Yemini, Eviatar et al. (2015) International Worm Meeting "Automated Neural Identification in C. elegans."

Yemini, Eviatar, Hobert, Oliver

[International Worm Meeting, 2015]

Stereotypy of the C. elegans nervous system affords single-neuron registration across animals and, consequently, robust statistics for neuronal transcriptomics and neural coding for behavior. Fast methods of whole-brain imaging in worm exist but, unfortunately, determining the identity of neurons within these images remains a long, difficult, ad hoc process. We are developing 2 landmark strains with accompanying software to perform automated neural identification. These strains will provide an unambiguous label for each neuron while preserving a channel free for signals reporting transcription (GFP and RFP) and neural activity (GCaMP and RCaMP); thus permitting whole-brain, single-neuron registration across animals for these signals. At present, we have tried 15 potential fluorescent colors as landmarks and settled on 3 per strain that perform well within the given constraints: differentiable, bright, stable without confounding the signal channel. Moreover, we have developed an initial landmark strain, using targeted promoters driving fluorescent proteins, that unambiguously identifies nearly 80% of C. elegans neurons.

Yemini, Eviatar I. et al. (2013) International Worm Meeting "A database of C. elegans behavioral phenotypes."

Brown, Andre E.X., Yemini, Eviatar I., Grundy, Laura J., Schafer, William R., Jucikas, Tadas

[International Worm Meeting, 2013]

Previous single and multi worm tracking experiments have produced summary statistics to phenotype small worm sets. We introduce a database of extensive and intensive single-worm phenotypes for over 300 strains of C. elegans with nervous system and locomotory defects as well as a reference of N2 variability composed of more than 1,200 young-adult hermaphrodites examined over the course of 3 years. The data is available online at http://wormbehavior.mrc-lmb.cam.ac.uk and includes a link to Worm Tracker 2.0 (our single-worm tracker used for data collection). Our phenomic database provides multiple levels of representation, from high-level statistical strain summaries all the way down to detailed time-series measurements for over 10,000 single-worm experiments. Within our database are 76 mutants with no previously characterized phenotype, 15 genes with multiple allelic representation, and 13 double or triple mutant combinations (the majority of which are accompanied by single mutant representation as well). Annotated experimental videos are easily accessible alongside their data, with various degrees of processing, from the skeleton and outline coordinates to the time series of extracted features, their histograms, and an in-depth view of collective strain statistics. For computational researchers, the database is a rich source of processed measures and raw data for developing new algorithms for segmentation, behavioral quantification, and bioinformatic approaches which link complex phenotypes with genetic perturbations. For neurogeneticists, the summary statistics and visualizations make it possible to identify behavioral phenotypes in mutants of interest. Free Worm Tracker 2.0 software and simple plans to build its inexpensive hardware are available at http://www.mrc-lmb.cam.ac.uk/wormtracker/.

Eviatar I Yemini et al. (2007) International Worm Meeting "A Simple, Inexpensive, and Feature-Rich Single-Worm Tracker."

Ryan C Lustig, Eviatar I Yemini, Christopher J Cronin, Paul W Sternberg, William R Schafer

[International Worm Meeting, 2007]

Past attempts at building a software and hardware solution to tracking single-worms have yielded complex and expensive results. We are developing a new open-source software solution that is simple to use, operates with a variety of inexpensive hardware, and offers a rich set of features that allow a researcher to follow and record video of a single, freely-behaving worm. This new program, titled Worm Tracker 2.0, will install with ease. It will run on a standard personal computer with any modern operating system (Windows, Macintosh, Unix variants, etc.). We expect to support most inexpensive cameras and the only major cost is incurred should be in purchasing a motorized stage and ordinary microscope, if necessary. Worm Tracker 2.0 has a rich set of features. Among them, the program can record video of a moving worm at 30 frames/second with duration limited only by the size of the hard drive. Video can be captured intermittently in the event that a researcher wishes to observe long-term changes. The program is robust and should track worms partially obscured by food and the edges of their enclosing dish. Analysis software is included to extract over 100 features describing the worms morphology and behavior. The code is abundantly documented and supported by a growing community of developers that provide software additions, updates, bug-fixes, and general troubleshooting. We hope to build a large network of labs using this new software and hardware solution to worm tracking.

