Technical approaches to induce selective cell death of pluripotent stem cells

Abstract

Despite the recent promising results of clinical trials using human pluripotent stem cell (hPSC)-based cell therapies for age-related macular degeneration (AMD), the risk of teratoma formation resulting from residual undifferentiated hPSCs remains a serious and critical hurdle for broader clinical implementation. To mitigate the tumorigenic risk of hPSC-based cell therapy, a variety of approaches have been examined to ablate the undifferentiated hPSCs based on the unique molecular properties of hPSCs. In the present review, we offer a brief overview of recent attempts at selective elimination of undifferentiated hPSCs to decrease the risk of teratoma formation in hPSC-based cell therapy.

Introduction

The first clinical trial of human embryonic stem cell (hESC)-based cell therapy was approved by the FDA on January 23, 2009 and was launched by the Geron Corporation, a biotechnology company in the United States. The clinical trial was performed with oligodendrocyte progenitors (GRNOPC1) derived from hESCs to treat acute spinal cord injury patients through remyelination of damaged axons [1].

After the sudden discontinuation of the clinical trial for undisclosed reasons, the safety and efficacy of hESC-based cell therapy have been actively discussed in a number of articles [2–5]. Since the recent outstanding clinical results of retinal pigment epithelial (RPE) cells derived from hESCs used to regenerate vision following age-related macular degeneration (AMD) or Stargardt’s disease [6, 7], enormous efforts have been made to examine the efficacy and safety of this approach in phase I/II trials. At this time, ten hESC-based clinical trials are being undertaken worldwide for treatment of AMD or Stargardt’s disease, type I diabetes mellitus, severe heart failure, and spinal cord injury [8, 9].

The pluripotency and active cell proliferation capacity (or high clonogenic capacity with telomerase activity) are incomparable technical advantages of human pluripotent stem cells (hPSCs, which include hESCs and human induced pluripotent stem cells, iPSCs) over adult stem cells in terms of not only pluripotent differentiation potential but also a theoretically unlimited supply of desirable cell types. Therefore, hPSCs have been considered as a promising cell source for regenerative medicine. However, two characteristics of hPSCs (i.e., pluripotency and high clonogenic capacity with telomerase activity) are responsible for the formation of teratomas, benign tumors composed of three germ layers, which has been used to determine the ‘pluripotency’ of PSCs in vivo [10–12]. Ironically, the teratoma-forming capacity of the hPSCs that remain undifferentiated after the differentiation process (therefore, tumorigenic hPSCs) due to these unique characteristics contributes to one of the major hurdles to broader clinical implementation of hPSC-based cell therapy [5, 13].

It is notable that none of the preclinical studies with differentiated cells derived from hPSCs in mouse models indicate teratoma formation, although it is generally accepted that teratomas frequently develop after transplantation of cells derived from mouse ESCs into mouse models, regardless of the cell sorting and long-term culture techniques used after differentiation to minimize the possible engraftment of residual undifferentiated mESCs [14–16]. This difference in tumorigenicity between human and mouse PSCs that occurs after the engraftment of the cells derived from human or mouse ESCs in a mouse model results from host-dependent tumorigenesis. For example, as few as 500 mESCs can cause a 100% teratoma development rate in a mouse model (3 out of 3); however, 80,000 mESCs have only a 9% teratoma development rate in a rat model (2 out of 22) [17]. Importantly, it was also shown that a minimum of 10,000 hESCs are required to form a teratoma in a mouse model [18].

For hPSC-based cell therapy in human subjects, of which the tumorigenic effect cannot be determined using different species (such as mouse or rat), the risk of teratoma formation remains an important technical issue that must be fully resolved before the expansion of its clinical application [5]. Therefore, a variety of approaches using antibodies, suicide genes, stem cell-killing agents (defined as ‘stemotoxic agents’), and hypothetical ‘stem cells without tumorigenicity’ (stem cells that do not form a tumor in vivo) have been previously proposed to overcome this hurdle [19]. Since then, diverse approaches to this end have been actively studied, and one ‘stem cell-killing agent’ is even well defined for practical use [20]. Although there are already a few excellent review articles that summarize the strategies used to inhibit teratoma formation [21, 22], we aim to accentuate the important safety concerns of hPSC-based cell therapy and summarize the wide range of recent advances in preventing teratoma formation, such as the use of (1) small molecule-based selective elimination (2) genetic approach to introduce a suicide gene or miRNA switch, (3) antibodies targeting a surface-specific antigen (or antibody-guided toxins), (4) phototoxic approach, and (5) live detection and quantification of the residual hPSCs.

