Publications

Animal behavior is principally expressed through neural control of muscles. Therefore understanding how the brain controls behavior requires mapping neuronal circuits all the way to motor neurons. We have previously established technology to collect large-volume electron microscopy data sets of neural tissue and fully reconstruct the morphology of the neurons and their chemical synaptic connections throughout the volume. Using these tools we generated a dense wiring diagram, or connectome, for a large portion of the Drosophila central brain. However, in most animals, including the fly, the majority of motor neurons are located outside the brain in a neural center closer to the body, i.e. the mammalian spinal cord or insect ventral nerve cord (VNC). In this paper, we extend our effort to map full neural circuits for behavior by generating a connectome of the VNC of a male fly.

Our companion paper (Takemura et al., 2023) introduces the first completely proofread connectome of the nerve cord of an animal that can walk or fly. The base connectome consists of neuronal morphologies and the connections between them. However, in order to efficiently navigate and understand this connectome, it is crucial to have a system of annotations that systematically categorises and names neurons, linking them to the existing literature. In this paper we describe the comprehensive annotation of the VNC connectome, first by a system of hierarchical coarse annotations, then by grouping left-right and serially homologous neurons and eventually by defining systematic cell types for the intrinsic interneurons and sensory neurons of the VNC; descending and motor neurons are typed in (Cheong et al., 2023). We assign a sensory modality to over 5000 sensory neurons, cluster them by connectivity, and identify serially homologous cell types and a layered organisation likely corresponding to peripheral topography. We identify the developmental neuroblast of origin of the large majority of VNC neurons and confirm that (in most cases) all secondary neurons of each hemilineage express a single neurotransmitter. Neuroblast hemilineages are serially repeated along the segments of the nerve cord and generally exhibit consistent hemilineage-to-hemilineage connectivity across neuromeres, supporting the idea that hemilineages are a major organisational feature of the VNC. We also find that more than a third of individual neurons belong to serially homologous cell types, which were crucial for identifying motor neurons and sensory neurons across leg neuropils. Categorising interneurons by their neuropil innervation patterns provides an additional organisation axis. Over half of the intrinsic neurons of the VNC appear dedicated to the legs, with the majority restricted to single leg neuropils; in contrast, inhibitory interneurons connecting different leg neuropils, especially those crossing the midline, appear rarer than anticipated by standard models of locomotor circuitry. Our annotations are being released as part of the neuprint.janelia.org web application and also serve as the basis of programmatic analysis of the connectome through dedicated tools that we describe in this paper.

In most animals, a relatively small number of descending neurons (DNs) connect higher brain centers in the animal’s head to motor neurons (MNs) in the nerve cord of the animal’s body that effect movement of the limbs. To understand how brain signals generate behavior, it is critical to understand how these descending pathways are organized onto the body MNs. In the fly, Drosophila melanogaster, MNs controlling muscles in the leg, wing, and other motor systems reside in a ventral nerve cord (VNC), analogous to the mammalian spinal cord. In companion papers, we introduced a densely-reconstructed connectome of the Drosophila Male Adult Nerve Cord (MANC, Takemura et al., 2023), including cell type and developmental lineage annotation (Marin et al., 2023), which provides complete VNC connectivity at synaptic resolution. Here, we present a first look at the organization of the VNC networks connecting DNs to MNs based on this new connectome information. We proofread and curated all DNs and MNs to ensure accuracy and reliability, then systematically matched DN axon terminals and MN dendrites with light microscopy data to link their VNC morphology with their brain inputs or muscle targets. We report both broad organizational patterns of the entire network and fine-scale analysis of selected circuits of interest. We discover that direct DN-MN connections are infrequent and identify communities of intrinsic neurons linked to control of different motor systems, including putative ventral circuits for walking, dorsal circuits for flight steering and power generation, and intermediate circuits in the lower tectulum for coordinated action of wings and legs. Our analysis generates hypotheses for future functional experiments and, together with the MANC connectome, empowers others to investigate these and other circuits of the Drosophila ventral nerve cord in richer mechanistic detail.

