Human neuronal circuits

Abstract

The present invention relates to a human cortico-striato-nigral circuit on chip comprising: i) a population of human striatal medium spiny neurons (MSNs) MSNs expressing MEIS2, BCL11B and GABAergic lineage markers GAD2 and DLX6-AS1, said MSNs thereby forming a medium spiny neuronal circuit, wherein said medium spiny neuronal circuit comprises at least two subpopulations of MSNs expressing dopaminergic D1- or D2- receptors; ii) a population of nigral dopaminergic neurons expressing FOXA1, LMX1A, EN1 and NR4A2, thereby forming a nigral dopaminergic neuronal circuit; and iii) a population of cortical glutamatergic neurons expressing LHX2 and SLC17A6 and / or SLC17A7, thereby forming cortical circuit; a multi-electrode array (MEA) chip for recording neuronal activity, said neuronal activity resulting at least in part from dopaminergic, GABAergic and / or glutamatergic neurotransmission; wherein said medium spiny neuronal circuit is spatially arranged onto said MEA chip for receiving excitatory and / or modulatory inputs from said cortical circuit iii) and / or said nigral dopaminergic circuit ii); and wherein each one of said populations i)-iii) are differentiated from a human pluripotent stem cell. The present invention further relates to a method for producing said cortico-striato-nigral circuit on chip, and a method of determining the effect of a candidate agent on neuronal activity of the cortico-striato- nigral circuits on chip.

Description

[0001] PaVer / CSNOC / 870

[0002] HUMAN NEURONAL CIRCUITS

[0003] Field of the invention

[0004] The present invention relates to field of human neuronal circuits and devices comprising thereof. In particular, the invention relates to in vitro human circuits comprising, for example, a striatal circuit, arranged on a chip, for use in neurological disorder diagnosis and treatment, drug discovery and / or patient stratification.

[0005] Background

[0006] Neurodegenerative diseases, such as Parkinson's disease (PD), Huntington's disease, and others, pose significant challenges to global health due to their progressive nature and the lack of effective treatments. Early identification of individuals at risk of developing these disorders is advantageous for improving patient outcomes and / or facilitating disease-modifying interventions. Detecting these conditions at preclinical and / or prodromal stages is a constant challenge in the field, as timely therapeutic interventions may have greater potential for success.

[0007] Striatum is part of the human brain gaining more interest in pharmacological research for role in neurodegenerative diseases with the motor and / or movement disorder component. Striatal nuclei control cognition, motor planning and reward processing. Neural projections from pyramidal neurons in the cerebral cortex towards the striatum are essential components of the neural circuits controlling motivated behavior. The striatum is part of the basal ganglia and coordinates functions such as reward, motor planning, and execution (Haber, J Chem Neuroanat. 2003: 26, 317-330). The location of its nuclei deep within cerebral hemispheres, as well as the complexity of its neuronal signalling, make the understanding of the striatal disfunctions and search for new therapies challenging. Medium spiny neurons (MSNs), the primary cell type in the striatum, integrate excitatory signals from the cortex and thalamus, and modulatory input from dopaminergic neurons in the substantia nigra pars compacta (SNpc) (Gerfen & Surmeier, Annu Rev Neurosci. 2011: 34, 441-466). MSNs project to other basal ganglia nuclei via direct and indirect pathways (Albin et al., Trends Neurosci. 1989: 12, 366-375). Striatal neurons are precisely modulated by dopamine and any failure in said dopamine modulation causes neurobehavioral and motor disorders. Dopamine (DA) affects these pathways oppositely: it increases excitability in direct pathway MSNs and decreases it in indirect pathway MSNs (Gerfen & Surmeier, Annu Rev Neurosci. 2011: 34, 441-466, Calabresi et al., Nat Neurosci. 2014: 17, 1022-1030, Gerfen et al., Science. 1990: 250, 1429-1432). Neurodegenerative diseases like Huntington's (HD) and Parkinson's (PD)PaVer / CSNOC / 870

[0008] disrupt these pathways, impairing motor functions (Albin et al., Trends Neurosci. 1989: 12, 366-375), yet it is very difficult to study this circuit in a reliable human model and / or a patient specific manner.

[0009] In the prior art, human basal ganglia and in particular, the human striatum has been functionally analysed in patients with movement disorders undergoing deep brain stimulation (DBS) (Singh et al. PNAS. 2016: 113, 9629-9634). While informative, these recordings are short, they have low resolution, are typically limited to one brain regions (not involving the substantia nigra), invariantly represent late stages of the diseases and do not include recordings from unaffected individuals (Singh et al. PNAS. 2016: 113, 9629-9634; Wozny et al. Neuroimage. 2020: 217, 116904; Swinnen et al., Neuroimage. 2022: 254, 119147). Hence, the state of art understanding of the basal ganglia in health and disease largely comes from experimental animal models, but these models have two significant limitations. First, they do not replicate the cell type vulnerability observed in complex disorders like PD and HD. Second, the complex etiology of these disorders involves both common and rare human genetic variations, often in noncoding regions, which are challenging to model with animal genomes.

[0010] The 3D organogenesis using human induced pluripotent stem cells (iPSCs) has been shown to drive cells towards maturer states than 2D cultures (reviewed in Pasca, 2018). For example, construction of human 3D striato-nigral assembloids which were shown to recapitulate medium spiny neuronal circuits and / or projections, was reported previously Wu et al., PNAS. 2024: 121(22), e2316176121).

[0011] US2022 / 0364053 discloses human striatal organoids or spheroids (hStrS) and human midbrain organoids or spheroids (hMbS) which may be generated entirely from human pluripotent stem (hPS) cells. However, the current 3D technology is hampered by the technically difficult procedure of bringing neurons of distantly developing human brain areas together (in so-called assembloids), thereby failing to recapitulate anatomical connectivity of the human brain. Do et al. (npj Parkinson's disease. 2024: 10:82) disclosed in vitro striatal circuits on a microfluidic chip, however, said model was limited to long times for measuring neuronal activity of single neurons.

[0012] There is a need in the art for new models for studying the diseases and / or disorders which cannot be recapitulated in animal models. There is a need for reliable human models for striatum related diseases and / or disorders, as well as new biomarker and therapy development to target said diseases and / or disorders. There is a need for models allowing for a clinical translation of therapeutics that can alter the output of the striatal circuit in a human specific context. There is a need in the art for new methods which could allow early and reliable detection and / or stratification of a cell state, for example a healthy and / or diseased cell state.PaVer / CSNOC / 870

[0013] Summary

[0014] The present invention aims, in the first place, to provide an alternative, improved model, i.e. a system to capture neural circuits formation, function and / or dysfunction in physiological and / or pathophysiological context.

[0015] In a first aspect, the present invention relates to a human cortico-striato-nigral circuit on chip comprising:

[0016] i) a population of human striatal medium spiny neurons (MSNs) MSNs expressing MEIS2, BCL11B and GABAergic lineage markers GAD2 and DLX6-AS1, said MSNs thereby forming a medium spiny neuronal circuit, wherein said medium spiny neuronal circuit comprises at least two subpopulations of MSNs expressing dopaminergic DI- or D2- receptor; ii) a population of nigral dopaminergic neurons expressing FOXA1, LMX1A, EN1 and NR4A2, thereby forming a nigral dopaminergic neuronal circuit; and

[0017] iii) a population of cortical glutamatergic neurons expressing LHX2 and SLC17A6 and / or SLC17A7, thereby forming cortical circuit;

[0018] a multi-electrode array (MEA) chip for recording neuronal activity, said neuronal activity resulting at least in part from dopaminergic, GABAergic and / or glutamatergic neurotransmission;

[0019] wherein said medium spiny neuronal circuit is spatially arranged onto said MEA chip for receiving excitatory and / or modulatory inputs from said cortical circuit iii) and / or said nigral dopaminergic circuit ii); and

[0020] wherein said populations i)-iii) are differentiated from human pluripotent stem cells.

[0021] The present invention, according to its first aspect, relates to a in vitro cortico-striato-nigral circuit on a multielectrode array (MEA) chip comprising populations human cortical excitatory, medium spiny (MSN) and nigral dopamine neuron. Surprisingly, the differentiated populations of MSNs, nigral dopaminergic and cortical glutamatergic neurons showed a functional connectivity and / or maturation across the network, thereby recapitulating anatomical, physiological and / or pathophysiological connectivity. Preferably, the patterns in observed connectivity and / or synchronization of the populations could be used in further characterization of physiological and / or pathological condition, for example, the presence of specific genetic background by examining the electrophysiological features of the cortico-striato-nigral circuit on the chip of the invention. Moreover, said circuits allow for simulation of diseaselike conditions, allowing for a new model for testing and / or screening the biomarkers and / or drugs for the diseases and / or disorders related to pathological striatal output. The obtained circuits show that the cortical, striatal, and dopamine neurons closely resemble their developing brain counterparts, with distinct DI- and D2-MSN subtypes within the striatal neurons. When placed on multielectrode arrays (MEA), these circuits form networks that become increasingly active, regular, and / or synchronized overPaVer / CSNOC / 870

[0022] time. Using genetic and pharmacological tools, relevant connectivity within the circuits can be demonstrated. Fort the first time, an in vitro model showing isolated the contributions of glutamate and dopamine to circuit activity was generated, with demonstrated activity for example, during the period of up to 5 months. Advantageously, a human cortico-striato-nigral circuit on chip allows for recreating dopamine modulation and observing its effect on distinct DI- and D2-MSN subtypes within the striatal neurons, which was not possible, using the known alternative models in the art. The population of MSNs receives abundant innervation from cortical and nigral neurons and the generated in vitro human neuronal network is modulated by dopamine. The present invention, according to its first aspect, provides for the first time a high throughput system allowing to measure and integrate the electrophysiological activity of the more complex, neural activities of the neural networks e.g., neural circuits, which are integrated in a complex cortico-striato-nigral circuits. It has been surprisingly shown that the generated human cortico-striato-nigral circuit on chip offers a promising avenue for the development of new therapeutic strategies targeting human striatal dysfunction and / or dopamine modulation thereof.

[0023] In a second aspect, the present invention relates to a method for producing a cortico-striato-nigral circuit on chip according to the present invention, the method comprising:

[0024] (a) providing (i) a lateral ganglionic eminence progenitor cell from a first human pluripotent stem cell; (ii) a ventral midbrain progenitor from a second human pluripotent stem cell; and (iii) a dorsal forebrain progenitor cell from a third human pluripotent stem cell;

[0025] (b) seeding the lateral ganglionic eminence (i), ventral midbrain (ii), and dorsal forebrain (iii) progenitor cells at three different locations of the MEA chip, so that the lateral ganglionic eminence (i) progenitor cell is spatially arranged onto said MEA chip for receiving excitatory and / or modulatory inputs from dorsal forebrain iii) progenitor cell and / or ventral midbrain (ii) progenitor cell;

[0026] (c) culturing the lateral ganglionic eminence (i), ventral midbrain (ii) and dorsal forebrain (iii) progenitor cells in conditions allowing for formation of differentiated and / or matured populations of medium spiny (i), nigral dopaminergic (ii) and cortical glutamatergic (iii) neurons; and / or in conditions allowing for synaptic activity between the cells.

[0027] The method for producing a cortico-striato-nigral circuit on chip according to the second aspect, allows for provision of a model which can be used for longitudinal, i.e. measurements of cortico-striato-nigral circuits electro physiological activity over time. The method of the second aspect allows for a provision of a relevant model for recreating pathophysiological context of disorders connected to striatal disfunction.PaVer / CSNOC / 870

[0028] In a third independent aspect, the present invention relates to a method of determining the effect of a candidate agent on neuronal activity of the cortico-striato-nigral circuits on chip according to the invention or obtainable by the method for producing a cortico-striato-nigral circuit on chip of the invention, the method comprising:

[0029] a) contacting the candidate agent with cortico-striato-nigral circuits on chip of the invention or obtainable by the method for producing a cortico-striato-nigral circuit on chip of the invention; b) observing the change of in a neuronal activity and / or function of circuits i)-ii) and iii),

[0030] c) determining the effect of the candidate agent based on resulting change in the neuronal activity and / or function of circuits i)-ii) and iii).

[0031] The method of determining the effect of a candidate agent on neuronal activity of the cortico-striato-nigral circuits on chip of the invention allows for alternative, more precise, longitudinal and / or specific evaluation of the candidate agent, for example a drug, on activity cortico-striato-nigral circuits. The method is advantageous for its reproducibility and / or specificity, compared to the known methods in the art.

[0032] In a fourth independent aspect, the present invention pertains to a method to detect a genetic mutation or a cell state, said mutation or said cell state being preferably related to a disease and / or disorder, the method comprising the steps of:

[0033] providing the human cortico-striato-nigral circuit on a chip of the invention, or obtainable by the method for producing the human cortico-striato-nigral circuit on chip of the invention; recording the electrophysiological activity of at least one circuit of the human cortico-striato- nigral circuit on a chip of the invention or obtainable by the method for producing the human cortico-striato-nigral circuit on chip of the invention; and

[0034] using a computer implemented method to associate the recorded electrophysiological activity with presence of said genetic mutation or said state.

[0035] The method according to a fourth aspect allows using the recorded electrophysiological signatures obtainable by the readout of the neuronal activity of at least one circuit of the cell populations on the chip to detect the underlying molecular etiology. For example, the method may allow for detecting a presence of a specific genetic mutation or a cell state, for example a healthy or diseased cell state in a cell in the circuit on the chip of the invention. The method of the fourth aspect may allow using the electrophysiological readouts of the chip of the invention to detect the specific mutation or the cell state, thereby enabling the molecular stratification of a cell in the circuit on the chip of the invention.PaVer / CSNOC / 870

[0036] Brief description of the Figures

[0037] The figures described are only schematic and are non-limiting. In the figures, the size of some of the elements may be exaggerated and not drawn on scale for illustrative purposes.

[0038] Figure 1. Generation of cortico-striato-nigral (CSN) circuits on top of high-density multielectrode arrays. (A) Schematic depicting the three major components of the cortico-striato-nigral circuitry and the approach undertaken to reconstitute it in vitro. (B) CRISPR / Cas9-based gene editing strategy followed to generate a triple reporter line to track the three major subtypes of neuronal populations present in the circuit. Note that SLC17A7 reporter requires a viral vector to enable the visualization or genetic targeting of cortical excitatory neurons. (C-C") Immunofluorescence validation of the triple reporter line in dopaminergic, striatal, and cortical neurons. Upper arrowhead in (C') points to a mClover3+ DAN not positive for GAD65 nor mScarlet. Lower arrow in (C') points to specific co-labeling of the mScarlet and GAD65 in a striatal neuron. Arrowhead in (C") points to a BFP- and mScarlet+ neuron in between BFP2+ and VGLUT1+ cortical neurons. (D) Photographs of the seeding strategy followed to deposit the three region-specific NPCs onto the MEA chips. (E) Live imaging of >=100-day old CSN circuits showing the expected enrichment of each neuronal subtype in their corresponding lane. Scale bar: 500 pm. (F-F") Grayscale images showing each of the channels separately. Note the pervasive innervation of the striatal lane by dopaminergic and glutamatergic axons emanating from the ventral midbrain and cortical layers, respectively. (G) Live imaging of the striatal lane of a 117-day old CSN circuit showing several striatal neurons completely surrounded by dopaminergic and cortical terminals. Scale bar: 50 pm. (H) Uniform Manifold Approximation and Projection (UMAP) plot of the single nuclei sequencing data obtained from 4 CSN circuits and 3 striatal-only monocultures totaling 25.241 cells obtained from 2 independent hiPSCs. The three major regional identities are shade and / or color-coded and enclosed within a dashed line. (I) Bubble plot showing the percent of cells that express a given regional marker and the expression levels of such marker. (J) The same UMAP plot showing the three major MSN subtypes shade and / or color-coded and indicated by a dashed line. (K) Bubble plot showing the percent of cells that express a given MSN subtype-specific marker and the expression levels of such marker. (L) UMAP plots of CSN circuits (n=4) and Str-only cultures (n=3) (M) Compositional analysis of the MSN component in CSN circuits and Str-only cultures. (N) Gene ontology (GO) analysis of those genes significantly upregulated in MSNs from CSN circuits versus those from Str-only cultures.

[0039] Figure 2. Cortical and dopaminergic neurons establish presynaptic connections onto striatal neurons. (A-A') Structural illumination microscopy imaging of a DARPP32+ MSN within a triple reporter cortico-striato-nigral (CSN) circuit contacted by dopaminergic and glutamatergic terminals. Scale bar: 5pm. (B-D) Correlated light and electron microscopy imaging of presynaptic contacts of cortical (C-C" and D-D'")PaVer / CSNQC / 870

[0040] and dopamine (D-D'") neurons on a striatal neuron from a 100-day old CSN circuit. Scale bar in (B): 10 pm. (E) Schematics representing the experimental layout followed to retrogradely trace presynaptic neurons to striatal neurons using recombinant rabies virus. (F-F") Live imaging of a day 110 circuit before fixation showing the specific labelling of striatal neurons with the recombinant TVA receptor and the pre- and postsynaptic neurons with the recombinant rabies vector expressing tdTomato. Scale bar: 500 pm. (G) Immunofluorescence analysis of striatal neurons reveals specific labeling of CTIP2+ and DARPP32+ neurons with the recombinant TVA receptor. Scale bar: 20 pm. (H) Quantification of the number of double-positive DARPP32+ / CTIP2+ or single-positive CTIP2+ striatal neurons positive for the recombinant TVA receptor (coexpressed with mCloverS). n=3 independent experiments. (I) Immunofluorescence analysis of ventral midbrain neurons reveals specific labeling of TH+ and FOXA2+ neurons with the recombinant tdTomato-expressing rabies vector. Scale bar: 20 pm. (J) Quantification of the number of TH+ dopaminergic neurons retrogradely traced with the recombinant tdTomato-expressing rabies vector. n=4 independent experiments. (K) Immunofluorescence analysis of cortical neurons reveals specific labeling of layer specific markers CTIP2+ and SATB2+ neurons with the recombinant tdTomato-expressing rabies vector. Scale bar: 20 pm. (L) Quantification of the number of positive CTIP2+ and SATB2+ cortical neurons retrogradely traced with the recombinant tdTomato-expressing rabies vector. n=3 independent experiments.

[0041] Figure 3. Longitudinal characterization of cortico-striato-nigral (CSN) circuits allows training Al-based classifiers to predict neuronal identity.

[0042] (A-C) Registered neuronal activity in each of the three areas coincides with the presence of neurons. Grey arrowheads indicate detected spikes while white arrowhead points to the neuron whose waveform is reconstructed and shown in the bottom-left corner. (D) Raster plots of detected spikes along 300 seconds from 5 randomly selected cells per area across 4 different time points (D= number of days from the iPSC stage). A 10-second zoomed-in clip from the day 180 raster plot highlights synchronized activity at the sub second scale. (E) Dot plot showing the evolution of circuit's network burst rate across 4 time points. Data obtained from 89 recordings from 15 independent experiments and 4 hiPSC lines. Black dots represent the average and the grey and / or green shade represents the S.E.M. The same data is used for (F-N) (F) Dot plot showing the evolution of time the network is bursting across 4 time points. (G) Dot plot showing the evolution network's synchrony as measured by the STTC approach across 4 time points. (H) Dot plot showing the evolution of the average of the population coupling across 4 time points. (I) Dot plot showing the evolution of time the average firing rate across 4 time points. (J) Dot plot showing the evolution of the firing irregularity across 4 time points. (K) Dot plot showing the evolution of the average burst rate across 4 time points. (L) Dot plot showing the evolution of time the average burst durationPaVer / CSNQC / 870

[0043] across 4 time points. (M) Dot plot showing the evolution of the average spikes per burst across 4 time points. (N) Dot plot showing the evolution of the average non-burst firing rate across 4 time points. (0) Summary of the datasets used for training the random forest (RF) classifier. (P) Receiver operator curve (ROC) showing the prediction accuracy for every cell type. (Q) 3-dimensional multidimensional scaling analysis plot showing the dispersion of the different neuronal subtypes according to their electrophysiological profiles. (R-R') Example of the cell type identity prediction of spike sorted neurons from a 100-day old triple reporter CSN circuit.

