Visualizing and Modulating Mitophagy for Therapeutic Studies of Neurodegeneration.

Dysfunctional mitochondria accumulate in many human diseases. Accordingly, mitophagy, which removes these mitochondria through lysosomal degradation, is attracting broad attention. Due to uncertainties in the operational principles of conventional mitophagy probes, however, the specificity and quantitativeness of their readouts are disputable. Thorough investigation of the behaviors and fates of fluorescent proteins inside and outside lysosomes enabled us to develop an indicator for mitophagy, mito-SRAI. Through strict control of its mitochondrial targeting, we were able to monitor mitophagy in fixed biological samples more reproducibly than before. Large-scale image-based high-throughput screening led to the discovery of a hit compound that induces selective mitophagy of damaged mitochondria. In a mouse model of Parkinsons disease, we found that dopaminergic neurons selectively failed to execute mitophagy that promoted their survival within lesions. These results show that mito-SRAI is an essential tool for quantitative studies of mitochondrial quality control.

A hypothalamic novelty signal modulates hippocampal memory.

The ability to recognize information that is incongruous with previous experience is critical for survival. Novelty signals have therefore evolved in the mammalian brain to enhance attention, perception and memory1,2. Although the importance of regions such as the ventral tegmental area3,4 and locus coeruleus5 in broadly signalling novelty is well-established, these diffuse monoaminergic transmitters have yet to be shown to convey specific information on the type of stimuli that drive them. Whether distinct types of novelty, such as contextual and social novelty, are differently processed and routed in the brain is unknown. Here we identify the supramammillary nucleus (SuM) as a novelty hub in the hypothalamus6. The SuM region is unique in that it not only responds broadly to novel stimuli, but also segregates and selectively routes different types of information to discrete cortical targets-the dentate gyrus and CA2 fields of the hippocampus-for the modulation of mnemonic processing. Using a new transgenic mouse line, SuM-Cre, we found that SuM neurons that project to the dentate gyrus are activated by contextual novelty, whereas the SuM-CA2 circuit is preferentially activated by novel social encounters. Circuit-based manipulation showed that divergent novelty channelling in these projections modifies hippocampal contextual or social memory. This content-specific routing of novelty signals represents a previously unknown mechanism that enables the hypothalamus to flexibly modulate select components of cognition.

Ras-like Gem GTPase induced by Npas4 promotes activity-dependent neuronal tolerance for ischemic stroke.

Ischemic stroke, which results in loss of neurological function, initiates a complex cascade of pathological events in the brain, largely driven by excitotoxic Ca2+ influx in neurons. This leads to cortical spreading depolarization, which induces expression of genes involved in both neuronal death and survival; yet, the functions of these genes remain poorly understood. Here, we profiled gene expression changes that are common to ischemia (modeled by middle cerebral artery occlusion [MCAO]) and to experience-dependent activation (modeled by exposure to an enriched environment [EE]), which also induces Ca2+ transients that trigger transcriptional programs. We found that the activity-dependent transcription factor Npas4 was up-regulated under MCAO and EE conditions and that transient activation of cortical neurons in the healthy brain by the EE decreased cell death after stroke. Furthermore, both MCAO in vivo and oxygen-glucose deprivation in vitro revealed that Npas4 is necessary and sufficient for neuroprotection. We also found that this protection involves the inhibition of L-type voltage-gated Ca2+ channels (VGCCs). Next, our systematic search for Npas4-downstream genes identified Gem, which encodes a Ras-related small GTPase that mediates neuroprotective effects of Npas4. Gem suppresses the membrane localization of L-type VGCCs to inhibit excess Ca2+ influx, thereby protecting neurons from excitotoxic death after in vitro and in vivo ischemia. Collectively, our findings indicate that Gem expression via Npas4 is necessary and sufficient to promote neuroprotection in the injured brain. Importantly, Gem is also induced in human cerebral organoids cultured under an ischemic condition, revealing Gem as a new target for drug discovery.

