Learning Enhances Sensory and Multiple Non-sensory Representations in Primary Visual Cortex

Hippocampus-Dependent Goal Localization by Head-Fixed Mice in Virtual Reality

The demonstration of the ability of rodents to navigate in virtual reality (VR) has made it an important behavioral paradigm for studying spatially modulated neuronal activity in these animals. However, their behavior in such simulated environments remains poorly understood. Here, we show that encoding and retrieval of goal location memory in mice head-fixed in VR depends on the postsynaptic scaffolding protein Shank2 and the dorsal hippocampus. In our newly developed virtual cued goal location task, a head-fixed mouse moves from one end of a virtual linear track to seek rewards given at a target location along the track. The mouse needs to visually recognize the target location and stay there for a short period of time to receive the reward. Transient pharmacological blockade of fast glutamatergic synaptic transmission in the dorsal hippocampus dramatically and reversibly impaired performance of this task. Encoding and updating of virtual cued goal location memory was impaired in mice deficient in the postsynaptic scaffolding protein Shank2, a mouse model of autism that exhibits impaired spatial learning in a real environment. These results highlight the crucial roles of the dorsal hippocampus and postsynaptic protein complexes in spatial learning and navigation in VR.

Context- and Output Layer-Dependent Long-Term Ensemble Plasticity in a Sensory Circuit

Sensory information is translated into ensemble representations by various populations of projection neurons in brain circuits. The dynamics of ensemble representations formed by distinct channels of output neurons in diverse behavioral contexts remains largely unknown. We studied the two output neuron layers in the olfactory bulb (OB), mitral and tufted cells, using chronic two-photon calcium imaging in awake mice. Both output populations displayed similar odor response profiles. During passive sensory experience, both populations showed reorganization of ensemble odor representations yet stable pattern separation across days. Intriguingly, during active odor discrimination learning, mitral but not tufted cells exhibited improved pattern separation, although both populations showed reorganization of ensemble representations. An olfactory circuitry model suggests that cortical feedback on OB interneurons can trigger both forms of plasticity. In conclusion, we show that different OB output layers display unique context-dependent long-term ensemble plasticity, allowing parallel transfer of non-redundant sensory information to downstream centers.

Video Abstract

•

Mitral and tufted cells in the olfactory bulb show similar odor-evoked responses

•

Passive odor experience reorganizes ensemble odor representations in both cell types

•

Associative odor learning specifically improves pattern separation in mitral cells

•

Cortical feedback can trigger both forms of plasticity in a network model

Virtual reality systems for rodents

Over the last decade virtual reality (VR) setups for rodents have been developed and utilized to investigate the neural foundations of behavior. Such VR systems became very popular since they allow the use of state-of-the-art techniques to measure neural activity in behaving rodents that cannot be easily used with classical behavior setups. Here, we provide an overview of rodent VR technologies and review recent results from related research. We discuss commonalities and differences as well as merits and issues of different approaches. A special focus is given to experimental (behavioral) paradigms in use. Finally we comment on possible use cases that may further exploit the potential of VR in rodent research and hence inspire future studies.

Random synaptic feedback weights support error backpropagation for deep learning

The brain processes information through multiple layers of neurons. This deep architecture is representationally powerful, but complicates learning because it is difficult to identify the responsible neurons when a mistake is made. In machine learning, the backpropagation algorithm assigns blame by multiplying error signals with all the synaptic weights on each neuron's axon and further downstream. However, this involves a precise, symmetric backward connectivity pattern, which is thought to be impossible in the brain. Here we demonstrate that this strong architectural constraint is not required for effective error propagation. We present a surprisingly simple mechanism that assigns blame by multiplying errors by even random synaptic weights. This mechanism can transmit teaching signals across multiple layers of neurons and performs as effectively as backpropagation on a variety of tasks. Our results help reopen questions about how the brain could use error signals and dispel long-held assumptions about algorithmic constraints on learning.

Functional and structural underpinnings of neuronal assembly formation in learning

Learning and memory are associated with the formation and modification of neuronal assemblies: populations of neurons that encode what has been learned and mediate memory retrieval upon recall. Functional studies of neuronal assemblies have progressed dramatically thanks to recent technological advances. Here we discuss how a focus on assembly formation and consolidation has provided a powerful conceptual framework to relate mechanistic studies of synaptic and circuit plasticity to behaviorally relevant aspects of learning and memory. Neurons are likely recruited to particular learning-related assemblies as a function of their relative excitabilities and synaptic activation, followed by selective strengthening of pre-existing synapses, formation of new connections and elimination of outcompeted synapses to ensure memory formation. Mechanistically, these processes involve linking transcription to circuit modification. They include the expression of immediate early genes and specific molecular and cellular events, supported by network-wide activities that are shaped and modulated by local inhibitory microcircuits.

