Transcriptome and epigenome dynamics of the clonal heterogeneity of human induced pluripotent stem cells for cardiac differentiation

Abstract

Human induced pluripotent stem cells (hiPSCs) generate multiple clones with inherent heterogeneity, leading to variations in their differentiation capacity. Previous studies have primarily addressed line-to-line variations in differentiation capacity, leaving a gap in the comprehensive understanding of clonal heterogeneity. Here, we aimed to profile the heterogeneity of hiPSC clones and identify predictive biomarkers for cardiomyocyte (CM) differentiation capacity by integrating transcriptomic, epigenomic, endogenous retroelement, and protein kinase phosphorylation profiles. We generated multiple clones from a single donor and validated that these clones exhibited comparable levels of pluripotency markers. The clones were classified into two groups based on their differentiation efficiency to CMs—productive clone (PC) and non-productive clone (NPC). We performed RNA sequencing (RNA-seq) and assay for transposase-accessible chromatin with sequencing (ATAC-seq). NPC was enriched in vasculogenesis and cell adhesion, accompanied by elevated levels of phosphorylated ERK1/2. Conversely, PC exhibited enrichment in embryonic organ development and transcription factor activation, accompanied by increased chromatin accessibility near transcription start site (TSS) regions. Integrative analysis of RNA-seq and ATAC-seq revealed 14 candidate genes correlated with cardiac differentiation potential. Notably, TEK and SDR42E1 were upregulated in NPC. Our integrative profiles enhance the understanding of clonal heterogeneity and highlight two novel biomarkers associated with CM differentiation. This insight may facilitate the identification of suboptimal hiPSC clones, thereby mitigating adverse outcomes in clinical applications.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1007/s00018-024-05493-9.

Introduction

Human induced pluripotent stem cells (hiPSCs) are widely used as in vitro models for rare diseases and cell therapies, offering insights into individual genetic backgrounds and the potential for differentiation into diverse cell types. Especially, hiPSC-derived cardiomyocytes (CMs) hold remarkable promise for drug development and drug toxicity testing [1, 2]. The reprogramming of somatic cells involves various methods, such as those based on lentiviruses, episome, and Sendai viruses, resulting in the formation of distinct colonies known as clones. Previous studies have highlighted the heterogeneity among these clones, including variability in epigenetic remodeling [3] and differentiation [4]. Studies by the Panopoulos group have revealed clone-specific effects of differentially methylated CpG sites, which affect the expression patterns of genes and key long non-coding RNAs associated with pluripotent stem cell function [5, 6]. Pluripotency gene expression and cell morphology [7] were empirically used to select high-quality clones, but these criteria may not ensure high differentiation capacity. Several studies have attempted to differentiate multiple hiPSC clones to identify those suitable for clinical applications [4, 8]. However, because differentiation rates are typically assessed post-differentiation, there is a growing need to evaluate the differentiation capacity of undifferentiated hiPSCs to minimize resource expenditure on suboptimal clones. To the best of our knowledge, few efforts have been made to profile the clonal heterogeneity, which emphasizes the need to identify predictive biomarkers to control cell fate determination at the undifferentiated stage.

Variations in differentiation outcomes have been reported based on factors such as reprogramming methods, cell origin, and molecular status [9–11]. For example, our previous paper investigated the cardiac differentiation potential of hiPSCs derived from various reprogramming methods, with each line including two clones [9]. Their findings indicated that the most significant differences in differentiation efficiency were associated with the reprogramming methods rather than the clonal variation. Other previous studies have also predominantly focused on parental cell types or line-to-line variation [10, 12, 13], which may not adequately explain the impact of clonal heterogeneity within hiPSC from a single donor.

Here, we aimed to investigate gene expression patterns and chromatin accessibility landscapes in hiPSC clones by integrating RNA sequencing (RNA-seq), assay for transposase-accessible chromatin by sequencing (ATAC-seq), and phospho-kinase array analysis. Our results unveiled distinct molecular signatures among clones with different differentiation potentials. We identified TEK and SDR42E1 as novel predictive biomarkers for selecting optimal hiPSC clones, providing comprehensive insights into predicting CM differentiation.

Materials and method

Blood collection and ethics statement

Human blood samples were obtained from healthy individuals. Ethical approval for the use of peripheral blood mononuclear cells (PBMCs) and hiPSCs was obtained from the Institutional Review Board of Sungkyunkwan University (IRB no. 2019-11-016-002 and 2020-01-020-001) and Samsung Medical Center (IRB no. 2016-11-025-015). Our study was conducted in accordance with the principles outlined in the Declaration of Helsinki, following the acquisition of written consent. For the reprogramming, PBMCs were obtained using the CPT mononuclear cell preparation tube (BD, 362761) and the provided protocol.

Reprogramming and cell culture

Human PBMCs were cultured and expanded in StemPro-34 SFM (Gibco, Waltham, MA, USA) supplemented with StemPro-34 Nutrient Supplement, 0.1 × penicillin–streptomycin (Gibco), 100 ng/mL stem cell factor (Peprotech, Cranbury, NJ, USA), 100 ng/mL FMS-related tyrosine kinase 3 ligand (FLT3L, Peprotech), 20 ng/mL interleukin-3 (IL-3, Peprotech), and 20 ng/mL interleukin-6 (IL-6, Peprotech). Reprogramming was performed by following the feeder-free protocol of the CytoTune™-iPS 2.0 Sendai Reprogramming Kit (Invitrogen, Carlsbad, CA, USA). The generated human iPSCs were maintained in TeSR™-E8™ medium (Stemcell Technologies, Vancouver, Canada) on Matrigel (Corning Inc., Corning, NY, USA)-coated plates in a 5% CO₂ incubator at 37 °C. All clones were cultured using EDTA dissociation (Invitrogen) following published protocols.

Karyotype analysis

hiPSCs were cultured in T25 flasks coated with Matrigel in TeSR™-E8™ medium for 3 days. The karyotype analysis was conducted to standard protocols for cell harvest and chromosomal Giemsa (G)-banding in GenDix Co., Ltd.(Seoul, Korea).

