree network features), where each data entry includes the expression value of each gene associated with a network feature value (n) having a spatial location (m).

[0167]

[0168] Where n is the spatial location point, m is the gene expression related to network features, and p is the number of factors.

[0169] The resulting decomposition basis matrix W contains spatial gene expression patterns and their locations associated with functional features, and the coefficient matrix H represents the contribution of these basis vectors to each spatial location and n-Ephys feature. To optimize the distance between V and the product matrices H and W, the inventors implemented the widely used distance optimization function, the squared Frobenius norm (F), which adds sparsity constraints (63) to the factors.

[0170]

[0171] in, and Non-negative values ​​( , ). and These are the corresponding regularization parameters for H and W.

[0172] Mutual Information

[0173] To represent the collective of points in a multilayer network based on gene expression, the gene expression distribution is calculated for each list of target genes. Next, mutual information distance scores are calculated for the gene information from each list of target genes at each point, compared between points, and sorted by cluster (27). The inventors use a modified function from the Scikit-learn 1.2.2 python package (sklearn.metrics.normalized_mutual info_score) to calculate the mutual information (63).

[0174] Functional Connections

[0175] As previously described, in order to infer large-scale statistically dependent connections of functional discharge activity on multilayer hippocampal networks, cross-covariance was calculated between discharge n-Ephys electrode pairs using Pearson correlation coefficient (PCC), subsequent directional transfer function (DTF), and multivariate Granger causality (13, 16).

[0176] Graphical visualization

[0177] To visualize large-scale network connectivity in both the SRT and n-Ephys datasets, mutual information distance scores and functional connectivity data architectures were constructed, respectively, to include nodes and edges as previously described (16). The data were converted to (.gexf) file format and directly read and visualized in Gephi program version 9.2 (https: / / gephi.org). To investigate the functional interactions of selected genes on the spatial network array, the inventors filtered the mutual information scores between all paired points to include the mean and two standard deviations of the mutual information scores. Thus, a threshold was set to include the mutual information scores for each list of target genes. The genes. To examine the functional connectivity of the n-Ephys electrode, the top 2% of total functional links were included. SRT connectivity maps and n-Ephys connectivity. Figure 2 The graph is drawn using similar edge weights and degree range queries.

[0178] Network topology metrics

[0179] Graph theory was used to characterize the overall network topology and interconnectivity based on functional connectivity from LFP events detected by n-Ephys or mutual information scores from SRT gene expression. The inventors computed topological metrics in custom-written Python code, as previously reported (13, 16). Briefly, the network connectivity topology metric describes the network by treating nodes n as central components of a graph, where nodes n may or may not be connected to each other. In the inventors' case, nodes n correspond to specific n-Ephys electrodes or SRT capture points in the sensing array, where edges e are functional links or connections between each node n. To represent the overall network topology and characteristics, the inventors chose the following graph theory topology parameters:

[0180] degree

[0181] In order to characterize different representations of network connectivity, the inventors characterize the degree k of node n to describe the number of edges connected to the node, as previously described (16).

[0182]

[0183] in, This indicates the degree of node i. This represents the connection between nodes i and j. N is the set of all computing nodes in the network.

[0184] Hub nodes and rich club nodes

[0185] To identify key nodes in a network and reveal its topology, hub nodes and rich club nodes are analyzed. Hub nodes are detected based on three node metrics—node strength, clustering coefficient, and network efficiency. The metrics for each node are calculated and compared to determine if a node's value is in the top 20% of all nodes (16). To constrain the definition of a hub node, the inventors set a limit using a hub score. The inventors' hub scores are valued between 0 and 3, where a node satisfies the top 20% in none, 1, 2, or all three node metrics. Within the hub node group is a subgroup of nodes with dense connections, which are assigned rich club nodes and described as hub nodes with a higher degree than average, and are determined by the rich club coefficient. Provide (16).

[0186]

[0187] Where k represents degree, This represents the number of nodes whose degree is greater than a given value k, and... Indicates including The number of connections in the subnet.

[0188] Network topology representation

[0189] To determine the potential impact of hub nodes on network functionality and the organizational processes that form network topology, the inventors characterized the degree distribution P(k) of nodes detected in the n-Ephys dataset and the SRT dataset, which yielded a decaying distribution with a power-law tail (64).

[0190]

[0191] To estimate the low-level power distribution P(k) to describe the scale-free topology with small-world properties, the inventors used a log-normal model for fitting.

[0192]

[0193] in, and These represent the mean and standard deviation of the distribution, respectively. To visualize the best-fit network representation, the complementary cumulative distribution function (cCDF) is used instead of the probability density of node degree, and the cCDF is plotted on a logarithmic axis for more robust visualization of high-k regions. Goodness-of-fit tests are performed between the actual data and the fitted model, and the coefficients R0 are determined. 2 To estimate. Finally, the power-law distribution is discretized using Pareto linear merging (scipy.stats.pareto) (65).

[0194] diffusion pseudotime

[0195] To precisely locate dynamic transcriptional changes from static, spatially resolved sequencing data and to determine the impact of intrinsic and extrinsic influences on different dynamic processes under study, the inventors employed diffusion pseudo-time (DPT) (35). DPT reveals the underlying dynamics of biological processes and, in this case, the temporal trajectories of specific gene expression from spatially resolved hippocampal transcriptomes. While conventional diffusion maps effectively denoise data while preserving local and global structure, the resulting maps often encode information in higher dimensions, limiting visualization. To overcome this prior to DPT analysis, the inventors implemented a thermal diffusion potential based on affinity transition embedding (PHATE) on spatial transcriptome data, presenting information in lower dimensions (https: / / github.com / KrishnaswamyLab / PHATE) (66). PHATE encodes both local and global data in a manifold structure. Local data relationship similarities are encoded by applying a kernel function to Euclidean distance. Global data relationships are encoded via latent distance, where local similarities are transformed into probabilities. These diffusion probabilities are determined by transforming local information into probabilities of transitioning from one data point to another in a single step of a random walk. This can power the t-step to give the t-step probability of both local and global distances. In the inventors’ dataset, each spatial point has a defined relationship with each nearest or farthest point in the weighted graph (66). Then, the DPT analysis sorts the transcriptome points according to the probability of differentiation toward different points (35).