Mice Can Use Second-Order, Contrast-Modulated Stimuli to Guide Visual Perception

Visual processing along the primate ventral stream takes place in a hierarchy of areas, characterized by an increase in both complexity of neuronal preferences and invariance to changes of low-level stimulus attributes. A basic type of invariance is form-cue invariance, where neurons have similar preferences in response to first-order stimuli, defined by changes in luminance, and global features of second-order stimuli, defined by changes in texture or contrast. Whether in mice, a now popular model system for early visual processing, visual perception can be guided by second-order stimuli is currently unknown. Here, we probed mouse visual perception and neural responses in areas V1 and LM using various types of second-order, contrast-modulated gratings with static noise carriers. These gratings differ in their spatial frequency composition and thus in their ability to invoke first-order mechanisms exploiting local luminance features. We show that mice can transfer learning of a coarse orientation discrimination task involving first-order, luminance-modulated gratings to the contrast-modulated gratings, albeit with markedly reduced discrimination performance. Consistent with these behavioral results, we demonstrate that neurons in area V1 and LM are less responsive and less selective to contrast-modulated than to luminance-modulated gratings, but respond with broadly similar preferred orientations. We conclude that mice can, at least in a rudimentary form, use second-order stimuli to guide visual perception.

SIGNIFICANCE STATEMENT

To extract object boundaries in natural scenes, the primate visual system does not only rely on differences in local luminance but can also take into account differences in texture or contrast. Whether the mouse, which has a much simpler visual system, can use such second-order information to guide visual perception is unknown. Here we tested mouse perception of second-order, contrast-defined stimuli and measured their neural representations in two areas of visual cortex. We find that mice can use contrast-defined stimuli to guide visual perception, although behavioral performance and neural representations were less robust than for luminance-defined stimuli. These findings shed light on basic steps of feature extraction along the mouse visual cortical hierarchy, which may ultimately lead to object recognition.

Experience-dependent spatial expectations in mouse visual cortex

In generative models of brain function, internal representations are used to generate predictions of sensory input, yet little is known about how internal models influence sensory processing. Here we show that, with experience in a virtual environment, the activity of neurons in layer 2/3 of mouse primary visual cortex (V1) becomes increasingly informative of spatial location. We found that a subset of V1 neurons exhibited responses that were predictive of the upcoming visual stimulus in a spatially dependent manner and that the omission of an expected stimulus drove strong responses in V1. Stimulus-predictive responses also emerged in V1-projecting anterior cingulate cortex axons, suggesting that anterior cingulate cortex serves as a source of predictions of visual input to V1. These findings are consistent with the hypothesis that visual cortex forms an internal representation of the visual scene based on spatial location and compares this representation with feed-forward visual input.

Induction and Consolidation of Calcium-Based Homo- and Heterosynaptic Potentiation and Depression

The adaptive mechanisms of homo- and heterosynaptic plasticity play an important role in learning and memory. In order to maintain plasticity-induced changes for longer time scales (up to several days), they have to be consolidated by transferring them from a short-lasting early-phase to a long-lasting late-phase state. The underlying processes of this synaptic consolidation are already well-known for homosynaptic plasticity, however, it is not clear whether the same processes also enable the induction and consolidation of heterosynaptic plasticity. In this study, by extending a generic calcium-based plasticity model with the processes of synaptic consolidation, we show in simulations that indeed heterosynaptic plasticity can be induced and, furthermore, consolidated by the same underlying processes as for homosynaptic plasticity. Furthermore, we show that by local diffusion processes the heterosynaptic effect can be restricted to a few synapses neighboring the homosynaptically changed ones. Taken together, this generic model reproduces many experimental results of synaptic tagging and consolidation, provides several predictions for heterosynaptic induction and consolidation, and yields insights into the complex interactions between homo- and heterosynaptic plasticity over a broad variety of time (minutes to days) and spatial scales (several micrometers).

