A Retinal Circuit Generating a Dynamic Predictive Code for Oriented Features

Temporal pattern recognition in retinal ganglion cells is mediated by dynamical inhibitory synapses

A fundamental task for the brain is to generate predictions of future sensory inputs, and signal errors in these predictions. Many neurons have been shown to signal omitted stimuli during periodic stimulation, even in the retina. However, the mechanisms of this error signaling are unclear. Here we show that depressing inhibitory synapses shape the timing of the response to an omitted stimulus in the retina. While ganglion cells, the retinal output, responded to an omitted flash with a constant latency over many frequencies of the flash sequence, we found that this was not the case once inhibition was blocked. We built a simple circuit model and showed that depressing inhibitory synapses were a necessary component to reproduce our experimental findings. A new prediction of our model is that the accuracy of the constant latency requires a sufficient amount of flashes in the stimulus, which we could confirm experimentally. Depressing inhibitory synapses could thus be a key component to generate the predictive responses observed in the retina, and potentially in many brain areas.

Selective decision-making and collective behavior of fish by the motion of visual attention

Collective motion provides a spectacular example of self-organization in Nature. Visual information plays a crucial role among various types of information in determining interactions. Recently, experiments have revealed that organisms such as fish and insects selectively utilize a portion, rather than the entirety, of visual information. Here, focusing on fish, we propose an agent-based model where the direction of attention is guided by visual stimuli received from the images of nearby fish. Our model reproduces a branching phenomenon where a fish selectively follows a specific individual as the distance between two or three nearby fish increases. Furthermore, our model replicates various patterns of collective motion in a group of agents, such as vortex, polarized school, swarm, and turning. We also discuss the topological nature of the visual interaction, as well as the positional distribution of nearby fish and the map of pairwise and three-body interactions induced by them. Through a comprehensive comparison with existing experimental results, we clarify the roles of visual interactions and issues to be resolved by other forms of interactions.

Neural Circuits Underlying Multifeature Extraction in the Retina

Classic ON-OFF direction-selective ganglion cells (DSGCs) that encode the four cardinal directions were recently shown to also be orientation-selective. To clarify the mechanisms underlying orientation selectivity, we employed a variety of electrophysiological, optogenetic, and gene knock-out strategies to test the relative contributions of glutamate, GABA, and acetylcholine (ACh) input that are known to drive DSGCs, in male and female mouse retinas. Extracellular spike recordings revealed that DSGCs respond preferentially to either vertical or horizontal bars, those that are perpendicular to their preferred-null motion axes. By contrast, the glutamate input to all four DSGC types measured using whole-cell patch-clamp techniques was found to be tuned along the vertical axis. Tuned glutamatergic excitation was heavily reliant on type 5A bipolar cells, which appear to be electrically coupled via connexin 36 containing gap junctions to the vertically oriented processes of wide-field amacrine cells. Vertically tuned inputs are transformed by the GABAergic/cholinergic "starburst" amacrine cells (SACs), which are critical components of the direction-selective circuit, into distinct patterns of inhibition and excitation. Feed-forward SAC inhibition appears to "veto" preferred orientation glutamate excitation in dorsal/ventral (but not nasal/temporal) coding DSGCs "flipping" their orientation tuning by 90° and accounts for the apparent mismatch between glutamate input tuning and the DSGC's spiking response. Together, these results reveal how two distinct synaptic motifs interact to generate complex feature selectivity, shedding light on the intricate circuitry that underlies visual processing in the retina.

Distributed feature representations of natural stimuli across parallel retinal pathways

How sensory systems extract salient features from natural environments and organize them across neural pathways is unclear. Combining single-cell and population two-photon calcium imaging in mice, we discover that retinal ON bipolar cells (second-order neurons of the visual system) are divided into two blocks of four types. The two blocks distribute temporal and spatial information encoding, respectively. ON bipolar cell axons co-stratify within each block, but separate laminarly between them (upper block: diverse temporal, uniform spatial tuning; lower block: diverse spatial, uniform temporal tuning). ON bipolar cells extract temporal and spatial features similarly from artificial and naturalistic stimuli. In addition, they differ in sensitivity to coherent motion in naturalistic movies. Motion information is distributed across ON bipolar cells in the upper and the lower blocks, multiplexed with temporal and spatial contrast, independent features of natural scenes. Comparing the responses of different boutons within the same arbor, we find that axons of all ON bipolar cell types function as computational units. Thus, our results provide insights into the visual feature extraction from naturalistic stimuli and reveal how structural and functional organization cooperate to generate parallel ON pathways for temporal and spatial information in the mammalian retina.

