End-to-end deep learning approach to mouse behavior classification from cortex-wide calcium imaging.

Deep learning is a powerful tool for neural decoding, broadly applied to systems neuroscience and clinical studies. Interpretable and transparent models that can explain neural decoding for intended behaviors are crucial to identifying essential features of deep learning decoders in brain activity. In this study, we examine the performance of deep learning to classify mouse behavioral states from mesoscopic cortex-wide calcium imaging data. Our convolutional neural network (CNN)-based end-to-end decoder combined with recurrent neural network (RNN) classifies the behavioral states with high accuracy and robustness to individual differences on temporal scales of sub-seconds. Using the CNN-RNN decoder, we identify that the forelimb and hindlimb areas in the somatosensory cortex significantly contribute to behavioral classification. Our findings imply that the end-to-end approach has the potential to be an interpretable deep learning method with unbiased visualization of critical brain regions.

Social circuits and their dysfunction in autism spectrum disorder.

Social behaviors, how individuals act cooperatively and competitively with conspecifics, are widely seen across species. Rodents display various social behaviors, and many different behavioral paradigms have been used for investigating their neural circuit bases. Social behavior is highly vulnerable to brain network dysfunction caused by neurological and neuropsychiatric conditions such as autism spectrum disorders (ASDs). Studying mouse models of ASD provides a promising avenue toward elucidating mechanisms of abnormal social behavior and potential therapeutic targets for treatment. In this review, we outline recent progress and key findings on neural circuit mechanisms underlying social behavior, with particular emphasis on rodent studies that monitor and manipulate the activity of specific circuits using modern systems neuroscience approaches. Social behavior is mediated by a distributed brain-wide network among major cortical (e.g., medial prefrontal cortex (mPFC), anterior cingulate cortex, and insular cortex (IC)) and subcortical (e.g., nucleus accumbens, basolateral amygdala (BLA), and ventral tegmental area) structures, influenced by multiple neuromodulatory systems (e.g., oxytocin, dopamine, and serotonin). We particularly draw special attention to IC as a unique cortical area that mediates multisensory integration, encoding of ongoing social interaction, social decision-making, emotion, and empathy. Additionally, a synthesis of studies investigating ASD mouse models demonstrates that dysfunctions in mPFC-BLA circuitry and neuromodulation are prominent. Pharmacological rescues by local or systemic (e.g., oral) administration of various drugs have provided valuable clues for developing new therapeutic agents for ASD. Future efforts and technological advances will push forward the next frontiers in this field, such as the elucidation of brain-wide network activity and inter-brain neural dynamics during real and virtual social interactions, and the establishment of circuit-based therapy for disorders affecting social functions.

[Analysis of cortical network dynamics in the behavioral state of a mouse model of autism].

Autism spectrum disorder (ASD) is a neurodevelopmental disorder characterized by behavioral abnormalities such as poor social communication and stereotyped/repetitive behaviors. Functional dynamics among multiple cortical areas are associated with processing sensory information and planning and executing behavioral expressions. However, the reconfiguration of large-scale functional network dynamics during behaviors remains to be elucidated in ASD. In this review, we describe our virtual reality (VR) based real-time imaging system which allowed us to investigate wide-field cortical activity in voluntarily behaving mice. We previously generated a mouse model of ASD with chromosome 15q11-13 duplication (15q dup), one of the most frequent genomic abnormalities, and reported that 15q dup mice display ASD-like behaviors. Using this system, we examined the functional cortical network during behaviors in 15q dup mice. Pair-wise correlation of cortical area activity on a time scale of a second was calculated to represent the dynamic state of cortical functional connectivity (FC). A graph theoretical network analysis was then conducted to illustrate rapid and robust behavior-state-dependent cortical network reconfiguration. Our VR-based real-time imaging system provides invaluable information to understand FC dynamics linked to a behavioral abnormality of neuropsychiatric disorders.

