How do cell assemblies encode information in the brain? (2023)

The brain is organized into anatomically distinct structures consisting of a large number of projection neurons. While this evolutionarily conserved neural circuit organization underlies the innate ability of animals to rapidly adapt to their environment, it can lead to biased perception and behavior. Although recent studies have begun to address the causal significance of projection neuron types as distinct computational units, how projection types are functionally organized among encoding variables during cognitive tasks remains unclear. This review focuses on the neural computation of decision making in the prefrontal cortex and discusses which decision variables are encoded by individual neurons, populations of neurons, and projection types, and how certain projection types limit decision making. In particular, we focus on the 'overrepresentation' of certain decision variables in the prefrontal cortex, reflecting the biological and subjective importance of the variable to the decision maker. We propose that task-specific overrepresentation in the prefrontal cortex is required to improve a given decision, whereas generalized overrepresentation of fundamental decision variables is associated with suboptimal decision biases, including pathological biases such as those found in patients with psychiatric disorders. This overrepresentation of fundamental decision variables in the prefrontal cortex appears to be severely limited by afferent and efferent connections that can be influenced by optogenetics. These ideas could provide important insights into potential therapeutic targets for psychiatric disorders, including addiction and depression.

The periventricular/periaqueductal gray matter (PAG/PVG) is critical to pain perception and is associated with pain-induced emotional perception. However, the electrophysiological properties of PAG/PVG in patients with chronic pain have been poorly studied. This study analyzed the oscillatory properties of local field potentials (LFPs) in the PAG/PVG of 18 patients with neuropathic pain. Power spectrum analysis and neural state analysis were applied to PAG/PVG-LFP. Neural state analysis is based on a dynamic approach to neural state identification and differentiates LFPs into distinct neural states, including single neural states based on one oscillation and combined neural states based on two paired oscillations. The duration and frequency of occurrence are used to quantify the dynamic characteristics of neuronal states. The results show that the combined neural states form three local networks based on neural oscillations responsible for the perceptual, sensory and affective components of pain. The first network arises from the interaction of delta vibrations with other vibrations and is responsible for encoding the perception of pain. The second network, responsible for encoding sensory pain information, has high gamma as the main node and is closely connected with other neural oscillations. A third network is responsible for encoding emotional pain information, in which beta oscillations play an important role. This study demonstrates that the combination of two neural oscillations in the PAG/PVG is critical for encoding the perceived, sensory and affective measures of pain.

In this updated article, we focus on "memory engrams", which are traces of long-term memories in the brain, and emphasize that they are dynamic rather than static. We begin by presenting key findings from neuroscience and psychology showing that memory traces are sometimes diffuse and unstable, suggesting that they are dynamically changing processes of consolidation and reconsolidation. Second, we introduce and discuss the concepts of cell assembly and imprinted cells. The former has been studied by psychological experiments and behavioral electrophysiology, and the latter is defined by the local combination of activity-dependent cellular markers and optogenetics to reveal the causal relationship between the activity of cell populations and changes in behavior. Third, we discuss the similarities and differences between cellular arrangements and the concept of engram cells to show the dynamics of memory engrams. We also discuss the advantages and issues of live-cell imaging, which has recently been developed to visualize multi-neuron activity. The final section proposes experimental strategies and background hypotheses for future engram research. The former facilitates the recording of cell assemblies from different brain regions during memory consolidation and reconsolidation, while the latter emphasizes the multipotency of neurons and regions that contribute to the dynamics of memory engrams in the working brain.