High-fat diet-induced activation of SGK1 promotes Alzheimer's disease-associated tau pathology.

Type 2 diabetes mellitus (T2DM) has long been considered a risk factor for Alzheimer's disease (AD). However, the molecular links between T2DM and AD remain obscure. Here, we reported that serum-/glucocorticoid-regulated kinase 1 (SGK1) is activated by administering a chronic high-fat diet (HFD), which increases the risk of T2DM, and thus promotes Tau pathology via the phosphorylation of tau at Ser214 and the activation of a key tau kinase, namely, GSK-3ß, forming SGK1-GSK-3ß-tau complex. SGK1 was activated under conditions of elevated glucocorticoid and hyperglycemia associated with HFD, but not of fatty acid-mediated insulin resistance. Elevated expression of SGK1 in the mouse hippocampus led to neurodegeneration and impairments in learning and memory. Upregulation and activation of SGK1, SGK1-GSK-3ß-tau complex were also observed in the hippocampi of AD cases. Our results suggest that SGK1 is a key modifier of tau pathology in AD, linking AD to corticosteroid effects and T2DM.

Uts2b is a microbiota-regulated gene expressed in vagal afferent neurons connected to enteroendocrine cells producing cholecystokinin.

Enteroendocrine cells (EECs) are the primary sensory cells that sense the gut luminal environment and secret hormones to regulate organ function. Recent studies revealed that vagal afferent neurons are connected to EECs and relay sensory information from EECs to the brain stem. To date, however, the identity of vagal afferent neurons connected to a given EEC subtype and the mode of their gene responses to its intestinal hormone have remained unknown. Hypothesizing that EEC-associated vagal afferent neurons change their gene expression in response to the microbiota-related extracellular stimuli, we conducted comparative gene expression analyses of the nodose-petrosal ganglion complex (NPG) using specific pathogen-free (SPF) and germ-free (GF) mice. We report here that the Uts2b gene, which encodes a functionally unknown neuropeptide, urotensin 2B (UTS2B), is expressed in a microbiota-dependent manner in NPG neurons. In cultured NPG neurons, expression of Uts2b was induced by AR420626, the selective agonist for FFAR3. Moreover, distinct gastrointestinal hormones exerted differential effects on Uts2b expression in NPG neurons, where cholecystokinin (CCK) significantly increased its expression. The majority of Uts2b-expressing NPG neurons expressed CCK-A, the receptor for CCK, which comprised approximately 25% of all CCK-A-expressing NPG neurons. Selective fluorescent labeling of Uts2b-expressing NPG neurons revealed a direct contact of their nerve fibers to CCK-expressing EECs. This study identifies the Uts2b as a microbiota-regulated gene, demonstrates that Uts2b-expressing vagal afferent neurons transduce sensory information from CCK-expressing EECs to the brain, and suggests potential involvement of UTS2B in a modality of CCK actions.

Generation of a MOR-CreER knock-in mouse line to study cells and neural circuits involved in mu opioid receptor signaling.

Mu opioid receptor (MOR) is involved in various brain functions, such as pain modulation, reward processing, and addictive behaviors, and mediates the main pharmacologic effects of morphine and other opioid compounds. To gain genetic access to MOR-expressing cells, and to study physiological and pathological roles of MOR signaling, we generated a MOR-CreER knock-in mouse line, in which the stop codon of the Oprm1 gene was replaced by a DNA fragment encoding a T2A peptide and tamoxifen (Tm)-inducible Cre recombinase. We show that the MOR-CreER allele undergoes Tm-dependent recombination in a discrete subtype of neurons that express MOR in the adult nervous system, including the olfactory bulb, cerebral cortex, striosome compartments in the striatum, hippocampus, amygdala, thalamus, hypothalamus, interpeduncular nucleus, superior and inferior colliculi, periaqueductal gray, parabrachial nuclei, cochlear nucleus, raphe nuclei, pontine and medullary reticular formation, ambiguus nucleus, solitary nucleus, spinal cord, and dorsal root ganglia. The MOR-CreER mouse line combined with a Cre-dependent adeno-associated virus vector enables robust gene manipulation in the MOR-enriched striosomes. Furthermore, Tm treatment during prenatal development effectively induces Cre-mediated recombination. Thus, the MOR-CreER mouse is a powerful tool to study MOR-expressing cells with conditional gene manipulation in developing and mature neural tissues.