High mitochondrial priming in hPSCs

One of the distinct characteristics of hPSCs is their high susceptibility to DNA damage [23]. When differentiated somatic cells are challenged with genotoxic stimuli, such as ultraviolet light, ionizing radiation, or chemotherapeutic reagents, the cell cycle is arrested at G1/S to allow DNA damage repair before entering the S phase, unless the DNA damage is too severe to be repaired and triggers cell death [24]. The cell cycle checkpoint and DNA damage repair system are important for ensuring the genomic integrity of somatic cells. However, unlike in somatic cells, cell cycle checkpoints are absent or attenuated in mouse and human ESCs [25, 26]. Instead of activating cell cycle arrest, hESCs commit cell death even under low genotoxic stress [27]. High sensitivity to DNA damage and rapid apoptosis after even low-damage insults is some of the unique features of hESCs that allow them to avoid deleterious genomic mutations in their differentiated progeny [23].

Recent studies demonstrate that the high sensitivity to DNA damage in hESCs results from high induction of the mitochondrial cell death mechanism (referred to as ‘high mitochondrial priming’), which is mediated by cytoplasmic p53 [28] or prompt mitochondrial translocation of constitutively active BAX localized at the Golgi complex [29]. The details of mitochondria-dependent cell death in hESCs were extensively reviewed recently [30]. The high susceptibility of hPSCs to mitochondrial cell death compared with differentiated somatic cells is closely associated with the induction of selective cell death of hPSCs by a variety of small molecule-based approaches.

Small molecule-based selective elimination

Small molecules that induce selective cell death of hESCs to inhibit teratoma formation were first reported in 2004. The ceramide analog N-oleoyl serinol (S18) was shown to eliminate mouse and human ESCs through apoptosis and promote neural differentiation [31]. Since then, a variety of small molecules have been demonstrated to induce ESC-specific cell death and inhibit (or reduce) teratoma formation.

Activating ‘high mitochondrial priming’ for selective cell death of hESCs

As mentioned above, the high susceptibility of hESCs to cell death is due to the active mitochondria-dependent cell death mechanism that results from cytoplasmic p53 [28] (or mitochondrial translocation of p53 [32]) or rapid mitochondrial translocation of constitutively active BAX [29]. Additionally, pro-apoptotic proteins are highly upregulated, while a few anti-apoptotic proteins such as baculoviral IAP repeat-containing 5 (BIRC5) [32, 33] and B-cell CLL/lymphoma 10 (BCL10) [32] are expressed to maintain balance (or promote survival) [34]. Therefore, suppression or inhibition of hESC-specific anti-apoptotic proteins, such as BIRC5, with YM155 or quercetin (QC) was able to induce selective cell death of hPSCs while sparing differentiated smooth muscle cells and dopaminergic neurons [32]. Later, YM155 treatment was found to be more highly selective in eradicating teratoma formation by human iPSCs while sparing CD34⁺ hematopoietic stem cells (HSCs) in mouse models than other genetic approaches using a chemically inducible suicide gene system [35].

Similarly, a combination of purvalanol A (Cdk1 inhibitor) and Taxol was used to suppress the expression of survivin (the protein encoded by BIRC5) to induce cell death of hESCs [33]. However, inhibition of survivin, whose expression is normally high in a variety of cancers [32, 33], for selective cell death of hPSCs could not be used to isolate HSCs because survivin is required for HSC survival [36].

A recent study also demonstrated that QC treatment produces reactive oxygen species (ROS) in hESCs but not in human dermal fibroblasts (hDFs), leading to activation of the mitochondrial cell death pathway through cyclophilin D, which is highly expressed in the mitochondria of hESCs [37].

