Learning spiking neuronal networks with artificial neural networks: neural oscillations

First-principles-based modelings have been extremely successful in providing crucial insights and predictions for complex biological functions and phenomena. However, they can be hard to build and expensive to simulate for complex living systems. On the other hand, modern data-driven methods thrive at modeling many types of high-dimensional and noisy data. Still, the training and interpretation of these data-driven models remain challenging. Here, we combine the two types of methods to model stochastic neuronal network oscillations. Specifically, we develop a class of artificial neural networks to provide faithful surrogates to the high-dimensional, nonlinear oscillatory dynamics produced by a spiking neuronal network model. Furthermore, when the training data set is enlarged within a range of parameter choices, the artificial neural networks become generalizable to these parameters, covering cases in distinctly different dynamical regimes. In all, our work opens a new avenue for modeling complex neuronal network dynamics with artificial neural networks.

Stress relief as a natural resilience mechanism against depression-like behaviors

Relief, the appetitive state after the termination of aversive stimuli, is evolutionarily conserved. Understanding the behavioral role of this well-conserved phenomenon and its underlying neurobiological mechanisms are open and important questions. Here, we discover that the magnitude of relief from physical stress strongly correlates with individual resilience to depression-like behaviors in chronic stressed mice. Notably, blocking stress relief causes vulnerability to depression-like behaviors, whereas natural rewards supplied shortly after stress promotes resilience. Stress relief is mediated by reward-related mesolimbic dopamine neurons, which show minute-long, persistent activation after stress termination. Circuitry-wise, activation or inhibition of circuits downstream of the ventral tegmental area during the transient relief period bi-directionally regulates depression resilience. These results reveal an evolutionary function of stress relief in depression resilience and identify the neural substrate mediating this effect. Importantly, our data suggest a behavioral strategy of augmenting positive valence of stress relief with natural rewards to prevent depression.

Multi-band oscillations emerge from a simple spiking network

In the brain, coherent neuronal activities often appear simultaneously in multiple frequency bands, e.g., as combinations of alpha (8-12 Hz), beta (12.5-30 Hz), and gamma (30-120 Hz) oscillations, among others. These rhythms are believed to underlie information processing and cognitive functions and have been subjected to intense experimental and theoretical scrutiny. Computational modeling has provided a framework for the emergence of network-level oscillatory behavior from the interaction of spiking neurons. However, due to the strong nonlinear interactions between highly recurrent spiking populations, the interplay between cortical rhythms in multiple frequency bands has rarely been theoretically investigated. Many studies invoke multiple physiological timescales (e.g., various ion channels or multiple types of inhibitory neurons) or oscillatory inputs to produce rhythms in multi-bands. Here, we demonstrate the emergence of multi-band oscillations in a simple network consisting of one excitatory and one inhibitory neuronal population driven by constant input. First, we construct a data-driven, Poincaré section theory for robust numerical observations of single-frequency oscillations bifurcating into multiple bands. Then, we develop model reductions of the stochastic, nonlinear, high-dimensional neuronal network to capture the appearance of multi-band dynamics and the underlying bifurcations theoretically. Furthermore, when viewed within the reduced state space, our analysis reveals conserved geometrical features of the bifurcations on low-dimensional dynamical manifolds. These results suggest a simple geometric mechanism behind the emergence of multi-band oscillations without appealing to oscillatory inputs or multiple synaptic or neuronal timescales. Thus, our work points to unexplored regimes of stochastic competition between excitation and inhibition behind the generation of dynamic, patterned neuronal activities.

Multilevel monte carlo for cortical circuit models

Multilevel Monte Carlo (MLMC) methods aim to speed up computation of statistics from dynamical simulations. MLMC is easy to implement and is sometimes very effective, but its efficacy may depend on the underlying dynamics. We apply MLMC to networks of spiking neurons and assess its effectiveness on prototypical models of cortical circuitry under different conditions. We find that MLMC can be very efficient for computing reliable features, i.e., features of network dynamics that are reproducible upon repeated presentation of the same external forcing. In contrast, MLMC is less effective for complex, internally generated activity. Qualitative explanations are given using concepts from random dynamical systems theory.