[0044] Figure 4. Glutamate synchronizes circuit's activity through its excitatory role.

[0045] (A) Traces of 6 representative neurons before and after the application of the glutamate receptor blockers. (B-B'") Population overall firing rate (B), network burst (NB) firing rate (B') and inter-NB firing rate (B") and raster plot 1 minute before and after the application of the ionotropic glutamate receptor blockers of D-APV (100 pM) and DNQX (10 pM). Dark grey and / or magneta and light grey and / or yellow lines in (B) represent the averaged intra- and inter-NB firing rates over time. The grey and / or red horizontal bars on top of the raster plot (B'") label identified NBs while dark grey and / or magenta ones label single neuron bursts. (C) Box & whiskers graph showing the network burst rate before and after DAPV and DNQX. Data obtained from 7 experiments performed in 6 circuits from 3 hiPSC lines. Same data are used for (C-J), the grey or black box & whiskers graph indicating before and / or after DAPV and DNQX. (D) Box & whiskers graph showing circuit's synchronization (as measured by spike contrast) before and after DAPV and DNQX. (E) Box & whiskers graph showing the average population coupling before and after D-APV and DNQX. (F) Box & whiskers graph showing the average firing rate before and after D-APV and DNQX. (G) Box & whiskers graph showing the average non-burst firing rate before and after D-APV and DNQX. (H) Box & whiskers graph showing average burst rate before and after D-APV and DNQX. (I) Box & whiskers graph showing the average burst duration before and after D-APV and DNQX. (J) Box & whiskers graph showing the average number of spikes per bursts before and after D-APV and DNQX. (K) Schematics depicting the experimental procedure to directly test excitatory connectivity by selectively stimulating cortical neurons. (L) Raster plots and traces of a cortical and a striatal neuron during optogenetic stimulation of cortical neurons before and after the application of the glutamate receptor blockers. These two neurons are connected as shown by the cross-correlograms (CCG) plots on the right. (M-M") Pie charts summarizing the effects of optogenetic stimulation on putative striatal, ventral midbrain and cortical ! units in the presence or absence of glutamate blockers. Data obtained from 2 experiments performed in 2 independent circuits. (N) Heatmaps showing the CCGs across all pairs of putative connected neurons before and after the application of glutamate blockers. Data obtained from 3 experiments performed in 2 independent circuits. Grey and / or colored linesPaVer / CSNOC / 870

[0046] crossing the center of the CCGs indicate the identity of the post-synaptic neuron (same grey shade and / or colors as in i). (0) Putative excitatory connections revealed by optogenetic stimulation of LV-transduced SLC17A7-expressing neurons in 2 experiments conducted in a single circuit. The dots represent electrodes where activity was registered and their grey shade and / or color corresponds to the physical location: the top grey cluster and / or blue is cortical area; the middle grey cluster and / or magenta is striatal and the down grey cluster and / or yellow is ventral midbrain. The lines represent putative connections with their lighter edge representing the putative presynaptic neuron and their darker edge representing the putative postsynaptic one. (P) Distribution of the number of synapses in each connection stemming from a cortical excitatory neuron. The line represents an analytic model prediction using the equations described in Lynn et al. (Nat Phys. 2024: 20, 484-491).

[0047] Figure 5. Dopamine modulates neuronal excitability and network synchronization through dopamine DI receptors.

[0048] (A-A'") Population overall firing rate (A), network burst (NB) firing rate (A') and inter-NB firing rate (A") and raster plot 1 minute before and after the application of the dopamine receptor 1 blocker SCH23390 (0.5 pM). Dark grey and / or magenta and light grey and / or yellow lines in (A) represent the averaged intra- and inter-NB firing rates over time. Horizontal bars on top of the raster plot (A'") label identified NBs while dark grey / magenta ones label single neuron bursts. (B) Box & whiskers graph showing the average firing rate before and after SCH23390. Data obtained from 7 experiments performed in 6 circuits from 4 hiPSC lines. Same data are used for (C-J). (C) Box & whiskers graph showing the average non-burst firing rate before and after SCH23390. (D) Box & whiskers graph showing the average burst rate before and after SCH23390. (E) Box & whiskers graph showing the network burst rate before and after SCH23390. (F) Box & whiskers graph showing the change in circuit's synchrony (as measured by the Spike-contrast) before and after SCH23390. (G) Box & whiskers graph showing the change the average population coupling before and after SCH23390. (H) Box & whiskers graph showing the average internetwork burst firing rate before and after SCH23390. (I) Box & whiskers graph showing the average intranetwork burst firing rate before and after SCH23390. (J) Box & whiskers graph showing the difference in change between the intra- and the inter-network burst firing rate before and after SCH23390.

[0049] Figure 6. Alpha-synuclein fibrils disrupt cortico-striato-nigral (CSN) circuit's network dynamics.

[0050] (A) Experimental procedure for inoculating alpha-synuclein fibrils and measure their effects on the CSN circuits. (B) Immunofluorescence analysis of neurons in CSN circuits 24 hours after being inoculated with alpha-synuclein fibrils labelled with ATTO594 reveals efficient neuronal uptake and perinuclear (arrows) and neuritic localization (arrowheads). Scale bar: 10 pm. (C) Summary of the datasets used for trainingPaVer / CSNOC / 870

[0051] the random forest (RF) classifier. (D) Classification accuracy of the trained RF classifier as a function of the certainty or probability of the prediction (E) Receiver operator curve (ROC) showing the prediction accuracy forthe treatment status. (F) Out-of-bag importance of the metrics used to train the RF classifier.

[0052] (G) Multidimensional scaling analysis shows an effective separation of the treated and non-treated neurons according to their electrophysiological characteristics (H) Challenging the RF classifier with the test set 1 consisting of the pre-treatment baseline recordings. (I) Challenging the RF classifier with the test set 2 consisting of the recordings used for the longitudinal analysis of Fig. 3.

[0053] Figure 7. Machine learning predicts the mutational status of cortico-striato-nigral circuits out of electrophysiological features. (A) Scheme depicting the generation of the two isogenic series carrying the Parkinson's disease-related mutations LRRK2 G2019S and GBA N409S from two independent hiPSC lines. (B) Table summarizing the number of recordings and neurons identified per recording in the multiwell HD-MEAs assigned to either the train or the test sets. (C-C") Confusion matrices showing the classification accuracies of neurons from the entire test set or from each of the hiPSC lines separately (C'-C"). (D-D') Multidimensional scaling plots show an effective separation of the WT, LRRK2 and GBA neurons (D) across the two different parental lines (D') according to the proximity matrices extracted from the random forest classifier. (E-E”) SHAP values for the 10 most important features to discriminate between WT, LRRK2 and GBA neurons. (F) Significance of the Genotype x Culture. Age interaction term from LMMs fitted to features defined per each neuronal layer. Only ventral midbrain layer features yielded significant results. (G-G') Ventral midbrain-specific mean burst duration and population coupling significantly correlate with mean network burst duration and network burst total time in the longitudinal recording series described in Fig. 3.

[0054] Detailed description

[0055] Definitions

[0056] The present invention will be described with respect to particular embodiments and with reference to certain figures, but the invention is not limited thereto. Any reference signs in the claims shall not be construed as limiting the scope. It is to be understood that not necessarily all aspects or advantages may be achieved in accordance with any particular embodiment of the invention. Thus, for example those skilled in the art will recognize that the invention may be embodied or carried out in a manner that achieves or optimizes one advantage or group of advantages as taught herein without necessarily achieving other aspects or advantages as may be taught or suggested herein. The drawings described are only schematic and are non-limiting. In the drawings, the size of some of the elements may be exaggerated and not drawn on scale for illustrative purposes.PaVer / CSNOC / 870

[0057] The invention, both as to organization and method of operation, together with features and advantages thereof, may best be understood by reference to the following detailed description when read in conjunction with the accompanying figures. The aspects and advantages of the invention will be apparent from and elucidated with reference to the embodiment(s) described hereinafter. Reference throughout this specification to "one embodiment" or "an embodiment" means that a particular feature, structure or characteristic described in connection with the embodiment is included in at least one embodiment of the present invention. Thus, appearances of the phrases "in one embodiment" or "in an embodiment" or "in embodiments" in various places throughout this specification are not necessarily all referring to the same embodiment, but may. Similarly, it should be appreciated that in the description of exemplary embodiments of the invention, various features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the disclosure and aiding in the understanding of one or more of the various inventive aspects. This method of disclosure, however, is not to be interpreted as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive aspects lie in less than all features of a single foregoing disclosed embodiment.

[0058] Where an indefinite or definite article is used when referring to a singular noun e.g. "a" or "an", "the", this includes a plural of that noun unless something else is specifically stated. Where the term "comprising" is used in the present description and claims, it does not exclude other elements or steps. Furthermore, the terms first, second, third and the like in the description and in the claims, are used for distinguishing between similar elements and not necessarily for describing a sequential or chronological order. It is to be understood that the terms so used are interchangeable under appropriate circumstances and that the embodiments, of the invention described herein are capable of operation in other sequences than described or illustrated herein.

[0059] The definitions provided herein should not be construed to have a scope less than understood by a person of ordinary skill in the art.

[0060] The term "population" or "cell population" as used herein, relates to at least two, at least three, at least four, preferably a group of more than five cells, or a plurality of cells that share common characteristics, such as type, function, morphology, expression markers, or origin. The cells of the same population can interact among each other and / or cells of a different population.PaVer / CSNOC / 870

[0061] The terms "circuit", "neuronal circuit" or "neural circuit" which are interchangeably used through the description, mean populations or groups of neurons which are in synaptic communication with each other. These the synapses or the synaptic communication can be excitatory or inhibitory or modulatory influencing the activity of the connected neurons. The term "circuit" as used herein, can relate to simple circuits, involving just a few neurons, as well as highly complex networks / circuits involving thousands or millions of neurons. Circuits, as used herein can interact, i.e., be in synaptic communication with other circuits and / or other brain parts, to form larger brain networks that underpin different neural functions.

[0062] The term "Medium spiny neurons" (MSNs) relates to a type of inhibitory neuron found primarily in the striatum, a key part of the basal ganglia in the brain. They make up about 90-95% of the neurons of striatum. MSNs play a crucial role in motor control, reward, and various aspects of behavior. They are involved in both the direct and indirect pathways of the basal ganglia, which help regulate movement and other functions. MSNs are GABAergic, meaning they primarily use gamma-aminobutyric acid (GABA) as their neurotransmitter, which has inhibitory effects on other neurons activities. MSNs can be classified into two main types based on the dopamine receptors they express: Dl-type and D2-type. Dl-type MSNs are part of the direct pathway and can facilitate movement, while D2-type MSNs are part of the indirect pathway and can inhibit movement.

[0063] The term "nigral dopaminergic neurons" relates to nerve cells located in the substantia nigra, a region in the midbrain. These neurons produce and release dopamine, a crucial neurotransmitter involved in regulating movement, reward, and various other functions. The term "substantia nigra”, as used herein, can relate to: Pars compacta (SNc) which contains the dopaminergic neurons that project to the striatum, forming the nigrostriatal pathway. This pathway is essential for motor control and is significantly affected in Parkinson's disease, and Pars reticulata (SNr) primarily containing GABAergic neurons and plays a role in conveying signals from the basal ganglia to other brain regions. In some embodiments, the nigral dopaminergic neurons originate preferably exclusively from Substantia nigra pars compacta (SNc). The loss of nigral dopaminergic neurons, particularly in the pars compacta, is a hallmark of Parkinson's disease.

[0064] The term "cortical glutamatergic neurons" relates to excitatory neurons found in the cerebral cortex. These neurons use glutamate as their primary neurotransmitter, which can play a crucial role in synaptic transmission, plasticity, and overall brain function. The cortical glutamatergic neurons as referred to herein, can have following characteristics: they may have: a) excitatory function, meaning, they can excite other neurons; b) may be projection neurons by exerting the functions of sending long-rangePaVer / CSNOC / 870

[0065] connections to other parts of the brain and spinal cord; c) may show layer distribution, meaning they may be distributed across different layers of the cortex, each layer having distinct types of glutamatergic neurons with specific functions; d) may show activity-dependent plasticity, meaning that they can undergo changes in response to activity, which is important for adapting to new information and experiences.

[0066] The gene names and / or abbreviations used in this invention are approved by Human Genome Organisation (HUGO) including HUGO Gene Nomenclature Committee (HGNC). The assigned unique and standardized names and symbols of the human genes referred herein encompass the known synonyms, which could be retrieved on HUGO approved database https: / / www.genenames.org / . The gene names as used in the description and claims of the present invention include the gene name synonyms referred to under the same HGNC ID in the HUGO approved database https: / / www.genenames.org / . The following genes are referred to shown as "gene name" and "(HGCN ID number)" followed by at least one other possible alternative name, if applicable: MEIS2 (HGNC:7001), with alternative names MRG1 and HST18361; BCL11B (HGNC:13222) with alternative names CTIP-2, CTIP2, hRITl-alpha and SMARCM2; GAD2 (HGNC:4093), with an alternative name GAD65; DLX6-AS1 (HGNC:37151) with alternative names FU34048 and Evf-2; F0XA1 (HGNC:5021); LMX1A (HGNC:6653) with an alternative name LMX1.1; EN1 (HGNC:3342); NR4A2 (HGNC:7981) with alternative names TINUR, NOT, RNR1, and HZF-3; LHX2 (HGNC:6594) with alternative names LH-2, hLhx2; SLC17A6 (HGNC:16703) with alternative names DNPI and VGLUT2; SLC17A7 (HGNC:16704) with alternative names BNPI and VGLUT1; F0XP1 (HGNC:3823) with alternative names QRF1, 12CC4, HSPC215 and hFKHIB; PDYN (HGNC:8820) with alternative names PENKB and ADCA; TSHZ1 (HGNC:10669) with alternative names NY-CO-33 and TSH1; F0XP2 (HGNC:13875) with an alternative name CAGH44; SATB2 (HGNC:21637) with alternative names KIAA1034 and FU21474; KCNJ6 (HGNC:6267) with alternative names Kir3.2, GIRK2, KATP2, BIR1, hiG IRK2; S0X6 (HGNC:16421).

[0067] Gene names as referred to in claims and the description and any one of their above-mentioned, alternative names, may be interchangeably used in this invention.

[0068] The terms "neuronal activity" and / or "electrophysiological activity" as used herein, relate to an electrophysiological impulse, i.e., an electrical activity. Said neuronal activity is resulting, at least in part from a dopaminergic, GABAergic and / or glutamatergic neurotransmission between the neuronal cells within the circuits.PaVer / CSNOC / 870

[0069] The term "human pluripotent stem cell" relates to a human cell able to develop into almost any cell type in the body. There are two main types of human pluripotent stem cells: a) Embryonic Stem Cells (ESCs), which are derived from early-stage embryos. Said ESCs can differentiate into all cell types of the body, making them highly versatile; and b) Induced Pluripotent Stem Cells (iPSCs), which are adult cells that have been genetically reprogrammed to an embryonic stem cell-like state. iPSCs share the same ability to differentiate into various cell types. "hiPSCs" and "iPSCs" are interchangeably used in this specification.

[0070] "hiPSCs" can be derived from, for example, human fibroblasts, blood cells, other skin cells, or any cell of the human body.

[0071] Said "ESCs" can be, for example, derived from human embryos.

[0072] The term "Multi Electrode Array" or "MEA" relates to a chip and / r a system featuring plurality of electrodes arranged to detect changes in voltage. Preferably, said electrodes are arranged to detect neural activity or the activity from cardiac cells generating such change in voltage. The term may include alternative MEA designs, which are arranged to record other types of phenomena but they also use electrical pulses for that and measure cells' properties related to, e.g., electrical conductivity. In some embodiments, MEA relates to a complementary-metal-oxide-semiconductor multi-electrode array (CMOS-MEA) chips, which allow for high-resolution recording of neuronal activity. These chips may possess a high electrode density, allowing for detailed analysis from single-cell level dynamics to population-level behaviours.

[0073] The terms "cortico-striato-nigral circuit" and "CSN circuit" are used interchangeably within the description of the present invention.

[0074] The term "cortico-striato-nigral circuits on chip of the invention" refers to the cortico-striato-nigral circuits on chip according to the first aspect of the invention, and / or any of the embodiments according to the first aspect.

[0075] The term "method for producing a cortico-striato-nigral circuit on chip according to the present invention" refers to the method according to the second aspect of the invention and / or any embodiment of the second aspect, which method is used to produce the cortico-striato-nigral circuits on chip of the invention.

[0076] The term "lineage" or "neuronal lineage" refers to the developmental pathway, including regional specification, that neural progenitor cells follow to become mature neurons. This process may involve several stages, including: Neural Progenitor Cells (NPCs), which are multipotent cells capable of selfrenewal and differentiation into various types of neural cells. NPCs further differentiate into neuroblasts; Neuroblasts which are immature neurons. Neuroblasts undergo maturation into mature neurons; and Mature Neurons, developing the complex structures and functions characteristic of fully developedPaVer / CSNOC / 870

[0077] neurons. The term "lineage" may relate to a specific region-specific population, in particular striatal, nigral and / or cortical circuits which crucial for the proper formation and function of the distinct circuits of the invention.

[0078] The terms "candidate agent" or "drug candidate" are interchangeably used in the description of the aspects and the embodiment of the present invention, and mean any therapeutic or prophylactic compound which may be used to maintain health, cure and / or alleviate a disease. Some non-limiting examples include a mall molecule, an antisense oligonucleotide (ASO), a gapmer, a siRNA, a shRNA, a zinc-finger nuclease, a meganuclease, a TAL effector nuclease, a CRISPR-Cas effector, an antibody or a fragment thereof, an alpha-body, a nanobody, an intrabody, an aptamer, a DARPin, an affibody, an affitin, an anticalin, and a monobody.

[0079] "Higher" or "increased" as used herein refers to at least 10%, at least 20%, at least 30%, at least 40%, at least 50%, at least 60%, at least 70%, at least 80%, at least 90%, at least 100%, at least 1.5 fold, at least 2 fold, at least 3 fold, at least 5 fold or at least 10 fold higher quantity or an effect. In some embodiments, "higher" or "increased" refers to a statistically significant difference. "Predominantly" as used herein, means at least 50%, at least 60%, at least 70%, at least 80%, at least 90%, at least 95% of quantity or an effect.

[0080] "Lower" or "decreased" as used herein is defined herein as a statistically significantly decreased, more particularly an at least 10%, at least 20%, at least 30%, at least 40%, at least 50%, at least 60%, at least 70%, at least 80%, at least 90%, at least 100%, at least 1.5 fold, at least 2 fold, at least 3 fold, at least 5 fold or at least 10 fold lower quantity or an effect. In some embodiments, "lower" or "decreased" refers to a statistically significant difference.

[0081] A LRRK2 mutation, as used herein, relates to a genetic change in the human LRRK2 gene (leucine-rich repeat kinase 2, Gene ID 120892). The mutations in LRRK2 have been associated to Parkinson's disease (PD).