Alternatively, etoposide treatment, which leads to DNA damage and activates ‘high mitochondrial priming’ [29], was sufficient to ‘purge the teratoma risk’ of mouse ESCs [38]. Recent studies from Hurskey N. E. et al. revealed that inhibition of CDK1 with small molecules, such as purvalanol A, Ro-3306, or dinaciclib, induces DNA damage and achieves selective cell death of mouse and human ESCs by inhibiting the anti-apoptotic molecule MCL1 in ESCs [39]. Likewise, PluriSIn#2, one of 15 pluripotent cell-specific inhibitors (PluriSIns), which were identified by Dr. Nissim Benvenisty’s group through high-throughput screening [40], induced selective cell death by suppressing the expression of topoisomerase II, which is important for maintaining DNA integrity [41]. It is also noteworthy that YM155, a known BIRC5 suppressant [42], which was used to induce hPSC cell death [32], was shown to induce DNA damage in cancer cells expressing high levels of solute carrier family 35 member F2 (SLC35F2) [43]. In line with this, SLC35F2 was previously shown to be a specific surface marker of hESCs [44], suggesting that the high sensitivity of hPSCs to YM155-dependent cell death [32] may result from the DNA damage-mediated mitochondrial priming occurring in hPSCs. If so, the use of YM155 to selectively eliminate hPSCs would be only applicable when the differentiated cells had low levels of SLC35F2.

Inhibiting the specific metabolism of hPSCs

PluriSIn#1, identified as the most selective compound for achieving hPSC-specific cell death among the other 15 PluriSIns identified via high-throughput screening, induces ER stress and apoptosis in hPSCs by inhibiting stearoyl-coA desaturase (SCD1) [40]. Later, PluriSIn#1 was found to be effective for isolating cardiomyocytes derived from iPSCs by selectively eliminating Nanog-positive cells [45]. Importantly, oleic acid biosynthesis via SCD1 is important for mouse embryonic development, suggesting that the unique metabolic processes of hPSCs would be a plausible target for purging residual hPSCs [40]. Similarly, an inhibitor of Erv1 oxidase (MitoBloCK-6, a mitochondrial protein import blocker from the laboratory of Carla Koehler), which is important in the mitochondrial disulfide relay system, was identified through chemical screening and selectively induces apoptotic cell death via cytochrome c release (a key event in mitochondrial cell death) in hESCs [46]. Similarly, differences in the glucose metabolism of hPSCs (e.g., high dependence on glutamine) were also used to selectively ablate residual hPSCs after cardiomyocyte differentiation using glucose- and glutamine-depleted culture medium supplemented with lactate [47].

However, considering the diverse molecular characteristics of desirable cells differentiated from hPSCs, this approach, targeting key metabolic enzymes or depleting a nutrient essential for hPSCs survival, would be limited to differentiated cells with low dependency on these enzymes or nutrients for their survival.

Other classes of small molecule-based selective elimination

In addition, through in-house compound library screening under hESCs and mouse embryonic fibroblast (MEF) co-culture conditions, JC011 was identified as a selective cell death inducer of hESCs to inhibit teratoma formation [48]. In a similar approach, screening of an in-house chemical library of cytotoxic small molecules, a derivative of okadaic acid, identified as 27-deoxy-27-oxookadaate, was found to have selective cytotoxicity to hESCs due to the low expression of ATP-binding cassette (ABC) transporters ABCB1 and ABCG2 in hESCs, leading to the accumulation of 27-deoxy-27-oxookadaate until the cytotoxic concentration was reached [49]. In this case, ABCB1 and ABCG2 expression in a certain type of differentiated cell would be an important indicator for the cytotoxicity of this compound to the differentiated cells, such as astrocytes, which are moderately sensitive to this compound [49]. Additionally, metformin treatment in vivo was shown to decrease teratoma size in an apoptosis-independent manner, although the mechanism of this finding was not clearly addressed [50].

Although the approach with small molecules would be highly effective and relatively simple, it would be difficult to guarantee the functional safety of all types of cells that differentiate from hPSCs considering their individual biological properties.