Efficient coding of natural scenes improves neural system identification

Neural system identification aims at learning the response function of neurons to arbitrary stimuli using experimentally recorded data, but typically does not leverage normative principles such as efficient coding of natural environments. Visual systems, however, have evolved to efficiently process input from the natural environment. Here, we present a normative network regularization for system identification models by incorporating, as a regularizer, the efficient coding hypothesis, which states that neural response properties of sensory representations are strongly shaped by the need to preserve most of the stimulus information with limited resources. Using this approach, we explored if a system identification model can be improved by sharing its convolutional filters with those of an autoencoder which aims to efficiently encode natural stimuli. To this end, we built a hybrid model to predict the responses of retinal neurons to noise stimuli. This approach did not only yield a higher performance than the "stand-alone" system identification model, it also produced more biologically plausible filters, meaning that they more closely resembled neural representation in early visual systems. We found these results applied to retinal responses to different artificial stimuli and across model architectures. Moreover, our normatively regularized model performed particularly well in predicting responses of direction-of-motion sensitive retinal neurons. The benefit of natural scene statistics became marginal, however, for predicting the responses to natural movies. In summary, our results indicate that efficiently encoding environmental inputs can improve system identification models, at least for noise stimuli, and point to the benefit of probing the visual system with naturalistic stimuli.

Amacrine cells differentially balance zebrafish color circuits in the central and peripheral retina

The vertebrate inner retina is driven by photoreceptors whose outputs are already pre-processed; in zebrafish, outer retinal circuits split "color" from "grayscale" information across four cone-photoreceptor types. It remains unclear how the inner retina processes incoming spectral information while also combining cone signals to shape grayscale functions. We address this question by imaging the light-driven responses of amacrine cells (ACs) and bipolar cells (BCs) in larval zebrafish in the presence and pharmacological absence of inner retinal inhibition. We find that ACs enhance opponency in some bipolar cells while at the same time suppressing pre-existing opponency in others, so that, depending on the retinal region, the net change in the number of color-opponent units is essentially zero. To achieve this "dynamic balance," ACs counteract intrinsic color opponency of BCs via the On channel. Consistent with these observations, Off-stratifying ACs are exclusively achromatic, while all color-opponent ACs stratify in the On sublamina.

Hierarchical retinal computations rely on hybrid chemical-electrical signaling

Bipolar cells (BCs) are integral to the retinal circuits that extract diverse features from the visual environment. They bridge photoreceptors to ganglion cells, the source of retinal output. Understanding how such circuits encode visual features requires an accounting of the mechanisms that control glutamate release from bipolar cell axons. Here, we demonstrate orientation selectivity in a specific genetically identifiable type of mouse bipolar cell-type 5A (BC5A). Their synaptic terminals respond best when stimulated with vertical bars that are far larger than their dendritic fields. We provide evidence that this selectivity involves enhanced excitation for vertical stimuli that requires gap junctional coupling through connexin36. We also show that this orientation selectivity is detectable postsynaptically in direction-selective ganglion cells, which were not previously thought to be selective for orientation. Together, these results demonstrate how multiple features are extracted by a single hierarchical network, engaging distinct electrical and chemical synaptic pathways.

