High-resolution single-cell RNA-seq data and heterogeneity analysis of human ESCs and ffEPSCs

Abstract

This study presents a comprehensive transcriptomic analysis of feeder-free extended pluripotent stem cells (ffEPSCs) and their parental human embryonic stem cells (ESCs), providing new insights into understanding human early development and cellular heterogeneity of pluripotency. Leveraging Smart-seq2-based single-cell RNA sequencing (scRNA-seq), we have compared gene expression profiles between ESCs and ffEPSCs and uncovered distinct subpopulations within both groups. Through pseudotime analysis, we have mapped the transition process from ESCs to ffEPSCs, revealing critical molecular pathways involved in the shift from a primed pluripotency to an extended pluripotent state. Additionally, we have employed repeat sequence analysis based on the latest T2T database and identified the stage-specific repeat elements contributing to regulating pluripotency and developmental transitions. This dataset deepens our understanding on early pluripotency and highlights the role of repeat sequences in early embryonic development. Our findings thus offer valuable resources for researchers in stem cell biology, pluripotency, early embryonic development, and potential cell therapy and regenerative medical applications.

Background & Summary

Human pluripotent stem cells (PSCs), including embryonic stem cells (ESCs) and induced pluripotent stem cells (iPSCs), represent powerful tools for studying early human development and hold great promise for cell therapy and regenerative medicine. While human PSCs exhibit the capacity to self-renew and differentiate into all three germ layers, they are limited by their resemblance to later stages of post-implantation embryos, known as the “primed” state¹. There are pluripotent and totipotent states during early development²; more importantly, scientists have established human stem cell lines with different pluripotent states, including naïve state³, extended pluripotent state^4,5, and 8-cell-like state⁶. Our group reported the establishment of feeder-free extended pluripotent stem cells (ffEPSCs) and characterized the identity including the super chimeric ability to embryonic and extra-embryonic tissues⁵, positioning them closer to earlier pluripotent states and making them a unique resource for investigating earlier stages of development.

Dissecting the underlying mechanisms driving the transition between these pluripotent states is crucial for basic developmental biology and potential therapeutic applications, thus being emergingly required. However, cell heterogeneity within PSCs represents the complexity of pluripotency regulation and hinders molecular understanding. Here, using the ESCs-ffEPSCs transition system, we can explore the regulatory networks governing the transition from the primed pluripotency to an earlier pluripotent state. In this study, following the strategy outlined in Fig. 1a, we perform Smart-seq2-based high-resolution scRNA-seq to compare ESCs and ffEPSCs at the transcriptional level, uncovering distinct subpopulations within each cell type. We then perform the pseudotime analysis to track the dynamic progression from ESCs to ffEPSCs and align this transition with key stages of human early embryonic development. By combining clustering, gene expression profiling, and repeat element analysis, we have not only deepened our understanding of pluripotency regulation/transition but also provided a valuable resource for the study of early human development.

Fig. 1.

Overview of quality assessment and data analysis. (a) Schematic overview of the study workflow. 1. Experimental transition from ESC to ffEPSC and sequencing: ESCs were reprogrammed to ffEPSCs using the LCDM-IY condition as previously reported. After cell dissociation and RNA extraction, scRNA-seq libraries were prepared using the Smart-seq2 protocol for paired-end sequencing on the Illumina HiSeq 2000 platform. 2. Data preprocessing: Sequencing reads were independently processed using the GRCh38 and T2T reference genomes, respectively. Reads were aligned using HISAT2, and gene expression quantification was performed with featureCounts based on the GRCh38 reference gene annotation. Repeat sequence quantification was conducted using RepeatMasker annotations from the T2T reference genome. 3. Downstream analysis: Normalization was performed to correct the library size differences, followed by UMAP/t-SNE for visualization of overall data structure. Correlation analysis assessed consistency across cells, while heterogeneity analysis examined variability within ESC and ffEPSC populations. Gene expression was quantified using both GRCh38- and T2T-based annotations to ensure robustness and reproducibility of the dataset. (b) Mean quality scores across all data, shown as a line graph, illustrating the distribution of average sequencing quality per base. (c) Per sequence quality scores showed the quality score distribution for each sequence across all samples, displayed as a line graph. (d) Per sequence GC content, a line graph representing the GC content distribution across all sequences. (e) Sequence quality (Phred score) demonstrated the Phred score distribution as a line graph for all data. (f) Heatmap showed correlation analysis of all cells, using Pearson correlation coefficients, represented as pairwise cell relationships. (g) Scatter plot showed the number of detected gene features for each cell. (h) Scatter plot showed the RNA counts for each cell.