The neural circuits responsible for animal behavior remain largely unknown. We summarize new methods and present the circuitry of a large fraction of the brain of the fruit fly . Improved methods include new procedures to prepare, image, align, segment, find synapses in, and proofread such large data sets. We define cell types, refine computational compartments, and provide an exhaustive atlas of cell examples and types, many of them novel. We provide detailed circuits consisting of neurons and their chemical synapses for most of the central brain. We make the data public and simplify access, reducing the effort needed to answer circuit questions, and provide procedures linking the neurons defined by our analysis with genetic reagents. Biologically, we examine distributions of connection strengths, neural motifs on different scales, electrical consequences of compartmentalization, and evidence that maximizing packing density is an important criterion in the evolution of the fly's brain.

Understanding memory formation, storage and retrieval requires knowledge of the underlying neuronal circuits. In Drosophila, the mushroom body (MB) is the major site of associative learning. We reconstructed the morphologies and synaptic connections of all 983 neurons within the three functional units, or compartments, that compose the adult MB’s α lobe, using a dataset of isotropic 8-nm voxels collected by focused ion-beam milling scanning electron microscopy. We found that Kenyon cells (KCs), whose sparse activity encodes sensory information, each make multiple en passant synapses to MB output neurons (MBONs) in each compartment. Some MBONs have inputs from all KCs, while others differentially sample sensory modalities. Only six percent of KC>MBON synapses receive a direct synapse from a dopaminergic neuron (DAN). We identified two unanticipated classes of synapses, KC>DAN and DAN>MBON. DAN activation produces a slow depolarization of the MBON in these DAN>MBON synapses and can weaken memory recall.

Animal behaviour arises from computations in neuronal circuits, but our understanding of these computations has been frustrated by the lack of detailed synaptic connection maps, or connectomes. For example, despite intensive investigations over half a century, the neuronal implementation of local motion detection in the insect visual system remains elusive. Here we develop a semi-automated pipeline using electron microscopy to reconstruct a connectome, containing 379 neurons and 8,637 chemical synaptic contacts, within the Drosophila optic medulla. By matching reconstructed neurons to examples from light microscopy, we assigned neurons to cell types and assembled a connectome of the repeating module of the medulla. Within this module, we identified cell types constituting a motion detection circuit, and showed that the connections onto individual motion-sensitive neurons in this circuit were consistent with their direction selectivity. Our results identify cellular targets for future functional investigations, and demonstrate that connectomes can provide key insights into neuronal computations.

Understanding the circuit mechanisms behind motion detection is a long-standing question in visual neuroscience. In , recent synapse-level connectomes in the optic lobe, particularly in ON-pathway (T4) receptive-field circuits, in concert with physiological studies, suggest an increasingly intricate motion model compared with the ubiquitous Hassenstein-Reichardt model, while our knowledge of OFF-pathway (T5) has been incomplete. Here we present a conclusive and comprehensive connectome that for the first time integrates detailed connectivity information for inputs to both T4 and T5 pathways in a single EM dataset covering the entire optic lobe. With novel reconstruction methods using automated synapse prediction suited to such a large connectome, we successfully corroborate previous findings in the T4 pathway and comprehensively identify inputs and receptive fields for T5. While the two pathways are likely evolutionarily linked and indeed exhibit many similarities, we uncover interesting differences and interactions that may underlie their distinct functional properties.

