The foveal retina is a unique primate specialization and a promising target for the first connectome of a human central nervous system structure. In the fovea, neural cells and circuits are miniaturized and compressed to sample the visual image at high spatial resolution, initiating form, color, and motion perception. Here, we report a working-draft human foveal connectome. We used deep learning to segment all cells and their synaptic connections in a piece of human fovea. We classified ~3,000 cells into 51 neuronal types and three glial types based on morphology and connectivity. Some cell types and synaptic motifs-including the midget and small bistratified ganglion cell circuits-are conserved between human and nonhuman primates. However, most other cell types and related synaptic connectivity remain unexplored. We find that in comparison to peripheral retina, the human foveal retina shows reduced cell type diversity with markedly unequal densities and distinct patterns of synaptic connectivity. For example, foveal H1 horizontal cells are present at high density relative to H2 cells and receive unexpected synaptic input from S cones, predicting fovea-specific chromatic properties of downstream receptive fields. Retinal ganglion cells could be divided into only 13 types-a contrast to the ~20 types in peripheral retina and the 40+ types recognized in mouse retina-but consistent with recent transcriptomic studies of human fovea. These findings support the hypothesis that the human fovea has traded a repertoire of feature detectors for high-precision, saccade-supported active vision centered around a small number of high-density visual pathways.

The fovea is a unique specialization of the primate retina and is a promising site for obtaining the first complete connectome of a human central nervous system (CNS) structure. Within the fovea, neural cells and circuits have been miniaturized and compressed during evolution to sample the visual image at highest spatial resolution and begin the neural processing that serves human form, color, and motion perception. Here we present a comprehensive analysis of a sample of human foveal retina using deep learning-based segmentation to reconstruct all cells and synaptic connections at nanoscale resolution. We classified ~3,000 cells into 51 distinct morphological types based on their structural features and connectivity patterns. Our observations reveal novel synaptic pathways absent in non-human primates, suggesting specialized circuits contribute uniquely to human trichromatic color vision. A biophysical model of the distinct connectomes made by gap junctions (electrical synapses) between short- (S) and medium-long- (ML) wavelength-sensitive cone photoreceptors, suggests chromatic interactions between S and ML cones prior to the first chemical synapse. Segmentation of retinal ganglion cells (RGCs) suggests the presence of only 11 visual pathways, with 5 high-density RGC pathways accounting for over 95% of foveal output to the brain: a dramatic contrast to the 40+ ganglion cell types recognized in mouse retina. Our connectomic analysis reveals distinctive features of human neural circuitry and demonstrates how AI-based computational approaches can advance understanding of human brain structure and function.

With over 10[14][1] synapses, the human brain presents a seemingly insurmountable challenge to a nano-scale circuit-level understanding of its diverse neural systems. The foveal retina however presents a feasible site for a complete connectome of a key human CNS structure. Foveal cells and circuits are miniaturized and compressed to densely sample the visual image at highest resolution to initiate form, color and motion perception. Here we use computational methods first applied to the fly brain to provide a draft connectome of all neurons in a foveal volume. We found synaptic connections distinct to humans linking short-wavelength sensitive cones to color vision pathways. Moreover, by reconstructing excitatory synaptic pathways arising from cone photoreceptors we found that 96% of foveal ganglion cells contribute to only three major pathways to the brain. This new resource reveals unique features of a human neural system and opens a door to its complete connectome. In Brief Deep-learning based reconstruction of cells and synapses in a human fovea reveals visual pathway origins and human-specific wiring for color vision, opening the door to a complete synaptic-level connectome of this critical locus in the human central nervous system. Highlights ### Competing Interest Statement K.L., N.K., D.I., T.N., R.L., S.P., A.H., J.A.B., J.S. and T.M. declare financial interests in Zetta AI. S.G. is owner of Aware LLC and developed the NeuroMaps.app. The remaining authors declare no competing interests. [1]: #ref-14

In a recent study, visual signals were recorded for the first time in starburst amacrine cells of the macaque retina, and, as for mouse and rabbit, a directional bias observed in calcium signals was recorded from near the dendritic tips. Stimulus motion from the soma toward the tip generated a larger calcium signal than motion from the tip toward the soma. Two mechanisms affecting the spatiotemporal summation of excitatory postsynaptic currents have been proposed to contribute to directional signaling at the dendritic tips of starbursts: (1) a "morphological" mechanism in which electrotonic propagation of excitatory synaptic currents along a dendrite sums bipolar cell inputs at the dendritic tip preferentially for stimulus motion in the centrifugal direction; (2) a "space-time" mechanism that relies on differences in the time-courses of proximal and distal bipolar cell inputs to favor centrifugal stimulus motion. To explore the contributions of these two mechanisms in the primate, we developed a realistic computational model based on connectomic reconstruction of a macaque starburst cell and the distribution of its synaptic inputs from sustained and transient bipolar cell types. Our model suggests that both mechanisms can initiate direction selectivity in starburst dendrites, but their contributions differ depending on the spatiotemporal properties of the stimulus. Specifically, the morphological mechanism dominates when small visual objects are moving at high velocities, and the space-time mechanism contributes most for large visual objects moving at low velocities.

