Introduction

Natural rubber, primarily sourced from the rubber tree (Hevea brasiliensis), is a critical raw material in various industries worldwide due to its unique mechanical properties and versatility1. However, rubber tree cultivation faces increasing challenges posed by global climate change, particularly the intensification of drought conditions, which adversely affect latex yield, tree growth, and overall plantation productivity 2,3,4. Water-deficit stress triggers a complex network of molecular responses in plants, altering gene expression patterns, hormonal signaling, and metabolic pathways, ultimately influencing plant morphology, physiology, and development5,6.

Most drought-stress studies in rubber trees have focused on bulk tissues or specific organs, providing valuable but limited insights into the cell-type-specific dynamics of stress responses7,8. Plant tissues are composed of diverse cell populations, each with specialized functions and distinct transcriptional profiles. These differences are often obscured in traditional bulk RNA-seq studies, where signals from rare or specialized cell types may be diluted by more abundant cell populations9. To surmount this limitation, single-cell RNA sequencing (scRNA-seq) technologies have recently emerged as powerful tools in plant research, enabling the comprehensive profiling of cellular heterogeneity and gene expression dynamics at unprecedented resolution10,11.

In rubber trees, the bark region is a critical interface with the external environment, encompassing tissues involved in water transport, nutrient allocation, protection, and latex production12,13. Bark tissue organization and function are tightly regulated by complex gene networks, and understanding how each cell type responds to water-deficit stress is vital for unraveling the adaptive mechanisms and improving rubber tree resilience. Previous studies in model plants like Arabidopsis and crop species have demonstrated that single-cell transcriptomic analyses can successfully identify cell-type-specific responses, highlight key regulatory genes and networks14,15,16.

Here, we present a single-cell transcriptomic atlas of rubber tree bark under water-deficit stress conditions. By applying scRNA-seq to protoplasts derived from bark tissues of rubber tree saplings subjected to different durations of water-deficit stress, we (1) construct a high-resolution cell atlas and identify key cell types, (2) elucidate cell-type-specific transcriptional changes in response to water-deficit stress, and (3) provide insights into the molecular mechanisms underpinning drought adaptation at single-cell resolution. Our study provides a valuable resource for advancing rubber research and enhancing plant tolerance to water-deficit stress for agricultural applications.

Results

Single-cell transcriptomic atlas of rubber tree bark

To systematically characterize gene expression dynamics in response to water-deficit stress at single-cell resolution, we performed scRNA-seq on protoplasts isolated from the bark of rubber tree saplings. Protoplasts were extracted from control samples (CK), as well as from bark tissues subjected to 4-day (DS1) and 7-day (DS2) water-deficit stress, respectively. Using droplet-based scRNA-seq, we generated a comprehensive single-cell transcriptomic atlas, enabling us to investigate the cellular and molecular mechanisms underpinning the response to different stages of water-deficit stress (Fig. 1A). A total of 17,994 individual bark cells were successfully labeled. cDNA libraries were then generated for each sample, followed by sequencing. The resulting data were filtered at both the cell and gene levels to ensure high quality. The median number of UMIs per cell was 3730, and the median number of genes detected per cell was 2457, which provided the foundation for further downstream analysis (Supplementary Table 1).

Fig. 1: Isolation and cluster analysis of single-cell transcriptomes from rubber tree bark in response to water-deficit stress.
figure 1

A Protocol for single-cell library preparation. Protoplasts were isolated from bark tissues of rubber tree saplings under control, 4-day water-deficit stress, and 7-day water-deficit stress conditions using enzymatic hydrolysis, and gel bead-in-emulsions (GEMs) were created for single-cell mRNA reverse transcription and cDNA library generation. B tSNE visualization of 12 cell clusters identified through unsupervised clustering analysis, representing cells from different water-deficit stress conditions. C Expression patterns of key marker genes across 12 cell clusters, the average expression levels (represented by color intensity) and the proportion of cells expressing each gene (indicated by dot size) across the identified clusters.

Full size image

Identification of key cell types in the bark of rubber tree saplings

To generate a transcriptomic atlas of rubber tree saplings bark response to water-deficit stress, we merged three experimental conditions: control, 4-day water-deficit stress, and 7-day water-deficit stress for cell clustering and annotation. A total of 17,994 single-cell transcriptomes were obtained and clustered into 12 distinct cell populations, with differences observed in the expression levels of genes across clusters (Supplementary Fig. 1). These clusters were then analyzed for differential gene expression across the different water-deficit stress conditions. The t-distributed stochastic neighbor embedding (t-SNE) method was used to visualize the clustering results (Fig. 1B).