Due to advances in electron microscopy and deep learning, it is now practical to reconstruct a connectome, a description of neurons and the chemical synapses between them, for significant volumes of neural tissue. Smaller past reconstructions were primarily used by domain experts, could be handled by downloading data, and performance was not a serious problem. But new and much larger reconstructions upend these assumptions. These networks now contain tens of thousands of neurons and tens of millions of connections, with yet larger reconstructions pending, and are of interest to a large community of non-specialists. Allowing other scientists to make use of this data needs more than publication-it requires new tools that are publicly available, easy to use, and efficiently handle large data. We introduce neuPrint to address these data analysis challenges. Neuprint contains two major components-a web interface and programmer APIs. The web interface is designed to allow any scientist worldwide, using only a browser, to quickly ask and answer typical biological queries about a connectome. The neuPrint APIs allow more computer-savvy scientists to make more complex or higher volume queries. NeuPrint also provides features for assessing reconstruction quality. Internally, neuPrint organizes connectome data as a graph stored in a neo4j database. This gives high performance for typical queries, provides access though a public and well documented query language Cypher, and will extend well to future larger connectomics databases. Our experience is also an experiment in open science. We find a significant fraction of the readers of the article proceed to examine the data directly. In our case preprints worked exactly as intended, with data inquiries and PDF downloads starting immediately after pre-print publication, and little affected by formal publication later. From this we deduce that many readers are more interested in our data than in our analysis of our data, suggesting that data-only papers can be well appreciated and that public data release can speed up the propagation of scientific results by many months. We also find that providing, and keeping, the data available for online access imposes substantial additional costs to connectomics research.

Open-source software development has skyrocketed in part due to community tools like github.com, which allows publication of code as well as the ability to create branches and push accepted modifications back to the original repository. As the number and size of EM-based datasets increases, the connectomics community faces similar issues when we publish snapshot data corresponding to a publication. Ideally, there would be a mechanism where remote collaborators could modify branches of the data and then flexibly reintegrate results via moderated acceptance of changes. The DVID system provides a web-based connectomics API and the first steps toward such a distributed versioning approach to EM-based connectomics datasets. Through its use as the central data resource for Janelia's FlyEM team, we have integrated the concepts of distributed versioning into reconstruction workflows, allowing support for proofreader training and segmentation experiments through branched, versioned data. DVID also supports persistence to a variety of storage systems from high-speed local SSDs to cloud-based object stores, which allows its deployment on laptops as well as large servers. The tailoring of the backend storage to each type of connectomics data leads to efficient storage and fast queries. DVID is freely available as open-source software with an increasing number of supported storage options.

Recent advances in automatic image segmentation and synapse prediction in electron microscopy (EM) datasets of the brain enable more efficient reconstruction of neural connectivity. In these datasets, a single neuron can span thousands of images containing complex tree-like arbors with thousands of synapses. While image segmentation algorithms excel within narrow fields of views, the algorithms sometimes struggle to correctly segment large neurons, which require large context given their size and complexity. Conversely, humans are comparatively good at reasoning with large objects. In this paper, we introduce several semi-automated strategies that combine 3D visualization and machine guidance to accelerate connectome reconstruction. In particular, we introduce a strategy to quickly correct a segmentation through merging and cleaving, or splitting a segment along supervoxel boundaries, with both operations driven by affinity scores in the underlying segmentation. We deploy these algorithms as streamlined workflows in a tool called Neu3 and demonstrate superior performance compared to prior work, thus enabling efficient reconstruction of much larger datasets. The insights into proofreading from our work clarify the trade-offs to consider when tuning the parameters of image segmentation algorithms.

We propose a method to facilitate exploration and analysis of new large data sets. In particular, we give an unsupervised deep learning approach to learning a latent representation that captures semantic similarity in the data set. The core idea is to use data augmentations that preserve semantic meaning to generate synthetic examples of elements whose feature representations should be close to one another.
We demonstrate the utility of our method applied to nano-scale electron microscopy data, where even relatively small portions of animal brains can require terabytes of image data. Although supervised methods can be used to predict and identify known patterns of interest, the scale of the data makes it difficult to mine and analyze patterns that are not known a priori. We show the ability of our learned representation to enable query by example, so that if a scientist notices an interesting pattern in the data, they can be presented with other locations with matching patterns. We also demonstrate that clustering of data in the learned space correlates with biologically-meaningful distinctions. Finally, we introduce a visualization tool and software ecosystem to facilitate user-friendly interactive analysis and uncover interesting biological patterns. In short, our work opens possible new avenues in understanding of and discovery in large data sets, arising in domains such as EM analysis.