Direct cardiac differentiation of hiPSCs

A previously published protocol involving the modulation of the canonical Wnt signaling pathway was used for hiPSC-CM differentiation [14]. At day 0, 4–6 µM CHIR99021 (Tocris, Bristol, UK) in RPMI1640, supplemented with B27 minus insulin (Gibco), was administered. On days 1–2, 1 mL B27 was added to the insulin differentiation medium to expose the cells to a gradient concentration of CHIR99021. On day 3, after aspiration of the previous medium, 5 µM endo-IWR1 (Tocris) in B27 minus insulin differentiation media was administered. On days 5 and 7, the B27 minus insulin differentiation medium was administered after aspiration of the previous medium. During days 9–15, after aspiration of the previous medium, RPMI1640 supplemented with B27 (Gibco) was administered. To compare the differentiation efficiencies of the clones, we did not proceed with metabolic purification on days 11–15. At day 15, beating cardiomyocytes were treated with TrypLE Select Enzyme 10 × without phenol red (Gibco) and incubated at 37 °C for 10–30 min. Single-cell suspensions were transferred to a 15 mL conical tube containing RPMI1640, and the cells were centrifuged at 300 × g for 5 min. After aspiration of the supernatant, the cell pellet was resuspended in RPMI1640 supplemented with B27 containing 10% Knock Out Serum Replacement (Gibco) and 10 µM Y-27632 (Biogems, Westlake Village, CA, USA) for further analysis.

Flow cytometry

Cells were fixed in 4% paraformaldehyde (Thermo Fisher Scientific, Waltham, MA, USA) for 15 min at room temperature, washed once with washing buffer (1% fetal bovine serum/phosphate-buffered saline [PBS]), and centrifuged at 300 × g for 5 min. The cells were permeabilized with 0.1% Triton X-100 in PBS for 15 min at room temperature, washed twice with washing buffer, and centrifuged at 300 × g for 5 min. Incubation with a diluted primary antibody (1:20 dilution in 3% bovine serum albumin (BSA)/PBS) was performed for 1 h on ice in the dark. The cells were washed twice with washing buffer and centrifuged at 300 × g for 5 min at 4 °C. Cell pellets were resuspended in 500 µL of 3% BSA/PBS and filtered through a cell strainer tube (Falcon, Corning, Inc.). Samples were analyzed using a FACS Aria3 SORP flow cytometer (Becton Dickinson, Franklin Lakes, NJ, USA) by BIORP of the Korea Basic Science Institute. For each experiment, 10,000 events were collected; all data were analyzed using the FlowJo software (Becton Dickinson).

Field potential measurements using a Multielectrode array (MEA)

To coat the surface of CytoView MEA 24-well plates (Axion BioSystems, Atlanta, GA, USA), hESC-qualified Matrigel was diluted in RPMI1640 and applied to the wells. On day 15, hiPSC-derived CMs were seeded on Matrigel-coated plates. See supplementary materials online for detailed method.

Quantitative reverse transcription polymerase chain reaction (qRT-PCR)

Total RNA was extracted from all hiPSC clones using the easy-BLUE™ Total RNA Extraction Kit (iNtRON, Seongnam, Korea) for qRT-PCR. Reverse transcription of 1 µg total RNA was performed using Maxime™ RT Premix (iNtRON). qPCR was performed using KAPA SYBR^® FAST Master Mix 2× Universal (KAPA BIOSYSTEMS, Wilmington, MA, USA), a CFX86™ Real-Time System (Bio-Rad, Hercules, CA, USA) or a CFX Connect™ Real-Time System (Bio-Rad) according to the manufacturer’s instructions. Normalized expression levels for each gene (ΔΔCt) were calculated based on the levels of the housekeeping genes GAPDH or RNA18S1. The primer sequences are listed in the Supplementary Table 1.

Immunocytochemistry

Cells were seeded on Matrigel-coated cell chamber slide. Next day, cells were fixed in 4% paraformaldehyde (Thermo Fisher Scientific). Cells were seeded on Matrigel-coated cell chamber slide. Next day, cells were fixed in 4% paraformaldehyde (Thermo Fisher Scientific) for 15 min at room temperature, washed thrice with PBS, permeabilized with 0.1% Triton X-100/PBS for 15 min at room temperature, and washed thrice with PBS. The cells were blocked with 1% BSA in PBS for 1 h at room temperature on a shaker. Incubation with diluted primary antibody was carried out overnight at 4 °C. After 16 h, the cells were washed thrice for 10 min each with PBS at room temperature. Alexa-conjugated secondary antibodies were diluted 1:600 in blocking buffer and incubated with cells for 1 h and 30 min at room temperature on a shaker. After incubation, the cells were washed thrice with PBS for 10 min. Nuclei were counterstained with Hoechst stain (Gibco). Imaging was performed using a TCS SP8 confocal microscope (Leica, Wetzlar, Germany). The antibodies are listed in the Supplementary Table 2.

Protein extraction and immunoblotting

Cells were lysed in the PRO-PREP protein extraction solution (iNtRON) with 1× phosphatase inhibitor cocktail solution (GenDEPOT, Katy, TX, USA) after washing once with cold 1× PBS. Lysates were homogenized and centrifuged at 13,000 rpm at 4 °C for 20 min. Protein concentration was measured, and equal amounts of each protein sample were loaded onto a polyacrylamide gel and separated via sodium dodecyl sulfate-polyacrylamide gel electrophoresis. Cells were lysed in the PRO-PREP protein extraction solution (iNtRON) with 1× phosphatase inhibitor cocktail solution (GenDEPOT, Katy, TX, USA) after washing once with cold 1× PBS. Lysates were homogenized using ultrasonic homogenizers for 3 s at 12% amplitude and centrifuged at 13,000 rpm at 4 °C for 20 min. Protein concentration was measured on a spectrophotometer using the Bradford method. Equal amounts of each protein sample were loaded onto a polyacrylamide gel and separated via sodium dodecyl sulfate-polyacrylamide gel electrophoresis. The proteins were subsequently transferred onto polyvinylidene difluoride membranes (Millipore, Burlington, MA, USA) using a wet transfer system (Bio-Rad). The membranes were blocked with 5% skim milk in Tris-buffered saline with Tween 20 and incubated overnight at 4 °C with diluted primary antibodies. Horseradish peroxidase (HRP)-conjugated secondary antibodies were added in a species-dependent manner and incubated at room temperature for 1 h. The HRP signals were detected using AbSignal (Abclon, Seattle, WA, USA) or SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific) using a ChemiDoc Imaging system (Bio-Rad). Bands were quantified using the FIJI software. The antibodies are listed in the Supplementary Table 2.

