An operating principle of the cerebral cortex, and a cellular mechanism for attentional trial-and-error pattern learning and useful classification extraction.

A feature of the brains of intelligent animals is the ability to learn to respond to an ensemble of active neuronal inputs with a behaviorally appropriate ensemble of active neuronal outputs. Previously, a hypothesis was proposed on how this mechanism is implemented at the cellular level within the neocortical pyramidal neuron: the apical tuft or perisomatic inputs initiate "guess" neuron firings, while the basal dendrites identify input patterns based on excited synaptic clusters, with the cluster excitation strength adjusted based on reward feedback. This simple mechanism allows neurons to learn to classify their inputs in a surprisingly intelligent manner. Here, we revise and extend this hypothesis. We modify synaptic plasticity rules to align with behavioral time scale synaptic plasticity (BTSP) observed in hippocampal area CA1, making the framework more biophysically and behaviorally plausible. The neurons for the guess firings are selected in a voluntary manner via feedback connections to apical tufts in the neocortical layer 1, leading to dendritic Ca2+ spikes with burst firing, which are postulated to be neural correlates of attentional, aware processing. Once learned, the neuronal input classification is executed without voluntary or conscious control, enabling hierarchical incremental learning of classifications that is effective in our inherently classifiable world. In addition to voluntary, we propose that pyramidal neuron burst firing can be involuntary, also initiated via apical tuft inputs, drawing attention toward important cues such as novelty and noxious stimuli. We classify the excitations of neocortical pyramidal neurons into four categories based on their excitation pathway: attentional versus automatic and voluntary/acquired versus involuntary. Additionally, we hypothesize that dendrites within pyramidal neuron minicolumn bundles are coupled via depolarization cross-induction, enabling minicolumn functions such as the creation of powerful hierarchical "hyperneurons" and the internal representation of the external world. We suggest building blocks to extend the microcircuit theory to network-level processing, which, interestingly, yields variants resembling the artificial neural networks currently in use. On a more speculative note, we conjecture that principles of intelligence in universes governed by certain types of physical laws might resemble ours.

Dynamic predictive coding: A model of hierarchical sequence learning and prediction in the neocortex.

We introduce dynamic predictive coding, a hierarchical model of spatiotemporal prediction and sequence learning in the neocortex. The model assumes that higher cortical levels modulate the temporal dynamics of lower levels, correcting their predictions of dynamics using prediction errors. As a result, lower levels form representations that encode sequences at shorter timescales (e.g., a single step) while higher levels form representations that encode sequences at longer timescales (e.g., an entire sequence). We tested this model using a two-level neural network, where the top-down modulation creates low-dimensional combinations of a set of learned temporal dynamics to explain input sequences. When trained on natural videos, the lower-level model neurons developed space-time receptive fields similar to those of simple cells in the primary visual cortex while the higher-level responses spanned longer timescales, mimicking temporal response hierarchies in the cortex. Additionally, the network's hierarchical sequence representation exhibited both predictive and postdictive effects resembling those observed in visual motion processing in humans (e.g., in the flash-lag illusion). When coupled with an associative memory emulating the role of the hippocampus, the model allowed episodic memories to be stored and retrieved, supporting cue-triggered recall of an input sequence similar to activity recall in the visual cortex. When extended to three hierarchical levels, the model learned progressively more abstract temporal representations along the hierarchy. Taken together, our results suggest that cortical processing and learning of sequences can be interpreted as dynamic predictive coding based on a hierarchical spatiotemporal generative model of the visual world.

Enhancing GABAergic Tonic Inhibition Reduces Seizure-Like Activity in the Neonatal Mouse Hippocampus and Neocortex.