A data-informed mean-field approach to mapping of cortical parameter landscapes

Constraining the many biological parameters that govern cortical dynamics is computationally and conceptually difficult because of the curse of dimensionality. This paper addresses these challenges by proposing (1) a novel data-informed mean-field (MF) approach to efficiently map the parameter space of network models; and (2) an organizing principle for studying parameter space that enables the extraction biologically meaningful relations from this high-dimensional data. We illustrate these ideas using a large-scale network model of the Macaque primary visual cortex. Of the 10-20 model parameters, we identify 7 that are especially poorly constrained, and use the MF algorithm in (1) to discover the firing rate contours in this 7D parameter cube. Defining a "biologically plausible" region to consist of parameters that exhibit spontaneous Excitatory and Inhibitory firing rates compatible with experimental values, we find that this region is a slightly thickened codimension-1 submanifold. An implication of this finding is that while plausible regimes depend sensitively on parameters, they are also robust and flexible provided one compensates appropriately when parameters are varied. Our organizing principle for conceptualizing parameter dependence is to focus on certain 2D parameter planes that govern lateral inhibition: Intersecting these planes with the biologically plausible region leads to very simple geometric structures which, when suitably scaled, have a universal character independent of where the intersections are taken. In addition to elucidating the geometry of the plausible region, this invariance suggests useful approximate scaling relations. Our study offers, for the first time, a complete characterization of the set of all biologically plausible parameters for a detailed cortical model, which has been out of reach due to the high dimensionality of parameter space.

Model Reduction Captures Stochastic Gamma Oscillations on Low-Dimensional Manifolds

Gamma frequency oscillations (25–140 Hz), observed in the neural activities within many brain regions, have long been regarded as a physiological basis underlying many brain functions, such as memory and attention. Among numerous theoretical and computational modeling studies, gamma oscillations have been found in biologically realistic spiking network models of the primary visual cortex. However, due to its high dimensionality and strong non-linearity, it is generally difficult to perform detailed theoretical analysis of the emergent gamma dynamics. Here we propose a suite of Markovian model reduction methods with varying levels of complexity and apply it to spiking network models exhibiting heterogeneous dynamical regimes, ranging from nearly homogeneous firing to strong synchrony in the gamma band. The reduced models not only successfully reproduce gamma oscillations in the full model, but also exhibit the same dynamical features as we vary parameters. Most remarkably, the invariant measure of the coarse-grained Markov process reveals a two-dimensional surface in state space upon which the gamma dynamics mainly resides. Our results suggest that the statistical features of gamma oscillations strongly depend on the subthreshold neuronal distributions. Because of the generality of the Markovian assumptions, our dimensional reduction methods offer a powerful toolbox for theoretical examinations of other complex cortical spatio-temporal behaviors observed in both neurophysiological experiments and numerical simulations.

[Impact of complicated myocardial injury on the clinical outcome of severe or critically ill COVID-19 patients]

Objective: To analyze the clinical characteristics of the severe or critically ill patients with novel coronavirus pneumonia (COVID-19), and evaluate the impact of complicated myocardial injury on the prognosis of these patients. Methods: A retrospective study was conducted in 54 patients who admitted to Tongji hospital from February 3, 2020 to February 24, 2020 and met the criteria of severe or critical conditions of COVID-19. The clinical characteristics and hospital mortality rate were analyzed and compared between the patients with or without myocardial injury, which was defined with 3 times higher serum cardiac troponin value. Results: The age of the 54 patients was 68.0(59.8, 74.3) years. Among all the patients, 24 (44.4%) patients were complicated with hypertension, 13 (24.1%) with diabetes, 8 (14.8%) with coronary heart disease, and 3 (5.6%) with previous cerebral infarction. During hospitalization, 24 (44.4%) of the patients were complicated with myocardial injury and 26 (48.1%) patients died in hospital. In-hospital mortality was significantly higher in patients with myocardial injury than in patients without myocardial injury (14 (60.9%) vs. 8 (25.8%), P=0.013). Moreover, the levels of C-reactive protein (153.6 (80.3, 240.7) ng/L vs. 49.8 (15.9, 101.9) ng/L) and N-terminal pro-B-type natriuretic peptide (852.0 (400.0, 2 315.3) ng/L vs. 197.0 (115.3, 631.0) ng/L) were significantly higher than patients without myocardial injury (all P<0.01). Conclusions: Prevalence of myocardial injury is high among severe or critically ill COVID-19 patients. Severe or critically ill COVID-19 patients with myocardial injury face a significantly higher risk of in-hospital mortality. The study suggests that it is important to monitor and manage the myocardial injury during hospitalization for severe or critically ill COVID-19 patients.

