Mapping model units to visual neurons reveals population code for social behaviour.

The rich variety of behaviours observed in animals arises through the interplay between sensory processing and motor control. To understand these sensorimotor transformations, it is useful to build models that predict not only neural responses to sensory input1-5 but also how each neuron causally contributes to behaviour6,7. Here we demonstrate a novel modelling approach to identify a one-to-one mapping between internal units in a deep neural network and real neurons by predicting the behavioural changes that arise from systematic perturbations of more than a dozen neuronal cell types. A key ingredient that we introduce is 'knockout training', which involves perturbing the network during training to match the perturbations of the real neurons during behavioural experiments. We apply this approach to model the sensorimotor transformations of Drosophila melanogaster males during a complex, visually guided social behaviour8-11. The visual projection neurons at the interface between the optic lobe and central brain form a set of discrete channels12, and prior work indicates that each channel encodes a specific visual feature to drive a particular behaviour13,14. Our model reaches a different conclusion: combinations of visual projection neurons, including those involved in non-social behaviours, drive male interactions with the female, forming a rich population code for behaviour. Overall, our framework consolidates behavioural effects elicited from various neural perturbations into a single, unified model, providing a map from stimulus to neuronal cell type to behaviour, and enabling future incorporation of wiring diagrams of the brain15 into the model.

Visual and motor signatures of locomotion dynamically shape a population code for feature detection in Drosophila.

Natural vision is dynamic: as an animal moves, its visual input changes dramatically. How can the visual system reliably extract local features from an input dominated by self-generated signals? In Drosophila, diverse local visual features are represented by a group of projection neurons with distinct tuning properties. Here, we describe a connectome-based volumetric imaging strategy to measure visually evoked neural activity across this population. We show that local visual features are jointly represented across the population, and a shared gain factor improves trial-to-trial coding fidelity. A subset of these neurons, tuned to small objects, is modulated by two independent signals associated with self-movement, a motor-related signal, and a visual motion signal associated with rotation of the animal. These two inputs adjust the sensitivity of these feature detectors across the locomotor cycle, selectively reducing their gain during saccades and restoring it during intersaccadic intervals. This work reveals a strategy for reliable feature detection during locomotion.

Adaptation in cone photoreceptors contributes to an unexpected insensitivity of primate On parasol retinal ganglion cells to spatial structure in natural images.

Neural circuits are constructed from nonlinear building blocks, and not surprisingly overall circuit behavior is often strongly nonlinear. But neural circuits can also behave near linearly, and some circuits shift from linear to nonlinear behavior depending on stimulus conditions. Such control of nonlinear circuit behavior is fundamental to neural computation. Here, we study a surprising stimulus dependence of the responses of macaque On (but not Off) parasol retinal ganglion cells: these cells respond nonlinearly to spatial structure in some stimuli but near linearly to spatial structure in others, including natural inputs. We show that these differences in the linearity of the integration of spatial inputs can be explained by a shift in the balance of excitatory and inhibitory synaptic inputs that originates at least partially from adaptation in the cone photoreceptors. More generally, this highlights how subtle asymmetries in signaling - here in the cone signals - can qualitatively alter circuit computation.

Neuroscience: Convergence of biological and artificial networks.

A new study shows that an artificial neural network trained to predict visual motion reproduces key properties of motion detecting circuits in the fruit fly.

The connectome predicts resting-state functional connectivity across the Drosophila brain.

Anatomical connectivity can constrain both a neural circuit's function and its underlying computation. This principle has been demonstrated for many small, defined neural circuits. For example, connectome reconstructions have informed models for direction selectivity in the vertebrate retina1,2 as well as the Drosophila visual system.3 In these cases, the circuit in question is relatively compact, well-defined, and has known functions. However, how the connectome constrains global properties of large-scale networks, across multiple brain regions or the entire brain, is incompletely understood. As the availability of partial or complete connectomes expands to more systems and species4-8 it becomes critical to understand how this detailed anatomical information can inform our understanding of large-scale circuit function.9,10 Here, we use data from the Drosophila connectome4 in conjunction with whole-brain in vivo imaging11 to relate structural and functional connectivity in the central brain. We find a strong relationship between resting-state functional correlations and direct region-to-region structural connectivity. We find that the relationship between structure and function varies across the brain, with some regions displaying a tight correspondence between structural and functional connectivity whereas others, including the mushroom body, are more strongly dependent on indirect connections. Throughout this work, we observe features of structural and functional networks in Drosophila that are strikingly similar to those seen in mammalian cortex, including in the human brain. Given the vast anatomical and functional differences between Drosophila and mammalian nervous systems, these observations suggest general principles that govern brain structure, function, and the relationship between the two.

Stimulus- and goal-oriented frameworks for understanding natural vision.