Methods

Maintenance and transition of human ESCs to ffEPSCs

The transition of human ESCs into ffEPSCs was performed according to our previous reports^5,7. Briefly, human ESC line H9 (from Wicell Research Institute)⁸ was used in this study, which was cultured on Matrigel-coated plates (1:100 dilution) in mTeSR1 medium (STEMCELL Technologies) supplemented with 1% penicillin-streptomycin. To initiate the transition, single ESCs dissociated with Accutase were transferred to Matrigel-coated plates in mTeSR1. On the following day, the medium was replaced with LCDM-IY, which consisted of a 1:1 mixture of knockout DMEM/F12 and neurobasal medium, supplemented with 0.5 × B27, 0.5 × N2, 5% knockout serum replacement, 1% GlutaMAX, 1% non-essential amino acids, 1% penicillin-streptomycin, 0.1 mM β-mercaptoethanol, and a combination of six chemical compounds: recombinant human LIF (10 ng/mL), CHIR99021 (1 μM), (S)-(+)-dimethindene maleate (2 μM), minocycline hydrochloride (2 μM), IWR-endo-1 (1 μM), and Y-27632 (2 μM). After one or two days, cells were harvested with TrypLE and reseeded onto Matrigel-coated plates (1:30 dilution) in LCDM-IY medium, which was refreshed daily. Established ffEPSCs were passaged every 3 days using TrypLE, whereas conventional H9 cells were passaged with Accutase every 5 days.

Smart-Seq2 cDNA generation, library preparation, and sequencing

RNA profiles of ESCs and transitioned ffEPSCs were generated through deep sequencing. Single cells were manually dissociated carefully. Each single cell was placed into lysis buffer provided by GEEK-Gene company for RNA extraction and library construction. The single-cell libraries were constructed using the Smart-seq2 protocol⁹ Briefly, first-strand cDNA synthesis was primed with UP1 primers, containing poly(dT) tails to capture mRNA, followed by pre-amplification. PCR was performed in two stages: an initial 20 cycles and an additional 9 cycles for further cDNA amplification, ensuring sufficient yield for sequencing. The cDNA was fragmented using Covaris and 3′ fragments were captured with Dynabeads. A second round of PCR was performed using NH2-blocked primers to prevent the carryover of small fragments, ensuring library integrity. Library preparation was completed with the Kapa Hyper Prep Kit, and paired-end sequencing was performed on the Illumina HiSeq 2000.

Preprocessing and alignment of Smart-seq2 data

scRNA-seq data went through quality assessment using FastQC first. For transcript expression analysis, alignment was performed using HISAT2 with the GRCh38 reference genome, which includes both protein-coding and non-coding RNA¹⁰. Format conversion was conducted using SAMtools¹¹, and transcript quantification was performed with featureCounts¹² based on GRCh38 gene annotation. For repeat sequence analysis, alignment was performed using the T2T reference genome (https://s3-us-west-2.amazonaws.com/human-pangenomics/T2T/CHM13/assemblies/analysis_set/chm13v2.0.fa.gz)¹³, and quantification was conducted using annotations from the RepeatMasker reference file (https://s3-us-west-2.amazonaws.com/human-pangenomics/T2T/CHM13/assemblies/annotation/chm13v2.0_RepeatMasker_4.1.2p1.2022Apr14.bed), which specifically captures repeat elements.

