The cells were then transduced using the Episomal hiPSC Reprogramming Vectors (Life Technologies), followed by culturing on matrigel-coated dishes using StemSpan™ SFEM/StemSpan™ CD34+ Expansion Supplement (Stem Cell Technologies). The half medium was changed with Essential 8 medium (Life Technologies) each day from day 1 to day 7 until small colonies were formed. From the 7th day, Essential 8 was used as the main median and was changed daily for 2–3 weeks. When colonies developed, they were individually retrieved (for the first 8 passages), mechanically dissociated, and used for expansion into individual hiPSC lines. Every 5 days, the hiPSCs were passaged with 0.5 mM EDTA (Sigma). Three colonies were chosen for further experiments based on molecular testing and morphology.

Matrigel plates (Corning) were used for hiPSCs and H9 cell culture. The medium used was the Essential 8 medium supplemented with penicillin (50 U/mL) and streptomycin (50 mg/mL) (Life Technologies). When the cell confluency reached 70–80%, EDTA solution (Sigma) was used for undifferentiated colonies passage.

Here, 0.5 mM EDTA was used to detach hiPSCs, suspended in the Essential 8 medium containing 10 $\mu$M of Y-27632 (Abcam), seeded into low attachment dishes (Corning), then incubated at 37 °C with 5% CO${}_2$ for 24 h. Following EB formation, they were cultured in an E6 medium (Life Technologies), which was changed every 3 days. Following an incubation period of 10–14 days, spontaneous differentiated EBs were obtained and used for further experiments.

The differentiation of cardiomyocytes from hiPSCs was induced, as reported by Cao et al. [16] but with some adjustements. In short, the hiPSCs were incubated for 5 min in 0.5 mM EDTA and PBS, then transferred into Corning low attachment dishes containing Essential 8 medium for 2 days before cardiomyocyte induction. Induction was performed as follows: On day 0, the EBs were treated using a cocktail of 3 $\mu$M CHIR99021 (Stemgent), 25 ng/mL BMP4, 50 $\mu$g/mL L-ascorbic acid (Sigma), 400 $\mu$M 1-thioglycerol, DMEM/F12 combined with B27-vitamin A (Life Technologies) for 2 days; on day 2, the initial medium was replaced with fresh medium; day 3: treatment with 2 $\mu$M C59 (Abcam), 10 ng/mL BMP4 (Peprotech) in RPMI combined with B27-insulin for 2 days; on day 5, the medium was replaced with RPMI combined with B27 (Life Technologies) and replaced every 3 days. After induction, the differentiated cells were retrieved and stored for further analysis.

Hepatocyte-like cell differentiation was derived from hiPSCs, as previously reported [17]. Briefly, the hPSCs were seeded for hepatocyte differentiation using LN-521 (BioLamina) for 24 h in Essential 8 medium. On day 1, the medium was changed to fresh RPMI 1640 medium (Life Technologies), to which 50 ng/mL of Wnt 3a (Peprotech), 100 ng/mL of Activin A (Peprotech), 2% of B27 (Life Technologies) and 1% of penicillin or streptomycin (Life Technologies) were added. The medium was replaced with fresh medium on a daily basis for 5 days. On day 6, the medium was switched to a cocktail of KSR/DMSO combined with 0.5% GlutaMAX (Life Technologies), 1% DMSO (Sigma), 1% non-essential amino acids (Life Technologies), 1% penicillin/streptomycin, 80% knockout DMEM (Life Technologies), 20% knockout serum replacement (Life Technologies) and 0.1 mM beta-mercaptoethanol (Life Technologies); which were replaced every day for 5 days. Then, the cocktail medium consisting of HepatoZYME basal (Life Technologies) combined with 20 ng/mL OSM (Peprotech), 10 ng/mL HGF (Peprotech), 10 $\mu$M hydrocortisone 21-hemisuccinate sodium salt (Sigma), 1% GlutaMAX and 1% penicillin/streptomycin was used and replaced with fresh ones every 2 days for 7–10 days. When the cells reached standard differentiation, they were retrieved and stored for further use.