Genetic approach to introduce suicide genes and miRNA switches

The typical suicide gene approach, which uses selective expression of the thymidine kinase gene of the herpes simplex virus (HSVtk) in undifferentiated PSCs in combination with the guanosine analog prodrug ganciclovir (GCV) [51], was extensively applied to achieve selective cell death by GCV treatment [52–57]. To prevent the undesirable cytotoxicity of GCV treatment from activating the suicide gene system in the differentiated cells, which may cause cytotoxicity to a normal cell type, such as lymphoblastoid [58] or corneal endothelial cells [59], visual light (540–560 nm of green light) was used instead to activate a novel photosensitizer suicide gene system, inducing selective cell death of mouse and human ESCs but not endothelial cells derived from PSCs. This selective cell death in mouse ESCs is sufficient to completely inhibit teratoma formation in the mouse model [60]. Such high phototoxic selectivity was achieved by introducing the KillerRed (KR) gene (a genetically encoded photosensitizer) [61] to the EOS [early transposon promoter and Oct-4 (Pou5f1) and Sox2 enhancers] promoter, an artificial promoter initially designed to select fully reprogrammed iPSCs [62] and allow KR to be specifically expressed only in undifferentiated PSCs [60]. Importantly, endothelial cells derived from KR-expressing mESCs remained fully functional even in vivo and were able to repair the ischemic damage from the visual light exposure used to purge the undifferentiated mESCs [60].

Alternatively, inducible caspase-9 (iCasp9: iC9), which becomes an active dimer upon treatment with a synthetic chemical inducer of dimerization (CID) and which was initially developed as a safeguard system for T-cell therapy [63], was applied to PSCs for teratoma inhibition not only in vitro but also in vivo [64, 65]. This suicide system was also used to reduce the size of the teratoma and ablate iPSC-derived rejuvenated cytotoxic T lymphocytes, serving as a possible safeguard system in vivo [66]. A recent study reported that chemical inducers (e.g., CID) for activating the iCasp9 suicide systems were cytotoxic to CD34⁺ HSCs, while GCV itself showed a bystander effect on normal iPSCs [35]. Therefore, a chemical inducer to activate the suicide system should be carefully selected based on the differentiated cells to ensure the safety of the differentiated cells.

Furthermore, genetic approaches using a suicide gene have been largely criticized due to the additional risk of random insertion of the foreign gene into the hPSCs, which may cause unexpected genetic aberration, unless the insertion of the foreign gene is tightly controlled by new genetic editing technologies to the genomic safe harbor sites, such as AAVS1, ROSA26, or CCR5, as previously proposed [54, 60, 66]. Until then, this genetic approach with an inducible suicide system may remain at the ‘proof of concept’ level.

Alternatively, micro-RNA (miRNA) switch technology, a transgene-free genetic approach targeting cell-specific miRNA, which was originally developed for the purification of desirable cells [67], was applied to selectively eliminate undifferentiated hPSCs with puromycin resistance using an miR-302a switch [68].

Targeting the specific surface markers of hPSCs with antibodies or proteins

Cell sorting using antibodies for surface proteins specific to the differentiated cell types has been primarily used for isolating desirable cell types after differentiation. However, certain types of cells, such as ventricular cardiomyocytes, lack a specific surface protein, making pure isolation technically challenging after differentiation [69]. Alternatively, molecular beacons (MBs), oligonucleotide hybridization probes that specifically bind to intracellular mRNAs, have been developed to isolate or enrich a desired cell type that lacks surface markers [69–72].

On the other hand, considering the unique surface marker expression profile of hPSCs [44], a set of antibodies recognizing the unique surface proteins of hPSCs has been produced [73, 74]. Antibodies against stage-specific embryonic antigens (SSEAs), such as SSEA-3 and SSEA-4, and tumor-related antigen (TRA)-1-60 and TRA-1-81 were used not only to identify but also to sort out the undifferentiated PSCs [75].