Normalization and feature selection of Smart-seq2 data

Normalization was performed using count depth scaling to 10,000 total counts per cell, resulting in the cp10k (counts per 10,000) unit. Count values were log-transformed using natural logarithm: ln(cp10k + 1). For transcript expression analysis, highly variable genes were identified using FindVariableFeatures with 4,500 genes selected, and all detected genes were used for scaling through ScaleData. For repeat sequence analysis, 1,500 highly variable repeat elements were selected, and all detected repeat features were included in the scaling process. All these steps were performed using the Seurat¹⁴ package in R.

Dimensionality reduction and clustering analysis

Dimensionality reduction analysis was conducted using the Seurat package in R. Principal component analysis (PCA) was applied to both GRCh38-based and repeat sequence data, with 40 principal components retained in each case. For clustering analysis, we used the first 20 PCs to construct FindNeighbors, followed by clustering with FindClusters. For GRCh38-based gene expression data, a resolution parameter of 1.3 was applied, while for repeat sequence-based clustering, a resolution of 1.0 was used to define cell populations.

Uniform manifold approximation and projection (UMAP) was performed using the “RunUMAP” function to visualize clustering results. And clusters were annotated using known markers representing naïve pluripotency, primed pluripotency, classical pluripotency, and ZGA-related signatures. Furthermore, we performed second-level clustering of the ESC and ffEPSC populations by gradually increasing the resolution parameter using “FindClusters” function, adjusting it until the populations were distinctly separated into two clusters, and the UMAP plots were adopted to display the results of second-level clustering using the abovementioned method.

Correlation analysis of gene expression profiles

RNA expression data was extracted from the Seurat object, and genes with low expression (mean expression <0.1) were filtered out. Metadata was merged with cell names, and row means were calculated for each unique cell type in parallel using the mclapply function. The resulting means were combined into a single matrix and further filtered to retain rows with mean expression above 0.1. A Spearman correlation matrix was then generated from the filtered mean expression matrix.

Heterogeneity analysis of cell populations

The silhouette score for each population was calculated as¹⁵: $s\left(i\right)=\frac{b\left(i\right)-a\left(i\right)}{\max (a\left(i\right)-b\left(i\right))}$, where a(i) represents the mean intra-cluster distance, defined as the average distance between a cell i and all other cells within the same cluster, and b(i) is the mean nearest-cluster distance, defined as the average distance between a cell i and the nearest neighbouring cluster. Silhouette scores range from −1 to 1, with higher values indicating cells that are well-clustered, and negative values signifying cells that may have been incorrectly clustered. For both the UMAP and t-SNE cluster, the average silhouette score was calculated to provide an overall assessment of clustering quality across the population.

Differentially expressed gene analysis

Differentially expressed genes (DEG) of different cell clusters, as well as those of different samples, were identified using the “FindMarkers” function of the Seurat package. DEGs were selected if the average logarithm of gene expression (fold-change) was 0.1 higher than the other clusters and the p-value was less than 0.05. Heatmaps were performed to show the DEGs of different cell clusters.

Gene set enrichment analysis (GSEA)

GSEA was conducted to assess whether predefined sets of genes exhibit statistically significant differences between two biological states. The analysis utilized the fgsea R package, following standard protocols. Gene expression data were ranked based on fold-change values, and predefined gene sets were derived from the top 500 feature genes associated with various stages of early embryonic development, as obtained from our analysis of in vivo early embryo data. The enrichment score was calculated to determine the extent to which each gene set is overrepresented at the extremes of the ranked list. Statistical significance was evaluated through permutation testing, with false discovery rate (FDR) correction applied to account for multiple comparisons. The results were visualized using enrichment plots, highlighting the key pathways differentially regulated between the analysed conditions.

