An ensemble code in medial prefrontal cortex links prior events to outcomes during learning

Prefrontal neuronal ensembles link prior knowledge with novel actions during flexible action selection

We make decisions based on currently perceivable information or an internal model of the environment. The medial prefrontal cortex (mPFC) and its interaction with the hippocampus have been implicated in the latter, model-based decision-making; however, the underlying computational properties remain incompletely understood. We have examined mPFC spiking and hippocampal oscillatory activity while rats flexibly select new actions using a known associative structure of environmental cues and outcomes. During action selection, the mPFC reinstates representations of the associative structure. These awake reactivation events are accompanied by synchronous firings among neurons coding the associative structure and those coding actions. Moreover, their functional coupling is strengthened upon the reactivation events leading to adaptive actions. In contrast, only cue-coding neurons improve functional coupling during hippocampal sharp wave ripples. Thus, the lack of direct experience disconnects the mPFC from the hippocampus to independently form self-organized neuronal ensemble dynamics linking prior knowledge with novel actions.

Activity Subspaces in Medial Prefrontal Cortex Distinguish States of the World

Medial prefrontal cortex (mPfC) activity represents information about the state of the world, including present behavior, such as decisions, and the immediate past, such as short-term memory. Unknown is whether information about different states of the world are represented in the same mPfC neural population and, if so, how they are kept distinct. To address this, we analyze here mPfC population activity of male rats learning rules in a Y-maze, with self-initiated choice trials to an arm end followed by a self-paced return during the intertrial interval (ITI). We find that trial and ITI population activity from the same population fall into different low-dimensional subspaces. These subspaces encode different states of the world: multiple features of the task can be decoded from both trial and ITI activity, but the decoding axes for the same feature are roughly orthogonal between the two task phases, and the decodings are predominantly of features of the present during the trial but features of the preceding trial during the ITI. These subspace distinctions are carried forward into sleep, where population activity is preferentially reactivated in post-training sleep but differently for activity from the trial and ITI subspaces. Our results suggest that the problem of interference when representing different states of the world is solved in mPfC by population activity occupying different subspaces for the world states, which can be independently decoded by downstream targets and independently addressed by upstream inputs.SIGNIFICANCE STATEMENT Activity in the medial prefrontal cortex plays a role in representing the current and past states of the world. We show that during a maze task, the activity of a single population in medial prefrontal cortex represents at least two different states of the world. These representations were sequential and sufficiently distinct that a downstream population could separately read out either state from that activity. Moreover, the activity representing different states is differently reactivated in sleep. Different world states can thus be represented in the same medial prefrontal cortex population but in such a way that prevents potentially catastrophic interference between them.

Choice-selective sequences dominate in cortical relative to thalamic inputs to NAc to support reinforcement learning

How are actions linked with subsequent outcomes to guide choices? The nucleus accumbens, which is implicated in this process, receives glutamatergic inputs from the prelimbic cortex and midline regions of the thalamus. However, little is known about whether and how representations differ across these input pathways. By comparing these inputs during a reinforcement learning task in mice, we discovered that prelimbic cortical inputs preferentially represent actions and choices, whereas midline thalamic inputs preferentially represent cues. Choice-selective activity in the prelimbic cortical inputs is organized in sequences that persist beyond the outcome. Through computational modeling, we demonstrate that these sequences can support the neural implementation of reinforcement-learning algorithms, in both a circuit model based on synaptic plasticity and one based on neural dynamics. Finally, we test and confirm a prediction of our circuit models by direct manipulation of nucleus accumbens input neurons.

Learning differentially shapes prefrontal and hippocampal activity during classical conditioning

The ability to use sensory cues to inform goal-directed actions is a critical component of behavior. To study how sounds guide anticipatory licking during classical conditioning, we employed high-density electrophysiological recordings from the hippocampal CA1 area and the prefrontal cortex (PFC) in mice. CA1 and PFC neurons undergo distinct learning-dependent changes at the single-cell level and maintain representations of cue identity at the population level. In addition, reactivation of task-related neuronal assemblies during hippocampal awake Sharp-Wave Ripples (aSWRs) changed within individual sessions in CA1 and over the course of multiple sessions in PFC. Despite both areas being highly engaged and synchronized during the task, we found no evidence for coordinated single cell or assembly activity during conditioning trials or aSWR. Taken together, our findings support the notion that persistent firing and reactivation of task-related neural activity patterns in CA1 and PFC support learning during classical conditioning.