[0196] Cell type deconvolution

[0197] To determine cell type colocalization and examine differences between two hippocampal transcriptomes, conditional autoregressive deconvolution (CARD) was performed using a single-cell sequencing reference (39, 40). CARD-based analysis is available on GitHub (https: / / github.com / YinglVla0107 / CARD) and is applicable to spatial transcriptome data in Python. Within the references, hippocampal cell types were broadly classified into groups, including astrocytes, endothelial cells, ependymal cells, macrophages, microglia, neurogenic cells, neurons, oligodendrocytes, and multidendrocytes, as well as appendage groups. Low-count cell type filtering reduced the hippocampal cell types from 85 to 76.

[0198] Prediction using the XGBoost algorithm

[0199] To determine whether specific gene expression values ​​can predict relevant HC network feature parameters such as LFP rate, magnitude, and duration for each SRT point, the inventors implemented the Gradient Boosting (XGBoost) algorithm, which integrates multiple tree models and has strong interpretability (47). XGBoost is described in the Scikit-learn 1.2.2 Python package (sklearn.ensemble.GgradientBoostingClassier), where the training and test datasets are implemented as described in the package (sklearn.model_selection.train_test_split) (63). These datasets contain spatially resolved gene expression values ​​based on a specific gene list and relevant network features from the functional n-Ephys dataset. To balance the input information between the SD dataset and the ENR dataset for better comparison, half of the ENR dataset was randomly generated to be used as input to the training and test datasets. This normalization was chosen as previously described (16) because ENR has a two-fold difference in network feature parameters. The input datasets were equally divided into an initial training dataset and a test dataset, with real data points trained through 100 iterations. The final predicted data points and actual data points were used to determine the predictability, accuracy, and significance of network feature predictions from transcriptome data (67). Packages for statistical implementation included the Scikit-learn 1.2.2 Python package (63) for calculating prediction accuracy (sklearn.metrics.explained_variance_score) and Scipy 1.10.1 (65) for calculating Pearson correlation coefficients (scipy.stats.pearsonr) and (scipy.stats.ttest_ind). Accuracy results were compared on multiple final data outputs, and values ​​within the mean and three standard deviations were determined to be within the chance threshold.

[0200] Statistical analysis

[0201] All statistical analyses were performed using Originlab 2020 or as described in the package appendix. Unless otherwise expressed as standard deviation, data in this work are expressed as mean ± standard mean error (SEM). Box plots were determined by the 25th to 75th percentiles, and whisker plots by the 5th to 95th percentiles, with lengths within the quartile range (1.5|QR). Additionally, lines depict the median and square of the means. Differences between groups were examined for statistical significance using the Kolmogorov-Smirnov test, one-way ANOVA, or two-way ANOVA followed by Tukey's post-hoc test, where appropriate. P < 0.05 was considered significant.

[0202] result

[0203] Interface Technology and Information Integration: From Transcriptome to Functional Networks

[0204] MEA-seqX integrates brain-on-a-chip technology via network electrophysiology (n-Ephys) of a high-density CMOS microelectrode array (CMOS-MEA) (13–16), bioimaging via optical microscopy, and spatial sequencing via spatial resolution transcriptomics (SRT) (Figure 1, a–d). Specifically, from 300… High-resolution spatiotemporal recordings of extracellular firing patterns obtained from acute HC sections of the mouse hippocampus cortex were docked with 4,096 on-chip sensors. Simultaneously, precise anatomical localization was achieved using optical imaging, and high-resolution transcriptional profiling data were obtained from the same cell ensemble within the circuit (Figure 1, e).

[0205] When visualized via the Uniform Manifold Approximation and Projection (UMAP) method, the inventors’ quality control of the transcriptome dataset revealed similar nFeature and nCount RNA statistics and organization (22) Figure 7 (a to c). Network-wide activity in the HC was evaluated by principal component analysis (PCA) and K-means clustering algorithms, which provided different features of the oscillating waveforms and their shapes in each interconnected HC layer (13). Figure 7 d). Multidimensional readings are processed through a Python-based computational pipeline to quantify the molecular dynamics of the mapped loops with high spatiotemporal resolution.

[0206] A multi-scale spatial rescaling and alignment process was developed to establish a direct correspondence between the n-Ephys electrode-SRT point interface and its corresponding network-range functional electrical activity and transcriptome signature readouts and their localization in tissues.

[0207] This is achieved through an automatic scaling algorithm, employing image resizing and rotation based on optical imaging, the physical size of the n-Ephys electrode-SRT point interface, and two anatomical landmarks in the dentate gyrus. Figure 1 (f; see Methods for details) to achieve this. The resulting superposition allows for the identification of network features from transcriptomic reads to be aligned with neural activity reads (f; see Methods for details). Figure 1 (g). The method according to the invention generates an accurate topology map of multimodal data connections to represent both the local and global relationships between the SRT points and the underlying discharge electrodes. Figure 1 h). From a large amount of high-dimensional data, the transcriptional pseudo-temporal dynamics of the underlying discharge information flow are derived ( Figure 1 i). By integrating deconvolution methods, the inventors inferred cell type resolution from spatial transcriptome data and correlated neuronal heterogeneity with its firing characteristics. Figure 1 Finally, the method according to the invention provides an automated machine learning algorithm for predicting network electrophysiological features with high accuracy from spatial transcription profiles. Figure 1 (k).