Parvalbumin-producing striatal interneurons receive excitatory inputs onto proximal dendrites from the motor thalamus in male mice.

In rodents, the dorsolateral striatum regulates voluntary movement by integrating excitatory inputs from the motor-related cerebral cortex and thalamus to produce contingent inhibitory output to other basal ganglia nuclei. Striatal parvalbumin (PV)-producing interneurons receiving this excitatory input then inhibit medium spiny neurons (MSNs) and modify their outputs. To understand basal ganglia function in motor control, it is important to reveal the precise synaptic organization of motor-related cortical and thalamic inputs to striatal PV interneurons. To examine which domains of the PV neurons receive these excitatory inputs, we used male bacterial artificial chromosome transgenic mice expressing somatodendritic membrane-targeted green fluorescent protein in PV neurons. An anterograde tracing study with the adeno-associated virus vector combined with immunodetection of pre- and postsynaptic markers visualized the distribution of the excitatory appositions on PV dendrites. Statistical analysis revealed that the density of thalamostriatal appositions along the dendrites was significantly higher on the proximal than distal dendrites. In contrast, there was no positional preference in the density of appositions from axons of the dorsofrontal cortex. Population observations of thalamostriatal and corticostriatal appositions by immunohistochemistry for pathway-specific vesicular glutamate transporters confirmed that thalamic inputs preferentially, and cortical ones less preferentially, made apposition on proximal dendrites of PV neurons. This axodendritic organization suggests that PV neurons produce fast and reliable inhibition of MSNs in response to thalamic inputs and process excitatory inputs from motor cortices locally and plastically, possibly together with other GABAergic and dopaminergic dendritic inputs, to modulate MSN inhibition.

Ventral medial nucleus neurons send thalamocortical afferents more widely and more preferentially to layer 1 than neurons of the ventral anterior-ventral lateral nuclear complex in the rat.

Not only inhibitory afferent-dominant zone (IZ) of the ventral anterior-ventral lateral thalamic complex (VA-VL) but also the ventral medial nucleus (VM) is known to receive strong inputs from the basal ganglia and send axons to motor areas. We previously reported differences in axonal arborization between IZ neurons and the other VA-VL neurons in rats by single-neuron tracing with viral vectors. In the present study, the axonal arborization of single VM neurons was visualized by the same method, and compared with that of IZ neurons. VM neurons formed fewer axon collaterals in the striatum, but sent axon fibers more widely and more preferentially (79% of fibers) to layer 1 of cortical areas than IZ neurons. Furthermore, the VM seemed to contain at least 2 types of neurons; a major population of VM neurons sent axon fibers principally to motor-associated areas as VA-VL neurons did, and the other population projected mainly to orbital or cingulate areas. Although both VM and IZ neurons receive strong basal ganglia inputs, these results suggest that VM neurons, at a single neuron level, innervate the apical dendrites of cortical pyramidal neurons more intensely and more widely than IZ neurons.

Different cortical projections from three subdivisions of the rat lateral posterior thalamic nucleus: a single-neuron tracing study with viral vectors.

