Differentiable simulations of optical systems can be combined with deep learning-based reconstruction networks to enable high performance computational imaging via end-to-end (E2E) optimization of both the optical encoder and the deep decoder. This has enabled imaging applications such as 3D localization microscopy, depth estimation, and lensless photography via the optimization of local optical encoders. More challenging computational imaging applications, such as 3D snapshot microscopy which compresses 3D volumes into single 2D images, require a highly non-local optical encoder. We show that existing deep network decoders have a locality bias which prevents the optimization of such highly non-local optical encoders. We address this with a decoder based on a shallow neural network architecture using global kernel Fourier convolutional neural networks (FourierNets). We show that FourierNets surpass existing deep network based decoders at reconstructing photographs captured by the highly non-local DiffuserCam optical encoder. Further, we show that FourierNets enable E2E optimization of highly non-local optical encoders for 3D snapshot microscopy. By combining FourierNets with a large-scale multi-GPU differentiable optical simulation, we are able to optimize non-local optical encoders 170× to 7372× larger than prior state of the art, and demonstrate the potential for ROI-type specific optical encoding with a programmable microscope.

Learning which stimuli (classical conditioning) or which actions (operant conditioning) predict rewards or punishments can improve chances of survival. However, the circuit mechanisms that underlie distinct types of associative learning are still not fully understood. Automated, high-throughput paradigms for studying different types of associative learning, combined with manipulation of specific neurons in freely behaving animals, can help advance this field. The Drosophila melanogaster larva is a tractable model system for studying the circuit basis of behaviour, but many forms of associative learning have not yet been demonstrated in this animal. Here, we developed a high-throughput (i. e. multi-larva) training system that combines real-time behaviour detection of freely moving larvae with targeted opto- and thermogenetic stimulation of tracked animals. Both stimuli are controlled in either open- or closed-loop, and delivered with high temporal and spatial precision. Using this tracker, we show for the first time that Drosophila larvae can perform classical conditioning with no overlap between sensory stimuli (i. e. trace conditioning). We also demonstrate that larvae are capable of operant conditioning by inducing a bend direction preference through optogenetic activation of reward-encoding serotonergic neurons. Our results extend the known associative learning capacities of Drosophila larvae. Our automated training rig will facilitate the study of many different forms of associative learning and the identification of the neural circuits that underpin them.

Deconstructing the mechanism by which the 3D genome encodes genetic information to generate diverse cell types during animal development is a major challenge in biology. The contrast between the elimination of chromatin loops and domains upon Cohesin loss and the lack of downstream gene expression changes at the cell population level instigates intense debates regarding the structure-function relationship between genome organization and gene regulation. Here, by analyzing single cells after acute Cohesin removal with sequencing and spatial genome imaging techniques, we discover that, instead of dictating population-wide gene expression levels, 3D genome topology mediated by Cohesin safeguards long-range gene co-expression correlations in single cells. Notably, Cohesin loss induces gene co-activation and chromatin co-opening between active domains in cis up to tens of megabase apart, far beyond the typical length scale of enhancer-promoter communication. In addition, Cohesin separates Mediator protein hubs, prevents active genes in cis from localizing into shared hubs and blocks intersegment transfer of diverse transcriptional regulators. Together, these results support that spatial organization of the 3D genome orchestrates dynamic long-range gene and chromatin co-regulation in single living cells.

Most animals have compound eyes, with tens to thousands of lenses attached rigidly to the exoskeleton. A natural assumption is that all of these species must resort to moving either their head or their body to actively change their visual input. However, classic anatomy has revealed that flies have muscles poised to move their retinas under the stable lenses of each compound eye. Here we show that Drosophila use their retinal muscles to smoothly track visual motion, which helps to stabilize the retinal image, and also to perform small saccades when viewing a stationary scene. We show that when the retina moves, visual receptive fields shift accordingly, and that even the smallest retinal saccades activate visual neurons. Using a head-fixed behavioural paradigm, we find that Drosophila perform binocular, vergence movements of their retinas-which could enhance depth perception-when crossing gaps, and impairing the physiology of retinal motor neurons alters gap-crossing trajectories during free behaviour. That flies evolved an ability to actuate their retinas suggests that moving the eye independently of the head is broadly paramount for animals. The similarities of smooth and saccadic movements of the Drosophila retina and the vertebrate eye highlight a notable example of convergent evolution.

Electron microscopy (EM) allows for the reconstruction of dense neuronal connectomes but suffers from low throughput, limiting its application to small numbers of reference specimens. We developed a protocol and analysis pipeline using tissue expansion and lattice light-sheet microscopy (ExLLSM) to rapidly reconstruct selected circuits across many samples with single synapse resolution and molecular contrast. We validate this approach in Drosophila, demonstrating that it yields synaptic counts similar to those obtained by EM, can be used to compare counts across sex and experience, and to correlate structural connectivity with functional connectivity. This approach fills a critical methodological gap in studying variability in the structure and function of neural circuits across individuals within and between species.

