Mistargeted retinal axons form synaptically segregated subcircuits in the visual thalamus of albino mice.

A Hebbian model of circuit remodeling predicts that two sets of inputs with sufficiently distinct activity patterns will synaptically capture separate sets of target cells. Mice in which a subset of retinal ganglion cells (RGCs) target the wrong region of the dorsal lateral geniculate nucleus (dLGN) provide the conditions for testing this prediction. In albino mice, mistargeted RGC axons form an island of terminals that is distinct from the surrounding neuropil. Blocking retinal activity during development prevents the formation of this island. However, the synaptic connectivity of the island was unknown. Here, we combine light and electron microscopy to determine if this activity-dependent island of axon terminals represent a synaptically segregated subcircuit. We reconstructed the microcircuitry of the boundary between the island and non-island RGCs and found a remarkably strong segregation within retinogeniculate connectivity. We conclude that, when sets of retinal input are established in the wrong part of the dLGN, the developing circuitry responds by forming a synaptically isolated subcircuit from the otherwise fully connected network. The fact that there is a developmental starting condition that can induce a synaptically segregated microcircuit has important implications for our understanding of the organization of visual circuits and for our understanding of the implementation of activity dependent development.

Evaluating the Quality of Serial EM Sections with Deep Learning.

Automated image acquisition can significantly improve the throughput of serial section scanning electron microscopy (ssSEM). However, image quality can vary from image to image depending on autofocusing and beam stigmation. Automatically evaluating the quality of images is, therefore, important for efficiently generating high-quality serial section scanning electron microscopy (ssSEM) datasets. We tested several convolutional neural networks for their ability to reproduce user-generated evaluations of ssSEM image quality. We found that a modification of ResNet-50 that we term quality evaluation Network (QEN) reliably predicts user-generated quality scores. Running QEN in parallel to ssSEM image acquisition therefore allows users to quickly identify imaging problems and flag images for retaking. We have publicly shared the Python code for evaluating images with QEN, the code for training QEN, and the training dataset.

Subcellular pathways through VGluT3-expressing mouse amacrine cells provide locally tuned object-motion-selective signals in the retina.

VGluT3-expressing mouse retinal amacrine cells (VG3s) respond to small-object motion and connect to multiple types of bipolar cells (inputs) and retinal ganglion cells (RGCs, outputs). Because these input and output connections are intermixed on the same dendrites, making sense of VG3 circuitry requires comparing the distribution of synapses across their arbors to the subcellular flow of signals. Here, we combine subcellular calcium imaging and electron microscopic connectomic reconstruction to analyze how VG3s integrate and transmit visual information. VG3s receive inputs from all nearby bipolar cell types but exhibit a strong preference for the fast type 3a bipolar cells. By comparing input distributions to VG3 dendrite responses, we show that VG3 dendrites have a short functional length constant that likely depends on inhibitory shunting. This model predicts that RGCs that extend dendrites into the middle layers of the inner plexiform encounter VG3 dendrites whose responses vary according to the local bipolar cell response type.

Diversity in homeostatic calcium set points predicts retinal ganglion cell survival following optic nerve injury in vivo.

Retinal ganglion cell (RGC) degeneration drives vision loss in blinding conditions. RGC death is often triggered by axon degeneration in the optic nerve. Here, we study the contributions of dynamic and homeostatic Ca2+ levels to RGC death from axon injury. We find that axonal Ca2+ elevations from optic nerve injury do not propagate over distance or reach RGC somas, and acute and chronic Ca2+ dynamics do not affect RGC survival. Instead, we discover that baseline Ca2+ levels vary widely between RGCs and predict their survival after axon injury, and that lowering these levels reduces RGC survival. Further, we find that well-surviving RGC types have higher baseline Ca2+ levels than poorly surviving types. Finally, we observe considerable variation in the baseline Ca2+ levels of different RGCs of the same type, which are predictive of within-type differences in survival.

Reconstructing neural circuits using multiresolution correlated light and electron microscopy.

Correlated light and electron microscopy (CLEM) can be used to combine functional and molecular characterizations of neurons with detailed anatomical maps of their synaptic organization. Here we describe a multiresolution approach to CLEM (mrCLEM) that efficiently targets electron microscopy (EM) imaging to optically characterized cells while maintaining optimal tissue preparation for high-throughput EM reconstruction. This approach hinges on the ease with which arrays of sections collected on a solid substrate can be repeatedly imaged at different scales using scanning electron microscopy. We match this multiresolution EM imaging with multiresolution confocal mapping of the aldehyde-fixed tissue. Features visible in lower resolution EM correspond well to features visible in densely labeled optical maps of fixed tissue. Iterative feature matching, starting with gross anatomical correspondences and ending with subcellular structure, can then be used to target high-resolution EM image acquisition and annotation to cells of interest. To demonstrate this technique and range of images used to link live optical imaging to EM reconstructions, we provide a walkthrough of a mouse retinal light to EM experiment as well as some examples from mouse brain slices.

