Lateral Connections Improve Generalizability of Learning in a Simple Neural Network.

To navigate the world around us, neural circuits rapidly adapt to their environment learning generalizable strategies to decode information. When modeling these learning strategies, network models find the optimal solution to satisfy one task condition but fail when introduced to a novel task or even a different stimulus in the same space. In the experiments described in this letter, I investigate the role of lateral gap junctions in learning generalizable strategies to process information. Lateral gap junctions are formed by connexin proteins creating an open pore that allows for direct electrical signaling between two neurons. During neural development, the rate of gap junctions is high, and daughter cells that share similar tuning properties are more likely to be connected by these junctions. Gap junctions are highly plastic and get heavily pruned throughout development. I hypothesize that they mediate generalized learning by imprinting the weighting structure within a layer to avoid overfitting to one task condition. To test this hypothesis, I implemented a feedforward probabilistic neural network mimicking a cortical fast spiking neuron circuit that is heavily involved in movement. Many of these cells are tuned to speeds that I used as the input stimulus for the network to estimate. When training this network using a delta learning rule, both a laterally connected network and an unconnected network can estimate a single speed. However, when asking the network to estimate two or more speeds, alternated in training, an unconnected network either cannot learn speed or optimizes to a singular speed, while the laterally connected network learns the generalizable strategy and can estimate both speeds. These results suggest that lateral gap junctions between neurons enable generalized learning, which may help explain learning differences across life span.

Ensemble dynamics and information flow deduction from whole-brain imaging data.

The recent advancements in large-scale activity imaging of neuronal ensembles offer valuable opportunities to comprehend the process involved in generating brain activity patterns and understanding how information is transmitted between neurons or neuronal ensembles. However, existing methodologies for extracting the underlying properties that generate overall dynamics are still limited. In this study, we applied previously unexplored methodologies to analyze time-lapse 3D imaging (4D imaging) data of head neurons of the nematode Caenorhabditis elegans. By combining time-delay embedding with the independent component analysis, we successfully decomposed whole-brain activities into a small number of component dynamics. Through the integration of results from multiple samples, we extracted common dynamics from neuronal activities that exhibit apparent divergence across different animals. Notably, while several components show common cooperativity across samples, some component pairs exhibited distinct relationships between individual samples. We further developed time series prediction models of synaptic communications. By combining dimension reduction using the general framework, gradient kernel dimension reduction, and probabilistic modeling, the overall relationships of neural activities were incorporated. By this approach, the stochastic but coordinated dynamics were reproduced in the simulated whole-brain neural network. We found that noise in the nervous system is crucial for generating realistic whole-brain dynamics. Furthermore, by evaluating synaptic interaction properties in the models, strong interactions within the core neural circuit, variable sensory transmission and importance of gap junctions were inferred. Virtual optogenetics can be also performed using the model. These analyses provide a solid foundation for understanding information flow in real neural networks.

Electrical synapse structure requires distinct isoforms of a postsynaptic scaffold.

Electrical synapses are neuronal gap junction (GJ) channels associated with a macromolecular complex called the electrical synapse density (ESD), which regulates development and dynamically modifies electrical transmission. However, the proteomic makeup and molecular mechanisms utilized by the ESD that direct electrical synapse formation are not well understood. Using the Mauthner cell of zebrafish as a model, we previously found that the intracellular scaffolding protein ZO1b is a member of the ESD, localizing postsynaptically, where it is required for GJ channel localization, electrical communication, neural network function, and behavior. Here, we show that the complexity of the ESD is further diversified by the genomic structure of the ZO1b gene locus. The ZO1b gene is alternatively initiated at three transcriptional start sites resulting in isoforms with unique N-termini that we call ZO1b-Alpha, -Beta, and -Gamma. We demonstrate that ZO1b-Beta and ZO1b-Gamma are broadly expressed throughout the nervous system and localize to electrical synapses. By contrast, ZO1b-Alpha is expressed mainly non-neuronally and is not found at synapses. We generate mutants in all individual isoforms, as well as double mutant combinations in cis on individual chromosomes, and find that ZO1b-Beta is necessary and sufficient for robust GJ channel localization. ZO1b-Gamma, despite its localization to the synapse, plays an auxiliary role in channel localization. This study expands the notion of molecular complexity at the ESD, revealing that an individual genomic locus can contribute distinct isoforms to the macromolecular complex at electrical synapses. Further, independent scaffold isoforms have differential contributions to developmental assembly of the interneuronal GJ channels. We propose that ESD molecular complexity arises both from the diversity of unique genes and from distinct isoforms encoded by single genes. Overall, ESD proteomic diversity is expected to have critical impacts on the development, structure, function, and plasticity of electrical transmission.