[0208] Linking spatial transcriptomics with network-wide neurodynamics

[0209] To apply the pipeline to specific, previously unsolved research problems, the inventors used the MEA-seqX framework to reveal the effects of experience-dependent plasticity (16) on the coordinated activity of neuronal assemblages and its interaction with orchestrated transcriptional activity. HC slices from mice raised in standard (SD) and enriched environment (ENR) conditions were prepared for recording oscillation patterns of local field potentials (LFP), optical imaging, and SRT sequencing. To assess how spatial patterns of gene expression (SRT) correspond to functional network electrophysiological features (n-Ephys) of the same brain tissue, the inventors measured Spearman correlations (23) to quantify transcriptional similarity of gene expression profiles from SRT points and examined their relationship with functional network features (i.e., amplitude, delay, energy, LFP rate, negative peak count, and positive peak count). The inventors found a significant enhancement in gene expression patterns corresponding to functionally coupled dentate gyrus (DG) and CA3 regions in the ENR compared to SD. Figure 2 a). Specifically, compared to SD, the LFP rates in the DG, CA3, and CA1 subdomains of the ENR were shown to be 2.2-fold, 2.5-fold, and 0.6-fold increases, respectively, in the associated transcriptomes. Figure 2 (b).

[0210] The inventors then sought to identify which specific genes would drive stronger causal links between relevant molecular and functional networks. Based on gene ontology clusters, the inventors categorized relevant genes into six target gene families, including the immediate early gene "IEG," hippocampal neurogenesis, hippocampal signaling pathways, receptors and channels, synaptic plasticity, synaptic vesicles, and adhesion (24, 25). Examination of transcripts from these families revealed enhanced expression of IEG, ion channel activity, synaptic function, and neurogenesis in the ENR compared to SD. Compared to SD, genes essential for hippocampal activity and function (such as Bdnf, Egrl, Homer 1, Npas4, Gria2, and Campk2a) showed higher expression levels in the hippocampal transcriptome of the ENR. Figure 8 ).

[0211] Next, to identify different spatiotemporal patterns across combined SRT and n-Ephys modalities, the inventors implemented an unsupervised machine learning algorithm using sparse-constrained nonnegative matrix factorization (NMF) (26). This approach allows modality decomposition into differentially expressed subnetworks of genes, spatial locations, and electrophysiological features to provide dimensionality reduction and interpretability. Figure 2 c). Because IEG has shown a significant contribution to linking spatial transcriptional patterns with LFP activity ( Figure 8 Therefore, the number of factors in the NMF decomposition is determined based on the size of the IEG list (i.e., 12 factors). The input expression V matrix (which contains combined information from SRT and n-Ephys (i.e., LFP rate) data, where each entry represents the expression level of a gene associated with a network functional characteristic at a specific spatial location) Figure 2 The d) is decomposed into two nonnegative matrices. The fundamental matrix W contains spatial gene expression patterns and their locations associated with functional characteristics. Figure 2 e), and the coefficient matrix H represents the contribution of these fundamental vectors to each spatial location and n-Ephys feature ( Figure 2 (f). This analysis revealed increased normalized gene expression and network eigenvalues ​​in the ENR network, as well as higher spatial resolution components driving gene expression intensity. Furthermore, the inventors identified spatially specific subnetworks in both the SD and ENR networks to highlight genes such as Bdnf, Egrl, Fosb, and Npas4.

[0212] MEA-seqX demonstrates the computational role of empirically dependent dynamics in the coordinated interaction of neuronal ensemble activity and its corresponding transcriptional patterns.

[0213] Spatial analysis of the coordinated topological network organization of transcriptome and activity patterns

[0214] The inventors employed quantitative measurements to comprehensively examine the interconnections between HC subnetworks derived from SRT data and neural n-Ephys recordings under SD and ENR conditions. The inventors calculated “mutual information” (27), i.e., a measure of the interdependence between two variables, to assess the degree of interdependence in gene expression within a specific gene family. Figure 9 a) and Pearson correlation coefficient (PCC) (13, 16) to measure the cross covariance among discharge electrode pairs ( Figure 9 b). These calculations allow the inventors to build connectivity matrices at multiple scales. Next, the inventors calculate mutual information distance scores for each target gene family at each point to measure the differences in interactions within different points. These scores are then compared between points and organized into clusters (27). Figure 10 (a to f). Meanwhile, the inventors quantified the differences in the correlation matrix by analyzing the PCC of concurrent LFP activities across interconnected HC layers (13, 16). Figure 10 (g). After analyzing various gene families, it became apparent that the ENR transcriptome exhibited higher mutual information compared to SD. This indicates a more robust statistical relationship in the coordination of activities and communication within gene expression patterns within the HC subnetwork. Figure 10 (a to f). Importantly, as according to ( Figure 10 The LFP cross-correlation diagram shown in g) 16 clearly shows that this finding is highly consistent with the significant enhancement of both the local and global strengths of spatiotemporal interactions in the ENR and SD networks (16).

[0215] Subsequently, in order to describe and quantify topological organization, information flow, and communication properties within multimodal transcriptional functional readouts, MEA-seqX employs graph theory methods to evaluate multi-scale network metrics. Figure 3 (a). This involves constructing detailed wiring diagrams that depict both local and global interconnections between transcriptomics and patterns of neural activity. Transcriptomic maps are formed using mutual information scores, while functional maps are generated from connectivity patterns within sets of co-firing neurons captured by LFP, such as under SD and ENR conditions. The connectivity maps derived between transcriptomics and neural function reveal the spatial arrangement of interconnected subnetworks derived from spatial IEG and spatiotemporal functional neural activity. Notably, these maps demonstrate similar spatial distributions and connectivity patterns at different scales and patterns. Figure 3 (bc). Comparison of SD and HC networks.