Efficient Coding by Midget and Parasol Ganglion Cells in the Human Retina.

In humans, midget and parasol ganglion cells account for most of the input from the eyes to the brain. Yet, how they encode visual information is unknown. Here, we perform large-scale multi-electrode array recordings from retinas of treatment-naive patients who underwent enucleation surgery for choroidal malignant melanomas. We identify robust differences in the function of midget and parasol ganglion cells, consistent asymmetries between their ON and OFF types (that signal light increments and decrements, respectively) and divergence in the function of human versus non-human primate retinas. Our computational analyses reveal that the receptive fields of human midget and parasol ganglion cells divide naturalistic movies into adjacent spatiotemporal frequency domains with equal stimulus power, while the asymmetric response functions of their ON and OFF types simultaneously maximize stimulus coverage and information transmission and minimize metabolic cost. Thus, midget and parasol ganglion cells in the human retina efficiently encode our visual environment.

An Individual Interneuron Participates in Many Kinds of Inhibition and Innervates Much of the Mouse Visual Thalamus.

One way to assess a neuron's function is to describe all its inputs and outputs. With this goal in mind, we used serial section electron microscopy to map 899 synaptic inputs and 623 outputs in one inhibitory interneuron in a large volume of the mouse visual thalamus. This neuron innervated 256 thalamocortical cells spread across functionally distinct subregions of the visual thalamus. All but one of its neurites were bifunctional, innervating thalamocortical and local interneurons while also receiving synapses from the retina. We observed a wide variety of local synaptic motifs. While this neuron innervated many cells weakly, with single en passant synapses, it also deployed specialized branches that climbed along other dendrites to form strong multi-synaptic connections with a subset of partners. This neuron's diverse range of synaptic relationships allows it to participate in a mix of global and local processing but defies assigning it a single circuit function.

A Fine-Scale Functional Logic to Convergence from Retina to Thalamus.

Numerous well-defined classes of retinal ganglion cells innervate the thalamus to guide image-forming vision, yet the rules governing their convergence and divergence remain unknown. Using two-photon calcium imaging in awake mouse thalamus, we observed a functional arrangement of retinal ganglion cell axonal boutons in which coarse-scale retinotopic ordering gives way to fine-scale organization based on shared preferences for other visual features. Specifically, at the ∼6 µm scale, clusters of boutons from different axons often showed similar preferences for either one or multiple features, including axis and direction of motion, spatial frequency, and changes in luminance. Conversely, individual axons could "de-multiplex" information channels by participating in multiple, functionally distinct bouton clusters. Finally, ultrastructural analyses demonstrated that retinal axonal boutons in a local cluster often target the same dendritic domain. These data suggest that functionally specific convergence and divergence of retinal axons may impart diverse, robust, and often novel feature selectivity to visual thalamus.

Digital tissue and what it may reveal about the brain.

Imaging as a means of scientific data storage has evolved rapidly over the past century from hand drawings, to photography, to digital images. Only recently can sufficiently large datasets be acquired, stored, and processed such that tissue digitization can actually reveal more than direct observation of tissue. One field where this transformation is occurring is connectomics: the mapping of neural connections in large volumes of digitized brain tissue.

A connectomic approach to the lateral geniculate nucleus.

Although the core functions and structure of the lateral geniculate nucleus (LGN) are well understood, this core is surrounded by questions about the integration of feedforward and feedback connections, interactions between different channels of information, and how activity dependent development restructures synaptic networks. Our understanding of the organization of the mouse LGN is particularly limited given how important it has become as a model system. Advances in circuit scale electron microscopy (cellular connectomics) have made it possible to reconstruct the synaptic connectivity of hundreds of neurons within in a circuit the size of the mouse LGN. These circuit reconstructions can reveal cell type-to-cell type canonical wiring diagrams as well as the higher order wiring motifs that are only visible in reconstructions of intact networks. Connectomic analysis of the LGN therefore not only can answer longstanding questions about the organization of the visual thalamus but also presents unique opportunities for investigating fundamental properties of mammalian circuit formation.

