Single‐cell insights into plant growth, adaptation, and evolution

ABSTRACT

The cellular heterogeneity associated with plant form and function is often overlooked in traditional bulk‐tissue analyses. Emerging single‐cell technologies provide new opportunities for dissecting this complexity at higher resolution. In this review, we summarize how single‐cell multi‐omics approaches, integrating transcriptomics, epigenomics, and spatial omics, can be used to characterize the regulatory landscapes associated with crop development, stress responses, and evolution. We discuss the application of these technologies across the crop life cycle, with a focus on identifying cell‐type‐specific programs related to key agronomic traits and tracing developmental trajectories. Furthermore, we describe how single‐cell tools contribute to the analysis of plant responses to abiotic and biotic stresses and provide insights into the evolution of specialized cell types. We also discuss current challenges, including technical difficulties in protoplast isolation, the computational integration of multi‐modal data, and scalability across diverse species. Finally, we outline potential future directions for combining machine learning and spatial transcriptomics to connect cellular‐level observations with tissue‐level functions, thereby supporting advances in functional genomics, precision breeding, and crop improvement.

Single‐cell multi‐omics is transforming crop research by revealing how distinct cell types shape development, stress responses, and evolution. This review highlights advances in transcriptomic, epigenomic, and spatial analyses across the crop life cycle, discusses current technical challenges, and outlines opportunities to link cellular regulation with precision breeding and crop improvement.

INTRODUCTION

Achieving global food security under climate change and population growth requires continued improvement of crop performance. For decades, traditional breeding and quantitative genetics, often supported by molecular biology, have contributed substantially to gains in yield, quality, and stress adaptation. However, these approaches have typically relied on bulk‐tissue analyses, which average molecular signals from diverse cell types and may overlook cell‐type‐specific regulatory mechanisms associated with complex agronomic traits. Plants, as sessile organisms, consist of highly specialized tissues and cells, each with distinct functions and responses to developmental and environmental cues. A better understanding of this cellular heterogeneity is therefore important for advancing plant biology and crop improvement. The development of single‐cell technologies has provided new opportunities to address this limitation by enabling the analysis of gene expression and regulatory landscapes at the level of individual cells.

In recent years, single‐cell multi‐omics approaches have been increasingly applied in major crops, leading to the construction of cellular atlases for key developmental stages and organs in species such as rice (Wang et al., 2025e), maize (Marand et al., 2021), and soybean (Zhang et al., 2025a). These studies have also expanded to comparative and integrative analyses, including unified atlases across vascular plants (Xue et al., 2025) and pan‐species transcriptomic datasets (Guillotin et al., 2023), which have helped characterize conserved and divergent cellular features. These approaches have been used to investigate gene regulatory networks associated with important aspects of plant growth and development, including somatic embryogenesis (Tang et al., 2025), seed development (Pelletier et al., 2025), and grain development (Li et al., 2025e). In addition, single‐cell multi‐omics have contributed to the analysis of cell‐type‐specific responses to environmental stresses by identifying molecular features in particular cell types, such as root hairs under salt stress (Liu et al., 2025) and root cortex cells under heat stress (Wang et al., 2025b).

These advances have strengthened the connection between molecular research and agricultural applications. By linking cell‐type‐specific gene expression and chromatin accessibility with agronomic traits (Wang et al., 2025e) and identifying genes associated with yield and quality in specific cell types (Li et al., 2025e), single‐cell omics offer useful approaches for crop genetic improvement. This review summarizes these recent advances and their relevance to crop breeding. We first provide an overview of single‐cell multi‐omics technologies currently applied in plants. We then discuss the major applications of these technologies in studies of plant growth and development, followed by their use in understanding cell‐type‐specific responses to abiotic and biotic stresses. Finally, we discuss future challenges and opportunities related to the integration of single‐cell biology into modern crop improvement strategies.

SINGLE‐CELL SEQUENCING TECHNOLOGIES IN PLANT

Single‐cell RNA sequencing technologies in plant

The development of single‐cell transcriptomics has created new opportunities to examine the cellular heterogeneity present in complex plant tissues. Traditional bulk transcriptomics, while informative, averages gene expression signals across cell types, which can obscure cell‐type‐specific responses and reduce the detectability of rare populations (Li and Wang, 2021; Baul et al., 2024). Single‐cell RNA sequencing (scRNA‐seq) addresses this limitation by capturing transcriptomes at the level of individual cells (Figure 1A), making it possible to classify cell types, identify previously unrecognized cell states, and reconstruct developmental trajectories (Shaw et al., 2021; Nolan and Shahan, 2023; Zhang et al., 2023a). At the same time, these approaches involve technical and interpretive considerations that should be taken into account when drawing biological conclusions.

Figure 1.

Overview of single‐cell and spatial multi‐omics technologies

(A) Single‐cell transcriptomics (scRNA‐seq/snRNA‐seq): Tissue is dissociated into cells or nuclei for barcoding and NGS, producing gene expression matrices and UMAP clusters. (B) Single‐cell epigenomics (scATAC‐seq): This method uses Tn5 tagmentation to map cell‐type‐specific chromatin accessibility and regulatory networks. (C) Spatial transcriptomics: Gene expression is mapped within the tissue context using sequencing‐based (Visium/Stereo‐seq) or imaging‐based (seqFISH and MERFISH) technologies. Figure created with Biorender.com.

A major methodological challenge in plant single‐cell studies is the presence of the rigid cell wall, which usually needs to be enzymatically removed to release individual cells, or protoplasts, for high‐throughput capture (Birnbaum et al., 2003; Grones et al., 2024). However, the protoplasting step introduces several technical considerations. The enzymatic digestion process can induce a transcriptional stress response and thereby alter the native gene expression profile before capture (Van den Brink et al., 2017; Tenorio Berrío and Dubois, 2024). This issue is especially relevant in studies of biotic and abiotic stress, where the stress signature introduced during sample preparation may overlap with the biological response under investigation (Tenorio Berrío et al., 2022). Furthermore, the process can introduce sampling bias, as certain cell types, such as the stele or those with more resistant cell walls, may be underrepresented in the final protoplast population (Denyer et al., 2019; Jean‐Baptiste et al., 2019; Ryu et al., 2019). The large and highly variable size of plant cells, which can exceed 100 μm, may also limit compatibility with some microfluidic capture systems (Shaw et al., 2021). Together, these factors may lead to incomplete representation of tissue composition and indicate the importance of careful validation of computational cell‐type annotations.

To reduce the limitations associated with protoplasting, single‐nucleus RNA‐seq (snRNA‐seq) has become a widely used alternative (Habib et al., 2017; Farmer et al., 2021; Cervantes‐Pérez et al., 2022). In this method, nuclei are mechanically isolated from fresh or frozen tissues. This approach largely avoids the transcriptional stress response associated with enzymatic digestion and is therefore well suited to tissues that are difficult to protoplast, as well as to the analysis of transient transcriptional states in response to environmental stimuli (Tian et al., 2020; Tenorio Berrío and Dubois, 2024). However, recent work suggests that nuclear isolation during snRNA‐seq library preparation may still induce a mild protoplasting‐like response, which should be considered in stress‐related experimental design (Ming et al., 2025).

More fundamentally, snRNA‐seq involves trade‐offs that may influence biological interpretation. Compared with protoplast‐based scRNA‐seq, snRNA‐seq typically captures fewer unique transcripts per cell or nucleus (Bakken et al., 2018; Farmer et al., 2021). In addition, snRNA‐seq preferentially captures nuclear transcripts enriched in unspliced pre‐mRNAs and does not fully represent the pool of cytoplasmic mature mRNAs (Gaidatzis et al., 2015). Although comparative studies have reported high overall correlation between nuclear and whole‐cell profiles for cell‐type clustering (Farmer et al., 2021; Guillotin et al., 2023), these differences may affect the interpretation of gene activity, particularly for rapidly regulated transcripts or transcripts with strong nuclear retention. These observations suggest that functional inference based solely on nuclear data should be interpreted with appropriate caution. Following isolation, single cells or nuclei are partitioned and barcoded for sequencing using several high‐throughput platforms. Droplet‐based systems are the most commonly used approach. Commercial platforms such as those developed by 10× Genomics and Fluent Biosciences use microfluidics to encapsulate each cell or nucleus in a nanoliter‐scale droplet together with a bead carrying unique barcodes (Klein et al., 2015; Macosko et al., 2015; Zheng et al., 2017). Microwell‐based platforms, including BD Rhapsody and Microwell‐seq, partition cells or nuclei into picoliter‐sized wells (Shum et al., 2019). Plate‐based methods, such as Smart‐seq. 2, are often used in lower‐throughput but higher‐sensitivity applications because they capture full‐length transcripts and enable isoform‐level analysis (Picelli et al., 2014). Split‐pool combinatorial barcoding avoids the need for complex microfluidic instruments by using cells or nuclei themselves as compartments. This strategy involves multiple rounds of splitting the sample pool, adding a unique barcode, and pooling the samples again. It is highly scalable, compatible with fixed cells, and relatively cost‐effective (Rosenberg et al., 2018; Cao et al., 2019). However, each platform introduces its own characteristics in capture efficiency, noise structure, and throughput. Therefore, computational inferences derived from these datasets, including cell clustering, trajectory analysis, and regulatory network inference, should ideally be supported by independent validation, such as in situ hybridization or reporter lines, to strengthen biological interpretation.