Therefore, separation based on an antibody against a specific surface protein (e.g., SSEA-5) through fluorescence-activated cell sorting (FACS) [76] or selective cell death with a cytotoxic antibody (e.g., against claudin-6 [77] or podocalyxin-like protein-1 (PODXL) [74]) would be a valid approach to reduce the potential for teratoma formation in heterogeneously differentiated cultures.

Similar to the antibody approach, rBC2LCN (recombinant N-terminal domain of the BC2L-C lectin derived from Burkholderia cenocepacia) was identified as a lectin probe that specifically binds to hyperglycosylated podocalyxin as a cell surface ligand in hPSCs [78]. Therefore, strategies involving the addition of a recombinant toxin protein (catalytic domain of Pseudomonas aeruginosa exotoxin A) conjugated to the lectin probe (rBC2LCN-PE23) were found to be effective at selectively eliminating undifferentiated hPSCs [79]. Importantly, rBC2LCN-PE23 has been recently commercialized as an ‘undifferentiated hPSCs elimination reagent’ (http://www.wako-chem.co.jp).

Phototoxic approach

An antibody conjugated with a gold nanoparticle, which absorbs the energy from laser pulses to produce heat, was previously developed to induce lethal membrane damage exclusively to the labeled cells [80]. Using an antibody against TRA-1-60 and TRA-1-81, a surface antigen specific to hPSCs, laser exposure and subsequent photothermolysis were sufficient to induce selective cell death and inhibit teratoma formation [81]. Alternatively, a fluorescence rhodamine compound (rosamine, a compound designated yellow1: CDy1), which was found to specifically interact with a protein in the mitochondria of PSCs [82], was used to induce selective cell death of both mouse and human PSCs [83]. Of note, exposure to green light at 532 nm produces ROS and selectively kills CDy1-stained PSCs but not the endothelial cells derived from PSCs. Importantly, teratoma formation after transplantation of mouse ESCs in the mouse model was completely blocked by CDy1 staining and light exposure [83].

It is also noteworthy that the amount of green light used to induce hPSC death with a photosensitizing gene (e.g., KR) [60] or fluorescence probe [83] while ensuring the functional safety of endothelial cells would not be suitable for RPE cells considering the high photosensitivity of RPE cells [84].

Detecting and evaluating residual hPSCs

After differentiation, the residual undifferentiated hPSCs should be carefully quantified to determine whether an additional step is required to eliminate the undifferentiated hPSCs with the aforementioned techniques. Furthermore, even after the treatment, further validation to specifically quantify the residual hPSCs as a safeguard to ensure complete ablation can also be performed, as shown in Fig. 1.

Fig. 1.

An example of the procedure of human pluripotent stem cell-based cell therapy for assuring teratoma-free cell therapy. After differentiation to the desirable cell type, an assay should be performed to check for possible contamination of tumorigenic hPSCs, even after enriching the desirable cell types. Likewise, even after additional steps have been taken for ablating residual tumorigenic hPSCs, cell therapy should only be conducted after a highly sensitive assay has been performed that validates that no tumorigenic hPSCs remain in order to guarantee teratoma-free cell therapy

Invasive approaches

Conventionally, semi quantitative real-time PCR or flow cytometry has been widely used to monitor the residual undifferentiated PSCs [85]. Dr. Yoji Sato’s group recently developed highly sensitive real-time PCR [86, 87] or droplet digital PCR [88] with a PSC-specific gene (e.g., Lin28) to examine the possible contamination of undifferentiated hPSCs at the single-cell level following their differentiation into RPE cells or cardiomyocytes. Although flow cytometry using mostly PSC-specific surface antibodies has been widely employed to monitor undifferentiated PSCs after differentiation, the proportion of the marker-positive population is markedly affected by the gating technique [86]. Therefore, several strict controls should be prepared to define the desirable population by flow cytometry. However, despite their high sensitivity, both real-time PCR and flow cytometry techniques are invasive approaches that irreversibly consume a large quantity of the differentiated cells (more than 10,000 cells). Because considerable time and resources are required to differentiate desirable cell types from hPSCs (for example, 30–50 days for differentiation of RPE cells from hESCs [89]), unrecoverable consumption of differentiated cells to ensure safety should be reconsidered.