Hunger-Dependent Enhancement of Food Cue Responses in Mouse Postrhinal Cortex and Lateral Amygdala

The needs of the body can direct behavioral and neural processing toward motivationally relevant sensory cues. For example, human imaging studies have consistently found specific cortical areas with biased responses to food-associated visual cues in hungry subjects, but not in sated subjects. To obtain a cellular-level understanding of these hunger-dependent cortical response biases, we performed chronic two-photon calcium imaging in postrhinal association cortex (POR) and primary visual cortex (V1) of behaving mice. As in humans, neurons in mouse POR, but not V1, exhibited biases toward food-associated cues that were abolished by satiety. This emergent bias was mirrored by the innervation pattern of amygdalo-cortical feedback axons. Strikingly, these axons exhibited even stronger food cue biases and sensitivity to hunger state and trial history. These findings highlight a direct pathway by which the lateral amygdala may contribute to state-dependent cortical processing of motivationally relevant sensory cues.

Air-Track: a real-world floating environment for active sensing in head-fixed mice

Natural behavior occurs in multiple sensory and motor modalities and in particular is dependent on sensory feedback that constantly adjusts behavior. To investigate the underlying neuronal correlates of natural behavior, it is useful to have access to state-of-the-art recording equipment (e.g., 2-photon imaging, patch recordings, etc.) that frequently requires head fixation. This limitation has been addressed with various approaches such as virtual reality/air ball or treadmill systems. However, achieving multimodal realistic behavior in these systems can be challenging. These systems are often also complex and expensive to implement. Here we present "Air-Track," an easy-to-build head-fixed behavioral environment that requires only minimal computational processing. The Air-Track is a lightweight physical maze floating on an air table that has all the properties of the "real" world, including multiple sensory modalities tightly coupled to motor actions. To test this system, we trained mice in Go/No-Go and two-alternative forced choice tasks in a plus maze. Mice chose lanes and discriminated apertures or textures by moving the Air-Track back and forth and rotating it around themselves. Mice rapidly adapted to moving the track and used visual, auditory, and tactile cues to guide them in performing the tasks. A custom-controlled camera system monitored animal location and generated data that could be used to calculate reaction times in the visual and somatosensory discrimination tasks. We conclude that the Air-Track system is ideal for eliciting natural behavior in concert with virtually any system for monitoring or manipulating brain activity.

Secondary visual cortex is critical to the expression of surprise-induced enhancements in cue associability in rats

Considerable evidence indicates that reinforcement prediction error, the difference between the obtained and expected reinforcer values, modulates attention to potential cues for reinforcement. The surprising delivery or omission of a reinforcer enhances the associability of the stimuli that were present when the error was induced, so that they more readily enter into new associations in the future. Previous research from our laboratory identified brain circuit elements critical to the enhancement of stimulus associability by omission of an expected event and to the subsequent expression of that altered associability in more rapid learning. A key finding was that the rat posterior parietal cortex was essential during the encoding, consolidation and retrieval of associability memories that were altered by the surprising omission of an expected event in a serial prediction task. Here, we found that the function of adjacent secondary visual cortex was critical only to the expression of altered cue associability in that same task. This specialization of function is discussed in the context of broader cortical and subcortical networks for modulation of attention in associative learning, as well as recent anatomical investigations that suggest that the rodent posterior parietal cortex overlaps with and may subsume secondary visual cortex.

A Mouse Model of Visual Perceptual Learning Reveals Alterations in Neuronal Coding and Dendritic Spine Density in the Visual Cortex

Visual perceptual learning (VPL) can improve spatial vision in normally sighted and visually impaired individuals. Although previous studies of humans and large animals have explored the neural basis of VPL, elucidation of the underlying cellular and molecular mechanisms remains a challenge. Owing to the advantages of molecular genetic and optogenetic manipulations, the mouse is a promising model for providing a mechanistic understanding of VPL. Here, we thoroughly evaluated the effects and properties of VPL on spatial vision in C57BL/6J mice using a two-alternative, forced-choice visual water task. Briefly, the mice underwent prolonged training at near the individual threshold of contrast or spatial frequency (SF) for pattern discrimination or visual detection for 35 consecutive days. Following training, the contrast-threshold trained mice showed an 87% improvement in contrast sensitivity (CS) and a 55% gain in visual acuity (VA). Similarly, the SF-threshold trained mice exhibited comparable and long-lasting improvements in VA and significant gains in CS over a wide range of SFs. Furthermore, learning largely transferred across eyes and stimulus orientations. Interestingly, learning could transfer from a pattern discrimination task to a visual detection task, but not vice versa. We validated that this VPL fully restored VA in adult amblyopic mice and old mice. Taken together, these data indicate that mice, as a species, exhibit reliable VPL. Intrinsic signal optical imaging revealed that mice with perceptual training had higher cut-off SFs in primary visual cortex (V1) than those without perceptual training. Moreover, perceptual training induced an increase in the dendritic spine density in layer 2/3 pyramidal neurons of V1. These results indicated functional and structural alterations in V1 during VPL. Overall, our VPL mouse model will provide a platform for investigating the neurobiological basis of VPL.