Approximately one-third of neonatal seizures do not respond to first-line anticonvulsants, including phenobarbital, which enhances phasic inhibition. Whether enhancing tonic inhibition decreases seizure-like activity in the neonate when GABA is mainly depolarizing at this age is unknown. We evaluated if increasing tonic inhibition using THIP [4,5,6,7-tetrahydroisoxazolo(5,4-c)pyridin-3-ol, gaboxadol], a δ-subunit-selective GABAA receptor agonist, decreases seizure-like activity in neonatal C57BL/6J mice (postnatal day P5-8, both sexes) using acute brain slices. Whole-cell patch-clamp recordings showed that THIP enhanced GABAergic tonic inhibitory conductances in layer V neocortical and CA1 pyramidal neurons and increased their rheobase without altering sEPSC characteristics. Two-photon calcium imaging demonstrated that enhancing the activity of extrasynaptic GABAARs decreased neuronal firing in both brain regions. In the 4-aminopyridine and the low-Mg2+ model of pharmacoresistant seizures, THIP reduced epileptiform activity in the neocortex and CA1 hippocampal region of neonatal and adult brain slices in a dose-dependent manner. We conclude that neocortical layer V and CA1 pyramidal neurons have tonic inhibitory conductances, and when enhanced, they reduce neuronal firing and decrease seizure-like activity. Therefore, augmenting tonic inhibition could be a viable approach for treating neonatal seizures.

Spontaneous slow wave oscillations in extracellular field potential recordings reflect the alternating dominance of excitation and inhibition.

In the resting state, cortical neurons can fire action potentials spontaneously but synchronously (Up state), followed by a quiescent period (Down state) before the cycle repeats. Extracellular recordings in the infragranular layer of cortex with a micro-electrode display a negative deflection (depth-negative) during Up states and a positive deflection (depth-positive) during Down states. The resulting slow wave oscillation (SWO) has been studied extensively during sleep and under anaesthesia. However, recent research on the balanced nature of synaptic excitation and inhibition has highlighted our limited understanding of its genesis. Specifically, are excitation and inhibition balanced during SWOs? We analyse spontaneous local field potentials (LFPs) during SWOs recorded from anaesthetised rats via a multi-channel laminar micro-electrode and show that the Down state consists of two distinct synaptic states: a Dynamic Down state associated with depth-positive LFPs and a prominent dipole in the extracellular field, and a Static Down state with negligible ( ≈ 0 mV $ \approx 0{\mathrm{\;mV}}$ ) LFPs and a lack of dipoles extracellularly. We demonstrate that depth-negative and -positive LFPs are generated by a shift in the balance of synaptic excitation and inhibition from excitation dominance (depth-negative) to inhibition dominance (depth-positive) in the infragranular layer neurons. Thus, although excitation and inhibition co-tune overall, differences in their timing lead to an alternation of dominance, manifesting as SWOs. We further show that Up state initiation is significantly faster if the preceding Down state is dynamic rather than static. Our findings provide a coherent picture of the dependence of SWOs on synaptic activity. KEY POINTS: Cortical neurons can exhibit repeated cycles of spontaneous activity interleaved with periods of relative silence, a phenomenon known as 'slow wave oscillation' (SWO). During SWOs, recordings of local field potentials (LFPs) in the neocortex show depth-negative deflection during the active period (Up state) and depth-positive deflection during the silent period (Down state). Here we further classified the Down state into a dynamic phase and a static phase based on a novel method of classification and revealed non-random, stereotypical sequences of the three states occurring with significantly different transitional kinetics. Our results suggest that the positive and negative deflections in the LFP reflect the shift of the instantaneous balance between excitatory and inhibitory synaptic activity of the local cortical neurons. The differences in transitional kinetics may imply distinct synaptic mechanisms for Up state initiation. The study may provide a new approach for investigating spontaneous brain rhythms.

Layer 1 neocortex: Gating and integrating multidimensional signals.

Layer 1 (L1) of the neocortex acts as a nexus for the collection and processing of widespread information. By integrating ascending inputs with extensive top-down activity, this layer likely provides critical information regulating how the perception of sensory inputs is reconciled with expectation. This is accomplished by sorting, directing, and integrating the complex network of excitatory inputs that converge onto L1. These signals are combined with neuromodulatory afferents and gated by the wealth of inhibitory interneurons that either are embedded within L1 or send axons from other cortical layers. Together, these interactions dynamically calibrate information flow throughout the neocortex. This review will primarily focus on L1 within the primary sensory cortex and will use these insights to understand L1 in other cortical areas.

Responses to Pattern-Violating Visual Stimuli Evolve Differently Over Days in Somata and Distal Apical Dendrites.

