Partial information decomposition reveals that synergistic neural integration is greater downstream of recurrent information flow in organotypic cortical cultures.

The directionality of network information flow dictates how networks process information. A central component of information processing in both biological and artificial neural networks is their ability to perform synergistic integration-a type of computation. We established previously that synergistic integration varies directly with the strength of feedforward information flow. However, the relationships between both recurrent and feedback information flow and synergistic integration remain unknown. To address this, we analyzed the spiking activity of hundreds of neurons in organotypic cultures of mouse cortex. We asked how empirically observed synergistic integration-determined from partial information decomposition-varied with local functional network structure that was categorized into motifs with varying recurrent and feedback information flow. We found that synergistic integration was elevated in motifs with greater recurrent information flow beyond that expected from the local feedforward information flow. Feedback information flow was interrelated with feedforward information flow and was associated with decreased synergistic integration. Our results indicate that synergistic integration is distinctly influenced by the directionality of local information flow.

Revealing the Dynamics of Neural Information Processing with Multivariate Information Decomposition.

The varied cognitive abilities and rich adaptive behaviors enabled by the animal nervous system are often described in terms of information processing. This framing raises the issue of how biological neural circuits actually process information, and some of the most fundamental outstanding questions in neuroscience center on understanding the mechanisms of neural information processing. Classical information theory has long been understood to be a natural framework within which information processing can be understood, and recent advances in the field of multivariate information theory offer new insights into the structure of computation in complex systems. In this review, we provide an introduction to the conceptual and practical issues associated with using multivariate information theory to analyze information processing in neural circuits, as well as discussing recent empirical work in this vein. Specifically, we provide an accessible introduction to the partial information decomposition (PID) framework. PID reveals redundant, unique, and synergistic modes by which neurons integrate information from multiple sources. We focus particularly on the synergistic mode, which quantifies the "higher-order" information carried in the patterns of multiple inputs and is not reducible to input from any single source. Recent work in a variety of model systems has revealed that synergistic dynamics are ubiquitous in neural circuitry and show reliable structure-function relationships, emerging disproportionately in neuronal rich clubs, downstream of recurrent connectivity, and in the convergence of correlated activity. We draw on the existing literature on higher-order information dynamics in neuronal networks to illustrate the insights that have been gained by taking an information decomposition perspective on neural activity. Finally, we briefly discuss future promising directions for information decomposition approaches to neuroscience, such as work on behaving animals, multi-target generalizations of PID, and time-resolved local analyses.

Abstract knowledge gets concrete in the hippocampus.

Nieh, Schottdorf, and colleagues (Nature, 595, 80-84, 2021) recently showed that hippocampal neurons of mice form a low-dimensional representation that integrates concrete and abstract task information. The reported findings confirm a long-standing prediction regarding the nature of the cognitive map and potentially offer a tantalizing glimpse of the cognitive map itself.

Overview of computational models of hippocampus and related structures: Introduction to the special issue.

Extensive computational modeling has focused on the hippocampal formation and associated cortical structures. This overview describes some of the factors that have motivated the strong focus on these structures, including major experimental findings and their impact on computational models. This overview provides a framework for describing the topics addressed by individual articles in this special issue of the journal Hippocampus.

Dorsal hippocampus not always necessary in a radial arm maze delayed win-shift task.

Spatial working memory is important for foraging and navigating the environment. However, its neural underpinnings remain poorly understood. The hippocampus, known for its spatial coding and involvement in spatial memory, is widely understood to be necessary for spatial working memory when retention intervals increase beyond seconds into minutes. Here, we describe new evidence that the dorsal hippocampus is not always necessary for spatial working memory for retention intervals of 8 min. Rats were trained to perform a delayed spatial win shift radial arm maze task with an 8-min delay between study and test phases. We then tested whether bilateral inactivation of the dorsal hippocampus between the study and test phases impaired behavioral performance at test. Inactivation was achieved through a bilateral infusion of lidocaine. Performance following lidocaine was compared to control trials, in which, sterile phosphate buffered saline (PBS) was infused. Test performance did not differ between the lidocaine and PBS conditions, remaining high in each. To explore the possibility that this insensitivity to inactivation was a result of overtraining, a second cohort of animals received substantially less training prior to the infusions. In this second cohort, lidocaine infusions did significantly impair task performance. These data indicate that successful performance of a spatial win-shift task on the 8-arm maze need not always be hippocampally dependent.