[0216] ENR further highlights the enhanced level of cross-scale causal coordination. Compared to SD, transcriptional functional connectons are enhanced in ENR.

[0217] The inventors also analyzed the constructed graphs to identify highly interconnected nodes (referred to as “hub complexes”) and densely connected hubs (referred to as “rich club organizations”) within the transcriptional functional connectome (28). The ENR subnetworks derived from both SRT and n-Ephys data and their SD counterparts ( Figure 3 Compared to d, e; gray, it exhibits a larger interconnected hub complex and rich club organization ( Figure 3 (d, e; dark gray). This indicates that rich experience leads to enhanced specialization of coordination and interactions, increased resilience, and increased global communication across transcriptional functional scales.29 These results reveal new insights into the dynamic interactions and mutual influences between molecular hub complexes and functional hub complexes, which significantly contribute to the overall coordination of multi-scale topological network organization.

[0218] Many biological networks are characterized by small-world topologies (30), which are defined by a scale-free architecture comprising highly connected hub nodes and degree distributions that decay with a power-law tail (31). By analyzing the cumulative degree distributions of interconnect links in transcriptomic and functional connectosomes from SD and ENR networks, the inventors found that these multiscale distributions do indeed follow a power-law function, as previously hypothesized in transcriptomic networks of adult hippocampal neurogenesis (32) and network-wide activity (16) in ENR. Both the SRT and n-Ephys distributions in the ENR network showed heavier tails compared to the SD, indicating a more significant number of densely connected hubs ( Figure 3 (f and illustration). This finding is supported by sparse-constrained NMF analysis of multi-scale degree distributions (i.e., similar to f and illustration). Figure 2 The analysis was implemented in c to f. The inventors quantified the multi-scale decomposition set, spatial location, and network connectivity features of variable expression subnetworks of genes to identify changes in network topology metrics associated with the expression and network connectivity of spatially resolved IEGs. Figure 11 ).

[0219] By revealing empirically induced hippocampal connectomics across scales and its complex multilayered dynamics in a single experiment, the inventors' results demonstrate the ability of MEA-seqX to integrate and capture coordinated transcriptomics and functional data. This integration allows for new insights into neural communication, resilience levels, hierarchical organization, and specialization across multiple scales that could previously only be studied independently and often based on limited data (29, 33, 34).

[0220] Multiscale dynamics for evaluating multimodal information

[0221] To address the challenge of revealing the synchronization dynamics across scales and modalities, the inventors combined two cutting-edge computational methods—the diffusion pseudo-time (DPT) of SRT (35) and the activity center trajectory (CAT) of n-Ephys (13, 16). This integrated approach aims to reveal the temporal progression of gene expression and network-wide neural activity in the hippocampal circuit.

[0222] The inventors applied DPT to static snapshot SRT data to achieve pseudo-temporal ranking by assembling points based on expression similarity. This allows the inventors to construct a network representation of the SRT evolution trajectory. The difference probability is calculated using Euclidean distance from vector-based random distances in the diffusion graph space, which helps identify low-dimensional variations from high-dimensional observations. Here, the inventors focus on IEG expression in SD and ENR to reveal significant regional differences in DPT based on IEG expression. Figure 4 (a, b). Simultaneously, the inventors quantified the spatiotemporal propagation pathways within the hippocampus by constructing a CAT of all n-Ephys loop-range oscillations, and thus calculated the spatiotemporal displacement rates of those discharge modes (a, b). Figure 4 (c, d). As previously reported (16), analysis of CAT durations showed that firing events in the ENR propagated faster (i.e., shorter durations) compared to SD. Notably, comparisons of the SD and ENR transcriptomes indicated faster DPT in all four hippocampal regions of the ENR transcriptome compared to SD, and thus reflected the faster spatiotemporal propagation patterns observed in the ENR from n-Ephys CATsl6 ( Figure 4 (c to e). By integrating the spatial map of DPT with the temporal progression of neural activity trajectories, modulated by the influence of rich experience, the inventors were allowed to instantiate a multi-scale perspective on how molecular and electrical processes unfold and interact simultaneously within biological systems. In-depth studies of experience-dependent activity dynamics make the comparison between SD and ENR particularly illuminating. Such comparisons may reveal causal relationships between dynamic changes at the molecular level and those at the functional scale. Different experiential environments, represented by SD and ENR, induce different responses at the transcriptomic level, which are inextricably linked to neural activity within the same cellular set. These distinct activity patterns link transcriptomics to function, and function to transcriptomics (36, 37).

[0223] Spatiotemporal cell type identification

[0224] Next, to understand the transcriptional diversity of neuronal cell types and their role in the firing patterns of the hippocampal circuit (38), the inventors employed a conditional autoregressive deconvolution (CARD) method using single-cell sequencing references (39, 40). This allowed the inventors to determine cell types and local tissue composition based on deconvolutioned gene expression patterns to construct a multi-scale spatial map of neuronal heterogeneity and its firing characteristics within the same hippocampal cortex tissue. Initial application of CARD yielded a broad group classification of hippocampal cell types. This diverse group included astrocytes, endothelial cells, ependymal cells, macrophages, microglia, neurogenic cells, neurons, oligodendrocytes, and NG2 cells. Prior to any filtering, the inventors identified 85 distinct cell types, highlighting the experimental effectiveness of their slicing technique. After removing low-count cell types, the inventors retained a robust group of 76 cell types.

[0225] Notably, when these cell types were exposed to two different transcriptomic inputs from SD and ENR, the inventors observed a consistent distribution of prominent cell types across both transcriptomics. Figure 5 a). This result highlights the robustness of the inventor's method and the reproducibility of the inventor's discovery.