Genetic databases currently lack a comprehensive set of cell-specific markers with well-characterized biological functions and expression profiles for rubber trees. To accurately annotate the 12 cell clusters identified in this study, we searched for homologous genes in rubber trees by referencing established single-cell markers (Supplementary Table 2). For instance, the transcription factors TDR and WOX4, known to regulate meristematic activity and auxin responsiveness, have been identified as markers for the cambium17,18. Similarly, the epidermis-specific genes FDH, CUT1, and KCS19, each implicated in cuticle formation and very-long-chain fatty acid biosynthesis, were used to annotate epidermal tissue19,20,21. The HD-ZIP IV transcription factor GL2, known to regulate trichome development, served as a reliable marker for trichome cells22. In the phloem, the phloem-specific sucrose transporter SUC2, along with SEOB and AGL12—genes involved in sieve element function and vascular differentiation—were identified as relevant markers9,20,23. For xylem cells, UXS5, LAC17, IRX9, and CTL2, each associated with cell wall biosynthesis and lignification, served as robust markers to distinguish xylem tissue10,24,25,26. Additionally, the small rubber particle protein SRPP was identified as a key latex-specific marker due to its role in rubber biosynthesis27, while the GRAS-family transcription factor SCR was used to define endodermal cells28. Based on the expression patterns of marker genes (Fig. 1C and Supplementary Fig. 2), we identified 7 distinct cell types in rubber tree bark, including epidermis cells (cluster 2), trichome cells (cluster 3), cambium cells (cluster 4), phloem cells (cluster 5), xylem cells (cluster 6), latex cells (cluster 7), and endodermis cells (clusters 8). The cell types for clusters 0, 1, 9, 10, and 11 could not be determined due to the lack of specific marker genes and were labeled as undefined cells (Fig. 1B). This single-cell transcriptomic atlas offers valuable insights into the functional roles of these cell types and provides a framework for analyzing cellular heterogeneity in rubber tree bark.

Water-deficit stress induced transcriptomic changes vary among cell types

To elucidate the cell type-specific transcriptional responses to water-deficit stress in rubber tree bark, we performed differential gene expression analysis across the identified cell populations under control, 4-day water-deficit stress, and 7-day water-deficit stress conditions. A total of 17,994 single-cell transcriptomes were analyzed, encompassing 6126 cells from CK, 6107 from DS1, and 5761 from DS2 samples. All 12 identified cell clusters contained cells from both control and drought-stressed saplings, indicating that cell type identities were not affected by water-deficit stress (Fig. 2A). However, the proportions of cell types varied significantly across the different conditions (Fig. 2B and Supplementary Table 3). Cambium cells (clusters 4), phloem cells (cluster 5), xylem cells (cluster 6), latex cells (cluster 7), and endodermis cells (cluster 8) were more abundant in the control group compared to drought-stressed plants, while epidermis (cluster 2) and trichome cells (cluster 3) showed higher proportions in the drought-stressed samples at both 4-day and 7-day stress conditions. Specifically, water-deficit stress led to a reduction in the proportion of cambium, phloem, xylem, latex, and endodermis cells, while the abundance of epidermis and trichome cells increased, suggesting a shift in cellular composition as the plants adapt to water deficit.

Fig. 2: Differentially expressed genes across cell types under water-deficit stress conditions.
figure 2

A Distribution and numbers of cells for bark samples in the control and water-deficit stress groups. B Cell numbers in 12 cell clusters. C DEGs between the control and water-deficit stress treatments across cell types. D Petal diagram illustrating the overlapping DEGs among different cell types.

Full size image

To further investigate the cell type-specific transcriptional responses to water-deficit stress in rubber tree bark, we identified differentially expressed genes (DEGs) for each cell type by comparing the CK sample with the DS1 and DS2 samples. DEGs were defined using a threshold of |log2FC | ≥ 0.36 and FDR < 0.05 (Supplementary Data 1 and 2). Among the known cell types, cambium, epidermis, and trichome cells exhibited more DEGs compared to the other cell types (Fig. 2C). Notably, only 171 and 182 DEGs were shared across all seven major cell types in two comparisons (Fig. 2D), indicating the presence of a core transcriptional response to water-deficit stress. In contrast, most DEGs were specific to individual cell types and drought durations, reflecting cell type-specific and time-dependent transcriptional dynamics under water deficit conditions.