The lateral posterior thalamic nucleus (LP) is one of the components of the extrageniculate pathway in the rat visual system, and is cytoarchitecturally divided into three subdivisions--lateral (LPl), rostromedial (LPrm), and caudomedial (LPcm) portions. To clarify the differences in the dendritic fields and axonal arborisations among the three subdivisions, we applied a single-neuron labeling technique with viral vectors to LP neurons. The proximal dendrites of LPl neurons were more numerous than those of LPrm and LPcm neurons, and LPrm neurons tended to have wider dendritic fields than LPl neurons. We then analysed the axonal arborisations of LP neurons by reconstructing the axon fibers in the cortex. The LPl, LPrm and LPcm were different from one another in terms of the projection targets--the main target cortical regions of LPl and LPrm neurons were the secondary and primary visual areas, whereas those of LPcm neurons were the postrhinal and temporal association areas. Furthermore, the principal target cortical layers of LPl neurons in the visual areas were middle layers, but that of LPrm neurons was layer 1. This indicates that LPl and LPrm neurons can be categorised into the core and matrix types of thalamic neurons, respectively, in the visual areas. In addition, LPl neurons formed multiple axonal clusters within the visual areas, whereas the fibers of LPrm neurons were widely and diffusely distributed. It is therefore presumed that these two types of neurons play different roles in visual information processing by dual thalamocortical innervation of the visual areas.

Efficient and graded gene expression in glia and neurons of primary cerebellar cultures transduced by lentiviral vectors.

Lentiviral vectors are valuable tools to express genes of interest in living animals and stem cell cultures. The use of promoters in lentiviral constructs has been successfully used to drive gene expression in particular cell types including neurons and glia of the central nervous system in vivo. However, their suitability in cell culture is less well documented. In this paper, we describe lentiviral vectors containing neuronal promoters of the murine stem cell virus, of the synapsin 1 gene, the tubulin alpha 1 gene, and the calmodulin kinase II gene, and the glial promoter of the glial fibrillary acidic protein gene to drive reporter gene expression in primary dissociated cerebellar cell cultures and in slice cultures. While the glial promoter was highly specific for glia, the neuronal promoters were active in neurons and glia of dissociated cultures to a comparable extent. In slice cultures, neuronal and glial promoters demonstrated higher, but not absolute selectivity for particular cell types. In addition, the promoters allowed for an efficient and graded expression of genes in dissociated cultures. By using selected combinations of vectors, it was also possible to drive the expression of two genes in one cell type with high efficiency. A gene of interest in combination with a reporter gene can thus be expressed in a graded manner to reveal gene function in a rather short time and in a complex cellular environment.

Cell type-specific inhibitory inputs to dendritic and somatic compartments of parvalbumin-expressing neocortical interneuron.

Parvalbumin (PV)-producing fast-spiking neurons are well known to generate gamma oscillation by mutual chemical and electrical connections in the neocortex. Although it was clearly demonstrated that PV neurons form a dense gap junction network with each other not only at the proximal sites but also at the distal dendrites, comprehensive quantitative data on the chemical connections are still lacking. To elucidate the connectivity, we investigated inhibitory inputs to PV neurons in the somatosensory cortex, using the transgenic mice in which the dendrites and cell bodies of PV neurons were clearly visualized. We first examined GABAergic inputs to PV neurons by labeling postsynaptic and presynaptic sites with the immunoreactivities for gephyrin and vesicular GABA transporter. The density of GABAergic inputs was highest on the cell bodies, and almost linearly decreased to the distal dendrites. We then investigated inhibitory inputs from three distinct subgroups of GABAergic interneurons by visualizing the axon terminals immunopositive for PV, somatostatin (SOM), or vasoactive intestinal polypeptide (VIP). PV and SOM inputs were frequently located on the dendrites with the ratio of 2.5:1, but much less on the cell bodies. By contrast, VIP inputs clearly preferred the cell bodies to the dendrites. Consequently, the dendritic and somatic compartments of PV neurons received ∼60 and 62% of inhibitory inputs from PV and VIP neurons, respectively. This compartmental organization of inhibitory inputs suggests that PV neurons, together with gap junctions, constitute mutual connections at the dendrites, and that their activities are negatively controlled by the somatic inputs of VIP neurons.

