Fast Hebbian plasticity and working memory

Maintenance of delay-period activity in working memory task is modulated by local network structure

Revealing the relationship between neural network structure and function is one central theme of neuroscience. In the context of working memory (WM), anatomical data suggested that the topological structure of microcircuits within WM gradient network may differ, and the impact of such structural heterogeneity on WM activity remains unknown. Here, we proposed a spiking neural network model that can replicate the fundamental characteristics of WM: delay-period neural activity involves association cortex but not sensory cortex. First, experimentally observed receptor expression gradient along the WM gradient network is reproduced by our network model. Second, by analyzing the correlation between different local structures and duration of WM activity, we demonstrated that small-worldness, excitation-inhibition balance, and cycle structures play crucial roles in sustaining WM-related activity. To elucidate the relationship between the structure and functionality of neural networks, structural circuit gradients in brain should also be subject to further measurement. Finally, combining anatomical data, we simulated the duration of WM activity across different brain regions, its maintenance relies on the interaction between local and distributed networks. Overall, network structural gradient and interaction between local and distributed networks are of great significance for WM.

Working memory expedites the processing of visual signals within the extrastriate cortex

Working memory is the ability to maintain information in the absence of sensory input. In this study, we investigated how working memory benefits processing in visual areas. Using a measure of phase consistency to detect the arrival time of visual signals to the middle temporal (MT) area, we assessed the impact of working memory on the speed of sensory processing. We recorded from MT neurons in two monkeys during a spatial working memory task with visual probes. When the memorized location closely matches the receptive field center of the recording site, visual input arrives sooner, but if the memorized location does not match the receptive field center then the arrival of visual information is delayed. Thus, working memory expedites the arrival of visual input in MT. These results reveal that even in the absence of firing rate changes, working memory can still benefit the processing of information within sensory areas.

Incremental induction of NMDAR-STP and NMDAR-LTP in the CA1 area of ventral hippocampal slices relies on graded activation of discrete NMDA receptors

N-methyl-d-aspartate receptor (NMDAR)-dependent short- and long-term types of potentiation (STP and LTP, respectively) are frequently studied in the CA1 area of dorsal hippocampal slices (DHS). Far less is known about the NMDAR dependence of STP and LTP in ventral hippocampal slices (VHS), where both types of potentiation are smaller in magnitude than in the DHS. Here, we first briefly review our knowledge about the NMDAR dependence of STP and LTP and some other forms of synaptic plasticity. We then show in new experiments that the decay of NMDAR-STP in VHS, similar to dorsal hippocampal NMDAR-STP, is not time- but activity-dependent. We also demonstrate that the induction of submaximal levels of NMDAR-STP and NMDAR-LTP in VHS differs from the induction of saturated levels of plasticity in terms of their sensitivity to subunit-preferring NMDAR antagonists. These data suggest that activation of distinct NMDAR subtypes in a population of neurons results in an incremental increase in the induction of different phases of potentiation with changing sensitivity to pharmacological agents. Differences in pharmacological sensitivity, which arise due to differences in the levels of agonist-evoked biological response, might explain the disparity of the results concerning NMDAR subunit involvement in the induction of NMDAR-dependent plasticity.This article is part of a discussion meeting issue 'Long-term potentiation: 50 years on'.

Promiscuous involvement of metabotropic glutamate receptors in the storage of N-methyl-d-aspartate receptor-dependent short-term potentiation

Short- and long-term forms of N-methyl-d-aspartate receptor (NMDAR)-dependent potentiation (most commonly termed short-term potentiation (STP) and long-term potentiation (LTP)) are co-induced in hippocampal slices by theta-burst stimulation, which mimics naturally occurring patterns of neuronal activity. While NMDAR-dependent LTP (NMDAR-LTP) is said to be the cellular correlate of long-term memory storage, NMDAR-dependent STP (NMDAR-STP) is thought to underlie the encoding of shorter-lasting memories. The mechanisms of NMDAR-LTP have been researched much more extensively than those of NMDAR-STP, which is characterized by its extreme stimulation dependence. Thus, in the absence of low-frequency test stimulation, which is used to test the magnitude of potentiation, NMDAR-STP does not decline until the stimulation is resumed. NMDAR-STP represents, therefore, an inverse variant of Hebbian synaptic plasticity, illustrating that inactive synapses can retain their strength unchanged until they become active again. The mechanisms, by which NMDAR-STP is stored in synapses without a decrement, are unknown and we report here that activation of metabotropic glutamate receptors may be critical in maintaining the potentiated state of synaptic transmission. This article is part of a discussion meeting issue 'Long-term potentiation: 50 years on'.