Place cell assemblies remain intact, despite reduced phase precession, after cholinergic disruption.

The hippocampal theta rhythm is frequently viewed as a clocking mechanism that coordinates the spiking activity of neurons across the hippocampus to form coherent neural assemblies. Phase precession is a form of temporal coding evidencing this mechanism and is degraded following systemic pharmacological disruption of cholinergic signaling. However, whether neural assemblies are commensurately degraded, as would be predicted from a clocking mechanism hypothesis, remains unknown. To address this, we recorded the spiking activity of hippocampal place cells as rats completed laps on a circle track for chocolate drink before versus during the influence of a systemic muscarinic acetylcholine receptor antagonist. We compared the integrity of hippocampal ensembles using three approaches. The first approach used cross-correlogram (CCG) analyses to ask if the relative spike-timing between pairs of cells became less reliable. The second used a general linear model based analysis to ask whether the activity of simultaneously recorded neurons became any less predictive of the spiking activity of single neurons. Finally, the third approach used a reconstruction analysis to ask if the population activity was any less informative regarding the environmental position of the animal and whether theta sequences were impaired. The results of all three analyses paint a consistent picture: systemic cholinergic disruption did not degrade assembly integrity. These data demonstrate that place cell assemblies do not depend upon high quality phase precession.

Precise spike timing dynamics of hippocampal place cell activity sensitive to cholinergic disruption.

New memory formation depends on both the hippocampus and modulatory effects of acetylcholine. The mechanism by which acetylcholine levels in the hippocampus enable new encoding remains poorly understood. Here, we tested the hypothesis that cholinergic modulation supports memory formation by leading to structured spike timing in the hippocampus. Specifically, we tested if phase precession in dorsal CA1 was reduced under the influence of a systemic cholinergic antagonist. Unit and field potential were recorded from the dorsal CA1 of rats as they completed laps on a circular track for food rewards before and during the influence of the systemically administered acetylcholine muscarinic receptor antagonist scopolamine. We found that scopolamine significantly reduced phase precession of spiking relative to the field theta, and that this was due to a decrease in the frequency of the spiking rhythmicity. We also found that the correlation between position and theta phase was significantly reduced. This effect was not due to changes in spatial tuning as tuning remained stable for those cells analyzed. Similarly, it was not due to changes in lap-to-lap reliability of spiking onset or offset relative to either position or phase as the reliability did not decrease following scopolamine administration. These findings support the hypothesis that memory impairments that follow muscarinic blockade are the result of degraded spike timing in the hippocampus.

Examination of rhythmicity of extracellularly recorded neurons in the entorhinal cortex.

A number of studies have examined the theta-rhythmic modulation of neuronal firing in the hippocampal circuit. For extracellular recordings, this is often done by examining spectral properties of the spike-time autocorrelogram, most significantly, for validating the presence or absence of theta modulation across species. These techniques can show significant rhythmicity for high firing rate, highly rhythmic neurons; however, they are substantially biased by several factors including the peak firing rate of the neuron, the amount of time spent in the neuron's receptive field, and other temporal properties of the rhythmicity such as cycle-skipping. These limitations make it difficult to examine rhythmic modulation in neurons with low firing rates or when an animal has short dwell times within the firing field and difficult to compare rhythmicity under disparate experimental conditions when these factors frequently differ. Here, we describe in detail the challenges that researchers face when using these techniques and apply our findings to recent recordings from bat entorhinal grid cells, suggesting that they may have lacked enough data to examine theta rhythmicity robustly. We describe a more sensitive and statistically rigorous method using maximum likelihood estimation (MLE) of a parametric model of the lags within the autocorrelation window, which helps to alleviate some of the problems of traditional methods and was also unable to detect rhythmicity in bat grid cells. Using large batteries of simulated data, we explored the boundaries for which the MLE technique and the theta index can detect rhythmicity. The MLE technique is less sensitive to many features of the autocorrelogram and provides a framework for statistical testing to detect rhythmicity as well as changes in rhythmicity in individual sessions providing a substantial improvement over previous methods.