Diurnal changes in the efficiency of information transmission at a sensory synapse

Neuromodulators adapt sensory circuits to changes in the external world or the animal’s internal state and synapses are key control sites for such plasticity. Less clear is how neuromodulation alters the amount of information transmitted through the circuit. We investigated this question in the context of the diurnal regulation of visual processing in the retina of zebrafish, focusing on ribbon synapses of bipolar cells. We demonstrate that contrast-sensitivity peaks in the afternoon accompanied by a four-fold increase in the average Shannon information transmitted from an active zone. This increase reflects higher synaptic gain, lower spontaneous “noise” and reduced variability of evoked responses. Simultaneously, an increase in the probability of multivesicular events with larger information content increases the efficiency of transmission (bits per vesicle) by factors of 1.5-2.7. This study demonstrates the multiplicity of mechanisms by which a neuromodulator can adjust the synaptic transfer of sensory information.

Neuromodulators can adjust how sensory signals are processed. In this study, the authors demonstrate how time of day affects the way information is transmitted in the zebrafish retina.

Feature Detection by Retinal Ganglion Cells

Retinal circuits transform the pixel representation of photoreceptors into the feature representations of ganglion cells, whose axons transmit these representations to the brain. Functional, morphological, and transcriptomic surveys have identified more than 40 retinal ganglion cell (RGC) types in mice. RGCs extract features of varying complexity; some simply signal local differences in brightness (i.e., luminance contrast), whereas others detect specific motion trajectories. To understand the retina, we need to know how retinal circuits give rise to the diverse RGC feature representations. A catalog of the RGC feature set, in turn, is fundamental to understanding visual processing in the brain. Anterograde tracing indicates that RGCs innervate more than 50 areas in the mouse brain. Current maps connecting RGC types to brain areas are rudimentary, as is our understanding of how retinal signals are transformed downstream to guide behavior. In this article, I review the feature selectivities of mouse RGCs, how they arise, and how they are utilized downstream. Not only is knowledge of the behavioral purpose of RGC signals critical for understanding the retinal contributions to vision; it can also guide us to the most relevant areas of visual feature space. Expected final online publication date for the Annual Review of Vision Science, Volume 8 is September 2022. Please see http://www.annualreviews.org/page/journal/pubdates for revised estimates.

Opposite forms of adaptation in mouse visual cortex are controlled by distinct inhibitory microcircuits

Sensory processing in the cortex adapts to the history of stimulation but the mechanisms are not understood. Imaging the primary visual cortex of mice we find here that an increase in stimulus contrast is not followed by a simple decrease in gain of pyramidal cells; as many cells increase gain to improve detection of a subsequent decrease in contrast. Depressing and sensitizing forms of adaptation also occur in different types of interneurons (PV, SST and VIP) and the net effect within individual pyramidal cells reflects the balance of PV inputs, driving depression, and a subset of SST interneurons driving sensitization. Changes in internal state associated with locomotion increase gain across the population of pyramidal cells while maintaining the balance between these opposite forms of plasticity, consistent with activation of both VIP->SST and SST->PV disinhibitory pathways. These results reveal how different inhibitory microcircuits adjust the gain of pyramidal cells signalling changes in stimulus strength.

The authors describe the role of inhibitory microcircuits in the visual cortex of mice in adaptation to contrast. They show how external stimuli and internal state interact to adjust processing in the visual cortex.

Direction selectivity in retinal bipolar cell axon terminals

The ability to encode the direction of image motion is fundamental to our sense of vision. Direction selectivity along the four cardinal directions is thought to originate in direction-selective ganglion cells (DSGCs) because of directionally tuned GABAergic suppression by starburst cells. Here, by utilizing two-photon glutamate imaging to measure synaptic release, we reveal that direction selectivity along all four directions arises earlier than expected at bipolar cell outputs. Individual bipolar cells contained four distinct populations of axon terminal boutons with different preferred directions. We further show that this bouton-specific tuning relies on cholinergic excitation from starburst cells and GABAergic inhibition from wide-field amacrine cells. DSGCs received both tuned directionally aligned inputs and untuned inputs from among heterogeneously tuned glutamatergic bouton populations. Thus, directional tuning in the excitatory visual pathway is incrementally refined at the bipolar cell axon terminals and their recipient DSGC dendrites by two different neurotransmitters co-released from starburst cells.