To investigate the functional responses of genes to water-deficit stress, GO analysis was performed to identify enriched biological processes at different stages of stress exposure. At 4 days of water-deficit stress, upregulated GO terms were significantly enriched in processes such as response to abscisic acid and alpha-amino acid metabolic process, reflecting the early activation of water-deficit sensing and metabolic adjustment pathways (Supplementary Fig. 3A). Meanwhile, downregulated terms included organelle subcompartment and cellulose biosynthetic process, suggesting an early suppression of photosynthesis and cell wall biosynthesis to conserve resources under limited water availability (Supplementary Fig. 3B). By the 7th day of water-deficit stress, upregulated GO terms shifted towards phosphatase activity and organonitrogen compound catabolic process, indicating a transition to enhanced catabolic processes aimed at sustaining energy balance and cellular homeostasis during prolonged stress (Supplementary Fig. 3C). In contrast, downregulated terms, such as chloroplast and photosystem, indicated continued suppression of photosynthesis-related pathways (Supplementary Fig. 3D). Notably, GO terms explicitly related to response to water and response to abscisic acid were enriched across both stages, illustrating their critical roles in mediating water-deficit stress responses.

Pseudotime trajectory analysis of rubber tree bark under water-deficit stress

To explore the dynamic transcriptional responses of rubber tree bark cells under water-deficit stress, we performed a pseudotime trajectory analysis of cells from control, 4-day water-deficit stress and 7-day water-deficit stress conditions. The analysis aimed to capture cell fate progression and identify key genes involved in water-deficit stress responses across different time points.

The expression matrices of bark cells from CK, DS1, and DS2 samples were used to construct pseudotime trajectories. The results revealed three major branch points within the pseudotime trajectory, dividing the cells into three distinct states (Fig. 3A). Notably, as processing time increased, cells shifted from earlier states (state 1) to more advanced states (state 2) along the pseudotime trajectory (Fig. 3B). To identify DEGs along the pseudotime trajectory, we performed clustering analysis on the expression patterns of all cells within the trajectory. The DEGs were categorized into five clusters based on their expression trends over time (Fig. 3C and Supplementary Data 3). A significant proportion of these DEGs were associated with processes related to stress response (e.g., detoxification, response to water deprivation, and abscisic acid) and metabolic adaptation (e.g., maltose and disaccharide metabolic processes, asparaginyl-tRNA aminoacylation, and mycotoxin biosynthesis), highlighting their potential roles in the bark’s response to water deficit.

Fig. 3: Pseudotime trajectory analysis of rubber tree bark cells under water-deficit stress.
figure 3

A Pseudotime trajectory. Each dot represents a single cell along the pseudotime axis. B Distribution of cells along the trajectory. Cells are distributed across different states, with colors representing various cell states. C Heatmap of significantly changed genes. The heatmap shows the expression patterns of genes that significantly changed along the pseudotime trajectory.

Full size image

Expression of ABFs across different cell types

In higher plants, the phytohormone abscisic acid (ABA) has been well established as a key regulator of water-deficit stress responses29,30. ABA accumulation triggers the transcriptional activation of numerous stress-responsive genes, among which the ABRE-binding factors (ABFs) serve as essential bZIP transcription factors31,32 (Fig. 4A). Previous studies have demonstrated that ABF2, ABF3, and ABF4 are key transcription factors in ABA signaling, playing crucial roles in water-deficit stress tolerance33,34.

Fig. 4: Expression patterns of ABFs in response to water-deficit stress across different cell types.
figure 4

A Schematic representation of ABA signaling and the role of ABF transcription factors in water-deficit stress response. B Expression of ABF genes—ABF2, ABF3, and ABF4—across seven distinct cell types. C Single-cell-level analysis showing upregulation of ABFs in all seven cell types under water-deficit stress. D The fold change and statistical significance of upregulation for each ABF gene across the cell types. E In situ hybridization analysis of ABF2, ABF3, and ABF4 expression in rubber tree bark under drought stress. Scale bars are indicated in each panel.