Virtual reality-based real-time imaging reveals abnormal cortical dynamics during behavioral transitions in a mouse model of autism.

Functional connectivity (FC) can provide insight into cortical circuit dysfunction in neuropsychiatric disorders. However, dynamic changes in FC related to locomotion with sensory feedback remain to be elucidated. To investigate FC dynamics in locomoting mice, we develop mesoscopic Ca2+ imaging with a virtual reality (VR) environment. We find rapid reorganization of cortical FC in response to changing behavioral states. By using machine learning classification, behavioral states are accurately decoded. We then use our VR-based imaging system to study cortical FC in a mouse model of autism and find that locomotion states are associated with altered FC dynamics. Furthermore, we identify FC patterns involving the motor area as the most distinguishing features of the autism mice from wild-type mice during behavioral transitions, which might correlate with motor clumsiness in individuals with autism. Our VR-based real-time imaging system provides crucial information to understand FC dynamics linked to behavioral abnormality of neuropsychiatric disorders.

Genetic dissection identifies Necdin as a driver gene in a mouse model of paternal 15q duplications.

Maternally inherited duplication of chromosome 15q11-q13 (Dup15q) is a pathogenic copy number variation (CNV) associated with autism spectrum disorder (ASD). Recently, paternally derived duplication has also been shown to contribute to the development of ASD. The molecular mechanism underlying paternal Dup15q remains unclear. Here, we conduct genetic and overexpression-based screening and identify Necdin (Ndn) as a driver gene for paternal Dup15q resulting in the development of ASD-like phenotypes in mice. An excess amount of Ndn results in enhanced spine formation and density as well as hyperexcitability of cortical pyramidal neurons. We generate 15q dupΔNdn mice with a normalized copy number of Ndn by excising its one copy from Dup15q mice using a CRISPR-Cas9 system. 15q dupΔNdn mice do not show ASD-like phenotypes and show dendritic spine dynamics and cortical excitatory-inhibitory balance similar to wild type animals. Our study provides an insight into the role of Ndn in paternal 15q duplication and a mouse model of paternal Dup15q syndrome.

Optogenetic Approaches to Understand the Neural Circuit Mechanism of Social Deficits Seen in Autism Spectrum Disorders.

Individuals with neurodevelopmental disorders, such as autism spectrum disorders (ASDs), are diagnosed based on nonquantitative objective parameters such as behavioral phenotypes. It is still unclear how any neural mechanism affects such behavioral phenotypes in these patients. In human genetics, a large number of genetic abnormalities including single nucleotide variation (SNV) and copy number variation (CNV) have been found in individuals with ASDs. It is thought that influence of such variations converges on dysfunction of neural circuit resulting in common behavioral phenotypes of ASDs such as deficits in social communication and interaction. Recent studies suggest that an excitatory/inhibitory (E/I) imbalanced state, which induces disruption of neural circuit activities, is one of the pathophysiological abnormalities in ASD brains. To assess the causal relationship between brain abnormalities and behavioral deficits, we can take advantage of optogenetics with animal models of ASDs that recapitulate human genetic mutations. Here, we review optogenetics studies being utilized to dissect neural circuit mechanisms associated with social deficits in model mice of ASD. Optogenetic manipulation of disrupted neural activities would help us understand how neural circuits affect behavioral deficits observed in ASDs.

Imaging the Neural Circuit Basis of Social Behavior: Insights from Mouse and Human Studies.