Visual pathways from the compound eye of an insect relay to four neuropils, successively the lamina, medulla, lobula, and lobula plate in the underlying optic lobe. Among these neuropils, the medulla, lobula, and lobula plate are interconnected by the complex second optic chiasm, through which the anteroposterior axis undergoes an inversion between the medulla and lobula. Given their complex structure, the projection patterns through the second optic chiasm have so far lacked critical analysis. By densely reconstructing axon trajectories using a volumetric scanning electron microscopy (SEM) technique, we reveal the three-dimensional structure of the second optic chiasm of , which comprises interleaving bundles and sheets of axons insulated from each other by glial sheaths. These axon bundles invert their horizontal sequence in passing between the medulla and lobula. Axons connecting the medulla and lobula plate are also bundled together with them but do not decussate the sequence of their horizontal positions. They interleave with sheets of projection neuron axons between the lobula and lobula plate, which also lack decussations. We estimate that approximately 19,500 cells per hemisphere, about two thirds of the optic lobe neurons, contribute to the second chiasm, most being Tm cells, with an estimated additional 2,780 T4 and T5 cells each. The chiasm mostly comprises axons and cell body fibers, but also a few synaptic elements. Based on our anatomical findings, we propose that a chiasmal structure between the neuropils is potentially advantageous for processing complex visual information in parallel. The EM reconstruction shows not only the structure of the chiasm in the adult brain, the previously unreported main topic of our study, but also suggest that the projection patterns of the neurons comprising the chiasm may be determined by the proliferation centers from which the neurons develop. Such a complex wiring pattern could, we suggest, only have arisen in several evolutionary steps.

Using FIB-SEM we report the entire synaptic connectome of glomerulus VA1v of the right antennal lobe in . Within the glomerulus we densely reconstructed all neurons, including hitherto elusive local interneurons. The -positive, sexually dimorphic VA1v included >11,140 presynaptic sites with ~38,050 postsynaptic dendrites. These connected input olfactory receptor neurons (ORNs, 51 ipsilateral, 56 contralateral), output projection neurons (18 PNs), and local interneurons (56 of >150 previously reported LNs). ORNs are predominantly presynaptic and PNs predominantly postsynaptic; newly reported LN circuits are largely an equal mixture and confer extensive synaptic reciprocity, except the newly reported LN2V with input from ORNs and outputs mostly to monoglomerular PNs, however. PNs were more numerous than previously reported from genetic screens, suggesting that the latter failed to reach saturation. We report a matrix of 192 bodies each having 50 connections; these form 88% of the glomerulus' pre/postsynaptic sites.

Reconstructing a connectome from an EM dataset often requires a large effort of proofreading automatically generated segmentations. While many tools exist to enable tracing or proofreading, recent advances in EM imaging and segmentation quality suggest new strategies and pose unique challenges for tool design to accelerate proofreading. Namely, we now have access to very large multi-TB EM datasets where (1) many segments are largely correct, (2) segments can be very large (several GigaVoxels), and where (3) several proofreaders and scientists are expected to collaborate simultaneously. In this paper, we introduce NeuTu as a solution to efficiently proofread large, high-quality segmentation in a collaborative setting. NeuTu is a client program of our high-performance, scalable image database called DVID so that it can easily be scaled up. Besides common features of typical proofreading software, NeuTu tames unprecedentedly large data with its distinguishing functions, including: (1) low-latency 3D visualization of large mutable segmentations; (2) interactive splitting of very large false merges with highly optimized semi-automatic segmentation; (3) intuitive user operations for investigating or marking interesting points in 3D visualization; (4) visualizing proofreading history of a segmentation; and (5) real-time collaborative proofreading with lock-based concurrency control. These unique features have allowed us to manage the workflow of proofreading a large dataset smoothly without dividing them into subsets as in other segmentation-based tools. Most importantly, NeuTu has enabled some of the largest connectome reconstructions as well as interesting discoveries in the fly brain.