Single‐cell epigenome sequencing in plant

To understand the regulatory mechanisms associated with cell‐type‐specific gene expression, it is important to profile the epigenomic landscape. Single‐cell epigenomic technologies provide detailed views of regulatory landscapes and help connect cellular identity with transcriptional regulation (Figure 1B).

In plants, one of the most widely used platforms is the assay for transposase‐accessible chromatin using sequencing (ATAC‐seq), which is generally implemented in a single‐nucleus format (snATAC‐seq). This technique utilizes the Tn5 transposase to preferentially fragment and tag open chromatin regions (ACRs), which often correspond to active regulatory elements such as promoters and enhancers (Buenrostro et al., 2015). An important advantage of snATAC‐seq is its compatibility with a range of sample types. It can reduce some of the difficulties associated with tissue dissociation by allowing direct extraction of nuclei from complex or dense tissues, as well as from frozen samples. This makes it useful for the analysis of cellular heterogeneity while preserving information on chromatin accessibility. Early studies in maize used snATAC‐seq to generate genome‐wide cis‐regulatory atlases, identifying cell‐type‐specific ACRs and enriched transcription factor (TF) binding motifs (Marand et al., 2021; Jiang et al., 2025a). This approach has also been extended to major crops, including rice, soybean, and sorghum, for which multi‐organ snATAC‐seq atlases have been reported (Swift et al., 2024; Zhang et al., 2025a; Wang et al., 2025e).

However, snATAC‐seq depends strongly on the quality of nuclear extraction, is affected by data sparsity, and does not capture cytoplasmic regulatory information (Kwok et al., 2025). In addition, snATAC‐seq involves relatively high experimental costs and requires substantial bioinformatic analysis, which places corresponding demands on both experimental design and downstream data processing. Recent technological developments now allow the simultaneous profiling of transcriptomes and chromatin accessibility from the same nucleus. This multiome approach, which has recently been applied to Arabidopsis root tips under osmotic stress, enables a more direct comparison between chromatin state and transcriptional output within the same cell than approaches based on separately profiled datasets (Liu et al., 2024b; Wang et al., 2025e). Further improvement in the interpretation of single‐cell ATAC‐seq (scATAC‐seq) at single‐cell and single‐locus resolution will depend on both computational and experimental progress. This includes optimization of experimental workflows, such as scTurboATAC, to improve the cleavage efficiency of Tn5 transposase. Such improvements may help increase signal intensity and reduce data sparsity (Kwok et al., 2025). In animals, a long‐read single‐cell transposase‐accessible chromatin sequencing technology (scNanoATAC‐seq. 2) has been developed for limited samples and can be applied to single cells. This method has shown high sensitivity and low contamination in single‐cell analysis and has been used to support studies of early embryonic samples, which are often difficult to obtain (Li et al., 2025a).

Beyond chromatin accessibility, other single‐cell epigenomic technologies are emerging in plant science, including methods for profiling DNA methylation (Wang et al., 2021b; Luo et al., 2022) and histone modifications (Xie et al., 2023). These approaches are expected to further support the study of gene regulation at single‐cell resolution.

Spatial genomics technologies in plant

A major limitation of scRNA‐seq is that the dissociation process required for cell isolation results in the loss of the original spatial context of cells. This information is essential for understanding how cell‐cell interactions, tissue microenvironments, and positional cues govern development and stress responses (Giacomello et al., 2017; Longo et al., 2021). Spatial transcriptomics (ST) technologies have emerged as a powerful solution, enabling transcriptome‐wide analysis while preserving the native tissue architecture (Ståhl et al., 2016; Figure 1C). In plants, sequencing‐based in situ capturing is currently the most commonly used spatial transcriptomics approach. These techniques involve placing a tissue cryosection onto a slide or chip that is arrayed with spatially barcoded oligonucleotide probes. The tissue is permeabilized, allowing mRNA to be “captured” by the probes at their original locations (You et al., 2024). The resulting barcoded cDNA is then sequenced to generate a spatial map of gene expression across the tissue section.

10× Genomics Visium is a popular commercial platform based on an array of spots with a diameter of 55 μm (Ståhl et al., 2016; Vickovic et al., 2019; Figure 1C). Because this resolution usually captures transcripts from small groups of cells rather than single cells, its spatial resolution remains limited. However, it has been used in studies of the spatial organization of gene expression in complex plant tissues, including poplar stems (Du et al., 2023), germinating barley grains (Peirats‐Llobet et al., 2023), and developing wheat grains under heat stress (Wang et al., 2024a). Spatial enhanced resolution omics‐sequencing (Stereo‐seq), developed by BGI, provides higher spatial resolution by using DNA nanoball‐patterned arrays with spots as small as 220 nm in diameter (Chen et al., 2022). This finer resolution has supported the generation of detailed spatial atlases. In plants, Stereo‐seq has been used to distinguish highly similar cell subtypes in Arabidopsis leaves (Xia et al., 2022) and to build high‐resolution spatial transcriptome maps of developing maize ears (Wang et al., 2024c) and soybean nodules (Liu et al., 2023b).

Imaging‐based methods, such as sequential fluorescence in situ hybridization (seqFISH) and multiplexed error‐robust fluorescence in situ hybridization (MERFISH), provide another route to spatial analysis at high resolution (Chen et al., 2015; Eng et al., 2019; Figure 1C). SeqFISH involves repeated cycles of hybridization, imaging, and signal removal on the same sample. By using combinations of fluorescence signals for encoding, it can identify large numbers of mRNA molecules in situ and support transcript detection at subcellular resolution while preserving tissue structure (Eng et al., 2019). An advantage of this method is its high detection sensitivity and spatial precision. However, repeated imaging cycles can prolong the experimental process, and regions with high transcript abundance may suffer from signal crowding, which increases the difficulty of image analysis and data processing.

Multiplexed error‐robust fluorescence in situ hybridization (MERFISH) is another imaging‐based method that uses fluorescently labeled probes to visualize individual RNA molecules directly within cells. It provides high‐resolution spatial data, but unlike sequencing‐based methods, it generally targets predefined gene panels rather than transcriptome‐wide profiles. Its main strengths are high throughput, high spatial resolution, and good quantitative performance, which allow the spatial positions of many transcripts to be recorded at subcellular resolution. Through combinatorial encoding and error‐correction strategies, MERFISH can also support the detection of relatively low‐abundance transcripts and the analysis of intracellular spatial organization and local microenvironmental variation (Fang et al., 2022; Androvic et al., 2023; Sarfatis et al., 2025).

The MERSCOPE platform, which is based on MERFISH, has been applied to in situ spatial transcriptomic analysis at single‐cell resolution in plant tissues. In Arabidopsis thaliana, this platform has supported the analysis of organ‐specific heterogeneity and transient transcriptional programs (Lee et al., 2025b). In another study using MERFISH, expression patterns of core genes associated with drought recovery were characterized in Arabidopsis leaf tissue (Illouz‐Eliaz et al., 2025). In crop research, MERFISH has also been adapted for technically challenging tissues, such as the wheat inflorescence, through adjustments including multi‐stage photobleaching and extended tissue clearing (Long et al., 2026).

The development of spatial transcriptomics has made it possible to visualize gene expression directly in tissue sections and has improved the analysis of tissue regionalization. However, current methods still face a trade‐off between single‐cell resolution and transcriptome‐wide coverage. Sequencing‐based methods can provide relatively unbiased transcriptome‐wide maps, but their spatial resolution is often limited. By contrast, imaging‐based methods enable more precise localization at the single‐cell or subcellular level, but they are usually restricted to predefined gene sets.

An important direction in the field is the computational integration of scRNA‐seq and spatial transcriptomics. These two methods are highly complementary: scRNA‐seq provides deep transcriptional information at true single‐cell resolution, allowing for robust cell‐type classification, while ST provides the spatial coordinates (Longo et al., 2021). Integrating cell identities inferred from scRNA‐seq with spatial data can support the construction of higher‐resolution spatial cell atlases (Satija et al., 2015; Biancalani et al., 2021). Nevertheless, effective integration of these two data types remains methodologically challenging. To address this issue, several computational frameworks have been developed. For example, the multi‐task learning framework stSCI projects heterogeneous single‐cell and spatial transcriptomic data into a batch‐corrected unified low‐dimensional space. In addition, the computational toolkit Spotiphy was developed to integrate deep single‐cell transcriptomic data with high‐resolution spatial gene expression data through data imputation approaches (Yang et al., 2025; Demesa‐Arevalo et al., 2026). These developments support the construction of spatial cell atlases with improved resolution and broader transcript coverage.