Social behavior includes a variety of behaviors that are expressed between two or more individuals. In humans, impairment of social function (i.e., social behavior and social cognition) is seen in neurodevelopmental and neurological disorders including autism spectrum disorders (ASDs) and stroke, respectively. In basic neuroscience research, fluorescence monitoring of neural activity, such as immediate early gene (IEG)-mediated whole-brain mapping, fiber photometry, and calcium imaging using a miniaturized head-mounted microscope or a two-photon microscope, and non-fluorescence imaging such as functional magnetic resonance imaging (fMRI) are increasingly used to measure the activity of many neurons and multiple brain areas in animals during social behavior. In this review, we overview recent rodent studies that have investigated the dynamics of brain activity during social behavior at the whole-brain and local circuit levels and studies that explored the neural basis of social function in healthy, in brain-injured, and in autistic human subjects. A synthesis of such findings will advance our understanding of brain mechanisms underlying social behavior and facilitate the development of pharmaceutical and functional neurosurgical interventions for brain disorders affecting social function.

Encoding of social exploration by neural ensembles in the insular cortex.

The insular cortex (IC) participates in diverse complex brain functions, including social function, yet their cellular bases remain to be fully understood. Using microendoscopic calcium imaging of the agranular insular cortex (AI) in mice interacting with freely moving and restrained social targets, we identified 2 subsets of AI neurons-a larger fraction of "Social-ON" cells and a smaller fraction of "Social-OFF" cells-that change their activity in opposite directions during social exploration. Social-ON cells included those that represented social investigation independent of location and consisted of multiple subsets, each of which was preferentially active during exploration under a particular behavioral state or with a particular target of physical contact. These results uncover a previously unknown function of AI neurons that may act to monitor the ongoing status of social exploration while an animal interacts with unfamiliar conspecifics.

Behavioral neuroscience of autism.

Autism spectrum disorder (ASD) is a neurodevelopmental disorder. Several genetic causes of ASD have been identified and this has enabled researchers to construct mouse models. Mouse behavioral tests reveal impaired social interaction and communication, as well as increased repetitive behavior and behavioral inflexibility in these mice, which correspond to core behavioral deficits observed in individuals with ASD. However, the connection between these behavioral abnormalities and the underlying dysregulation in neuronal circuits and synaptic function is poorly understood. Moreover, different components of the ASD phenotype may be linked to dysfunction in different brain regions, making it even more challenging to chart the pathophysiological mechanisms involved in ASD. Here we summarize the research on mouse models of ASD and their contribution to understanding pathophysiological mechanisms. Specifically, we emphasize abnormal serotonin production and regulation, as well as the disruption in circadian rhythms and sleep that are observed in a subset of ASD, and propose that spatiotemporal disturbances in brainstem development may be a primary cause of ASD that propagates towards the cerebral cortex.

Altered microbiota composition reflects enhanced communication in 15q11-13 CNV mice.

Autism spectrum disorder (ASD) is a complex and heterogeneous neurodevelopmental disorder. In addition to the core symptoms of ASD, many patients with ASD also show comorbid gut dysbiosis, which may lead to various gastrointestinal (GI) problems. Intriguingly, there is evidence that gut microbiota communicate with the central nervous system to modulate behavioral output through the gut-brain axis. To investigate how the microbiota composition is changed in ASD and to identify which microbes are involved in autistic behaviors, we performed a 16S rRNA gene-based metagenomics analysis in an ASD mouse model. Here, we focused on a model with human 15q11-13 duplication (15q dup), the most frequent chromosomal aberration or copy number variation found in ASD. Species diversity of the microbiome was significantly decreased in 15q dup mice. A combination of antibiotics treatment and behavioral analysis showed that neomycin improved social communication in 15q dup mice. Furthermore, comparison of the microbiota composition of mice treated with different antibiotics enabled us to identify beneficial operational taxonomic units (OTUs) for ultrasonic vocalization.

Common Defects of Spine Dynamics and Circuit Function in Neurodevelopmental Disorders: A Systematic Review of Findings From in Vivo Optical Imaging of Mouse Models.