Full size image

In this study, we investigated the expression patterns of three ABF genes—ABF2, ABF3, and ABF4—which were expressed in all seven cell types (Fig. 4B). Our single-cell-level analyses revealed that the ABFs were significantly upregulated across the examined cell clusters: cambium cells, phloem cells, xylem cells, latex cells, endodermis cells, epidermis cells, and trichome cells (Fig. 4C). Although the overall expression of ABFs was enhanced across all seven cell types, the magnitude and statistical significance of upregulation varied for each ABF gene (Fig. 4D and Supplementary Data 4). For instance, ABF2 was consistently and strongly upregulated across all clusters, emphasizing its central role in the drought response. ABF3 showed moderate upregulation, primarily in xylem, phloem and epidermis cells. ABF4 was broadly upregulated across all clusters during DS1 and maintained elevated expression in cambium, trichome, epidermis, latex, and xylem cells under DS2. These findings underscore the divergent yet coordinated roles of ABFs in regulating ABA-dependent stress responses across multiple cell types.

To further validate the transcriptional upregulation of ABF genes under drought stress, we conducted in situ hybridization to detect the spatial expression of ABF2, ABF3, and ABF4 in rubber tree bark tissues (Fig. 4E and Supplementary Fig. 4). The results confirmed that all three ABFs were clearly induced by water-deficit treatment, showing comparable levels of hybridization signals across samples. These observations are consistent with the scRNA-seq data and support the conclusion that ABF2, ABF3, and ABF4 are broadly upregulated in response to water-deficit stress at the tissue level.

Discussion

This study presents a comprehensive single-cell transcriptomic atlas of Hevea brasiliensis bark in response to water-deficit stress, providing unprecedented insights into the cellular and molecular mechanisms underpinning drought adaptation in rubber trees. By employing droplet-based scRNA-seq on protoplasts isolated from control, 4-day, and 7-day drought-stressed bark tissues, we successfully profiled 17,994 individual cells, identifying 12 distinct cell clusters. Among these, seven major cell types—cambium, epidermis, trichome, phloem, xylem, latex, and endodermis—were annotated based on homologous marker genes from model plants. This single-cell atlas not only elucidates the functional roles of these cell types but also highlights the cellular heterogeneity and dynamic transcriptional responses to water-deficit stress in rubber tree bark.

The establishment of the single-cell atlas revealed significant cellular diversity within the bark tissue. The identification of cambium, epidermis, trichome, phloem, xylem, latex, and endodermis cells aligns with the complex anatomical structure of rubber tree bark, which comprises various cell types involved in growth, transport, defense, and rubber biosynthesis35,36. The use of specific marker genes, such as TDR and WOX4 for cambium cells, FDH, CUT1, and KCS19 for epidermis cells, and SRPP for latex cells, ensured accurate annotation and provided a reliable foundation for subsequent analyses. However, five clusters remained undefined due to the lack of specific markers, indicating the presence of potentially novel or less-characterized cell types that warrant further investigation.

Water-deficit stress induced significant alterations in the cellular composition of rubber tree bark. Under both 4-day and 7-day water-deficit conditions, there was a noticeable decrease in the proportions of cambium, phloem, xylem, latex, and endodermis cells, while epidermis and trichome cells became more abundant. This shift suggests an adaptive strategy where the bark enhances its protective barrier by increasing epidermal and trichome cells to reduce water loss and bolster defense mechanisms. The reduction in cambium and vascular cell proportions may reflect a downregulation of growth and transport activities to conserve energy and resources under water-limited conditions. Similar trends have been observed in other plant species subjected to abiotic stress, where changes in cell type abundance contribute to overall stress resilience37,38.

Differential gene expression analysis across cell types revealed that cambium, epidermis, and trichome cells exhibited the most extensive transcriptional changes in response to water-deficit stress. This highlights their pivotal roles in orchestrating the bark’s adaptive responses. GO enrichment analysis further delineated the functional dynamics of these responses. At the early stage of water-deficit stress (4 days), there was an upregulation of genes involved in abscisic acid (ABA) signaling and amino acid metabolism, alongside a downregulation of photosynthesis-related processes and cellulose biosynthesis. This indicates an initial phase where the plant actively senses water deficit and adjusts its metabolic pathways to conserve energy and maintain cellular homeostasis39,40. By the seventh day, the transcriptomic landscape shifted towards the upregulation of phosphatase activity and catabolic processes, suggesting a transition to sustained metabolic adjustments necessary for prolonged stress conditions. Concurrently, the continued suppression of photosynthesis underscores the plant’s commitment to minimizing energy expenditure under severe water-deficit stress.