Pseudotime trajectory inference

The Monocle R package (version 2.30.0) was used for the developmental trajectory inference. The analysis began by separating the cell clusters of interest using the “subset” command in Seurat. A CellDataSet object was then created using the “newCellDataSet” function in Monocle. After calculating size factors and estimating dispersions, genes were selected for pseudotime definition in a semi-supervised manner. Specifically, genes were filtered based on mean expression level and variability across cells, using the following criteria: mean expression > = 0.5 and dispersion empirical > = 2 * dispersion fit. Dimension reduction was performed using the “reduceDimension” function with the “DDRTree” method. Cells were ordered in pseudotime, and visualization was performed using the “plot cell trajectory” plot genes in “pseudotime” and “plot genes branched heatmap” functions.

Integration of single-cell RNA sequencing datasets

For the integration of single-cell RNA sequencing datasets, we employed Seurat v5 and utilized the integrated.rpca method for dimensionality reduction and data integration. Initially, we segmented the Seurat object by batch and applied the SCTransform normalization method to each subset. Integration anchors were identified through canonical correlation analysis (CCA), and data integration was performed using the IntegrateData function with the new.reduction = “integrated.rpca” parameter to address batch effects and align datasets. Following integration, dimensionality reduction was conducted using principal component analysis (PCA), selecting the top 10 principal components (PCs) to capture the most significant sources of variation across the integrated datasets. This step was executed using the RunPCA function. For visualization, we applied Uniform Manifold Approximation and Projection (UMAP) based on the PCA results. UMAP was computed using the RunUMAP function with dims = 1:10 to include the top 10 PCs.

Data Records

The raw data files and processed matrix files generated in this study are publicly accessible through the NCBI Gene Expression Omnibus (GEO) under the accession number GSE279316¹⁶. The processed data includes two parts: one based on the GRCh38 reference genome and the other based on repeat element annotation from the T2T reference genome.

In addition, differential gene expression results related to early embryonic cell populations and ESC versus ffEPSC comparisons are available via Zenodo at 10.5281/zenodo.15192674¹⁷ and 10.5281/zenodo.15192743¹⁸, respectively.

Technical Validation

Cell preparation, sequencing, and quality control

We constructed scRNA-seq libraries for the ESC (16 cells) and ffEPSC (25 cells) samples (Table S1). The workflow of sequencing, quality control and analysis are illustrated in Fig. 1a. Key quality control metrics were evaluated, including the “Mean Quality Score,” “Per Sequence Quality Scores,” and “Per Sequence GC Content.” All samples exhibited mean and per-sequence quality scores above 30, reflecting consistently high base quality across reads (Fig. 1b,c). The GC content followed a normal distribution, closely matching theoretical expectations, which suggests minimal contamination in the samples (Fig. 1d). Additionally, more than 90% of bases in all samples achieved a Q30 score or higher, meaning fewer than one error per 1,000 base pairs sequenced (Table S2). The sequence duplication level analysis indicated that less than 20% of the library consisted of duplicated reads, reflecting high library complexity and superior sequencing quality (Fig. 1e). Notably, Pearson correlation analysis further supported the robustness of the data, demonstrating a strong correlation between ESC and ffEPSC populations, confirming the reliability and reproducibility of the sequencing data (Fig. 1f). The number of detected genes ranged from 1 × 10⁴ to 1.6 × 10⁴, while total RNA counts ranged from 1.2 × 10⁷ to 2.0 × 10⁷, which is higher than typical 10X Genomics-based scRNA-seq data, reflecting the deep transcriptome coverage achieved with Smart-seq2 (Fig. 1g,h).