The Fuzzy Logic of Network Connectivity in Mouse Visual Thalamus.

In an attempt to chart parallel sensory streams passing through the visual thalamus, we acquired a 100-trillion-voxel electron microscopy (EM) dataset and identified cohorts of retinal ganglion cell axons (RGCs) that innervated each of a diverse group of postsynaptic thalamocortical neurons (TCs). Tracing branches of these axons revealed the set of TCs innervated by each RGC cohort. Instead of finding separate sensory pathways, we found a single large network that could not be easily subdivided because individual RGCs innervated different kinds of TCs and different kinds of RGCs co-innervated individual TCs. We did find conspicuous network subdivisions organized on the basis of dendritic rather than neuronal properties. This work argues that, in the thalamus, neural circuits are not based on a canonical set of connections between intrinsically different neuronal types but, rather, may arise by experience-based mixing of different kinds of inputs onto individual postsynaptic cells.

Saturated Reconstruction of a Volume of Neocortex.

We describe automated technologies to probe the structure of neural tissue at nanometer resolution and use them to generate a saturated reconstruction of a sub-volume of mouse neocortex in which all cellular objects (axons, dendrites, and glia) and many sub-cellular components (synapses, synaptic vesicles, spines, spine apparati, postsynaptic densities, and mitochondria) are rendered and itemized in a database. We explore these data to study physical properties of brain tissue. For example, by tracing the trajectories of all excitatory axons and noting their juxtapositions, both synaptic and non-synaptic, with every dendritic spine we refute the idea that physical proximity is sufficient to predict synaptic connectivity (the so-called Peters' rule). This online minable database provides general access to the intrinsic complexity of the neocortex and enables further data-driven inquiries.

Imaging ATUM ultrathin section libraries with WaferMapper: a multi-scale approach to EM reconstruction of neural circuits.

The automated tape-collecting ultramicrotome (ATUM) makes it possible to collect large numbers of ultrathin sections quickly-the equivalent of a petabyte of high resolution images each day. However, even high throughput image acquisition strategies generate images far more slowly (at present ~1 terabyte per day). We therefore developed WaferMapper, a software package that takes a multi-resolution approach to mapping and imaging select regions within a library of ultrathin sections. This automated method selects and directs imaging of corresponding regions within each section of an ultrathin section library (UTSL) that may contain many thousands of sections. Using WaferMapper, it is possible to map thousands of tissue sections at low resolution and target multiple points of interest for high resolution imaging based on anatomical landmarks. The program can also be used to expand previously imaged regions, acquire data under different imaging conditions, or re-image after additional tissue treatments.

The spatial structure of a nonlinear receptive field.

Understanding a sensory system implies the ability to predict responses to a variety of inputs from a common model. In the retina, this includes predicting how the integration of signals across visual space shapes the outputs of retinal ganglion cells. Existing models of this process generalize poorly to predict responses to new stimuli. This failure arises in part from properties of the ganglion cell response that are not well captured by standard receptive-field mapping techniques: nonlinear spatial integration and fine-scale heterogeneities in spatial sampling. Here we characterize a ganglion cell's spatial receptive field using a mechanistic model based on measurements of the physiological properties and connectivity of only the primary excitatory circuitry of the retina. The resulting simplified circuit model successfully predicts ganglion-cell responses to a variety of spatial patterns and thus provides a direct correspondence between circuit connectivity and retinal output.

Coating particles with dextran-conjugated fluorescent dyes or other hydrophilic compounds.

Imaging and reconstruction of developing neurons require cells that are labeled in a way that distinguishes them from their neighbors. This can be achieved with ballistic labeling, which refers to the delivery of a cell label by means of carrier particles (tungsten or gold) propelled from a pressurized gun. Ballistic delivery can reach many dispersed cells in one shot and can deploy a wide variety of cell markers to neurons in diverse preparations. The three most commonly used types of ballistic labels are carbocyanine dyes, dextran-conjugated fluorescent markers, and DNA plasmids. The primary advantage of ballistic labeling is that multiple dispersed cells can be labeled quickly in live or fixed tissue. This article describes a protocol for coating tungsten particles with dextran-conjugated fluorescent dyes or ion indicators. Such hydrophilic compounds conjugated to dextran are water soluble, and therefore they are excellent indicators for functional studies within living cells. This protocol was developed for labeling ganglion cells in retinal flat mounts.

Coating gold particles with DNA (biolistics).