Loss of cell adhesion molecule CHL1 improves homeostatic adaptation and survival in hypoxic stress

Close homologue of L1 (CHL1) is a transmembrane cell adhesion molecule that is critical for brain development and for the maintenance of neural circuits in adults. Recent studies revealed that CHL1 has diverse roles and is involved in the regulation of recovery after spinal cord injury. CHL1 expression was downregulated in the cerebral cortex, hypothalamus, and brain stem after the induction of acute hypoxia (AH). In the current study, we sought to address the role of CHL1 in regulating homeostasis responses to hypoxia using CHL1-knockout (CHL1^−/−) mice. We found that, compared with wild-type littermates, CHL1^−/− mice showed a dramatically lower mortality rate and an augmented ventilatory response after they were subjected to AH. Immunofluorescence staining revealed that CHL1 was expressed in the carotid body (CB), the key oxygen sensor in rodents, and CHL1 expression level in the CB as assayed by western blot was decreased after hypoxic exposure. The number of glomus cells and the expression of tyrosine hydroxylase (a marker for glomus cells) in the CB of CHL1^−/− mice appeared to be increased compared with CHL1^+/+ mice. In addition, in the ex vivo CB preparation, hypoxia induced a significantly greater afferent nerve discharge in CHL1^−/− mice compared with CHL1^+/+ mice. Furthermore, the arterial blood pressure and plasma catecholamine levels of CHL1^−/− mice were also significantly higher than those of CHL1^+/+ mice. Our findings first demonstrate that CHL1 is a novel intrinsic factor that is involved in CB function and in the ventilatory response to AH.

Axoglial interaction via the notch receptor in oligodendrocyte differentiation

INTRODUCTION

Increasing evidence has revealed that the Notch signalling pathway is one of the pivotal systems that mediate oligodendrocyte development. The Notch receptor is a type I transmembrane molecule that represents a novel cellular signalling paradigm, namely, regulated intramembrane proteolysis (RIP).

METHOD

The typical Notch ligands, such as Delta, Serrate/Jagged and Lag2 (DSL), promote the formation of oligodendocyte precursor cells (OPCs) and maintain them in an uncommitted stage, thus retarding oligodendrocyte appearance in the central nervous system (CNS).

RESULTS

In contrast, our recent studies have revealed that F3/contactin, a GPI-linked neural adhesion molecule, interacts with Notch and speeds up the generation and maturation of oligodendrocytes.

CONCLUSIONS

Considering the distinct, albeit somewhat overlapping expression patterns of F3 and DSL in the CNS, the Notch receptor appears to function ligand-dependently during oligodendrocyte development. This multipotentiality may well designate the Notch receptor as one of the therapeutic targets that one can manoeuvre to treat demyelinating diseases, such as multiple sclerosis, that is characterised by chronic myelin degeneration.

Contactin associates with Na+ channels and increases their functional expression

Contactin (also known as F3, F11) is a surface glycoprotein that has significant homology with the beta2 subunit of voltage-gated Na(+) channels. Contactin and Na(+) channels can be reciprocally coimmunoprecipitated from brain homogenates, indicating association within a complex. Cells cotransfected with Na(+) channel Na(v)1.2alpha and beta1 subunits and contactin have threefold to fourfold higher peak Na(+) currents than cells with Na(v)1.2alpha alone, Na(v)1.2/beta1, Na(v)1.2/contactin, or Na(v)1.2/beta1/beta2. These cells also have a correspondingly higher saxitoxin binding, suggesting an increased Na(+) channel surface membrane density. Coimmunoprecipitation of different subunits from cell lines shows that contactin interacts specifically with the beta1 subunit. In the PNS, immunocytochemical studies show a transient colocalization of contactin and Na(+) channels at new nodes of Ranvier forming during remyelination. In the CNS, there is a particularly high level of colocalization of Na(+) channels and contactin at nodes both during development and in the adult. Contactin may thus significantly influence the functional expression and distribution of Na(+) channels in neurons.