Cholinergic blockade reduces theta-gamma phase amplitude coupling and speed modulation of theta frequency consistent with behavioral effects on encoding.

Large-scale neural activation dynamics in the hippocampal-entorhinal circuit local field potential, observable as theta and gamma rhythms and coupling between these rhythms, is predictive of encoding success. Behavioral studies show that systemic administration of muscarinic acetylcholine receptor antagonists selectively impairs encoding, suggesting that they may also disrupt the coupling between the theta and gamma bands. Here, we tested the hypothesis that muscarinic antagonists selectively disrupt coupling between theta and gamma. Specifically, we characterized the effects of systemically administered scopolamine on movement-induced theta and gamma rhythms recorded in the superficial layers of the medial entorhinal cortex (MEC) of freely moving rats. We report the novel result that gamma power at the peak of theta was most reduced following muscarinic blockade, significantly shifting the phase of maximal gamma power to occur at later phases of theta. We also characterize the existence of multiple distinct gamma bands in the superficial layers of the MEC. Further, we observed that theta frequency was significantly less modulated by movement speed following muscarinic blockade. Finally, the slope relating speed to theta frequency, a correlate of familiarity with a testing enclosure, increased significantly less between the preinjection and recovery trials when scopolamine was administered during the intervening injection session than when saline was administered, suggesting that scopolamine reduced encoding of the testing enclosure. These data are consistent with computational models suggesting that encoding and retrieval occur during the peak and trough of theta, respectively, and support the theory that acetylcholine regulates the balance between encoding versus retrieval.

Grid cell spatial tuning reduced following systemic muscarinic receptor blockade.

Grid cells of the medial entorhinal cortex exhibit a periodic and stable pattern of spatial tuning that may reflect the output of a path integration system. This grid pattern has been hypothesized to serve as a spatial coordinate system for navigation and memory function. The mechanisms underlying the generation of this characteristic tuning pattern remain poorly understood. Systemic administration of the muscarinic antagonist scopolamine flattens the typically positive correlation between running speed and entorhinal theta frequency in rats. The loss of this neural correlate of velocity, an important signal for the calculation of path integration, raises the question of what influence scopolamine has on the grid cell tuning as a read out of the path integration system. To test this, the spatial tuning properties of grid cells were compared before and after systemic administration of scopolamine as rats completed laps on a circle track for food rewards. The results show that the spatial tuning of the grid cells was reduced following scopolamine administration. The tuning of head direction cells, in contrast, was not reduced by scopolamine. This is the first report to demonstrate a link between cholinergic function and grid cell tuning. This work suggests that the loss of tuning in the grid cell network may underlie the navigational disorientation observed in Alzheimer's patients and elderly individuals with reduced cholinergic tone.

Phase coding by grid cells in unconstrained environments: two-dimensional phase precession.

Action potential timing is thought to play a critical role in neural representation. For example, theta phase precession is a robust phenomenon exhibited by spatial cells of the rat entorhinal-hippocampal circuit. In phase precession, the time a neuron fires relative to the phase of theta rhythm (6-10 Hz) oscillations in the local field potential reduces uncertainty about the position of the animal. This relationship between neural firing and behavior has made precession an important constraint for hypothetical mechanisms of temporal coding. However, challenges exist in identifying what regulates the spike timing of these cells. We have developed novel analytical techniques for mapping between behavior and neural firing that provide sufficient sensitivity to examine features of grid cell phase coding in open environments. Here, we show robust, omnidirectional phase precession by entorhinal grid cells in openfield enclosures. We present evidence that full phase precession persists regardless of how close the animal comes to the center of a firing field. Many conjunctive grid cells, previously thought to be phase locked, also exhibited phase coding. However, we were unable to detect directional- or field-specific phase coding predicted by some variants of models. Finally, we present data that suggest bursting of layer II grid cells contributes to the bimodality of phase precession. We discuss implications of these observations for models of temporal coding and propose the utility of these techniques in other domains where behavior is aligned to neural spiking.

Moderate excitation leads to weakening of perceptual representations.