[0226] Furthermore, integrating the CARD method into MEA-seqX has proven helpful in achieving spatial resolution of cell type composition and linking it to large-scale oscillatory electrophysiological characteristics. Figure 5 b). The inventors identified specific high-proportion cell types within the DG and CA3 regions based on unique marker genes for these cell types. In the DG, granular cells (GCs) are characterized by Cck and Penk expression, while CA3 pyramidal cells exhibit Nosl and Inhba markers. Based on previous research (41), the inventors observed both common Cck expression and less frequent Penk expression in the molecular layers of the DG and upper pyramidal lobes of the GC. Figure 5 c). Notably, compared to SD, the expression levels of both marker genes in ENR were significantly higher ( Figure 5 e).

[0227] When these markers are overlaid on the DG functional network data, the ENR samples show enhanced discharge modes and signal amplitudes, particularly within the upper cone of the DG. Figure 5(c, e). Previous studies have linked Penk to the enrichment of DG imprinted cells and their involvement in hippocampal-associated behavior (41), while Cck is associated with the dynamic selection and control of cell assemblages in the DG (42). The inventors’ data are consistent with existing reports and highlight the significant increase in DG excitability and temporal dynamics in the ENR, suggesting potential pathways for exploring how rich experience can influence transcriptional functional interactions in the hippocampal circuit to contribute to understanding the cellular and molecular basis of memory (43). A similar analysis was conducted, focusing on the spatial distribution of pyramidal cell layers in CA3 ( Figure 5 The inventors found that the Nosl marker gene and the Inhba marker gene showed significantly higher expression in ENR compared to SD. Figure 5 (f). This transcriptome reading matched the increased LFP rate and unique waveform characteristics recognized within the CA3 region (f). Figure 5 (d, f). Nosl has been linked to key neural mechanisms, including long-term potentiation (LTP), synaptic plasticity, and modulated neural circuit dynamics (44, 45), while Inhba has been linked to neuroprotection and neuronal survival (46). These reports provide strong evidence for the inventors' discovery of enhanced transcriptional function within a clearly empirically dependent paradigm in the ENR group. This, in turn, paves the way for a deeper exploration of the specific roles of these marker genes in CA3 pyramidal neurons and their potential impact on understanding neural function and cross-scale dysregulation.

[0228] Prediction across scales and modes

[0229] To investigate whether expression profiles of a single spatially resolved gene can predict electrophysiological features across the HC network, the inventors employed the gradient boosting (XGBoost) algorithm, known for its strong interpretability through the integration of multiple tree models (47). While previous approaches have focused on gene properties related to electrophysiological and morphological diversity across cell types using low-resolution transcriptomics and electrophysiology (e.g., single-cell RNA sequencing and patch-clamp) (4, 5, 38), the potential experimental aim of this invention is to evaluate whether specific electrophysiological features can be predicted using spatially resolved transcriptomics data. The XGBoost model was trained using 70% of the spatial transcriptomics data points detected in the SRTs (i.e., 333 genes across the spatial background of HC tissues and six gene families) for each quantitative n-Ephys measurement. Three spatiotemporal n-Ephys metrics (LFP rate, amplitude, and time delay) were successfully predicted based on differential spatial gene expression. The relationship between cross-validation predictions and true values ​​was evaluated using the Pearson correlation coefficient (r) of the SRT-spot-n-Ephys electrodes (r). Figure 6 a and Figure 12By applying XGBoost to specific gene families that exhibit higher expression in ENR compared to the SD loop, the inventors observed significantly higher prediction accuracy on the ENR dataset. The XGBoost classifier achieved approximately 93% accuracy on the ENR dataset and approximately 70% accuracy on the SD dataset. Figure 6 (b). Such predicted interactions between individual genes or gene families at the transcriptomic and network levels support the idea that brain function is orchestrated via multiscale networks following fundamental organizational principles (18). These results highlight multiscale causal links between neural activity, plasticity, and the expression of genes in different spatial locations within specific gene families.

[0230] This aligns with predictions of network-level function modulated by prior experience (16). This approach can reveal key genes and cellular pathways that shape neuronal responses and overall brain function, while also elucidating regulatory mechanisms controlling neurodynamics, plasticity, and disease pathogenesis. The inventors' integrative approach significantly reduces experimental complexity, thus improving the interpretability of results and providing guidance for future research.

[0231] References cited

[0232] 1. Frackowiak, R. and Markram, H. The future of human cerebral cartography: A novel approach. Proceedings of the Royal Society B: Biological Sciences 370, (2015).

[0233] 2. Parent!, I., Rabaneda, LG, Schoen, H., and Novarino, G. Neurodevelopmental Disorders: From Genetics to Functional Pathways. Trends in Neuroscience, 43, 608–621 (2020).

[0234] 3. Palop, JJ, and Mucke, L. Network abnormalities and interneurondysfunction in Alzheimer disease. Nature Review Neuroscience 17, 777-792 (2016).

[0235] 4. Cadwell, CR, et al. Electrophysiological, transcriptomic and morphologic profiling of single neurons using Patch-seq. Nature Biotechnology 34, 199-203 (2016).

[0236] 5. Fuzik, J. et al. Integration of electrophysiological recordings with single-cell RNA-seq data identifies neuronal subtypes. Nature Biotechnology 34, 175-183 (2016).

[0237] 6. Li, Q. et al. Multimodal charting of molecular and functional cell states via in situ electrosequencing. Cell 186, 2002–2017.e21 (2023).

[0238] 7. Stähl, PL et al. Visualization and analysis of gene expression intissue sections by spatial transcriptomics. Science (1979) 353, 78-82 (2016).

[0239] 8. Vickovic, S. et al. High-definition spatial transcriptomics for in situtissue profiling. Natural Methods 16, 987-990 (2020).