Layers of inhibitory networks shape receptive field properties of AII amacrine cells.

In the retina, rod and cone pathways mediate visual signals over a billion-fold range in luminance. AII ("A-two") amacrine cells (ACs) receive signals from both pathways via different bipolar cells, enabling AIIs to operate at night and during the day. Previous work has examined luminance-dependent changes in AII gap junction connectivity, but less is known about how surrounding circuitry shapes AII receptive fields across light levels. Here, we report that moderate contrast stimuli elicit surround inhibition in AIIs under all but the dimmest visual conditions, due to actions of horizontal cells and at least two ACs that inhibit presynaptic bipolar cells. Under photopic (daylight) conditions, surround inhibition transforms AII response kinetics, which are inherited by downstream ganglion cells. Ablating neuronal nitric oxide synthase type-1 (nNOS-1) ACs removes AII surround inhibition under mesopic (dusk/dawn), but not photopic, conditions. Our findings demonstrate how multiple layers of neural circuitry interact to encode signals across a wide physiological range.

Sources of Calcium at Connexin 36 Gap Junctions in the Retina.

Synaptic plasticity is a fundamental feature of the CNS that controls the magnitude of signal transmission between communicating cells. Many electrical synapses exhibit substantial plasticity that modulates the degree of coupling within groups of neurons, alters the fidelity of signal transmission, or even reconfigures functional circuits. In several known examples, such plasticity depends on calcium and is associated with neuronal activity. Calcium-driven signaling is known to promote potentiation of electrical synapses in fish Mauthner cells, mammalian retinal AII amacrine cells, and inferior olive neurons, and to promote depression in thalamic reticular neurons. To measure local calcium dynamics in situ, we developed a transgenic mouse expressing a GCaMP calcium biosensor fused to Connexin 36 (Cx36) at electrical synapses. We examined the sources of calcium for activity-dependent plasticity in retina slices using confocal or Super-Resolution Radial Fluctuations imaging. More than half of Cx36-GCaMP gap junctions responded to puffs of glutamate with transient increases in fluorescence. The responses were strongly dependent on NMDA receptors, in keeping with known activity-dependent signaling in some amacrine cells. We also found that some responses depended on the activity of voltage-gated calcium channels, representing a previously unrecognized source of calcium to control retinal electrical synaptic plasticity. The high prevalence of calcium signals at electrical synapses in response to glutamate application indicates that a large fraction of electrical synapses has the potential to be regulated by neuronal activity. This provides a means to tune circuit connectivity dynamically based on local activity.

Restoring connexin-36 function in diabetogenic environments precludes mouse and human islet dysfunction.

The secretion of insulin from ß-cells in the islet of Langerhans is governed by a series of metabolic and electrical events, which can fail during the progression of type 2 diabetes (T2D). ß-cells are electrically coupled via connexin-36 (Cx36) gap junction channels, which coordinates the pulsatile dynamics of [Ca2+ ] and insulin release across the islet. Factors such as pro-inflammatory cytokines and free fatty acids disrupt gap junction coupling under in vitro conditions. Here we test whether gap junction coupling and coordinated [Ca2+ ] dynamics are disrupted in T2D, and whether recovery of gap junction coupling can recover islet function. We examine islets from donors with T2D, from db/db mice, and islets treated with pro-inflammatory cytokines (TNF-α, IL-1ß, IFN-É£) or free fatty acids (palmitate). We modulate gap junction coupling using Cx36 over-expression or pharmacological activation via modafinil. We also develop a peptide mimetic (S293) of the c-terminal regulatory site of Cx36 designed to compete against its phosphorylation. Cx36 gap junction permeability and [Ca2+ ] dynamics were disrupted in islets from both human donors with T2D and db/db mice, and in islets treated with pro-inflammatory cytokines or palmitate. Cx36 over-expression, modafinil treatment and S293 peptide all enhanced Cx36 gap junction coupling and protected against declines in coordinated [Ca2+ ] dynamics. Cx36 over-expression and S293 peptide also reduced apoptosis induced by pro-inflammatory cytokines. Critically, S293 peptide rescued gap junction coupling and [Ca2+ ] dynamics in islets from both db/db mice and a sub-set of T2D donors. Thus, recovering or enhancing Cx36 gap junction coupling can improve islet function in diabetes. KEY POINTS: Connexin-36 (Cx36) gap junction permeability and associated coordination of [Ca2+ ] dynamics is diminished in human type 2 diabetes (T2D) and mouse models of T2D. Enhancing Cx36 gap junction permeability protects against disruptions to the coordination of [Ca2+ ] dynamics. A novel peptide mimetic of the Cx36 c-terminal regulatory region protects against declines in Cx36 gap junction permeability. Pharmacological elevation in Cx36 or Cx36 peptide mimetic recovers [Ca2+ ] dynamics and glucose-stimulated insulin secretion in human T2D and mouse models of T2D.