Extracting a connectome from an electron microscopy (EM) data set requires identification of neurons and determination of synapses between neurons. As manual extraction of this information is very time-consuming, there has been extensive research effort to automatically segment the neurons to help guide and eventually replace manual tracing. Until recently, there has been comparatively less research on automatically detecting the actual synapses between neurons. This discrepancy can, in part, be attributed to several factors: obtaining neuronal shapes is a prerequisite first step in extracting a connectome, manual tracing is much more time-consuming than annotating synapses, and neuronal contact area can be used as a proxy for synapses in determining connections.
However, recent research has demonstrated that contact area alone is not a sufficient predictor of synaptic connection. Moreover, as segmentation has improved, we have observed that synapse annotation is consuming a more significant fraction of overall reconstruction time. This ratio will only get worse as segmentation improves, gating overall possible speed-up. Therefore, we address this problem by developing algorithms that automatically detect pre-synaptic neurons and their post-synaptic partners. In particular, pre-synaptic structures are detected using a Deep and Wide Multiscale Recursive Network, and post-synaptic partners are detected using a MLP with features conditioned on the local segmentation.
This work is novel because it requires minimal amount of training, leverages advances in image segmentation directly, and provides a complete solution for polyadic synapse detection. We further introduce novel metrics to evaluate our algorithm on connectomes of meaningful size. These metrics demonstrate that complete automatic prediction can be used to effectively characterize most connectivity correctly.

We reconstructed the synaptic circuits of seven columns in the second neuropil or medulla behind the fly's compound eye. These neurons embody some of the most stereotyped circuits in one of the most miniaturized of animal brains. The reconstructions allow us, for the first time to our knowledge, to study variations between circuits in the medulla's neighboring columns. This variation in the number of synapses and the types of their synaptic partners has previously been little addressed because methods that visualize multiple circuits have not resolved detailed connections, and existing connectomic studies, which can see such connections, have not so far examined multiple reconstructions of the same circuit. Here, we address the omission by comparing the circuits common to all seven columns to assess variation in their connection strengths and the resultant rates of several different and distinct types of connection error. Error rates reveal that, overall, <1% of contacts are not part of a consensus circuit, and we classify those contacts that supplement (E+) or are missing from it (E-). Autapses, in which the same cell is both presynaptic and postsynaptic at the same synapse, are occasionally seen; two cells in particular, Dm9 and Mi1, form ≥20-fold more autapses than do other neurons. These results delimit the accuracy of developmental events that establish and normally maintain synaptic circuits with such precision, and thereby address the operation of such circuits. They also establish a precedent for error rates that will be required in the new science of connectomics.

Recent results have shown the possibility of both reconstructing connectomes of small but biologically interesting circuits and extracting from these connectomes insights into their function. However, these reconstructions were heroic proof-of-concept experiments, requiring person-months of effort per neuron reconstructed, and will not scale to larger circuits, much less the brains of entire animals. In this paper we examine what will be required to generate and use substantially larger connectomes, finding five areas that need increased attention: firstly, imaging better suited to automatic reconstruction, with excellent z-resolution; secondly, automatic detection, validation, and measurement of synapses; thirdly, reconstruction methods that keep and use uncertainty metrics for every object, from initial images, through segmentation, reconstruction, and connectome queries; fourthly, processes that are fully incremental, so that the connectome may be used before it is fully complete; and finally, better tools for analysis of connectomes, once they are obtained.