Tenascin-R is a functional modulator of sodium channel beta subunits

Voltage-gated sodium channels isolated from mammalian brain are composed of alpha, beta1, and beta2 subunits. The alpha subunit forms the ion conducting pore of the channel, whereas the beta1 and beta2 subunits modulate channel function, as well as channel plasma membrane expression levels. beta1 and beta2 each contain a single, extracellular Ig-like domain with structural similarity to the neural cell adhesion molecule (CAM), myelin Po. beta2 contains strong amino acid homology to the third Ig domain and to the juxtamembrane region of F3/contactin. Many CAMs of the Ig superfamily have been shown to interact with extracellular matrix molecules. We hypothesized that beta2 may interact with tenascin-R (TN-R), an extracellular matrix molecule that is secreted by oligodendrocytes during myelination and that binds F3-contactin. We show here that cells expressing sodium channel beta1 or beta2 subunits are functionally modulated by TN-R. Transfected cells stably expressing beta1 or beta2 subunits initially recognized and then were repelled from TN-R substrates. The cysteine-rich amino-terminal domain of TN-R expressed as a recombinant peptide, termed EGF-L, appears to be responsible for the repellent effect on beta subunit-expressing cells. The epidermal growth factor-like repeats and fibronectin-like repeats 6-8 are most effective in the initial adhesion of beta subunit-expressing cells. Application of EGF-L to alphaIIAbeta1beta2 channels expressed in Xenopus oocytes potentiated expressed sodium currents without significantly altering current time course or the voltage dependence of current activation or inactivation. Thus, sodium channel beta subunits appear to function as CAMs, and TN-R may be an important regulator of sodium channel localization and function in neurons.

Role for myelin-associated glycoprotein as a functional tenascin-R receptor

The expression of the immunoglobulin superfamily member myelin-associated glycoprotein (MAG) and the extracellular matrix glycoprotein tenascin-R (TN-R) by oligodendrocytes overlaps in time and space. The two molecules can be neurite outgrowth-inhibitory or -promoting depending on the neuronal cell type and the environment in which they are presented. Here we show that the two molecules directly bind to each other in vitro and that binding sites on TN-R localize to two domains, the fibrinogen domain and the epidermal growth factor-like repeat domain with the N-terminal cysteine-rich stretch. We further show by a functional assay, namely the repulsion of MAG-transfected Chinese hamster ovary cells (CHO) cells from a TN-R substrate, that MAG is part of the signalling pathway of TN-R for cell repulsion. When coated as a uniform substrate, MAG was inhibitory for neurite outgrowth of hippocampal and cerebellar neurons in vitro, when compared to poly-L-lysine, while TN-R enhanced neurite outgrowth. When added to MAG, TN-R neutralized the neurite outgrowth-inhibitory effects of MAG, presumably by blocking the neurite outgrowth-inhibitory domain of MAG.

Defasciculation of neurites is mediated by tenascin-R and its neuronal receptor F3/11

Fasciculation and defasciculation of axons are major morphogenetic events in the formation of neuronal pathways during development. We have identified the extracellular matrix glycoprotein tenascin-R (TN-R) and its neuronal receptor, the immunoglobulin superfamily recognition molecule F3, as promoters of neurite defasciculation in cerebellar explant cultures. Perturbation of the interaction between these two molecules using both antibodies and an antisense oligonucleotide resulted in increased neurite fasciculation. The domains involved in defasciculation were identified as the N-terminal region of TN-R containing the cysteine-rich stretch and the 4.5 epidermal growth factor-like repeats and the immunoglobulin-like domains of F3. Fasciculation induced by antibodies and the antisense oligonucleotide could be reverted by a phorbol ester activator of protein kinase C, whereas the protein kinase inhibitor staurosporine increased fasciculation. Our observations indicate that defasciculated neurite outgrowth does not only depend on the reduction of the expression of fasciculation enhancing adhesion molecules, such as L1 and the neural cell adhesion molecule (NCAM), but also on recognition molecules that actively induce defasciculation by triggering second messenger systems.