Cage effects on synaptic plasticity and its modulation in a mouse model of fragile X syndrome

Fragile X syndrome (FXS) is characterized by impairments in executive function including different types of learning and memory. Long-term potentiation (LTP), thought to underlie the formation of memories, has been studied in the Fmr1 mouse model of FXS. However, there have been many discrepancies in the literature with inconsistent use of littermate and non-littermate Fmr1 knockout (KO) and wild-type (WT) control mice. Here, the influence of the breeding strategy (cage effect) on short-term potentiation (STP), LTP, contextual fear conditioning (CFC), expression of N-methyl-d-aspartate receptor (NMDAR) subunits and the modulation of NMDARs, were examined. The largest deficits in STP, LTP and CFC were found in KO mice compared with non-littermate WT. However, the expression of NMDAR subunits was unchanged in this comparison. Rather, NMDAR subunit (GluN1, 2A, 2B) expression was sensitive to the cage effect, with decreased expression in both WT and KO littermates compared with non-littermates. Interestingly, an NMDAR-positive allosteric modulator, UBP714, was only effective in potentiating the induction of LTP in non-littermate KO mice and not the littermate KO mice. These results suggest that commonly studied phenotypes in Fmr1 KOs are sensitive to the cage effect and therefore the breeding strategy may contribute to discrepancies in the literature.This article is part of a discussion meeting issue 'Long-term potentiation: 50 years on'.

A single fast Hebbian-like process enabling one-shot class addition in deep neural networks without backbone modification

One-shot learning, the ability to learn a new concept from a single instance, is a distinctive brain function that has garnered substantial interest in machine learning. While modeling physiological mechanisms poses challenges, advancements in artificial neural networks have led to performances in specific tasks that rival human capabilities. Proposing one-shot learning methods with these advancements, especially those involving simple mechanisms, not only enhance technological development but also contribute to neuroscience by proposing functionally valid hypotheses. Among the simplest methods for one-shot class addition with deep learning image classifiers is "weight imprinting," which uses neural activity from a new class image data as the corresponding new synaptic weights. Despite its simplicity, its relevance to neuroscience is ambiguous, and it often interferes with original image classification, which is a significant drawback in practical applications. This study introduces a novel interpretation where a part of the weight imprinting process aligns with the Hebbian rule. We show that a single Hebbian-like process enables pre-trained deep learning image classifiers to perform one-shot class addition without any modification to the original classifier's backbone. Using non-parametric normalization to mimic brain's fast Hebbian plasticity significantly reduces the interference observed in previous methods. Our method is one of the simplest and most practical for one-shot class addition tasks, and its reliance on a single fast Hebbian-like process contributes valuable insights to neuroscience hypotheses.

Enhanced expression of activity-regulated cytoskeleton-associated protein in the medial prefrontal cortex is involved in working memory performance

Working memory (WM) is a cognitive function important for guiding the on-going or upcoming behavior. A memory-related protein Arc (activity-regulated cytoskeleton-associated protein) is implicated in long-term memory consolidation. Recent evidence further suggests the involvement of hippocampal Arc in spatial WM. The medial prefrontal cortex (mPFC) is a key brain region mediating WM. However, the role of mPFC Arc in WM is still uncertain. To investigate whether mPFC Arc protein is involved in WM performance, delayed non-match to sample (DNMS) T-maze task was performed in rats with or without blocking new synthesis of mPFC Arc. In DNMS task, a 10-s or 30-s delay between the sample run and the choice run was given to evaluate WM performance. To block new Arc protein synthesis during the DNMS task, Arc antisense oligodeoxynucleotides (ODNs) were injected to the bilateral mPFC. The results show that, in rats without surgery for cannula implantation and subsequent intracerebral injection of ODNs, WM was functioning well during the DNMS task with a delay of 10 s but not 30 s, which was accompanied with a significantly increased level of mPFC Arc protein, indicating a possible link between enhanced Arc protein expression and the performance of WM. After preventing the enhancement of mPFC Arc protein expression with Arc antisense ODNs, rat's WM performance was impaired. These findings support enhanced mPFC Arc protein expression playing a role during WM performance.