This integrative strategy has been applied in several crop studies. For example, the integration of snRNA‐seq and Stereo‐seq has been used to characterize transitional states during soybean nodule maturation (Liu et al., 2023b) and to map the cellular architecture of soybean seeds (Zhang et al., 2025a). Similarly, snRNA‐seq has been combined with spatial transcriptomics platforms to study shoot regeneration in tomato callus (Song et al., 2023) and to construct a spatiotemporal cell atlas of the wheat root tip (Ke et al., 2025). These integrated analyses can help identify cell types, define their spatial positions, and examine how neighboring cell relationships change during development or under stress conditions.

UNDERSTANDING PLANT GROWTH AND DEVELOPMENT

Crop development is a highly coordinated, dynamic, and continuous process that extends throughout the life cycle, from seed germination and vegetative growth to flowering, fruiting, and the formation of new seeds. Traditional population‐level omics studies often obscure the spatiotemporal heterogeneity among different tissues, organs, and even cell types within the same tissue, making it difficult to accurately characterize the dynamic changes in cell‐type‐specific programs and developmental trajectories associated with key agronomic traits. The development of single‐cell technologies has made it possible to systematically characterize the diversity, state transitions, interaction networks, and gene regulatory programs of major crop organs, including roots, stems, leaves, flowers, and seeds, across the life cycle (Marand et al., 2021; Zhang et al., 2025a; Wang et al., 2025e; Figure 2; Table 1). The construction of single‐cell maps spanning crop development not only improves our understanding of developmental biology in crops but also provides a useful basis for investigating the genetic regulation of complex agronomic traits and for supporting molecular breeding strategies.

Figure 2.

Summary of cell populations in various plant organs identified by single‐cell and spatial technologies

Using maize as an example, the figure illustrates the cell clusters that have been identified across the whole developmental stages based on single‐cell RNA‐seq. Figure created with Biorender.com.

Table 1.

Application of single‐cell technologies in plant growth

Tissue	Species	Technology	Reference
Leaf, pod, hypocotyl, nodule, root	Soybean	ScATAC‐seq, snRNA‐seq	Zhang et al. (2025a)
Seed	Soybean	SnRNA‐seq	Pelletier et al. (2025)
Endosperm	Maize	ScRNA‐seq	Yuan et al. (2024)
Pod	Peanut	SnRNA‐seq, snATAC‐seq	Cui et al. (2024)
Root	Rice	ScRNA‐seq, ATAC‐seq	Zhang et al. (2021)
Root	Maize, Setaria	ScRNA‐seq	Ortiz‐Ramírez et al. (2021)
Root	Maize, sorghum, Setaria	ScRNA‐seq, snRNA‐seq	Guillotin et al. (2023)
Seed, axillary bud, ear primorida, tasse primorida, embryonic roots, crown roots	Maize	SnATAC‐seq	Marand et al. (2021)
Root	Wheat	SnRNA‐seq, snATAC‐seq	Zhang et al. (2023b)
Root	Wheat	ScRNA‐seq, spRNA‐seq	Ke et al. (2025)
Root	Wheat	ScRNA‐seq	Du et al. (2025a)
Root	Medicago	SnRNA‐seq	Liu et al. (2023a)
Root	Soybean	SnRNA‐seq	Cervantes‐Pérez et al. (2024)
Rhizome	Oryza longistaminata	Stereo‐seq	Lian et al. (2024)
Shoot apex	Maize	ScRNA‐seq	Ma et al. (2025)
Shoot	Tomato	ScRNA‐seq	Omary et al. (2022)
Shoot apex	Tomato	SnRNA‐seq	Tian et al. (2020)
Callus	Tomato	ScStereo‐seq, snRNA‐seq	Song et al. (2023)
Lateral meristem	Cotton	ScRNA‐seq	Zhu et al. (2023b)
Stem	Peanut	SnRNA‐seq	Wang et al. (2025d)
Stem	Poplar	ScRNA‐seq	Chen et al. (2021)
Leaves and ligules	Maize	ScRNA‐seq	Satterlee et al. (2023)
Leaf epidermis	Maize	SnRNA‐seq	Sun et al. (2022a)
Leaf ligular	Maize	SnRNA‐seq	Wang et al. (2024b)
Leaves	Rice; sorghum	SnRNA‐seq, snATAC‐seq	Swift et al. (2024)
Leaves	Peanut	SnRNA‐seq	Liu et al. (2024a)
Leaves	Brassica rapa	ScRNA‐seq	Guo et al. (2022)
Floret and inflorescence meristems	Rice	ScRNA‐seq	Zong et al. (2022)
Pistils	Rice	SnRNA‐seq	Li et al. (2023)
Shoot apex, seed, flag leaf, panicle, culm, leaf, crown root, tiller bud	Rice	SnRNA‐seq, snATAC‐seq	Wang et al. (2025e)
Ears	Maize	ScRNA‐seq	Xu et al. (2021)
Ears	Maize	ScStereo‐seq	Wang et al. (2024c)

Application of single‐cell technologies in plant seeds

Understanding seed development provides an important basis for the improvement of agronomically important traits. Single‐cell multi‐omics technologies offer a cell‐level perspective for studying seed development by characterizing cellular heterogeneity and examining the molecular basis of coordinated multicellular development. Seeds are complex structures consisting of three regions, namely the embryo, endosperm, and seed coat, each of which can be further divided into subregions with distinct tissues, cell layers, and cell types.

Different cell types were identified by sequencing individual cells from soybean seeds at the cotyledon stage. In addition, some co‐expression networks were cell‐type or region‐specific, whereas others were distributed across multiple regions, suggesting that different seed components are regulated through both region‐specific and coordinated developmental programs (Pelletier et al., 2025). A single‐cell multi‐omics map of rice showed that traits related to seed quality were associated with embryo cells (Wang et al., 2025e). The endosperm is an important nutrient source for both humans and animals. In soybean, the temporal and spatial dynamics of three cell subtypes, microbud, periphery, and hypocotyl, during endosperm development were characterized through the integration of single‐cell and spatial transcriptomic data (Zhang et al., 2025a). The same study further showed that the transcription factor GmATHB13 is involved in the differentiation of mid‐embryonic pulp tissue in soybean by inhibiting axial development and activating the cotyledon pulp gene program, thereby contributing to the coordination of seed nutrient accumulation. These findings provide a potential target for the improvement of yield‐related traits. Single‐cell transcriptomic analysis of differentiated maize endosperm cells showed that the endosperm could be classified into five major cell types (Yuan et al., 2024). By integrating single‐cell transcriptome data and DAP‐seq‐based analysis of regulatory networks and transcription factor binding sites, a high‐confidence gene regulatory network of maize endosperm was constructed, which contained 181 transcription factors and target genes. CRISPR‐Cas9 technology was used to generate mybr29, mybr19, and ereb108 mutants, which verified that these three genes were important regulators in BETL cell types. This study provides a useful framework for understanding cereal endosperm development and function at single‐cell resolution.

Peanut is an important legume crop for oil production and agricultural use, with flowering occurring above ground and pod development taking place below ground. The development of underground pods has a strong influence on peanut yield. snRNA‐seq and snATAC‐seq were performed on aerpeg, subpeg, and exppod tissues or cell populations, respectively. The results suggested that AGL5, YAB5, and IAA play important roles in pod development, and that these factors may act together to regulate parenchyma cell development and support the continued expansion of peanut pods (Cui et al., 2024). Overall, single‐cell sequencing technologies have improved our understanding of the molecular basis of crop seed formation by characterizing cellular heterogeneity and spatiotemporal dynamics during seed development. These findings provide potential targets for the improvement of seed quality and yield and support the development of molecular breeding methods.

Application of single‐cell technologies in plant roots

Roots are essential for water and nutrient uptake; understanding their development requires detailed analysis. Single‐cell multi‐omics techniques systematically reveal conserved and divergent mechanisms of root development and cell‐type‐specific responses to symbiotic nitrogen fixation by analyzing root cell heterogeneity.

Root development depends on the establishment and activity of meristems, which give rise to distinct cell types. Zhang et al. reconstructed continuous developmental trajectories of epidermal and ground tissue lineages by analyzing single cells from root tips (Zhang et al., 2021), revealing both conserved and divergent root development pathways between dicots and monocots. Root structural diversity across species is largely caused by differences in cortical cell number (Esau, 1977), and this cortical diversity can influence how plants respond to environmental stresses. Ortiz‐Ramírez et al. showed that SHR could regulate cortex expansion in maize and Setaria viridis using histological staining together with scRNA‐seq (Ortiz‐Ramírez et al., 2021). Key traits in different species are mediated by specific cell types, and single‐cell studies indicate that whole‐genome duplication (WGD) events provide raw material for the generation of new genes, facilitating the rapid evolution of cell‐type functions. Comparative analyses of genes involved in mucilage synthesis show similar expression patterns of cortex cells in sorghum and Setaria, and of root cap cells in maize and Setaria (Guillotin et al., 2023).