Connexins Biology in the Pathophysiology of Retinal Diseases.

Connexins (Cx) are a family of transmembrane proteins that form gap junction intercellular channels that connect neighboring cells. These channels allow the passage of ions and other biomolecules smaller than 1 kDa, thereby synchronizing the cells both electrically and metabolically. Cxs are expressed in all retinal cell types and the diversity of Cx isoforms involved in the assembly of the channels provides a functional syncytium required for visual transduction. In this chapter, we summarize the status of current knowledge regarding Cx biology in retinal tissues and discuss how Cx dysfunction is associated with retinal disease pathophysiology. Although the contribution of Cx deficiency to retinal degeneration is not well understood, recent findings present Cx as a potential therapeutic target. Therefore, we will briefly discuss pharmacological approaches and gene therapies that are being explored to modulate Cx function and fight sight-threatening eye diseases.

Electrophysiological In Vitro Study of Long-Range Signal Transmission by Astrocytic Networks.

Astrocytes are diverse brain cells that form large networks communicating via gap junctions and chemical transmitters. Despite recent advances, the functions of astrocytic networks in information processing in the brain are not fully understood. In culture, brain slices, and in vivo, astrocytes, and neurons grow in tight association, making it challenging to establish whether signals that spread within astrocytic networks communicate with neuronal groups at distant sites, or whether astrocytes solely respond to their local environments. A multi-electrode array (MEA)-based device called AstroMEA is designed to separate neuronal and astrocytic networks, thus allowing to study the transfer of chemical and/or electrical signals transmitted via astrocytic networks capable of changing neuronal electrical behavior. AstroMEA demonstrates that cortical astrocytic networks can induce a significant upregulation in the firing frequency of neurons in response to a theta-burst charge-balanced biphasic current stimulation (5 pulses of 100 Hz × 10 with 200 ms intervals, 2 s total duration) of a separate neuronal-astrocytic group in the absence of direct neuronal contact. This result corroborates the view of astrocytic networks as a parallel mechanism of signal transmission in the brain that is separate from the neuronal connectome. Translationally, it highlights the importance of astrocytic network protection as a treatment target.

Locomotion modulates olfactory learning through proprioception in C. elegans.

Locomotor activities can enhance learning, but the underlying circuit and synaptic mechanisms are largely unknown. Here we show that locomotion facilitates aversive olfactory learning in C. elegans by activating mechanoreceptors in motor neurons, and transmitting the proprioceptive information thus generated to locomotion interneurons through antidromic-rectifying gap junctions. The proprioceptive information serves to regulate experience-dependent activities and functional coupling of interneurons that process olfactory sensory information to produce the learning behavior. Genetic destruction of either the mechanoreceptors in motor neurons, the rectifying gap junctions between the motor neurons and locomotion interneurons, or specific inhibitory synapses among the interneurons impairs the aversive olfactory learning. We have thus uncovered an unexpected role of proprioception in a specific learning behavior as well as the circuit, synaptic, and gene bases for this function.

Sodium channel subpopulations with distinct biophysical properties and subcellular localization enhance cardiac conduction.