A fundamental goal of memory research is to specify the conditions that lead to the strengthening and weakening of neural representations. Several computational models of memory formation predict that learning effects should vary as a nonmonotonic function of the amount of excitation received by a neural representation. Specifically, moderate excitation should result in synaptic weakening, while strong excitation should result in synaptic strengthening. In vitro investigations of plasticity in rodents have provided support for this prediction at the level of single synapses. However, it remains unclear whether this principle scales beyond the synapse to cortical representations and manifests changes in behavior. To address this question, we used electroencephalogram pattern classification in human subjects to measure trial-by-trial fluctuations in stimulus processing, and we used a negative priming paradigm to measure learning effects. In keeping with the idea that moderate excitation leads to weakening, moderate levels of stimulus processing were associated with negative priming (slower subsequent responding to the stimulus), but higher and lower levels of stimulus processing were not associated with negative priming. These results suggest that the same principles that account for synaptic weakening in rodents can also account for diminished accessibility of perceptual representations in humans.

Cellular networks underlying human spatial navigation.

Place cells of the rodent hippocampus constitute one of the most striking examples of a correlation between neuronal activity and complex behaviour in mammals. These cells increase their firing rates when the animal traverses specific regions of its surroundings, providing a context-dependent map of the environment. Neuroimaging studies implicate the hippocampus and the parahippocampal region in human navigation. However, these regions also respond selectively to visual stimuli. It thus remains unclear whether rodent place coding has a homologue in humans or whether human navigation is driven by a different, visually based neural mechanism. We directly recorded from 317 neurons in the human medial temporal and frontal lobes while subjects explored and navigated a virtual town. Here we present evidence for a neural code of human spatial navigation based on cells that respond at specific spatial locations and cells that respond to views of landmarks. The former are present primarily in the hippocampus, and the latter in the parahippocampal region. Cells throughout the frontal and temporal lobes responded to the subjects' navigational goals and to conjunctions of place, goal and view.

Medial entorhinal cortex activates in a traveling wave in the rat.

Traveling waves are hypothesized to support the long-range coordination of anatomically distributed circuits. Whether separate strongly interacting circuits exhibit traveling waves remains unknown. The hippocampus exhibits traveling 'theta' waves and interacts strongly with the medial entorhinal cortex (MEC). To determine whether the MEC also activates in a traveling wave, we performed extracellular recordings of local field potentials (LFP) and multi-unit activity along the MEC. These recordings revealed progressive phase shifts in activity, indicating that the MEC also activates in a traveling wave. Variation in theta waveform along the region, generated by gradients in local physiology, contributed to the observed phase shifts. Removing waveform-related phase shifts left significant residual phase shifts. The residual phase shifts covaried with theta frequency in a manner consistent with those generated by weakly coupled oscillators. These results show that the coordination of anatomically distributed circuits could be enabled by traveling waves but reveal heterogeneity in the mechanisms generating those waves.

Correlated activity favors synergistic processing in local cortical networks in vitro at synaptically relevant timescales.

Neural information processing is widely understood to depend on correlations in neuronal activity. However, whether correlation is favorable or not is contentious. Here, we sought to determine how correlated activity and information processing are related in cortical circuits. Using recordings of hundreds of spiking neurons in organotypic cultures of mouse neocortex, we asked whether mutual information between neurons that feed into a common third neuron increased synergistic information processing by the receiving neuron. We found that mutual information and synergistic processing were positively related at synaptic timescales (0.05-14 ms), where mutual information values were low. This effect was mediated by the increase in information transmission-of which synergistic processing is a component-that resulted as mutual information grew. However, at extrasynaptic windows (up to 3,000 ms), where mutual information values were high, the relationship between mutual information and synergistic processing became negative. In this regime, greater mutual information resulted in a disproportionate increase in redundancy relative to information transmission. These results indicate that the emergence of synergistic processing from correlated activity differs according to timescale and correlation regime. In a low-correlation regime, synergistic processing increases with greater correlation, and in a high-correlation regime, synergistic processing decreases with greater correlation.

Computation is concentrated in rich clubs of local cortical networks.