Asymmetric expression of homologous genes is common in wheat, but the extent to which this asymmetry reflects cell heterogeneity remains unclear. Zhang et al. (2023b) explored asymmetric gene expression patterns in wheat roots at single‐cell resolution, constructed cell‐type‐specific networks that help define cell identity, and combined cell‐type expression with chromatin accessibility to identify transcription factors associated with the transition from meristem cells to root hairs. Single‐cell and spatial transcriptome analyses of wheat root tip meristems grown in soil enabled annotation of wheat root cell populations through cross‐species homology and spatial verification, and identified many known and previously unreported cell‐type marker genes and developmental regulators (Ke et al., 2025). Du et al. identified the main cell types of wheat roots, clarified aquaporin‐mediated water transport mechanisms in different root cell types, and compared conserved and divergent expression patterns in wheat and rice root tips (Du et al., 2025a).

Shoot‐borne roots play important roles in plant adaptation to different environments. Omary et al. described molecular features of shoot‐borne root formation using single‐cell approaches, showing that roots arising from stems derive from differentiated primary phloem parenchyma cells, whereas lateral roots originate from pericycle‐derived cells (Omary et al., 2022). The LBD transcription factor SBRL regulates stem‐to‐root transition cell formation; evidence suggests this mechanism is conserved across angiosperms.

Legumes perform symbiotic nitrogen fixation through infection of root cells by rhizobia, and nitrogen fixation occurs in infected cells that host the bacteria. How cis‐regulatory and chromatin features of these cells change after infection is still being explored. TF motifs enriched in infected cells show cell‐specific accessibility, including the known nodule regulator NLP7 motif; two additional motifs, STREME‐7 and STREME‐9, were identified with infection‐specific accessibility and may represent candidate regulators of nodule development (Zhang et al., 2025a). Transcriptome analysis of root cells in Medicago truncatula showed that MtFER and MtLYK3 have similar expression responses to nodulation factors, and that phosphorylation of MtFER by MtLYK3 may coordinate developmental, immune, and symbiotic gene expression during rhizobial invasion (Liu et al., 2023a).

Single‐cell transcriptomic maps of soybean root systems and mature nodules indicate that mature nodules are heterogeneous, comprising multiple cell subpopulations (Cervantes‐Pérez et al., 2024). Gene co‐expression analysis of 28‐d nodule cells identified six candidate genes, including NAC family members, potentially involved in symbiotic regulation. A membrane‐associated protein, GMFWL3, was also identified; loss‐of‐function reduced nodule number, supporting a role in nodulation. In summary, single‐cell approaches that combine cross‐species comparison, spatial epigenetics, and network analysis can define cell‐type‐specific regulatory frameworks and identify molecular targets relevant to root adaptation and symbiotic engineering.

Application of single‐cell technologies in plant shoot apex

The shoot apical meristem is the source of aboveground organs. Single‐cell multi‐omics help to reveal the cellular basis of organogenesis and regeneration by resolving shoot apical meristem dynamics and reprogramming mechanisms. In maize, ZmEREB14 was identified as a regulator associated with meristem development and yield‐related processes using single‐cell data (Ma et al., 2025). Comparative spatiotemporal transcriptomics of rhizomes and tillers in perennial wild rice identified two vascular bundle clusters and two parenchyma clusters as major distinguishing features between these organs, and localized meristem initiation cells to a parenchyma depression at the internode base (Lian et al., 2024). snRNA‐seq in tomato produced high‐resolution expression maps of stem apical cells and helped reconstruct developmental trajectories, identify regulators, and outline gene regulatory networks involved in cell differentiation (Tian et al., 2020).

Callus is a reprogrammed cell mass relevant to regeneration and genetic transformation. Spatial and nuclear transcriptome profiling of tomato callus during stem regeneration revealed cellular heterogeneity and indicated that vascular tissues and chloroplast‐containing cells near light‐induced primordia support adventitious bud formation (Song et al., 2023). High‐resolution single‐cell maps of hypocotyls in cotton varieties identified primary vascular cells as major contributors to callus formation (Zhu et al., 2023b). Single‐cell profiling of non‐embryogenic and primary embryogenic calli in cotton showed that SE‐ASSOCIATED LIPID TRANSFER PROTEIN (SELTP) gene marks embryogenic states in a dose‐dependent manner and can serve as a quantitative single‐cell marker of embryogenesis (Guo et al., 2024). Spatial transcriptome and metabolome integration in cotton somatic embryo development highlighted genes such as AATP1 and DOX2 that are associated with pluripotency acquisition and embryo development (Sun et al., 2025). Collectively, these studies indicate that single‐cell omics provide resources and molecular insights relevant to meristem organization, early organogenesis, and stem‐cell homeostasis.

Application of single‐cell technologies in plant stem

The stem provides mechanical support and conduits for transport. Single‐nucleus transcriptomics has been used to dissect cell‐type‐specific regulatory networks in stems and to identify transcription factors associated with tissue differentiation. In peanut, combined single‐cell and bulk transcriptome data indicated that AhWRKY70 is enriched in stem cortex and xylem and is associated with stem growth regulation (Wang et al., 2025d). Single‐cell mapping in poplar clarified molecular features of phloem and xylem differentiation, profiled expression of hormone‐related genes, and suggested potential cases of gene redundancy (Chen et al., 2021). These findings locate regulatory hubs of stem development and identify candidate targets for stem‐structure optimization and stress resilience.

Application of single‐cell technologies in plant leaves

Single‐cell multi‐omics reveal cell‐specific networks underlying leaf morphology and photosynthetic cell differentiation, which are relevant to plant architecture and photosynthetic performance. In maize, leaf and ligule development share genetic programs, and Wox3‐module activity contributes to planar growth of both organs (Satterlee et al., 2023). Leaf angle is one of the key factors determining planting density and yield. Histology and single‐nucleotide RNA sequencing revealed the central role of bHLH30 and bHLH155 in regulating leaf angle (Wang et al., 2024b). Photosynthesis in C4 plant leaves is an efficient photosynthetic pathway, about 50% more efficient than the C3 pathway. By constructing gene expression and chromatin accessibility maps of rice and sorghum, the regulatory mechanism of gene expression in bundle sheath cells in C4 photosynthesis was revealed, and cis‐regulatory elements of the DOF transcription factor family were found, which enhanced the expression of photosynthetic genes and improved photosynthetic efficiency (Swift et al., 2024).

Single‐cell data have resolved differences between palisade and spongy mesophyll cells and characterized stomatal cell lineages, including guard cells and subsidiary cells, in crops such as maize (Guo et al., 2022). The leaf epidermis, as the outermost layer of the cell, regulates the exchange of gases and water between the plant and the environment, and protects the plant from both biotic and abiotic stresses. Through the single‐nucleotide RNA sequencing (snRNA‐seq) technique, the signaling network of corn stomata was comprehensively described. The types of mature and developing stomatal cells were identified, including guard cells (GCs) and subsidiary guard cells (SCs) (Sun et al., 2022a). By decoding leaf‐cell heterogeneity, developmental programs, and response networks, single‐cell studies supply candidate targets and datasets for trait improvement aimed at higher light‐use efficiency and stress tolerance.

Application of single‐cell technologies in plant ears

During reproductive growth, inflorescence and panicle development influence grain yield and quality through coordinated meristem activity, organ differentiation, and grain formation. Single‐cell multi‐omics have begun to resolve transcriptional programs and regulatory factors that underlie meristem fate transitions and floral organ specification. Single‐cell profiling of rice inflorescences produced a transcriptional atlas covering the transition from inflorescence to floret and implicated the WOX factor DWARF TILLER1 and OsAUX1 in meristem activity and inflorescence development (Zong et al., 2022). Single‐cell multi‐omics across panicle stages identified transitional cell types in floral meristems (Wang et al., 2025e). In rice pistils, single‐cell analyses identified markers of ovule and carpel progenitor cells (Li et al., 2023a). High‐resolution transcriptome mapping of developing maize ears combined with ChIP‐seq identified direct targets of transcription factors such as ZmHDZIV6 and ZmM16, and integration with GWAS nominated candidate genes associated with ear morphology and yield (Xu et al., 2021). Stereo‐seq‐based spatial mapping of maize ear development highlighted genes, including ZmMADS8 and ZmMADS14, and supported their roles in ear morphogenesis and yield traits (Wang et al., 2024c).

In summary, single‐cell atlases of inflorescence and ear development, together with transcription factor networks and spatial expression modules, clarify regulators of meristem activity, pistil cell fate, and ear morphogenesis. Continued integration of multi‐omics, cross‐species comparison, and gene‐editing validation should help define conserved and species‐specific pathways relevant to yield improvement.

APPLICATION OF SINGLE‐CELL TECHNOLOGIES IN PLANT ENVIRONMENTAL ADAPTABILITY

Environmental stresses negatively affect plant growth, development, metabolic homeostasis, and the transport of water and nutrients. Investigating how plants respond to these stresses is essential not only for understanding the molecular basis of stress adaptation but also for guiding the development of stress‐tolerant crop varieties and improving yield and quality under adverse conditions (Jiang et al., 2025b). Single‐cell technologies are increasingly being applied to uncover cell‐type‐specific gene expression and regulatory networks in plants under environmental stress, providing valuable insights for crop improvement under stressful conditions (Cao et al., 2023; Li et al., 2023b; Li et al., 2024; Wang et al., 2025b; Figure 3; Table 2).