Pixel and superpixel classifiers have become essential tools for EM segmentation algorithms. Training these classifiers remains a major bottleneck primarily due to the requirement of completely annotating the dataset which is tedious, error-prone and costly. In this paper, we propose an interactive learning scheme for the superpixel classifier for EM segmentation. Our algorithm is 'active semi-supervised' because it requests the labels of a small number of examples from user and applies label propagation technique to generate these queries. Using only a small set (< 20%) of all datapoints, the proposed algorithm consistently generates a classifier almost as accurate as that estimated from a complete groundtruth. We provide segmentation results on multiple datasets to show the strength of these classifiers.

We aim to improve segmentation through the use of machine learning tools during region agglomeration. We propose an active learning approach for performing hierarchical agglomerative segmentation from superpixels. Our method combines multiple features at all scales of the agglomerative process, works for data with an arbitrary number of dimensions, and scales to very large datasets. We advocate the use of variation of information to measure segmentation accuracy, particularly in 3D electron microscopy (EM) images of neural tissue, and using this metric demonstrate an improvement over competing algorithms in EM and natural images.

The ability to automatically segment an image into distinct regions is a critical aspect in many visual processing applications. Because inaccuracies often exist in automatic segmentation, manual segmentation is necessary in some application domains to correct mistakes, such as required in the reconstruction of neuronal processes from microscopic images. The goal of the automated segmentation tool is traditionally to produce the highest-quality segmentation, where quality is measured by the similarity to actual ground truth, so as to minimize the volume of manual correction necessary. Manual correction is generally orders-of-magnitude more time consuming than automated segmentation, often making handling large images intractable. Therefore, we propose a more relevant goal: minimizing the turn-around time of automated/manual segmentation while attaining a level of similarity with ground truth. It is not always necessary to inspect every aspect of an image to generate a useful segmentation. As such, we propose a strategy to guide manual segmentation to the most uncertain parts of segmentation. Our contributions include 1) a probabilistic measure that evaluates segmentation without ground truth and 2) a methodology that leverages these probabilistic measures to significantly reduce manual correction while maintaining segmentation quality.

The most established method of reconstructing neural circuits from animals involves slicing tissue very thin, then taking mosaics of electron microscope (EM) images. To trace neurons across different images and through different sections, these images must be accurately aligned, both with the others in the same section and to the sections above and below. Unfortunately, sectioning and imaging are not ideal processes - some of the problems that make alignment difficult include lens distortion, tissue shrinkage during imaging, tears and folds in the sectioned tissue, and dust and other artifacts. In addition the data sets are large (hundreds of thousands of images) and each image must be aligned with many neighbors, so the process must be automated and reliable. This paper discusses methods of dealing with these problems, with numeric results describing the accuracy of the resulting alignments.

A central problem in neuroscience is reconstructing neuronal circuits on the synapse level. Due to a wide range of scales in brain architecture such reconstruction requires imaging that is both high-resolution and high-throughput. Existing electron microscopy (EM) techniques possess required resolution in the lateral plane and either high-throughput or high depth resolution but not both. Here, we exploit recent advances in unsupervised learning and signal processing to obtain high depth-resolution EM images computationally without sacrificing throughput. First, we show that the brain tissue can be represented as a sparse linear combination of localized basis functions that are learned using high-resolution datasets. We then develop compressive sensing-inspired techniques that can reconstruct the brain tissue from very few (typically 5) tomographic views of each section. This enables tracing of neuronal processes and, hence, high throughput reconstruction of neural circuits on the level of individual synapses.

Reconstructing neuronal circuits at the level of synapses is a central problem in neuroscience, and the focus of the nascent field of connectomics. Previously used to reconstruct the C. elegans wiring diagram, serial-section transmission electron microscopy (ssTEM) is a proven technique for the task. However, to reconstruct more complex circuits, ssTEM will require the automation of image processing. We review progress in the processing of electron microscopy images and, in particular, a semi-automated reconstruction pipeline deployed at Janelia. Drosophila circuits underlying identified behaviors are being reconstructed in the pipeline with the goal of generating a complete Drosophila connectome.