Butler, Victoria et al. (2009) International Worm Meeting "Worm Tracker 2.0: Generating a phenome database."

Sternberg, Paul, Schafer, William, Yemini, Eviatar, Cronin, Chris, Butler, Victoria, Jucikas, Tadas

[International Worm Meeting, 2009]

The nematode Caenorhabditis elegans is widely used for the genetic analysis of nervous system function. This analysis relies on the precise description of behavioural phenotypes but standard methods for classifying the behavioural patterns of mutants are qualitative and imprecise. Together with the Sternberg lab, we have developed the Worm Tracker 2.0 system; a low cost, feature-rich single worm tracker that allows the rapid and consistent quantification of the effects of mutations on behaviour. This system allows us to obtain measurements of a wide range of morphological and behavioural features, including velocity, flex, bending frequency, track amplitude and track wavelength, and we will present data collected from crawling and swimming assays. A standardized phenotyping system makes it possible to compare behavioural data collected by different researchers in different labs and we are using this system to start the generation of a comprehensive C. elegans phenotypic database that could be used to explore the clustering and relative similarities of mutant phenotypes.

Yan, Wenwei Richard et al. (2021) International Worm Meeting "Real-time volumetric whole-animal imaging at cellular resolution with SCAPE microscopy in NeuroPAL worms"

Yemini, Eviatar, Yan, Wenwei Richard, Hillman, Elizabeth M. C., Li, Wenzi, Hobert, Oliver

[International Worm Meeting, 2021]

Recording neural activity at single cell resolution during unrestrained behavior holds tremendous potential for investigating the C. elegans neural code on a global scale. As fluorescent calcium indicators and 3D microscopy speeds have improved, recording from 100's of cells in moving worms has become feasible. However, 2 problems remain 1) being able to extract robust information from individual cells in a moving worm, and 2) knowing the defined identity of each individual tracked cell. Recently, a novel C. elegans strain termed NeuroPAL was developed that labels individual neuronal identities via a stereotypical multi-color fluorescence map. To harness the power of this worm we developed a high-speed multispectral volumetric microscopy platform with sub-cellular resolution, optimized for the NeuroPAL worms. Our SCAPE microscopy-based approach uses a scanning oblique light sheet which provides low phototoxicity and optical sectioning capabilities in a convenient single-objective geometry compatible with common C. elegans sample mounting procedures. The system's multi-laser launch, spectral image splitter and high-speed intensified camera make it possible to rapidly acquire a fully 3D NeuroPAL image in under 0.3 s. These scans can be interspersed with dual channel imaging of GCaMP and RFP with 0.33 x 0.67 x 0.25 microm sampling density over a 310 x 210 microm FEP-covered agarose arena at 13 volumes per second. The resulting data suggests much simpler tracking of uniquely identifiable cells throughout the worm, and analysis of the cellular calcium dynamics during free behavior.

Varol, Erdem et al. (2021) International Worm Meeting "A computational approach linking neuron-specific gene expression with connectivity"

Barrett, Alec, McWhirter, Rebecca, Vidal, Berta, Tavazoie, Saeed, Hobert, Oliver, Weinreb, Alexis, Miller, David, Xu, Chuan, Taylor, Seth, Paninski, Liam, Yemini, Eviatar, Sestan, Nenad, Basavaraju, Manasa, Litwin-Kumar, Ashok, Cros, Cyril, Reilly, Molly, Santpere, Gabriel, Poff, Abigail, Glenwinkel, Lori, Abrams, Alexander, Hammarlund, Marc, Rafi, Ibnul, Varol, Erdem, Oikonomou, Panos, Cook, Steven