Figure 3.

Application of single‐cell technologies in plant stress resistance

(A–D) Single‐cell multi‐omics is a powerful tool for research on plant stress. This approach facilitates the analysis of cell‐type‐specific mechanisms of plant stress responses, including identifying marker genes (A), exploring differentiation trajectories under stress conditions (B), analyzing single‐cell differences between varieties (C), and constructing regulatory networks for stress response (D). Figure created with Biorender.com.

Table 2.

Application of single‐cell technologies in plant stress resistance

Stress	Species	Tissue	Technology	Reference
Heat stress (42°C, 2 h)	Maize	Primary root tips (4 d old)	ScRNA‐seq	Wang et al. (2025b)
Heat stress (45°C, 3 h)	Rice	Primary root tips (3 d old)	SnATAC‐seq	Feng et al. (2022)
Heat stress (40°C, 12 h)	Chinese cabbage	Shoot apices and developing leaves	ScRNA‐seq	Sun et al. (2022b)
High temperature stress (28°C to 31°C, 7 d)	Cotton	The stage 7 anthers	SnRNA‐seq, snATAC‐seq	Li et al. (2024)
Heat stress (40°C, 2 h)	Pearl millet	Leaves	ScRNA‐seq	Jin et al. (2025)
High‐salinity (200 mM NaCl), low‐nitrogen ((NH4)2SO4 dropped out), and iron‐deficiency (Fe(II)‐EDTA dropped out)	Rice	Proximal shoots, root tips	ScRNA‐seq	Wang et al. (2021a)
Salt (100 mM NaCl, 0.5 h or 1 h, or 150 mM NaCl, 0.5 h)	Cotton	Root lateral root tips	ScRNA‐seq	Li et al. (2023b)
Salt (150 mM NaCl, 12 h)	Chinese cabbage	Primary root tips (6 d old)	SnRNA‐seq, snATAC‐seq	Liu et al. (2025)
Drought	Maize	Lateral root	ScRNA‐seq	Li et al. (2025b)
Dark and light	Peanuts	Leaves (3, 5, and 7 d old)	ScRNA‐seq	Deng et al. (2024)
Soil stress	Rice	Primary root tips (3 d old)	ScRNA‐seq, stRNA‐seq	Zhu et al. (2025)
Nitrate	Maize	Primary root tips (4 d old)	ScRNA‐seq	Li et al. (2022b)
Boron (25 μM H3BO3)	Pea	Shoot (10 d old)	ScRNA‐seq	Chen et al. (2024)
Fungal (Puccinia polysora, 12 h post‐inoculation)	Maize	Third leaf tissues	ScRNA‐seq	Yan et al. (2025b)
Fungal (Fusarium verticillioides, 48 h post‐inoculation)	Maize	Primary root tips (7 d old)	ScRNA‐seq	Cao et al. (2023)
Fungal (Magnaporthe oryzae, 0, 12, 24, and 48 h post‐inoculation)	Rice	Leaves (2 weeks old)	SnRNA‐seq, StRNA‐seq	Wang et al. (2025c)
Bacteria (Ralstonia solanacearum)	Peanuts	Primary root tips (4 d old)	SnRNA‐seq	Yin et al. (2025)
Fungal (Sporisorium scitamineum)	Sugarcane	Buds	ScRNA‐seq	Zang et al. (2025)
Virus (Sugarcane mosaic virus, 2 h on 5 d post‐inoculation)	Maize	Leaves (8 d old)	ScRNA‐seq	Chen et al. (2025)

Heat stress

Heat stress is an important abiotic stress that affects plant growth, development, and reproduction and can lead to yield losses (Bita and Gerats, 2013). Single‐cell technologies have contributed to the analysis of complex and heterogeneous cellular responses to high temperature, which are often difficult to resolve using bulk‐tissue analyses that average signals across diverse cell types (Figure 3A, B; Jean‐Baptiste et al., 2019; Sun et al., 2022b).

In Arabidopsis roots, an early scRNA‐seq study showed that responses to heat stress are highly cell‐type‐specific (Jean‐Baptiste et al., 2019). The greatest transcriptional changes occurred in the outer cell layers, including the epidermis, hair cells, and cortex, which are directly exposed to the stress. This response involved a redistribution of transcriptional activity, with a general downregulation of genes related to development and growth and an upregulation of canonical heat‐shock genes, including heat shock proteins (HSPs) (Jean‐Baptiste et al., 2019). Similar cell‐type‐specific responses have also been reported in major crops. In Chinese cabbage leaves, scRNA‐seq revealed a heterogeneous response to heat, with mesophyll and epidermal cells showing relatively strong transcriptional activity (Sun et al., 2022b). The same study also found that different heat shock factor (HSF) genes were induced in a cell‐type‐specific manner.

In maize roots, a spatiotemporal single‐cell RNA‐seq atlas identified cortex cells as an important cell type in the heat stress response (Wang et al., 2025b). Beyond transcriptomics, single‐cell epigenomics studies have provided additional mechanistic insights. A scATAC‐seq study in rice roots revealed dynamic and cell‐type‐specific changes in chromatin accessibility under heat stress and identified regulatory elements associated with the transcriptional response (Feng et al., 2022). Integrated analyses have also been applied to heat‐induced male sterility, an important cause of yield reduction. In cotton anthers, combined snRNA‐seq and snATAC‐seq analysis showed that heat stress strongly affected tapetal cells, which are required for pollen development (Li et al., 2024). Under high temperatures, a specific tapetal subpopulation involved in pollen wall synthesis was no longer detected. This change was associated with reduced chromatin accessibility at loci of key pollen wall synthesis genes, including QRT3 and CYP703A2, and with abnormal pollen wall development and subsequent sterility (Li et al., 2024). Pearl millet is a typical C4 heat‐tolerant crop. By comparing single‐cell transcriptome maps of pearl millet leaves under heat stress and control conditions, vascular tissue cells were identified as an important responsive cell type, showing the largest numbers of differentially expressed genes and heat stress memory genes (Jin et al., 2025). Overall, single‐cell omics technologies have supported the analysis of cell‐type‐specific heat stress responses in both model and crop plants and have improved our understanding of transcriptional and epigenetic changes associated with thermotolerance and heat‐induced male sterility.

Osmotic stress

Osmotic stress, primarily caused by drought and high salinity, is a major environmental factor limiting crop productivity worldwide (Zhu, 2016). Single‐cell technologies have helped move beyond whole‐organ analyses by enabling the characterization of transcriptional and regulatory responses of individual cell types to these stresses (Rhaman et al., 2024).

Single‐cell transcriptomics has been applied in crops to construct high‐resolution atlases of salt stress responses. In cotton, a single‐cell RNA‐seq atlas of roots under salt stress identified numerous cell‐type‐specific differentially expressed genes (Figure 3C; Li et al., 2023b). This study identified the auxin‐responsive gene GaGH3.6 as an important regulator of salt tolerance. Similarly, a single‐cell atlas of Brassica rapa roots under salt stress showed that root hairs, a critical cell type for environmental sensing and nutrient uptake, are highly sensitive to salinity (Liu et al., 2025). Salt stress inhibited root hair elongation and repressed a root‐hair‐specific gene regulatory network (GRN) involving BcRAP2.11 and BcIRT2, thereby identifying a developmental pathway associated with the salt response.

Beyond transcriptomics, multiome approaches provide additional mechanistic insights by simultaneously profiling the epigenome and transcriptome from the same nucleus. A recent study on Arabidopsis root tips utilized a single‐nucleus multiome approach to investigate the immediate response to osmotic stress (Liu et al., 2024b). This method enabled the comparison of stress‐induced gene expression with changes in chromatin accessibility. Root hair cells and epidermal cells showed pronounced cell‐type‐specific responses. An osmotic stress‐activated gene regulatory network further identified several stress‐related regulators in defined NHCC root cell types. This type of multi‐omic analysis provides a useful framework for linking epigenomic changes with transcriptional networks involved in osmotic stress adaptation. A high‐resolution transcriptome analysis of leaves and roots from rice seedlings showed that abiotic stresses, including low nitrogen, high salinity, and iron deficiency, mainly affected gene expression in a cell‐type‐specific manner. Under different stress conditions, the same cell type showed some common patterns of gene expression change, suggesting that plant cells may use related transcriptional strategies to respond to multiple stresses (Wang et al., 2021a). The formation and distribution of lateral roots are important for the efficient uptake of water and nutrients and for adaptation to abiotic stresses such as drought. By integrating multi‐omics data, the transcription factor ZmbZIP89, which is associated with lateral root development and drought tolerance in maize, was identified and cloned (Li et al., 2025b).

In summary, single‐cell multi‐omic technologies have revealed cell‐type‐specific transcriptional and epigenetic changes associated with plant responses to osmotic stress and have supported the identification of key regulators and gene networks involved in salt and drought tolerance in crops.

Light

Light is an important environmental signal for plants and regulates the developmental transition from skotomorphogenesis to photomorphogenesis. scRNA‐seq has helped move beyond whole‐organ analyses by enabling the study of how this signal is perceived and processed at cell‐type‐specific resolution in both model plants and crops.