Using pseudotime trajectory analysis, we observed that rubber tree bark cells transition from an early stress-responsive state to more advanced phases of drought adaptation, consistent with earlier single-cell findings in other plant species37,38. The presence of three major branch points and five distinct gene expression clusters underscores the temporal complexity of water-deficit stress responses, in which detoxification, water deprivation response, and abscisic acid signaling genes become increasingly critical over time. This progressive activation of stress-related pathways aligns with the notion of an early stress perception phase followed by a more pronounced defense phase41.

A focal point of our study was the expression of ABF family genes (ABF2, ABF3, and ABF4), which are critical components of the ABA signaling pathway. ABA is a central phytohormone regulating plant responses to water-deficit stress29,30. Our single-cell and in situ hybridization analysis revealed that all three ABF genes were upregulated under water-deficit stress across the majority of cell types. Specifically, ABF2 was significantly upregulated in cambium, phloem, xylem, latex, endodermis, epidermis, and trichome cells, underscoring its central role in orchestrating ABA-dependent drought responses. ABF3 and ABF4 also showed upregulation, albeit with varying magnitudes across different cell types, indicating a nuanced and potentially redundant regulatory mechanism. This differential expression suggests that while ABF2 may serve as a primary regulator, ABF3 and ABF4 could provide additional layers of control, enhancing the flexibility and robustness of the drought response.

In conclusion, our single-cell transcriptomic atlas of rubber tree bark under water-deficit stress provides an unprecedented view of the cellular and molecular mechanisms involved in drought adaptation. Through scRNA-seq analysis, we identified seven distinct cell types and captured their dynamic transcriptional responses to different drought durations. The cell type-specific expression profiles revealed significant shifts in cellular composition and highlighted the central roles of abscisic acid signaling and stress-responsive genes across different stages of water-deficit stress. The pseudotime trajectory analysis further revealed the temporal progression of cellular responses, with key genes associated with detoxification, metabolic adaptation, and stress response becoming progressively activated. This study offers valuable insights into the molecular basis of drought tolerance in rubber trees and lays the foundation for future research aimed at improving drought resilience in rubber plantations.

Materials and methods

Plant materials and growth conditions

In this study, we utilized Hevea brasiliensis GT1 saplings, a widely cultivated variety known for its robust growth and rubber yield. The saplings were obtained from the experimental plantation of the Yunnan Institute of Tropical Crops (Xishuangbanna, China) and cultivated in a controlled greenhouse environment. Plants were maintained at 25–31 °C during the day and 20–25 °C at night, with a relative humidity of approximately 80–90% and a 13-h light/11-h dark photoperiod. Saplings were watered regularly to ensure optimal growth and development until the onset of the water-deficit stress experiments.

Water-deficit stress treatment

Six-month-old GT1 saplings with uniform stem diameter and height were randomly assigned to three treatment groups: control (CK), 4-day water-deficit stress (DS1), and 7-day water-deficit stress (DS2). To ensure comparability across groups, all samples were harvested on the same day. For the DS2 group, watering was withheld for seven consecutive days; for the DS1 group, drought treatment was initiated three days after the DS2 group so that both could be sampled simultaneously. For the control group, saplings were watered daily with 200 ml per plant to maintain optimal soil moisture. At the end of each treatment period, bark tissues from the mid-stem region were harvested in the early morning to minimize transcriptional variations caused by diurnal rhythms.

Protoplast isolation and Single-cell library preparation

The extraction of rubber tree bark protoplasts was performed according to a standard protocol42. An appropriate amount of single-cell suspension was mixed with 0.4% trypan blue staining solution at a 9:1 ratio, and then cell counting and viability calculation were performed using the Countess® II Automated Cell Counter. To ensure the viability was ≥90%, the cell concentration was adjusted to the desired concentration (no lower than 1000 cells/μL).

For single-cell transcriptomic analysis, we employed 10X Genomics 3’ RNA sequencing, which utilizes short-read sequencing and microfluidic technology to simultaneously profile the transcriptome of 500–10,000 cells per sample. The process begins by mixing the cell suspension with a gel bead containing barcode information and enzymes. This mixture is encapsulated within droplets of oil and surfactant in the microfluidic “double-cross” system, forming Gel Beads-In-Emulsions (GEMs). The GEMs are collected in a reservoir, where the gel beads dissolve to release the barcode sequences, reverse transcribe cDNA fragments, and label the samples. The oil droplets are then broken, and cDNA is amplified using PCR. The resulting products from all GEMs are pooled together to construct a standard sequencing library. For library construction, cDNA is fragmented into 200–300 bp segments, and traditional library preparation steps, including the addition of sequencing adapters (P5) and primer (R1), are performed, followed by PCR amplification. The prepared library is then sequenced using the Illumina NovaSeq X Plus platform in paired-end mode. Read 1 contains 16 bp barcode information and 10 bp UMI (Unique Molecular Identifier) sequences for cell identification and expression quantification. Read 2 includes cDNA fragments that are aligned to the reference genome for gene expression analysis.