Sodium (Na+) current is responsible for the rapid depolarization of cardiac myocytes that triggers the cardiac action potential upstroke. Recent studies have illustrated the presence of multiple pools of Na+ channels with distinct biophysical properties and subcellular localization, including clustering of channels at the intercalated disk and along the lateral membrane. Computational studies predict that Na+ channel clusters at the intercalated disk can regulate cardiac conduction via modulation of the narrow intercellular cleft between electrically coupled myocytes. However, these studies have primarily focused on the redistribution of Na+ channels between intercalated disk and lateral membranes and have not considered the distinct biophysical properties of the Na+ channel subpopulations. In this study, we use computational modeling to simulate computational models of single cardiac cells and one-dimensional cardiac tissues and predict the function of distinct Na+ channel subpopulations. Single-cell simulations predict that a subpopulation of Na+ channels with shifted steady-state activation and inactivation voltage dependency promotes an earlier action potential upstroke. In cardiac tissues that account for distinct subcellular spatial localization, simulations predict that shifted Na+ channels contribute to faster and more robust conduction in response to changes in tissue structure (i.e., cleft width), gap junctional coupling, and rapid pacing rates. Simulations predict that the intercalated disk-localized shifted Na+ channels contribute proportionally more to total Na+ charge than lateral membrane-localized Na+ channels. Importantly, our work supports the hypothesis that Na+ channel redistribution may be a critical mechanism by which cells can respond to perturbations to support fast and robust conduction.

Gap junctions regulate the development of neural circuits in the neocortex.

Gap junctions between cells are ubiquitously expressed in the developing brain. They are involved in major steps of neocortical development, including neurogenesis, cell migration, synaptogenesis, and neural circuit formation, and have been implicated in cortical column formation. Dysfunctional gap junctions can contribute to or even cause a variety of brain diseases. Although the role of gap junctions in neocortical development is better known, a comprehensive understanding of their functions is far from complete. Here we explore several critical open questions surrounding gap junctions and their involvement in neural circuit development. Addressing them will greatly impact our understanding of the fundamental mechanisms of neocortical structure and function as well as the etiology of brain disease.

Role of ephaptic coupling in discordant alternans domain sizes and action potential propagation in the heart.

Discordant alternans, the spatially out-of-phase alternation of the durations of propagating action potentials in the heart, has been linked to the onset of fibrillation, a major cardiac rhythm disorder. The sizes of the regions, or domains, within which these alternations are synchronized are critical in this link. However, computer models employing standard gap junction-based coupling between cells have been unable to reproduce simultaneously the small domain sizes and rapid action potential propagation speeds seen in experiments. Here we use computational methods to show that rapid wave speeds and small domain sizes are possible when a more detailed model of intercellular coupling that accounts for so-called ephaptic effects is used. We provide evidence that the smaller domain sizes are possible, because different coupling strengths can exist on the wavefronts, for which both ephaptic and gap-junction coupling are involved, in contrast to the wavebacks, where only gap-junction coupling plays an active role. The differences in coupling strength are due to the high density of fast-inward (sodium) channels known to localize on the ends of cardiac cells, which are only active (and thus engage ephaptic coupling) during wavefront propagation. Thus, our results suggest that this distribution of fast-inward channels, as well as other factors responsible for the critical involvement of ephaptic coupling in wave propagation, including intercellular cleft spacing, play important roles in increasing the vulnerability of the heart to life-threatening tachyarrhythmias. Our results, combined with the absence of short-wavelength discordant alternans domains in standard gap-junction-dominated coupling models, also provide evidence that both gap-junction and ephaptic coupling are critical in wavefront propagation and waveback dynamics.

Fish-specific N-terminal domain sequence in Connexin 39.4 plays an important role in zebrafish stripe formation by regulating the opening and closing of gap junctions and hemichannels.