A time‐series scRNA‐seq study in Arabidopsis seedlings comparing dark‐grown and light‐grown (de‐etiolated) cotyledons revealed that the response to light is highly heterogeneous across cell types (Han et al., 2023). Cell‐type‐specific expression atlases revealed differential activity of key light‐signaling hubs, including the transcription factors HY5 and PIFs (PIF1, PIF3, PIF4, PIF5). In peanut, an scRNA‐seq atlas of etiolated and de‐etiolated seedlings was used to dissect the transcriptional changes during photomorphogenesis (Deng et al., 2024). This study reveals that peanut leaf cells exhibit clear transcriptional differences under light and dark conditions. Under dark conditions, cell division and cell‐cycle progression were reduced, together with lower expression of genes related to chlorophyll synthesis. Under light conditions, specific developmental trajectories of epidermal cells were altered, and this response may be associated with auxin signaling (Figure 3B). To further examine the underlying regulatory mechanism, heterologous overexpression of the peanut gene AhAHL17 in Arabidopsis thaliana was shown to promote the expansion of leaf epidermal cells. Together, these findings provide useful insights into how light signals regulate leaf cell development.

Nutrient

Nutrient availability is a fundamental determinant of plant growth and crop yield. Single‐cell transcriptomics provides a detailed approach for examining how plants perceive and respond to nutrient fluctuations, showing that these responses are heterogeneous and coordinated across different cell types.

In crop roots, scRNA‐seq has been applied to map the complex response to nitrate, a critical macronutrient. A study in maize root tips constructed a single‐cell transcriptional atlas to identify cell‐type‐specific nitrate‐response genes (Li et al., 2022b). By comparing nitrate‐treated and nitrate‐free conditions, this study showed that epidermal and meristematic zone cells are important responders. Genes involved in nitrate uptake and assimilation, such as ZmGS2 and ZmNAR2.1 (Buoso et al., 2021), were specifically induced in these outer cell layers, highlighting the spatial organization of the primary nitrate response (Li et al., 2022b).

Single‐cell analysis in shoots has also provided novel insights into micronutrient deficiency. In pea, scRNA‐seq was employed to study the shoot apex under boron deficiency (Chen et al., 2024). This analysis identified cell‐type‐specific gene expression changes, including the downregulation of photosynthesis‐related genes in mesophyll cells, which is a known physiological consequence of boron deficiency (Camacho‐Cristóbal et al., 2008). The study also suggested that boron deficiency affects shoot apical meristem development by reducing the expression of chromatin‐remodeling genes, including homologs of the SWI/SNF complex (Chen et al., 2024). These findings illustrate how scRNA‐seq can move beyond whole‐organ analysis and identify specific cellular pathways and regulatory factors associated with nutrient stress adaptation.

Soil compaction

Soil compaction, which presents significant mechanical impedance to root penetration, is a major abiotic stress that limits root system architecture and crop yield. Single‐cell technology, combined with soil‐simulating gel systems, has supported the analysis of cell‐type‐specific responses to this physical stress (Zhu et al., 2025).

Biotic stress

Plant–pathogen interactions are dynamic and localized processes. Infections are often initiated in specific cells, and host responses differ between infected cells and neighboring bystander cells (Tang et al., 2023; Zhu et al., 2023a). Traditional bulk RNA‐seq averages these distinct signals and can therefore obscure cell‐type‐specific mechanisms associated with both resistance and susceptibility (Jones and Dangl, 2006). Single‐cell technologies have become useful for resolving this heterogeneity by enabling the capture of distinct transcriptional states of individual cells during infection (Zhu et al., 2023a).

ScRNA‐seq studies have shown that plant immune responses are often cell‐type‐specific. In leaf tissues, epidermis and mesophyll cells are often among the earliest responders and show marked transcriptional changes following fungal or bacterial infection (Tang et al., 2023; Zhu et al., 2023a). For example, in Arabidopsis infected with Pseudomonas syringae, pseudotime analysis revealed a continuum of cellular states, tracing a trajectory from an initial immune response to a later susceptible state (Zhu et al., 2023a). Similarly, scRNA‐seq of Arabidopsis infected with the fungus Colletotrichum higginsianum identified distinct cell populations, including a specific infected epidermal cell cluster that co‐expressed both defense genes and susceptibility‐associated genes (Tang et al., 2023). These results illustrate the capacity of scRNA‐seq to capture the interaction between host defense and pathogen influence within individual cells.

This high‐resolution approach is also useful for identifying cell‐type‐specific genes associated with immunity in crops, thereby providing candidate targets for crop improvement. In maize roots infected with Fusarium verticillioides (causal agent of stalk rot), scRNA‐seq identified a cell‐type‐specific immune regulatory network in the root apical meristem (Cao et al., 2023). This study identified ZmWOX5b and the auxin transporter ZmPIN1a as key regulators whose overexpression enhanced resistance.

In leaves of resistant and susceptible maize varieties, several immune‐related genes were found to be pre‐activated in guard cells and epidermal cells of the resistant varieties (Yan et al., 2025b). In rice, an integrated scRNA‐seq and spatial transcriptomics study of Magnaporthe oryzae infection mapped the spatial progression of the fungus and identified candidate targets associated with resistance breeding (Wang et al., 2025c). This strategy has also been applied to viral and oomycete pathogens. In sugarcane infected with smut disease, scRNA‐seq identified cell‐type‐specific regulators such as ScNPR3‐ScTGA2 (Zang et al., 2025). In maize leaves infected with SCMV during early systemic infection before symptom onset, five cell types were resolved, and mesophyll‐4 cells showed the highest levels of viral accumulation (Chen et al., 2025). In peanut resistance to Phytophthora, single‐cell analysis suggested an important role for suberin deposition in cortical cells (Yin et al., 2025).

APPLICATION OF SINGLE‐CELL TECHNOLOGIES IN PLANT EVOLUTION

Single‐cell technologies are increasingly used in the study of plant evolution because they enable the analysis of cell‐type diversity, regulatory networks, and lineage innovation across phylogenetic scales (Figure 3D). Recent studies suggest that conserved cell identities and their associated gene modules can be traced across vascular plant evolution. For example, a unified cell atlas of vascular plants integrated single‐cell transcriptomes from lycophytes, ferns, gymnosperms, and angiosperms, and identified shared cell populations and conserved genes associated with epidermal, xylem, and phloem lineages, as well as companion‐cell‐like populations in ferns and gymnosperms (Xue et al., 2025). These findings provide molecular evidence relevant to the early establishment of vascular innovations and suggest convergent or parallel patterns in the evolution of phloem complexity.

Beyond vascular tissues, single‐cell atlases in bryophytes such as Marchantia polymorpha and Physcomitrium patens show how tissue‐specific programs diversified during land plant evolution (Zeng et al., 2025). Single‐cell transcriptomics also enables the construction of pan‐species cellular atlases that support comparative analysis of the conservation and divergence of cell differentiation trajectories across grasses. By comparing homologous cell populations, these approaches help reveal how core gene regulatory networks are either maintained or rewired during plant evolution, contributing to morphological diversity and specialized crop traits (Guillotin et al., 2023).

By constructing multi‐species single‐cell atlases, it is also possible to trace the evolutionary trajectories of specific cell types and examine physiological innovations, such as the transition from C3 to C4 photosynthesis in grasses (Mendieta et al., 2024). For instance, single‐cell dual‐omics sequencing in rice, a C3 crop, and sorghum, a C4 crop, showed that DOF transcription factors are expressed in bundle sheath cells of both C3 and C4 plants, but with stronger expression in the C4 species (Swift et al., 2024).

Comparative single‐cell profiling of shoot apical meristems between divergent lineages, such as Arabidopsis and maize, has also identified both conserved stem‐cell regulators and rewired gene networks associated with morphological diversity and yield‐related variation (Zeng et al., 2025). In addition, integrating single‐cell chromatin accessibility data across multiple plant species suggests that cis‐regulatory evolution is cell‐type‐specific rather than uniform, with some lineages, such as epidermal cells, showing relatively rapid regulatory divergence (Yan et al., 2025a).

In summary, single‐cell technologies, including transcriptomic and multi‐omic approaches, provide useful tools for plant evolutionary research by supporting multi‐species cell atlas construction, analysis of key innovations and regulatory changes, and characterization of cell‐type‐specific cis‐regulatory evolution associated with plant adaptation and diversification.

FUTURE PERSPECTIVE

Technological breakthrough

Advances in plant single‐cell technologies are helping to address long‐standing technical challenges in crop research. The integration of improved sample preparation, multi‐modal sequencing, and high‐throughput barcoding has supported increasingly detailed characterization of cellular landscapes. In addition, developments such as semipermeable droplets and imaging‐free spatial analysis provide scalable and relatively cost‐effective approaches for studying regulatory networks. Together, these developments support multi‐omics analyses of plant development and spatial heterogeneity (Figure 4A).

Figure 4.

