High-throughput image segmentation and machine learning approaches in the plant sciences across multiple scales.

Agriculture has benefited greatly from the rise of big data and high-performance computing. The acquisition and analysis of data across biological scales have resulted in strategies modeling inter- actions between plant genotype and environment, models of root architecture that provide insight into resource utilization, and the elucidation of cell-to-cell communication mechanisms that are instrumental in plant development. Image segmentation and machine learning approaches for interpreting plant image data are among many of the computational methodologies that have evolved to address challenging agricultural and biological problems. These approaches have led to contributions such as the accelerated identification of gene that modulate stress responses in plants and automated high-throughput phenotyping for early detection of plant diseases. The continued acquisition of high throughput imaging across multiple biological scales provides opportunities to further push the boundaries of our understandings quicker than ever before. In this review, we explore the current state of the art methodologies in plant image segmentation and machine learning at the agricultural, organ, and cellular scales in plants. We show how the methodologies for segmentation and classification differ due to the diversity of physical characteristics found at these different scales. We also discuss the hardware technologies most commonly used at these different scales, the types of quantitative metrics that can be extracted from these images, and how the biological mechanisms by which plants respond to abiotic/biotic stresses or genotypic modifications can be extracted from these approaches.

MAGIC: Live imaging of cellular division in plant seedlings using lightsheet microscopy.

Imaging technologies have been used to understand plant genetic and developmental processes, from the dynamics of gene expression to tissue and organ morphogenesis. Although the field has advanced incredibly in recent years, gaps remain in identifying fine and dynamic spatiotemporal intervals of target processes, such as changes to gene expression in response to abiotic stresses. Lightsheet microscopy is a valuable tool for such studies due to its ability to perform long-term imaging at fine intervals of time and at low photo-toxicity of live vertically oriented seedlings. In this chapter, we describe a detailed method for preparing and imaging Arabidopsis thaliana seedlings for lightsheet microscopy via a Multi-Sample Imaging Growth Chamber (MAGIC), which allows simultaneous imaging of at least four samples. This method opens new avenues for acquiring imaging data at a high temporal resolution, which can be eventually probed to identify key regulatory time points and any spatial dependencies of target developmental processes.

BioVision Tracker: A semi-automated image analysis software for spatiotemporal gene expression tracking in Arabidopsis thaliana.

Fluorescence microscopy can produce large quantities of data that reveal the spatiotemporal behavior of gene expression at the cellular level in plants. Automated or semi-automated image analysis methods are required to extract data from these images. These data are helpful in revealing spatial and/or temporal-dependent processes that influence development in the meristematic region of plant roots. Tracking spatiotemporal gene expression in the meristem requires the processing of multiple microscopy imaging channels (one channel used to image root geometry which serves as a reference for relating locations within the root, and one or more channels used to image fluorescent gene expression signals). Many automated image analysis methods rely on the staining of cell walls with fluorescent dyes to capture cellular geometry and overall root geometry. However, in long time-course imaging experiments, dyes may fade which hinders spatial assessment in image analysis. Here, we describe a procedure for analyzing 3D microscopy images to track spatiotemporal gene expression signals using the MATLAB-based BioVision Tracker software. This software requires either a fluorescence image or a brightfield image to analyze root geometry and a fluorescence image to capture and track temporal changes in gene expression.

Stem-cell-ubiquitous genes spatiotemporally coordinate division through regulation of stem-cell-specific gene networks.

Stem cells are responsible for generating all of the differentiated cells, tissues, and organs in a multicellular organism and, thus, play a crucial role in cell renewal, regeneration, and organization. A number of stem cell type-specific genes have a known role in stem cell maintenance, identity, and/or division. Yet, how genes expressed across different stem cell types, referred to here as stem-cell-ubiquitous genes, contribute to stem cell regulation is less understood. Here, we find that, in the Arabidopsis root, a stem-cell-ubiquitous gene, TESMIN-LIKE CXC2 (TCX2), controls stem cell division by regulating stem cell-type specific networks. Development of a mathematical model of TCX2 expression allows us to show that TCX2 orchestrates the coordinated division of different stem cell types. Our results highlight that genes expressed across different stem cell types ensure cross-communication among cells, allowing them to divide and develop harmonically together.

Automated Imaging, Tracking, and Analytics Pipeline for Differentiating Environmental Effects on Root Meristematic Cell Division.

Exposure of plants to abiotic stresses, whether individually or in combination, triggers dynamic changes to gene regulation. These responses induce distinct changes in phenotypic characteristics, enabling the plant to adapt to changing environments. For example, iron deficiency and heat stress have been shown to alter root development by reducing primary root growth and reducing cell proliferation, respectively. Currently, identifying the dynamic temporal coordination of genetic responses to combined abiotic stresses remains a bottleneck. This is, in part, due to an inability to isolate specific intervals in developmental time where differential activity in plant stress responses plays a critical role. Here, we observed that iron deficiency, in combination with temporary heat stress, suppresses the expression of iron deficiency-responsive pPYE::LUC (POPEYE::luciferase) and pBTS::LUC (BRUTUS::luciferase) reporter genes. Moreover, root growth was suppressed less under combined iron deficiency and heat stress than under either single stress condition. To further explore the interaction between pathways, we also created a computer vision pipeline to extract, analyze, and compare high-dimensional dynamic spatial and temporal cellular data in response to heat and iron deficiency stress conditions at high temporal resolution. Specifically, we used fluorescence light sheet microscopy to image Arabidopsis thaliana roots expressing CYCB1;1:GFP, a marker for cell entry into mitosis, every 20 min for 24 h exposed to either iron sufficiency, iron deficiency, heat stress, or combined iron deficiency and heat stress. Our pipeline extracted spatiotemporal metrics from these time-course data. These metrics showed that the persistency and timing of CYCB1;1:GFP signal were uniquely different under combined iron deficiency and heat stress conditions versus the single stress conditions. These metrics also indicated that the spatiotemporal characteristics of the CYCB1;1:GFP signal under combined stress were more dissimilar to the control response than the response seen under iron deficiency for the majority of the 24-h experiment. Moreover, the combined stress response was less dissimilar to the control than the response seen under heat stress. This indicated that pathways activated when the plant is exposed to both iron deficiency and heat stress affected CYCB1;1:GFP spatiotemporal function antagonistically.

Tracking Gene Expression via Light Sheet Microscopy and Computer Vision in Living Organisms.

Automated tracking of spatiotemporal gene expression using in vivo microscopy images have given great insight into understanding developmental processes in multicellular organisms. Many existing analysis tools rely on the fluorescent tagging of cell wall or cell nuclei localized proteins to assess position, orientation, and overall shape of an organism; information necessary for determining locations of gene expression activity. Particularly in plants, organism lines that have fluorescent tags can take months to develop, which can be time consuming and costly. We propose an automated solution for analyzing spatial characteristics of gene expression without the necessity of fluorescent tagged cell walls or cell nuclei. Our solution indicates, segments, and tracks gene expression using a fluorescent imaging channel of a light sheet microscope while determining gene expression location within an organism from a Brightfield (non-fluorescent) imaging channel. We use the images obtained from the Arabidopsis thaliana root as a proof of concept for our solution by studying the effects of heat shock stress on CYCLIN B1 protein production.