[0082] AGBA mutation, as used herein, relates to a genetic change of the human GBAgene (glucocerebrosidase or GCase, Gene ID 2629), one of the known, and most common genetic risk factors for developing Parkinson's disease (PD).

[0083] The term "isogenic mutation" as used herein, means a genetic change introduced into a cell, while keeping the rest of the genome identical to the original, i.e., "parent" cell. Introducing the isogenic mutation preferably allows for observations of the effects of the single mutation without genetic background interference.

[0084] Terms "state" and "cell state" are interchangeably used within this description, and herein refer to the functional, molecular, and physiological condition of a cell at a given moment. The cell state, as usedPaVer / CSNOC / 870

[0085] herein, may refer to healthy and / or diseased and / or the type of diseased cell state. The cell state, as defined herein, may be affected by different factors such as genetic expression, genetic makeup; epigenetic factors, for example chemical modifications on DNA and / or histones which could influence genetic activity; metabolic status and / or activity of the cells, for example energy pathways; signalling environment, for example responses to hormones, cytokines, growth factors; and functional physiological or pathophysiological behaviour, such as dividing, differentiating, repairing, resting, an the like.

[0086] The term "confounding factors" as used herein, refers to variables or conditions that may influence electrophysiological measurements or the association between electrophysiological features and the health condition, potentially leading to biased or misleading results. These factors are unrelated variables that can affect neuronal activity or data analysis. Examples include genetic variations, environmental exposures, differences in experimental conditions, patient age, concurrent diseases, pharmaceutical use, or lifestyle factors such as diet.

[0087] As used herein, the term "cross-correlogram" or "CCG" refers to a measure of the temporal relationship between spike trains of two neurons, calculated by determining the distribution of time intervals between spikes in one neuron relative to spikes in another neuron. A sharp peak in the cross-correlogram at a short delay (e.g., less than 6 ms) indicates consistent temporal firing relationships between the neurons, which may be indicative of synaptic connectivity. Cross-correlograms may be calculated between pairs of neurons within the same neuronal subtype or between neurons of different subtypes within the neuronal circuit.

[0088] The term "spike time tiling coefficient" or "STTC", as used herein refers to a measure that quantifies spike train correlations between neurons while accounting for firing rate differences. The STTC may be used to assess correlations between neurons within the same layer or across different layers of the neuronal circuit, thereby providing a measure of inter-layer functional connectivity.

[0089] The term "spike contrast", as used herein refers to a measure of overall synchronicity in a neuronal network, quantifying the degree to which neuronal firing is temporally coordinated across the circuit. As used herein, and unless otherwise specified, the term "network burst" refers to a transient period of elevated activity involving multiple neurons across the neuronal circuit firing in a coordinated manner. Network burst features may include network burst rate, network burst duration, network burst total time, and activity levels between network bursts.

[0090] As used herein, and unless otherwise specified, the term "linear mixed-effects model" or "LMM" refers to a statistical model that accounts for both fixed effects (e.g., genotype, health condition) and random effects (e.g., individual variability, batch effects) when analysing electrophysiological features. LinearPaVer / CSNOC / 870

[0091] mixed-effects models are particularly suited for analysing longitudinal data and for accounting for hierarchical data structures.

[0092] Detailed description of embodiments

[0093] The aim of the present invention is to provide a reliable in vitro human neuronal network model, relevant to various neurodegenerative disorders related to striatal dysfunction, for example motor disorders such as Parkinson's and / or Huntington's disease. Preferably the model can be used to assess patient-specific functional deficits and link them to the genetics or environmental factors associated with these diseases and / or conditions.

[0094] The present invention, according to a first aspect, relates to a human cortico-striato-nigral circuit on chip comprising:

[0095] i) a population of human striatal medium spiny neurons (MSNs) MSNs preferably expressing MEIS2, BCL11B and GABAergic lineage markers GAD2 and DLX6-AS1, said MSNs thereby forming a medium spiny neuronal circuit, wherein said medium spiny neuronal circuit comprises at least two subpopulations of MSNs expressing dopaminergic DI- or D2- receptors;

[0096] ii) a population of nigral dopaminergic neurons preferably expressing FOXA1, LMX1A, EN1 and NR4A2, thereby forming a nigral dopaminergic neuronal circuit; and

[0097] iii) a population of cortical glutamatergic neurons preferably expressing LHX2 and SLC17A6 and / or SLC17A7, thereby forming cortical circuit;

[0098] a multi-electrode array (MEA) chip for recording neuronal activity, said neuronal activity resulting at least in part from a dopaminergic, GABAergic and / or glutamatergic neurotransmission; wherein said medium spiny neuronal circuit is spatially arranged onto said MEA chip for receiving excitatory and / or modulatory inputs from said cortical circuit iii) and / or said nigral dopaminergic circuit ii); and

[0099] wherein said populations i)-iii) are differentiated from human pluripotent stem cells.

[0100] It has been surprisingly found that the human cortico-striato-nigral circuit on chip according to a first aspect, can be used as a model and / or tool which can allow to define biomarkers from susceptible genomes and / or screening for therapeutic agents. The human cortico-striato-nigral circuit on chip of the present invention is centered around striatal MSNs as they are the main post-synaptic output of the nigral dopaminergic neurons, which are, for example, lost in PD. The human cortico-striato-nigral circuit on chip shows demonstrable glutamatergic and dopaminergic input onto MSN neurons. Herein the present invention provides a model and / or tool able to record from a large number of neurons toPaVer / CSNOC / 870

[0101] maximize data acquisition and / or to include methods to detect deviations in the composition of the different neuronal populations arising from various sources.

[0102] The human cortico-striato-nigral circuit on chip of the invention can overcome the issues observed with 3D organoids utilizing human induced pluripotent stem cells (iPSCs) which fail to bring distantly developing human brain areas together (in so-called assembloids). Moreover, the human cortico-striato-nigral circuit on chip of the present invention allows for conducting longitudinal recordings of neuronal activity across high cell numbers.

[0103] In the present invention, three physiologically distinct populations of 2D cultures combined with microelectrode arrays (MEA) were shown to be capable of providing highly detailed functional information at both single-neuron and network levels. Such large-scale neuronal recording enables multiple applications, including the development of computational models to identify cell types within diverse cultures, and / or to evaluate the effects of genetic mutations or pharmacological interventions based on single-cell and network characteristics. This invention teaches of optimized generation of enriched populations of cortical glutamatergic, ventral midbrain dopaminergic, and striatal medium spiny neurons, which were seeded at specific locations on a MEA, preferably a high-density MEA (HD-MEA). Surprisingly, the enriched populations of cortical glutamatergic, ventral midbrain dopaminergic, and striatal medium spiny neurons were shown to be more mature than mono-cultures. Moreover, the present invention allows for recording the activity of said neuronal populations forming circuits over a period of up to five months. A classifying tool such as an Al classifier can be utilized to distinguish cortical, striatal, and ventral midbrain neurons and validated this using an engineered human pluripotent cell expressing, upon differentiation, fluorescent markers linked to the three neurotransmitter classes. The human cortico-striato-nigral circuit on chip of the invention allows for screening / observing the pharmacological intervention, due to dopamine and glutamate contributions to network dynamics The present invention provides a reliable human microcircuit that recapitulates cardinal, expected features, including the capacity to distinguish neuronal subtypes on chip and reveal cell-specific neurotransmitter contributions to network behavior, and thereby may be used as a high throughput, patient specific tool to search for biomarkers and / or new therapeutic agents that could be effective in early-stages of striatal disfunction.

[0104] The human cortico-striato-nigral circuit on chip according to the present invention possesses features that make it well-suited for phenotyping at scale. Unlike 3D models such as organoids known in the art, the cortico-striato-nigral circuit is assembled from quality-controlled regionalized progenitor populations, ensuring there is little or no progeny from neighbouring regions. The high number of neurons sampled per experiment allows to generate machine learning models with the ability to predict cell type identity from electrophysiological parameters. Preferably the model can predict the neuronalPaVer / CSNOC / 870

[0105] identity with accuracies of >75%. This outperforms currently available non-invasive methods to interrogate electrophysiological function of organoid cultures, which yield information on 30-50 neurons at a time (Yang et al. Nat Biotechnol. 2024: doi:10.1038 / s41587-023-02081-3; Miura et al. Nat Biotechnol. 2020: 38, 1421-1430; Andersen et al. Cell. 2020: 183, 1913-1929. e26), while often not reaching single cell resolution.

[0106] In a preferred embodiment, the human cortico-striato-nigral circuit on chip comprises human induced pluripotent stem cells (hiPSCs). Said hiPSCs allow for easy obtainable, uniform cell cultures, which allow optimization of the human cortico-striato-nigral circuit on chip to a specific need.

[0107] In a preferred embodiment, said hiPSC are derived from human fibroblasts, skin cells and / or blood cells. In an embodiment, the hiPSCs can be derived from other somatic cells.

[0108] In an embodiment, said human pluripotent stem cell is derived from a human embryo.

[0109] Preferably, said human pluripotent stem cells is derived from a single human subject. This may allow for monitoring striatal function and / or dysfunction in a subject specific manner, identification for a new, patient-population specific biomarkers, patient stratification and / or proposing patient specific therapy. Preferably, the human cortico-striato-nigral circuit is provided on the chip which is a high density MEA chip, preferably featuring a plurality of electrodes wherein at least one of electrodes is arranged to detect a single neuron electrophysiological activity. In some embodiments, multiple electrodes are arranged to detect a single neuron's electrophysiological activity.

[0110] Preferably, said chip is a high density MEA chip, featuring a plurality of electrodes arranged to detect electrophysiological activity of said circuits i-iii). Preferably the chip used in some embodiment of the first aspect allows for individual measuring of the electrophysiological activity, but also a neural network activity, i.e. the circuit neural activity of medium spiny, nigral dopaminergic and cortical neural networks. Preferably said high density MEA chip allows for data analysis strategies that can fully capture complexities and / or nuances associated with disease-related neuronal dysfunctions

[0111] In embodiments, the multi-electrode array chip may be a complementary-metal-oxide-semiconductor multi-electrode array chip. This may allow for a high-density, high-resolution platform for recording neuronal activity.

[0112] In embodiments, the multi-electrode array may be adapted for single-cell level recordings of neuronal activity. This allows for the precise characterization of individual neuronal responses.

[0113] In embodiments, the multi-electrode array chip may have an electrode density exceeding 1000 electrodes per mm2. This enables high-resolution mapping of neuronal activity across the circuit.

[0114] In embodiments, the multi-electrode array chip features electrodes which may be individually addressable. This allows for targeted stimulation and recording from specific regions of the neuronal circuit.PaVer / CSNOC / 870

[0115] In embodiments, the electrode density of the multi-electrode array may be such that each cell, for example, a neuron is sensed across a plurality of electrodes simultaneously. Preferably, each cell is sensed across 3 to 5 electrodes. This multi-electrode sensing is advantageous, as it may provide redundant measurements of individual cell activity, reducing the impact of noise or electrode failure on signal quality. It may facilitate single unit reconstruction by providing multiple spatially distinct measurements of the same cell's activity, which can be used to distinguish between signals from different cells based on their spatial profiles. Furthermore, it may enable spatial localization of individual cells within the neuronal circuit, which may be correlated with the defined architecture of the circuit. High-density complementary-metal-oxide-semiconductor multi-electrode arrays, such as those with inter-electrode spacing of approximately 17.5 pm or less, are particularly suited for achieving multielectrode sensing of individual neurons.

[0116] In embodiments, the multi-electrode array electrodes may have their largest lateral dimension measuring from 2 to 500 pm, preferably from 2 to 50 pm, more preferably form 2 to 25 pm. This enables the recording of neuronal activity at different spatial scales.

[0117] In embodiments, the electrophysiological features may comprise cell-level features derived from action potential waveform characteristics. This allows for the characterization of individual neuronal firing properties.

[0118] In embodiments, the electrophysiological features may include circuit-level features derived from spike train data. This enables the analysis of network-level activity patterns.

[0119] In embodiments, the electrophysiological features may be selected from network synchrony, rhythmicity, connectivity, and firing patterns. This captures key aspects of neuronal circuit dynamics. In embodiments, the electrophysiological features may be derived using a spike sorting algorithm to decompose the electrophysiological recording data into contributions from individual neurons. This enables the isolation of single-unit activity from the multi-electrode array recordings.

[0120] In embodiments, the electrophysiological features may be derived using a biophysical model-based fitting process to determine activation timepoints. This allows for the precise estimation of neuronal firing times.

[0121] In embodiments, the confounding factors may comprise batch-specific variability and / or experimental covariates. This enables accounting for potential sources of experimental noise or bias.

[0122] In embodiments, the performance of comparing the associated data with at least one reference associated data and determining whether the reference electrophysiological feature, said at least one of the derived one or more cell-level electrophysiological features and / or one or more circuit-level electrophysiological features is a biomarker specific for the health condition may comprise using mutualPaVer / CSNOC / 870

[0123] nearest neighbour models. This provides a robust approach for identifying biomarkers from highdimensional data.

[0124] In an embodiment, said MSNs further express FOXP1. This confirms accurate MSNs specification.

[0125] In a preferred embodiment, said MSNs express dopaminergic DI receptor which further express PDYN or TSHZ1. This may allow testing specific DI receptor agonists or antagonists in the context of the disorder or disease linked with the striatal output. Surprisingly, the human cortico-striato-nigral circuit on chip of the invention may allow for observing of DI receptor antagonism resulting in a mild but significant reduction of the CSN circuit's firing frequency, in line with the previously reported role of dopamine as a modulator of neuronal excitability in cortical and medium spiny neurons (Lahiri & Bevan, Neuron. 2020: 106, 277-290. e6; Yano et al. Nat Common. 2018: 9, 486). Also, blockade of DI receptors increased circuit synchronization and neuronal coupling, which is reminiscent of what is observed in individuals with PD and believed to occur as a result of a decrease in brain dopamine levels (Ortone et al. PLoS Comput Biol. 2023: 19, el010645). Preferably, human cortico-striato-nigral circuit on chip of the invention provides a model which may recapitulate pathology linked to a striatal output and / or can be used for biomarker and therapeutic search for disorders such as PD.

[0126] In an embodiment, said population of cortical glutamatergic neurons comprises at least a dorsal forebrain (DF) cluster reminiscent of deeper layers of cortical neurons expressing Layer IV marker FOXP2 and / or layer V / VI marker CTIP2 and / or upper cortical neurons expressing layers ll / lll markers such as SATB2. This may allow obtaining more mature and / or human-brain like cortical circuits. This may also enable to recapitulate cortico-striatal connectivity observed in the brain.

[0127] In an embodiment, said population of dopaminergic neurons further express KCNJ6 or SOX6, which markers define the nigral neurons that are the most susceptible to PD.

[0128] In a preferred embodiment, said at least one of populations i)-iii) comprises at least one allele or genetic event associated with a neurologic or psychiatric disorder. The presence of said allele or genetic event allows for generating a disease model associated with a neurologic or psychiatric disorder which could be used for defining a biomarkers, as well as therapeutic agents for said neurologic or psychiatric disorder.

[0129] In a preferred embodiment, said at least one of populations i)-iii) comprises at least one mutant allele, preferably an allele comprising an isogenic mutation. Said isogenic mutation may be linked to symptoms of a disease or a disorder, preferably Parkinson's Disease or Huntington's disease. Preferably, the presence of said mutant allele may be linked to motor symptoms of a disease or disorder. In an embodiment said mutant allele is an LRRK2 or GBA mutant allele.PaVer / CSNOC / 870

[0130] In a preferred embodiment, said neurologic or psychiatric disorder is selected from the group consisting of: Parkinson's disease, Huntington's disease, Multiple System Atrophy (MSA), dystonia syndromes, motor dyskinesia, Alzheimer disease, amyotrophic lateral sclerosis (ALS), Attention deficit hyperactivity disorder (ADHD), schizophrenia, affective disorder and autism spectrum disorder (ASD).

[0131] In a preferred embodiment said medium spiny neuronal circuit i) is spatially arranged on the MEA chip to receive excitatory stimuli form said cortical circuit iii) and modulatory stimuli from said nigral dopaminergic circuit ii).

[0132] In a preferred embodiment said circuits i-iii) are arranged so that the electrophysiological activities thereof is synchronised with glutamate.

[0133] In the human cortico-striato-nigral circuit on chip according to the present invention, MSNs receive abundant connections from both dopamine and cortical neurons and that these neurotransmitters have excitatory and neuromodulatory effects on these neurons. This may allow for recapitulating the disorders in neuronal activity from single neuron to a circuit e.g., neural network scale, which allows for monitoring early symptoms and / or markers of striatal disfunction, as well as testing for new agents active in early stages of neurodegenerative disorders connected with striatal disfunction.

[0134] In a preferred embodiment, said circuits i-iii) are arranged so that an electrophysiological activity of at least circuit i), circuit ii) or circuit iii) is disrupted by alpha-synuclein. Introducing Parkinson-relevant alpha-synuclein fibrils into the circuit of the human cortico-striato-nigral circuit on chip of the invention leads to observable changes in network activity and synchronization while sparing single-neuron firing patterns, simulating disease-like conditions. This allows using complex circuit parameters when modelling neurodegenerative diseases, such as PD and may allow for observing network defects as early biomarkers in PD. Alpha-synuclein fibrils preferentially impact network activity and circuit synchronization while leaving spike waveform and single-cell electrophysiological features unaltered. The human cortico-striato-nigral circuit on chip according to the present invention revealed that network dysfunction might be one of the earliest disease phenotypes and underscores the need to investigate PD pathogenesis with a circuit perspective.

[0135] Preferably, said circuits i-iii) are arranged so that an electrophysiological activity of any one of said circuits i-iii) can be measured for up to five months. The human cortico-striato-nigral circuit on chip according to the present invention allows for longitudinal study in order to define biomarkers and / or therapeutic agents relevant for diseases and disorders linked to striatal disfunction.PaVer / CSNOC / 870

[0136] In a second aspect, the present invention relates to a method for producing a cortico-striato-nigral circuit on chip according to any one of the above-mentioned embodiments of the first aspect, the method comprising:

[0137] (a) providing (i) a lateral ganglionic eminence progenitor cell from a first human pluripotent stem cell; (ii) a ventral midbrain progenitor from a second human pluripotent stem cell; and (iii) a dorsal forebrain progenitor cell from a third human pluripotent stem cell;

[0138] (b) seeding the lateral ganglionic eminence (i), ventral midbrain (ii), and dorsal forebrain (iii) progenitor cells at three different locations of the MEA chip, so that the lateral ganglionic eminence (i) progenitor cell is spatially arranged onto said MEA chip for receiving excitatory and / or modulatory inputs from dorsal forebrain iii) progenitor cell and / or ventral midbrain (ii) progenitor cell;

[0139] (c) culturing the lateral ganglionic eminence (i), ventral midbrain (ii) and dorsal forebrain (iii) progenitor cells in conditions allowing for formation of differentiated, matured populations of medium spiny (i), nigral dopaminergic (ii) and cortical glutamatergic (iii) neurons; and / or in conditions allowing for synaptic activity between the cells.

[0140] In some embodiments of the present invention, terms "dorsal forebrain progenitors" and "cortical progenitors" are used interchangeably.

[0141] In some embodiments of the present invention, terms "lateral ganglionic eminence progenitors" and "striatal progenitors" are used interchangeably.

[0142] In some embodiments of the present invention, terms "ventral midbrain progenitors" and "nigral progenitors" are used interchangeably.