A schematic diagram illustrates the integration of single‐cell omics and crop science to guide the design of the next generation of crops

(A) Multimodal analysis technology: Highlights key omics techniques, including protoplast preparation from samples, high‐resolution spatial multi‐omics, and multimodal analysis of the same nucleus. (B) Predictive modeling: Illustrates the use of deep learning for cross‐modal data integration, building crop and pangenome cell atlases, and enabling the “Design‐Build‐Test‐Learn” cycle for in silico crop design. (C) Precision breeding: Shows applications of single‐cell technology, including analyzing heterosis (hybrid vigor), discovering important rare cell groups, and screening for cell‐type‐specific genetic markers (from QTL/GWAS). (D) Crop improvement: Focuses on specific improvement targets, including analyzing complex polyploid genomes, comparing cultivated crops with crop wild relatives to find beneficial genes, and analyzing plant–microbe interactions. Figure created with Biorender.com.

Sample preparation methods

The plant cell wall remains a major challenge for single‐cell sequencing. Traditional enzymatic digestion methods are often unsuitable for lignified tissues and may induce stress responses. Single‐nucleus sequencing reduces limitations associated with the cell wall; however, it may show relatively low gene‐capture sensitivity. FX‐Cell technology addresses difficulties associated with hard‐to‐digest tissues and frozen samples by combining chemical fixation with high‐temperature enzymatic digestion, making it applicable to different plant materials and experimental settings (Ming et al., 2025). However, in crop research, several practical issues remain, including the fragility of highly vacuolated cells, instrument clogging caused by large cells, the need for customized enzymatic systems for particular cell walls, and protein denaturation associated with fixation. Further development of preparation systems compatible with multi‐omics approaches and fluorescence‐activated cell sorting (FACS) will be important.

Multi‐omics co‐profiling technologies

Most plant multi‐omics studies are still conducted in parallel, which can limit the interpretation of gene regulatory networks. Simultaneous capture of transcriptomic and epigenomic information from the same nucleus has therefore become an important technical direction (Guo et al., 2025; Lee et al., 2025a). Building on multi‐omics strategies used in animal studies, multimodal integration may provide a useful framework for analyzing plant cell differentiation and development and for constructing more comprehensive plant cell maps.

Combinatorial droplet barcoding technologies

Traditional droplet‐based sequencing methods are affected by reagent consumption and throughput limitations. Combinatorial droplet barcoding technologies, such as Oak‐seq and UDA‐seq, use a two‐step labeling strategy, increasing single‐channel cell‐processing capacity to approximately 150,000 while maintaining comparable data quality (Wu et al., 2024; Li et al., 2025d). These methods are applicable to multiple analytical modalities and provide efficient and relatively cost‐effective options for large‐scale single‐cell transcriptomic analysis and gene regulatory network interpretation.

Semipermeable droplet technologies

Traditional droplets have limitations in reagent exchange and multi‐step reactions. Semipermeable droplet technologies such as CAGEs use polymer gel shells with selective permeability, allowing small molecules to diffuse while retaining large molecules (Baronas et al., 2025; Mazelis et al., 2025). These systems are compatible with flow sorting and live‐cell culture and may support more complex molecular experiments relevant to dynamic regulatory network analysis in crops.

Imaging‐free spatial analysis

Traditional spatial transcriptomics often depends on expensive imaging equipment and may be limited in scalability. New imaging‐free approaches based on in situ barcodes and mathematical inference estimate spatial positions computationally, thereby reducing costs and improving scalability (Hu et al., 2026). These methods can support efficient reconstruction of large tissue maps while maintaining useful data quality, providing an additional approach for studying organ development and spatial heterogeneity in crops.

Algorithmic developments

Algorithmic progress is increasingly supporting plant single‐cell research, extending analyses from descriptive mapping toward more predictive and system‐level modeling. Advanced computational frameworks, including generative models and transformer‐based methods, are being developed to address issues such as polyploidy and data sparsity. Through multi‐modal integration and virtual‐cell modeling, these tools may help resolve regulatory networks and cellular behaviors. This transition from observation to prediction may improve the connection between cellular heterogeneity and agronomic traits and support future data‐driven breeding strategies (Figure 4B).

Whole‐organism reference mapping

Whole‐organism reference mapping is becoming increasingly relevant for the study of complex plant traits. To address biological heterogeneity arising from WGD, repetitive sequences, and fluctuations in mapping accuracy caused by sparse data (Wang et al., 2025e), future research may increasingly move from isolated single‐tissue atlases toward more unified reference‐mapping frameworks. Inspired by models such as Pan‐human Azimuth, future efforts may focus on establishing hierarchical cell ontologies and ortholog‐based integration strategies to improve alignment and standardized annotation across platforms and species (Zhang et al., 2025b). Such computational architectures, together with predictive generative models, may help characterize gene regulatory dynamics across diverse genetic backgrounds. In this way, they may support analysis linking single‐cell heterogeneity with agronomic traits and provide a computational basis for molecular breeding.

Specialized computational frameworks for complex crop genomes

To address the challenges posed by highly repetitive, polyploid, and structurally variable crop genomes, the development of specialized computational frameworks will be important. Future algorithmic innovations are likely to combine traditional statistical models with deep‐learning architectures, including self‐attention mechanisms and graph‐based pangenomic methods (Deng et al., 2025; Zhou et al., 2025; Du et al., 2025b). By improving genomic scaffolds, orthology annotations, and ancestral haplotype context, these approaches may improve the alignment accuracy of single‐cell data in polyploid systems. They may also support analysis of cell‐type evolutionary trajectories and facilitate studies connecting cellular heterogeneity with broader evolutionary patterns.

AI‐driven denoising and imputation

To address the pronounced zero inflation and data sparsity often observed in plant single‐nucleus RNA sequencing, generative deep‐learning architectures are increasingly being explored as tools for improving data quality and recovering biological signals. Future algorithmic developments may use the nonlinear modeling capabilities of variational autoencoders (VAEs) (Yan et al., 2025a), generative adversarial networks, and diffusion models to extract latent features and impute missing values in high‐noise datasets. By incorporating heterogeneous data integration frameworks such as HIVE (Horizontal Integration Analysis using Variational AutoEncoders) to reduce batch effects (Calia et al., 2025), and transformer‐based models such as scPlantLLM to identify regulatory hubs in stress responses, researchers may more accurately reconstruct gene‐response networks under multiple stress conditions (Zeng and Yang, 2025). In addition, integration with perturbation‐prediction tools such as Monae and CoupleVAE may support analyses of the molecular basis of complex crop traits and help construct high‐resolution single‐cell atlases while examining cell‐type evolutionary trajectories (Tang et al., 2024; Wu et al., 2025).

Multi‐modal joint embedding and integration

To address heterogeneity across transcriptomic, chromatin‐accessibility, and spatial‐omics datasets, future algorithms are likely to rely increasingly on unified learning frameworks, including models such as stSCI (Shu et al., 2026) and generative approaches such as Cisformer (Ji et al., 2025). These models may support data integration, spatial deconvolution, and batch correction across modalities. In studies of plant evolutionary diversity, deep‐learning methods such as SATURN may help construct universal cell‐embedding spaces for functional alignment across species and for tracing cell‐type evolution (Rosen et al., 2024). In addition, conversational AI systems such as CellWhisperer may reduce barriers to the interpretation of large heterogeneous datasets (Schaefer et al., 2025). Together, these developments may support a transition from single‐omics description toward broader cross‐species analyses of regulatory mechanisms relevant to agronomic traits such as seed development and symbiotic nitrogen fixation.

Inference algorithms for cell‐specific gene regulatory networks

Precise inference of cell‐specific gene regulatory networks is increasingly important for studying plant developmental trajectories and environmental adaptation. Current GRN inference methods are shifting from traditional co‐expression frameworks, such as WGCNA and GENIE3, toward multidimensional models with improved biological interpretability. Approaches based on chromatin accessibility and regulatory interactions, including Single‐Cell rEgulatory Network Inference and Clustering (SCENIC) (Aibar et al., 2017), DeepTFni (Li et al., 2022a), LINGER (Lifelong Neural Network for Gene Regulation) (Yuan and Duren, 2025), and single‐cell Mutual Information‐based Network Engineering Ranger (scMINER) (Pan et al., 2025), integrate scATAC‐seq data with prior biological knowledge to characterize cell‐specific regulatory landscapes and interactions between transcription factors and target genes. In parallel, graph neural network‐based methods, including scMultiomeGRN (Xu et al., 2025), MultiGATE (Miao et al., 2025), SpaGRN (Li et al., 2025c), and GCLink (Wang et al., 2025a), incorporate spatial information and multimodal datasets to predict regulatory directionality and analyze the effects of spatial heterogeneity. Collectively, these computational approaches may support the construction of dynamic regulatory views of crop growth and stress responses and facilitate the study of evolutionary processes.