Functional validation of sequencing data

After counting the sequencing reads based on the GRCh38 reference genome, clustering was performed using UMAP and t-SNE methods (Fig. 2a). The clustering patterns revealed distinct groups corresponding to ESC and ffEPSC populations. The silhouette score was used to evaluate clustering quality. Notably, the ESC populations showed lower silihoutette coefficients compared to ffEPSC populations, suggesting higher heterogeneity of the ESC populations (Fig. 2b). Differential gene expression analysis between ESCs and ffEPSCs confirmed the expected expression patterns, supporting the reliability of the data¹⁸. ESCs exhibited high expression of primed pluripotency markers, along with genes associated with ectoderm and endoderm differentiation, consistent with their well-characterized role in maintaining pluripotency⁵. In contrast, ffEPSCs exhibited higher expression of zygotic genome activation (ZGA)-related genes, and classical/naïve pluripotency genes (Fig. 2c,d, Table S3). These results align with known biological features of ESCs and ffEPSCs, reinforcing the validity of the data.

Fig. 2.

Transcriptome analysis reveals the gene expression characteristics of ESCs and ffEPSCs. (a) Clustering of ESCs and ffEPSCs based on gene expression (GRCh38 annotation): The left panel showed clustering based on t-SNE, and the right panel showed clustering based on UMAP. (b) Table comparing Silhouette coefficient for ESCs and ffEPSCs derived from the clustering in (a). (c) Boxplot showed the expression of early embryonic development and pluripotency markers in ESCs and ffEPSCs. (d) Heatmap displayed the expression of early embryonic development markers and pluripotency markers across ESCs and ffEPSCs. (e) Clustering of ESCs and ffEPSCs based on repeat sequence expression (T2T annotation): The left panel showed t-SNE-based clustering, and the right panel showed UMAP-based clustering. (f) Table of Silhouette coefficient for ESCs and ffEPSCs based on the clustering in (e). (g) Boxplot showed the expression levels of repeat elements in ESCs, ffEPSCs, and in vivo early embryonic stages. (h) Heatmap illustrated the expression patterns of repeat elements across ESCs, ffEPSCs, and in vivo early embryo data.

Repeat sequences, including endogenous retroviral elements, play a vital role in mammalian in vivo early embryonic development^19–21. Using repeat sequences from the T2T reference genome, clustering analysis revealed higher heterogeneity in ffEPSCs, as indicated by a lower silhouette coefficient (Fig. 2e-f). Additionally, ffEPSCs displayed higher expression of 2-cell to 8-cell stage-specific repeat elements, including LTR12C, LTR14B, LTR78, and LTR6 (Fig. 2g), aligning their repeat expression pattern with the zygote to morula stages²². In contrast, the repeat element profile of ESCs was more similar to the late blastocyst stage (Fig. 2h). The alignment of repeat expression patterns with known developmental stages supports the biological relevance of the dataset. The full-length transcript coverage of Smart-seq2 allows for precise quantification of repeat elements, overcoming limitations of conventional transcriptomic approaches. Together, these findings reinforce the reliability of the data in capturing distinct transcriptional landscapes of ESCs and ffEPSCs.

Heterogeneity analysis of ESCs and ffEPSCs reveals distinct subpopulations

To investigate the heterogeneity of ESCs and ffEPSCs, we identified two distinct subpopulations within the ESC group (Fig. 3a). In contrast, ffEPSCs exhibited a more continuous distribution along the analyzed dimensions, without clearly demarcated subpopulations (Fig. 3b), suggesting lower heterogeneity within the ffEPSC population. ESC_C1(10/15 cells) is characterized by a primed pluripotent state, showing high expression of primed pluripotency genes such as DNMT3B²³, DUSP6, THY1²⁴ (Fig. 3c,d). On the other hand, ESC_C2(5/15 cells) represents an earlier developmental stage, with elevated expression of naïve pluripotency gene TFAP2C²⁵ and ZGA-related gene YAP1^26,27 (Fig. 3c,d). To further validate these findings, we processed the in vivo early embryo data²², which confirmed that ESC_C1 exhibits high expression of late blastocyst-related genes¹⁷ (Fig. 3e). GSEA also revealed that ESC_C1 is enriched in pathways associated with the late blastocyst stage (Fig. 3f). This subpopulation analysis underscores the heterogeneity within the ESC population and highlights the distinct developmental states represented by the two subgroups.