[0143] Preferably, said first, second and third human pluripotent stem cell are from the same subject, thereby allowing patient-specific biomarker characterisation and / or therapy optimisation.

[0144] Preferably, the step of c) culturing is performed for at least 70 days, preferably at least 75, 80 or 85 days, thereby allowing formation of functional differentiated striatal (i), nigral dopaminergic (ii) and cortical (iii) circuits. In embodiments, said culturing is performed for at least 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99,100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130 days, or. more. By culturing time, it is considered the number of days from the hPSC (human pluripotent stem cells) to developed nigral dopaminergic, cortical and / or striatal neurons exerting neurophysiological activity. Said culturing time for striatal neurons may be shorter, for example for 5, 6, 7, 8, 9, or 10 days.PaVer / CSNOC / 870

[0145] In an embodiment, said culturing may comprise the following steps: a) pre-differentiating of the cells from hPSC, preferably hiPSC until the obtaining of dorsal forebrain, lateral ganglionic eminence and / or ventral midbrain progenitor cells and cryopreserving of said cells; b) thawing the cryopreserved progenitor cells and seeding them on the MEA chip; c) growing the seeded progenitor cells on the chip to obtain functional differentiated striatal (i), nigral dopaminergic (ii) and / or cortical (iii) circuits. In a preferred embodiment, the step a) of pre-differentiating the cells from human pluripotent stem cells until obtaining of the dorsal forebrain and / or ventral midbrain progenitor cells and cryopreserving thereof is done for between 20 and 34 days, preferably for 21, 22, 23, 24, 25, 26, 1 , 28, 29, 30, 31, 32, or 33 days. In a preferred embodiment, the step a) pre-differentiating the cells from human pluripotent stem cells until obtaining of the lateral ganglionic eminence progenitors and cryopreserving thereof may last for between 12 and 26 days, preferably, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, and 25 days. The step of b) thawing may take for about 1, 2, 3, 4, 5 or 6 days. The step c) growing the seeded progenitor cells on the chip to obtain functional differentiated striatal (i), nigral dopaminergic (ii) and / or cortical (iii) circuits may last for up to 5 months, as shown in the following examples. In some embodiments, "DIV" or "days in vitro culture" are used interchangeably and refer to the total number of days spent on steps a) pre-differentiating and cryopreserving of the dorsal forebrain and / or ventral midbrain progenitor cells, and b) thawing and c) growing the seeded cells.

[0146] In an embodiment, within said differentiated, matured populations of medium spiny (i), nigral dopaminergic (ii) and cortical (iii) neurons comprise at least one allele or genetic event can be associated with a neurologic or psychiatric disorder.

[0147] Preferably, the neurologic or psychiatric disorder is selected from the group consisting of: Parkinson's disease, Huntington's disease, Multiple System Atrophy (MSA), dystonia syndromes, motor dyskinesia, Alzheimer disease, amyotrophic lateral sclerosis (ALS), Attention deficit hyperactivity disorder (ADHD), schizophrenia, affective disorder and autism spectrum disorder (ASD).

[0148] In a third independent aspect, the invention pertains to a method of determining the effect of a candidate agent on neuronal activity of the cortico-striato-nigral circuits on chip according to any one of above-mentioned embodiments of a first aspect or obtainable by the method according to any one of the above-mentioned embodiments of a second aspect, the method comprising:

[0149] a) contacting the candidate agent with cortico-striato-nigral circuits on chip according to any one of embodiments of a first aspect or obtainable by the method according to any one of embodiments according to a second aspect;

[0150] b) observing the change of in a neuronal activity and / or function of circuits i)-ii) and iii), c) determining the effect of the candidate agent based on resulting change in the neuronal activity and / or function of circuits i)-ii) and iii).PaVer / CSNOC / 870

[0151] In a preferred embodiment, said change in the neuronal activity is resulting at least in part from dopamine modulation. This allows for a screening for drug candidates linked to altered dopaminergic neurotransmission.

[0152] In a preferred embodiment, said candidate agent is tested for use in a neurologic or psychiatric disorder, further optionally wherein the neurologic or psychiatric disorder is selected from the group consisting of: Parkinson's disease, Huntington's disease, Multiple System Atrophy (MSA), dystonia syndromes, motor dyskinesia, Alzheimer disease, amyotrophic lateral sclerosis (ALS), Attention deficit hyperactivity disorder (ADHD), schizophrenia, affective disorder and autism spectrum disorder (ASD).

[0153] In a preferred embodiment said candidate agent is a selective DI agonist. In an embodiment, said candidate agent is a DI antagonist, D2-agonist and / or D2-antagonist.

[0154] In a preferred embodiment, the method of determining the effect of a candidate agent on neuronal activity of the cortico-striato-nigral circuits on chip the effect of a candidate agent on neuronal activity is measured over several time-points and / or a time interval. The method allows for longitudinal study of the effects of various candidate agents on human cortico-striato-nigral circuits, measured as an electrophysiological outputs obtainable by MEA chip recordings.

[0155] In a fourth independent aspect, the present invention relates to a method to detect a genetic mutation or a cell state, which mutation or cell state may related to a disease and / or disorder, the method comprising the steps of:

[0156] providing the human cortico-striato-nigral circuit on a chip of the invention, or obtainable by the method of the invention;

[0157] recording the electrophysiological activity of at least one circuit of the human cortico-striato-nigral circuit on a chip of the invention, or obtainable by the method of the invention; and

[0158] using a computer implemented method to associate the recorded electrophysiological activity with presence of said genetic mutation or cell state.

[0159] The method according to the forth aspect allows for the detection of different molecular alterations which may lead to neuronal network defects. Said molecular alterations may correspond to symptomatic or asymptomatic diseases and / or disorders, for example said molecular alterations may correspond to motor symptoms of Parkinson's disease. Provision of the method according to the fourth aspect is particularly favourable in case wherein the symptoms do not reveal the underlying molecular etiology. The measured electrophysiological activity of at least one human circuit of the human CSN circuit on a chip of the invention may encode characteristic functional signatures, preferably on a circuit level, which functional signatures may enable stratification of the disease and / or proposing mechanism-specific therapies. The method according to the fourth aspect may allow provision of a mechanism-specificPaVer / CSNOC / 870

[0160] therapy even in case of absent, unclear and / or overlapping clinical symptoms of a disease and / or a disorder. Advantageously, the method of the fourth aspect enables to stratify a disease and / or a disorder beyond clinical symptomatology. The method according to the fourth aspect can be leveraged to stratify patient subgroups to enrich and benchmark clinical trials and to quantify drug efficacy.

[0161] In embodiments, the recorded electrophysiological activity is derived from one or more cell-level electrophysiological features and / or one or more circuit-level electrophysiological features. Said one or more cell-level electrophysiological features may comprise at least one of: overall firing rate, burst rate, non-burst firing rate, spike waveform characteristics, and interspike interval distributions for individual electrogenic cells. In embodiments, deriving one or more circuit-level electrophysiological features may comprise deriving at least one of: spike train correlations, population synchrony measures, population burst rate, cross-population synchronization between distinct neuronal subtypes, inter-layer connectivity measures, and layer-specific firing patterns. The combination of cell-level and circuit-level features provides a comprehensive characterization of neuronal circuit function.

[0162] In an embodiment, the method according to the fourth independent aspect comprises additional step of comparing the recorded electrophysiological activity with at least one reference activity and determining whether the recorded activity significantly differs from said reference activity. Said step may allow for a more precise associating of one or more cell-level electrophysiological features and / or one or more circuit-level electrophysiological features with a certain genetic mutation and / or a cell state, for example a healthy or diseased cell state.

[0163] In an embodiment, the method according to the fourth aspect may comprise obtaining longitudinal recordings of the electrophysiological activity of at least one circuit of the human cortico-striato-nigral circuit on a chip of the invention, or obtainable by the method of the invention, for example recordings at multiple time points during the maturation or development of the circuit in vitro. For example, recordings may be obtained at multiple days in vitro (DIV), such as at 95 DIV or at regular intervals over weeks or months. Longitudinal recordings enable tracking of temporal evolution of electrophysiological features during circuit development.

[0164] In some embodiments, the method according to a fourth independent aspect allows detecting a cell state. Said cell state may be connected to a physiological cell state, i.e., developmental stage, cell type and / or differentiation stage, and / or a cell reprogramming stage.

[0165] Said cell state may be connected to a pathological and / or diseased cell state, which may be used to follow up a disease progression. Advantageously, the method allows for obtaining the important information about physiological and / or pathophysiological condition of a cell, based on the electrophysiological recordings using the human cortico-striato-nigral circuit on a chip of the invention, or obtainable the method according to any embodiment of the second aspect of the invention. ThisPaVer / CSNOC / 870

[0166] method may allow for better monitoring and / or diagnostic of a physiological and / or pathophysiological condition in a cell, thereby allowing detecting a disease, and / or disease progression. In some embodiments, the method of according to the fourth aspect may be used in a disease and / or disorder stratification, for example, when said stratification is linked to a genetic background. In some embodiments, the method according to a fourth aspect allows detecting the genetic and / or molecular background related to a disease and / or disorder connected to a cell state.

[0167] In some embodiments, the method of the fourth aspect allows for recording of a electrophysiological activity in the presence of an agent. Said agent may be a pharmacological modality and / or therapeutic agent.

[0168] In some embodiments, the method of the fourth aspect may be used in drug testing and / or screening. Preferably, the computer implemented method used in the method according to the fourth aspect is a machine learning (ML) classifier model. Preferably, said ML classifier model is trained to classify the mutant and control circuits with high confidence and / or is trained for distinguishing between the disease-related mutants based on unbiased electrophysiological screening. The method according to the fourth aspect preferably combines stem cell-based circuit models with computer implemented, preferably Al-powered computational approaches, enables to connect molecular etiologies of disease with the circuit dysfunction.

[0169] In an embodiment, the method may comprise using linear mixed-effects models (LMM) to analyze electrophysiological features across a genetic mutation and / or a cell state over time. The linear mixed-effects models may be fitted to features defined per layer of the neuronal circuit to identify which neuronal subtypes show significant changes in a genetic mutation and / or a cell state.

[0170] In an embodiment, the method may further comprise performing post-hoc interpretability analysis on the trained machine learning classifier. SHapley Additive exPlanations (SHAP) analysis may be performed to identify which electrophysiological features have highest discriminative power for distinguishing between a genetic mutation and / or a cell state. Such analysis may reveal that network-level parameters, such as network burst rate and network burst duration, have greater discriminative power than cell-level parameters, enabling identification of circuit-level functional signatures characteristic of specific genetic mutation and / or a cell state.

[0171] In embodiments, the method according to a fourth aspect may include a machine learning classifier, for example a random forest classifier. Training may comprise group K-fold cross validation. This avoids information leakage between training and validation sets. Where class imbalance exists, synthetic minority oversampling technique (SMOTE) may be used. Preferably, said machine learning classifier allows predictive modelling which can reveal mutation- and / or state- specific electrophysiological

[0172] 1PaVer / CSNOC / 870

[0173] signatures, enabling at least 55%, 60%, 65%, 70%, 75%, 80%, 81%, 82%, 83%, 84%, 85%, 86%, 87%, 88%, 89% or 95% accurate classification.

[0174] In embodiment, the classification may further incorporate genetic mutation- and / or cell state-specific phenotypes obtained from clinical observations, enabling to link electrophysiological signatures with clinical manifestations. This approach may allows linking the recorded in vitro electrophysiological characteristics and the in vivo clinical phenotypes associated with the health condition.

[0175] In an embodiment, the circuits of the CSN circuit on chip of the invention are isogenic circuits.

[0176] In an embodiment, said circuits preferably carry Parkinson's disease relevant mutant circuits. Some nonlimiting examples for such mutants are LRRK2 and GBA mutations associated with PD disease. It is surprisingly shown that low penetrance, Parkinson's disease mutations, such as LRRK2 and GBA display early and distinct electrophysiological defects in CSN circuits corresponding thereto. Studies in nonmanifesting individuals carrying pathogenic mutations in the two most-commonly mutated genes in Parkinson's Disease (LRRK2 and GBA) reveal subtle motor impairments 4-6 years before clinical symptom onset, indicative of compromised nigrostriatal circuit function. The method of the fourth aspect may be particularly advantageous to diagnose and / or stratify patients with or without these mutations who do not show clinical symptoms of a disease or disorder yet(i.e., presymptomatic or prodromal individuals), for example Parkinson's disease.

[0177] In a preferred embodiment, the circuits on the chip utilised in a method according to the fourth aspect are constructed using hiPSCs from a single subject. Advantageously, such method may allow early and accurate diagnosis of a disease and / or disorder, for example in cases where symptoms are unclear or not present in a subject.

[0178] The following non-limiting Examples describe methods and means according to the invention. Unless stated otherwise in the Examples, all techniques are carried out according to protocols standard in the art. The following examples are included to illustrate embodiments of the invention. Those of skill in the art should, in light of the present disclosure, appreciate that many changes can be made in the specific embodiments which are disclosed and still obtain a like or similar result without departing from the concept, spirit and scope of the invention. More specifically, it will be apparent that certain chemical analogues and / or derivatives of the compounds may be substituted for the agents described herein while the same or similar results would be achieved. All such similar substitutes and modifications apparent to those skilled in the art are deemed to be within the spirit, scope and concept of the invention as defined by the appended claims.

[0179] EXAMPLES

[0180] Table 1 explains the meaning of the terms used in the Experimental section of the present invention.PaVer / CSNOC / 870

[0181] Table 1. The explanation of the terms used in Experiments 1-9.

[0182] Metric Description

[0183] Culture Age Days from the iPSC stage for cortical and ventral midbrain neurons

[0184] Cell Line hiPSC line

[0185] AB Ratio The average ratio of the PtP-to-PtP identified for a single unit

[0186] ACG Asymptote The asymptote of the fit to the ACG identified for a single unit

[0187] ACG C an amplitude fit to the ACG identified for a single unit

[0188] ACG D an amplitude fit to the ACG identified for a single unit

[0189] ACG H the goodness of the fit to the ACG identified for a single unit

[0190] ACG Refractory Period an amplitude fit to the ACG identified for a single unit

[0191] ACG Tau Burst Period the refractory period (in ms) derived from the fit to the ACG identified for a single unit

[0192] ACG Tau Decay the tau burst period (in ms) derived from the fit to the ACG identified for a single unit

[0193] ACG Tau Rise the tau decay period (in ms) derived from the fit to the ACG identified for a single unit

[0194] Avg. PtP Voltage (Amplitude) the tau rise period (in ms) derived from the fit to the ACG identified for a single unit

[0195] Exponential Fit Avg. PtP Voltage a fitted value through the average PtP amplitude of the filtered waveform identified for a single unit Exponential Fit Length Constant Avg. PtP Voltage a fitted value through the average PtP amplitude of the filtered waveform identified for a single unit PtP Voltage Amplitude (PHY-derived) relative PtP waveform amplitude obtained from PHY Waveform Polarity waveform polarity identified for a single unit Refractory Period Violation refractory period violation (1 / 1000) defined as the units fraction of ISIs less then 2 ms

[0196] Avg. Waveform TtP Interval the average waveform trough to PtP interval in us Derivative of the Avg. Waveform TtP Interval the derivative of the average waveform trough to PtP interval in usPaVer / CSNOC / 870

[0197] Metric Description

[0198] Bursts per Minute (BPM) Burst per minute

[0199] Burst Spikes per Minute (BSPM) Burst spikes per minute

[0200] Burst Total Time (BTT) Burst Time to Total recording time: the ratio of the accumulated burst interval time identified for a single unit to the total recording time

[0201] Inter-spike Interval CV (ISICV) InterSpike Interval Coefficient of Variation: the ratio of the standard deviation to the average ISI identified for a single unit

[0202] Mean Inter-spike Interval (ISI M) Mean InterBurst Interval: the average time from the end of one burst to the beginning of the next burst identified for a single unit

[0203] Median Inter-Spike Interval (ISIM2) Median InterSpike Interval: the median time between spiking events identified for a single unit Inter-Spike Variance (ISIV) InterSpike Interval variance of ISI calculated for a single unit

[0204] Mean Burst Duration (MBD) Mean Burst Duration: the average time from the start to the end of a burst identified for a single unit Mean Spikes per Burst (MBS) Mean Burst Spikes: the average number of spikes in a burst identified for a single unit

[0205] Single-Cell Regularity Frequency (SCRF) Single-Cell Regularity Frequency: the frequency with the highest magnitude of the Fourier-transformed activity identified for a single unit

[0206] Single-Cell Regularity Magnitude (SCRM) Single-Cell Regularity Magnitude: the magnitude of the PtP frequency of the Fourier-transformed activity identified for a single unit

[0207] Burst Index Burst index calculated according to Mizuseki 2012 paper. The burst index refers to the fraction of spikes with a neighboring ISI < 6 ms

[0208] Firing Rate Burst index calculated according to Royer 2012 paper

[0209] Burst Firing Rate the number of spikes in a burst identified for a single unit normalized by the accumulated burst intervals Network Burst Firing Rate (or network burst rate): the number of spikes in a network burst identified for a single unit normalized by the accumulated network burst intervals Firing Rate CV the standard deviation of the single unit firing rate across time divided by the mean firing rate Firing Rate Gini Coefficient the Gini coefficient of the firing rate across timePaVer / CSNOC / 870

[0210] Metric Description

[0211] Firing Rate Inter-Spike Interval firing rate in Hz, spike count normalized by the interval between the first and the last spike Firing Rate Instability the average of the absolute differential firing rate across time divided by the mean firing rate Non-burst Firing Rate (or inter-burst rate) the number of spikes in the inter-burst phase identified for a single unit normalized by the accumulated inter-burst intervals Non-Network Burst Firing Rate (or network inter-burst rate): the number of spikes in the network inter-burst phase identified for a single unit normalized by the accumulated network inter-burst intervals

[0212] Firing Rate during Non-Synchronized Phase the number of spikes in the inter-synchronized phase normalized by the accumulated intersynchronized intervals.

[0213] Firing Rate during Synchronized Phase the number of spikes in a network synchronized phase normalized by the accumulated synchronized intervals. The synchronized phase is determined as the population firing rate (all units) exceeding the mean plus a factor (2) multiplied with the standard deviation on the population firing rate

[0214] Mean Network Burst Duration (MNBT) Sorry Carles forgot to add a description here Mean Neurons in a Network Burst (MNNB) Mean Neurons in a Network Burst: the average number of units participating in a network burst Mean Spikes per Network Burst (MSNB) Mean Spikes in a Network Burst: the average number of spikes from a single unit participating in a network burst

[0215] Network Bursts per Minute (NBPM) Network Bursts Per Minute: the rate of network burst events

[0216] Network Burst Spikes per Minute (NBSPM) Network Burst Spikes Per Minute: the rate of network burst spikes for a single unit participating in that network burst

[0217] Network Burst Total Time (NBTT) Network Burst Time to Total recording time: the ratio of the accumulated network burst interval time to the total recording time

[0218] Spike Contrast (SCt) Spike-Contrast: a time-scale independent measure of maximal synchronicity identified from the network activity for different bin sizes

[0219] Bin Size at Maximal Spike Contrast (SbinMax) Bin size at Maximal Spike-Contrast

[0220] Population Coupling Population Coupling: a measure of how correlated an individual neuron activity is with the rest of the population. It is dependent on the smoothed firing rates of the single units (Gaussian kernal of halfPaVer / CSNOC / 870

[0221] Metric Description

[0222] width 12 / sqrt(2) ms, its mean firing rate and the number of spikes fired (norm)

[0223] Incoming Synaptic Connections count of incoming synaptic connections

[0224] Outgoing Synaptic Connections count of outgoing synaptic connections

[0225] Spike Count Total number of spikes detected during the recording session

[0226] Example 1: Modular generation of cortical, lateral ganglionic eminence and caudal ventral 107 midbrain progenitor cells.