Single‐cell pre‐trained large models and cross‐species transfer learning

The integration of single‐cell pre‐trained foundational models with cross‐species transfer learning is becoming an important direction for addressing data sparsity in crop research and supporting breeding applications. By using deep‐learning architectures pre‐trained on large datasets, including Geneformer and GeneCompass, researchers can apply self‐supervised learning to capture gene interaction patterns and regulatory features. Plant‐oriented models such as scPlantLLM, together with homology‐independent cross‐species transfer strategies such as SATL, have supported automated cell‐type annotation and shown potential for cross‐modal knowledge transfer, including from transcriptomic data to spatial omics (Park et al., 2024). Future incorporation of transfer‐learning and perturbation–simulation capabilities from models such as CellPolaris may help researchers reconstruct dynamic regulatory networks using limited task‐specific data and may support in silico perturbation screening for agronomic traits (Feng et al., 2023). This shift from descriptive atlas construction toward predictive modeling may provide computational support for intelligent and precision‐oriented molecular breeding.

AI‐based virtual cells

Under the “Data + AI” framework, the construction of virtual cells with predictive capacity is becoming an important approach in precision crop breeding (Bunne et al., 2024). As computational models, virtual cells can simulate cellular responses to developmental differentiation, genetic perturbation, and environmental change, thereby helping to address the high cost and limited efficiency of traditional experiments under complex combinatorial conditions. Similar large‐scale models have already been reported in the biopharmaceutical field, such as STATE, which was trained on data from hundreds of millions of cellular perturbations (Adduri et al., 2025). In crops, virtual knockout simulations across more than one hundred thousand cells have been used to predict and validate the negative regulatory effect of RSR1 on cortical cell fate (Wang et al., 2025e). The same study also identified regulatory hubs such as OsF3H, which is associated with carbon and nitrogen metabolism, and OsLTPL120, which influences plant architecture. Looking ahead, virtual cells may increasingly integrate information from spatial omics and epigenomics to support the analysis of cell‐type‐specific regulatory networks and the identification of important regulatory factors and favorable alleles, thereby contributing to AI‐assisted molecular breeding.

Mechanism analysis

Mechanism analysis is moving from descriptive mapping toward more function‐oriented studies of plant development and evolution. By combining cross‐species single‐cell data with multi‐omics approaches, researchers can reconstruct cell‐lineage trajectories and examine the regulatory logic associated with phenotypic plasticity. This framework also supports the study of cellular specialization in metabolism and environmental responses, including host–microbe interactions. These analyses may provide a basis for cell‐type‐specific engineering and may support a shift from whole‐plant intervention toward more targeted and modular crop improvement (Figure 4C).

Analysis of the construction and evolution of a universal cell atlas

In future plant single‐cell research, the construction of pan‐cell atlases across species and tissues is likely to become an important direction for understanding plant complexity (Guillotin et al., 2023; Xue et al., 2025). This direction is not simply an expansion in species coverage, but also aims to reconstruct evolutionary trajectories of cell types through comparative single‐cell genomics.

By integrating multi‐dimensional single‐cell data from early‐diverging plants to major crops, researchers may identify core cell lineages that have been conserved over long evolutionary periods. Such cross‐species comparison can help define transcriptional regulatory units associated with specific cell identities and may clarify how vascular tissues, stomata, or specialized metabolic cells evolved from ancestral cell types through functional differentiation and modular reorganization.

A key aspect of pan‐cell atlas analysis is the study of dynamic changes in cis‐regulatory elements (CREs) among species. In the context of crop domestication and climate change, plants from different evolutionary lineages may have acquired distinct stress responses through the rewiring of regulatory networks. Single‐cell multi‐omics may help identify regulatory nodes gained or lost during evolution and may therefore help explain why some wild relatives retain strong stress adaptation, whereas cultivated varieties show different regulatory features shaped by domestication.

This type of cross‐species pan‐cell analysis may support molecular breeding by extending interpretation from the genetic level to the cellular level. By understanding the constraints and patterns of cell‐type evolution, breeders may move beyond single‐gene insertion or deletion and instead examine how favorable regulatory pathways might be reconstructed in defined cell types.

Analysis of gene regulatory networks during dynamic development

As plant single‐cell omics moves beyond the initial mapping stage, analysis of dynamic regulatory networks during developmental differentiation is becoming an increasingly important topic (Leong et al., 2025). Current dynamic analyses often rely on algorithm‐derived pseudotime. Future studies may integrate spatial transcriptomics, single‐cell multi‐omics, time‐series sampling, and live imaging to reconstruct developmental trajectories from stem‐cell states to differentiated functional cells in a more continuous temporal and spatial framework. Such analyses may help identify key transition points associated with cell‐fate change. Single‐dimensional transcriptomic data alone are often insufficient to explain differentiation mechanisms. Future studies will likely focus more strongly on multi‐omics integration, especially the joint analysis of chromatin accessibility, histone modifications, and transcriptional activity (Bartosovic et al., 2021; Wu et al., 2021). By examining how epigenetic priming precedes transcriptional activation, researchers may identify core cis‐regulatory elements and transcription factor cascades associated with differentiation direction. This transition from correlation‐based analysis toward more causal interpretation may improve understanding of how regulatory networks maintain developmental stability or generate phenotypic variation through nonlinear regulation.

In crop trait improvement, understanding dynamic regulatory networks may support molecular design methods. By using deep learning and systems biology models, researchers may simulate the responses of specific regulatory nodes to gene editing or environmental stress. This may allow more precise modulation of differentiation timing or of the proportion of specific cell types, thereby supporting modular improvement of crop yield and stress adaptation through targeted cellular programs.

Analysis of cell‐specific interactions between plants and the environment

Future plant single‐cell research is increasingly focusing on the spatiotemporal coordination of environmental interactions, moving beyond bulk‐level averages to reveal cellular division of labor. Under abiotic stress, one objective is to characterize how specialized cell lineages, such as the root epidermis and vascular tissues, use distinct ion‐transport and signaling programs to support collective adaptation. At the same time, high‐resolution sequencing can reveal transcriptional and organellar reprogramming in early host cells during symbiotic accommodation or pathogen entry (Feng et al., 2023; Liu et al., 2023b; Serrano et al., 2024), thereby offering candidate targets relevant to synthetic biology and disease resistance improvement. A further direction is dual scRNA‐seq, which may allow simultaneous analysis of host and microbiome transcriptomes and may help characterize cross‐kingdom molecular interactions relevant to plant health. Through integration of these multi‐omics datasets, future studies may support breeding strategies that consider both plants and their associated microbiota in the design of crops with improved environmental adaptability.

Analysis of biosynthetic metabolic pathways in individual cells

An important topic in plant biology is understanding how plants coordinate complex biochemical synthesis through cell‐type‐specific division of labor. Many valuable specialized metabolites are produced through spatially organized processes (Yu et al., 2023), in which precursors are transferred across multiple cell types, potentially reducing toxicity and feedback inhibition from intermediate products. By integrating single‐cell transcriptomics with spatial metabolomics, researchers can identify metabolic core cell types and regulatory features associated with energy allocation, especially in the balance between growth and defense (Zhan et al., 2024). This level of resolution may support a shift from whole‐plant overexpression strategies toward cell‐type‐specific metabolic engineering. By using specialized promoters to restrict complex pathways to cells with suitable metabolic or storage capacity, breeders may improve the accumulation of bioactive compounds while maintaining stable yield, thereby supporting crop nutritional improvement.

Crop improvement

Single‐cell omics research is expected to support a transition in crop breeding from large‐scale population screening toward more precise improvement strategies based on cellular information (Figure 4D). To achieve this goal, future research may focus on four major directions. First, it will be important to localize and functionally validate key regulatory loci for complex traits at the cellular level. Through single‐cell analytical approaches, researchers may identify specific cell populations associated with agronomic traits such as yield, quality, and stress adaptation, and may combine single‐cell isolation with micro‐ and nanoscale manipulation methods to establish workflows linking correlation with mechanistic validation. Second, efficient cell‐specific gene editing systems will need to be developed. One focus will be the identification of cell‐type‐specific promoters to direct editing tools such as CRISPR at defined times and in defined tissues, thereby reducing growth inhibition or pleiotropic effects associated with whole‐plant editing. Third, pan‐cell atlases for staple crops and economically important species may facilitate cross‐species transfer of favorable trait targets. Based on the evolutionary conservation of cell types among species, researchers may analyze shared regulatory modules associated with useful traits, thereby providing an evolutionary basis for improving complex polyploid crops or for the targeted use of traits from closely related wild species. Fourth, standardized breeding resource platforms and integrated multi‐omics databases will be needed. By integrating single‐cell multi‐omics data, large‐scale genotype information, and high‐throughput field phenotyping data, future research may promote the translation of single‐cell technology from basic discovery to breeding practice and provide support for molecular breeding of crops with improved yield stability and environmental adaptability.

CONFLICTS OF INTEREST

The authors declare no conflicts of interest.

AUTHOR CONTRIBUTIONS

L.P. and L.L. planned and designed the manuscript. F.H.W., F.Z.W., Y.W., and L.L. wrote the first draft of the manuscript, and F.H.W. and F.Z.W. prepared the figures. All authors have read and approved the final version of the manuscript.

Biographies

Wang, F. , Wang, F. , Wu, Y. , Pu, L. , and Le, L. (2026). Single‐cell insights into plant growth, adaptation, and evolution. J. Integr. Plant Biol. 68: 2889–2911.