Scientists have long conjectured that the neocortex learns patterns in sensory data to generate top-down predictions of upcoming stimuli. In line with this conjecture, different responses to pattern-matching vs pattern-violating visual stimuli have been observed in both spiking and somatic calcium imaging data. However, it remains unknown whether these pattern-violation signals are different between the distal apical dendrites, which are heavily targeted by top-down signals, and the somata, where bottom-up information is primarily integrated. Furthermore, it is unknown how responses to pattern-violating stimuli evolve over time as an animal gains more experience with them. Here, we address these unanswered questions by analyzing responses of individual somata and dendritic branches of layer 2/3 and layer 5 pyramidal neurons tracked over multiple days in primary visual cortex of awake, behaving female and male mice. We use sequences of Gabor patches with patterns in their orientations to create pattern-matching and pattern-violating stimuli, and two-photon calcium imaging to record neuronal responses. Many neurons in both layers show large differences between their responses to pattern-matching and pattern-violating stimuli. Interestingly, these responses evolve in opposite directions in the somata and distal apical dendrites, with somata becoming less sensitive to pattern-violating stimuli and distal apical dendrites more sensitive. These differences between the somata and distal apical dendrites may be important for hierarchical computation of sensory predictions and learning, since these two compartments tend to receive bottom-up and top-down information, respectively.

Primate neocortex performs balanced sensory amplification.

The sensory cortex amplifies relevant features of external stimuli. This sensitivity and selectivity arise through the transformation of inputs by cortical circuitry. We characterize the circuit mechanisms and dynamics of cortical amplification by making large-scale simultaneous measurements of single cells in awake primates and testing computational models. By comparing network activity in both driven and spontaneous states with models, we identify the circuit as operating in a regime of non-normal balanced amplification. Incoming inputs are strongly but transiently amplified by strong recurrent feedback from the disruption of excitatory-inhibitory balance in the network. Strong inhibition rapidly quenches responses, thereby permitting the tracking of time-varying stimuli.

Artificial sharp-wave-ripples to support memory and counter neurodegeneration.

Information processed in our sensory neocortical areas is transported to the hippocampus during memory encoding, and between hippocampus and neocortex during memory consolidation, and retrieval. Short bursts of high-frequency oscillations, so called sharp-wave-ripples, have been proposed as a potential mechanism for this information transfer: They can synchronize neural activity to support the formation of local neural networks to store information, and between distant cortical sites to act as a bridge to transfer information between sensory cortical areas and hippocampus. In neurodegenerative diseases like Alzheimer's Disease, different neuropathological processes impair normal neural functioning and neural synchronization as well as sharp-wave-ripples, which impairs consolidation and retrieval of information, and compromises memory. Here, we formulate a new hypothesis, that artificially inducing sharp-wave-ripples with noninvasive high-frequency visual stimulation could potentially support memory functioning, as well as target the neuropathological processes underlying neurodegenerative diseases. We also outline key challenges for empirical tests of the hypothesis.

Holistic bursting cells store long-term memory in auditory cortex.

The sensory neocortex has been suggested to be a substrate for long-term memory storage, yet which exact single cells could be specific candidates underlying such long-term memory storage remained neither known nor visible for over a century. Here, using a combination of day-by-day two-photon Ca2+ imaging and targeted single-cell loose-patch recording in an auditory associative learning paradigm with composite sounds in male mice, we reveal sparsely distributed neurons in layer 2/3 of auditory cortex emerged step-wise from quiescence into bursting mode, which then invariably expressed holistic information of the learned composite sounds, referred to as holistic bursting (HB) cells. Notably, it was not shuffled populations but the same sparse HB cells that embodied the behavioral relevance of the learned composite sounds, pinpointing HB cells as physiologically-defined single-cell candidates of an engram underlying long-term memory storage in auditory cortex.

Abnormal patterns of sleep and waking behaviors are accompanied by neocortical oscillation disturbances in an Ank3 mouse model of epilepsy-bipolar disorder comorbidity.