BACKGROUND: Connexin 39.4 (Cx39.4) is involved in zebrafish (Danio rerio) skin patterning; when mutated, zebrafish exhibit a wavy stripe/labyrinth pattern instead of stripes. Cx39.4 is unique in that it has two additional serine/arginine (SR) residues, Ser2 and Arg3, at positions 2 and 3. Here, I investigated the role of these SR residues in Cx39.4 function. METHODS: To examine the SR residues in Cx39.4, mutants of the SR residues were generated. Voltage-clamp recordings were performed using Xenopus oocytes to characterize the channel properties of the mutants. Transgenic zebrafish expressing each mutant were generated, and the effects of each mutation on fish skin patterning were evaluated. RESULTS: The Cx39.4R3K mutant showed essentially the same properties as the wild-type (Cx39.4WT) in both electrophysiological analyses, leading to transgenic, complete phenotype rescue. Both the Cx39.4R3A mutant and deletion mutant of SR residues (Cx39.4delSR) showed a faster decay of gap junction activity and abnormal hemichannel activity, resulting in wide stripes and interstripes that indicate instability. Although the Cx39.4R3D mutant showed no channel activity in gap junctions or hemichannels, it caused unstable phenotypes in the transgene, namely a completely rescued phenotype in some individuals and loss of melanophores in others. CONCLUSIONS: The SR residues in the NT domain of Cx39.4 are critical for the regulation of channel function, which appears to determine skin patterning. GENERAL SIGNIFICANCE: These results elucidate the roles of the two SR residues unique to the NT domain of Cx39.4 in its channel function, which is important for zebrafish stripe pattern formation.

Genesis of a functional astrocyte syncytium in the developing mouse hippocampus.

Astrocytes are increasingly shown to operate as an isopotential syncytium in brain function. Protoplasmic astrocytes acquire this ability to functionally go beyond the single-cell level by evolving into a spongiform morphology, cytoplasmically connecting into a syncytium, and expressing a high density of K+ conductance. However, none of these cellular/functional features exist in neonatal newborn astrocytes, which imposes a basic question of when a functional syncytium evolves in the developing brain. Our results show that the spongiform morphology of individual astrocytes and their spatial organization all reach stationary levels by postnatal day (P) 15 in the hippocampal CA1 region. Functionally, astrocytes begin to uniformly express a mature level of passive K+ conductance by P11. We next used syncytial isopotentiality measurement to monitor the maturation of the astrocyte syncytium. In uncoupled P1 astrocytes, the substitution of endogenous K+ by a Na+ -electrode solution ([Na+ ]p ) resulted in the total elimination of the physiological membrane potential (VM ), and outward K+ conductance as predicted by the Goldman-Hodgkin-Katz (GHK) equation. As more astrocytes are coupled to each other through gap junctions during development, the [Na+ ]p -induced loss of physiological VM and the outward K+ conductance is progressively compensated by the neighboring astrocytes. By P15, a stably established syncytial isopotentiality (-73 mV), and a fully compensated outward K+ conductance appeared in all [Na+ ]p -recorded astrocytes. Thus, in view of the developmental timeframe wherein a singular syncytium is anatomically and functionally established for intra-syncytium K+ equilibration, an astrocyte syncytium becomes fully operational at P15 in the mouse hippocampus.

Interneuronal network model of theta-nested fast oscillations predicts differential effects of heterogeneity, gap junctions and short term depression for hyperpolarizing versus shunting inhibition.

Theta and gamma oscillations in the hippocampus have been hypothesized to play a role in the encoding and retrieval of memories. Recently, it was shown that an intrinsic fast gamma mechanism in medial entorhinal cortex can be recruited by optogenetic stimulation at theta frequencies, which can persist with fast excitatory synaptic transmission blocked, suggesting a contribution of interneuronal network gamma (ING). We calibrated the passive and active properties of a 100-neuron model network to capture the range of passive properties and frequency/current relationships of experimentally recorded PV+ neurons in the medial entorhinal cortex (mEC). The strength and probabilities of chemical and electrical synapses were also calibrated using paired recordings, as were the kinetics and short-term depression (STD) of the chemical synapses. Gap junctions that contribute a noticeable fraction of the input resistance were required for synchrony with hyperpolarizing inhibition; these networks exhibited theta-nested high frequency oscillations similar to the putative ING observed experimentally in the optogenetically-driven PV-ChR2 mice. With STD included in the model, the network desynchronized at frequencies above ~200 Hz, so for sufficiently strong drive, fast oscillations were only observed before the peak of the theta. Because hyperpolarizing synapses provide a synchronizing drive that contributes to robustness in the presence of heterogeneity, synchronization decreases as the hyperpolarizing inhibition becomes weaker. In contrast, networks with shunting inhibition required non-physiological levels of gap junctions to synchronize using conduction delays within the measured range.