[0227] To create cortico-striato-nigral (CSN) circuits in vitro (Fig 1A) protocols to generate cortical, striatal and dopaminergic nigral neurons from hiPSCs were optimized. In the developing brain, these three neuronal subtypes originate from different regions. The dorsal forebrain produces cortical glutamatergic neurons (CN) that populate the cortex (Ctx), the lateral ganglionic eminence striatal (Str) medium spiny neurons (MSN) and the ventral midbrain (VM) substantia nigra dopaminergic neurons (DAN). Parallelized neuronal progenitor (NPC) production for the three cell types were facilitated by optimizing and standardizing different parameters across established protocols (Kriks, et al. Nature. 2011: 480, 547-51; Kirkeby, etal. Cell Stem Cell. 2017: 20, 135-148; Strano, et al., Cell Rep. 2020: 31, 107732; Miura, et al., Nat Biotechnol. 2020: 38). This enabled the generation of highly enriched populations of dorsal forebrain expressing FOXG1, PAX6 and TBR2 (Englund, et al The Journal of Neuroscience. 2005: 25, 247-251) of lateral ganglionic eminence expressing FOXG1, GSX2, and DLX2 (Wang, et al. Journal of Comparative Neurology. 2013: 521, 1561-1584; Zhao, Z. et al., Cell Res. 2022: 32, 425-436) and of ventral midbrain expressing FOXA2, LMX1A, EN1 and OTX2. There were also low numbers of other cell types, usually with an identity akin to those found in neighbouring developing brain regions (Strano, et al., Cell Rep. 2020: 31, 107732; Floruta, et al., Stem Cell Reports. 2017: 9). Expression of the early born neuron markers was also assessed: TBR1, which labels cortical neurons, and DLX2, which marks intermediate progenitors and early neurons from the ganglionic eminences (Long, et al. Journal of Comparative Neurology. 2009: 512, 556-572) in 30-35-day-old cultures and discarded cortical or striatal batches with more than 10% DLX2+ orTBRl+ cells, respectively. In summary, efficient quality-controlled protocols for the generation of N PCs from different human brain regions were established.

[0228] Example 2. A triple reporter system to visualize terminal differentiation.

[0229] To facilitate the monitoring of neurogenesis when building CSN circuits starting from a single iPSC line (SFC065), a "triple reporter system" was introduced. CRISPR-Cas9-dependent genome engineering was used to insert three cassettes at different locations in the same cell line: one encodes T2A-mClover3 andPaVer / CSNOC / 870

[0230] is inserted in the last exon of TH (expressed in DAN), another encodes T2A-mScarlet and is inserted in the last exon of GAD2 (GAD65; expressed in MSN) and a final one encodes T2A-CRE recombinase and is inserted in the last exon of SLC17A7 (VGLUT1; expressed in CN) (Fig. IB) to specifically activate CRE-responsive payloads in CN, including BFP2 (Fig. IB and C), TDsmURFP or optogenetic tools (see Fig. 4). Comparative genomic hybridization arrays was used to verify and confirm that these multiple rounds of genome engineering did not introduce genomic rearrangements. The fluorescent reporter system faithfully reports on cellular identity following differentiation. When these cells were differentiated into DAN and labeled with anti- TH, the mClover3-positive cells were found also TH-positive (Fig. 1C). When differentiated into MSN and labeled with anti-GAD65, mScarlet-positive cells were GAD65-positive (Fig.

[0231] 1C'). Finally, when the neurons were transduced with a CRE148 responsive BFP2 cassette (LV-hSyn::cFLEX-BFP2), the cells differentiated into CN and labeled with anti-VGLUTl, all BFP2-labeled cells (Fig. 1C"). This reporter system also enabled us to assess cell morphology showing differences between the three neuronal subtypes: CN had larger and MSN smaller soma, in line with their previously reported morphologies (Braak & Braak, Cell Tissue Res. 1982: 227, 319-342; Sal'kov, et al., Neurosci Behav Physiol.

[0232] 2017: 47, 366-369; Petanjek, Cerebral Cortex. 2008: 18, 915-929; Rajkowska, et al., Arch Gen Psychiatry.

[0233] 1998: 55, 215). This latter observation opens the possibility of harnessing morphological differences among neurons to determine their identity.

[0234] Example 3. Building human cortico-striato-nigral circuits on HD-MEA arrays.

[0235] To create integrated CSN circuits quality-controlled cortical, striatal and ventral midbrain neural progenitors obtained from the "triple reporter" hiPSC were seeded at three specific locations on a HD-MEA with the help of custom-built easy-to-use detachable silicone inserts (Fig. ID). Immunofluorescence labeling one day after plating with specific progenitor markers for each of the NPC-subtypes showed that cells specifically occupied the three defined regions, indicating separated deposition of the NPCs. The cultures were then treated with DAPT (N-[N-(3,5-difluorophenacetyl)-L-alanyl]-S-phenylglycine t-butyl ester, a y-secretase inhibitor) to accelerate neuronal differentiation and 5-Fluoro-2'-deoxyuridine to stop cellular division. Primary rat astrocytes were placed on top of the co-cultures (10 days post-seeding) facilitating the development of active synaptic contacts. Live fluorescent imaging at >30 days of differentiation on the HD-MEAs showed the expected lineage-specific reporter expression in each of the three areas (Fig. IE) and also dense neuritic invasion across the HD-MEA (Fig. 1F-F"). GAD2-positive MSN were surrounded by TH-positive and VGLUTl-positive neurites (Fig. 1G). It was feasible to maintain these cultures for up to 5 months, without substantial cell loss or adverse effects on viability. The identity of the differentiated cells was verified in the CSN circuits. Single nuclei sequencing analysis performed in 100-day old circuits revealed the presence of the three major neural lineages: dorsal forebrain, lateralPaVer / CSNOC / 870

[0236] ganglionic eminence and ventral midbrain (Fig. 1H and I). The three major lineages were present in similar proportions in CSN circuits obtained from 2 independent iPSC lines, indicating the robustness and reproducibility of the differentiation protocols. Cells in the dorsal forebrain clusters expressed markers of cortical excitatory neurons such as LHX2 and SLC17A6 and / or SLC17A7. Moreover, at this stage some clusters reminiscent of deeper layer cortical neurons could be found including a cluster expressing Layer VI marker FOXP2 (cluster 7), while others express markers of upper cortical neurons such as SATB2 (cluster 23) (Fig. 4A, D and E'). Neurons in the ventral midbrain layer expressed the typical ventral midbrain DAN makers FOXA1, LMX1A, EN1 and NR4A2 together with other markers specific to DA metabolism and neurotransmission (Fig. II) (Arenas, et al., Development. 2015: 142, 1918-1936). Interestingly, most of the DAN also expressed KCNJ6 (GIRK2) and SOX6 that are both highly expressed in the ventral tier of the SNpc and define the most vulnerable subset of ventral midbrain DAN in PD (Pereira Luppi, et al. Cell Rep. 2021: 37, 109975; Kamath, et al. Nat Neurosci. 2022: 25, 588-595; Brichta & Greengard, Front Neuroanat.2014: 8). Of note, neurons expressing markers of the subthalamic nucleus (STN) (e.g. PITX2) were not found, which is a common contaminant in iPSC-derived ventral midbrain preparations (Kirkeby et al. Cell Stem Cell. 2017: 20, 135-148), further indicating our optimized differentiation protocol produces the desired midbrain neurons on the HD-MEA. Lateral ganglionic eminence neurons expressed MEIS2, the GABAergic lineage markers GAD2 and DLX6-AS1 (Fig. II), and within the lateral ganglionic eminence, MSNs could be identified by the co-expression of FOXP1 and BCL11B (Fig. II). These cells could also be divided further into three defined major subgroups that corresponded to the known MSN-subtypes expressing DI or D2 dopamine receptors that give rise to the direct and indirect pathways, respectively: The two types of Dl-expressing MSNs were identified: MSNs that express either PDYN (Dla) or TSHZ1 (Dlb) and the D2-expressing MSN (Fig.lJ and K) (Zhao et al. Cell Res. 2022: 32, 425-436; Shi etal. Science. 2021: 374, eabj6641; Xiao etal. Cell. 2020: 183, 211-227.e20). TSHZ1+ / PDYN- Dl-MSNs are known to be present in the striatum (Xiao et al. Cell. 2020: 183, 211-227. e20) as well as in the amygdala (Kuerbitz et al. Cerebral Cortex. 2021: 31, 1744-1762) of adult rodents. In contrast to previous reports describing the generation of lateral ganglionic eminence-derived populations from human stem cells (Reumann et al. Nat Methods: 2023: 20, 2034-2047), our protocol resulted in fewer undesired cells expressing markers specific to the medial (LHX6+, LHX8+ and / or NKX2-1+; less than 2% of cells expressing GABAergic markers) and caudal ganglionic eminences (NR2F1 / 2+; less than 3%). The effect of culturing MSNs in CSN circuits where two of their cognate presynaptic counterparts are present was then assessed versus culturing MSN of identical age and differentiation batch in isolation (Str-only). First, it was noticed that the relative proportion of cells belonging to the different MSN clusters differed between CSN circuits and Str-only. Striatal cells in circuits were enriched in those cell clusters showing the highest expression of key markers of the three MSN-subtypes (Fig. ILPaVer / CSNOC / 870

[0237] and M). Importantly, the relative proportion between these MSN subtypes, with similar numbers of Dla and D2 and less of Dlb, resembled that of the developing human brain much better than when MSN were produced in isolation (Str-only) (Shi et al. Science. 2021: 374, eabj6641). Second, differential gene expression testing of MSN in CSN circuits compared to Str-only revealed an enrichment of genes categorized by gene ontologies (GOs) like "synapse", "neuron projection" or "cell communication" across all three MSN lineages (Fig. IN;) or "glutamatergic synapse" along with many other synapse related GOs in the Dl-MSN and D2-MSNs clusters. Last, striatal MSN are located in striosomes or in the matrix of the Str (Graybiel & Ragsdale. PNAS. 1978: 75, 5723-5726). Neurons in these regions have specific projection patterns (e.g., striosomal PDYN+ Dla-MSN project to the substantia nigra (Crittenden & Graybiel, Front Neuroanat. 2011: 5; McGregor et al. Cell Rep. 2019: 29, 1419-1428. e5) and they have specific expression profiles ((Matsushima, et al. Nat Commun. 2023: 14, 282; He, et al. Current Biology.

[0238] 2021: 31, 5473-5486. e6). In co-cultured MSN the upregulation of several genes associated with the striosome compartment in PDYN+ Dla-MSN such as GPC6 and NETO1 (and several others) was noted (Matsushima, et al. Nat Commun. 2023: 14, 282). Gene set enrichment analysis on the DEGs of each of the MSN subtypes shows there is increased expression of striosomal genes in Dla- and DlbMSN (the latter are only found in striosomes (Xiao et al. Cell. 2020: 183, 211-227. e20); and both striosomal and matrix genes in D2-MSN. Taken together, MSN in our CSN circuits were differentiated into three classes of dopamine receptor-expressing neurons. These MSNs were more mature when grown in circuits versus in isolation, and the cell populations resembled those found in the human brain, including that they expressed transcriptomic features of their known subdivision into striosome and matrix compartments.

[0239] Example 4. MSN receive presynaptic input from DAN and CN.

[0240] To assess connectivity of MSN the CSN circuits of the invention were labelled with VGLUT1 and VMAT2 to visualize presynaptic terminals of CN and DAN and with DARPP32, to label the cytoplasm of the MSN. Structural illumination super resolution microscopy (SIM) imaging revealed numerous glutamatergic and dopaminergic presynaptic terminals in close proximity to MSN neurites (Fig. 2A). It was then assessed if neurons are morphologically connected at the ultrastructural level with correlative light and electron microscopy. The triple reporter CSN circuits were used and imaged areas in the mScarlet red fluorescent MSN layer, that also had mClover3 (DAN) and BFP2 (CN) positive neurites. These same areas were then identified in TEM images based on cellular morphology. Overlaying the TEM micrographs with the fluorescent images identified areas of synaptic connectivity between the 3 cell types (Fig 2B-E). TEM micrographs showed numerous typical synaptic connections with pre- and post-synaptic specializations and synaptic vesicles in the BFP2 and mClover3 CN and DAN terminals (Fig. 2B-D). These data indicated the existence of physical synaptic contacts of CN and DAN onto MSN in our CSN circuits. To furtherPaVer / CSNOC / 870

[0241] establish that MSN are in close physical contact with CN and DAN, rabies tracing (Zampieri et al. Neuron: 2014, 81, 766-778) was used. This system involves a modified rabies virus that uses a specific receptor (avian leukosis and sarcoma virus subgroup A or TVA) that was expressed in MSN, enabling to specifically infect these neurons. The virus (and its payload) in the infected MSN then transfers trans-synaptically using an optimized G protein (oG, also expressed in MSN) to presynaptic neurons (CN and DAN), activating a reporter (Reardon et al. Neuron. 2016: 89, 711-724). To target TVA and oG to LGE progenitors (that give rise to MSN) the FLP recombinase-controlled FLEX system was used (Atasoy et al. The Journal of Neuroscience. 2008: 28, 7025-7030). Using lentiviral vectors, FLPase (FLPo) was delivered under the control of the GABA lineage-specific Dlx5 / 6 enhancer (Dimidschstein etal. Nat Neurosci. 2016: 19, 1743-1749) together with a FLPase-responsive lentiviral vector that encodes TVA and oG (hSynl::fFLEX- mClover3-TVA-oG). Together this system causes controlled TVA and oG expression in MSN. The CSN circuits with these infected LGE progenitors and with uninfected DF and VM progenitors were created. 100-day-old circuits were incubated with envelope protein of the avian sarcoma and leukosis virus (EnvA)-coated tdTomato-encoding recombinant rabies virus that recognizes the TVA receptor on MSNs (Fig. 2F-H). The cultures were labelled with cell-specific markers and found numerous TH-labelled DAN that then expressed the rabies virus-derived tdTomato (Fig. 21-J). Likewise, many CN, labeled by Ctx layer V marker BCL11B (also known as CTIP2) or by Ctx layer Il-Ill marker SATB2, were found to express virus-derived tdTomato (Fig. 2K-L). Note that specifically Ctx layer V and Ctx layer Il-Ill neurons are known to innervate the Str and this is partly recapitulated in the system, i.e. the circuit of the present invention (Rosell & Gimenez-Amaya, Neurosci Res. 1999: 34, 257-269). These data indicate that MSN form physical synaptic contacts with DAN and CN in our CSN circuits as they do in the human brain (Haber, J Chem Neuroanat. 2003: 26, 317-330).

[0242] Experiment 5. CSN circuits increase their synchronized activity during maturation and aging.

[0243] The MEA chips used herein are HD-MEAs containing 26.400 electrodes spaced 17.5 pm apart, enabling longitudinal single cell resolution recordings (Rosell & Gimenez-Amaya, Neurosci Res. 1999: 34, 257-269). The triple reporter fluorescence overlay onto the HD-MEA area showed good alignment of electrical activity with the areas populated by neurons (Fig. 3A-C). Waveform reconstruction revealed that every neuron is sensed across 3-5 electrodes (Fig. 3A-C), resulting in signal overlap and necessitating spike sorting for single neuron-centered analyses (Rosell & Gimenez-Amaya, Neurosci Res. 1999: 34, 257-269). Recordings at days in vitro (DIV) 95, showed neuronal activity in all three regions of the HD-MEA (Fig. 3D). These activities manifested as alternating phases of isolated neuronal firing interspersed with neuronal bursts (bouts of high-frequency firing). Often, these bursts were synchronized across neurons localized in the three zones, a feature that was named network bursts (NB; shown in Fig. 3D).PaVer / CSNOC / 870

[0244] The frequency of these network bursts increased over the approximately 100 days during which it was recorded from them (Fig. 3D, E). Also, as CSN circuits aged, they spent more time within this synchronized phase (Fig. 3F). Several metrics to investigate the CSN circuits' degree of synchronization (Cutts & Eglen, The Journal of Neuroscience. 2014: 34, 14288-14303) were considered. Spike-time tilling coefficient (STTC) did not show a steady increase over time (Fig. 3G), but this synchronization metric does not account for the accumulated time the circuit is actively firing synchronously (as Fig. 3F does). Alternatively, neuronal (or population) coupling, a measure of how closely the firing of single neurons resembles that of the entire network (Okun et al. Nature. 2015: 521, 511-515), registered a gradual increase over time (Fig. 3H). At the single neuron level, an increase in neuronal firing rate and regularity was observed (Fig. 31 and J). This increase in neuronal firing rate was reflected in both the burst rate (Fig.

[0245] 3K) and the single-spike firing rate (Fig. 3N). Single neuron bursts also lasted longer and consisted of more spikes as cultures grew older (Fig. 3L and M). Further assessment revealed that, as expected, neuronal activity was most synchronized within the same layer, followed by synchronization between cortical and striatal layers, with the ventral midbrain layer exhibiting the lowest levels of synchronization with the other layers. Taken together the data indicate that our system enables to measure high resolution and longitudinal CSN network properties and there is functional connectivity and maturation across the network. The robustness of CSN circuits of the present invention was defined by using iPSC from different unaffected individuals (K0LF2-1J, SFC065, SFC065 triple reporter and L2135) and compared network properties from longitudinal recordings between DIV85 and DIV180. Our analyses did not reveal significant differences between the different lines in the way these parameters evolved. Hence, similar CSN network properties emerge across different iPSC from healthy individuals. Altogether, the evolution of these parameters indicates that neuron firing properties evolve with increased firing frequencies, longer and more frequent bursts, which, through synaptic integration, progressively integrate into the network's coordinated activity.

[0246] Experiment 6. Al-based neuronal electrophysiological analysis accurately predicts neuronal identity.

[0247] Next, the question whether single-cell electrophysiology (ePhys) parameters contained sufficient information to differentiate the three neuron subtypes on the MEA was addressed. 13.001 putative single units (neurons) from 15 CSN circuits and labelled each unit as cortical, striatal or dopaminergic were identified based on the spatial position where the cells were physically deposited on the MEA. For each single unit, 54 parameters were derived including spike waveform (the extracellular electrical signal produced by every action potential and studied through the autocorrelogram or ACG) descriptors, single neuron firing characteristics, metrics of the neuron's synaptic integration, and circuit network parameters (see Table lfor a definition of these parameters). These metrics were used to train a randomPaVer / CSNOC / 870

[0248] forest (RF) classifier (Fig. 3N), which achieved a classification accuracy of neuron identity of 69.54%, significantly above random chance (33.33%). Every prediction made with a RF classifier is accompanied by a certainty or confidence value. Le., predictions made with higher certainty are expected to be more accurate. The variation of the classification accuracy was investigated as a function of the certainty of the cell-type prediction and found that the prediction accuracy for each neuronal subtype increased above 75% with certainty thresholds above jS0.4; with maximum accuracies of 78.44% for CN, 77.11% for MSN and 79.57% for DAN. Not all parameters contributed equally to discriminating between neuronal subtypes. Among the highest contributors there were found: culture age; waveform descriptors such as the ratio between the amplitudes of the two positive deflections of a spike (AB ratio); single-neuron activity descriptors such as the coefficient of variation of the inter-spike intervals (ISI-CV), which informs about the firing regularity; the population coupling; or the time window in which the maximum synchronized activity is detected (Bin size at maximal SCt). Using these parameters, a multidimensional scaling analysis was conducted and projected the neurons into a 3-dimensional space using the three main scaled coordinates and color-coded them according to the area they were sampled from. This showed that different neuronal subtypes occupied 3 different edges of the cloud (Fig. 3P). This result implies that the different neuronal subtypes as placed in the three regions on the HD-MEA can be detected as functionally different (i.e. the parameters predict neuronal identity) and that these parameters capture the strongest ePhys differences that include both cell specific as well as networkspecific features. The ePhys-phenotype prediction tool was validated by applying it to independent recordings from our triple-reporter cultures. To that end, single cell ePhys responses were recorded from CSN circuits, predicted the cell-type identities of the spike-sorted neurons using our classifier and assessed whether the predictions matched with the fluorescent marker (rather with just their location on the MEA) they expressed based on images captured from the same cultures (Fig. 3D-E). The predicted identities strongly overlapped with the neuronal subtypes expected to be found in each of the three areas (Fig. 3Q-Q'). Taken together, these results have several implications: 1) different subtypes of human induced neurons exhibit distinct intrinsic neuronal ePhys properties, 2) the spiking features specific to each neuronal subtype enable accurate cell type prediction, and 3) the integration of these neurons within a network strongly influences their firing properties, as network properties have strong predictive value.