In situ hybridization validation

Stem tissues were fixed in paraformaldehyde fixative at 4 °C for at least 12 hours, then dehydrated through an ethanol series, embedded in paraffin, and sectioned at 4 μm thickness. Sections were mounted on slides and baked at 62 °C for 2 hours. After dewaxing and rehydration, slides were treated with proteinase K (20 μg/mL) at 40 °C to improve probe penetration.

Pre-hybridization was carried out at 40 °C for 1 hour, followed by hybridization with DIG-labeled RNA probes specific to ABF2, ABF3, and ABF4 at 40 °C for 45 minutes in a humid chamber. After hybridization, slides were washed in SSC buffers of decreasing concentrations at 40 °C and stained with NBT/BCIP to visualize probe signals. Slides were examined under a light microscope.

scRNA-seq data analysis

Initial quality control was performed using the Cell Ranger Single Cell Software Suite (v6.1) for sample demultiplexing, barcode processing, and single-cell 3’ gene counting. Raw sequencing data were first demultiplexed into FASTQ format using bcl2fastq software. Quality control of FASTQ files was conducted using FastQC, and subsequent alignment was done against the Nucleotide Sequence Database using the NCBI Basic Local Alignment Search Tool (BLAST). Low-quality sequences, including those with incomplete barcodes or unique molecular identifiers (UMIs), were removed.

The raw scRNA-seq data were aligned to the Hevea brasiliensis genome using our previously published reference genome (GCA_001654055.1)43. Barcodes and UMIs were filtered and corrected to ensure high accuracy in the gene-barcode matrix. CellRanger’s pipeline was used for UMI counting, and only confidently mapped, non-PCR duplicates with valid barcodes and UMIs were included in the analysis. Data from different libraries were normalized by equalizing the read depth before merging to minimize batch effects. For higher-depth libraries, samples were normalized to the sample sequencing depth. Cell quality was further filtered using Seurat (v4.0.4)44, with the following criteria: (1) gene counts >3000 per cell, (2) UMI counts >12,000 per cell, and (3) mitochondrial gene percentage <15%.

Differential gene expression analysis and cell clustering

Gene-barcode matrices were normalized and filtered using canonical correlation analysis (CCA), followed by clustering in 2D space using tSNE. Cells with similar expression profiles were grouped into 12 unsupervised clusters based on marker gene expression. Differentially expressed genes (DEGs) were identified using Seurat, with a threshold of |log2FC | ≥ 0.36 and a minimum gene expression in ≥10% of cells within any group. The MAST45 obstacle model was applied to test for statistical significance. P-values were corrected for multiple testing using the BH method in Seurat, and genes with corrected p-values ≤ 0.05 were considered significantly differentially expressed. These significant genes were then used for further clustering analysis, distribution assessment, and enrichment analyses. DEGs results were visualized with heatmaps using Seurat (v4.0.4), and specific gene expression patterns across clusters were represented with tSNE plots.

Single-cell pseudotime analysis

Pseudotime trajectories of single cells were analyzed using Monocle (v2.20.0)46, which constructs cell trajectories based on gene expression data. The gene-cell matrix was reduced to two dimensions, enabling the ordering of cells along a continuous pseudotime axis. This approach revealed a tree-like structure of the cell trajectory, consisting of branching points and terminal ends, allowing for the visualization of cell progression and differentiation paths within the reduced dimensional space.

Gene functional enrichment analysis

DEGs were mapped to GO terms from the GO database47, and a hypergeometric test was used to determine the enrichment. P-values were corrected for false discovery rate (FDR) with a threshold of FDR ≤ 0.05. GO terms meeting this criterion were considered significantly enriched in the DEGs.

Statistics and reproducibility

All numerical data and p values are reported in the Supplementary Tables. Statistical significance is defined as p ≤ 0.05. Sample sizes are specified and explained in each corresponding figure legend.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.