Our knowledge of sensory processing has advanced dramatically in the last few decades, but this understanding remains far from complete, especially for stimuli with the large dynamic range and strong temporal and spatial correlations characteristic of natural visual inputs. Here we describe some of the issues that make understanding the encoding of natural images a challenge. We highlight two broad strategies for approaching this problem: a stimulus-oriented framework and a goal-oriented one. Different contexts can call for one framework or the other. Looking forward, recent advances, particularly those based in machine learning, show promise in borrowing key strengths of both frameworks and by doing so illuminating a path to a more comprehensive understanding of the encoding of natural stimuli.

GABA release selectively regulates synapse development at distinct inputs on direction-selective retinal ganglion cells.

Synaptic inhibition controls a neuron's output via functionally distinct inputs at two subcellular compartments, the cell body and the dendrites. It is unclear whether the assembly of these distinct inhibitory inputs can be regulated independently by neurotransmission. In the mammalian retina, γ-aminobutyric acid (GABA) release from starburst amacrine cells (SACs) onto the dendrites of on-off direction-selective ganglion cells (ooDSGCs) is essential for directionally selective responses. We found that ooDSGCs also receive GABAergic input on their somata from other amacrine cells (ACs), including ACs containing the vasoactive intestinal peptide (VIP). When net GABAergic transmission is reduced, somatic, but not dendritic, GABAA receptor clusters on the ooDSGC increased in number and size. Correlative fluorescence imaging and serial electron microscopy revealed that these enlarged somatic receptor clusters are localized to synapses. By contrast, selectively blocking vesicular GABA release from either SACs or VIP ACs did not alter dendritic or somatic receptor distributions on the ooDSGCs, showing that neither SAC nor VIP AC GABA release alone is required for the development of inhibitory synapses in ooDSGCs. Furthermore, a reduction in net GABAergic transmission, but not a selective reduction from SACs, increased excitatory drive onto ooDSGCs. This increased excitation may drive a homeostatic increase in ooDSGC somatic GABAA receptors. Differential regulation of GABAA receptors on the ooDSGC's soma and dendrites could facilitate homeostatic control of the ooDSGC's output while enabling the assembly of the GABAergic connectivity underlying direction selectivity to be indifferent to altered transmission.

Receptive field center-surround interactions mediate context-dependent spatial contrast encoding in the retina.

Antagonistic receptive field surrounds are a near-universal property of early sensory processing. A key assumption in many models for retinal ganglion cell encoding is that receptive field surrounds are added only to the fully formed center signal. But anatomical and functional observations indicate that surrounds are added before the summation of signals across receptive field subunits that creates the center. Here, we show that this receptive field architecture has an important consequence for spatial contrast encoding in the macaque monkey retina: the surround can control sensitivity to fine spatial structure by changing the way the center integrates visual information over space. The impact of the surround is particularly prominent when center and surround signals are correlated, as they are in natural stimuli. This effect of the surround differs substantially from classic center-surround models and raises the possibility that the surround plays unappreciated roles in shaping ganglion cell sensitivity to natural inputs.

Synaptic Rectification Controls Nonlinear Spatial Integration of Natural Visual Inputs.

A central goal in the study of any sensory system is to predict neural responses to complex inputs, especially those encountered during natural stimulation. Nowhere is the transformation from stimulus to response better understood than the vertebrate retina. Nevertheless, descriptions of retinal computation are largely based on stimulation using artificial visual stimuli, and it is unclear how these descriptions map onto the encoding of natural stimuli. We demonstrate that nonlinear spatial integration, a common feature of retinal ganglion cell (RGC) processing, shapes neural responses to natural visual stimuli in primate Off parasol RGCs, whereas On parasol RGCs exhibit surprisingly linear spatial integration. Despite this asymmetry, both cell types show strong nonlinear integration when presented with artificial stimuli. We show that nonlinear integration of natural stimuli is a consequence of rectified excitatory synaptic input and that accounting for nonlinear spatial integration substantially improves models that predict RGC responses to natural images.

Direction-Selective Circuits Shape Noise to Ensure a Precise Population Code.

Neural responses are noisy, and circuit structure can correlate this noise across neurons. Theoretical studies show that noise correlations can have diverse effects on population coding, but these studies rarely explore stimulus dependence of noise correlations. Here, we show that noise correlations in responses of ON-OFF direction-selective retinal ganglion cells are strongly stimulus dependent, and we uncover the circuit mechanisms producing this stimulus dependence. A population model based on these mechanistic studies shows that stimulus-dependent noise correlations improve the encoding of motion direction 2-fold compared to independent noise. This work demonstrates a mechanism by which a neural circuit effectively shapes its signal and noise in concert, minimizing corruption of signal by noise. Finally, we generalize our findings beyond direction coding in the retina and show that stimulus-dependent correlations will generally enhance information coding in populations of diversely tuned neurons.

Type I intrinsically photosensitive retinal ganglion cells of early post-natal development correspond to the M4 subtype.