In the retina, ON starburst amacrine cells (SACs) play a crucial role in the direction-selective circuit, but the sources of inhibition that shape their response properties remain unclear. Previous studies demonstrate that ∼95% of their inhibitory synapses are GABAergic, yet we find that the light-evoked inhibitory currents measured in SACs are predominantly glycinergic. Glycinergic inhibition is extremely slow, relying on non-canonical glycine receptors containing α4 subunits, and is driven by both the ON and OFF retinal pathways. These attributes enable glycine inputs to summate and effectively control the output gain of SACs, expanding the range over which they compute direction. Serial electron microscopic reconstructions reveal three specific types of ON and OFF narrow-field amacrine cells as the presumptive sources of glycinergic inhibition. Together, these results establish an unexpected role for specific glycinergic amacrine cells in the retinal computation of stimulus direction by SACs.

Experience-dependent modulation of neuronal responses is a key attribute in sensory processing. In the mammalian retina, the On-Off direction-selective ganglion cell (DSGC) is well known for its robust direction selectivity. However, how the On-Off DSGC light responsiveness dynamically adjusts to the changing visual environment is underexplored. Here, we report that On-Off DSGCs tuned to posterior motion direction [i.e. posterior DSGCs (pDSGCs)] in mice of both sexes can be transiently sensitized by prior stimuli. Notably, distinct sensitization patterns are found in dorsal and ventral pDSGCs. Although responses of both dorsal and ventral pDSGCs to dark stimuli (Off responses) are sensitized, only dorsal cells show the sensitization of responses to bright stimuli (On responses). Visual stimulation to the dorsal retina potentiates a sustained excitatory input from Off bipolar cells, leading to tonic depolarization of pDSGCs. Such tonic depolarization propagates from the Off to the On dendritic arbor of the pDSGC to sensitize its On response. We also identified a previously overlooked feature of DSGC dendritic architecture that can support dendritic integration between On and Off dendritic layers bypassing the soma. By contrast, ventral pDSGCs lack a sensitized tonic depolarization and thus do not exhibit sensitization of their On responses. Our results highlight a topographic difference in Off bipolar cell inputs underlying divergent sensitization patterns of dorsal and ventral pDSGCs. Moreover, substantial crossovers between dendritic layers of On-Off DSGCs suggest an interactive dendritic algorithm for processing On and Off signals before they reach the soma. SIGNIFICANCE STATEMENT Visual neuronal responses are dynamically influenced by the prior visual experience. This form of plasticity reflects the efficient coding of the naturalistic environment by the visual system. We found that a class of retinal output neurons, On-Off direction-selective ganglion cells, transiently increase their responsiveness after visual stimulation. Cells located in dorsal and ventral retinas exhibit distinct sensitization patterns because of different adaptive properties of Off bipolar cell signaling. A previously overlooked dendritic morphologic feature of the On-Off direction-selective ganglion cell is implicated in the cross talk between On and Off pathways during sensitization. Together, these findings uncover a topographic difference in the adaptive encoding of upper and lower visual fields and the underlying neural mechanism in the dorsal and ventral retinas.

From mouse to primate, there is a striking discontinuity in our current understanding of the neural coding of motion direction. In non-primate mammals, directionally selective cell types and circuits are a signature feature of the retina, situated at the earliest stage of the visual process. In primates, by contrast, direction selectivity is a hallmark of motion processing areas in visual cortex, but has not been found in the retina, despite significant effort. Here we combined functional recordings of light-evoked responses and connectomic reconstruction to identify diverse direction-selective cell types in the macaque monkey retina with distinctive physiological properties and synaptic motifs. This circuitry includes an ON-OFF ganglion cell type, a spiking, ON-OFF polyaxonal amacrine cell and the starburst amacrine cell, all of which show direction selectivity. Moreover, we discovered that macaque starburst cells possess a strong, non-GABAergic, antagonistic surround mediated by input from excitatory bipolar cells that is critical for the generation of radial motion sensitivity in these cells. Our findings open a door to investigation of a precortical circuitry that computes motion direction in the primate visual system.