[International Worm Meeting, 2021]

There is strong prior evidence for genetic encoding of synaptogenesis, axon guidance, and synaptic pruning in neural circuits. Despite these foundational observations, the transcriptional codes that drive neural connectivity have not been elucidated. The C. elegans nervous system is a particularly useful model for studying the interplay between genetics and connectivity since its wiring diagram is highly stereotyped and uniquely well-defined by electron microscopy. Furthermore, recent evidence in C. elegans has suggested that a unique combination of transcription factors specifies each of the 118 neuron classes[1]. Motivated by evidence for the stereotypy of neural circuits and for the genetic encoding of neural identity, we introduce a novel statistical technique, termed Network Differential Gene expression analysis (nDGE), to test the hypotheses that neuron-specific gene expression dictates connectivity. Specifically, we test the hypothesis that pre-synaptic neural identity is defined by a "key" gene combination whose post-synaptic targets are determined by a "lock" gene combination. For our approach, we utilize neuron-specific gene expression profiles from the CeNGEN project[2] to investigate transcriptional codes for connectivity in the nerve ring[3]. We hypothesize that the expression of specific cell adhesion molecules (CAM) among synaptically-connected neurons drives synaptic maintenance in the mature nervous system. We posit that CAMs mediating synaptic stability would be more highly expressed in synaptically-connected neurons than in adjacent neurons with membrane contacts but no synapses. Thus, for each neuron, we compare the expression of all possible combinations of pairs of CAMs in the neuron and its synaptic partners relative to the neuron and its non-synaptic adjacent neurons. Two independent comparisons are generated, one for presynaptic neurons and a second result for postsynaptic neurons. Our nDGE analysis reveals that specific combinations of CAMs are correlated with connectivity in different subsets of neurons and thus provides a uniquely comprehensive road map for investigating the genetic blueprint for the nerve ring wiring diagram. Open source software of Network Differential Gene Expression (nDGE) is publicly available at https://github.com/cengenproject/connectivity_analysis along with a vignette showcasing the CAM results. 1. Reilly, M. B., Cros, C., Varol, E., Yemini, E., & Hobert, O. (2020). Unique homeobox codes delineate all the neuron classes of C. elegans. Nature, 584(7822), 595-601. 2. Taylor, S. R., Santpere, G., Weinreb, A., Barrett, A., Reilly, M. B., Xu, C. Varol, E., ... & Miller, D. M. (2020). Molecular topography of an entire nervous system. bioRxiv. 3. Cook, S. J.,... & Emmons, S. W. (2019). Whole-animal connectomes of both Caenorhabditis elegans sexes. Nature, 571(7763), 63-71.

Endo, Yuto et al. (2021) International Worm Meeting "Whole-brain imaging with neuronal identities to elucidate the mechanism of a sensory processing"

Miyachi, Miyu, Kimura, Kotaro, Wen, Chentao, Endo, Yuto

[International Worm Meeting, 2021]

A neural circuit in the brain processes sensory stimuli such as an odor to elicit a motor command. To understand the mechanism of such a sensory processing, a number of studies have sought to identify a neural circuit which processes a specific sensory stimulus (functional neural circuit). For identification of a functional neural circuit, it is required to conduct two types of experiments throughout the brain: 1) locating neurons that respond to the stimulus, and 2) identifying their cell types (cell-IDs) to estimate their connectivity based on anatomical connections. Locating stimulus-responsive neurons can be achieved by whole-brain calcium imaging, which record all neuronal activities throughout the brain. However, due to the lack of way to efficiently identifying cell-IDs, it has been difficult to identify a complete functional neural circuit with whole-brain imaging and cell identification. To overcome the limitation, the method for cell identification of all neurons, called NeuroPAL, was developed in C. elegans (Yemini et al., Cell, 2020). Together with the connectome information as well as the records of all neuronal activities, we are now able to estimate a functional neural circuit based on cell-IDs of stimulus-responsive neurons. To identify a complete functional neural circuit by integrating those techniques, we established a whole-brain imaging system combined with efficient cell identification system using NeuroPAL. This system consists of a 3D confocal microscopy, a multi-color imaging system, a microfluidic device, and a pipeline of an image analysis (Chronis et al., Nat. Meth., 2007; Wen et al., eLife, 2021). By using this system, we successfully recorded activities of most head neurons simultaneously with an information of each cell-ID. Now we are identifying a set of odor-responsive neurons using this system with stimulation by a repellent odor 2-nonanone (Kimura et al., J. Neurosci., 2010). This study aims to identify a complete functional neural circuit for specific sensory processing pathways and it may provide clues for how a sensory stimulus is processed in the whole of a neural circuit. We thank Ev Yemini and Oliver Hobert for the help on NeuroPAL.