The hiPSCs were plated onto Geltrex-coated six-well plates after splitting into cell clumps at a density of 2–2.5 $\times$ 10${}^4$ cells per cm${}^2$. After a day, the Gibco PSC Neural Induction Medium (Life Technologies) containing a Neurobasal medium and Gibco PSC neural induction supplement was used. From days 0–4, medium was changed every other day and changed to a daily basis as from day 4 because the cells reached full confluence quickly. On day 7, the primitive NSCs were detached with Accutase (Life Technologies) transferred onto Geltrex-coated dishes, at a density of 0.5–1 $\times$ 10${}^5$ cells per cm${}^2$, and maintained in an NSC expansion medium consisting of a neural induction supplement and 50% of Neurobasal and Advanced DMEM/F12 medium. The NSC expansion medium was changed on a daily basis until they reached confluence on day 5 since the NSC plating.

Total RNA was isolated from the cells (hiPSCs, H9 and EBs) with Trizol (Life Technologies) following the manufacturer’s protocol, which was then purified using the column cleanup (Qiagen). cDNA was synthesized by reverse transcription following the manufacturer’s instructions using TransScript One-Step gDNA Removal and cDNA Synthesis SuperMix (Transgene).

The BioRad QX200 Droplet Digital PCR instrument with corresponding EvaGreen ddPCR SuperMix (Bio-Rad) was used. After preparing the ddPCR reaction mixtures, the QX200 droplet generator (Bio-Rad) was used to produce droplets using the following conditions: 10 min at 95 °C, followed by 40 cycles of denaturation for 30 s at 95 °C, 60 s of annealing/extension at 60 °C followed by 1 min at 72 °C, 5 min at 4 °C and 5 min at 90 °C, then maintained at 12 °C. Following the PCR reaction, the data were assessed using QuantaSoft (Bio-Rad). The corresponding fluorescence amplitude threshold of each gene was determined manually by comparing their distribution in distilled water.

Cells were fixed with 4% paraformaldehyde (Sigma-Aldrich) for 15 min at room temperature, permeabilized with 0.1% Triton-X (Sigma-Aldrich) for 10 min at room temperature, blocked in 5% BSA (Sigma-Aldrich) 1h at room temperature, followed by immunocytochemistry staining with the Embryonic Stem Cell Marker Panel (ab109884, abcam), Mouse monoclonal [1F11] to cTnT (ab10214, abcam), Nestin Polyclonal antibody (19483-1-AP, proteintech), AFP Polyclonal antibody (14550-1-AP, proteintech), Albumin Polyclonal antibody (16475-1-AP, proteintech) overnight at 4 °C in 1% BSA. Cells were washed two times, for 10 min, and then incubated for 1 h at room temperature in the dark with secondary antibodies Alexa Fluor 488 goat anti-mouse IgG, AlexaFluor 488 goat anti-rabbit IgG (H+L) or Alexa Fluor 594 goat anti-rabbit IgG (H+L) (all Life Technologies). Cells were washed two times, for 10 min, nuclei were stained with DAPI (Sigma-Aldrich) for 5 min , and washed three times, for 10 min. then The cells were imaged with ImageXpress Pico (MolecularDevices).

Total RNA was obtained from the samples with TRIzol (Invitrogen, Carlsbad, CA, USA). RNA integrity was determined on the Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA). RNA quality and purity were determined using 1% agarose gel and the NanoDrop equipment (Thermo Fisher Scientific Inc., Waltham, MA, USA). NGS library were prepared using 1 $\mu$g of total RNA at RIN value $>$6.5. The Poly(A) mRNA Magnetic Isolation Module was used for mRNA isolation. The First-Strand Synthesis Reaction Buffer was used for mRNA fragmentation. Random Primers were used for mRNA priming. To synthesize the first-strand of cDNA, ProtoScript II Reverse Transcriptase was used, and for the second-strand cDNA, the Second Strand Synthesis Enzyme Mix was used. The End Prep Enzyme Mix was then used to repair the opposite ends of the purified double-stranded cDNA and add a dA-tail, in one reaction. Next, T-A ligation was performed to both ends to add adaptors. An appropriate size of the adaptor-ligated DNA was chosen with Beads, and fragments with a size of approximately ~420 bp (approximate insert size, 300 bp) were recaptured. Next, amplification with 13 cycles of PCR using P5 and P7 primers was performed for each sample. The P7 primers, which contained a six-base index, were used for multiplexing. After this, cleaning of the PCR products were performed with beads, Qsep100 was used for validation (Biopticm, Taiwan, China) and a Fluorometer (Qubit 3.0; Invitrogen) was used for quantification. Libraries having different indexes were multiplexed and loaded onto the Illumina HiSeq instrument (Illumina, San Diego, CA, USA). The configuration was set at 2 $\times$ 150 bp paired-end configuration for sequencing. The HiSeq Control Software (HCS) + OLB + GAPipeline-1.6 (Illumina) on the HiSeq instrument was used for image analysis and base calling. Lastly, GENEWIZ (Suzhou, China) was used to analyze the sequences.

