Deep learning-based high-throughput phenotyping can drive future discoveries in plant reproductive biology.

KEY MESSAGE: Advances in deep learning are providing a powerful set of image analysis tools that are readily accessible for high-throughput phenotyping applications in plant reproductive biology. High-throughput phenotyping systems are becoming critical for answering biological questions on a large scale. These systems have historically relied on traditional computer vision techniques. However, neural networks and specifically deep learning are rapidly becoming more powerful and easier to implement. Here, we examine how deep learning can drive phenotyping systems and be used to answer fundamental questions in reproductive biology. We describe previous applications of deep learning in the plant sciences, provide general recommendations for applying these methods to the study of plant reproduction, and present a case study in maize ear phenotyping. Finally, we highlight several examples where deep learning has enabled research that was previously out of reach and discuss the future outlook of these methods.

A Maize Male Gametophyte-Specific Gene Encodes ZmLARP6c1, a Potential RNA-Binding Protein Required for Competitive Pollen Tube Growth.

Members of the La-related protein family (LARPs) contain a conserved La module, which has been associated with RNA-binding activity. Expression of the maize gene GRMZM2G323499/Zm00001d018613, a member of the LARP family, is highly specific to pollen, based on both transcriptomic and proteomic assays. This suggests a pollen-specific RNA regulatory function for the protein, designated ZmLARP6c1 based on sequence similarity to the LARP6 subfamily in Arabidopsis. To test this hypothesis, a Ds-GFP transposable element insertion in the ZmLarp6c1 gene (tdsgR82C05) was obtained from the Dooner/Du mutant collection. Sequencing confirmed that the Ds-GFP insertion is in an exon, and thus likely interferes with ZmLARP6c1 function. Tracking inheritance of the insertion via its endosperm-expressed GFP indicated that the mutation was associated with reduced transmission from a heterozygous plant when crossed as a male (ranging from 0.5 to 26.5% transmission), but not as a female. Furthermore, this transmission defect was significantly alleviated when less pollen was applied to the silk, reducing competition between mutant and wild-type pollen. Pollen grain diameter measurements and nuclei counts showed no significant differences between wild-type and mutant pollen. However, in vitro, mutant pollen tubes were significantly shorter than those from sibling wild-type plants, and also displayed altered germination dynamics. These results are consistent with the idea that ZmLARP6c1 provides an important regulatory function during the highly competitive progamic phase of male gametophyte development following arrival of the pollen grain on the silk. The conditional, competitive nature of the Zmlarp6c1::Ds male sterility phenotype (i.e., reduced ability to produce progeny seed) points toward new possibilities for genetic control of parentage in crop production.

A cost-effective maize ear phenotyping platform enables rapid categorization and quantification of kernels.

High-throughput phenotyping systems are powerful, dramatically changing our ability to document, measure, and detect biological phenomena. Here, we describe a cost-effective combination of a custom-built imaging platform and deep-learning-based computer vision pipeline. A minimal version of the maize (Zea mays) ear scanner was built with low-cost and readily available parts. The scanner rotates a maize ear while a digital camera captures a video of the surface of the ear, which is then digitally flattened into a two-dimensional projection. Segregating GFP and anthocyanin kernel phenotypes are clearly distinguishable in ear projections and can be manually annotated and analyzed using image analysis software. Increased throughput was attained by designing and implementing an automated kernel counting system using transfer learning and a deep learning object detection model. The computer vision model was able to rapidly assess over 390 000 kernels, identifying male-specific transmission defects across a wide range of GFP-marked mutant alleles. This includes a previously undescribed defect putatively associated with mutation of Zm00001d002824, a gene predicted to encode a vacuolar processing enzyme. Thus, by using this system, the quantification of transmission data and other ear and kernel phenotypes can be accelerated and scaled to generate large datasets for robust analyses.

High expression in maize pollen correlates with genetic contributions to pollen fitness as well as with coordinated transcription from neighboring transposable elements.

In flowering plants, gene expression in the haploid male gametophyte (pollen) is essential for sperm delivery and double fertilization. Pollen also undergoes dynamic epigenetic regulation of expression from transposable elements (TEs), but how this process interacts with gene expression is not clearly understood. To explore relationships among these processes, we quantified transcript levels in four male reproductive stages of maize (tassel primordia, microspores, mature pollen, and sperm cells) via RNA-seq. We found that, in contrast with vegetative cell-limited TE expression in Arabidopsis pollen, TE transcripts in maize accumulate as early as the microspore stage and are also present in sperm cells. Intriguingly, coordinate expression was observed between highly expressed protein-coding genes and their neighboring TEs, specifically in mature pollen and sperm cells. To investigate a potential relationship between elevated gene transcript level and pollen function, we measured the fitness cost (male-specific transmission defect) of GFP-tagged coding sequence insertion mutations in over 50 genes identified as highly expressed in the pollen vegetative cell, sperm cell, or seedling (as a sporophytic control). Insertions in seedling genes or sperm cell genes (with one exception) exhibited no difference from the expected 1:1 transmission ratio. In contrast, insertions in over 20% of vegetative cell genes were associated with significant reductions in fitness, showing a positive correlation of transcript level with non-Mendelian segregation when mutant. Insertions in maize gamete expressed2 (Zm gex2), the sole sperm cell gene with measured contributions to fitness, also triggered seed defects when crossed as a male, indicating a conserved role in double fertilization, given the similar phenotype previously demonstrated for the Arabidopsis ortholog GEX2. Overall, our study demonstrates a developmentally programmed and coordinated transcriptional activation of TEs and genes in pollen, and further identifies maize pollen as a model in which transcriptomic data have predictive value for quantitative phenotypes.