EphA4-dependent axon retraction and midline localization of Ephrin-B3 are disrupted in the spinal cord of mice lacking mDia1 and mDia3 in combination.

mDia is an actin nucleator and polymerization factor regulated by the small GTPase Rho and consists of three isoforms. Here, we found that mice lacking mDia1 and mDia3, two isoforms expressed in the brain, in combination (mDia-DKO mice) show impaired left-right limb coordination during locomotion and aberrant midline crossing of axons of corticospinal neurons and spinal cord interneurons. Given that mice lacking Ephrin-B3-EphA4 signaling show a similar impairment in locomotion, we examined whether mDia is involved in Ephrin-B3-EphA4 signaling for axon repulsion. In primary cultured neurons, mDia deficiency impairs growth cone collapse and axon retraction induced by chemo-repellants including EphA ligands. In mDia-DKO mice, the Ephrin-B3-expressing midline structure in the spinal cord is disrupted, and axons aberrantly cross the spinal cord midline preferentially through the region devoid of Ephrin-B3. Therefore, mDia plays multiple roles in the proper formation of the neural network in vivo.

Transient sensorimotor projections in the developmental song learning period.

Memory recall and guidance are essential for motor skill acquisition. Like humans learning to speak, male zebra finches learn to sing by first memorizing and then matching their vocalization to the tutor's song (TS) during specific developmental periods. Yet, the neuroanatomical substrate supporting auditory-memory-guided sensorimotor learning has remained elusive. Here, using a whole-brain connectome analysis with activity-dependent viral expression, we identified a transient projection into the motor region, HVC, from neuronal ensembles responding to TS in the auditory forebrain, the caudomedial nidopallium (NCM), in juveniles. Virally induced cell death of the juvenile, but not adult, TS-responsive NCM neurons impaired song learning. Moreover, isolation, which delays closure of the sensory, but not the motor, learning period, did not affect the decrease of projections into the HVC from the NCM TS-responsive neurons after the song learning period. Taken together, our results suggest that dynamic axonal pruning may regulate timely auditory-memory-guided vocal learning during development.

A morphological analysis of thalamocortical axon fibers of rat posterior thalamic nuclei: a single neuron tracing study with viral vectors.

The rostral sector of the posterior thalamic nuclei (POm) is, together with the ventral posterior nuclei (VP), involved in somatosensory information processing in rodents. The POm receives inputs from the spinal cord and trigeminal nuclei and projects to the primary somatosensory (S1) cortex and other cortical areas. Although thalamocortical axons of single VP neurons are well known to innervate layer (L) 4 of the S1 cortex with distinct columnar organization, those of POm neurons have not been elucidated yet. In the present study, we investigated complete axonal and dendritic arborizations of single POm neurons in rats by visualizing the processes with Sindbis viruses expressing membrane-targeted fluorescent protein. When we divided the POm into anterior and posterior parts according to calbindin immunoreactivity, dendrites of posterior POm neurons were wider but less numerous than those of anterior neurons. More interestingly, axon fibers of anterior POm neurons were preferentially distributed in L5 of the S1 cortex, whereas those of posterior neurons were principally spread in L1 with wider and sparser arborization than those of anterior neurons. These results suggest that the POm is functionally segregated into anterior and posterior parts and that the 2 parts may play different roles in somatosensory information processing.

Specific AAV2/PHP.eB-mediated gene transduction of CA2 pyramidal cells via injection into the lateral ventricle.

Given its limited accessibility, the CA2 area has been less investigated compared to other subregions of the hippocampus. While the development of transgenic mice expressing Cre recombinase in the CA2 has revealed unique features of this area, the use of mouse lines has several limitations, such as lack of specificity. Therefore, a specific gene delivery system is required. Here, we confirmed that the AAV-PHP.eB capsid preferably infected CA2 pyramidal cells following retro-orbital injection and demonstrated that the specificity was substantially higher after injection into the lateral ventricle. In addition, a tropism for the CA2 area was observed in organotypic slice cultures. Combined injection into the lateral ventricle and stereotaxic injection into the CA2 area specifically introduced the transgene into CA2 pyramidal cells, enabling us to perform targeted patch-clamp recordings and optogenetic manipulation. These results suggest that AAV-PHP.eB is a versatile tool for specific gene transduction in CA2 pyramidal cells.

