Natural physical, chemical, and biological dynamical systems are often complex, with heterogeneous components interacting in diverse ways. We show that graph neural networks can be designed to jointly learn the interaction rules and the structure of the heterogeneity from data alone. The learned latent structure and dynamics can be used to virtually decompose the complex system which is necessary to parameterize and infer the underlying governing equations. We tested the approach with simulation experiments of moving particles and vector fields that interact with each other. While our current aim is to better understand and validate the approach with simulated data, we anticipate it to become a generally applicable tool to uncover the governing rules underlying complex dynamics observed in nature.

Behavior has molecular, cellular, and circuit determinants. However, because many proteins are broadly expressed, their acute manipulation within defined cells has been difficult. Here, we combined the speed and molecular specificity of pharmacology with the cell type specificity of genetic tools. DART (drugs acutely restricted by tethering) is a technique that rapidly localizes drugs to the surface of defined cells, without prior modification of the native target. We first developed an AMPAR antagonist DART, with validation in cultured neuronal assays, in slices of mouse dorsal striatum, and in behaving mice. In parkinsonian animals, motor deficits were causally attributed to AMPARs in indirect spiny projection neurons (iSPNs) and to excess phasic firing of tonically active interneurons (TANs). Together, iSPNs and TANs (i.e., D2 cells) drove akinesia, whereas movement execution deficits reflected the ratio of AMPARs in D2 versus D1 cells. Finally, we designed a muscarinic antagonist DART in one iteration, demonstrating applicability of the method to diverse targets.

Hunger is a complex behavioural state that elicits intense food seeking and consumption. These behaviours are rapidly recapitulated by activation of starvation-sensitive AGRP neurons, which present an entry point for reverse-engineering neural circuits for hunger. Here we mapped synaptic interactions of AGRP neurons with multiple cell populations in mice and probed the contribution of these distinct circuits to feeding behaviour using optogenetic and pharmacogenetic techniques. An inhibitory circuit with paraventricular hypothalamus (PVH) neurons substantially accounted for acute AGRP neuron-evoked eating, whereas two other prominent circuits were insufficient. Within the PVH, we found that AGRP neurons target and inhibit oxytocin neurons, a small population that is selectively lost in Prader-Willi syndrome, a condition involving insatiable hunger. By developing strategies for evaluating molecularly defined circuits, we show that AGRP neuron suppression of oxytocin neurons is critical for evoked feeding. These experiments reveal a new neural circuit that regulates hunger state and pathways associated with overeating disorders.

Serine hydrolases have diverse intracellular substrates, biological functions, and structural plasticity, and are thus important for biocatalyst design. Amongst serine hydrolases, the recently described ybfF enzyme family are promising novel biocatalysts with an unusual bifurcated substrate-binding cleft and the ability to recognize commercially relevant substrates. We characterized in detail the substrate selectivity of a novel ybfF enzyme from Vibrio cholerae (Vc-ybfF) by using a 21-member library of fluorogenic ester substrates. We assigned the roles of the two substrate-binding clefts in controlling the substrate selectivity and folded stability of Vc-ybfF by comprehensive substitution analysis. The overall substrate preference of Vc-ybfF was for short polar chains, but it retained significant activity with a range of cyclic and extended esters. This broad substrate specificity combined with the substitutional analysis demonstrates that the larger binding cleft controls the substrate specificity of Vc-ybfF. Key selectivity residues (Tyr116, Arg120, Tyr209) are also located at the larger binding pocket and control the substrate specificity profile. In the structure of ybfF the narrower binding cleft contains water molecules prepositioned for hydrolysis, but based on substitution this cleft showed only minimal contribution to catalysis. Instead, the residues surrounding the narrow binding cleft and at the entrance to the binding pocket contributed significantly to the folded stability of Vc-ybfF. The relative contributions of each cleft of the binding pocket to the catalytic activity and folded stability of Vc-ybfF provide a valuable map for designing future biocatalysts based on the ybfF scaffold.

Young birds learn to sing by using auditory feedback to compare their own vocalizations to a memorized or innate song pattern; if they are deafened as juveniles, they will not develop normal songs. The completion of song development is called crystallization. After this stage, song shows little variation in its temporal or spectral properties. However, the mechanisms underlying this stability are largely unknown. Here we present evidence that auditory feedback is actively used in adulthood to maintain the stability of song structure. We found that perturbing auditory feedback during singing in adult zebra finches caused their song to deteriorate slowly. This ’decrystallization’ consisted of a marked loss of the spectral and temporal stereotypy seen in crystallized song, including stuttering, creation, deletion and distortion of song syllables. After normal feedback was restored, these deviations gradually disappeared and the original song was recovered. Thus, adult birds that do not learn new songs nevertheless retain a significant amount of plasticity in the brain.

