Single-Cell Genomics Acts as a Microscope on Plant Biology

Single-cell genomic atlases provide scientists with a new molecular lens for identifying the regulatory programs that enable plants to tolerate environmental stress.

The Molecular Lens: A Microscope for the Genome

Water scarcity is projected to intensify across many major agricultural regions in the coming decades. According to the World Resources Institute, by 2040 more than one third of global cropland is expected to face high or extremely high water stress under current use scenarios.¹ Increased exposure to water limitation threatens crop yields and the stability of food systems.

To better understand how plants withstand environmental stress, scientists have traditionally used bulk RNA sequencing. This, however, provides only an average snapshot of a certain tissue (e.g. leaf, root, or flower), effectively blending the distinct signals of thousands of different cells into a single image. It is the molecular equivalent of trying to understand a symphony by listening to every instrument at once. Single-cell genomic atlases provide a new molecular lens for identifying the regulatory programs that enable plants to tolerate environmental stress.

Using high-throughput single-cell multi-omics platforms, scientists can now treat the genome like a microscope, peering into individual cells to see exactly which genes are turned on and off. This advance is especially transformative in plants, which are rooted in place and cannot escape environmental stress; instead, each cell must continuously sense and respond to fluctuating conditions such as drought, nutrient limitation, and pathogen attack. This is not just a marginal improvement—it is a data revolution.

Cracking the Code with AI

At the core of plant cellular responses are regulatory DNA elements that determine when and where genes are activated. Because these non-coding sequences can lie far from the genes they control, identifying them from sequence alone has been challenging. Deep learning models trained on single-cell chromatin accessibility and gene expression data now infer this regulatory logic directly from DNA sequences.^2-4 Much like a language model learns grammar from text, these systems learn the “syntax” of regulatory DNA. Recent work in plants demonstrates that such models can capture combinatorial transcription factor syntax and predict enhancer activity from sequence alone, indicating that plant regulatory elements follow structured, learnable rules.⁵

In animal systems, this approach has progressed beyond prediction: Deep learning-guided design has enabled the de novo creation of synthetic enhancers that drive cell-type-specific gene expression.⁶ In plants, rationally engineered synthetic promoters and genetic circuits have achieved tunable and combinatorial expression programs in Arabidopsis thaliana, supported by large-scale chromatin accessibility maps that link sequence features to cell-type-specific regulation.^7-11 Together, these advances suggest that integrating single-cell atlases with sequence-based modeling could enable the design of stress-responsive regulatory elements tailored to specific plant cell types. By combining genomic atlases with machine learning, plant science is beginning to shift from reactive stress response toward predictive and precision-guided crop improvement.

Anticipating the Unseen: Drought and Recovery

The ability to alter stress-responsive regulatory elements in plants is particularly urgent in the context of water scarcity. A central goal of predictive plant genomics is to understand how plants sense and respond to environmental change at single-cell resolution across the entire organism. In the reference plant A. thaliana, rehydration after drought triggers rapid, cell-state-specific immune activation, revealing that recovery involves distinct transcriptional programs that bulk profiling would obscure.¹² Identifying the regulatory sequences underlying these recovery programs is essential for designing resilience-associated regulatory elements.

We are now extending this work to Sorghum bicolor, a bioenergy feedstock, by subjecting plants to controlled drought and recovery treatments. Using artificial intelligence (AI)-driven predictions, we are working with synthetic biologists and computer scientists to identify the regulatory sequences that drive these transient recovery responses, moving beyond descriptive characterization of stress toward predicting a plant’s capacity for survival and resilience. Such predictive capability is essential for strengthening global food security, particularly because drought is S. bicolor’s dominant abiotic constraint, and its global production is concentrated in regions that experience multiple months of annual water scarcity.¹³

By identifying the regulatory sequences that control stress- and recovery-responsive programs, single-cell approaches begin to shift plant biology from mapping responses to designing interventions—from descriptive atlases toward predictive insight. This transition marks a turning point from understanding resilience to engineering it.

Building the Future of Agriculture

The transition from bulk tissue analysis to single-cell resolution is a paradigm shift for biotechnology. We are no longer throwing darts in the dark; we have a molecular roadmap for crop resilience. Our single-cell atlas of plants recovering from drought reveals that recovery-specific genes are activated within minutes of rehydration, indicating that plants are primed during drought for a rapid molecular switch back to growth.¹² This dataset now enables us to identify precise regulatory switches that determine when and where stress programs are deployed within the leaf.

Rather than broadly activating costly preventive immune pathways, such as drought recovery-induced immunity, across the entire tissue, we could theoretically instruct plants to restrict energetically costly immune activation to specific cell types, while allowing other cells, particularly those essential for photosynthesis and growth restoration, to prioritize productivity. In this way, we would not override evolutionarily conserved strategies such as initiating immunity at a vulnerable post-drought stage when pathogen risk is high but instead refine them to optimize both protection and recovery. Furthermore, emerging single-cell data reveal heterogeneity even within the same cell type, including distinct mesophyll subpopulations, the primary photosynthetic cells of the leaf. By understanding their spatial organization within a tissue, we could selectively activate defense programs in certain subsets of cells while others focus on resuming growth. Such spatially and temporally precise engineering would enable crops to maintain resilience without sacrificing yield, advancing the development of environmentally adaptable plants for future climates.

As we identify the exact genetic switches that fail under stress, we move closer to an era where we can program plants to thrive in the world’s most marginal soils, but considerable challenges remain. Discoveries made in reference plants do not automatically translate to crops such as sorghum, maize, or wheat, where genome complexity and environmental interactions introduce additional layers of regulation. Field validation is slow and resource-intensive, and the regulatory approval processes for engineered crops vary widely across regions. Moreover, predictive models are only as strong as the data used to train them, and biological systems often defy simple rules. Progress will require sustained experimentation and collaboration across disciplines. Still, integrating single-cell genomics with sequence-based modeling provides a clearer roadmap than has ever before been available.

The future of agriculture will depend not on a single breakthrough, but on the steady integration of genomic insight, optimization of AI-based predictive models trained on these data, and careful field validation. By learning the regulatory language plants use to survive stress, we take a deliberate step toward crops that are better prepared for the environments they face.