•

Cardinal direction selectivity emerges at types 7 and 2 bipolar cell axon terminals

•

Starburst amacrine cells are necessary for direction selectivity in bipolar cells

•

Cholinergic excitation and GABAergic inhibition are integrated at axon terminals

•

Direction-selective ganglion cells receive directionally aligned glutamate inputs

Adjudicating Between Local and Global Architectures of Predictive Processing in the Subcortical Auditory Pathway

Predictive processing, a leading theoretical framework for sensory processing, suggests that the brain constantly generates predictions on the sensory world and that perception emerges from the comparison between these predictions and the actual sensory input. This requires two distinct neural elements: generative units, which encode the model of the sensory world; and prediction error units, which compare these predictions against the sensory input. Although predictive processing is generally portrayed as a theory of cerebral cortex function, animal and human studies over the last decade have robustly shown the ubiquitous presence of prediction error responses in several nuclei of the auditory, somatosensory, and visual subcortical pathways. In the auditory modality, prediction error is typically elicited using so-called oddball paradigms, where sequences of repeated pure tones with the same pitch are at unpredictable intervals substituted by a tone of deviant frequency. Repeated sounds become predictable promptly and elicit decreasing prediction error; deviant tones break these predictions and elicit large prediction errors. The simplicity of the rules inducing predictability make oddball paradigms agnostic about the origin of the predictions. Here, we introduce two possible models of the organizational topology of the predictive processing auditory network: (1) the global view, that assumes that predictions on the sensory input are generated at high-order levels of the cerebral cortex and transmitted in a cascade of generative models to the subcortical sensory pathways; and (2) the local view, that assumes that independent local models, computed using local information, are used to perform predictions at each processing stage. In the global view information encoding is optimized globally but biases sensory representations along the entire brain according to the subjective views of the observer. The local view results in a diminished coding efficiency, but guarantees in return a robust encoding of the features of sensory input at each processing stage. Although most experimental results to-date are ambiguous in this respect, recent evidence favors the global model.

High Contrast Allows the Retina to Compute More Than Just Contrast

The goal of sensory processing is to represent the environment of an animal. All sensory systems share a similar constraint: they need to encode a wide range of stimulus magnitudes within their narrow neuronal response range. The most efficient way, exploited by even the simplest nervous systems, is to encode relative changes in stimulus magnitude rather than the absolute magnitudes. For instance, the retina encodes contrast, which are the variations of light intensity occurring in time and in space. From this perspective, it is easy to understand why the bright plumage of a moving bird gains a lot of attention, while an octopus remains motionless and mimics its surroundings for concealment. Stronger contrasts simply cause stronger visual signals. However, the gains in retinal performance associated with higher contrast are far more than what can be attributed to just a trivial linear increase in signal strength. Here we discuss how this improvement in performance is reflected throughout different parts of the neural circuitry, within its neural code and how high contrast activates many non-linear mechanisms to unlock several sophisticated retinal computations that are virtually impossible in low contrast conditions.

On the potential role of lateral connectivity in retinal anticipation

We analyse the potential effects of lateral connectivity (amacrine cells and gap junctions) on motion anticipation in the retina. Our main result is that lateral connectivity can-under conditions analysed in the paper-trigger a wave of activity enhancing the anticipation mechanism provided by local gain control (Berry et al. in Nature 398(6725):334-338, 1999; Chen et al. in J. Neurosci. 33(1):120-132, 2013). We illustrate these predictions by two examples studied in the experimental literature: differential motion sensitive cells (Baccus and Meister in Neuron 36(5):909-919, 2002) and direction sensitive cells where direction sensitivity is inherited from asymmetry in gap junctions connectivity (Trenholm et al. in Nat. Neurosci. 16:154-156, 2013). We finally present reconstructions of retinal responses to 2D visual inputs to assess the ability of our model to anticipate motion in the case of three different 2D stimuli.