ANK3 is a leading bipolar disorder (BD) candidate gene in humans and provides a unique opportunity for studying epilepsy-BD comorbidity. Previous studies showed that deletion of Ank3-1b, a BD-associated variant of Ank3 in mice leads to increased firing threshold and diminished action potential dynamic range of parvalbumin (PV) interneurons and absence epilepsy, thus providing a biological mechanism linking epilepsy and BD. To explore the behavioral overlap of these disorders, we characterized behavioral patterns of Ank3-1b KO mice during overnight home-cage activity and examined network activity during these behaviors using paired video and EEG recordings. Since PV interneurons contribute to the generation of high-frequency gamma oscillations, we anticipated changes in the power of neocortical EEG signals in the gamma frequency range (> 25 Hz) during behavioral states related to human BD symptoms, including abnormal sleep, hyperactivity, and repetitive behaviors. Ank3-1b KO mice exhibited an overall increase in slow gamma (~25-45 Hz) power compared to controls, and slow gamma power correlated with seizure phenotype severity across behaviors. During sleep, increased slow gamma power correlated with decreased time spent in the rapid eye movement (REM) stage of sleep. Seizures were more common during REM sleep compared to non-REM (NREM) sleep. We also found that Ank3-1b KO mice were hyperactive and exhibited a repetitive behavior phenotype that co-occurred with increased slow gamma power. Our results identify a novel EEG biomarker associating Ank3 genetic variation with BD and epilepsy and suggest modulation of gamma oscillations as a potential therapeutic target.

The structural and functional complexity of the integrative hypothalamus.

The hypothalamus ("hypo" meaning below, and "thalamus" meaning bed) consists of regulatory circuits that support basic life functions that ensure survival. Sitting at the interface between peripheral, environmental, and neural inputs, the hypothalamus integrates these sensory inputs to influence a range of physiologies and behaviors. Unlike the neocortex, in which a stereotyped cytoarchitecture mediates complex functions across a comparatively small number of neuronal fates, the hypothalamus comprises upwards of thousands of distinct cell types that form redundant yet functionally discrete circuits. With single-cell RNA sequencing studies revealing further cellular heterogeneity and modern photonic tools enabling high-resolution dissection of complex circuitry, a new era of hypothalamic mapping has begun. Here, we provide a general overview of mammalian hypothalamic organization, development, and connectivity to help welcome newcomers into this exciting field.

Temporal disparity of action potentials triggered in axon initial segments and distal axons in the neocortex.

Neural population activity determines the timing of synaptic inputs, which arrive to dendrites, cell bodies, and axon initial segments (AISs) of cortical neurons. Action potential initiation in the AIS (AIS-APs) is driven by input integration, and the phase preference of AIS-APs during network oscillations is characteristic to cell classes. Distal regions of cortical axons do not receive synaptic inputs, yet experimental induction protocols can trigger retroaxonal action potentials (RA-APs) in axons distal from the soma. We report spontaneously occurring RA-APs in human and rodent cortical interneurons that appear uncorrelated to inputs and population activity. Network-linked triggering of AIS-APs versus input-independent timing of RA-APs of the same interneurons results in disparate temporal contribution of a single cell to in vivo network operation through perisomatic and distal axonal firing.

A marmoset brain cell census reveals regional specialization of cellular identities.

The mammalian brain is composed of many brain structures, each with its own ontogenetic and developmental history. We used single-nucleus RNA sequencing to sample over 2.4 million brain cells across 18 locations in the common marmoset, a New World monkey primed for genetic engineering, and examined gene expression patterns of cell types within and across brain structures. The adult transcriptomic identity of most neuronal types is shaped more by developmental origin than by neurotransmitter signaling repertoire. Quantitative mapping of GABAergic types with single-molecule FISH (smFISH) reveals that interneurons in the striatum and neocortex follow distinct spatial principles, and that lateral prefrontal and other higher-order cortical association areas are distinguished by high proportions of VIP+ neurons. We use cell type-specific enhancers to drive AAV-GFP and reconstruct the morphologies of molecularly resolved interneuron types in neocortex and striatum. Our analyses highlight how lineage, local context, and functional class contribute to the transcriptional identity and biodistribution of primate brain cell types.

Comparative transcriptomics reveals human-specific cortical features.