The biophysical mechanisms by which input signals elicit neuronal responses are well understood (a sufficiently large input alters the neuron's membrane potential, producing electrical impulses called action potentials or spikes), but there is good understanding of how neurons generate these spikes using It is currently not possible to encode signal information. Recent theoretical studies have focused on how neurons encode weak periodic signals (which cannot produce spikes by themselves) in noisy environments where random electrical fluctuations occur that do not encode information. Analysis of spike trains generated by a single neuron and two coupled neurons (simulated using a FitzHugh-Nagumo stochastic model) revealed that the relative timing of spikes can encode signal information. Analysis of spike trains using a symbolic approach identified preferred and rare spike patterns whose probabilities varied with the amplitude and frequency of the signal. To investigate whether this encoding mechanism also applies to ensembles of neurons, we here analyze the activity of a group of neurons when they perceive weak periodic signals. We found that, as in the case of one or two coupled neurons, the spike pattern probability now computed from the spike trains of all neurons depends on the amplitude and period of the signal, and thus on the pattern probability that encodes information about the signal. We also found that when a group of neurons perceives a signal, the resonance with the signal period or noise level is more pronounced than when only one or two coupled neurons perceive the signal. Neural coupling facilitates signal encoding because a group of neurons is able to encode small-amplitude signals that cannot be encoded if only one or two coupled neurons perceive it. Interestingly, we found that for a group of neurons, even a few interconnections can significantly improve the encoding of small-amplitude signals. Our results suggest that information encoded in preferred and rare spike patterns is a plausible mechanism by which populations of neurons can exploit the presence of neural noise to encode weakly periodic inputs.

Hippocampal vibrations, especially in the theta (6–12 Hz) and gamma (30–90 Hz) frequency bands, play an important role in a variety of cognitive functions. Theta and gamma oscillations exhibit cross-frequency coupling (CFC), in which the phase of theta rhythms modulates the amplitude of gamma oscillations, and this CFC is thought to reflect the dynamics of cellular assembly during cognition. Previous studies have reported that CFC intensity is associated with learning. However, the details of these dynamic relationships could not be elucidated. In the present study, we analyzed local field potentials recorded from the rat hippocampus during a learned rule switching task. The modulation index, an indicator of CFC strength, was higher in rule-driven behavior than in irregular conditions. The enhanced coupling between theta and high gamma oscillations (60–90 Hz) changed later in learning. In contrast, the coupling between theta and low gamma oscillations (30–60 Hz) did not change during learning. These results suggest that coupling between theta and gamma bands occurs during rule learning and that high and low gamma bands play different roles in rule switching.

This chapter describes the process of information processing. In general, information processing can be defined as a change in information that can be detected in some way by an observer or a system. In terms of a biosensory system, it can be a process that describes everything that is happening outside or inside the body, e.g. B. Sudden noises, changes in heart rate after waking up quickly or running, headaches, etc. From a cognitive perspective, information processing is a tool for trying to understand the process of memory storage or learning in humans.

Journal of Molecular Biology, Band 427, Ausgabe 12, 2015, S. 2244–2255

The chaperone GroEL is a cylindrical complex composed of two stacked heptamer rings, and its lid-like cofactor GroES forms a nanocage in which one polypeptide chain is transiently trapped and freely folds through aggregation. GroEL and GroES undergo an ATP-regulated interaction cycle for closing and opening the folding cage. Recent reports have shown that the presence of non-native substrate proteins alters the GroEL/ES response, shifting it from an asymmetric complex to a symmetric complex. In the asymmetric reaction mode, only one GroEL ring is connected to GroES, and the two rings act sequentially, via negative allosteric coupling. In symmetric mode, both GroEL rings are bound to GroES and are folded at the same time. Here, we note that the results of recent fluorescence resonance energy transfer-based assays used to quantify symmetric complexes strongly depend on the fluorophore pair used. Therefore, we developed a novel assay based on fluorescence cross-correlation spectroscopy to accurately measure the GroEL:GroES stoichiometry. This assay avoids the fluorophore labeling of GroEL and the use of GroEL cysteine ​​mutants. Our results show that the symmetric GroEL:GroES₂The complex colonizes essentially only in the presence of unfolded model proteins such as α-lactalbumin and α-casein, which "overstimulate" GroEL ATPase and uncouple negative GroEL interstitial allosterism. In contrast, asymmetric complexes dominate both in the absence of substrate and in the presence of foldable substrate proteins. Furthermore, the GroEL rings decouple and form a symmetrical GroEL:GroES₂The complex is inhibited at physiological ATP:ADP concentrations. We conclude that the asymmetric GroEL:GroES complex represents the predominantly folded active form of the chaperone.

Journal of Theoretical Biology, Band 360, 2014, S. 95-101

Chemotaxis or gradient tracking is important in many biological systems but is subject to noise. The way receptors are positioned on cells or sensing devices affects the quality of the inferences they can support about gradients, suggesting that their configuration can be optimized. We show that, for elliptical sensor devices, uneven receptor placement could be a potential method of biasing the distribution of gradient orientations following cell cancellation. Using information theory, we calculated the mutual information between the gradient and the binding mode to find the optimal arrangement of receptors identified by the gradient.