Human phospho-kinase arrays

Cells were lysed in lysis buffer, containing 1 × protease inhibitor cocktail solution (GenDEPOT), after washing once with cold 1 × PBS, and phosphorylation analysis was performed according to the manufacturer’s instructions (R&D Systems, Minneapolis, MN, USA). Signals were detected using a ChemiDoc Imaging System (Bio-Rad), and the density of the dots was quantified using the Quick Spots Tool (Ideal Eyes Systems, Bountiful, UT, USA).

Sample preparation of bulk mRNA-sequencing and ATAC-sequencing

Cells were collected for bulk RNA-seq and ATAC-seq analysis. Briefly, hiPSCs were dissociated and collected in PBS. Cell counting was performed using a trypan blue solution (0.4%). For each hiPSC clone, 1.7 × 10⁶ cells were used for ATAC-seq, and the remaining were used for RNA extraction. See supplementary materials online for detailed method.

Analysis of bulk mRNA-sequencing data

Paired-end reads were aligned to the Gencode GRCh37.p13 reference genome (https://www.gencodegenes.org) using STAR (v.2.7.3a) with the default settings. Gene counts were quantified using featureCounts based on the hg19 GENCODE annotation reference. DESeq2 was used for differential expression analysis and to obtain normalized counts. See supplementary materials online for detailed method.

Analysis of long terminal repeats (LTRs)

RepeatMasker (RMSK) from the UCSC Genome Table Browser of hg19 were utilized in this study. STAR was employed to align raw FASTQ files to hg19 rmsk files with specific multimapper parameters, including winAnchorMultimapNmax 200, outFilterMatchNminOverLread 0.66, outFilterMismatchNoverLmax 0.05, and outFilterMultimapNmax 100. TEs were quantified following an established RNA sequencing analysis protocol using featureCounts. DESeq2 was then applied to identify differentially expressed TEs with a significance threshold of P_adj < 0.05 and log₂|FC|> 1. To gain deeper insights into the TE characteristics, each TE count was initially treated as repeat and presented in a bar plot. Subsequently, a heatmap was generated to annotate the TEs, with TE counts serving as repeat values.

Analysis of bulk ATAC-sequencing data

FASTQ files were aligned with the human genome (GRCh37.p13), and peak calling was executed using the PEPATAC pipeline. Adapter sequences were removed using Skewer, reads were mapped to the GRCh37 build of the human genome using bowtie2, and reads were sorted and isolated using Samtools while removing duplicates with samblaster. MACS2 was applied for peak calling, extending the peak summits by 250 bp in both directions and filtering based on the ENCODE blacklist. See supplementary materials online for detailed method.

Statistical analysis

Graphic presentation and statistical analysis were performed using GraphPad Prism software v10.1.1 (GraphPad Software, San Diego, CA, USA), and results are presented as the mean per group ± standard deviation (SD) unless otherwise indicated. Between-group comparisons were performed by unpaired Student’s t-tests, Welch’s t-tests, one-way analysis of variance (ANOVA), or Brown-Forsythe and Welch ANOVA. For one-way ANOVA and Brown-Forsythe or Welch ANOVA, multiplicity-adjusted (Tukey or Dunnett T3) P values are reported throughout, with the threshold of alpha set at 0.05. The p-values are indicated in each figure for statistically significant comparisons (*, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001). The specific tests used for each figure and number of replicates are mentioned in the figure legends.

Results

Human iPSC clones reprogrammed from a single donor represented comparable levels of pluripotency marker expression and cell proliferation

We generated multiple clones from a single donor using Sendai virus to investigate the heterogeneity among hiPSC clones (Figs. 1A and S1A). Approximately 20 colonies were observed, and individual colonies were manually picked and subsequently expanded on Matrigel-coated culture plates. The resulting six clones (CL1-6: clone 1–6) formed stable colonies devoid of differentiated cells and displayed typical hiPSC morphology [7], including prominent nucleoli and distinct borders (Fig. 1B). All clones were maintained under the same conditions to minimize the effect of subculture processes. Immunofluorescence confirmed the expression of pluripotency-associated markers, including SOX2, SSEA4, and OCT4 in all clones (Figs. 1C and S1B). The protein levels of SOX2 and OCT4 were not significantly different among the clones (Fig. 1D, F, and G). In addition, the six clones exhibited similar proliferation rates (~ 60%) and normal karyotypes (Figs. 1E, H, and S1C). In summary, we obtained six hiPSC clones that expressed pluripotency markers and exhibited comparable growth rates.

Fig. 1.

Generation and characterization of human induced pluripotent stem cell (hiPSC) clones. A Experimental design for the evaluation of clonal heterogeneity in hiPSC. B Phase-contrast images of the clones. Scale bar = 50 μm. C Representative immunostaining images for SOX2 (green), SSEA4 (red), OCT4 (yellow), and Hoechst (blue) staining of the clones. Scale bar = 50 μm. D Representative immunoblotting images for OCT4, SOX2, and H3. E Representative images for 5-ethynyl-2′-deoxyuridine (EdU) incorporation assay. Scale bar = 50 μm. F–G Quantification of signal intensity of the immunoblotting for OCT4 (F) and SOX2 (G). Data show two technical replicates per group, and all values were normalized to H3. H Quantification of EdU + proliferating cells stained with Hoechst. Data show six technical replicates per group. All statistical analysis were performed using one-way analysis of variance (ANOVA) with Tukey’s multiple comparisons test (F, H) or Brown-Forsythe and Welch ANOVA with Dunnett’s multiple comparisons test (G). Data are presented as bar graphs and error bars with mean ± standard deviation

Differentiation capacity of cardiomyocytes varied among clones

To assess the differentiation potential of the clones, we induced cardiomyocytes (CMs) from six hiPSC clones following an established protocol [14]. All clones were simultaneously treated with the same concentration of CHIR99021 (5 µM), a GSK-3 inhibitor. The proportion of TNNT2 (cardiac troponin-T)-positive cells were compared via flow cytometry between day 12 to 15 of differentiation (Fig. 2A). CL1 and 3 consistently exhibited robust differentiation rates (> 60%), whereas CL2 showed a lower differentiation rate (< 50%). CL4, 5, and 6 showed variability among batches, thereby these clones were excluded in further analysis (Fig. 2B, C, and S2A). The differentiated cells derived from CL1 and 3 showed spontaneous beating by day 15 (white arrows in Fig. 2D), whereas CL2 showed a higher proportion of non-beating cells (Videos S1 and 2). Immunofluorescence confirmed the coexistence of CMs (TNNT2-positive) and non-CMs (TNNT2-negative) in CL2 (Figs. 2E and S2B). Using quantitative reverse transcription-polymerase chain reaction (qRT-PCR), we compared the expression levels of CM markers (TNNT2 and MYH7/6: myosin heavy chain 7/6) and a non-CM marker (VIM: vimentin) (Fig. 2F). VIM was upregulated in CL2, indicating the presence of a mixed proportion of CMs and non-CMs. To mitigate the potential effects of CHIR99021 concentration, differentiation was also induced at 4 and 6 μM. At 4 μM of CHIR99021, the results were consistent with previous observations, while differentiation did not occur in any of the clones at 6 μM of CHIR99021 (Fig. S3A–C and Videos S3–S5).