The regulation of transcription is a complex process that involves binding of transcription factors (TFs) to specific sequences, recruitment of cofactors and chromatin remodelers, assembly of the pre-initiation complex and recruitment of RNA polymerase II. Increasing evidence suggests that TFs are highly dynamic and interact only transiently with DNA. Single molecule microscopy techniques are powerful approaches for tracking individual TF molecules as they diffuse in the nucleus and interact with DNA. Here we employ multifocus microscopy and highly inclined laminated optical sheet microscopy to track TF dynamics in response to perturbations in labile zinc inside cells. We sought to define whether zinc-dependent TFs sense changes in the labile zinc pool by determining whether their dynamics and DNA binding can be modulated by zinc. We used fluorescently tagged versions of the glucocorticoid receptor (GR), with two C4 zinc finger domains, and CCCTC-binding factor (CTCF), with eleven C2H2 zinc finger domains. We found that GR was largely insensitive to perturbations of zinc, whereas CTCF was significantly affected by zinc depletion and its dwell time was affected by zinc elevation. These results indicate that at least some transcription factors are sensitive to zinc dynamics, revealing a potential new layer of transcriptional regulation.

Nitrate (NO3-) uptake and distribution are critical to plant life. Although the upstream regulation of nitrate uptake and downstream responses to nitrate in a variety of cells have been well-studied, it is still not possible to directly visualize the spatial and temporal distribution of nitrate with high resolution at the cellular level. Here, we report a nuclear-localized, genetically encoded biosensor, nlsNitraMeter3.0, for the quantitative visualization of nitrate distribution in Arabidopsis thaliana. The biosensor tracked the spatiotemporal distribution of nitrate along the primary root axis and disruptions by genetic mutation of transport (low nitrate uptake) and assimilation (high nitrate accumulation). The developed biosensor effectively monitors nitrate concentrations at cellular level in real time and spatiotemporal changes during the plant life cycle.

Advances in microscopy hold great promise for allowing quantitative and precise measurement of morphological and molecular phenomena at the single-cell level in bacteria; however, the potential of this approach is ultimately limited by the availability of methods to faithfully segment cells independent of their morphological or optical characteristics. Here, we present Omnipose, a deep neural network image-segmentation algorithm. Unique network outputs such as the gradient of the distance field allow Omnipose to accurately segment cells on which current algorithms, including its predecessor, Cellpose, produce errors. We show that Omnipose achieves unprecedented segmentation performance on mixed bacterial cultures, antibiotic-treated cells and cells of elongated or branched morphology. Furthermore, the benefits of Omnipose extend to non-bacterial subjects, varied imaging modalities and three-dimensional objects. Finally, we demonstrate the utility of Omnipose in the characterization of extreme morphological phenotypes that arise during interbacterial antagonism. Our results distinguish Omnipose as a powerful tool for characterizing diverse and arbitrarily shaped cell types from imaging data.

RNA-guided systems, such as CRISPR-Cas, combine programmable substrate recognition with enzymatic function, a combination that has been used advantageously to develop powerful molecular technologies. Structural studies of these systems have illuminated how the RNA and protein jointly recognize and cleave their substrates, guiding rational engineering for further technology development. Recent work identified a new class of RNA-guided systems, termed OMEGA, which include IscB, the likely ancestor of Cas9, and the nickase IsrB, a homologue of IscB lacking the HNH nuclease domain. IsrB consists of only around 350 amino acids, but its small size is counterbalanced by a relatively large RNA guide (roughly 300-nt ωRNA). Here, we report the cryogenic-electron microscopy structure of Desulfovirgula thermocuniculi IsrB (DtIsrB) in complex with its cognate ωRNA and a target DNA. We find the overall structure of the IsrB protein shares a common scaffold with Cas9. In contrast to Cas9, however, which uses a recognition (REC) lobe to facilitate target selection, IsrB relies on its ωRNA, part of which forms an intricate ternary structure positioned analogously to REC. Structural analyses of IsrB and its ωRNA as well as comparisons to other RNA-guided systems highlight the functional interplay between protein and RNA, advancing our understanding of the biology and evolution of these diverse systems.

Brains must represent the outside world so that animals survive and thrive. In early sensory systems, neural populations have diverse receptive fields structured to detect important features in inputs, yet significant variability has been ignored in classical models of sensory neurons. We model neuronal receptive fields as random, variable samples from parameterized distributions and demonstrate this model in two sensory modalities using data from insect mechanosensors and mammalian primary visual cortex. Our approach leads to a significant theoretical connection between the foundational concepts of receptive fields and random features, a leading theory for understanding artificial neural networks. The modeled neurons perform a randomized wavelet transform on inputs, which removes high frequency noise and boosts the signal. Further, these random feature neurons enable learning from fewer training samples and with smaller networks in artificial tasks. This structured random model of receptive fields provides a unifying, mathematically tractable framework to understand sensory encodings across both spatial and temporal domains.