Harnessing Deep Learning to Analyze Cryptic Morphological Variability of Marchantia polymorpha.

Characterizing phenotypes is a fundamental aspect of biological sciences, although it can be challenging due to various factors. For instance, the liverwort Marchantia polymorpha is a model system for plant biology and exhibits morphological variability, making it difficult to identify and quantify distinct phenotypic features using objective measures. To address this issue, we utilized a deep-learning-based image classifier that can handle plant images directly without manual extraction of phenotypic features and analyzed pictures of M. polymorpha. This dioicous plant species exhibits morphological differences between male and female wild accessions at an early stage of gemmaling growth, although it remains elusive whether the differences are attributable to sex chromosomes. To isolate the effects of sex chromosomes from autosomal polymorphisms, we established a male and female set of recombinant inbred lines (RILs) from a set of male and female wild accessions. We then trained deep learning models to classify the sexes of the RILs and the wild accessions. Our results showed that the trained classifiers accurately classified male and female gemmalings of wild accessions in the first week of growth, confirming the intuition of researchers in a reproducible and objective manner. In contrast, the RILs were less distinguishable, indicating that the differences between the parental wild accessions arose from autosomal variations. Furthermore, we validated our trained models by an 'eXplainable AI' technique that highlights image regions relevant to the classification. Our findings demonstrate that the classifier-based approach provides a powerful tool for analyzing plant species that lack standardized phenotyping metrics.

In-Depth Quantification of Cell Division and Elongation Dynamics at the Tip of Growing Arabidopsis Roots Using 4D Microscopy, AI-Assisted Image Processing and Data Sonification.

One of the fundamental questions in plant developmental biology is how cell proliferation and cell expansion coordinately determine organ growth and morphology. An amenable system to address this question is the Arabidopsis root tip, where cell proliferation and elongation occur in spatially separated domains, and cell morphologies can easily be observed using a confocal microscope. While past studies revealed numerous elements of root growth regulation including gene regulatory networks, hormone transport and signaling, cell mechanics and environmental perception, how cells divide and elongate under possible constraints from cell lineages and neighboring cell files has not been analyzed quantitatively. This is mainly due to the technical difficulties in capturing cell division and elongation dynamics at the tip of growing roots, as well as an extremely labor-intensive task of tracing the lineages of frequently dividing cells. Here, we developed a motion-tracking confocal microscope and an Artificial Intelligence (AI)-assisted image-processing pipeline that enables semi-automated quantification of cell division and elongation dynamics at the tip of vertically growing Arabidopsis roots. We also implemented a data sonification tool that facilitates human recognition of cell division synchrony. Using these tools, we revealed previously unnoted lineage-constrained dynamics of cell division and elongation, and their contribution to the root zonation boundaries.

A novel red fluorescence dopamine biosensor selectively detects dopamine in the presence of norepinephrine in vitro.

Dopamine (DA) and norepinephrine (NE) are pivotal neuromodulators that regulate a broad range of brain functions, often in concert. Despite their physiological importance, untangling the relationship between DA and NE in the fine control of output function is currently challenging, primarily due to a lack of techniques to allow the observation of spatiotemporal dynamics with sufficiently high selectivity. Although genetically encoded fluorescent biosensors have been developed to detect DA, their poor selectivity prevents distinguishing DA from NE. Here, we report the development of a red fluorescent genetically encoded GPCR (G protein-coupled receptor)-activation reporter for DA termed 'R-GenGAR-DA'. More specifically, a circular permutated red fluorescent protein (cpmApple) was replaced by the third intracellular loop of human DA receptor D1 (DRD1) followed by the screening of mutants within the linkers between DRD1 and cpmApple. We developed two variants: R-GenGAR-DA1.1, which brightened following DA stimulation, and R-GenGAR-DA1.2, which dimmed. R-GenGAR-DA1.2 demonstrated a reasonable dynamic range (ΔF/F0 = - 43%), DA affinity (EC50 = 0.92 µM) and high selectivity for DA over NE (66-fold) in HeLa cells. Taking advantage of the high selectivity of R-GenGAR-DA1.2, we monitored DA in presence of NE using dual-color fluorescence live imaging, combined with the green-NE biosensor GRABNE1m, which has high selectivity for NE over DA (> 350-fold) in HeLa cells and hippocampal neurons grown from primary culture. Thus, this is a first step toward the multiplex imaging of these neurotransmitters in, for example, freely moving animals, which will provide new opportunities to advance our understanding of the high spatiotemporal dynamics of DA and NE in normal and abnormal brain function.

Reconstruction of global regulatory network from signaling to cellular functions using phosphoproteomic data.

Cellular signaling regulates various cellular functions via protein phosphorylation. Phosphoproteomic data potentially include information for a global regulatory network from signaling to cellular functions, but a procedure to reconstruct this network using such data has yet to be established. In this paper, we provide a procedure to reconstruct a global regulatory network from signaling to cellular functions from phosphoproteomic data by integrating prior knowledge of cellular functions and inference of the kinase-substrate relationships (KSRs). We used phosphoproteomic data from insulin-stimulated Fao hepatoma cells and identified protein phosphorylation regulated by insulin specifically over-represented in cellular functions in the KEGG database. We inferred kinases for protein phosphorylation by KSRs, and connected the kinases in the insulin signaling layer to the phosphorylated proteins in the cellular functions, revealing that the insulin signal is selectively transmitted via the Pi3k-Akt and Erk signaling pathways to cellular adhesions and RNA maturation, respectively. Thus, we provide a method to reconstruct global regulatory network from signaling to cellular functions based on phosphoproteomic data.

Trans-omic Analysis Reveals Selective Responses to Induced and Basal Insulin across Signaling, Transcriptional, and Metabolic Networks.

The concentrations of insulin selectively regulate multiple cellular functions. To understand how insulin concentrations are interpreted by cells, we constructed a trans-omic network of insulin action in FAO hepatoma cells using transcriptomic data, western blotting analysis of signaling proteins, and metabolomic data. By integrating sensitivity into the trans-omic network, we identified the selective trans-omic networks stimulated by high and low doses of insulin, denoted as induced and basal insulin signals, respectively. The induced insulin signal was selectively transmitted through the pathway involving Erk to an increase in the expression of immediate-early and upregulated genes, whereas the basal insulin signal was selectively transmitted through a pathway involving Akt and an increase of Foxo phosphorylation and a reduction of downregulated gene expression. We validated the selective trans-omic network in vivo by analysis of the insulin-clamped rat liver. This integrated analysis enabled molecular insight into how liver cells interpret physiological insulin signals to regulate cellular functions.

Reconstruction of insulin signal flow from phosphoproteome and metabolome data.

Cellular homeostasis is regulated by signals through multiple molecular networks that include protein phosphorylation and metabolites. However, where and when the signal flows through a network and regulates homeostasis has not been explored. We have developed a reconstruction method for the signal flow based on time-course phosphoproteome and metabolome data, using multiple databases, and have applied it to acute action of insulin, an important hormone for metabolic homeostasis. An insulin signal flows through a network, through signaling pathways that involve 13 protein kinases, 26 phosphorylated metabolic enzymes, and 35 allosteric effectors, resulting in quantitative changes in 44 metabolites. Analysis of the network reveals that insulin induces phosphorylation and activation of liver-type phosphofructokinase 1, thereby controlling a key reaction in glycolysis. We thus provide a versatile method of reconstruction of signal flow through the network using phosphoproteome and metabolome data.