Imaging and reconstruction of developing neurons require cells that are labeled in a way that distinguishes them from their neighbors. This can be achieved with ballistic labeling, which refers to the delivery of a cell label by means of carrier particles (tungsten or gold) propelled from a pressurized gun. Ballistic delivery can reach many dispersed cells in one shot and can deploy a wide variety of cell markers to neurons in diverse preparations. The three most commonly used types of ballistic labels are carbocyanine dyes, dextran-conjugated fluorescent markers, and DNA plasmids. The primary advantage of ballistic labeling is that multiple dispersed cells can be labeled quickly in live or fixed tissue. This article describes a protocol for coating gold particles with plasmid DNA, which can be used to label developing ganglion cells in retinal flat mounts.

Coating particles with carbocyanine dyes.

Imaging and reconstruction of developing neurons require cells that are labeled in a way that distinguishes them from their neighbors. This can be achieved with ballistic labeling, which refers to the delivery of a cell label by means of carrier particles (tungsten or gold) propelled from a pressurized gun. Ballistic delivery can reach many dispersed cells in one shot and can deploy a wide variety of cell markers to neurons in diverse preparations. The three most commonly used types of ballistic labels are carbocyanine dyes, dextran-conjugated fluorescent markers, and DNA plasmids. The primary advantage of ballistic labeling is that multiple dispersed cells can be labeled quickly in live or fixed tissue. This article describes a protocol for coating tungsten particles (∼1 µm in diameter) with carbocyanine dyes, which are widely used to label neurons in tissue and neural cells in suspension. These dyes are lipophilic and highly fluorescent within lipid bilayers. Because tissue damage worsens with the increasing pressure required for deeper bullet penetration, ballistic labeling of neurons is most effective when the target cells are near the surface of the preparation. This protocol was developed for labeling ganglion cells in retinal flat mounts.

Shooting DNA, dyes, or indicators into tissue slices using the gene gun.

Imaging and reconstruction of developing neurons require cells that are labeled in a way that distinguishes them from their neighbors. This can be achieved with ballistic labeling, which refers to the delivery of a cell label by means of carrier particles (tungsten or gold) propelled from a pressurized gun. Ballistic delivery can reach many dispersed cells in one shot and can deploy a wide variety of cell markers to neurons in diverse preparations. The three most commonly used types of ballistic labels are carbocyanine dyes, dextran-conjugated fluorescent markers, and DNA plasmids. This article describes a protocol for using a Helios Gene Gun (Bio-Rad Laboratories) to inject coated particles into cells located near the surface of a tissue preparation. Shooting particles coated with carbocyanine dyes or dextran-conjugated fluorescent markers requires that a filter be placed between the gene gun and the target tissue. The filter prevents unbound dye clumps from reaching the tissue and attenuates the pressure wave reaching the tissue. DNA-coated particles can be shot without a filter if the target cells are located near enough to the surface (<20 µm deep) for the particles to penetrate using low helium pressures (35-40 psi).

Neurotransmission selectively regulates synapse formation in parallel circuits in vivo.

Activity is thought to guide the patterning of synaptic connections in the developing nervous system. Specifically, differences in the activity of converging inputs are thought to cause the elimination of synapses from less active inputs and increase connectivity with more active inputs. Here we present findings that challenge the generality of this notion and offer a new view of the role of activity in synapse development. To imbalance neurotransmission from different sets of inputs in vivo, we generated transgenic mice in which ON but not OFF types of bipolar cells in the retina express tetanus toxin (TeNT). During development, retinal ganglion cells (RGCs) select between ON and OFF bipolar cell inputs (ON or OFF RGCs) or establish a similar number of synapses with both on separate dendritic arborizations (ON-OFF RGCs). In TeNT retinas, ON RGCs correctly selected the silenced ON bipolar cell inputs over the transmitting OFF bipolar cells, but were connected with them through fewer synapses at maturity. Time-lapse imaging revealed that this was caused by a reduced rate of synapse formation rather than an increase in synapse elimination. Similarly, TeNT-expressing ON bipolar cell axons generated fewer presynaptic active zones. The remaining active zones often recruited multiple, instead of single, synaptic ribbons. ON-OFF RGCs in TeNT mice maintained convergence of ON and OFF bipolar cells inputs and had fewer synapses on their ON arbor without changes to OFF arbor synapses. Our results reveal an unexpected and remarkably selective role for activity in circuit development in vivo, regulating synapse formation but not elimination, affecting synapse number but not dendritic or axonal patterning, and mediating independently the refinement of connections from parallel (ON and OFF) processing streams even where they converge onto the same postsynaptic cell.