In vivo optical imaging is a powerful tool for revealing brain structure and function at both the circuit and cellular levels. Here, we provide a systematic review of findings obtained from in vivo imaging studies of mouse models of neurodevelopmental disorders, including the monogenic disorders fragile X syndrome, Rett syndrome, and Angelman syndrome, which are caused by genetic abnormalities of FMR1, MECP2, and UBE3A, as well as disorders caused by copy number variations (15q11-13 duplication and 22q11.2 deletion) and BTBR mice as an inbred strain model of autism spectrum disorder (ASD). Most studies visualize the structural and functional responsiveness of cerebral cortical neurons to sensory stimuli and the developmental and experience-dependent changes in these responses as a model of brain functions affected by these disorders. The optical imaging techniques include two-photon microscopy of fluorescently labeled dendritic spines or neurons loaded with fluorescent calcium indicators and macroscopic imaging of cortical activity using calcium indicators, voltage-sensitive dyes or intrinsic optical signals. Studies have revealed alterations in the density, stability, and turnover of dendritic spines, aberrant cortical sensory responses, impaired inhibitory function, and concomitant failure of circuit maturation as common causes for neurological deficits. Mechanistic hypotheses derived from in vivo imaging also provide new directions for therapeutic interventions. For instance, it was recently demonstrated that early postnatal administration of a selective serotonin reuptake inhibitor (SSRI) restores impaired cortical inhibitory function and ameliorates the aberrant social behaviors in a mouse model of ASD. We discuss the potential use of SSRIs for treating ASDs in light of these findings.

Serotonin rebalances cortical tuning and behavior linked to autism symptoms in 15q11-13 CNV mice.

Serotonin is a critical modulator of cortical function, and its metabolism is defective in autism spectrum disorder (ASD) brain. How serotonin metabolism regulates cortical physiology and contributes to the pathological and behavioral symptoms of ASD remains unknown. We show that normal serotonin levels are essential for the maintenance of neocortical excitation/inhibition balance, correct sensory stimulus tuning, and social behavior. Conversely, low serotonin levels in 15q dup mice (a model for ASD with the human 15q11-13 duplication) result in impairment of the same phenotypes. Restoration of normal serotonin levels in 15q dup mice revealed the reversibility of a subset of ASD-related symptoms in the adult. These findings suggest that serotonin may have therapeutic potential for discrete ASD symptoms.

Translocated in liposarcoma regulates the distribution and function of mammalian enabled, a modulator of actin dynamics.

Translocated in liposarcoma/fused in sarcoma (TLS/FUS) is an RNA-binding protein that regulates the splicing pattern of mRNA transcripts and is known to cause a type of familial amyotrophic lateral sclerosis (ALS). In the absence of TLS, Mammalian enabled (Mena), an actin-regulatory protein and a target of TLS, undergoes preferential alternative splicing. In the present study, we show that the ablation of TLS dysregulates the subcellular location and functions of Mena. When TLS knockout (KO) mouse embryonic fibroblasts (MEFs) were transfected with wild-type Mena, it no longer accumulated at focal adhesions and peripheral structures, whereas the localization of the alternatively spliced form was maintained. Additionally, the ability of Mena to suppress the motility of cells was lost in TLS KO MEFs. Moreover, Mena failed to promote neurite outgrowth in TLS KO primary neurons. Taken together, TLS is intimately involved in the local cytoskeletal dynamics surrounding Mena in both fibroblasts and neurons. The robust change in cytoskeletal dynamics, as indicated by the dysregulation of Mena in TLS KO cells, provides a new insight into the pathogenesis of certain types of ALS.

Autism spectrum disorder model mice: Focus on copy number variation and epigenetics.

Autism spectrum disorder (ASD) is gathering concerns in socially developed countries. ASD is a neuropsychiatric disorder of genetic origin with high prevalence of 1%-2%. The patients with ASD characteristically show impaired social skills. Today, many genetic studies identify numerous susceptible genes and genetic loci associated with ASD. Although some genetic factors can lead to abnormal brain function linked to ASD phenotypes, the pathogenic mechanism of ASD is still unclear. Here, we discuss a new mouse model for ASD as an advanced tool to understand the mechanism of ASD.