Non-invasive approaches

Recently, a hPSC-specific glycoprotein was identified [90], and rBC2LCN, a recombinant lectin probe, was identified as a specific probe [91] for podocalyxin, a heavily glycosylated type 1 transmembrane protein prominent in hPSCs [92]. Specific interaction with rBC2LCN was sufficient for live-cell imaging of hPSCs in a cost-effective manner [93]. Furthermore, it was also demonstrated that hyperglycosylated podocalyxin is secreted into the hPSC culture medium. Therefore, simple determination of the concentration of hyperglycosylated podocalyxin in the culture medium using the rBC2LCN-based sandwich assay system (named the Glycostem test) can selectively quantify the teratoma-forming (or tumorigenic) undifferentiated hPSCs present after differentiation [78].

Alternatively, a few fluorescent chemical probes, such as the Kyoto probe 1 (KP1) [94] and the aforementioned CDy1, were demonstrated to be highly specific to hPSCs [82]. In particular, CDy1 was later applied to quantify or isolate the undifferentiated hPSCs using FACS [95] and to selectively kill the stained, undifferentiated PSCs using visible light [83].

On the other hand, dual fluorescence resonance energy transfer (FRET) MBs that specifically bind to Oct-4, originally developed for identification and isolation of hESCs [96], could also be used to quantify and even sort out the residual hPSCs. The efficiency of this approach has not yet been experimentally determined.

Due to recent advances in bioengineering techniques, a cell-chip system has been widely applied to monitor the cellular response, including differentiation, of the stem cells [97]. Similarly, a cell-chip system detecting the unique electrochemical potential of hPSCs has been designed. The intensity of the electrochemical potential generated from the live hPSCs without any labeling demonstrated a clear linear relationship with cell number, even in mixed cell conditions with differentiated progeny, allowing the extrapolation of the exact number of residual hPSCs in the mixed condition [98]. Very recently, a surface-enhanced Raman scattering (SERS) assay based on nanoparticles conjugated with the hPSC surface markers SSEA-5 and TRA-1-60 was demonstrated to trace as few as a single hPSC in 10⁶ cells [99]. Therefore, when such an assay system allowing live detection of hPSCs is optimized and further improved, it would be highly useful for monitoring the presence of residual hPSCs after differentiation and for deciding whether an additional step to ablate the residual hPSCs in the mixture is necessary for safety assurance. Finally, immediately prior to cell transplantation, the differentiated cell population would be again used for final validation of safety for teratoma-free cell therapy (Fig. 1).

Concluding remarks

As mentioned above, a variety of strategies, summarized in Table 1, have been examined to selectively ablate undifferentiated PSCs from differentiated cells for teratoma-free hPSC-based cell therapy with no or low cytotoxicity of the differentiated cells for quality assurance as well. However, considering the diversity of desirable cell types for future hPSC-based cell therapy, it would be hardly possible to presume that one methodology may selectively eliminate residual hPSCs without damaging the diverse types of differentiated cell type, of which properties would be varied. In this line, each methodology should be carefully selected depending on the molecular characteristics of desirable cells. Besides, not only low cytotoxicity of the differentiated cells, as listed in Table 1, but also the functional safety of each methodology to the differentiated cells in vivo, as shown previously [60, 100], should be intensively examined to apply each method to practical use. Accordingly, continuous effort should be applied further to develop novel strategies to ensure the safety of the differentiated cells as well as the efficacy of eliminating the undifferentiated hPSCs for future teratoma-free hPSC-based cell therapy.

Table 1.