Yu, Xinwei et al. (2021) International Worm Meeting "Fast deep learning correspondence for neuron tracking and identification in C. elegans using semi-synthetic training"

Sharma, Anuj K., Linderman, Scott W., Randi, Francesco, Yu, Xinwei, Leifer, Andrew M., Creamer, Matthew S.

[International Worm Meeting, 2021]

The nervous system of the nematode C.elegans is well characterized, such that each of the 302 neurons is named and has stereotyped locations across animals(Witvliet et al., 2020). The capability to find corresponding neurons across animals is essential to investigate neural coding and neural dynamics across animals. Despite the worm's overall stereotypy, the variability in neurons' spatial arrangement is sufficient to make predicting neural correspondence a challenge. We present an automated method to track and identify neurons in C. elegans, called "fast Deep Learning Correspondence" or fDLC, based on the transformer network architecture(Vaswani et al., 2017). The transformer has shown great success in natural language processing tasks by modeling the dependencies between words in a sentence. We reasoned this same architecture would be well-suited to extract spatial relationships between neurons in order to build a representation that facilitates finding correspondence to neurons in a template worm. The model is trained once on empirically derived semi-synthetic data and then predicts neural correspondence across held-out real animals via transfer learning. The same pre-trained model both tracks neurons across time and identifies corresponding neurons across individuals. Performance is evaluated against hand-annotated datasets, including NeuroPAL(Yemini et al., 2020). Using only position information, the method achieves 80.0% accuracy at tracking neurons within an individual and 65.8% accuracy at identifying neurons across individuals. Accuracy is even higher on a published dataset(Chaudhary et al., 2021). Accuracy reaches 76.5% when using color information from NeuroPAL. Unlike previous methods, fDLC does not require straightening or transforming the animal into a canonical coordinate system. The method is fast and predicts correspondence in 10 ms making it suitable for future real-time applications. Witvliet D, Mulcahy B, Mitchell JK, Meirovitch Y, Berger DR, Wu Y, et al.Connectomes across development reveal principles of brain maturation in C.elegans. bioRxiv. 2020;doi:10.1101/2020.04.30.066209. Vaswani A, Shazeer N, Parmar N, Uszkoreit J, Jones L, Gomez AN, et al.Attention is All You Need; 2017.Available from:https://arxiv.org/pdf/1706.03762.pdf. Yemini E, Lin A, Nejatbakhsh A, Varol E, Sun R, Mena GE, et al. NeuroPAL: A multicolor atlas for whole-brain neuronal identification in C. Elegans. Cell.2020;doi:10.1016/j.cell.2020.12.012. Chaudhary S, Lee SA, Li Y, Patel DS, Lu H. Graphical-model framework for automated annotation of cell identities in dense cellular images. eLife.2021;10:e60321. doi:10.7554/eLife.60321.