To understand how neural circuits process information, it is essential to identify the relationship between computation and circuit organization. Rich clubs, highly interconnected sets of neurons, are known to propagate a disproportionate amount of information within cortical circuits. Here, we test the hypothesis that rich clubs also perform a disproportionate amount of computation. To do so, we recorded the spiking activity of on average ∼300 well-isolated individual neurons from organotypic cortical cultures. We then constructed weighted, directed networks reflecting the effective connectivity between the neurons. For each neuron, we quantified the amount of computation it performed based on its inputs. We found that rich-club neurons compute ∼160% more information than neurons outside of the rich club. The amount of computation performed in the rich club was proportional to the amount of information propagation by the same neurons. This suggests that in these circuits, information propagation drives computation. In total, our findings indicate that rich-club organization in effective cortical circuits supports not only information propagation but also neural computation.

Switching between internal and external modes: A multiscale learning principle.

Brains construct internal models that support perception, prediction, and action in the external world. Individual circuits within a brain also learn internal models of the local world of input they receive, in order to facilitate efficient and robust representation. How are these internal models learned? We propose that learning is facilitated by continual switching between internally biased and externally biased modes of processing. We review computational evidence that this mode-switching can produce an error signal to drive learning. We then consider empirical evidence for the instantiation of mode-switching in diverse neural systems, ranging from subsecond fluctuations in the hippocampus to wake-sleep alternations across the whole brain. We hypothesize that these internal/external switching processes, which occur at multiple scales, can drive learning at each scale. This framework predicts that (a) slower mode-switching should be associated with learning of more temporally extended input features and (b) disruption of switching should impair the integration of new information with prior information.

A neural network model of retrieval-induced forgetting.

Retrieval-induced forgetting (RIF) refers to the finding that retrieving a memory can impair subsequent recall of related memories. Here, the authors present a new model of how the brain gives rise to RIF in both semantic and episodic memory. The core of the model is a recently developed neural network learning algorithm that leverages regular oscillations in feedback inhibition to strengthen weak parts of target memories and to weaken competing memories. The authors use the model to address several puzzling findings relating to RIF, including why retrieval practice leads to more forgetting than simply presenting the target item, how RIF is affected by the strength of competing memories and the strength of the target (to-be-retrieved) memory, and why RIF sometimes generalizes to independent cues and sometimes does not. For all of these questions, the authors show that the model can account for existing results, and they generate novel predictions regarding boundary conditions on these results.

Learning your way around town: how virtual taxicab drivers learn to use both layout and landmark information.

By having subjects drive a virtual taxicab through a computer-rendered town, we examined how landmark and layout information interact during spatial navigation. Subject-drivers searched for passengers, and then attempted to take the most efficient route to the requested destinations (one of several target stores). Experiment 1 demonstrated that subjects rapidly learn to find direct paths from random pickup locations to target stores. Experiment 2 varied the degree to which landmark and layout cues were preserved across two successively learned towns. When spatial layout was preserved, transfer was low if only target stores were altered, and high if both target stores and surrounding buildings were altered, even though in the latter case all local views were changed. This suggests that subjects can rapidly acquire a survey representation based on the spatial layout of the town and independent of local views, but that subjects will rely on local views when present, and are harmed when associations between previously learned landmarks are disrupted. We propose that spatial navigation reflects a hierarchical system in which either layout or landmark information is sufficient for orienting and wayfinding; however, when these types of cues conflict, landmarks are preferentially used.

Human theta oscillations related to sensorimotor integration and spatial learning.

oscillations in the rat hippocampus have been implicated in sensorimotor integration (Bland, 1986), especially during exploratory and wayfinding behavior. We propose that human cortical activity coordinates sensory information with a motor plan to guide wayfinding behavior to known goal locations. To test this hypothesis, we analyzed invasive recordings from epileptic patients while they performed a spatially immersive, virtual taxi driver task. Consistent with this hypothesis, we found oscillations during both exploratory search and goal-seeking behavior and, in particular, during virtual movement, when sensory information and motor planning were both in flux, compared with periods of self-initiated stillness. oscillations had different topographic and spectral characteristics during searching than during goal-seeking, suggesting that different cortical networks exhibit depending on which cognitive functions are driving behavior (spatial learning during exploration vs orienting to a learned representation during goal-seeking). In contrast, oscillations in the beta band appeared to be related to simple motor planning, likely a variant of the Rolandic mu rhythm. These findings suggest that human cortical oscillations act to coordinate sensory and motor brain activity in various brain regions to facilitate exploratory learning and navigational planning.