[0240] 9. Maniatis, S. et al. Spatiotemporal dynamics of molecular pathology in amytrophic lateral sclerosis. Science (1979) 364, 89-93 (2019).

[0241] 10. Asp, M. et al. A Spatiotemporal Organ-Wide Gene Expression and Cell Atlas of the Developing Human Heart. Cell 179, 1647 to 1660. el9 (2019).

[0242] 11. Urai, AE, Doiron, B., Leifer, AM, and Churchland, AK. Large-scale neural recordings call for new insights to link brain and behavior. Nature Neuroscience 25, (2021).

[0243] 12. Buzsäki, G. Large-scale recording of neuronal ensembles. Nature Neuroscience 7, 446-451 (2004).

[0244] 13. Hu, X., Khanzada, S., Klütsch, D., Calegari, F., and Amin, H. Implementation of biohybridol factory bulb on a high-density CMOS-chip to reveal large-scale spatiotemporal circuit information. Biosensors Bioelectronics 198, 113834 (2022).

[0245] 14. Amin, H., Nieus, T., Leonardoni, D., Maccione, A., and Berdondini, L. Sei. Rep., 2460. 7, (2017).

[0246] 15. Imfeld, K. et al. Large-scale, high-resolution data acquisition system for extracellular recording of electrophysiological activity. IEEE Transactions on Biomedical Engineering 55, 2064-73 (2008).

[0247] 16. Emery, BA, Hu, X., Khanzada, S., Kempermann, G., and Amin, H. High-resolution CMOS-based biosensor for assessing hippocampal circuit dynamics in experience-dependent plasticity. Biosensors Bioelectronics 115471 (2023) doi: 10.1016 / J.BIOS.2023.115471.

[0248] 17. Emery, B.A., et al. Large-scale Multimodal Neural Recordings on a High-density Neurochip: Olfactory Bulb and Hippocampal Networks, leee Embs 42 to 45 (2022) doi: 10.1109 / EMBC48229. 2022. 9871961.

[0249] 18.Luo, L., Callaway, EM & Svoboda, K. Genetic Dissection of NeuralCircuits: A Decade of Progress. Neuron 98, 256 to 281 (2018).

[0250] 19. Buzsäki, G. Large-scale recording of neuronal ensembles. Nature Neuroscience 7, 446-451 (2004).

[0251] 20. Panzeri, S., Macke, JH, Gross, J., and Kayser, C. Neural populationcoding: Combining insights from microscopic and mass signals. Trends Cogn Sei19, 162-172 (2015).

[0252] 21. Kempermann, G. Environmental enrichment, new neurons and the neurobiology of individuality. Nature Review Neuroscience 20, 235-245 (2019).

[0253] 22. Mclnnes, L., Healy, J., and Melville, J. UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction. (2018).

[0254] 23. Anderson, KM, et al. Gene expression links functional networks across cortex and striatum. Nature Communications 9, (2018).

[0255] 24. Zhang, J. and Jiao, J. Molecular Biomarkers for Embryonic and AdultNeural Stem Cell and Neurogenesis. Biomed Res Int 2015, 727542 (2015).

[0256] 25. Valor, L. M., Charlesworth, P., Humphreys, L, Anderson, CNG, and Grant, SGN Network activity-independent coordinated gene expression program for synapse assembly. www.pnas.orgcgidoil0.1073pnas. 0609071104 (2007).

[0257] 26. Brunet, JP, Tamayo, P., Golub, TR, and Mesirov, JP. Metagenes and molecular pattern discovery using matrix factorization. Proc Natl Acad Sei US 101, 4164-4169 (2004).

[0258] 27. Priness, I, Maimon, O. and Ben-Gal, I. Evaluation of gene-expression clustering via mutual information distance measure. BMC Bioinformatics 8, (2007).

[0259] 28. van den Heuvel, MP and Sporns, O. Rich-club organization of the human connectome. Journal of Neuroscience 31, 15775-15786 (2011).

[0260] 29. McAuley, JJ, Da Fontoura, Costa, L., and Caetano, T.S. Rich-club phenomenon across complex network hierarchies. Applied Physics Letters 91, 2–5 (2007).

[0261] 30. Watts, DJ, and Strogatz, SH. Collective dynamics of 'small-world' networks. Nature 393, 440-2 (1998).

[0262] 31. Barabäsi, A.-L. and Albert, R. Emergence of Scaling in Random Networks. Science (1979) 286, 509-512 (1999).

[0263] 32. Overall, RW, and Kempermann, G. The Small World of Adult Hippocampal Neurogenesis. Frontiers in Neuroscience 12, 1–12 (2018).

[0264] 33. Arnatkeviciute, A. et al. Genetic influences on hub connectivity of the human connectome. Nature Communications 12, 1–14 (2021).

[0265] 34. Zador, AM et al. Sequencing the Connectome. PLOS Biology 10, 1–7 (2012).

[0266] 35. Haghverdi, L., Büttner, M., Wolf, FA., Buettner, F., and Theis, FJ. Diffusion pseudotime robustly reconstructs lineage branching. Natural Methods 13, 845-848 (2016).

[0267] 36. Zhang, W. and Linden, DJ. The other side of the engram: experience-driven changes in neuronal intrinsic excitability. Nature Review Neuroscience 4, 885–900 (2003).

[0268] 37. Hübener, M. and Bonhoeffer, T. Searching for Engrams. Neuron 67, 363-371 (2010).

[0269] 38. Gouwens, NW et al. Classification of electrophysiological and morphological neuron types in the mouse visual cortex. Nature Neuroscience 22, 1182-1195 (2019).