Hypertensive Nephropathy: Unveiling the Possible Involvement of Hemichannels and Pannexons.

Hypertension is one of the most common risk factors for developing chronic cardiovascular diseases, including hypertensive nephropathy. Within the glomerulus, hypertension causes damage and activation of mesangial cells (MCs), eliciting the production of large amounts of vasoactive and proinflammatory agents. Accordingly, the activation of AT1 receptors by the vasoactive molecule angiotensin II (AngII) contributes to the pathogenesis of renal damage, which is mediated mostly by the dysfunction of intracellular Ca2+ ([Ca2+]i) signaling. Similarly, inflammation entails complex processes, where [Ca2+]i also play crucial roles. Deregulation of this second messenger increases cell damage and promotes fibrosis, reduces renal blood flow, and impairs the glomerular filtration barrier. In vertebrates, [Ca2+]i signaling depends, in part, on the activity of two families of large-pore channels: hemichannels and pannexons. Interestingly, the opening of these channels depends on [Ca2+]i signaling. In this review, we propose that the opening of channels formed by connexins and/or pannexins mediated by AngII induces the ATP release to the extracellular media, with the subsequent activation of purinergic receptors. This process could elicit Ca2+ overload and constitute a feed-forward mechanism, leading to kidney damage.

Gap junctions mediate discrete regulatory steps during fly spermatogenesis.

Gametogenesis requires coordinated signaling between germ cells and somatic cells. We previously showed that Gap junction (GJ)-mediated soma-germline communication is essential for fly spermatogenesis. Specifically, the GJ protein Innexin4/Zero population growth (Zpg) is necessary for somatic and germline stem cell maintenance and differentiation. It remains unknown how GJ-mediated signals regulate spermatogenesis or whether the function of these signals is restricted to the earliest stages of spermatogenesis. Here we carried out comprehensive structure/function analysis of Zpg using insights obtained from the protein structure of innexins to design mutations aimed at selectively perturbing different regulatory regions as well as the channel pore of Zpg. We identify the roles of various regulatory sites in Zpg in the assembly and maintenance of GJs at the plasma membrane. Moreover, mutations designed to selectively disrupt, based on size and charge, the passage of cargos through the Zpg channel pore, blocked different stages of spermatogenesis. Mutations were identified that progressed through early germline and soma development, but exhibited defects in entry to meiosis or sperm individualisation, resulting in reduced fertility or sterility. Our work shows that specific signals that pass through GJs regulate the transition between different stages of gametogenesis.

Temporal signals drive the emergence of multicellular information networks.

Coordinated responses to environmental stimuli are critical for multicellular organisms. To overcome the obstacles of cell-to-cell heterogeneity and noisy signaling dynamics within individual cells, cells must effectively exchange information with peers. However, the dynamics and mechanisms of collective information transfer driven by external signals are poorly understood. Here we investigate the calcium dynamics of neuronal cells that form confluent monolayers and respond to cyclic ATP stimuli in microfluidic devices. Using Granger inference to reconstruct the underlying causal relations between the cells, we find that the cells self-organize into spatially decentralized and temporally stationary networks to support information transfer via gap junction channels. The connectivity of the causal networks depends on the temporal profile of the external stimuli, where short periods, or long periods with small duty fractions, lead to reduced connectivity and fractured network topology. We build a theoretical model based on communicating excitable units that reproduces our observations. The model further predicts that connectivity of the causal network is maximal at an optimal communication strength, which is confirmed by the experiments. Together, our results show that information transfer between neuronal cells is externally regulated by the temporal profile of the stimuli and internally regulated by cell-cell communication.

Recording Synaptic Transmission from Auditory Mixed Synapses on the Mauthner Cells of Developing Zebrafish.

The Mauthner cells are a pair of large reticulospinal neurons that organize sensory-evoked tail flip responses in fishes. An identifiable group of auditory "mixed" (electrical and chemical) synaptic contacts known as "Large Myelinated Club endings" on these cells have provided a valuable model for the study of synaptic transmission in the vertebrate brain. While most of studies were performed in adult fish, we describe here methods that make possible recording synaptic transmission from these contacts in developing zebrafish, a genetically tractable vertebrate species which is uniquely amenable for combining synaptic physiology with live imaging and behavioral analysis.