The cognitive abilities of humans are distinctive among primates, but their molecular and cellular substrates are poorly understood. We used comparative single-nucleus transcriptomics to analyze samples of the middle temporal gyrus (MTG) from adult humans, chimpanzees, gorillas, rhesus macaques, and common marmosets to understand human-specific features of the neocortex. Human, chimpanzee, and gorilla MTG showed highly similar cell-type composition and laminar organization as well as a large shift in proportions of deep-layer intratelencephalic-projecting neurons compared with macaque and marmoset MTG. Microglia, astrocytes, and oligodendrocytes had more-divergent expression across species compared with neurons or oligodendrocyte precursor cells, and neuronal expression diverged more rapidly on the human lineage. Only a few hundred genes showed human-specific patterning, suggesting that relatively few cellular and molecular changes distinctively define adult human cortical structure.

An evolutionary conserved division-of-labor between archicortical and neocortical ripples organizes information transfer during sleep.

Systems-level memory consolidation during sleep depends on the temporally precise interplay between cardinal sleep oscillations. Specifically, hippocampal ripples constitute a key substrate of the hippocampal-neocortical dialog underlying memory formation. Recently, it became evident that ripples are not unique to archicortex, but constitute a wide-spread neocortical phenomenon. To date, little is known about the morphological similarities between archi- and neocortical ripples. Moreover, it remains undetermined if neocortical ripples fulfill distinct functional roles. Leveraging intracranial recordings from the human medial temporal lobe (MTL) and neocortex during sleep, our results reveal region-specific functional specializations, albeit a near-uniform morphology. While MTL ripples synchronize the memory network to trigger directional MTL-to-neocortical information flow, neocortical ripples reduce information flow to minimize interference. At the population level, MTL ripples confined population dynamics to a low-dimensional subspace, while neocortical ripples diversified the population response; thus, constituting an effective mechanism to functionally uncouple the MTL-neocortical network. Critically, we replicated the key findings in rodents, where the same division-of-labor between archi- and neocortical ripples was evident. In sum, these results uncover an evolutionary preserved mechanism where the precisely coordinated interplay between MTL and neocortical ripples temporally segregates MTL information transfer from subsequent neocortical processing during sleep.

Direct and indirect neurogenesis generate a mosaic of distinct glutamatergic projection neuron types in cerebral cortex.

Variations in size and complexity of the cerebral cortex result from differences in neuron number and composition, rooted in evolutionary changes in direct and indirect neurogenesis (dNG and iNG) that are mediated by radial glia and intermediate progenitors (IPs), respectively. How dNG and iNG differentially contribute to neuronal number, diversity, and connectivity are unknown. Establishing a genetic fate-mapping method to differentially visualize dNG and iNG in mice, we found that while both dNG and iNG contribute to all cortical structures, iNG contributes the largest relative proportions to the hippocampus and neocortex. Within the neocortex, whereas dNG generates all major glutamatergic projection neuron (PN) classes, iNG differentially amplifies and diversifies PNs within each class; the two pathways generate distinct PN types and assemble fine mosaics of lineage-based cortical subnetworks. Our results establish a ground-level lineage framework for understanding cortical development and evolution by linking foundational progenitor types and neurogenic pathways to PN types.

Target cell-specific synaptic dynamics of excitatory to inhibitory neuron connections in supragranular layers of human neocortex.

Rodent studies have demonstrated that synaptic dynamics from excitatory to inhibitory neuron types are often dependent on the target cell type. However, these target cell-specific properties have not been well investigated in human cortex, where there are major technical challenges in reliably obtaining healthy tissue, conducting multiple patch-clamp recordings on inhibitory cell types, and identifying those cell types. Here, we take advantage of newly developed methods for human neurosurgical tissue analysis with multiple patch-clamp recordings, post-hoc fluorescent in situ hybridization (FISH), machine learning-based cell type classification and prospective GABAergic AAV-based labeling to investigate synaptic properties between pyramidal neurons and PVALB- vs. SST-positive interneurons. We find that there are robust molecular differences in synapse-associated genes between these neuron types, and that individual presynaptic pyramidal neurons evoke postsynaptic responses with heterogeneous synaptic dynamics in different postsynaptic cell types. Using molecular identification with FISH and classifiers based on transcriptomically identified PVALB neurons analyzed by Patch-seq, we find that PVALB neurons typically show depressing synaptic characteristics, whereas other interneuron types including SST-positive neurons show facilitating characteristics. Together, these data support the existence of target cell-specific synaptic properties in human cortex that are similar to rodent, thereby indicating evolutionary conservation of local circuit connectivity motifs from excitatory to inhibitory neurons and their synaptic dynamics.