Rule learning enhances structural plasticity of long-range axons in frontal cortex

Rules encompass cue-action-outcome associations used to guide decisions and strategies in a specific context. Subregions of the frontal cortex including the orbitofrontal cortex (OFC) and dorsomedial prefrontal cortex (dmPFC) are implicated in rule learning, although changes in structural connectivity underlying rule learning are poorly understood. We imaged OFC axonal projections to dmPFC during training in a multiple choice foraging task and used a reinforcement learning model to quantify explore–exploit strategy use and prediction error magnitude. Here we show that rule training, but not experience of reward alone, enhances OFC bouton plasticity. Baseline bouton density and gains during training correlate with rule exploitation, while bouton loss correlates with exploration and scales with the magnitude of experienced prediction errors. We conclude that rule learning sculpts frontal cortex interconnectivity and adjusts a thermostat for the explore–exploit balance.

The orbitofrontal cortex is associated with foraging behaviour yet the structural changes underlying such rule-based learning remain unclear. Here, the authors imaged OFC axons throughout a digging-based odour discrimination task and found correlations between the rate of bouton turnover and the behavioural strategies of individual mice.

Arc-Expressing Neuronal Ensembles Supporting Pattern Separation Require Adrenergic Activity in Anterior Piriform Cortex: An Exploration of Neural Constraints on Learning

Arc ensembles in adult rat olfactory bulb (OB) and anterior piriform cortex (PC) were assessed after discrimination training on highly similar odor pairs. Nonselective α- and β-adrenergic antagonists or saline were infused in the OB or anterior PC during training. OB adrenergic blockade slowed, but did not prevent, odor discrimination learning. After criterion performance, Arc ensembles in anterior piriform showed enhanced stability for the rewarded odor and pattern separation for the discriminated odors as described previously. Anterior piriform adrenergic blockade prevented acquisition of similar odor discrimination and of OB ensemble changes, even with extended overtraining. Mitral and granule cell Arc ensembles in OB showed enhanced stability for rewarded odor only in the saline group. Pattern separation was not seen in the OB. Similar odor discrimination co-occurs with increased stability in rewarded odor representations and pattern separation to reduce encoding overlap. The difficulty of similar discriminations may relate to the necessity to both strengthen rewarded representations and weaken overlap across similar representations.

SIGNIFICANCE STATEMENT

We show for the first time that adrenoceptors in anterior piriform cortex (aPC) must be engaged for adult rats to learn to discriminate highly similar odors. Loss of adrenergic activation in olfactory bulb (OB) slows, but does not prevent, discrimination learning. Both increased stability of the rewarded odor representation and increased pattern separation of the rewarded and unrewarded odors in aPC accompany successful discrimination. In the OB, rewarded odors increase in ensemble stability, but there is no evidence of pattern separation. We suggest that the slow acquisition of similar odor discriminations is related to the differing plasticity requirements for increased stability and pattern separation.

Deletion of p75NTR enhances the cholinergic innervation pattern of the visual cortex

The cholinergic system is involved in cortical plasticity, attention, and learning. Within the visual cortex the cholinergic system seems to play a role in visual perception. The cholinergic neurons which project into the visual cortex are located in the basal forebrain. It has been shown that mice deficient for the low-affinity neurotrophin receptor p75NTR display increased numbers of cholinergic neurons in the basal forebrain and a denser cholinergic innervation of the hippocampus. This prompted us to analyze whether the cholinergic system is altered in adult p75NTR deficient mice. By analyzing the densities of cholinergic fibers within layer IV as well as within layer V of the visual cortex, we found that adult p75NTR deficient mice display increased cholinergic fiber densities. However, this increase was not accompanied by an increase in the density of local cholinergic neurons within the visual cortex. This indicates that the enhanced cholinergic innervation of the visual cortex is due to alteration of the cholinergic neurons located in the basal forebrain, projecting to the visual cortex. The increased cholinergic innervation of the visual cortex makes the p75NTR deficient mice an attractive model to study the necessity of the cholinergic system for the visual cortex.