[0249] Example 7. CSN circuits are synchronized by glutamate.

[0250] Glutamate is the most abundant excitatory transmitter and the only excitatory transmitter in the CSN circuit of the present invention. Blocking AMPA (a-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid) and NMDA (N-methyl-D-aspartate) glutamate receptors has been shown to disrupt synchronizedPaVer / CSNQC / 870

[0251] electrical network activity, e.g. in neonatal mouse cortex and in the hippocampus of adult rats (McCabe et al. Dev Neurobiol. 2007: 67, 1574-1588; Dalkara et al. Brain Res. 1989: 498, 123-130). To determine if glutamate also controls network synchronization in our CSN circuits (100 DIV) AMPA and NMDA ionotropic receptor antagonists a-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (DNQX; 10 pM) and (2R)-2-Amino-5-phosphonopentanoic acid (DAPV; 100 pM) were utilized. Together, DNQX and D-APV completely eliminated the network bursts in the population activity (Fig. 4A-C) with near-complete loss of synchronized activity (Fig. 4D and E). Blocking ionotropic glutamatergic transmission resulted in a significant reduction in the overall firing rate (Fig. 4F). This reduction of firing frequency mostly affected burst firing (Fig. 4H) while sparing the non-burst firing rate (Fig. 4G). Indeed, the remaining bursts were shorter and consisted of fewer spikes (Fig. 41 and J). This selective reduction of bursting activity while not affecting non-burst firing indicates that blocking ionotropic glutamate neurotransmission eliminated the synchronizing crosslinks between the neurons that nevertheless remained spontaneously active without firing in phase with each other (Fig. 4A and B). Thus, glutamate was responsible for synchronizing activity across neurons in the circuits and this activity manifested mainly through network bursts and not isolated spikes, the effect of glutamate signaling using optogenetics was independently assessed. Triple reporter cells (Fig. IB) were used and employed a lentivirus to transduce a floxed bicistronic TagBFP2-channelrhodopsin-2H134R(ChR2) (Nagel et al. Current Biology. 2005: 15, 2279-2284). Only CN expressed CRE (SLC17A7::CRE) and thus activated the expression of BFP2 and ChR2 (Fig. 4K). Exposure of these CSN circuits to 475-nm light indeed showed individual CN to increase their firing rates (Fig.

[0252] 4L,M"). 475-nm light also markedly increased neuronal activity in the MSN (Fig. 4L and M) and DAN layers (Fig. 4M'). This is specific to glutamatergic connectivity because when DAPV and DNQX were added, blue light still increases activity in the CN as expected (Fig. 4L and M") but blocks the increase in neuronal activity in MSN and DAN (Fig. 4L, M-M'), confirming that the light-evoked responses in the MSN and DAN layers were driven by glutamatergic connections from the CN region. Finally, a holistic approach was taken to detect and map these glutamate-mediated excitatory connections within the circuits. The spike train cross-correlation (i.e., cross-correlogram, CCG) was calculated between each pair of neurons. A sharp peak in the CCG (Fig. 4L, histograms on the right) at a delay shorter than 6 ms indicates that the second order neuron fires consistently at a fixed delay after the first, suggesting that they are connected synaptically and that the first neuron is excitatory. Optogenetics and pharmacological experiments were used to confirm that the computationally identified connections and their corresponding CCG peaks (Fig.

[0253] 4L, left panel) were eliminated after the addition of DAPV and DNQX (Fig. 4L, right panel). To characterise the connectivity pattern (Fig. 40), network analysis was performed to calculate how many connections (or degree) each unit had. Most neurons had just a few connections whereas a small number of units were more densely connected (Fig. 4P), and the degree distribution can be fit with a power-law function,PaVer / CSNOC / 870

[0254] a feature mirroring many nervous systems (Lynn et al. Nat Phys. 2024: 20, 484-491). These results are consistent with excitatory activity to initiate synchronized waves of activity in our CSN circuits, a feature also observed in vivo in the basal ganglia of rodents (Peters et al. Nature. 2021: 591, 420-425).

[0255] Experiment 8. Dopamine modulates the excitability of neurons in the CSN circuit.

[0256] Dopamine is a neuromodulator, and as such, it is thought to modulate the excitability of the neurons. This modulation occurs positively through the Dl-like family of dopamine receptors and negatively through the D2-like family. Our single-nuclei sequencing data confirmed the presence of distinct DI- and D2-expressing populations within LGE-derived neurons (Fig. 1J and K), suggesting that the CSN circuits of the invention could potentially be modulated by dopamine. To test whether dopamine is being secreted in the CSN circuit of the invention, sensor cells were exposed (referred to as Sniffer cells in Klein Herenbrink et al. Commun Biol. 2022: 5, 578) expressing the dopamine sensor rDA3h (Zhuo et al. Nat Methods. 2024: 21, 680-691) to the supernatants obtained from CSN circuits or Str-only cultures. Supernatants from CSN circuits treated with 50mM KCI to stimulate neurotransmitter release elicited an 80% increase in the sensor signal (over baseline) compared to only a 20% increase when the supernatants were obtained from Str-only cultures. This indicates that ventral midbrain dopamine neurons secreted dopamine in the CSN circuits of the invention. Changes in neuronal excitability usually manifest in the way neurons respond to external stimuli. For example, a more excitable neuron would respond more readily to excitatory input, exhibit a stronger response and recover faster; and that would ultimately materialize into a higher firing rate (Lahir & Bevan. Neuron. 2020: 106, 277-290.e6; Seong & Carter. Journal of Neuroscience. 2012: 32, 10516-10521). It was tested whether dopamine was modulating neuronal excitability in our cultures by applying the highly specific DI dopamine receptor (DIR) antagonist SCH23390 (0.5 pM)( (Lahir & Bevan. Neuron. 2020: 106, 277-290. e6; Seong & Carter. Journal of Neuroscience. 2012: 32, 10516-10521; Miyawaki et al. PLoS One. 2014: 9, el04438). Blocking the positive modulatory effects of DIR signaling would, in theory, result in a mild reduction of neuronal firing rate. Unlike the blockade of glutamate ionotropic receptors, DI antagonism did not result in any visually striking changes in population activity (Fig. 5A). Analysis of spike trains before and after the application of the blocker revealed a decrease in overall firing rate (Fig. 5B), due to a reduction in both the frequency of isolated spikes (Fig. 5C) and burst rate (Fig. 5D). This small reduction in firing rate in the presence of the drug is compatible with a blockade of the DlR-mediated increase in neuronal excitability elicited by the circuit's endogenous dopamine. A key characteristic of the basal ganglia in PD patients is the presence of abnormally synchronized oscillatory activity across various levels of the basal ganglia-cortical loop (Bergman et al. Trends Neurosci. 1998: 21, 32-38; Levy et al. Brain. 2002: 125, 1196-1209; Hammond et al., Trends Neurosci. 2007: 30, 357-364; Ortone et al., PLoS Comput Biol. 2023: 19,PaVer / CSNOC / 870

[0257] el010645). In our model, DIR antagonism also increases synchronicity and coupling between single neurons (Fig. 5F-G) without significantly affecting the frequency of network bursts (Fig. 5E). The small reduction of non-synchronous activity between network bursts likely contributes to the increased circuit synchronicity (as measured by the spike contrast (Ciba et al. J Neurosci Methods. 2018: 293, 136-143 synchronization metric) and neuronal coupling (Fig. 5A" and H-J). These results demonstrate that, on one hand, dopamine regulates neuronal excitability in the CSN circuits, and on the other hand, that a reduction of dopamine signaling increases circuit synchronization-a feature also observed in the brains of individuals with PD. Altogether, these data supports the use of hiPSC-derived CSN circuits to model aspects of the network dysfunction typical of PD.

[0258] Example 9. Alpha-synuclein fibrils disrupt network dynamics in CSN circuits.

[0259] PD patients usually receive a diagnosis when more than 70% of their nigral neurons are lost and the patterns of communication among the different basal ganglia are already derailed (Lees et al. Lancet.

[0260] 2009: 373, 2055-2066). Toxic aggregation of the neuronal protein alpha-synuclein into higher molecular weight assemblies including oligomers, fibrils and the proteinaceous inclusions called Lewy bodies (Spil lantin i et al. Nature. 1997: 388, 839-840) has been shown to correlate well with functional deficits and in many cases, neuronal loss in the substantia nigra and in other brain regions (Surmeier et al. Nat Rev Neurosci. 2017: 18, 101-113; Cookson. Mol Neurodegener. 2009: 4, 9). Whether alpha-synuclein aggregation is the cause or the consequence of an underlying cellular defect is a matter of intense investigation. What is clear is that mutations that increase its aggregation propensity, genetic variants or genomic rearrangements that increase alpha-synuclein transcription, are causal to disease (Polymeropoulos et al., Science. 1997: 276, 2045-2047; Zarranz et al., Ann Neurol. 2004: 55, 164-173; Singleton et al. Science. 2003: 302, 841-841; Chartier-Harlin et al. Lancet. 2004: 364, 1167-1169; Ibanez et al. Lancet. 2004: 364, 1169-1171). Moreover, overexpression of the SNCA gene (van der Putten et al. The Journal of Neuroscience. 2000: 20, 6021-6029) or inoculation of recombinant alpha-synuclein fibrils preassembled in vitro (Volpicell i-Daley et al. Neuron. 2011: 72, 57-71; Peelaerts et al. Nature. 2015: 522, 340-344) into experimental animal models, causes neuronal loss accompanied by some motor features. To investigate whether our model is sensitive to toxic forms of alpha-synuclein, the CSN circuits of the invention were inoculated with sonicated fluorescently tagged pre-formed alpha-synuclein fibrils into >115-day-old CSN circuits and recorded their neuronal activity at various timepoints for up to 4 weeks (Fig. 6A). Confocal microscopy confirmed efficient neuronal uptake of the fibrils. These localized both in the perinuclear area as well as in the neurites (Fig. 6B) in line with what was reported previously (Bayati et al. Nat Neurosci. 2024: 27, 2401-2416; Freundt et al. Ann Nurol. 2012: 72, 517-524). Inspection of circuit activity did not reveal any overt or striking difference, neither between treated and untreatedPaVer / CSNOC / 870

[0261] cultures, nor between the recording made at baseline and those made after the inoculation. These results are consistent with previous recordings made from neuron monocultures incubated with alpha-synuclein fibrils where also no obvious changes were found (Valderhaug et al., Am J Physiol Cell Physiol.

[0262] 2021: 320, C1141-C1152; Kapucu et al. npj Parkinson's Disease. 2024: 10:1, 1-20). It was resorted again to an RF classifier (Fig. 6C): (1) to predict whether a circuit has been exposed to fibrils or not; and (2) to investigate what electrophysiological and network parameters enable such prediction. To our surprise, our classifier predicted the treatment status with near-perfect classification accuracy (Fig. 6D-E). Among the metrics used to train the classifier, it was found that those quantifying network activity-such as the descriptors of the network bursts, and synchronization-were the best predictors of treatment status (Fig. 6F). The functional divergence of neurons treated with the fibrils was also evident from a multidimensional scaling analysis, as these clearly clustered separately from untreated samples (Fig. 6G). As the training dataset consisted of a limited number of experiments (n=4; Fig. 6C), such high prediction accuracy could have resulted from random functional commonalities among the treatment groups rather than from the treatment itself. To rule out this possibility two training datasets were used. The first one consisted of the baseline recordings made on the circuits before being inoculated with the fibrils. The second consisted of recordings made for the longitudinal characterization of the CSN circuits of the invention (Fig. 3). Both analyses yielded prediction accuracies of 100% and 98.1%, respectively (Fig. 6H-I), indicating that the detected differences were specific to the treatment with alpha-synuclein fibrils. Together these results indicate that alpha-synuclein fibrils induces functional changes in CSN circuits specifically at the network level.

[0263] Example 10. Electrophysiological measurements distinguish LRRK2 and GBA mutant circuits independent of the genomic background.

[0264] Parkinson's disease (PD) is clinically defined by motor symptoms, but its genetic architecture is polygenic and molecularly heterogeneous (Bandres-Ciga et al., Neurobiol Dis. 2020: 137, 104782). To investigate whether different molecular etiologies of disease leave an imprint on cortico-striato-nigral circuits, two isogenic mutant series in unrelated iPSC lines (male 303 K0LF2-1J and female SFC062) were generated. Two frequent, yet incompletely penetrant Parkinson's disease-associated mutations were selected: LRRK2 G2019S and GBA N409S. These genes were chosen because they represent the most common genetic contributors to Parkinson's disease in both familial and idiopathic disease, substantially increasing disease risk, as shown in Fig. 7A (Okun et al ., Nature. 2015: 521, 511-515; Graf et al., Elife.

[0265] 2022: 11; 36-40; Cutts & Eglen, J Neurosci. 2014: 34, 14288-14303; Souza et al., Brain Struct Funct. 2022: 227 , 2465-2487; Watabe-Uchida et al., Neuron. 2012: 74, 858-873). The activity from control and isogenic mutant cortico-striato-nigral circuits was recorded longitudinally (5 months) and it was assessedPaVer / CSNOC / 870

[0266] whether the recorded data contained information that could be exploited to discriminate between genotypes. Machine learning classifiers were trained to discriminate between three genotypes: wildtype, LRRK2 and GBA. The data were split into training and test sets with similar representation of both isogenic lines (Fig. 7B). A random forest classifier achieved a classification accuracy of 85% on the test set and a balanced classification accuracy across the three genotypes (MCC=76%; Fig. 7C and D).

[0267] Classification was accurate across the two isogenic series suggesting that the classifier can identify mutation-specific signatures independently of the genetic background of the carrier (Fig. 7C'-C').

[0268] Furthermore, different features were investigated for their contribution to the classification task. Post-hoc SHAP (SHapley Additive exPlanations, Lundberg, Erion & Lee, 2018, https: / / arxiv.org / pdf / 1802.03888) analysis on the trained classifier revealed that network parameters had the highest discriminative power between the three genotypes (Fig.7E). The values of these features enabled extracting electrophysiological signatures that define the behavior of every genotype. For example, the classifier associated low network burst rates with LRRK2 circuits, whereas high rates were associated with GBA circuits (Fig. 7E'-E"). Of note, iPSC line identity only ranked 9th in the SHAP analysis, further indicating that the classifier distinguishes LRRK2 and GBA genotypes despite the different genetic backgrounds.

[0269] To determine whether these differences in network behavior could be connected to changes in the activity patterns of the three neuronal subtypes, it was examined how the expression of the different electrophysiological features varied across genotypes over time. Linear mixed-effects models (LMM) fitted to features that were defined per layer revealed that only the activity patterns of dopamine neurons showed significant changes in the different genotypes (Fig. 7F). Of the 11 electrophysiological features that were significantly different in dopamine neurons among the genotypes, four descriptors of the bursting behavior of dopamine neurons, burst rise and decay times, burst total time and mean burst duration, were highly correlated with the three most discriminative network features: the mean duration of the network bursts and network burst rate, and network burst total time (Fig. 7G-G')- These results suggest that these changes may even be present from the moment these circuits are formed. The activity of cortico-striato-nigral circuits were monitored between 85 and 145 days of age, a period that epigenetically, transcriptomically and functionally corresponds to developing stages (Luo et al., Cell Rep. 2016: 17, 3369-3384; Trujillo et al., Cell Stem Cell. 2019: 25, 558-569.e7, He et al., Nature.

[0270] 2024: 635, 690-698). In said circuits, LRRK2- and GBA-induced electrophysiological dysfunction is subtle but measurably distinct, consistent with LRRK2 and GBA affecting different cellular processes: LRRK2 phosphorylates proteins orchestrating vesicular transport and endocytosis, while GBA encodes the lysosomal enzyme glucocerebrosidase. Yet, at the circuit level, the functional problems caused byPaVer / CSNOC / 870

[0271] mutations in these genes result in cortico-striato-nigral network problems, including dopamine neuron modulatory dysfunction.

[0272] These results suggest that Parkinson's disease mutations cause changes in the bursting behaviour of dopamine neurons, and this behaviour correlates with those network-level differences that enable the classification of Parkinson's disease mutant circuits. The machine learning approach used in the invention, not only classifies mutant and control circuits with high confidence but also distinguishes LRRK2 from GBA mutants, showing that unbiased electrophysiological screening can resolve molecularly defined subtypes of Parkinson's disease. This raises the possibility of using circuit-level signatures as functional markers to stratify patients and guide mechanism-specific therapies, even when their clinical symptoms overlap. The use of the CSN circuit on chip of the invention could help in detecting the asymptomatic subjects carrying out mutations affecting these pathways who may feature only subtle motor impairments suggestive of compromised nigrostriatal circuit function years before the diagnosis.

[0273] Materials and Methods

[0274] hiPSC maintenance and differentiation. iPSC from three control lines (SFC062, SFC065, L2135 and K0LF2-1J) and TH::mClover3 / GAD2::mScarlet / SLC17A7::CRE knock-in line engineered in the SFC065 line (hiPSC lines used: SFC065, Sendai virus-based reprogramming, age at biopsy 65 years; L2135, Retrovirus-based repgramming, age at biopsy 40 years, Seibier et al. J. Neurosc. 2011, 16, 5970-5976; K0LF2-1J, Sendai virus-based reprogramming, age at biopsy 55-59 years. Pantazis et al. Cell Stem Cell 2022, 12, 1685-1702) were differentiated into cortical excitatory, striatal medium spiny and ventral midbrain dopaminergic neurons as described in the methods section. Briefly, hiPSC were maintained on Geltrex-coated plates (ThermoFisher, A1413302) with StemMACS PSC-BrewXF, human medium (Miltenyi Biotec, 130-127-865) and medium was refreshed daily. All human iPSC lines were used in accordance with an MTA with the University of Lubeck (Germany) or with Jackson Laboratories (USA) and registered in the KU Leuven / UZ Leuven Biobank.