Pyramidal neurons form active, transient, multilayered circuits perturbed by autism-associated mutations at the inception of neocortex.

Cortical circuits are composed predominantly of pyramidal-to-pyramidal neuron connections, yet their assembly during embryonic development is not well understood. We show that mouse embryonic Rbp4-Cre cortical neurons, transcriptomically closest to layer 5 pyramidal neurons, display two phases of circuit assembly in vivo. At E14.5, they form a multi-layered circuit motif, composed of only embryonic near-projecting-type neurons. By E17.5, this transitions to a second motif involving all three embryonic types, analogous to the three adult layer 5 types. In vivo patch clamp recordings and two-photon calcium imaging of embryonic Rbp4-Cre neurons reveal active somas and neurites, tetrodotoxin-sensitive voltage-gated conductances, and functional glutamatergic synapses, from E14.5 onwards. Embryonic Rbp4-Cre neurons strongly express autism-associated genes and perturbing these genes interferes with the switch between the two motifs. Hence, pyramidal neurons form active, transient, multi-layered pyramidal-to-pyramidal circuits at the inception of neocortex, and studying these circuits could yield insights into the etiology of autism.

The temporal structure of REM sleep shows minute-scale fluctuations across brain and body in mice and humans.

Rapid eye movement sleep (REM) is believed to have a binary temporal structure with "phasic" and "tonic" microstates, characterized by motoric activity versus quiescence, respectively. However, we observed in mice that the frequency of theta activity (a marker of rodent REM) fluctuates in a nonbinary fashion, with the extremes of that fluctuation correlating with phasic-type and tonic-type facial motricity. Thus, phasic and tonic REM may instead represent ends of a continuum. These cycles of brain physiology and facial movement occurred at 0.01 to 0.06 Hz, or infraslow frequencies, and affected cross-frequency coupling and neuronal activity in the neocortex, suggesting network functional impact. We then analyzed human data and observed that humans also demonstrate nonbinary phasic/tonic microstates, with continuous 0.01 to 0.04-Hz respiratory rate cycles matching the incidence of eye movements. These fundamental properties of REM can yield insights into our understanding of sleep health.

Mechanisms of Dominant Electrophysiological Features of Four Subtypes of Layer 1 Interneurons.

Neocortical layer 1 (L1) consists of the distal dendrites of pyramidal cells and GABAergic interneurons (INs) and receives extensive long-range "top-down" projections, but L1 INs remain poorly understood. In this work, we systematically examined the distinct dominant electrophysiological features for four unique IN subtypes in L1 that were previously identified from mice of either gender: Canopy cells show an irregular firing pattern near rheobase; neurogliaform cells are late-spiking, and their firing rate accelerates during current injections; cells with strong expression of the α7 nicotinic receptor (α7 cells), display onset (rebound) bursting; vasoactive intestinal peptide (VIP) expressing cells exhibit high input resistance, strong adaptation, and irregular firing. Computational modeling revealed that these diverse neurophysiological features could be explained by an extended exponential-integrate-and-fire neuron model with varying contributions of a slowly inactivating K+ channel, a T-type Ca2+ channel, and a spike-triggered Ca2+-dependent K+ channel. In particular, we show that irregular firing results from square-wave bursting through a fast-slow analysis. Furthermore, we demonstrate that irregular firing is frequently observed in VIP cells because of the interaction between strong adaptation and a slowly inactivating K+ channel. At last, we reveal that the VIP and α7 cell models resonant with alpha/theta band input through a dynamic gain analysis.SIGNIFICANCE STATEMENT In the neocortex, ∼25% of neurons are interneurons. Interestingly, only somas of interneurons reside within layer 1 (L1) of the neocortex, but not of excitatory pyramidal cells. L1 interneurons are diverse and believed to be important in the cortical-cortex interactions, especially top-down signaling in the cortical hierarchy. However, the electrophysiological features of L1 interneurons are poorly understood. Here, we systematically studied the electrophysiological features within each L1 interneuron subtype. Furthermore, we build computational models for each subtype and study the mechanisms behind these features. These electrophysiological features within each subtype should be incorporated to elucidate how different L1 interneuron subtypes contribute to communication between cortexes.