Fig. 3.

Cluster analysis of ESCs and ffEPSCs characterizes heterogeneity and gene expression profiles. (a) UMAP clustering of ESC subpopulations. (b) UMAP clustering of ffEPSC subpopulations. (c) Dot plot showed the expression of primed and naïve pluripotency markers in ESC subpopulations. (d) Violin plot showed the expression of ZGA-related, classical pluripotency, naïve pluripotency, and primed pluripotency markers in ESC subpopulations. (e) Heatmap showed gene expression patterns in vivo embryonic development data, ESC subpopulations. (f) GSEA plot highlighted genes specifically enriched in the late blastocyst stage of in vivo embryonic development for ESC subpopulations.

Repeat markers clustering reveals enhanced subpopulation distinction and heterogeneity in ffEPSCs

We next performed clustering analysis based on repeat elements. Interestingly, as shown in the UMAP visualization, ffEPSCs distinctly separated into two subpopulations, whereas ESCs did not exhibit a similar segregation. This finding implies that repeat sequences could effectively capture the heterogeneity within the ffEPSCs (Fig. 4a,b). Differential gene expression analysis was performed to explore further the gene expression heterogeneity between the two ffEPSC subpopulations. ffEPSC_C1(8/25 cells) exhibited higher pluripotency gene expression, whereas ffEPSC_C2(17/25 cells) expressed more totipotency-related genes (Fig. 4c). At the level of repeat sequences, previous studies have shown that LTR14B and MLT2A1 specifically upregulated during pre-implantation development^28,29. Consistently, our analysis revealed elevated expression of these elements in ffEPSC_C1, suggesting that the two subpopulations may correspond to different stages of early embryonic development (Fig. 4d). Heatmap analysis revealed that the gene expression pattern of ffEPSC_C1 was more similar to the morulae and late blastocyst stages, while ffEPSC_C2 resembled the expression patterns of zygote to 8-cell stages (Fig. 4e). This combined analysis of repeat elements and gene expression suggests that ffEPSCs display heterogeneity, with each subpopulation corresponding to specific stages of early embryonic development. These findings are consistent with the expected transcriptional dynamics during development, reinforcing the robustness of our data and highlighting the ability of our approach to resolve cellular heterogeneity within the ffEPSC population.

Fig. 4.

Repeat sequence analysis revealed the heterogeneity of ESCs and ffEPSCs. (a) UMAP clustering of ESCs. (b) UMAP clustering of ffEPSCs. (c) Dot plot showed the expression of primed pluripotency, naïve pluripotency and totipotency markers in ffEPSC subpopulations. (d) Boxplot of LTR14B and MLT2A1 expression in ffEPSC subpopulations. (e) Heatmap showed gene expression patterns similar to in vivo early embryonic development for ffEPSC subpopulations.

Pseudotime reconstruction of the ESC-to-ffEPSC transition and developmental mapping to early human embryogenesis