Preferential arborization of dendrites and axons of parvalbumin- and somatostatin-positive GABAergic neurons within subregions of the mouse claustrum.

The claustrum coordinates the activities of individual cortical areas through abundant reciprocal connections with the cerebral cortex. Although these excitatory connections have been extensively investigated in three subregions of the claustrum-core region and dorsal and ventral shell regions-the contribution of GABAergic neurons to the circuitry in each subregion remains unclear. Here, we examined the distribution of GABAergic neurons and their dendritic and axonal arborizations in each subregion. Combining in situ hybridization with immunofluorescence histochemistry showed that approximately 10% of neuronal nuclei-positive cells expressed glutamic acid decarboxylase 67 mRNA across the claustral subregions. Approximately 20%, 30%, and 10% of GABAergic neurons were immunoreactive for parvalbumin (PV), somatostatin (SOM), and vasoactive intestinal polypeptide, respectively, in each subregion, and these neurochemical markers showed little overlap with each other. We then reconstructed PV and SOM neurons labeled with adeno-associated virus vectors. The dendrites and axons of PV and SOM neurons were preferentially localized to their respective subregions where their cell bodies were located. Furthermore, the axons were preferentially extended in a rostrocaudal direction, whereas the dendrites were relatively isotropic. The present findings suggest that claustral PV and SOM neurons might execute information processing separately within the core and shell regions.

Local connections of excitatory neurons to corticothalamic neurons in the rat barrel cortex.

Corticothalamic projection neurons in the cerebral cortex constitute an important component of the thalamocortical reciprocal circuit, an essential input/output organization for cortical information processing. However, the spatial organization of local excitatory connections to corticothalamic neurons is only partially understood. In the present study, we first developed an adenovirus vector expressing somatodendritic membrane-targeted green fluorescent protein. After injection of the adenovirus vector into the ventrobasal thalamic complex, a band of layer (L) 6 corticothalamic neurons in the rat barrel cortex were retrogradely labeled. In addition to their cell bodies, fine dendritic spines of corticothalamic neurons were well visualized without the labeling of their axon collaterals or thalamocortical axons. In cortical slices containing retrogradely labeled L6 corticothalamic neurons, we intracellularly stained single pyramidal/spiny neurons of L2-6. We examined the spatial distribution of contact sites between the local axon collaterals of each pyramidal neuron and the dendrites of corticothalamic neurons. We found that corticothalamic neurons received strong and focused connections from L4 neurons just above them, and that the most numerous nearby and distant sources of local excitatory connections to corticothalamic neurons were corticothalamic neurons themselves and L6 putative corticocortical neurons, respectively. These results suggest that L4 neurons may serve as an important source of local excitatory inputs in shaping the cortical modulation of thalamic activity.

Parvalbumin-producing cortical interneurons receive inhibitory inputs on proximal portions and cortical excitatory inputs on distal dendrites.

To examine inputs to parvalbumin (PV)-producing interneurons, we generated transgenic mice expressing somatodendritic membrane-targeted green fluorescent protein specifically in the interneurons, and completely visualized their dendrites and somata. Using immunolabeling for vesicular glutamate transporter (VGluT)1, VGluT2, and vesicular GABA transporter, we found that VGluT1-positive terminals made contacts 4- and 3.1-fold more frequently with PV-producing interneurons than VGluT2-positive and GABAergic terminals, respectively, in the primary somatosensory cortex. Even in layer 4, where VGluT2-positive terminals were most densely distributed, VGluT1-positive inputs to PV-producing interneurons were 2.4-fold more frequent than VGluT2-positive inputs. Furthermore, although GABAergic inputs to PV-producing interneurons were as numerous as VGluT2-positive inputs in most cortical layers, GABAergic inputs clearly preferred the proximal dendrites and somata of the interneurons, indicating that the sites of GABAergic inputs were more optimized than those of VGluT2-positive inputs. Simulation analysis with a PV-producing interneuron model compatible with the present morphological data revealed a plausible reason for this observation, by showing that GABAergic and glutamatergic postsynaptic potentials evoked by inputs to distal dendrites were attenuated to 60 and 87%, respectively, of those evoked by somatic inputs. As VGluT1-positive and VGluT2-positive axon terminals were presumed to be cortical and thalamic glutamatergic inputs, respectively, cortical excitatory inputs to PV-producing interneurons outnumbered the thalamic excitatory and intrinsic inhibitory inputs more than two-fold in any cortical layer. Although thalamic inputs are known to evoke about two-fold larger unitary excitatory postsynaptic potentials than cortical ones, the present results suggest that cortical inputs control PV-producing interneurons at least as strongly as thalamic inputs.