Upstream analyses of RNA-seq data were performed using a series of software, including fastp (version 0.14.1, https://github.com/OpenGene/fastp) was used to do the quality control of raw reads, star (version 2.7.10, https://github.com/alexdobin/STAR) was used to do the reads mapping to hg19, htseq (version 0.11.3, https://github.com/simon-anders/htseq) was used to do the reads counting, countToFPKM (version 1.0.0, https://github.com/AAlhendi1707/countToFPKM) was used to do the fpkm normalizing. Downstream analyses of RNA-seq data were performed using R (version 3.6.0, https://www.r-project.org/). pheatmap (version 1.0.12, https://github.com/raivokolde/pheatmap) was used to generate the heatmap plot. The description of the heatmap is: the color is drawn by the fpkm value,expressing the difference between samples, with the color bar depicting the extents of the expression value accordingly.

Marker genes were selected for detecting residual undifferentiated cells using the experimental scheme illustrated in Fig. 1A. Three hiPSC cell lines were constructed from PBMCs for further experiments based on molecular testing and morphology (Supplementary Fig. 1A). To identify candidate genes with high expression in hiPSCs but minimally expressed in differentiated cells, next-generation sequencing (NGS) was performed on the total RNAs isolated from the hiPSC cell lines and embryonic stem cells (H9), a panel of 250 stem cells markers was selected as a candidates pool [18], and a heat map was plotted (Fig. 1B), following which the top 30 expressed genes were picked.

Fig. 1.

Experimental scheme and stem cell genes analysis. (A) Experimental scheme for universal hiPSCs residue detection markers identification. Differentiate hiPSC into cardiomyocytes, neural stem cells and hepatocytes, then perform whole-exome sequencing on the hiPSCs and differentiated cells to analyze gene expression. Genes that were highly expressed in hiPSCs and were simultaneously low or not expressed in differentiated cells were selected as the candidate genes. NGS sequencing on spontaneous differentiated EB and GEtex query showed that the candidate genes with low expression in various tissues were the target genes. Finally, analyze the sensitivity of the target genes by digital PCR. (B) Expression heat map of a panel of 250 stem cell genes; sorting gene expression levels from high to bottom.

hiPSCs were differentiated into neural stem cells (ectoderm), hepatocytes-like cells (endoderm) and cardiomyocytes (mesoderm) to identify genes specifically expressed in differentiated cells. hiPSCs derived neural stem cells were identify by the expresion of NESTIN and SOX2 (Supplementary Fig. 2A), hiPSCs derived hepatocytes-like cells were identify by the expresion of AFP and ALB (Supplementary Fig. 3A), hiPSCs derived cardiomyocytes were identify by the expresion of cTnT (Supplementary Fig. 4A). Next-generation sequencing (NGS) was also performed on total RNAs isolated from the hiPSC differentiated cells following differentiation. Next, the mRNA seq data on the top 30 expressed genes was explored (Fig. 2A). Several genes, such as POU5F1, ESRG, DNMT3B, GAL, L1TD1, TDGF1, NSCAN10 and DPPA4, were highly expressed in hiPSCs and had minimal expression in differentiated cells.

Fig. 2.

hiPSCs residue detection candidate genes screening. (A) Expression heat map of 30 stem cell genes of the hiPSCs, H9 and hiPSC-derived hepatocyte-like cells (iHep), cardiomyocytes (iCM), and NSC (iNSC). The gene expression levels were sorted from high to bottom. (B) mRNA expression fold change between hiPSCs and hepatocyte-like cells. The genes marked in red are those whose expression fold change exceeded 20. (C) mRNA expression fold change between hiPSCs and NSC. The genes marked in red are those whose expression fold change exceeded 20. (D) mRNA expression fold change between hiPSCs and cardiomyocytes. The genes marked in red are those whose expression fold change exceeded 20. (E) Screening strategy for identifying candidate genes.