Given that ffEPSCs are derived from ESCs through a series of transition conditions, we explored the transition from ESCs to ffEPSCs through pseudotime analysis. Based on the previously defined ESC and ffEPSC subpopulations, we identified a pseudotime trajectory progressing from ESC_C1 → ESC_C2 → ffEPSC_C1 → ffEPSC_C2 (Fig. 5a). To further evaluate the robustness of the dataset, we compared the ESC-to-ffEPSC transition trajectory with gene expression patterns in early human embryonic development (Fig. 5b). Notably, the two distinct pseudotime gene expression patterns mirrored those observed during the ESC-to-ffEPSC transition (Fig. 5c, Table S4). GO enrichment analysis of these gene sets highlighted critical biological processes, including BMP signaling pathway, DNA binding, and mitosis bioprocess (Fig. 5d). For instance, the key SUMOylation gene SUMO3 demonstrated a gradual decrease in expression along the pseudotime axis. Interestingly, inhibition of SUMOylation in mouse ESCs had been demonstrated to be sufficient to drive the transition into a 2-cell-like state, highlighting the potential role of SUMO3 in regulating pluripotency and early developmental stages³⁰. Conversely, genes associated with the BMP signaling pathway (E2F1, EXT1) and mitosis-related genes such as EML1 exhibited a progressive increase (Fig. 5e). To further contextualize ffEPSCs within the developmental spectrum, we integrated our dataset with publicly available human embryo data spanning the zygote to day 14 post-fertilization^22,31,32, as well as publicly available ESC datasets²². UMAP clustering revealed that ffEPSC_C2 corresponded to the E3-E4 stages (morula to blastocyst), while ffEPSC_C1 and ESCs were positioned at the E6 stage. Additionally, two ESC subpopulations aligned with human E5-E6 stages (blastocyst to gastrulation) (Fig. 5f). This method of clustering based on differences in repeat sequence expression highlights its enhanced ability to elucidate the subtle distinctions associated with pluripotent states. Notably, the alignment between the inferred pseudotime trajectory of the ESC-to-ffEPSC transition and the transcriptional dynamics observed in early embryonic development further supports the reliability of our approach. These findings reinforce the robustness of our data and highlight the developmental relevance of ffEPSCs within the pluripotency spectrum.

Fig. 5.

Comparative pseudotime analysis of ESC-to-ffEPSC transition and early embryonic development. (a) Pseudotime trajectory clustering of ESC and ffEPSC subpopulations. (b) Pseudotime trajectory clustering of in vivo early embryonic development. (c) Pseudotime heatmap showed the expression patterns of totipotency, naïve, and primed pluripotency markers along the pseudotime trajectory. (d) GO term enrichment analysis of gene sets that display similar pseudotime trajectories between ESC and ffEPSC subpopulations and in vivo early embryonic development. The x-axis represents the number of genes in each term, and the color scale indicates p-value significance. (e) Spline plots of genes showed analogous expression patterns between in vivo embryonic development and ESC/ffEPSC subpopulations. (f) Clustering analysis of ESC/ffEPSC subpopulations with in vivo early embryo data.

Usage Notes

This dataset involves annotating and analyzing repeat sequences using the T2T reference genome. By aligning our data to the complete human genome assembly available, we performed comprehensive repeat element identification and counting using the T2T-provided repeat sequence BED file. Although the T2T reference and BED file are publicly accessible, our work incorporates an innovative aspect by applying repeat sequence analysis to ffEPSCs. This approach allowed us to detect unique patterns of repeat element expression linked to specific developmental stages and pluripotency states.

This resource offers valuable insights for researchers investigating ESCs-ffEPSCs comparisons and pluripotency-totipotency dynamics. The integration of pseudotime trajectory analysis offers a rigorous framework for tracing cell state transitions and developmental progression in pluripotent cells. Moreover, clustering based on repeat element expression enables a finer resolution of cellular heterogeneity, reinforcing the biological relevance and reproducibility of the dataset. These features establish a strong foundation for future studies exploring the regulatory role of repeat elements in early embryogenesis.

Author contributions

Research design, W.J.; Experiments conduction, R.Z. and S.W.; Smart-seq2 and library preparation, R.Z.; Bioinformatics analyses of the data, L.Z.; Paper drafting and revision, L.Z., W.J., R.Z., S.W.

Code availability

The scripts used for data processing and analysis are available for reproducibility. The code for upstream processing, including quality control, alignment, and quantification, as well as the downstream analysis pipeline, including normalization, clustering, differential expression analysis, and repeat element quantification, is publicly accessible on Zenodo³³.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41597-025-05024-6.