An in vitro model of neuronal ensembles

Abstract

Advances in 3D neuronal cultures, such as brain spheroids and organoids, are allowing unprecedented in vitro access to some of the molecular, cellular and developmental mechanisms underlying brain diseases. However, their efficacy in recapitulating brain network properties that encode brain function remains limited, thereby precluding development of effective in vitro models of complex brain disorders like schizophrenia. Here, we develop and characterize a Modular Neuronal Network (MoNNet) approach that recapitulates specific features of neuronal ensemble dynamics, segregated local-global network activities and a hierarchical modular organization. We utilized MoNNets for quantitative in vitro modelling of schizophrenia-related network dysfunctions caused by highly penetrant mutations in SETD1A and 22q11.2 risk loci. Furthermore, we demonstrate its utility for drug discovery by performing pharmacological rescue of alterations in neuronal ensembles stability and global network synchrony. MoNNets allow in vitro modelling of brain diseases for investigating the underlying neuronal network mechanisms and systematic drug discovery.

Introduction

Recent advances in three-dimensional (3D) neuronal cultures are providing effective means for non-invasive modeling of specific neuronal processes in vitro^1,2,3,4,5,6. For example, the brain spheroids and organoids have been successfully used to investigate the cell biological and developmental mechanisms underlying autism spectrum disorders (ASD)^7,8. Despite this progress, the state-of-the-art 3D neuronal cultures still lack complex (i.e. beyond single neuronal patterns) features^9,10 that are reminiscent of in vivo brain functional networks, limiting their utility for in vitro modeling of brain diseases such as schizophrenia (SCZ) and other neuropsychiatric disorders, where local and global network dysfunction appears to be a common convergence point of diverse etiological factors^11,12,13,14. Recent approaches of increasing the size of 3D neuronal cultures^8,15 may further increase the captured functional complexity, however, the inaccessibility of nutrients and oxygen supply to the inner cores remains a major technical hurdle. Moreover, the enhanced size may further limit the reagents penetrability and optical accessibility.

We explored an alternate 3D cell-culture approach to mitigate some of these limitations, with an eventual goal to develop quantitative in vitro models of the compromised signal propagation in brain networks associated with SCZ and other neuropsychiatric disorders¹². Our approach builds upon the facts that the mammalian brain architecture is hierarchically modular with small-world network architecture i.e. highly intra-connected modules (specific brain regions) with fewer inter-modular connections across the system¹⁶. For example, the dominant successful computational models simulate cortical function by implementing brain networks as highly intra-connected modules of neurons with fewer inter-modular connections¹⁷. To speedily test and validate the approach by robust comparisons to an extensive in vivo knowledge base, we used dissociated hippocampal cells from late stage mouse embryos to develop self-organized networks of spheroid-like modular units, termed Modular Neuronal Network (MoNNet). We performed exhaustive molecular, cellular and network function characterization of MoNNets by RNAseq experiments, immunostaining, systems-level cellular-resolution Ca²⁺ imaging over several weeks of culture age, and systematic pharmacological interventions. These experiments revealed diverse neuronal activity patterns, segregated local-global network activity, formation and maintenance of stable ensembles/modules and a hierarchical organization of modules with varied strength functional connections.

We then used the MoNNet approach to develop and characterize in vitro quantitative models of network dysfunctions induced by two bona fide SCZ-risk mutations. SCZ is characterized by psychosis, cognitive symptoms as well as negative symptoms such as social and emotional withdrawal. Research in humans and model organisms has identified a diverse set of changes in structural connectivity, synaptic plasticity and excitatory–inhibitory (E–I) balance as well as altered neuromodulation^18,19 reflecting the vast diversity of genetic risk²⁰. Acting alone or in combination, such changes may lead to local and long-range degradation of the ordered local and global connectivity and functional synchrony of neuronal networks needed to support the cognitive and perceptual operations, and hippocampal-dependent forms of memory (such as episodic or spatial working memory) affected in SCZ^21,22. Despite the onset of SCZ-associated psychosis in adolescence or adulthood, schizophrenia also has a strong neurodevelopmental component much before the presentation of psychotic symptoms. Therefore, it is critical to investigate the neuronal deficiencies in development as well as their manifestation in the mature functional networks.

We aimed to develop in vitro models of SCZ-related network pathophysiology by utilizing two well-characterized genetically engineered mouse models of SCZ-risk genes^{11,14,23,24,25}. First, Setd1a+/− mice, which model loss-of-function mutation in SETD1A, a lysine-methyltransferase that modulates the expression of a large number of genes expressed brain-wide but especially in the neocortex²⁴, in large part by mediating methylation on lysine 4 on the histone H3 protein (H3K4)^26,27. Second, Df(16)A+/− mice, which model a highly penetrant microdeletion in human chromosome 22q11.2 locus resulting in sporadic cases of SCZ in ~30% of carriers²⁸. Both mutations show the strongest association with SCZ in respective exome sequencing and structural variant scans agnostic to the genotype^29,30,31 (https://schema.broadinstitute.org/results) and, while mutation carriers exhibit additional behavioral and cognitive symptoms, SCZ is the most prominent psychiatric diagnosis for both mutations. These genetic models were recently shown to recapitulate the cognitive and circuitry deficits generally associated with SCZ^{11,14,23,24,32,33}. We derived MoNNet (SCZ-MoNNet) preparations from these systems, and performed exhaustive comparative characterization over several weeks, revealing significant alterations in the stability of modules/ensembles formation, much-reduced global network synchrony with much smaller impact on local synchrony, and lack of hierarchical modular organization due to severely reduced inter-modular functional interactions. To reveal the underlying molecular mechanisms, we performed RNAseq experiments for MoNNets derived from Setd1a+/− and WT siblings. The comparative analysis of gene expressions in Setd1a+/− vs. WT sibling MoNNets revealed dysregulation of cytoskeleton/neuronal structure and synapse signaling/function-related genes, in striking correlation with the alterations in structural connectivity, synaptic plasticity and neuromodulation generally associated with SCZ^18,19. Furthermore, we demonstrate the utility of this approach for drug discovery by performing retrospectively predictive pharmacological rescue experiments with antagonists of LSD1 demethylase activity (ORY-1001 and TCP) which were recently shown to effectively rescue some of the cognitive behavioral abnormalities in Setd1a+/− mice²⁴. We found that treatment with these compounds promoted stabilization of functional connections and modules in the later mature stages, revealing their network-level mechanisms of action…