[0270] 39. Ma, Y. and Zhou, X. Spatially informed cell-type deconvolution for spatial transcriptomics. Nature Biotechnology (2022) doi: 10.1038 / s41587-022-01273-7.

[0271] 40. Cable, DM, et al. Robust decomposition of cell type mixtures inspatial transcriptomics. Nature Biotechnology 40, 517-526 (2022).

[0272] 41. Erwin, SR et al. A Sparse, Spatially Biased Subtype of Mature Granule Cell Dominates Recruitment in Hippocampal-Associated Behaviors. Cell Reports 31, (2020).

[0273] 42. Klausberger, T. and Somogyi, P. Neuronal diversity and temporal dynamics: The unity of hippocampal circuit operations. Science (1979) 321, 53-57 (2008).

[0274] 43. Pignatelli, M. et al. Engram Cell Excitability State Determines theEfficacy of Memory Retrieval. Neuron 101, 274 to 284.e5 (2019).

[0275] 44. Steinert, JR, Chernova, T., and Forsythe, ID. Nitric oxide signaling in brain function, dysfunction, and dementia. Neuroscientists 16, 435–452 (2010).

[0276] 45. Bon, CLM, and Garthwaite, J. On the role of nitric oxide in hippocampal long-term potentiation. Journal of Neuroscience 23, 1941–1948 (2003).

[0277] 46. ​​Oberländer, K. et al. Dysregulation of Npas4 and Inhba expression and an altered excitation-inhibition balance are associated with cognitive deficits in DBA / 2 mice. Learning and Memory 29, 55-70 (2022).

[0278] 47. Li, W., Yin, Y., Quan, X., and Zhang, H. Gene Expression Value Prediction Based on XGBoost Algorithm. Frontiers in Genetics 10, 1–7 (2019).

[0279] 48. Asp, M., Bergensträhle, J., and Lundeberg, J. Spatially Resolved Transcriptomes-Next Generation Tools for Tissue Exploration. BioEssays 42, 1 to 16 (2020).

[0280] 49.Moore, H., Lega, BC, and Konopka, G. Riding brain “waves” to identify human memory genes. Curr Opin Cell Biol 78, 102118 (2022).

[0281] 50. He, BJ. Scale-free brain activity: Past, present, and future. Trends in Cognitive Science 18, 480-487 (2014).

[0282] 51. Sejnowski, TJ, Churchland, PS, and Movshon, JA. Putting big data to good use in neuroscience. Nature Neuroscience 17, 1440-1441 (2014).

[0283] 52. Siegel, M., Donner, TH, and Engel, AK. Spectral fingerprints of large-scale neuronal interactions. Nature Review Neuroscience 13, 121–134 (2012).

[0284] 53.Dean Ho, Stephen R. Quake, Edward RB McCabe, Wee Joo Chng, Edward K. Chow, Xianting Ding, Bruce D. Gelb, Geoffrey S. Ginsburg, Jason Hassenstab, Chih-Ming Ho, William C. Mobley, Garry P Nolan, Steven T. Rosen, AZ Enablingtechnologies for personalized and precision medicine. Trends in Biotechnology 38, 497-518 (2019).

[0285] 54. Berdondini, L. et al. Active pixel sensor array for high spatio-temporal resolution electrophysiological recordings from single cell to large-scale neuronal networks. Lab on a Chip 9, 2644-51 (2009).

[0286] 55. Heer, F. et al. Single-chip microelectronic system to interface with living cells. Biosensors Bioelectronics 2, 2546-53 (2007).

[0287] 56. Jun, JJ, et al. Fully integrated silicon probes for high-density recording of neural activity. Nature 551, 232-236 (2017).

[0288] 57. Angotzi, GN, et al. SiNAPS: An implantable active pixel sensor CMOS-probe for simultaneous large-scale neural recordings. Biosensors Bioelectronics 126, 355-364 (2019).

[0289] 58. Hao, Y. et al. Integrated analysis of multimodal single-cell data. Cell 184,3573 to 3587.e29 (2021).

[0290] 59. Korsunsky, I. et al. Fast, sensitive and accurate integration of single-cell data with Harmony. Natural Methods 16, 1289–1296 (2019).

[0291] 60. Wolf, FA, Angerer, P., and Theis, FJ. SCANPY: large-scale single-cell gene expression data analysis. Genome Biology 19, 15 (2018).

[0292] 61. Anderson, KM, et al. Gene expression links functional networks across cortex and striatum. Nature Communications 9, 1428 (2018).

[0293] 62. Benjamini, Y. and Hochberg, Y. Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society: Series B (Methodology) 57, 289-300 (1995).

[0294] 63. Pedregosa, F. et al. Scikit-learn: Machine Learning in Python. Journal of Machine Learning Research, Vol. 12. http: / / scikit-learn.sourceforge.net. (2011).

[0295] 64. Barabäsi, A.-L. and Albert, R. Emergence of Scaling in RandomNetworks Downloaded from. Mat. Res. Soc. Symp. Proc Volume 74 www. sciencemag.org http: / / science. sciencemag.org / (1995).

[0296] 65. Virtanen, P. et al. SciPy 1.0: fundamental algorithms for scientific computing in Python. Natural Methods 17, 261-272 (2020).

[0297] 66. Moon, KR, et al. Visualizing structure and transitions in high-dimensional biological data. Nature Biotechnology 37, 1482–1492 (2019).

[0298] 67.Chicco, D., Warrens, MJ, and Jurman, G. The coefficient of determination R-squared is more informative than SMAPE, MAE, MAPE, MSE and RMSE in regression analysis evaluation. PeerJ ComputSciT, 1 to 24 (2021).