[0275] hiPSC differentiation to cortical, striatal and ventral midbrain progenitors. The day before starting the neural induction, 60-70% confluent hiPSC were dissociated with accutase (ThermoFisher, A1110501), counted and seeded on top of freshly Matrigel (Sigma, CLS354277) coated multi-well plates at a density of 300.000 cells / cm2. The morning after, the confluent cultures started the neural induction and patterning following the following morphogen exposure regimes: For cortical neural induction, the cells were exposed to SB431542 (lOpM, Miltenyi Biotec, 130-106-543) from day 0 until day 6, LDN193189 (lOOnM, Sigma, 6053 / 10) from day 0 until day 12. The cells were cryopreserved on day 27 without any intermediate cells splitting. For striatal neural induction, the cells were exposed to SB431542 from day 0 until day 6, LDN193189 (500 nM) from day 0 until day 11, IWP2 (1 pM, Miltenyi Biotec, 130-105-335)PaVer / CSNOC / 870

[0276] from day 0 until day 4 and Activin A (25 pg / mL, Miltenyi Biotec, 130-115-009) from day 12 until day 19. The cells were split 1:1 on day 11 and 1:2 on day 16 before being cryopreserved on day 19. For ventral midbrain neural induction, the cells were exposed to SB431542 (lOpM, 1150 Miltenyi Biotec) from day 0 until day 6, LDN193189 (500 nM) from day 0 until day 11, SAG (500 nM, Tocris, 6390) or SHH-C24II (100 ng / mL, Miltenyi Biotec, 130-095-727) and Purmorphamine (2 pM, Sigma, SML0868) from day 0 until day 8, CHIR99021 (1.5 pM, Miltenyi Biotec, 130-103-926) from day 3 until day 16, and FGF8b (100 ng / mL, Miltenyi Biotec, 130-095-733) from day 9 until day 16. The cells were split 1:1 on day 11 and 1:2 on day 16 before being cryopreserved on day 27. In all cases, neural induction was initiated in Neural Maintenance Medium (NMM) followed by a gradual transition to N2B27 starting from day 11. NMM is based on a 1:1 mix of DMEM / F-12+GlutaMAX (ThermoFisher, 10565018) and Neurobasal (ThermoFisher, 21103049) supplemented with 0.5x B27 minus Vit. A (ThermoFisher, 12587010), 0.5x N2 (ThermoFisher, 17502048), 0.5x GlutaMAX (ThermoFisher, 35050038), 0.5x PenStrep (ThermoFisher, 15140122), 2.5 pg / mL Insulin (Sigma, I0516-5ML), 50 pM 2-Mercaptoethanol (ThermoFisher, 31350010), 0.5x NEAA (ThermoFisher, 11140035) and 0.5x Sodium Pyruvate (ThermoFisher, 11360070). N2B27 is based on Neurobasal supplemented with 0.5x B27 minus Vit. A, 0.5x N2, lx GlutaMAX, lx PenStrep and lx NEAA.

[0277] Multielectrode array plate preparation for seeding.

[0278] Six-well MaxTwo plates (Maxwell Biosystems) were pretreated following manufacturer's instructions. The day before the seeding, the wells were pre-coated with 0.002% PEI (Sigma) at 37 °C for 1 hour. PEI solution was washed with sterile deionized water, air-dried and custom-made 3-chamber inserts (acrylic adhesive-back silicone rubber sheets, McMaster-Carr) were glued on the top of the active area (area covered by the electrodes and amplifiers). 10 pg / mL recombinant human Iaminin521 (BioLamina) was added to each of the three chambers and the plates were kept at 4°C overnight followed by a 2-hour incubation at 37 °C immediately before the seeding.

[0279] Cortico-striato-nigral circuit assembly.

[0280] Cortical and ventral midbrain progenitors were thawed on Matrigel-coated wells at DIV27 in Terminal Differentiation (TD) medium. Three days later (DIV30), the cells were dissociated with accutase, counted and seeded at their respective position of the MaxTwo wells with the help of a custom-made detachable plastic insert. 45 to 60 minutes later, the plastic inserts were detached and additional hLam521-supplemented media was added to the wells. Five days after seeding, 50.000 primary rat astrocytes were added per well and 10 pM DAPT (Tocris, 2634 / 10) treatment was initiated. Primary rat astrocytes were chose over human ones as these have been shown to perform particularly well in electrophysiology studies in human neurons (Meijer et al. Cell Rep. 2019: 27 , 2199-2211. e6; Rhee et al. Cell Rep. 2019: 27, 2212-2228. e7). Five days later, the circuits were incubated with 75 pM FUdR (Sigma, 343333) for 24PaVer / CSNOC / 870

[0281] hours to prevent further cell division. From day 45 onwards, the medium was gradually changed to BrainPhys+++. TD was prepared by supplementing Neurobasal-A (ThermoFisher, 10888022) with lx B27 minus Vit. A, lx GlutaMAX, lx PenStrep, 10 ng / mL GDNF (Miltenyi-biotech, 130-129-544), 10 ng / mL BDNF (Miltenyi-biotech, 130-093-811), 0.2 mM dbcAMP (Sigma, D0627-250mg), 0.2 mM L-ascorbic acid (Sigma, A4544-25G), lOOnM SR11237 (Tocris, 3411 / 10). Brainphys+++ was prepared by supplementing Brainphys (Stem Cell Technologies, 05790) with lx B27 Plus (ThermoFisher, A3582801), lx GlutaMAX, lx PenStrep, 10 ng / mL GDNF, 10 ng / mL BDNF, 0.2 mM ascorbic acid and lOOnM SR11237.

[0282] CRISPR / Cas9 gene editing.

[0283] For engineering the reporter lines, the guidelines published in Skarnes et al. (Methods. 2019: 164-165, 18-28) were followed, with some modification as plasmids as donor templates were used. The day of the nucleofection 60-70% confluent cultures were pre-treated with Rl for 2h, dissociated with accutase, counted and 800.000 cells were nucleofected with the P3 Primary Cell 4D-Nucleofector™ X Kit L using the Amaxa Nucleofector 4D (both from Lonza). Nucleofected cells were reseeded on Matrigel-coated wells of a 24-well plate in the presence of RevitaCell (StemCell Technologies, A2644501) and the HDR enhancer v2 (IDT technologies) for 24 hours and 72 hours, respectively; and incubated at 32°C for 72 hours, moment in which they were returned to 37°C. Five to seven days post- nucleofection, the surviving cells were seeded as single cells at low density on Matrigel-coated plates to isolate monoclonal colonies. The colonies were taken to 96- well plates and screened by means of PCR followed by Sanger sequencing. These mutations were introduced in heterozygosis except LRRK2 G2019S in the K0LF2-1J background as we could only recover homozygous knock-ins after several attempts. We therefore used a homozygous K0LF2-1J LRRK2 G2019S line as epidemiological studies do not reveal significant clinical expression between homozygous and heterozygous carriers of LRRK2 G2019S (Ishihara et al., Arch Neurol. 2006: 63, 1250-1254). Correctly targeted clones were subjected to comparative genomic hybridization (CGH) arrays to rule out large chromosomal aberrations.

[0284] Nuclei isolation.

[0285] Recent advances in single nuclei RNA sequencing have demonstrated that with optimal buffer / detergent compositions, the assay performance and sensitivity of gene detection can be significantly enhanced (Slyper, et al. Nat Med. 2020 26, 792-802; Otero-Garcia, et al. Neuron. 2022: 110, 2929-2948. e8). A modified protocol for nuclei isolation was performed (Slyper, et al. Nat Med. 202026, 792-802).The 24 well plate that was stored at -80°C was kept on dry ice till the lysis step. The plate was thawed on wet ice for 5 mins, before adding the lysis buffer. 600 ul of ice-cold homogenization buffer (146mM NaCI; Thermo Fisher, lOmM Tris 7.5; Thermo Fisher, 1 mM CaCL; Thermo Fisher. 21 mM MgCL; Thermo Fisher, 0.03% Tween-20; Sigma Aldrich, 0.03% NP-40; Sigma Aldrich, 1% BSA; Sigma Aldrich, 25 mM KCI; Thermo Fisher, 250 mM sucrose; Sigma Aldrich, 1 mM mM DTT; Thermo Fisher, 0.5X complete protease inhibitor;PaVer / CSNQC / 870

[0286] Roche, 0.2U / ul RNAse In Plus; Promega) was added to the well with thwed cells in them. Using a cell scraper, scrape the surface of the well to detach and lyse the cells. The lysed cells were Incubated for 5 mins for lysis completion. The lysed cells from the plate were removed and the nuclei suspension was filtered using 40pm cell strainer. The cell filtrate was transferred to a low bind eppendorf and centrifuged at 500xg for 5 mins. The supernatant was discarded. The nuclei pellet was thoroughly resuspended in 1 ml of HB w / o Tween buffer (146mM NaCI; Thermo Fisher, lOmM Tris 7.5; Thermo Fisher, 1 mM CaCL; Thermo Fisher. 21 mM 1235 MgCL; Thermo Fisher, 1% BSA; Sigma Aldrich, 25 mM KCI; Thermo Fisher, 250 mM sucrose; Sigma Aldrich, 1 mM mM DTT; Thermo Fisher, 0.5X complete protease inhibitor; Roche, 0.2U / ul RNAse In Plus; Promega). The nuclei suspension was again centrifuged at 500xg and the supernatant was discarded. The nuclei pellet was resuspend in IX resuspension buffer (BSA 1%; Sigma Aldrich, 1 / ul RNAse In Plus; Promega, PBS; Thermo Fisher). The nuclei enrichment was performed by sorting nuclei in a flow sorter.

[0287] Single nuclei sequencing.

[0288] Library preparations for the single nuclei RNA-seq was performed using 10X Genomics Chromium Single Cell 3' Kit, v3.1 NextGEM chemistry (10X Genomics, 1246 Pleasanton, CA, USA). The cell count and the viability of the samples were accessed using LUNA dual florescence cell counter (Logos Biosystems) and a targeted Nuclei recovery of upto 10000 cells were aimed for each of the samples. Single nuclei sample enriched post FACS sorting was subjected to single nuclei encapsulation using either 10X controller or custom HyDrop microfluidics droplet generator (De Rop, et al. Elife. 2022: 11). The ability of fine tuning the liquid flowrates with custom HyDrop microfluidics setup allows to get higher cell encapsulation efficiency. Single nuclei RNAseq libraries were prepared using manufacturers recommendations (Single cell 3' reagent kits v3.1 user guide; CG000204 Rev D), and at the different check points the library quality was accessed using Qubit (ThermoFisher) and Bioanalyzer (Agilent). For a targeted sequencing saturation of 50-60%, sequencing were performed at a depth of 30,000 - 60,000 reads per cell and single cell libraries were sequenced either on Illumina's NovaSeq 6000 platform or MGISEQ-2000 sequencing platform using paired-end sequencing workflow and with recommended 10X; v3.1 read parameters (28-8-0-91 cycles). 10X libraries that were sequenced on MGISEQ-2000 sequencing platform was sequenced at the MGI Hong Kong SAR sequencing facility. Before being able to sequence on the MGI platforms, the 10X libraries need to undergo a conversion step using MGIEasy Universal DNA Library Prep kit. Briefly, the final 10X sequencing libraires were circularized using the splint-ligation step and the circularized libraries were converted to single stranded DNA copies. DNA nanoballs were prepared from the circularized ssDNA using Rolling Circle Amplification (RCA). DNB libraries generated were then flown through the patterned flow cell of the MGISEQ-2000RSPaVer / CSNOC / 870

[0289] High-Throughput Sequencing kit. A custom sequencing recipe of paired end read of 28 bps (read 1), 100 bps (read 2) and 10 bps (index 1) was used for sequencing.

[0290] Processing and Analysis of Single Nuclei Data.

[0291] A combined Human (GRCh38, release 98) and Rat (mRatBN7.2, release 109) CellRanger index was created using instructions from lOx Genomics for the creation of dual indexed samples and CellRanger (v7.1.0) mkfastq with default parameters. Raw BCL files were demultiplexed using Illumina's bcl2fastq (v2.20.0) into fastq files which were then aligned to the joint human and rat genome and quantified using CellRanger (v7.1.0) count with default settings. For samples containing cells originating from multiple genotypes, cellsnp-lite and vireo were used for cell demultiplexing. First, SNPs from the 1000 genomes project (GRCh38) were downloaded and filtered to include SNPs only found within gene regions as determined by the CellRanger gtf, as well as including only SNPs with an allele frequency >=0.10 and <= 0.90 to remove very rare variants, genomes were renamed to match those in the combined index, cellsnp-lite (vl.2.3; X. Huang & Huang, 2021)) was run using all chromosomes and the -genotype - cellTAG None options. Output data from cellsnp-lite was used with vireo (vO.5.8; Y. Huang et al., 2019) to call genotypes per cell, with the number of donors set to 2. Filtered count matrices were loaded into Scanpy (vl.9.1; Wolf, et al. Genome Biol. 2018: 19, 15; Python 3.8.10), any cells determined to be either Rat or multiplet by CellRanger were removed. For samples containing multiple genotypes, outputs from vireo were loaded and cells were annotated with their identified genotype, doublets and cells where a genotype could not be determined were removed. Scrublet was then used to score remaining cells for doublets and cells with a doublet score >=0.50 were removed, finally, cells with more than 15% of counts assigned to mitochondrial reads were also removed. Counts were normalized to 10,000 counts and log normalized for downstream analyses. All samples were combined into a single object and highly variable genes were detected (min_mean=0.0125, max_mean=3, min_disp=0.5), percentage of mitochondrial reads and total counts were regressed out and the counts were scaled to unit variance with a mean of zero and a max count of 10. A principal component analysis was then performed, selecting the top 90 components to be used for further analyses. Harmony (Korsunsky, et al. Nat Methods. 2019: 16, 1289-1296) was used to correct PCA eigenvalues for any batch effect between samples. Corrected PCA embeddings were used for the calculation of UMAP and tSNE dimensionality reductions as well as cluster identification using the Leiden algorithm. Clusters were manually annotated with cell types based on marker genes and these were used to perform pseudobulk based differential gene expression analyses in a cell type specific manner using DESeq2 (Love et al., Genome Biol. 2014: 15, 550). In short, cells were aggregated based on their sample of origin and cell type, providing several replicates per condition. Coculture status was compared with Batch and Genotype being included as covariates. Gene set enrichment analysis was performed using GSEApy (vl.1.3; Fang et al., 2023) prerank using the followingPaVer / CSNOC / 870

[0292] parameters: min_size=5, max_size=1000, permutation_num=1000, seed=42. To generate scores for gene sets, the score_genes function from Scanpy was used on the raw counts with default parameters. Immunofluorescence staining and confocal microscopy.

[0293] Cultures were fixed with 4% formaldehyde for 15 minutes at room temperature, blocked and permeabilized with 3% normal goat serum and 0.1% Triton X-100 or with 0.01% saponin in the case of membrane-bound antigens such as presynaptic markers. Primary antibodies were diluted in the blocking solution and the samples were incubated with these overnight at 4 °C, followed by a 2h incubation with the secondary antibodies at room temperature. Nuclei were counterstained with 4',6-diamidino-2-phenylindole (DAPI). Immunolabelled circuits were imaged with a Nikon TiE AIR equipped with lOx, 20x and 60x objectives. The following primary antibodies were used for imaging: anti-GFP (Sigma, G6539), anti-TH (Santa Cruz Biotechnologies, sc- 25269), anti-TH (Sigma, AB152), anti-GAD65 (Chemicon, MAB351R), anti-dsRed (Takara, 632496), anti-tRFP (Evrogen, AB233), anti-VGLUTl (Addgene, 180087-rAb), anti-VMAT2 (R&D Systems, MAB8327), anti-DARPP32 (Cell Signaling, 2306S), anti-CTIP2, i.e. anti-BCL11B (Abeam, abl8465), anti-tdTomato (SICGEN, AB8181-200), anti-SATB2 (Abeam, ab51502), anti-P3-tubulin (BioLegend, 802001), anti-FOXGl (Bioss, bs- 11557R), anti-OTX2 (Santa Cruz Biotechnologies, ), anti-GSX2 (Abeam, ABN162), anti-DLX2 (Santa Cruz Biotechnology, sc-393879), anti-EOMES (Abeam, AF6166), anti-PAX6 (Santa Cruz Biotechnologies, sc-81649), anti-TBRl (Merck-Millipore, abl0554), anti-LMX1 (Millipore, AB10533), anti-FOXA2 (Santa Cruz Biotechnologies, sc-101060) and anti-ENl (DSHB, 4G11), anti-SOX2 (Santa Cruz Biotechnologies, sc-365823), anti-OCT4 (Abeam, abl9857), anti-NANOG (Santa Cruz Biotechnologies, sc-293121) and anti-TRA-1-81 (Abeam, mab4381). Secondary antibodies from the Alexa series were purchased from Life Technologies and those conjugated to ATTO490 LS were purchased from HYPERMOL and they were used 1:1000 in all cases.

[0294] Lentiviral vector production.

[0295] Eight million HEK cells were transfected with a 1:5:6.5 mass mix of pMD2.G (a gift from Didier Trono; Addgene plasmid #12259), psPAX2 (a gift from Didier Trono; Addgene plasmid #12260) and the transfer vector using jetOPTIMUS (Polyplus- transfection). Fourteen hours later, medium was replaced with fresh OptiMEM medium (Life Technologies; 31985062). Supernatant containing viral particles was collected 24 and 48 hours after initial medium replacement and was subsequently concentrated 500x using Amicon Ultra-15 Centrifugal Filter Unit (Sigma; UFC910024) and stored at -70 °C until further use (Brum, 2015, Preprint at https: / / doi.org / 10.17504 / protocols.io.c54y8v ).

[0296] Molecular cloning.

[0297] Homology-directed repair (HDR) plasmid templates, transfer plasmids for lentiviral production and dopamine reporter expression plasmids were engineered using NEBuilder HiFi DNA Assembly (New England Biolabs; #E2621). Genomic regions used as HDR templates and the different genetic elementsPaVer / CSNOC / 870

[0298] introduced in the plasmids were amplified using Q5 High-Fidelity DNA Polymerase (New England Biolabs; #M0491L). The following plasmids were obtained from Addgene and used to extract coding sequences: #118760 (mClover3), #118431 (mScarlet), #116856 (mTagBFP2 / CAAX), #80342 (TDsmURFP), #26197 (oG and TV A), #26973 (ChR2H134R), #208703 (GRAB-rDA3h). Lentiviral transfer plasmids were grown in NEB Stable Competent E. coli (New England Biolabs; C3040H) at 30 °C for 24 hours. The rest of the plasmids were grown in home-made DH5alpha E. coli at 37 °C overnight (O / N).

[0299] Correlative light and electron microscopy.

[0300] The cells were grown on a 35 mm glass bottom petri dish (MatTek Corporation, P35G-1.5-14-C). Prior to fixation for electron microscopy, the entire circuits were imaged with a Zeiss LSM900 Airyscan2 equipped with 20x objective and selected areas z- sectioned with a 60x objective. After the live imaging, the cells were post-fixed over night with 2.5% GA in 0.1M sodium cacodylate trihydrate buffer (pH7.4) at 4°C. Subsequently the cells were washed with 0.1M sodium cacodylate trihydrate buffer (pH 7.4) and then osmicated for lh on ice (1% OsO4 and 1.5% potassium ferrocyanide). After a washing step (ddH2O) the cells were incubated O / N in 0,5% uranyl acetate / 25% methanol. The following day the samples were washed (ddH2O) and stained en bloc with lead aspartate (Walton's lead aspartate: 20mM lead nitrate in 30mM sodium aspartate, pH5.5) for 30 min at 60°C. After a final washing, the sample was dehydrated with solutions of increasing ethanol concentration (30%, 50%, 70%, 80%, 95% and twice with 100%). The cells then were infiltrated O / N with agar 100 resin (Laborimpex, AGR1031). After an additional 6h incubation with fresh resin, a BEEM capsule (Laborimpex, AGAR G362-1) was placed over the region of interest. Then the sample was placed in an oven for 48h at >60°C, followed by immersion liquid nitrogen to facilitate the detachment of the BEEM capsule containing the cells embedded in the resin. The resin block was sliced at 70 nm sections on an ultramicrotome (EM UC7, Leica). The sections were imaged using a JEM-1400 transmission electron microscope (Jeol) at 80 keV.