To identify highly selective markers for undifferentiated hiPSCs, expression fold change analysis on the RNA sequencing data comparing hiPSC cells to the differentiated cells was performed. An expression fold change $>$20 was set as the criteria for candidate genes. Our findings showed the expression fold change of ZSCAN10, ESRG, PTPRZ1, GAL, TDGF1, ZIC2, POU5F1, SOX2, L1TD1 and DNMT3B was $>$20 when differentiating hiPSCs to hepatocyte-like cells (Fig. 2B). Comparing hiPSCs with neural stem cells, TDGF1, ESRG, ZSCAN10, GAL, L1TD1, EPCAM, DNMT3B, PIM2 and POU5F1 met the selection criteria (Fig. 2C). ESRG, ZSCAN10, ZIC2, DPPA4, SOX2, DNMT3B, PTPRZ1, POU5F1, THY1, GAL, LIN28A, L1TD1, TDGF1, TERF1 and PIM2 were selected as the candidates when differentiating hiPSCs to cardiomyocytes (Fig. 2D). Seven overlapping genes were found between the three sets of analyses (Fig. 2E), among which the expression fold change of four genes (ZSCAN10, ESRG, GAL and TDGF1) was $>$50. Therefore, these seven genes were selected as candidates because they were highly expressed and abundant in hiPSC.

To determine whether the candidate marker genes could be used for identifying residual undifferentiated hiPSCs in any cell type in the body, Multi-Gene Query was conducted through The GTEx database (UTL: https://gtexportal.org/home/) (Fig. 3A). Five genes (ZSCAN10, ESRG, FOXH1, LIN28A, and DPPA4) which were abundantly expressed in hiPSCs but minimally expressed in other tissues (Fig. 3A, in red) were selected as suitable for universal residual hiPSCs detection. We also generated the tissue expression pattern of several genes (Fig. 3B). A slight expression of ZSCAN10 was observed in the testis, pituitary and cerebellum, while the expression of ESRG was high in the vagina but low in the other tissues. The other candidate genes were abundantly expressed in many tissues (Supplementary Fig. 5A). Combining the differentiation data above (Fig. 2E), ZSCAN10 and ESRG were selected as the ideal markers for universal residual hiPSCs detection.

Fig. 3.

Tissue expression pattern of the candidate genes. (A) Gene expression level in various tissues. Data were from the Multi-Gene Query through The Genotype-Tissue Expression (GTEx) project database. (B) ESRG and ZSCAN 10 expression level in various tissues. Data were from the Multi-Gene Query through The Genotype-Tissue Expression (GTEx) project database. (C) Expression heat map of 30 stem cell genes of the spontaneous differentiated EBs; Gene expression levels were sorted from high to bottom. (D) mRNA expression fold change between hiPSCs and spontaneous differentiated EBs. The genes marked in red are those whose expression fold change exceeded 20.

A spontaneous embryonic body (EB) differentiation was performed to validate the selection of ZSCAN10 and ESRG, as EB contains all three germ layers. For spontaneous differentiation, self-aggregated EBs were formed and transferred to Corning low attachment dishes. The medium was switched to E6 medium for 14 days. Following differentiation, NGS was also performed on total RNAs isolated from the EBs. Next, the mRNA seq data on the top 30 expressed genes was assessed (Fig. 3C). ZSCAN10 and ESRG were found highly expressed in hiPSCs but scarcely expressed in differentiated EBs, with an expression fold change of 146 and 574, respectively (Fig. 3D). These findings showed that ZSCAN10 and ESRG were suitable genes for universal residual hiPSCs detection.

Here Spike-in experiments were performed. A serial of 2, 20, 200, 2000 and 20,000 hiPSCs was mixed with 293T cells to get 2 million cells; then, the RNA was isolated for ddPCR. The copy number of ESRG and ZSCAN10 gene expression correlated with hiPSCs numbers (Fig. 4A,B). Two hiPSCs in a background of 2 $\times$ 10${}^6$ cells were detected with at least 0.0001% sensitivity. However, ESRG performed better with a higher copy number (Fig. 4A), possibly because of a higher expression level of ESRG in hiPSCs (Fig. 2D), which was consistent with the RNA seq data.

Fig. 4.

Detection sensitivity via spike-in experiments. (A) mRNA copy numbers of ESRG and ZSCAN10 in samples containing 2000, 200, 20 and 2 hiPSC samples in 2 million 293T cells. N = 3. (B) mRNA copy numbers of ESRG and ZSCAN10 in 2 million hiPSCs samples and samples containing 20,000 hiPSCs in 2 million 293T cells. N = 3.