Single-Cell and Spatial Omics Technologies in Rice Abiotic Stress Biology: A Methodological Review

Abstract

Abiotic stresses—drought, salinity, extreme temperature, flooding, and heavy-metal toxicity—constrain rice (Oryza sativa L.) yield worldwide, and the cellular programmes underlying them are unevenly distributed across cell types that bulk-tissue assays average together. This review examines, from a methodological standpoint, what single-cell and spatial omics technologies can and cannot establish about rice abiotic stress biology. We first define the modality space: single-cell omics measures RNA, chromatin accessibility, DNA methylation, protein, or metabolite features at the resolution of individual cells or nuclei, whereas spatial omics measures such features while retaining tissue coordinates; the two are complementary rather than interchangeable. We then treat each platform class—droplet-based scRNA-seq, combinatorial-indexing approaches including SPLiT-seq, nuclei-based snRNA-seq and multiome, sequencing-based and imaging-based spatial transcriptomics—under a common template covering measurement principle, the questions each can answer, applicability to rice tissues, dominant biases, and the inferences each cannot support. To make evidence strength comparable across a heterogeneous literature, we apply a four-tier scheme throughout: Tier A, direct rice cell-resolved or spatial evidence with functional or field validation; Tier B, robust rice functional and localization evidence without single-cell data; Tier C, cell-resolved evidence without causal validation; and Tier D, cross-species analogy or reasoned proposal. Applying this scheme shows that the genes with genuine breeding traction in rice—SUB1A, OsHKT1;5, OsHMA3, OsNRAMP5, DRO1—rest on Tier B evidence from classical genetics and field testing, whereas the most cell-resolved rice evidence concentrates in root outer layers and barrier formation at Tier C, and heat and cold stress, despite dominating yield loss, lack rice cell-resolved data almost entirely. We extend the discussion beyond transcriptomics to single-cell DNA methylome profiling, spatial proteomics and metabolomics, and three-dimensional analysis of thick plant tissues, in each case distinguishing demonstrated plant capability from mammalian-only capability, and we assess the expanding role of artificial intelligence in annotation, segmentation, batch correction, integration, and perturbation prediction alongside its documented failure modes. Rice, maize, and wheat are compared to identify transferable methodology. Cell-resolved omics has to date improved biological interpretation and candidate prioritization; demonstrating an incremental breeding advantage from it remains an unmet requirement.