Summary: Retinal bipolar cells are second-order neurons that transmit basic features of the visual scene to postsynaptic partners. However, their contribution to motion detection has not been fully appreciated. Here, we demonstrate that cholinergic feedback from starburst amacrine cells (SACs) to certain presynaptic bipolar cells via alpha-7 nicotinic acetylcholine receptors (α7-nAChRs) promotes direction-selective signaling. Patch clamp recordings reveal that distinct bipolar cell types making synapses at proximal SAC dendrites also express α7-nAChRs, producing directionally skewed excitatory inputs. Asymmetric SAC excitation contributes to motion detection in On-Off direction-selective ganglion cells (On-Off DSGCs), predicted by computational modeling of SAC dendrites and supported by patch clamp recordings from On-Off DSGCs when bipolar cell α7-nAChRs is eliminated pharmacologically or by conditional knockout. Altogether, these results show that cholinergic feedback to bipolar cells enhances direction-selective signaling in postsynaptic SACs and DSGCs, illustrating how bipolar cells provide a scaffold for postsynaptic microcircuits to cooperatively enhance retinal motion detection.

In the outer plexiform layer (OPL) of the mammalian retina, cone photoreceptors (cones) provide input to more than a dozen types of cone bipolar cells (CBCs). In the mouse, this transmission is modulated by a single horizontal cell (HC) type. HCs perform global signaling within their laterally coupled network but also provide local, cone-specific feedback. However, it is unknown how HCs provide local feedback to cones at the same time as global forward signaling to CBCs and where the underlying synapses are located. To assess how HCs simultaneously perform different modes of signaling, we reconstructed the dendritic trees of five HCs as well as cone axon terminals and CBC dendrites in a serial block-face electron microscopy volume and analyzed their connectivity. In addition to the fine HC dendritic tips invaginating cone axon terminals, we also identified “bulbs,” short segments of increased dendritic diameter on the primary dendrites of HCs. These bulbs are in an OPL stratum well below the cone axon terminal base and make contacts with other HCs and CBCs. Our results from immunolabeling, electron microscopy, and glutamate imaging suggest that HC bulbs represent GABAergic synapses that do not receive any direct photoreceptor input. Together, our data suggest the existence of two synaptic strata in the mouse OPL, spatially separating cone-specific feedback and feedforward signaling to CBCs. A biophysical model of a HC dendritic branch and voltage imaging support the hypothesis that this spatial arrangement of synaptic contacts allows for simultaneous local feedback and global feedforward signaling by HCs.

Previously, we found that in the mammalian retina, inhibitory inputs onto starburst amacrine cells (SACs) are required for robust direction selectivity of On-Off direction-selective ganglion cells (On-Off DSGCs) against noisy backgrounds (Chen et al., 2016). However, the source of the inhibitory inputs to SACs and how this inhibition confers noise resilience of DSGCs are unknown. Here, we show that when visual noise is present in the background, the motion-evoked inhibition to an On-Off DSGC is preserved by a disinhibitory motif consisting of a serially connected network of neighboring SACs presynaptic to the DSGC. This preservation of inhibition by a disinhibitory motif arises from the interaction between visually evoked network dynamics and short-term synaptic plasticity at the SAC-DSGC synapse. Although the disinhibitory microcircuit is well studied for its disinhibitory function in brain circuits, our results highlight the algorithmic flexibility of this motif beyond disinhibition due to the mutual influence between network and synaptic plasticity mechanisms.

While multicompartment models have long been used to study the biophysics of neurons, it is still challenging to infer the parameters of such models from data including uncertainty estimates. Here, we performed Bayesian inference for the parameters of detailed neuron models of a photoreceptor and an OFF- and an ON-cone bipolar cell from the mouse retina based on two-photon imaging data. We obtained multivariate posterior distributions specifying plausible parameter ranges consistent with the data and allowing to identify parameters poorly constrained by the data. To demonstrate the potential of such mechanistic data-driven neuron models, we created a simulation environment for external electrical stimulation of the retina and optimized stimulus waveforms to target OFF- and ON-cone bipolar cells, a current major problem of retinal neuroprosthetics.

Excitatory amino acid transporter 5 (EAAT5) glutamate transporters have a chloride channel that is strongly activated by glutamate, which modulates excitatory signaling. We found that EAAT5 is a major contributor to feedback inhibition on rod bipolar cells. Inhibition to rod bipolar cells is also mediated by GABA and glycine. GABA and glycine mediate the early phase of feedback inhibition, and EAAT5 mediates a more delayed inhibition. Together, inhibitory transmitters and EAAT5 coordinate to mediate feedback inhibition, controlling neuronal output.