Expression of gap junction protein connexin36 in multiple subtypes of GABAergic neurons in adult rat somatosensory cortex.

To characterize connexin36 (Cx36)-expressing neurons of the adult rat somatosensory cortex, we examined fluorescence signals for Cx36 messenger RNA (mRNA) in 3 nonoverlapping subpopulations of γ-aminobutyric acid (GABA)ergic interneurons, which showed immunoreactivity for 1) parvalbumin (PV); 2) somatostatin (SOM); and 3) either calretinin (CR), vasoactive intestinal polypeptide (VIP), cholecystokinin (CCK), or choline acetyltransferase (ChAT). About 80% of PV-, 52% of SOM-, 37% of CR/VIP/CCK/ChAT-immunoreactive cells displayed Cx36 signals across all cortical layers, and inversely 64%, 25%, and 9% of Cx36-expressing neurons were positive for PV, SOM, or CR/VIP/CCK/ChAT, respectively. Notably, although almost all Cx36-expressing neurons in layer (L) 4, L5, and L6 were positive for one of these markers, a substantial proportion of those in L1 (91%) and L2/3 (10%) were negative for the markers tested, suggesting that other types of neurons might express Cx36. We further investigated the colocalization of Cx36 mRNA and α-actinin2 immunoreactivity, as a marker for late-spiking GABAergic neurons, by using mirror-image sections. Surprisingly, more than 77% of α-actinin2-positive cells displayed Cx36 signals in L1-L3, and about 49% and 13% of Cx36-expressing neurons were positive for α-actinin2 in L1 and L2/3, respectively. These findings suggest that all the subtypes of GABAergic interneurons might form gap junctions in the neocortex.

Local Connections of Pyramidal Neurons to Parvalbumin-Producing Interneurons in Motor-Associated Cortical Areas of Mice.

Parvalbumin (PV)-producing neurons are the largest subpopulation of cortical GABAergic interneurons, which mediate lateral, feedforward, and feedback inhibition in local circuits and modulate the activity of pyramidal neurons. Clarifying the specific connectivity between pyramidal and PV neurons is essential for understanding the role of PV neurons in local circuits. In the present study, we visualized somas and dendrites of PV neurons using transgenic mice in which PV neurons specifically express membrane-targeted GFP, and intracellularly labeled local axons of 26 pyramidal neurons in layers 2-6 in acute slices of the motor-associated cortex from transgenic mice. We mapped morphologically distribution of inputs from a pyramidal neuron to PV neurons based on contact sites (appositions) between the axons from an intracellularly filled pyramidal neuron and the dendrites of PV neurons. Layer 6 corticothalamic (CT)-like pyramidal neurons formed appositions to PV neurons at a significantly higher rate than other pyramidal neurons. The percentage of apposed varicosities to all the labeled varicosities of layer 6 CT-like neurons was 28%, and that of the other pyramidal neurons was 12-19%. Layer 6 CT-like neurons preferentially formed appositions with PV neurons in layers 5b-6, while other pyramidal neurons uniformly formed appositions with PV neurons in all layers. Furthermore, both layer 6 CT-like and corticocortical-like neurons more frequently formed compound appositions, where two or more appositions were located on a dendritic branch, than other pyramidal neurons. Layer 6 CT neurons may contribute to intracortical information processing through preferential connections with PV neurons in layers 5b-6.