Thalamic nuclei convey diverse contextual information to layer 1 of visual cortex

Sensory perception depends on the context in which a stimulus occurs. Prevailing models emphasize cortical feedback as the source of contextual modulation. However, higher order thalamic nuclei, such as the pulvinar, interconnect with many cortical and subcortical areas, suggesting a role for the thalamus in providing sensory and behavioral context. Yet the nature of the signals conveyed to cortex by higher order thalamus remains poorly understood. Here we use axonal calcium imaging to measure information provided to visual cortex by the pulvinar equivalent in mice, the lateral posterior nucleus (LP), as well as the dorsolateral geniculate nucleus (dLGN). We found that dLGN conveys retinotopically precise visual signals, while LP provides distributed information from the visual scene. Both LP and dLGN projections carry locomotion signals. However, while dLGN inputs often respond to positive combinations of running and visual flow speed, LP signals discrepancies between self-generated and external visual motion. This higher order thalamic nucleus therefore conveys diverse contextual signals that inform visual cortex about visual scene changes not predicted by the animal's own actions.

Distinct Eligibility Traces for LTP and LTD in Cortical Synapses

In reward-based learning, synaptic modifications depend on a brief stimulus and a temporally delayed reward, which poses the question of how synaptic activity patterns associate with a delayed reward. A theoretical solution to this so-called distal reward problem has been the notion of activity-generated "synaptic eligibility traces," silent and transient synaptic tags that can be converted into long-term changes in synaptic strength by reward-linked neuromodulators. Here we report the first experimental demonstration of eligibility traces in cortical synapses. We demonstrate the Hebbian induction of distinct traces for LTP and LTD and their subsequent timing-dependent transformation into lasting changes by specific monoaminergic receptors anchored to postsynaptic proteins. Notably, the temporal properties of these transient traces allow stable learning in a recurrent neural network that accurately predicts the timing of the reward, further validating the induction and transformation of eligibility traces for LTP and LTD as a plausible synaptic substrate for reward-based learning.

Modulation of Neuronal Responses by Exogenous Attention in Macaque Primary Visual Cortex

Visual perception is influenced by attention deployed voluntarily or triggered involuntarily by salient stimuli. Modulation of visual cortical processing by voluntary or endogenous attention has been extensively studied, but much less is known about how involuntary or exogenous attention affects responses of visual cortical neurons. Using implanted microelectrode arrays, we examined the effects of exogenous attention on neuronal responses in the primary visual cortex (V1) of awake monkeys. A bright annular cue was flashed either around the receptive fields of recorded neurons or in the opposite visual field to capture attention. A subsequent grating stimulus probed the cue-induced effects. In a fixation task, when the cue-to-probe stimulus onset asynchrony (SOA) was <240 ms, the cue induced a transient increase of neuronal responses to the probe at the cued location during 40-100 ms after the onset of neuronal responses to the probe. This facilitation diminished and disappeared after repeated presentations of the same cue but recurred for a new cue of a different color. In another task to detect the probe, relative shortening of monkey's reaction times for the validly cued probe depended on the SOA in a way similar to the cue-induced V1 facilitation, and the behavioral and physiological cueing effects remained after repeated practice. Flashing two cues simultaneously in the two opposite visual fields weakened or diminished both the physiological and behavioral cueing effects. Our findings indicate that exogenous attention significantly modulates V1 responses and that the modulation strength depends on both novelty and task relevance of the stimulus. Significance statement: Visual attention can be involuntarily captured by a sudden appearance of a conspicuous object, allowing rapid reactions to unexpected events of significance. The current study discovered a correlate of this effect in monkey primary visual cortex. An abrupt, salient, flash enhanced neuronal responses, and shortened the animal's reaction time, to a subsequent visual probe stimulus at the same location. However, the enhancement of the neural responses diminished after repeated exposures to this flash if the animal was not required to react to the probe. Moreover, a second, simultaneous, flash at another location weakened the neuronal and behavioral effects of the first one. These findings revealed, beyond the observations reported so far, the effects of exogenous attention in the brain.