Fig. 2.

Evaluation of the differentiation efficiency and functionality of human induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CMs). A Experimental scheme for the evaluation of cardiac differentiation capacity. B Flow cytometry of cardiac troponin T (TNNT2) expression in hiPSC-CMs showed that the clones had different efficiencies, based on which they were divided into two groups—PC (blue) and NPC (red). Three or four technical replicates per group are shown. C Representative flow cytometric histograms showing the TNNT2 + (%) cells of hiPSC-CMs when hiPSC clones (CL1- 3) were treated with 5 µM of CHIR99021. Gray peaks represent the non-stained control. D Phase-contrast images of hiPSC-CMs derived from PC (left, CL1) and NPC (right, CL2); beating CMs were represented with white arrows. Scale bar = 1000 μm. E Representative images of Immunostaining for TNNT2 (green) and Hoechst (blue) staining of hiPSC-CMs (day 15). F Quantitative reverse transcription-polymerase chain reaction (qRT-PCR) of CM (TNNT2, MYH7, and MYH6) and non-CM (VIM) markers in hiPSC-CMs. All values were normalized to GAPDH. Data show three or four technical replicates per group. G Field potential analysis of hiPSC-CMs using a multielectrode (MEA) system. Quantification of the beating rate, spike amplitude, and spike slope. Data represent three technical replicates per group. H Representative heatmap images of conduction map. The cardiac beating begins where the plot is blue (shortest delay from the beat origin) and ends where the plot is red (longest delay from the beat origin). All statistical analysis were performed using unpaired Student’s t-tests or Welch’s t-tests; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Data are presented as bar graphs and error bars with mean ± standard deviation (B, G) or standard error of mean (F)

Functional assessment of CMs was performed using a multielectrode array system, and data were acquired using CMs at day 20, which was 5 days post-plating. We detected active beating of CMs in CL1 and 3, but not in CL2 (Fig. S3D), possibly due to the low proportion of CMs (Fig. 2B-F). There was no significant difference of beating rate between CL1 and 3, whereas CL3 exhibited an increase in spike amplitude and slope (Fig. 2G and S3D). The conduction map showed a delay in conduction, which was no significant difference of conduction velocity, maximum delay, and propagation consistency between CL1 and 3, with no observed arrhythmias (Fig. 2H and S3E). Recent studies have reported that hiPSC-derived CMs show conduction velocities ranging from 0.04 m/s to 1.0 m/s [15, 16], and CL1 and 3 showed the velocities within this range except for CL2. Overall, the hiPSC clones (CL1-3) were classified into two groups based on their differentiation capacity, with CL2 showing a low capacity, resulting in mixed proportions of CMs and non-CMs—CL1 and 3 as productive clone (PC) and CL2 as non-productive clone (NPC).

Phosphorylation state of ERK1/2 signaling pathway was elevated in NPC

Signaling pathways such as Wnt, BMP, and ERK play crucial roles in cardiac development [17, 18]. Recent studies have highlighted the effect of the phosphorylation status of signaling kinases in hiPSCs on their differentiation potential [19, 20]. To compare the signaling profiles of the hiPSC clones (PC and NPC), we conducted a human phospho-kinase array. We examined whether the phosphorylation levels of Wnt/β-catenin or GSK-3α/β differed between the two groups, because we used a GSK-3 inhibitor, a regulator of the Wnt/β -catenin pathway, to induce CM differentiation (Fig. 2A). However, the phosphorylation of GSK-3α/β (Ser21/Ser9) did not show significant differences among the clones (Fig. 3A and B). We also assessed the nuclear translocation of β-catenin using immunofluorescence, which revealed no discernible differences across all clones (Fig. S4A).

Fig. 3.

The phosphorylation state of ERK1/2 signaling pathway was elevated in non-productive clone (NPC). A Chemiluminescent images of human phosphokinase array revealed distinct phosphorylation levels of ERK1/2 (T202/Y204 and T185/Y187), eNOS (S1177), p38α (T180/Y182), and p53 (S46) among clones (CL1-3). B Mean pixel density of 37 kinase phosphorylation sites and two related total proteins, normalized to reference spots. Statistical analysis was performed using multiple unpaired Student’s t-tests, and the two-stage step-up method of Benjamini, Krieger, and Yekutieli was used for the correction of multiple comparisons; *q < 0.05. Data represent two technical replicates per clone (PC: CL1 and 3, NPC: CL2). C Volcano plot of multiple unpaired Student’s t-tests showing that four phosphorylation sites were significantly different between PC and NPC; the largest difference is highlighted with a pink box. D Representative immunoblotting images for ERK1/2, phosphorylated ERK1/2 (T202/Y204), and H3. T: threonine, Y: tyrosine, S: serine

The phosphorylation levels of ERK1/2 (Thr202/Tyr204 and T185/Y187), p38α (Thr180/Tyr182), and eNOS (Ser1177) were significantly elevated in NPC, whereas those of p53 (Ser46) were increased in PC (Fig. 3B and C). Of these pathways, previous studies have reported that ERK1/2 [18] and p38α [21, 22] pathways are involved in CMs differentiation. The difference in the phosphorylation level of ERK1/2 between PC and NPC was the most pronounced in this study (Fig. 3C), and we confirmed the increased phosphorylation of ERK1/2 in NPC using immunoblotting (Fig. 3D). Taken together, the level of phosphorylated ERK1/2 was elevated in NPC compared to that in PC, whereas no differences were observed in the regulation of the Wnt pathway.