[0301] Rabies tracing experiments.

[0302] Day 19-old striatal progenitors were transduced with a mix of VSV-coated hSynl::fFLEX-mC3-TVA-oG-Wpre and LV-Dlx5 / 6::FlpO lentivira to induce the expression of the receptor needed for the recombinant rabies to enter the cell only in cells of the GABAergic lineage. CSN circuits were assembled using these progenitors and transduced with EnvA-coated CVS-N2c(deltaG)-tdTomato recombinant rabies 70 days after circuit assembly. Circuits were fixed, stained and imaged 10 days after recombinant rabies virus transduction.

[0303] Electrophysiological recording with high-density multi-electrode arrays (MEA).

[0304] Cortico-striato-nigral (CSN) circuits were recorded bi-weekly once they reached 55 days on the HD-MEA (DIV85) using MaxTwo multiwell MEA plates and the MaxLab Live software (MaxWell 1400 Biosystems). Because of limited channel availability, it was possible only to record from 1020 electrodes per wellPaVer / CSNOC / 870

[0305] simultaneously, which were evenly delivered across the three layers in a grid of 2x2, with one electrode enabled to record every other position. To ensure correct circuit sampling, areas at the intersection between the layers were avoided. The recordings were performed using the "record" function at a 10-kHz sampling rate and 1024x gain. For those recordings including pharmacological manipulation of the cultures, drugs were applied during uninterrupted recording sessions after an initial baseline recording. The following drugs were used: DAPV (100 pM, Alomone labs, D-145), DNQX (10 pM, Alomone labs, D-130), SCH-23390 (0.5 pM, Tocris, 0925 / 10) and L-741,626 (100 nM, Tocris, 1003 / 10).

[0306] Data processing and spike-sorting.

[0307] To characterize the electrophysiological features of neurons at cellular resolution, single-unit activity was first resolved via spike-sorting analysis using Kilosort v2.5 (Pachitariu et al., Nat Methods. 2024: 21, 914-921) and then manually curated using Phy v2.0b5 (https: / / github.com / cortex-lab / phy). Parameters describing spike waveforms and firing patterns were extracted for each isolated neuron using the CellExplorer framework (CellExplorer vl.0) (Petersen et al., Neuron. 2021: 109, 3594-3608. e2). These parameters were utilized to characterize circuit maturation and to classify neuronal phenotypes. Scripts to evaluate the population coupling parameter, spike contrast synchronization, spike time tiling coefficient, burst and network burst activity and synchronized phases were implemented as custom calculations in the CellExplorer processing module. The population coupling parameter was calculated as described in Okun, et al. (Nature. 2015: 521, 511-515). The population rate for a given unit was computed by accumulating all detected spikes for the other resolved units in a recording, smoothing the obtained vector with a Gaussian window of half-width 12 ms, and subtracting the mean population rate from this vector. The similarity between the population rate and the smoothed firing rate of the individual unit was measured using the xcorr() function in MATLAB. The spike-triggered population rate was obtained from this similarity vector at 0-time lag. Finally, the population coupling was calculated as the spike-triggered population rate normalized by the number of spikes fired by the individual unit. The spike contrast synchronization parameter was implemented as described in Ciba et al., J Neurosci Methods., 2018: 293, 136-143. Briefly, this network parameter is defined as the maximum value of a synchronization curve, that is calculated as the product of a contrast curve and an active spike train curve. The contrast curve is obtained by creating a time histogram for various bin sizes counting the number of spikes per kth bin across all spike trains and accumulating the absolute differences between each neighboring pair of bins. The active spike train curve is created from a time histogram for various bin sizes counting the number of spike trains showing at least one spike per kth bin. Spike train correlations between pairs of neurons were calculated utilizing the spike time tiling coefficient (STTC), as previously reported in Cutts& Eglen (J Neurosci. 2014: 34, 14288-14303). This parameter is calculated as the difference between the proportion (P ) of spikes from unit A that fall within ±At of any spike fromPaVer / CSNOC / 870

[0308] unit B and the proportion of recording time that is present within ±At of any spike of unit B (TB). The obtained value is then divided by 1 minus the product of P and TB. The same is then applied after inverting unit A and B, and the mean between these 2 values is kept. Negative correlations were not investigated as the STTC cannot be straightforwardly applied to those. STTC values were computed at various time windows or delays At to investigate various timescales of correlations.

[0309] Inter-spike intervals were calculated on spike trains of individual units to identify neuronal bursting activity. This activity was obtained by thresholding the intervals below a 50 ms threshold for a minimum of 2 consecutive intervals, resulting in at least 3 spikes that were present in a neuronal burst. Representative ISI distributions and a definition of the bursting threshold. Network bursts were defined as concurrent bursting activity originating from the three neuron subtypes on the HD-MEA. To ensure that all subtypes were engaged in a network burst we excluded concurrent bursting activity originating from small subgroups of units containing only a single or two neuron subtypes. Such activity was evident from the multiple plateau phases when plotting concurrent unit bursting rate as a function of the number of units engaging in a network burst (i.e. circuit engagement). A threshold on the circuit engagement was calculated as the inflection point at the start of the last plateau phase. To segregate network activity into synchronized and non-synchronized phases, the population firing rate was calculated by counting the detected spikes in 200 ms bins, and interpolating with a cubic spline function to increase smoothness and resolution to 10 ms. The result was averaged by the number of detected units. Synchronized activity was defined by thresholding the firing rate above the median minus two times the "noise level". The "Noise level" was estimated using the median absolute deviation (MAD). Here, the median and MAD were used to account for the wide firing rate variability and "outlier" values due to the presence of high amplitude synchronized bursts. To characterize network behavior in pharmacological and optogenetic experiments, multi-unit activity (MUA) was used without spike sorting. Multi-unit spikes were obtained from the online spike detection program built into the Maxwell system. In the optogenetics experiment, to identify whether a neuron responded to light stimulation, a paired t-test (without correcting for multiple comparisons) was used to compare the firing rate before and after laser onset. The window of comparison was limited to 10 seconds before laser onset vs 5 - 15 seconds after laser onset. Excluding the initial 5 seconds from the analysis in order to reject potential laser-induced electrical artifacts. Neurons that exhibited significantly different firing rates before and after laser stimulation were further categorized into increased or decreased responses, depending on the polarity of the firing rate change. Putative monosynaptic connections were estimated based on the cross-correlation of spiking activity between each pair of neurons, as implemented in the ce_MonoSynConvClick() function in CellExplorer vl.0. Here, we restricted the maximally allowed monosynaptic delay to 6 ms, used a bin size of 0.5 ms to calculate cross-correlogram (CCG), convolvedPaVer / CSNOC / 870

[0310] the CCG with a Gaussian window with 10 ms standard deviation, and limited the significance level to 0.1 %. Only excitatory connections were searched for. To ensure that the algorithmically identified putative connections are indeed mediated by glutamate, optogenetics and glutamate receptor blockers (DAPV / DNQX) were additionally used to reject unlikely connections. Specifically, we screened for connections between a presynaptic neuron that responded to light stimulation both before and after DAPV / DNQX treatment, and a postsynaptic neuron that responded to light only before, but not after, DAPV / DNQX treatment. Only these algorithmically and experimentally identified putative connections were retained for further visualization and analysis.

[0311] Phenotype prediction methods.

[0312] Six learning algorithms were assessed in their performance to classify the different cell phenotypes. Linear and quadratic discriminant analysis (LDA and QDA), nearest neighbor classifiers (KNN), multiclass decision trees, and ensemble classifiers AdaBoost and Random Forest were compared using the MATLAB functions fitcdiscr(),fitcknn(), fitctree(), fitcensemble() and TreeBagger(), respectively. Model classification accuracies were evaluated using k-fold cross validation with a k value of 10, and stratified folds that were generated using the cvpartition() function in MATLAB. Test set accuracies were computed using the values in the confusion matrix, which was created using the true test data and predicted phenotypes. The split predictor selection technique accounted for differences in the number of unique values (i.e. levels) between various features, as most features derived for single cells (waveform, electrophysiology and synaptic integration) contained many more levels compared to circuit features (metadata and network). This was implemented in MATLAB by specifying the interaction test as 'PredictorSelection' argument which also accounted for potential associations between pairs of features and putative phenotype. Predictor importance estimates were obtained by permuting out-of-bag observations with the oobPermutedPredictorlmportance() function in MATLAB. Similarly, this function is also robust against large variations in levels between features. The posterior probability of out-of-bag units belonging to a phenotype class was computed as the fraction of units of the class in a tree leaf. Derived probabilities were averaged over all trees in the ensemble model using the oobPrediction() function in MATLAB. The classification model performance across a range of thresholds on the classification probabilities was calculated using the rocmetrics() function, which allowed us to derive a receiver operating characteristic (ROC) curve that shows the true positive rate versus the false positive rate for different thresholds of classification probabilities. The fraction of trees was computed in the random forest model for which any two units land on the same leaf using the fillProximities() function in MATLAB. The obtained proximity matrix was pairwise distance transformed with the pdist() function using the spearman method for distance calculation. Multidimensional nonmetric scaling with Kruskal's normalized stress formula as goodness-of-fit criterion and an embedding dimension of 2 (alphaPaVer / CSNOC / 870

[0313] synuclein fibril data) or 3 (neuron phenotype data) was then applied using the mdscale() function to identify potential clusters in the data.

[0314] Optogenetic stimulation.

[0315] The system used for optical stimulation was equipped with a blue LED at a wavelength of 470nm and an optical power of 809mW (Thorlabs, M470F4) controlled by a dedicated LED driver (Thorlabs, LEDD1B). The blue LED driver is connected to an Arduino micro, itself connected to a computer to control the excitation protocol (number of pulses and duration). The LED is followed by a condenser lens before and sent into a lm multimode optical fiber with a 1.5mm core diameter and a numerical aperture of 0.5 (M107L01). It is injected in the fiber thanks to a 30mm focal length lens (AC254-030-A). Finally, a multimode collimator (F950SMA-A) is placed at the output of the fiber to obtain a 10mm beam diameter. The optogenetics experiment consisted in 22 15-second-long light pulses followed by 15 seconds-long pauses. To minimize the electrical aberration induced by the light, the recording gain was set at 112x. Dopamine detection with dopamine sensor cells.

[0316] HEK 293T (Lenti-X™ 293T Cell Line; Takara) cells were transfected with a 8:1:1 mass ratio of empty pUC, a control pMAX-mTagBFP2-CAAX and the pMAX-rDA3h dopamine sensor (Zhuo et al., Nat Methods.

[0317] 2024: 21, 680-691) using jetOPTIMUS. Two days later, transfected cells were dissociated and reseeded in imaging 96-well plates in phenol-free DMEM supplemented with heat-inactivated fetal bovine serum (ThermoFisher; A5256701), GlutaMAX and PenStrep. The day of the experiment, a volume of cortico-striato-nigral (CSN) or striatal-only circuits' supernatant or fresh BrainPhys+++ medium equal to the volume in the wells was added to the different wells. Two minutes later, the plate was introduced into an Operetta CLS imaging system (Perkin Elmer) equipped with 355-385 and 530-560 nm LEDs and a 20x objective to acquire blue and red fluorescence from the control and dopamine reporter fluorescent proteins. The single datapoints result from dividing the fluorescence from the experimental sample by the one from those where fresh media was added.

[0318] Inoculation of CSN circuits with sonicated alpha-synuclein fibrils.

[0319] >115-day old cultures were inoculated with lpg / mL of sonicated ATTO594-conjugated Type 1 human alpha-synuclein pre-formed fibrils (Stressmarq; SPR-322-A594) or with vehicle. Circuits were recorded right before adding the fibrils, 3-, 7-, 14- and 28-days post-inoculation. Separately, a glass control was also treated to verify fibril uptake.

[0320] Quantification and statistical analysis.

[0321] All data were expressed as mean ± S.E.M. Graphs and statistical analyses were made in GraphPad Prism or with different Python or R packages. Distribution of the raw data was tested for normality; statistical analyses were conducted using two-tailed t-tests (paired when appropriate) or Linear mixed-effectsPaVer / CSNOC / 870

[0322] models depending on the structure of the experimental data. Linear mixed-effects models were fitted to each electrophysiological feature using the formula feature ~ Genotype x cultureAge + (11 CircuitID). Models were estimated with Ime4 using Satterthwaite degrees of freedom (ImerTest). p-values were FDR-corrected across features, and model diagnostics were inspected for each fit. Machine learning classifiers were trained in Python (using scikit-learn and pycaret). Input features of the training set were power transformed and training groups were created in which well 1, 2 and 3 and well 4, 5 and 6 (out of the 6-well plate multielectrode array) were combined, which were subsequently used for group K-fold cross validation to avoid leaking metrics between the train and validation set within each fold. The test set data was power transformed using only information from the training set. The synthetic minority oversampling technique (SMOTE) was used to remedy class imbalance.

Claims

PaVer / CSNOC / 870Claims1. A human cortico-striato-nigral circuit on chip comprising:i) a population of human striatal medium spiny neurons (MSNs) MSNs expressing MEIS2, BCL11B and GABAergic lineage markers GAD2 and DLX6-AS1, said MSNs thereby forming a medium spiny circuit, wherein said medium spiny circuit comprises at least two subpopulations of MSNs expressing dopaminergic DI- or D2- receptor;ii) a population of nigral dopaminergic neurons expressing FOXA1, LMX1A, EN1 and NR4A2, thereby forming a nigral dopaminergic circuit; andiii) a population of cortical glutamatergic neurons expressing LHX2 and SLC17A6 and / or SLC17A7, thereby forming cortical circuit;a multi-electrode array (MEA) chip for recording neuronal activity, said neuronal activity resulting at least in part from dopaminergic, GABAergic and / or glutamatergic neurotransmission;wherein said medium spiny neuronal circuit is spatially arranged onto said MEA chip for receiving excitatory and / or modulatory inputs from said cortical circuit iii) and / or said nigral dopaminergic circuit ii); andwherein each one of said populations i)-iii) are differentiated from a human pluripotent stem cell.

2. The human cortico-striato-nigral circuit on chip according to claim 1, wherein said human pluripotent stem cell is a human induced pluripotent stem cells (hiPSC).

3. The human cortico-striato-nigral circuit on chip according to any one of claims 1 or 2, wherein each one of said populations i)-iii) are differentiated from the human pluripotent stem cell derived from a single human subject.

4. The human cortico-striato-nigral circuit on chip according to any one of the preceding claims, wherein said chip is a high density MEA chip.

5. The human cortico-striato-nigral circuit on chip according to any one of the preceding claims, wherein said chip is a high density MEA chip, featuring a plurality of electrodes arranged to detect electrophysiological activity of said circuits i-iii).

6. The human cortico-striato-nigral circuit on chip according to any one of the preceding claims, wherein said at least one of populations i)-iii) comprises comprise at least one allele or genetic event associated with a neurologic or psychiatric disorder.

7. The human cortico-striato-nigral circuit on chip according to, wherein the neurologic or psychiatric disorder is selected from the group consisting of: Parkinson's disease, Huntington's56PaVer / CSNOC / 870disease, Multiple System Atrophy (MSA), motor dyskinesia, dystonia syndromes, Alzheimer disease, amyotrophic lateral sclerosis (ALS), Attention deficit hyperactivity disorder (ADHD), schizophrenia, affective disorder and autism spectrum disorder (ASD).

8. The human cortico-striato-nigral circuit on chip according to any one of the preceding claims, wherein said medium spiny neuronal circuit i) is spatially arranged on the MEA chip to receive excitatory stimuli form said cortical circuit iii) and modulatory stimuli from said nigral dopaminergic circuit ii).

9. The human cortico-striato-nigral circuit on chip according to any one of the preceding claims, wherein said circuits i-iii) are arranged so that an electrophysiological activity of at least circuit i), circuit ii) or circuit iii) is disrupted by alpha-synuclein.

10. A method for producing a cortico-striato-nigral circuit on chip according to any one claims 1-9, the method comprising:(a) providing (i) a lateral ganglionic eminence progenitor cell from a first human pluripotent stem cell; (ii) a ventral midbrain progenitor from a second human pluripotent stem cell; and (iii) a dorsal forebrain progenitor cell from a third human pluripotent stem cell;(b) seeding the lateral ganglionic eminence (i), ventral midbrain (ii), and dorsal forebrain (iii) progenitor cells at three different locations of the MEA chip, so that the lateral ganglionic eminence (i) progenitor cell is spatially arranged onto said MEA chip for receiving excitatory and / or modulatory inputs from dorsal forebrain iii) progenitor cell and / or ventral midbrain (ii) progenitor cell;(c) culturing the lateral ganglionic eminence (i), ventral midbrain (ii) and dorsal forebrain (iii) progenitor cells in conditions allowing for formation of differentiated populations of medium spiny (i), nigral dopaminergic (ii) and cortical glutamatergic (iii) neurons; and / or in conditions allowing for synaptic activity between the cells, thereby forming the medium spiny (i), nigral dopaminergic (ii) and cortical circuit (iii).

11. The method according to claim 10, wherein at least one of said differentiated, matured populations of medium spiny (i), nigral dopaminergic (ii) and cortical glutamatergic (iii) neurons comprise at least one allele or genetic event associated with a neurologic or psychiatric disorder.

12. The method according to claim 11, wherein the neurologic or psychiatric disorder is selected from the group consisting of: Parkinson's disease, Huntington's disease, Multiple System Atrophy (MSA), dystonia syndromes, motor dyskinesia, Alzheimer disease, amyotrophic lateral sclerosis (ALS), Attention deficit hyperactivity disorder (ADHD), schizophrenia, affective disorder and autism spectrum disorder (ASD).57PaVer / CSNOC / 87013. A method of determining the effect of a candidate agent on neuronal activity of the cortico- striato-nigral circuits on chip according to any one of claims 1 to 9 or obtainable by any one of claims 10 to 12, the method comprising:a) contacting the candidate agent with cortico-striato-nigral circuits on chip according to any one of claims 1 to 10 or obtainable by any one of claims 10 to 12;b) observing the change of in a neuronal activity and / or function of circuits i)-ii) and iii), c) determining the effect of the candidate agent based on resulting change in the neuronal activity and / or function of circuits i)-ii) and iii).

14. The method according to claim 13 wherein said candidate agent is tested for use in a neurologic or psychiatric disorder, further optionally wherein the neurologic or psychiatric disorder is selected from the group consisting of: Parkinson's disease, Huntington's disease, Multiple System Atrophy (MSA), motor dyskinesia, Alzheimer disease, amyotrophic lateral sclerosis (ALS), Attention deficit hyperactivity disorder (ADHD), schizophrenia, affective disorder and autism spectrum disorder (ASD).

15. The method according to any one of claims 13 or 14 wherein said candidate agent is a selective DI agonist.

16. A method to detect a genetic mutation or cell state preferably related to a disease and / or disorder, the method comprising the steps of:providing the human cortico-striato-nigral circuit on a chip according to any one of claims 1 to 9, or obtainable by the method according to any one of claims 10 to 12;recording the electrophysiological activity of at least one circuit of the human cortico-striato- nigral circuit on a chip according to any one of claims 1 to 9, or obtainable by the method according to any one of claims 10 to 12; andusing a computer implemented method to associate the recorded electrophysiological activity with presence of said genetic mutation or said state.58

Images