Taylor, Seth R et al. (2021) International Worm Meeting "Molecular topology of an entire nervous system"

Xu, Chuan, Barrett, Alec, Varol, Erdem, Cook, Steven J, Basavaraju, Manasa, Yemini, Eviatar, Hammarlund, Marc, Miller III, David M, Weinreb, Alexis, Abrams, Alexander, Rafi, Ibnul, Santpere, Gabriel, Vidal, Berta, Sestan, Nenad, Oikonomou, Panos, McWhirter, Rebecca, Cros, Cyril, Glenwinkel, Lori, Tavazoie, Saeed, Taylor, Seth R, Hobert, Oliver, Reilly, Molly B, Poff, Abigail

[International Worm Meeting, 2021]

Neurons are the fundamental structural and functional units of the nervous system. Although all neurons share many common features, they also display remarkably diverse morphological and functional characteristics. To uncover the underlying genetic programs that specify individual neuron identities, the CeNGEN consortium produced scRNA-Seq profiles of > 100,000 cells from the L4 stage C. elegans hermaphrodite, including all neuron classes and several non-neuronal cells (e.g., glia, muscle, hypodermis, reproductive tissues). In addition, we identified distinct subclasses for 10 of the 118 anatomically-defined classes. Our results suggest that individual neuron classes can be solely identified by combinatorial expression of specific gene families. For example, each neuron class expresses unique codes of ~23 neuropeptide genes and ~36 neuropeptide receptors thus pointing to an expansive "wireless" signaling network. To demonstrate the utility of this uniquely comprehensive gene expression catalog, we used computational approaches to identify cis-regulatory elements for neuron-specific gene expression. Because our scRNA-Seq data match the single cell resolution of the wiring diagram, we also sought to correlate expression of cell adhesion proteins with neuron-specific fasciculation and connectivity in the nerve ring. We expect that this neuron-specific directory of gene expression will spur investigations of underlying mechanisms that define anatomy, connectivity and function throughout the C. elegans nervous system. These data are available at cengen.org and can be interrogated with the web application CengenApp at cengen.shinyapps.io/CengenApp.

Chaudhary, S. et al. (2019) International Worm Meeting "A strategy for neuron identification in whole-brain videos."

Lee, S., Patel, D., Chaudhary, S., Lu, H.

[International Worm Meeting, 2019]

A current bottleneck in generating meaningful interpretation of the neuron activity dynamics extracted from C. elegans whole-brain videos is determining the identity of neurons. Previous strategies to determine neuron identity include, manual comparison of images with an atlas [1] and polychromatic barcoding of neurons [2]. The former strategy is labor intensive, prone to human bias and error, and requires expert supervision. The latter strategy requires custom designed microscope with several excitation-emission channels that may not be available to all researchers. Here we present an alternative strategy based on neuronal landmarks for assisting researchers in unbiased identification of neurons. We developed a framework [3] that uses known geometrical relationships among positions of neurons and automatically generates a probabilistic identity for each neuron in whole brain stacks. We tested the effect of neuron position variability on prediction accuracy by perturbing ideal data (atlas data) to mimic real images. Next, we established that either spatially distributed landmarks or landmarks in lateral ganglion will provide optimal prediction accuracy. Guided by our analysis, we generated whole-brain imaging strains with CyOFP-labeled neuronal landmarks in the head, which are easily identified in each worm. They establish a coordinate system with which neurons are identified by matching to the atlas. We also validated the automatic prediction accuracy on many real worm data sets. Further, using CyOFP enables 3-channel (GCaMP, RFP and landmarks) imaging with only 2 emission-excitation filter sets that are commonly available, thus freeing a channel for other purposes such as optogenetic manipulations. We use this strategy to compare neurons' responses to chemical stimulation in different individuals, and show how experimental context can be included in the prediction framework as additional constraint. We expect that our algorithm will help realize the full potential of whole-brain imaging techniques by enabling complex experiments and analyses using methods that require neuron identity. References [1] Kato S, et al. Cell. 2015;163: 656-669. doi:10.1016/j.cell.2015.09.034 [2] Personal communication (Ev Yemini and Oliver Hobert) [3] Lafferty J, et al. ICML '01 Proc Eighteenth Int Conf Mach Learn. 2001;8: 282-289. doi:10.1038/nprot.2006.61