Distinct transcriptomic profiles revealed clonal heterogeneity

Recent studies have reported that variations in the gene expression patterns of human pluripotent stem cells can affect their potential to differentiate into various cell lineages [23]. To analyze the transcriptomic profiles of the three clones (hiPSC #1 CL1-3), we performed bulk RNA-seq, followed by quality control of the raw read data (Fig. S5A). We compared our data with previously published data using principal component analysis (PCA) to ensure that all clones exhibited transcriptional signatures consistent with those of hiPSCs [24]. Our clones were closely clustered with hiPSCs, contrasting with other tissues, based on data from the Genotype-Tissue Expression (GTEx) portal (Fig. 4A).

Fig. 4.

Transcriptomic analysis of human induced pluripotent stem cell (hiPSC) clones using bulk RNA-sequencing. A Principal component analysis (PCA) plot showing that generated hiPSC clones were in a cluster of previous iPSC data and distant from data from other tissues, based on data from the Genotype-Tissue Expression [GTEx] portal. B Heatmap displayed differentially expressed genes (DEGs, log₂|FC|> 1 and adjusted P-values (P_adj) < 0.05) between PC and NPC. C, D Gene Ontology (GO) analysis of DEGs in the biological process (BP), molecular function (MF), and cellular component (CC) categories. The emapplots (C) and bar plots (D) showed significantly enriched GOs in the two groups. The blue and red dots and bars indicate upregulated GOs in PC and NPC, respectively. E Chord diagram illustrating inter-relationships between the genes related to the GOs identified in C. F The volcano plot depicted the DEGs shown in B, and the dashed lines indicated log₂|FC|= 1 and P_adj = 0.05. The top genes identified in E are marked with green dots

To investigate the molecular signatures that affect clonal heterogeneity, especially the differentiation capacity of CMs, we identified 416 differentially expressed genes (DEGs) between the two groups. Among these, 240 genes were upregulated in PCs, whereas 176 were upregulated in NPC (log₂|FC|> 1, adjusted P-value [P_adj] < 0.05) (Fig. 4B). The majority of the DEGs were annotated as protein-coding genes (90%, Fig. S5B). Gene Ontology (GO) analysis revealed that PC exhibited enrichment in the development of embryonic organs, including the heart and kidney, along with DNA-binding transcription activator activity (Fig. 4C, D, and S5C). In contrast, NPC was enriched for muscle cell development, vasculogenesis, transmembrane receptor protein tyrosine kinase activity, and actin filaments. Both groups exhibited significant enrichment of the ERK1/2 signaling pathway. Furthermore, gene set enrichment analysis (GSEA) indicated the enrichment of signaling pathways regulating the pluripotency of stem cells in PC, whereas NPC was enriched for cell adhesion molecules and vascular smooth muscle contraction (Fig. S5E and F). To figure out which genes are involved in clone-specific GO terms and the ERK1/2 signaling (Figs. 3A–D, 4C, D, S4C, and D), we created a chord plot showing the relationships between DEGs and enriched pathways (Fig. 4E and F). For example, NPC showed upregulated levels of the transmembrane protein receptor tyrosine kinases TIE1 and TEK (also known as TIE2), which are key regulators of vasculature development (e.g., endothelial cells) [25, 26] and ERK1/2 activation [27]. In contrast, PC showed upregulated levels of FOXC1, a member of the FOX family involved in embryonic tissue development [28] and differentiation of early CMs [29]. These findings highlight the transcriptomic profiles between PC and NPC with varying differentiation potential for CMs.

Dysregulation of endogenous retroelements modulated nearby gene transcription in clones

Endogenous retroelements (EREs), comprising long terminal repeats (LTRs) and non-LTR transposons (Fig. 5A), constitute almost half of the human genome [30]. Previous studies have highlighted the effect of ERE expression patterns in pluripotent stem cells on various characteristics of hiPSCs, including cell morphology, pluripotency, and lineage commitment [31–33]. To examine the expression patterns of EREs among the clones, we mapped transcript reads onto copies of transposable elements (TEs) (Fig. 5B). The total read counts of TE contents were quantified for each clone (4.6 × 10⁶, 4.7 × 10⁶ and 4.7 × 10⁶, respectively), showing consistent patterns across all clones (Fig. 5C). Differentially expressed TEs were identified between PC and NPC (log₂|FC|> 1 and P_adj < 0.05), with LTRs showing the highest proportions (CL1: 83.9%, CL2: 86.6%, and CL3: 84.9%). Notably, NPC exhibited an increase in the total number of significant copies of TEs when compared with PC (Fig. 5D). As copies of repeat classes are distributed across the human chromosomes [34], we compared the overall and locus-specific expression of LTRs to identify differentially expressed subtypes and their chromosomal locations. We detected 15 significant LTR subtypes (log₂|FC|> 1 and P_adj < 0.05) between PC and NPC. Of these, 5 subtypes (HERVK-int, HERVK11-int, LTR88c, THE1D-int, and MER52C) were highly expressed in PC, while 10 subtypes (HERVI-int, MER52C, HERV4_I-int, MER51B, LTR88b, etc.) were highly expressed in NPC (Fig. 5E). Additionally, 103 ERE loci were significantly dysregulated, with 49 of these loci belonging to LTRs from five subfamilies—ERV1 (63.3%), ERVK (2%), ERVL (8.2%), ERVL-MaLR (24.5%), and gypsy (2%) (Table S3).

Fig. 5.

Assessment of variations in expression levels of endogenous retroelements (EREs). A Classification of human transposable elements (TEs). B Experimental scheme for the analysis of TEs. C Bar plot showing the proportion of total read counts of TE contents identified in RNA-seq. D Bar plot showing the proportion of differentially expressed TEs (log₂|FC|> 1 and P_adj < 0.05), indicating that the long terminal repeats (LTRs) showed the highest proportion. Normalized counts of the differentially expressed LTRs are presented. E Heatmap showed the expression levels of significant LTR subtypes (e.g., LTR88c, HERVI-int, and HERV4_I-int) between PC and NPC. F Heatmap displayed the expression of genes located in proximity to the identified EREs in E. The genes corresponding to DEGs are marked with *. (G-H) Loci of the genes (ANKRD2, CLEC2A, OC90, and HHLA1) were identified in the University of California, Santa Cruz (UCSC) browsers and visualized in Integrative Genomics Viewer (IGV) tracks. Two examples of dysregulated MER77 and HERVI-int integrants and expression level of the closest genes: OC90 and HHLA1 for MER77 on chromosome 8 (G), and ANKRRD2 and CLEC2A for HERVI-int on chromosome 1 and 12, respectively (H)