Strategies to selectively eliminate the tumorigenic hPSCs

Strategies	Name	Mode of action or target	Cell model	Refs.
Small molecules	Quercetin	BIRC5 repression, CypD interaction	hPSCs vs hDFs, hASMCs, hiPSC-derived SMCs, hESC-derived dopaminergic neurons	[32, 37]
	YM155	BIRC5 repression or DNA damage (?)	hPSCs vs hDFs, hASMCs, hiPSC-derived SMCs, hESC-derived dopaminergic neurons	[32]
	Taxol and purvalanol A	BIRC5 repression	hESC-derived teratoma	[33]
	Purvalanol A, Ro-3306, Dinaciclib	CDK inhibition	Human and mouse ESCs vs mES-diff, hESC-derived pancreatic progenitor cells	[39]
	Etoposide	DNA damage	mPSCs vs MEFs	[38]
	PluriSIn#2	Suppression of Topoisomerase II alpha	hPSCs vs hPSC-derived various cell types, fibroblasts	[41]
	Metformin	Unknown	miPSCs vs MEFs	[50]
	JC011	Unknown	hPSCs vs MRC-5, human neonatal cardiomyocytes	[48]
	S18 (N-oleoyl serinol)	Ceramide analog, PKC delta inhibition	Human and mouse ESCs	[31]
	27-deoxy-27-oxookadaate	ABCB1 and ABCG2	hiPSCs vs adrenal gland, liver, bronchia and prostate cells	[49]
Targeting hPSCs’ specific metabolism	PluriSIn#1	Oleate synthesis inhibitor to inhibit stearoyl-coA desaturase (SCD1)	hPSCs vs fibroblasts, hepatocytes, cardiomyoctes, NSCs, MSCs	[40]
	MitoBloCK-6	Erv1 oxidase	hESCs vs hDFs	[46]
	Glucose and glutamine free medium with lactate supplement	High dependency of glucose and glutamine metabolism	hPSCs vs hPSC-derived cardiomyocytes	[47]
	HSVtk and GCV	Inhibition of DNA elongation	hPSCs vs MRC-5	[52–57]
Introducing suicide gene and miRNA switch	KillerRed with visual light	Oxidative damage	hESCs, mPSCs vs mESC-derived endothelial cells	[60]
	Inducible caspase 9	Apoptosis induction	hiPSCs	[64, 65]
	miR-302a switch	micro-RNA-302a	hiPSCs vs NHDF, hiPSC-derived dopaminergic-like neuronal cells	[68]
	SSEA-5 mAb	SSEA-5	hPSCs vs RA-induced differentiated cell mixtures	[76]
Targeting hPSCs’ specific surface markers with antibody or protein	Claudin-6 mAb	Claudin-6	hPSCs vs ectodermal and mesodermal cell types	[77]
	mAb 84	Podocalyxin-like Protein 1	hESCs, hECs vs hEBs, mESC, mouse fibroblasts	[74]
	rBC2LCN-PE23	Hyperglycosylated podocalyxin	hPSCs vs Human fibroblasts, hADSCs	[79]
	Antibody conjugated gold nanoparticle with laser exposure	Photothermolysis	hESCs vs hESC-derived neural precursors	[81]
Phototoxic approach	CDy1 with visual light	Oxidative damage	Human and mouse PSCs vs hESC-derived endothelial cells	[82]

PSCs pluripotent stem cells, ESCs embryonic stem cells, EB embryonic body, EC embryonic carcinoma, DFs dermal fibroblasts, ASMCs aortic smooth muscle cells, NSCs neural stem cells, MSCs mesenchymal stem cells, MRC-5 human lung fibroblasts, NHDF normal human dermal fibroblasts, MEFs mouse embryonic fibroblasts, ADSCs adipose-derived mesenchymal stem cells, RA retinoic acid

In addition, it is also important to develop an approach to selectively eliminate the hPSCs in vivo after accidental transplantation into patients. To this end, the iCasp9 suicide system was shown to reduce the teratoma size in vivo [64], and subsequent treatment with metformin lowered teratoma formation after the transplantation of mouse iPSCs [50]. Such a methodology to limit the teratoma formation in vivo should be more intensively verified to minimize unexpected side effects in human patients.

As with drugs, safety should be considered to be of a similarly high priority as efficacy in hPSC-based stem cell therapy for future clinical outcomes. Therefore, after serious assessment of the risk–benefit ratio, hPSC-based therapy should be performed when the benefit to the human patient is considered to be greater than the potential risks. From this point of view, continuous attempts to lower the risk of teratoma formation may improve the clinical application of hPSC-based therapy by increasing the therapeutic index in the future.

Compliance with ethical standards

Competing interests

None.