A novel knockout mouse model of the noncoding antisense Brain-Derived Neurotrophic Factor (Bdnf) gene displays increased endogenous Bdnf protein and improved memory function following exercise

Brain-derived neurotrophic factor (Bdnf) expression is tightly controlled at the transcriptional and post-transcriptional levels. Previously, we showed that inhibition of noncoding Bdnf antisense (Bdnf-AS) RNA upregulates Bdnf protein. Here, we generated a Bdnf-antisense knockout (Bdnf-AS KO) mouse model by deleting 6 kilobases upstream of Bdnf-AS. After verifying suppression of Bdnf-AS, baseline behavioral tests indicated no significant difference in knockout and wild type mice, except for enhanced cognitive function in the knockout mice in the Y-maze. Following acute involuntary exercise, Bdnf-AS KO mice were re-assessed and a significant increase in Bdnf mRNA and protein were observed. Following long-term involuntary exercise, we observed a significant increase in nonspatial and spatial memory in novel object recognition and Barnes maze tests in young and aged Bdnf-AS KO mice. Our data provides evidence for the beneficial effects of endogenous Bdnf upregulation and the synergistic effect of Bdnf-AS knockout on exercise and memory retention.

Reinforcement learning approaches to hippocampus-dependent flexible spatial navigation

Humans and non-human animals show great flexibility in spatial navigation, including the ability to return to specific locations based on as few as one single experience. To study spatial navigation in the laboratory, watermaze tasks, in which rats have to find a hidden platform in a pool of cloudy water surrounded by spatial cues, have long been used. Analogous tasks have been developed for human participants using virtual environments. Spatial learning in the watermaze is facilitated by the hippocampus. In particular, rapid, one-trial, allocentric place learning, as measured in the delayed-matching-to-place variant of the watermaze task, which requires rodents to learn repeatedly new locations in a familiar environment, is hippocampal dependent. In this article, we review some computational principles, embedded within a reinforcement learning framework, that utilise hippocampal spatial representations for navigation in watermaze tasks. We consider which key elements underlie their efficacy, and discuss their limitations in accounting for hippocampus-dependent navigation, both in terms of behavioural performance (i.e. how well do they reproduce behavioural measures of rapid place learning) and neurobiological realism (i.e. how well do they map to neurobiological substrates involved in rapid place learning). We discuss how an actor-critic architecture, enabling simultaneous assessment of the value of the current location and of the optimal direction to follow, can reproduce one-trial place learning performance as shown on watermaze and virtual delayed-matching-to-place tasks by rats and humans, respectively, if complemented with map-like place representations. The contribution of actor-critic mechanisms to delayed-matching-to-place performance is consistent with neurobiological findings implicating the striatum and hippocampo-striatal interaction in delayed-matching-to-place performance, given that the striatum has been associated with actor-critic mechanisms. Moreover, we illustrate that hierarchical computations embedded within an actor-critic architecture may help to account for aspects of flexible spatial navigation. The hierarchical reinforcement learning approach separates trajectory control via a temporal-difference error from goal selection via a goal prediction error and may account for flexible, trial-specific, navigation to familiar goal locations, as required in some arm-maze place memory tasks, although it does not capture one-trial learning of new goal locations, as observed in open field, including watermaze and virtual, delayed-matching-to-place tasks. Future models of one-shot learning of new goal locations, as observed on delayed-matching-to-place tasks, should incorporate hippocampal plasticity mechanisms that integrate new goal information with allocentric place representation, as such mechanisms are supported by substantial empirical evidence.