Neuropsychology, Group 82, 2016, Issues 18-30

People have varying degrees of success in acquiring a second language (L2). In many cases, lexical and syntactic skills remain insufficient, although some learners achieve native-language proficiency even if they start late in acquiring a second language. In this study, we used an online psycholinguistic language proficiency test and a neurophysiological measure of syntactic processing, Syntax Mismatch Negative (sMMN) to Local Protocol Violations, to compare the effects of syntactic processing among native English speakers (NS). Behavioral and neurophysiological markers and non-native speakers (NNS). Variable grammatical knowledge is measured by psycholinguistic tests. When NS hears ungrammatical phrases that do not agree between subject and verb (e.g. *we kick) MMN is extended to syntactically legal sentences (e.g.is the step). The more competent NNS also showed this difference, but the less competent NNS did not. The main cortical source of MMN responses was located in bilateral superior temporal regions, and the strength of the source of grammatical neural activity correlated significantly with the grammatical ability of individual L2 speakers, as shown by psycholinguistic testing. Because our results show that early MMN indices of morphosyntactic agreement violations are similar in native and non-native speakers with high grammatical proficiency, they seem consistent with the use of similar brain mechanisms for at least some aspects of L1 and L2 grammar.

Neuropsychology, Group 89, 2016, pp. 161-171

A growing body of research has implicated the left anterior temporal lobe (LATL) in combinatorial semantic processing. However, magnetoencephalography (MEG) studies have shown that this activity occurs quite early, at 200-250 ms, earlier than the most common time window for lexical-semantic effects. What kind of semantic composition can LATL perform in 200-250 milliseconds? We assume that LATL computes the early stages of synthesis, using only the most readily available lexical-semantic information as input. To test this, we varied the context sensitivity of pre-adjectives and assumed only context-independent intersecting adjectives (e.g.until,italian) should be formed in earlier time windows, whereas the composition of context-sensitive scalar adjectives (egquickly,big) shall be deferred until the explanation of the following nouns is fully resolved. Consistent with this, early combination effects in the left temporal cortex were only observed for cross-adjectives, although the effects in this study were more retrospective than previous reports. Collectively, our results suggest multiple stages of semantic formation, of which LATL may indicate the earliest.

Neural Computing, Band 454, 2021, S. 313-323

Domain matching methods aim to learn transferable models for unlabeled target domains. Recently, a wide range of domain matching models have been proposed to align representations by minimizing the distributional distance between different domains or adversarial training methods. However, most existing methods only adjust various features in the feature space and ignore the semantic features generated by classifiers, which may lead to misclassification of target samples near the source decision limits. In this paper, we propose a Semantic Adaptive Network (SAN) for unattended domain adaptation. SAN aligns domain representations at the functional and semantic level. SAN uses adversarial training to align different domain representations in feature space. In the semantic space, SAN obtains pseudo-labels of target samples via the nearest source semantic representation centroids, and then enforces the pseudo-labeled target semantic representations to be close to the corresponding source semantic representation centroids. Experiments with ImageCLEF-DA, Office-Home and VisDA datasets confirm the effectiveness and superiority of our model.

Neuron, Vol. 88, No. 4, 2015, pp. 805–818

The spatial arrangement of luminance increases (ON) and luminance decreases (OFF) falling on the retina provides a wealth of information that is used by the central visual pathway to construct a coherent representation of the visual scene. However, it was unclear how the polarity of brightness changes was reflected in the activity of cortical circuits. Using wide-field epifluorescence and two-photon imaging, we demonstrate a robust modular representation of luminance polarity (ON or OFF) in ferret primary visual cortex. Polarity-specific domains were found in both smooth changes in brightness and single bright and dark edges, and included neurons that were selective for orientation and direction of motion. The integration of orientation and polarity preferences is reflected in the selectivity and discriminative ability of most layer 2/3 neurons. We conclude that polarity selectivity is an integral feature of layer 2/3 neurons and ensures that the distinction between light and dark stimuli is available for further processing in downstream outer striate regions.