To integrate changing environmental cues with high spatial and temporal resolution is critical for animals to orient themselves. Drosophila larvae show an effective motor program to navigate away from light sources. How the larval visual circuit processes light stimuli to control navigational decision remains unknown. The larval visual system is composed of two sensory input channels, Rhodopsin5 (Rh5) and Rhodopsin6 (Rh6) expressing photoreceptors (PRs). We here characterize how spatial and temporal information are used to control navigation. Rh6-PRs are required to perceive temporal changes of light intensity during head casts, while Rh5-PRs are required to control behaviors that allow navigation in response to spatial cues. We characterize how distinct behaviors are modulated and identify parallel acting and converging features of the visual circuit. Functional features of the larval visual circuit highlight the principle of how early in a sensory circuit distinct behaviors may be computed by partly overlapping sensory pathways.

Single-molecule localization microscopy (SMLM) has had remarkable success in imaging cellular structures with nanometer resolution, but the need for activating only single isolated emitters limits imaging speed and labeling density. Here, we overcome this major limitation using deep learning. We developed DECODE, a computational tool that can localize single emitters at high density in 3D with highest accuracy for a large range of imaging modalities and conditions. In a public software benchmark competition, it outperformed all other fitters on 12 out of 12 data-sets when comparing both detection accuracy and localization error, often by a substantial margin. DECODE allowed us to take live-cell SMLM data with reduced light exposure in just 3 seconds and to image microtubules at ultra-high labeling density. Packaged for simple installation and use, DECODE will enable many labs to reduce imaging times and increase localization density in SMLM.Competing Interest StatementThe authors have declared no competing interest.

The most sophisticated existing methods to generate 3D isotropic super-resolution (SR) from non-isotropic electron microscopy (EM) are based on learned dictionaries. Unfortunately, none of the existing methods generate practically satisfying results. For 2D natural images, recently developed super-resolution methods that use deep learning have been shown to significantly outperform the previous state of the art. We have adapted one of the most successful architectures (FSRCNN) for 3D super-resolution, and compared its performance to a 3D U-Net architecture that has not been used previously to generate super-resolution. We trained both architectures on artificially downscaled isotropic ground truth from focused ion beam milling scanning EM (FIB-SEM) and tested the performance for various hyperparameter settings.

Our results indicate that both architectures can successfully generate 3D isotropic super-resolution from non-isotropic EM, with the U-Net performing consistently better. We propose several promising directions for practical application.

Optical aberrations hinder fluorescence microscopy of thick samples, reducing image signal, contrast, and resolution. Here we introduce a deep learning-based strategy for aberration compensation, improving image quality without slowing image acquisition, applying additional dose, or introducing more optics. Our method (i) introduces synthetic aberrations to images acquired on the shallow side of image stacks, making them resemble those acquired deeper into the volume and (ii) trains neural networks to reverse the effect of these aberrations. We use simulations and experiments to show that applying the trained ‘de-aberration’ networks outperforms alternative methods, providing restoration on par with adaptive optics techniques; and subsequently apply the networks to diverse datasets captured with confocal, light-sheet, multi-photon, and super-resolution microscopy. In all cases, the improved quality of the restored data facilitates qualitative image inspection and improves downstream image quantitation, including orientational analysis of blood vessels in mouse tissue and improved membrane and nuclear segmentation in C. elegans embryos.

Limited color channels in fluorescence microscopy have long constrained spatial analysis in biological specimens. Here, we introduce cycle Hybridization Chain Reaction (HCR), a method that integrates multicycle DNA barcoding with HCR to overcome this limitation. cycleHCR enables highly multiplexed imaging of RNA and proteins using a unified barcode system. Whole-embryo transcriptomics imaging achieved precise three-dimensional gene expression and cell fate mapping across a specimen depth of ~310 μm. When combined with expansion microscopy, cycleHCR revealed an intricate network of 10 subcellular structures in mouse embryonic fibroblasts. In mouse hippocampal slices, multiplex RNA and protein imaging uncovered complex gene expression gradients and cell-type-specific nuclear structural variations. cycleHCR provides a quantitative framework for elucidating spatial regulation in deep tissue contexts for research and potentially diagnostic applications.

bioRxiv preprint: 10.1101/2024.05.17.594641