Given that EREs act as cis-regulatory elements to regulate the expression of neighboring genes and generate chimeric transcripts [35], we investigated differentially expressed loci near gene transcripts. We identified 21 genes adjacent to these loci (Fig. 5F and Table S4), of which 8 genes (ANKRD45, LYST, CLEC2A, RAB17, WFDC3, CEP85L, HHLA1, and OC90) are protein-coding. Among these, 5 genes (CLEC2A, RAB17, CEP85L, HHLA1, and OC90) were identified as DEGs (Fig. 4B). Consistent with previous findings [32, 34, 35], the upregulation of HHLA1 and OC90 was affected by nearby ERE loci (Fig. 5G). Unexpectedly, no reads were mapped to exons 1–4 of CLEC2A, suggesting a new transcript potentially transcribed from LTR10B1 (chr12: 10,055260–10,055,764) and HERVI-int (chr12: 10,055,800–10,059,965) (Fig. 5H and Table S4). We observed a differentially expressed ERE loci near ANKRD45, which was not identified as a DEGs (chr1: 173,607,756–173,611,588). Collectively, our analysis of EREs revealed that profiles of EREs, especially LTRs, may contribute to clonal heterogeneity by regulating nearby genes and generating new transcripts.

Clones revealed differential chromatin accessibility and susceptibility to transcription factors

The reprogramming of hiPSCs entails global remodeling of the epigenetic state [36], where epigenetic variations, such as DNA methylation and chromatin state, play pivotal roles in the differentiation potential of hiPSCs [37]. To detect genomic chromatin accessibility between PC and NPC, we performed an ATAC-seq. An average of 9.6 × 10⁶ fragments per sample was obtained, with a consistent distribution of peaks across genomic regions in all clones (Fig. 6A). In total, 3.1 × 10⁵, 3.0 × 10⁵, and 3.6 × 10⁵ open chromatin peaks were identified in CL1, 2, and 3, respectively (Fig. 6B). PC exhibited elevated chromatin accessibility in the vicinity of transcription start site (TSS) regions (TSS ± 3 kb) compared to NPC, indicating a predisposition for active transcription (Fig. 6C). We identified 2869 differentially accessible regions (DARs) by comparing ATAC-seq peak tag counts. Among these, 2142 regions were detected in PC (log₂|FC|> 0.5 and P < 0.05), whereas 727 regions were detected in NPC (Fig. 6D). Next, we examined the genomic locations of DARs, and the distal regions (> 10 kb from the TSS) accounted for 43% and 37% in PC and NPC (Fig. 6E), respectively. The promoter-proximal regions (TSS ± 3 kb) accounted for 23% and 18% of DARs in PC and NPC. Although it has been reported that the chromatin states in distal regions can affect gene expression [38, 39], we focused on how DARs in promoter-proximal regions affect gene expression status.

Fig. 6.

Different chromatin statuses of human induced pluripotent stem cell (hiPSC) clones were identified using ATAC-sequencing. A Quality control of ATAC-seq. Stacked bar charts showed the distribution of peaks across genomic regions. B The circle plot of genome-wide chromatin accessibility identified in ATAC-seq. C The enriched heatmap displayed the normalized score of ATAC-seq peaks within ± 3 kb of the transcription start site (TSS). D Heatmap of differentially accessible regions (DARs, log₂|FC|> 0.5 and P < 0.05) between PC and NPC. E Pie charts illustrated the distribution of DARs across genomic regions. F The volcano plot depicted the genes annotated as promoter in (E); n = 454 for PC (red) and n = 128 for NPC (blue). The dashed lines indicated log₂|FC|> 0.5 and P < 0.05. G Gene Ontology (GO) analysis of the genes shown in F. H Loci of the genes (TEK and PDGFRB) were identified in University of California, Santa Cruz (UCSC) browsers and visualized in Integrative Genomics Viewer (IGV) tracks. I Heatmap displayed gene expression levels related to identified GOs (GO:1,905,772, 0007507) in G. The genes corresponding to DEGs are marked with *

We identified 454 and 128 genes with increased open chromatin around their promoters in PC and NPC, respectively (Fig. 6F). For example, chromatin accessibility (TSS ± 3 kb) of DPPA2, a key factor in resetting the epigenome to a pluripotent state during reprogramming [40], was increased in PC. We performed GO analysis to investigate the biological processes associated with the genes harboring DARs. We found that DAR-associated genes in PC were enriched in processes related to positive regulation of DNA/Nucleic acid-templated transcription (Fig. 6G), consistent with findings from our RNA-seq analysis (Fig. 4C). Despite the limited differentiation potential of NPC towards CMs, DAR-associated genes in NPC as well as PC were enriched in heart development (Fig. 6G). Specifically, 11 genes in NPC were related to heart development (GO: 0007507) (Fig. 6H and I). The expression levels of PDGFRB and TEK, which are implicated in endothelial cell differentiation [41] and atrioventricular canal development [42], were significantly upregulated in NPC. Consequently, this increased expression in NPC, driven by DARs in promoter-proximal regions, may impede CM differentiation, resulting in non-CMs with increased expression of VIM.

Next, we performed HOMER (Hypergeometric Optimization of motif EnRichment) de novo motif analysis using DARs to identify clone-specific transcription factor (TF) motifs. The ETV4 and GATA5 motifs were the top motifs in PC, whereas the TEAD2 and FOSL2 motifs were found in NPC (Fig. S6A). The ETV4-encoding ETS-domain-containing TF family plays a role in the maintenance of pluripotency in stem cells [43], and some GATA factors (GATA4, GATA5, and GATA6) are expressed in the precardiac mesoderm during heart development [44]. Although we further associated enriched motifs with the mRNA expression levels of TF families that could bind these motifs, we did not find a correlation between the motifs and the expression levels of their family members at the undifferentiated stage (Fig. S6B). Nevertheless, our results showed that the clones presented different susceptibilities to specific TFs, and they could regulate gene expression during cardiac differentiation, similar to the GATA family. Collectively, our epigenomic profile suggests that variability in the epigenetic state may influence the changes in gene expression, thereby contributing to clonal heterogeneity.