BACKGROUND: Intrinsically photosensitive retinal ganglion cells (ipRGCs) mediate circadian light entrainment and the pupillary light response in adult mice. In early development these cells mediate different processes, including negative phototaxis and the timing of retinal vascular development. To determine if ipRGC physiologic properties also change with development, we measured ipRGC cell density and light responses in wild-type mouse retinas at post-natal days 8, 15 and 30. RESULTS: Melanopsin-positive cell density decreases by 17% between post-natal days 8 and 15 and by 25% between days 8 and 30. This decrease is due specifically to a decrease in cells co-labeled with a SMI-32, a marker for alpha-on ganglion cells (corresponding to adult morphologic type M4 ipRGCs). On multi-electrode array recordings, post-natal day 8 (P8) ipRGC light responses show more robust firing, reduced adaptation and more rapid recovery from short and extended light pulses than do the light responses of P15 and P30 ipRGCs. Three ipRGC subtypes - Types I-III - have been defined in early development based on sensitivity and latency on multielectrode array recordings. We find that Type I cells largely account for the unique physiologic properties of P8 ipRGCs. Type I cells have previously been shown to have relatively short latencies and high sensitivity. We now show that Type I cells show have rapid and robust recovery from long and short bright light exposures compared with Type II and III cells, suggesting differential light adaptation mechanisms between cell types. By P15, Type I ipRGCs are no longer detectable. Loose patch recordings of P8 M4 ipRGCs demonstrate Type I physiology. CONCLUSIONS: Type I ipRGCs are found only in early development. In addition to their previously described high sensitivity and rapid kinetics, these cells are uniquely resistant to adaptation and recover quickly and fully to short and prolonged light exposure. Type I ipRGCs correspond to the SMI-32 positive, M4 subtype and largely lose melanopsin expression in development. These cells constitute a unique morphologic and physiologic class of ipRGCs functioning early in postnatal development.

Nonlinear dendritic integration of electrical and chemical synaptic inputs drives fine-scale correlations.

Throughout the CNS, gap junction-mediated electrical signals synchronize neural activity on millisecond timescales via cooperative interactions with chemical synapses. However, gap junction-mediated synchrony has rarely been studied in the context of varying spatiotemporal patterns of electrical and chemical synaptic activity. Thus, the mechanism underlying fine-scale synchrony and its relationship to neural coding remain unclear. We examined spike synchrony in pairs of genetically identified, electrically coupled ganglion cells in mouse retina. We found that coincident electrical and chemical synaptic inputs, but not electrical inputs alone, elicited synchronized dendritic spikes in subregions of coupled dendritic trees. The resulting nonlinear integration produced fine-scale synchrony in the cells' spike output, specifically for light stimuli driving input to the regions of dendritic overlap. In addition, the strength of synchrony varied inversely with spike rate. Together, these features may allow synchronized activity to encode information about the spatial distribution of light that is ambiguous on the basis of spike rate alone.

Visual space is represented by nonmatching topographies of distinct mouse retinal ganglion cell types.

The distributions of neurons in sensory circuits display ordered spatial patterns arranged to enhance or encode specific regions or features of the external environment. Indeed, visual space is not sampled uniformly across the vertebrate retina. Retinal ganglion cell (RGC) density increases and dendritic arbor size decreases toward retinal locations with higher sampling frequency, such as the fovea in primates and area centralis in carnivores [1]. In these locations, higher acuity at the level of individual cells is obtained because the receptive field center of a RGC corresponds approximately to the spatial extent of its dendritic arbor [2, 3]. For most species, structurally and functionally distinct RGC types appear to have similar topographies, collectively scaling their cell densities and arbor sizes toward the same retinal location [4]. Thus, visual space is represented across the retina in parallel by multiple distinct circuits [5]. In contrast, we find a population of mouse RGCs, known as alpha or alpha-like [6], that displays a nasal-to-temporal gradient in cell density, size, and receptive fields, which facilitates enhanced visual sampling in frontal visual fields. The distribution of alpha-like RGCs contrasts with other known mouse RGC types and suggests that, unlike most mammals, RGC topographies in mice are arranged to sample space differentially.

A method for detecting molecular transport within the cerebral ventricles of live zebrafish (Danio rerio) larvae.

The production and flow of cerebrospinal fluid performs an important role in the development and homeostasis of the central nervous system.However, these processes are difficult to study in the mammalian brain because the ventricles are situated deep within the parenchyma.In this communication we introduce the zebrafish larva as an in vivo model for studying cerebral ventricle and blood­brain barrier function. Using confocal microscopy we show that zebrafish ventricles are topologically similar to those of the mammalian brain.We describe a new method for measuring the dynamics of molecular transport within the ventricles of live zebrafish by means of the uncaging of a fluorescein derivative. Furthermore, we determine that in 5­6 days post-fertilization zebrafish, the dispersal of molecules in the ventricles is driven by a combination of ciliary motion and diffusion. The zebrafish presents a tractable system with the advantage of genetics, size and transparency for exploring ventricular physiology and for mounting large-scale high throughput experiments.