Claims

1. An in vitro method for detecting the spatiotemporal electrophysiological dynamics and transcriptional profiles of an ensemble of functional neurons, the in vitro method comprising the following steps: i) Provide the set of functional neurons to be characterized. ii) Using a suitable electrode array, record the patterns of local field potentials (LFPs) of the neurons in the ensemble with high spatial and temporal resolution, and iii) Using optical imaging, obtain the spatial positioning patterns of the LFPs for substantially the entire set. Alternatively, cryopreservation is performed on essentially the entire collection. iv) Perform optical imaging and spatial resolution transcriptome analysis of the cells in the set using a second array, and v) Processing the obtained data, including aligning the spatial localization pattern of the LFP with the spatially resolved transcriptome analysis, and This allows for the detection of spatiotemporal electrophysiological dynamics and spatial transcriptional spectra of cells within the aforementioned functional neuronal cell set. Preferably, the mode for recording the local field potential (LFP) includes using a high-density CMOS-based biosensing MEA chip (CMOS-MEA).

2. The method according to claim 1, wherein, The functional neuronal cell set to be characterized is selected from neuronal tissue, brain slices, neurocancer samples, cultures of interconnected brain cells, cultured neuronal differentiated iPSCs, spinal cord, bioelectric tissues such as olfactory bulbs and cardiac tissue, and parts or combinations thereof, particularly derived from neuronal tissue of mammals with behavioral conditions or disorders, mammals with neurological conditions or disorders such as neurodegenerative disorders or cancer, and / or mammals being treated for one of these conditions or disorders.

3. The method according to claim 1 or 2, wherein, The optical imaging includes, for example, using bright-field imaging, including imaging the entire network of cell assemblies using a microscope such as a modular stereomicroscope.

4. The method according to any one of claims 1 to 3, wherein, Simultaneously, recording of the LFP and optical imaging are performed.

5. The method according to any one of claims 1 to 4, wherein, Cryopreservation includes rapid freezing, preferably freezing on dry ice, and optionally also includes freezing with... storage.

6. The method according to any one of claims 1 to 5, wherein, The cell collection was appropriately stained prior to the transcriptome analysis.

7. The method according to any one of claims 1 to 6, wherein, The transcriptomics includes generating barcoded cDNA and performing sequencing on the second array, such as quantitative sequencing, such as spatial resolution transcriptomics (SRT) sequencing.

8. The method according to any one of claims 1 to 7, wherein, The evaluation of the recorded LFPs and the assignment of the recorded LFPs to specific characteristics and shapes of waveforms from interconnect network layers include grouping electrodes and using principal component analysis (PCA) and K-means clustering algorithms.

9. The method according to any one of claims 1 to 8, wherein, Processing the data and aligning the spatial positioning pattern of the LFP with the spatial resolution sequencing and the transcriptome analysis includes overlaying the optical image with the electrode array layout and overlaying the optical image with a second array layout, and the alignment of the two overlays includes image resizing and rotation.

10. A method for identifying the composition of a functional neuronal ensemble, comprising performing the method according to any one of claims 1 to 9, and further comprising the step of inferring the composition of the functional neuronal ensemble based on the spatiotemporal electrophysiological dynamics of detected cells and the transcriptional profile, wherein, Preferably, the identification includes mapping cells in the set, identifying clusters of cells in the set such as functional clusters, identifying different nerve cell types in the set, and / or identifying spectral characteristics of cells in the set.

11. A method for monitoring the spatiotemporal electrophysiological dynamics and transcriptional profile of a functional neuronal cell ensemble, comprising repeatedly performing the method according to any one of claims 1 to 9 at different time points, and monitoring the spatial electrophysiological and transcriptional information of the molecular neuronal cell ensemble based on comparing the spatiotemporal electrophysiological dynamics and the transcriptional profile of the cells detected at different time points.

12. A method for monitoring the cellular composition of a functional neuron cell set, comprising repeatedly performing the method of claim 10 at different time points, and monitoring the cellular composition of the functional neuron cell set based on a comparison of the cellular composition of the functional neuron cell set detected at the different time points.

13. A method for identifying compounds that influence the spatiotemporal electrophysiology and transcriptional profiles and / or composition of cells in a functional neuronal cell ensemble, comprising performing the method according to any one of claims 1 to 10 in the presence and absence of at least one candidate compound, wherein, In the presence or absence of at least one candidate compound, or when compared with suitable controls, differences in the spatiotemporal electrophysiological dynamics and transcriptional dynamics and / or composition of the cells in the functional neuronal cell set identify compounds that have an effect on the spatiotemporal electrophysiological dynamics and transcriptional profiles and / or composition of the cells in the functional neuronal cell set, wherein, preferably, the effect is a therapeutic effect, such as modification of the cell composition, alteration of gene expression, anticancer effect, neuroprotective effect, morphological effect, cognitive effect, and / or behavioral effect.

14. An apparatus for performing the method according to any one of claims 1 to 13, particularly, the apparatus comprising a collection of neuronal cells to be analyzed, an electrode array adapted to record patterns of local field potentials (LFPs) of the collection during optical imaging, a second array adapted to perform spatial resolution transcriptome analysis of the cells in the collection during optical imaging, and a computer program product for processing the acquired data, the processing of the acquired data including aligning the spatial localization patterns of the LFPs with the spatial resolution sequencing and the transcriptome analysis, the computer program product optionally having additional buffers and instructions for use.

15. Use of the apparatus according to claim 14, the apparatus being used to detect spatiotemporal electrophysiological dynamics and transcriptional profiles of cells in a functional neuronal cell set, to identify the composition of the functional neuronal cell set, to monitor the spatiotemporal electrophysiological dynamics and transcriptional dynamics of cells in the functional neuronal cell set, to monitor the cellular composition of the functional neuronal cell set, or to identify compounds that affect the spatiotemporal electrophysiological dynamics and the transcriptional profile, and / or to identify the cellular composition of the functional neuronal cell set.

Images