Isolation of a tenascin-R binding protein from mouse brain membranes. A phosphacan-related chondroitin sulfate proteoglycan

We have isolated a chondroitin sulfate proteoglycan from mouse brain by affinity chromatography with a fragment of the extracellular matrix glycoprotein tenascin-R (TN-R) that comprises the amino-terminal cysteine-rich stretch and the 4.5 epidermal growth factor-like repeats. The isolated chondroitin sulfate proteoglycan has a molecular mass of 500-600 kDa and carries the HNK-1 carbohydrate epitope. Treatment with chondroitinase ABC reveals a major band of approximately 400 kDa and two minor bands at 200 and 150 kDa. Immunoblot analysis relates the molecule to phosphacan but not to the chondroitin sulfate proteoglycans neurocan and versican. Binding of the phosphacan-related molecule to the epidermal growth factor-like repeats of TN-R is Ca2+-dependent. Co-localization of the molecule with TN-R in the retina and optic nerve by immunocytochemistry suggests a functional relationship between the two molecules in vivo. Inhibition of neurite outgrowth from hippocampal neurons by the phosphacan-related molecule in vitro is neutralized by TN-R when coated as a uniform substrate. Furthermore, the phosphacan-related molecule neutralizes growth cone repulsion induced by TN-R coated as a sharp substrate boundary with or without prior treatment with chondroitinase ABC. These observations indicate that TN-R can interact with a phosphacan-related molecule and thereby modulate its inhibitory influence on neuritogenesis.

Signaling events following the interaction of the neuronal adhesion molecule F3 with the N-terminal domain of tenascin-R

Interaction between the extracellular matrix protein tenascin-R and the neuronal adhesion molecule F3 might be involved in the formation of neuronal networks. In this study, the fragment of tenascin-R comprising epithelial growth factor (EGF)-like repeats and the cysteine-rich NH2 terminal stretch (EGF-L), known to be inhibitory for growing neurites and repellent for growth cones, was used to investigate the signaling events following the F3/EGF-L interaction. We addressed this question using an in vitro test with F3-transfected Chinese hamster ovary (CHO) cells that allowed us to measure the kinetics, magnitude and specificity of the repellent effect resulting from the specific F3/EGF-L interaction. We showed that the repellent effect was counteracted by addition of the serine/threonine kinase and -phosphatase modulators (staurosporine, okadaic acid and H7) but not by modulators of tyrosine kinase or -phosphatases. This result indicates that the intracellular signals activated by the repellent effect involve a serine/threonine kinase pathway. Furthermore, the repellent effect of the EGF-L fragment for growth cones of cultured cerebellar neurons was also abolished by the identical modulators of serine/threonine kinase and -phosphatases. The inhibition of neurite outgrowth from hippocampal neurons by EGF-L was abolished in the presence of the serine threonine-kinase inhibitor H7. These results strongly suggest that the F3/tenascin-R interaction through EGF-L involves an intracellular activation of serine/ threonine kinase(s) in all F3-expressing cells tested.

Distinct effects of recombinant tenascin-R domains in neuronal cell functions and identification of the domain interacting with the neuronal recognition molecule F3/11

We have identified distinct domains of the rat extracellular matrix glycoprotein tenascin-R using recombinant fragments of the molecule that confer neuronal cell functions. In short-term adhesion assays (0.5 h), cerebellar neurons adhered best to the fragment representing the fibrinogen knob (FG), but also the fibronectin type III (FN) repeats 1-2 and 6-8. FG, FN1-2 and FN3-5 were the most repellent fragments for neuronal cell bodies. Neurites and growth cones were strongly repelled from areas coated with fragments containing the cysteine-rich stretch and the EGF-like domains (EGF-L), FN1-2, FN3-5 and FG. Polarization of morphology of hippocampal neurons was exclusively associated with FG, while EGF-L prevented neurite outgrowth altogether. The binding site of the neuronal receptor for tenascin-R, the immunoglobulin superfamily adhesion molecule F3/11, was localized to EGF-L. The combined observations show distinct, but also overlapping functions for the different tenascin-R domains. They further suggest the existence of multiple neuronal tenascin-R receptors which influence the response of neurons to their extracellular matrix environment.