Neuroanatomical Phenotypes Are Consistent With Autism-Like Behavioral Phenotypes in the 15q11-13 Duplication Mouse Model.

Paternally and maternally inherited deletions and duplications of human chromosome 15q11-13 are relatively common in the human population. Furthermore, duplications in the 15q region are often associated with autism. Both maternal and paternal interstitial 15q11-13 duplication mouse models have been previously created, where several behavioral differences were found in the paternal duplication (patDp/+) mouse but not in the maternal duplication (matDp/+). These included decreased sociability, behavioral inflexibility, abnormal ultrasonic vocalizations, decreased spontaneous activity, and increased anxiety. Similarly, in the current study, we found several anatomical differences in the patDp/+ mice that were not seen in the matDp/+ mice. Regional differences that are evident only in the paternal duplication are a smaller dentate gyrus and smaller medial striatum. These differences may be responsible for the behavioral inflexibility. Furthermore, a smaller dorsal raphe nucleus could be responsible for the reported serotonin defects. This study highlights consistency that can be found between behavioral and anatomical phenotyping.

Decreased exploratory activity in a mouse model of 15q duplication syndrome; implications for disturbance of serotonin signaling.

Autism spectrum disorders (ASDs) have garnered significant attention as an important grouping of developmental brain disorders. Recent genomic studies have revealed that inherited or de novo copy number variations (CNVs) are significantly involved in the pathophysiology of ASDs. In a previous report from our laboratory, we generated mice with CNVs as a model of ASDs, with a duplicated mouse chromosome 7C that is orthologous to human chromosome 15q11-13. Behavioral analyses revealed paternally duplicated (patDp/+) mice displayed abnormal behaviors resembling the symptoms of ASDs. In the present study, we extended these findings by performing various behavioral tests with C57BL/6J patDp/+ mice, and comprehensively measuring brain monoamine levels with ex vivo high performance liquid chromatography. Compared with wild-type controls, patDp/+ mice exhibited decreased locomotor and exploratory activities in the open field test, Y-maze test, and fear-conditioning test. Furthermore, their decreased activity levels overcame increased appetite induced by 24 hours of food deprivation in the novelty suppressed feeding test. Serotonin levels in several brain regions of adult patDp/+ mice were lower than those of wild-type control, with no concurrent changes in brain levels of dopamine or norepinephrine. Moreover, analysis of monoamines in postnatal developmental stages demonstrated reduced brain levels of serotonin in young patDp/+ mice. These findings suggest that a disrupted brain serotonergic system, especially during postnatal development, may generate the phenotypes of patDp/+ mice.

Fezf1 is required for penetration of the basal lamina by olfactory axons to promote olfactory development.

In the development of the olfactory system, olfactory receptor neurons (ORNs) project their axons from the olfactory epithelium (OE) to the olfactory bulb (OB). The surface of the OB is covered by the central nervous system (CNS) basal lamina. To establish this connection, pioneer axons of the ORNs penetrate the CNS basal lamina at embryonic day 12.5 in mice. The importance of this penetration is highlighted by the Kallmann syndrome. However, little has been known about the molecular mechanism underlying this penetration process. Fezf1 (also called as Fez, Zfp312-like, and 3110069A13Rik) is a C2H2-type zinc-finger gene expressed in the OE and hypothalamic region in mice. In Fezf1-deficient mice, ORN axons (olfactory axons) do not reach the OB. Here we demonstrate that Fezf1-deficient olfactory axons do not penetrate the CNS basal lamina in vivo, and the penetration activity of the axons in Matrigel is impaired in vitro. Coculture experiments using the OE and OB reveal that axonal projection of ORNs is rescued in Fezf1-deficient mice in which the meninges including the CNS basal lamina are removed from the mutant OB. These data indicate that Fezf1 is required for the penetration of olfactory axons through the CNS basal lamina before they innervate the OB.