Synapse elimination activates a coordinated homeostatic presynaptic response in an autaptic circuit

The number of synapses present in a neuronal circuit is not fixed. Neurons must compensate for changes in connectivity caused by synaptic pruning, learning processes or pathological conditions through the constant adjustment of the baseline level of neurotransmission. Here, we show that cholinergic neurons grown in an autaptic circuit in the absence of glia sense the loss of half of their synaptic contacts triggered by exposure to peptide p4.2, a C-terminal fragment of SPARC. Synaptic elimination is driven by a reorganization of the periodic F-actin cytoskeleton present along neurites, and occurs without altering the density of postsynaptic receptors. Neurons recover baseline neurotransmission through a homeostatic presynaptic response that consists of the coordinated activation of rapid synapse formation and an overall potentiation of presynaptic calcium influx. These results demonstrate that neurons establishing autaptic connections continuously sense and adjust their synaptic output by tweaking the number of functional contacts and neurotransmitter release probability.

Cecilia Velasco and Artur Llobet study how autapses respond to synapse elimination. They employ microisland cultures free of glial cells, treat with a SPARC-derived peptide and show that neurons forming autaptic circuits continuously sense and regulate the number of contacts and neurotransmitter release.

Motor Behavior Selectively Inhibits Hair Cells Activated by Forward Motion in the Lateral Line of Zebrafish

How do sensory systems disambiguate events in the external world from signals generated by the animal's own motor actions? One strategy is to use an "efference copy" of the motor command to inhibit the sensory input caused by active behavior [1]. But does inhibition of self-generated inputs also block transmission of external stimuli? We investigated this question in the lateral line, a sensory system that allows fish and amphibians to detect water currents and that contributes to behaviors such as rheotaxis [2] and predator avoidance [3, 4]. This mechanical sense begins in hair cells grouped into neuromasts dotted along the animal's body [5]. Each neuromast contains two populations of hair cells, activated by deflection in either the anterior or posterior direction [6], as well as efferent fibers that are active during motor behavior to suppress afferents projecting to the brain [7-12]. To test how far the efference copy signal modulates responses to external stimuli, we imaged neural and synaptic activity in larval zebrafish during fictive swimming. We find that efferents transmit a precise copy of the motor signal and a single spike in the motor nerve can be associated with ∼50% inhibition of glutamate release. The efference copy signal acted with high selectivity on hair cells polarized to be activated by posterior deflections, as would occur during forward motion. During swimming, therefore, "push-pull" encoding of stimulus direction by afferents of opposite polarity is disrupted while still allowing a subset of hair cells to detect stimuli originating in the external world.

Understanding the retinal basis of vision across species

The vertebrate retina first evolved some 500 million years ago in ancestral marine chordates. Since then, the eyes of different species have been tuned to best support their unique visuoecological lifestyles. Visual specializations in eye designs, large-scale inhomogeneities across the retinal surface and local circuit motifs mean that all species' retinas are unique. Computational theories, such as the efficient coding hypothesis, have come a long way towards an explanation of the basic features of retinal organization and function; however, they cannot explain the full extent of retinal diversity within and across species. To build a truly general understanding of vertebrate vision and the retina's computational purpose, it is therefore important to more quantitatively relate different species' retinal functions to their specific natural environments and behavioural requirements. Ultimately, the goal of such efforts should be to build up to a more general theory of vision.

An arbitrary-spectrum spatial visual stimulator for vision research

Visual neuroscientists require accurate control of visual stimulation. However, few stimulator solutions simultaneously offer high spatio-temporal resolution and free control over the spectra of the light sources, because they rely on off-the-shelf technology developed for human trichromatic vision. Importantly, consumer displays fail to drive UV-shifted short wavelength-sensitive photoreceptors, which strongly contribute to visual behaviour in many animals, including mice, zebrafish and fruit flies. Moreover, many non-mammalian species feature more than three spectral photoreceptor types. Here, we present a flexible, spatial visual stimulator with up to six arbitrary spectrum chromatic channels. It combines a standard digital light processing engine with open source hard- and software that can be easily adapted to the experimentalist's needs. We demonstrate the capability of this general visual stimulator experimentally in the in vitro mouse retinal whole-mount and the in vivo zebrafish. With this work, we intend to start a community effort of sharing and developing a common stimulator design for vision research.