Canonical goal-selective representations are absent from prefrontal cortex in a spatial working memory task requiring behavioral flexibility

The prefrontal cortex (PFC)'s functions are thought to include working memory, as its activity can reflect information that must be temporarily maintained to realize the current goal. We designed a flexible spatial working memory task that required rats to navigate - after distractions and a delay - to multiple possible goal locations from different starting points and via multiple routes. This made the current goal location the key variable to remember, instead of a particular direction or route to the goal. However, across a broad population of PFC neurons, we found no evidence of current-goal-specific memory in any previously reported form - that is differences in the rate, sequence, phase, or covariance of firing. This suggests that such patterns do not hold working memory in the PFC when information must be employed flexibly. Instead, the PFC grouped locations representing behaviorally equivalent task features together, consistent with a role in encoding long-term knowledge of task structure.

A review of learning in biologically plausible spiking neural networks

Artificial neural networks have been used as a powerful processing tool in various areas such as pattern recognition, control, robotics, and bioinformatics. Their wide applicability has encouraged researchers to improve artificial neural networks by investigating the biological brain. Neurological research has significantly progressed in recent years and continues to reveal new characteristics of biological neurons. New technologies can now capture temporal changes in the internal activity of the brain in more detail and help clarify the relationship between brain activity and the perception of a given stimulus. This new knowledge has led to a new type of artificial neural network, the Spiking Neural Network (SNN), that draws more faithfully on biological properties to provide higher processing abilities. A review of recent developments in learning of spiking neurons is presented in this paper. First the biological background of SNN learning algorithms is reviewed. The important elements of a learning algorithm such as the neuron model, synaptic plasticity, information encoding and SNN topologies are then presented. Then, a critical review of the state-of-the-art learning algorithms for SNNs using single and multiple spikes is presented. Additionally, deep spiking neural networks are reviewed, and challenges and opportunities in the SNN field are discussed.

Learning of distant state predictions by the orbitofrontal cortex in humans

Representations of our future environment are essential for planning and decision making. Previous research in humans has demonstrated that the hippocampus is a critical region for forming and retrieving associations, while the medial orbitofrontal cortex (OFC) is an important region for representing information about recent states. However, it is not clear how the brain acquires predictive representations during goal-directed learning. Here, we show using fMRI that while participants learned to find rewards in multiple different Y-maze environments, hippocampal activity was highest during initial exposure and then decayed across the remaining repetitions of each maze, consistent with a role in rapid encoding. Importantly, multivariate patterns in the OFC-VPFC came to represent predictive information about upcoming states approximately 30 s in the future. Our findings provide a mechanism by which the brain can build models of the world that span long-timescales to make predictions.

Medial Prefrontal Cortex Population Activity Is Plastic Irrespective of Learning

The prefrontal cortex (PFC) is thought to learn the relationships between actions and their outcomes. But little is known about what changes to population activity in PFC are specific to learning these relationships. Here we characterize the plasticity of population activity in the medial PFC (mPFC) of male rats learning rules on a Y-maze. First, we show that the population always changes its patterns of joint activity between the periods of sleep either side of a training session on the maze, regardless of successful rule learning during training. Next, by comparing the structure of population activity in sleep and training, we show that this population plasticity differs between learning and nonlearning sessions. In learning sessions, the changes in population activity in post-training sleep incorporate the changes to the population activity during training on the maze. In nonlearning sessions, the changes in sleep and training are unrelated. Finally, we show evidence that the nonlearning and learning forms of population plasticity are driven by different neuron-level changes, with the nonlearning form entirely accounted for by independent changes to the excitability of individual neurons, and the learning form also including changes to firing rate couplings between neurons. Collectively, our results suggest two different forms of population plasticity in mPFC during the learning of action-outcome relationships: one a persistent change in population activity structure decoupled from overt rule-learning, and the other a directional change driven by feedback during behavior.SIGNIFICANCE STATEMENT The PFC is thought to represent our knowledge about what action is worth doing in which context. But we do not know how the activity of neurons in PFC collectively changes when learning which actions are relevant. Here we show, in a trial-and-error task, that population activity in PFC is persistently changing, regardless of learning. Only during episodes of clear learning of relevant actions are the accompanying changes to population activity carried forward into sleep, suggesting a long-lasting form of neural plasticity. Our results suggest that representations of relevant actions in PFC are acquired by reward imposing a direction onto ongoing population plasticity.