Directional responses in retinal ganglion cells are generated in large part by direction‐selective release of γ‐aminobutyric acid from starburst amacrine cells onto direction‐selective ganglion cells (DSGCs). The excitatory inputs to DSGCs are also widely reported to be direction‐selective, however, recent evidence suggests that glutamate release from bipolar cells is not directional, and directional excitation seen in patch‐clamp analyses may be an artifact resulting from incomplete voltage control. Here, we test this voltage‐clamp‐artifact hypothesis in recordings from 62 ON‐OFF DSGCs in the rabbit retina. The strength of the directional excitatory signal varies considerably across the sample of cells, but is not correlated with the strength of directional inhibition, as required for a voltage‐clamp artifact. These results implicate additional mechanisms in generating directional excitatory inputs to DSGCs.

An animal's ability to survive depends on its sensory systems being able to adapt to a wide range of environmental conditions, by maximizing the information extracted and reducing the noise transmitted. The visual system does this by adapting to luminance and contrast. While luminance adaptation can begin at the retinal photoreceptors, contrast adaptation has been shown to start at later stages in the retina. Photoreceptors adapt to changes in luminance over multiple time scales ranging from tens of milliseconds to minutes, with the adaptive changes arising from processes within the phototransduction cascade. Here we show a new form of adaptation in cones that is independent of the phototransduction process. Rather, it is mediated by voltage-gated ion channels in the cone membrane and acts by changing the frequency response of cones such that their responses speed up as the membrane potential modulation depth increases and slow down as the membrane potential modulation depth decreases. This mechanism is effectively activated by high-contrast stimuli dominated by low frequencies such as natural stimuli. However, the more generally used Gaussian white noise stimuli were not effective since they did not modulate the cone membrane potential to the same extent. This new adaptive process had a time constant of less than a second. A critical component of the underlying mechanism is the hyperpolarization-activated current, Ih, as pharmacologically blocking it prevented the long- and mid- wavelength sensitive cone photoreceptors (L- and M-cones) from adapting. Consistent with this, short- wavelength sensitive cone photoreceptors (S-cones) did not show the adaptive response, and we found they also lacked a prominent Ih. The adaptive filtering mechanism identified here improves the information flow by removing higher-frequency noise during lower signal-to-noise ratio conditions, as occurs when contrast levels are low. Although this new adaptive mechanism can be driven by contrast, it is not a contrast adaptation mechanism in its strictest sense, as will be argued in the Discussion.

A persistent change in illumination causes light-adaptive changes in retinal neurons. Light adaptation improves visual encoding by preventing saturation and by adjusting spatiotemporal integration to increase the signal-to-noise ratio (SNR) and utilize signaling bandwidth efficiently. In dim light, the visual input contains a greater relative amount of quantal noise, and vertebrate receptive fields are extended in space and time to increase SNR. Whereas in bright light, SNR of the visual input is high, the rate of synaptic vesicle release from the photoreceptors is low so that quantal noise in synaptic output may limit SNR postsynaptically. Whether and how reduced synaptic SNR impacts spatiotemporal integration in postsynaptic neurons remains unclear. To address this, we measured spatiotemporal integration in retinal horizontal cells and ganglion cells in the guinea pig retina across a broad illumination range, from low to high photopic levels. In both cell types, the extent of spatial and temporal integration changed according to an inverted U-shaped function consistent with adaptation to low SNR at both low and high light levels. We show how a simple mechanistic model with interacting, opponent filters can generate the observed changes in ganglion cell spatiotemporal receptive fields across light-adaptive states and postulate that retinal neurons postsynaptic to the cones in bright light adopt low-pass spatiotemporal response characteristics to improve visual encoding under conditions of low synaptic SNR.

Directionally tuned signalling in starburst amacrine cell (SAC) dendrites lies at the heart of the circuit that detects the direction of moving stimuli in the mammalian retina. The relative contributions of intrinsic cellular properties and network connectivity to SAC direction selectivity remain unclear. Here we present a detailed connectomic reconstruction of SAC circuitry in mouse retina and describe two previously unknown features of synapse distributions along SAC dendrites: input and output synapses are segregated, with inputs restricted to proximal dendrites; and the distribution of inhibitory inputs is fundamentally different from that observed in rabbit retina. An anatomically constrained SAC network model suggests that SAC–SAC wiring differences between mouse and rabbit retina underlie distinct contributions of synaptic inhibition to velocity and contrast tuning and receptive field structure. In particular, the model indicates that mouse connectivity enables SACs to encode lower linear velocities that account for smaller eye diameter, thereby conserving angular velocity tuning. These predictions are confirmed with calcium imaging of mouse SAC dendrites responding to directional stimuli.