TEK and SDR42E1 were identified as marker genes for clonal heterogeneity of hiPSC-CM differentiation

Recent studies have proposed reliable marker genes based on comprehensive analysis of transcriptional and epigenetic modifications [12, 13, 39]. To identify candidate genes that could predict cardiac differentiation, we assigned chromatin accessibility at the promoter-proximal regions to the nearest genes in the hg19 reference genome. By overlapping DEGs and DARs between the two groups, we identified a total of 56 genes with accessible regions that were significantly differentially expressed between PC and NPC (Fig. 7A). From these, we selected 14 candidate markers for PC (DPPA2, FOXK1, FOXO4, GBX2, HEY2, NANOG, PTCH1, PDGFα, and UTF1) and NPC (TEK, PDGFRβ, SDR42E1, TXNRD2, and JPH4) based on their basal expression levels in hiPSCs (Fig. 7A and B).

Fig. 7.

TEK and SDR42E1 were identified as marker genes for predicting cardiac differentiation capacity. A Venn diagram showed the overlap between DEGs and DARs identified via RNA-seq and ATAC-seq peaks, specifically in protein-coding genes. The positively correlated genes were colored in red (upregulated in PC) or blue (NPC), while the negatively correlated genes were colored in green (opposite). B Representative images of ATAC-seq (left) and normalized counts obtained from RNA-seq (right) for FOXK1 and TEK; upregulated in PC (blue) and NPC (red). The bar graphs for normalized counts represent two biological replicates (PC) and one biological replicate (NPC). C Experimental scheme for the validation of candidate genes. D Flow cytometry of TNNT2 expression in hiPSC-CMs at day 15 after differentiation. Three (hiPSC #2) or five (hiPSC #3) technical replicates per group are shown. E Immunostaining for TNNT2 (green) and Hoechst (blue) staining of hiPSC-CMs. (F) qRT-PCR of TEK and SDR42E1 in hiPSC clones (#2 and #3). All values were normalized to RNA18S1. Data show three technical replicates per group. All statistical analysis were performed using unpaired Student’s t-tests; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Data are presented as bar graphs and error bars with mean ± standard deviation (D) or standard error of mean (F)

To further refine the candidate genes, we included two additional hiPSC lines derived from healthy individuals, hiPSC #2 and #3. Each line contained two clones (CL1 and CL2) that were differentiated into CMs (Fig. 7C). The proportion of TNNT2-positive cells was assessed via flow cytometry. In hiPSC #2, CL1 showed a higher percentage of TNNT2-positive cells compared to CL2 when treated with 6 μM of CHIR99021, while both clones of hiPSC #3 showed no significant differences between each other (Fig. 7D, E). As expected, the expression of CM markers (TNNT2 and MYH7) was increased in CMs derived from #2 CL1, whereas hiPSC #2 CL2 showed increased expression of the non-CM marker VIM (Fig. S7A). Based on these data, we categorized the hiPSC lines into two groups: those with differing differentiation capacities within clones (hiPSC #2) and those with similar capacities (hiPSC #3).

In hiPSC #2, the non-productive markers TEK and SDR42E1 exhibited increased expression (4.3- and 23-fold increases, respectively) in hiPSC #2 CL2 (Fig. 7F). The expression of another non-productive marker, PDGFRB, was significantly increased in hiPSC #2 CL1, which did not align with the expression pattern observed in hiPSC #1. In contrast, all productive markers showed no significant differences between the two clones (Fig. S7B). These data indicated that most of the candidate genes, except for TEK and SDR42E1, could be false-positive markers and showed no consistent predisposition toward cardiac differentiation potential. On the other hand, in the clones of hiPSC #3, which had similar differentiation capacities, the expression of TEK and SDR42E1 was not significantly different between the two clones, confirming the correlation between TEK and SDR42E1 expression and differentiation capacities (Figs. 7F, S7B).

The angiopoietin-1/TEK axis plays a crucial role in vascular development, including the development of endothelial and muscle cells [25, 26, 45]. SDR42E1, a member of the short-chain dehydrogenase/reductase (SDR) family, is associated with multiple cellular processes, including NAD(P)-dependent oxidoreductase activity [46]. Although the direct effects of the SDR family on stem cell development have not been reported, the family is known to be involved in the metabolism of various compounds, including retinoids [47] and lipids, which sustain pluripotency and modulate myocardial cell differentiation [19, 48, 49]. In summary, our transcriptome and epigenome profiling data suggest that TEK and SDR42E1 serve as predictive biomarkers implicated in the cardiac differentiation of hiPSC clones.

Discussion

Human iPSCs represent a potential source for regenerative medicine and disease modeling, and stem cell therapy using hiPSC-CMs holds promise for clinical applicability [50]. For clinical applications, maintaining consistently high purity and yield across batches is important. Therefore, researchers have worked to optimize these aspects. For example, empirical efforts have been made for the selection of hiPSC clones expressing high levels of pluripotency marker genes [7] or the optimization of the CHIR99021 concentration [51]. However, to the best of our knowledge, few references serve as standards for the clonal selection. The efficient selection of clones with high differentiation capacity at an undifferentiated stage may significantly contribute to this field, and thereby predictive markers for clonal heterogeneity need to be investigated. In the present study, we established clonal heterogeneity profiles and identified TEK and SDR42E1 as predictive biomarkers of cardiac differentiation.

All the six generated clones expressed pluripotency markers (OCT4, SOX2, and SSEA-1) and exhibited similar proliferation rates. However, we observed that CL2 consistently exhibited a lower CM differentiation efficiency than the other clones. Moreover, the expression of the three germ layers showed similar patterns upon comparing the average expression levels of the marker genes using RNA-seq. These results suggest that conventional evaluation is not enough for predicting the differentiation potential of hiPSC clones and highlight the need for alternative indicators of clonal heterogeneity. We analyzed RNA-seq, ATAC-seq, endogenous retroelement, and phosphokinase array to identify the biomarkers required to minimize the likelihood of false positives. We identified 56 DEGs with an open chromatin state in TSS-proximal regions. By including two additional hiPSC lines, we validated our results using qRT-PCR and identified TEK and SDR42E1 as novel predictive biomarkers. Furthermore, there were no significant differences of OCT4 and SOX2 expression levels between two clones of hiPSC #2, suggesting comparable levels of pluripotency. Our data demonstrate that TEK and SDR42E1 were upregulated in the clones with reduced CM differentiation capacity. TEK, also known as TIE2, is a transmembrane receptor tyrosine kinase and a known factor affecting vascular development via angiopoietin-1 (Ang1). Ang1-TEK reportedly forms a complex with the extracellular matrix (ECM) [27], and NPC showed enrichment of cell–cell contacts, such as actin filaments and basement filaments. We found that NPC showed upregulated phosphorylation levels of ERK1/2 and eNOS, which are downstream signaling factors of ECM-anchored TEK [52]. SDR42E1, a member of the short-chain dehydrogenase/reductase family, is involved in steroid biosynthesis; a previous study reported that mutations in the SDR42E1 gene are associated with oculocutaneous genital syndrome [53]. However, its role in pluripotent stem cells has not yet been reported. Taken together, our findings indicate that the expression patterns of TEK and SDR42E1 serve as markers for the differentiation potential of hiPSC clones toward CMs, establishing them as new references for clonal heterogeneity.

Although previous studies have reported the molecular signatures of hiPSCs that regulate their differentiation capacity (e.g., CMs), those studies focused on variations in individual hiPSCs. For example, the modulatory effects of SALL3 [12] and CXCL4/PF4 [13] on the differentiation lineage in various hiPSCs have been demonstrated, indicating that SALL3 expression levels correlate with ectoderm differentiation, and PF4 levels correlate with mesoderm differentiation. Although these markers reflected variations affecting the cardiac differentiation propensity, we observed an increase in SALL3 expression levels in PC, whereas PF4 levels exhibited no significant differences among the clones (data not shown).

Given that high expression of TEs is a characteristic of naïve pluripotency [54], we investigated whether PC and NPC exhibit features of primed or naïve pluripotency at the transcriptional level. We compared our RNA-seq data with human primed and naïve ESCs (hESCs) from previously published datasets [55]. PCA and heatmap analysis confirmed that the expression patterns of naïve signature genes, including SUSD2, TFCP2L1, KLF5, and KLF4, in hiPSC clones were distinct from those observed in both primed and naïve ESCs (Fig. S8A and B). Furthermore, our epigenome profile data showed that 71% (1,384 of 1,948) and 67% (449 of 668) of the total DARs were distal regions (> 10 kb from the TSS) in PC and NPC, respectively. Some studies have reported that the chromatin state of distal regions as well as regions proximal to the TSS induces gene expression variability [38]. Similarly, we identified specific subtypes of EREs, such as HERVI-int, which were dysregulated among clones and were located across the genome. Previous studies have reported that copies of EREs act as distal enhancers [56]. Although it is not clear how clone-specific EREs in intergenic regions regulate gene expression, we identified several clone-specific intergenic regions, suggesting a possible role of open distal regions in gene regulation.

Our study has some limitations. Firstly, the number of clones analyzed for integrative analysis is restricted due to reprogramming efficiency. Distinguishing clones into PC and NPC categories is challenging due to batch-to-batch variability, potentially complicating statistical analysis and leading to false positives. However, additional hiPSCs recruited for candidate gene validation consistently showed increased expression of TEK and SDR42E1 in NPC. Secondly, the role of biomarkers in cardiac differentiation of hiPSCs remains unclear. Our results from bulk RNA-seq and ATAC-seq of undifferentiated cells do not elucidate expression pattern of biomarkers during cardiac differentiation. Therefore, future studies should investigate the molecular mechanisms to understand the roles of biomarkers during developmental stages. Thirdly, regulatory mechanisms governing TEK and SDR42E1 expression are not fully understood. Epigenetic remodeling occurred during reprogramming [35], and our results showed higher chromatin accessibility and enriched pathways such as RNA transcription factor in PC than NPC. It suggests potential effects on nearby gene transcription. However, we focused on the gene expression patterns explaining the clonal heterogeneity in fully reprogrammed cells, lacking precise mechanisms regulating the expression of TEK and SDR42E1. This underscores the importance of addressing this complex question in future studies.

In conclusion, by integrating transcriptomic and epigenomic profiles, we identified TEK and SDR42E1 as novel predictive biomarkers of clonal heterogeneity for cardiac differentiation. These findings provide a comprehensive understanding of the molecular signatures underlying the cardiac differentiation capacity of hiPSC clones.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

hiPSCs

Human induced pluripotent stem cells

ATAC-seq

Assay for transposase-accessible chromatin using sequencing

PC

Productive clone

NPC

Non-productive clone

CMs

Cardiomyocytes

CL1,2,3,4,5,6

Clone1,2,3,4,5,6

DEGs

Differentially expressed genes

DARs

Differentially accessible regions

PBMCs

Peripheral blood mononuclear cells

TEs

Transposable elements

EREs

Endogenous retroelements

LTRs

Long terminal repeats

TEK

TEK receptor tyrosine kinase

SDR42E1

Short chain dehydrogenase/reductase family 42E member 1

ERK1/2

Extracellular signal-regulated kinase 1/2

TNNT2

Cardiac Troponin T

HERV

Human endogenous retrovirus

Author contributions

Conceptualization, J.Y., J.S., and J.L.; methodology, J.Y. and J.S.; software, S.J.; validation, J.Y. and J.S.; formal analysis, J.Y., J.S., and S.J.; investigation, J.Y., J.S., J.J., and S.H.; resources, J.Y., J.S., J.J., K.L., and H.R.J.; data curation, J.Y., J.S., and S.J.; writing—original draft, J.Y.; writing—review & editing—J.Y., J.S., S.J., and J.L.; visualization, J.Y., J.S., and S.J.; supervision, J.L.; project administration, J.L.; funding acquisition, J.L. All authors have read and commented on the manuscript.

Funding

This work was supported by the Bio&Medical Technology Development Program of the National Research Foundation (NRF) funded by the Korean government (MSIT) [grant numbers 2020M3A9E4037847, 2020M3A9E4037902]; and a Korean Fund for Regenerative Medicine (KFRM) grant funded by the Korean government (Ministry of Science and ICT, Ministry of Health and Welfare) [Grant Number 22A0302L1].

Data availability

The datasets are available in the article, in its online supplementary material, and data repositories (GEO accession No. GSE260767, GSE260883).

Declarations

Conflict of interest

All authors declare to have no conflict of interest related to this manuscript.

Consent for publication

We confirmed that consent for publication was received from all individuals.

Ethical approval

Ethical approval for the use of PBMCs and hiPSCs in this study was obtained from the Institutional Review Board of Sungkyunkwan University (IRB no. 2019-11-016-002 and 2020-01-020-001) and Samsung Medical Center (IRB no. 2016-11-025-015). Our study was conducted in accordance with the principles outlined in the Declaration of Helsinki, following the acquisition of written consent.