Challenges to the clinical application of pluripotent stem cells: towards genomic and functional stability

Human embryonic stem cells (hESCs) can undergo unlimited self-renewal and are pluripotent, retaining the ability to differentiate into all cell types in the body. As a renewable source of various types of human cells, hESCs hold great therapeutic potential. Although significant advances have been achieved in defining the conditions needed to differentiate hESCs into various types of biologically active cells, many challenges remain in the clinical development of hESC-based cell therapy, such as the immune rejection of allogeneic hESC-derived cells by recipients. Breakthroughs in the generation of induced pluripotent stem cells (iPSCs), which are reprogrammed from somatic cells with defined factors, raise the hope that autologous cells derived from patient-specific iPSCs can be transplanted without immune rejection. However, recent genomic studies have revealed epigenetic and genetic abnormalities associated with induced pluripotency, a risk of teratomas, and immunogenicity of some iPSC derivatives. These findings have raised safety concerns for iPSC-based therapy. Here, we review recent advances in understanding the genomic and functional stability of human pluripotent stem cells, current challenges to their clinical application and the progress that has been made to overcome these challenges.

and infection [4]. The recent development of induced pluripotent stem cells (iPSCs) has raised the hope that these cells could become a renewable source of autolo gous cells for transplantation into patients [10]. However, recent studies have identified genomic instability, epi genetic abnormality and immunogenicity of iPSCs and their derivatives, raising safety concerns for their clinical development. In addition, contamination with undiffer en tiated pluripotent stem cells poses a cancer risk through the potential formation of teratomas in the recipient [11]. Here, we summarize recent progress in under standing the genomic and functional stability of human pluripotent stem cells, current challenges to their clinical application and recent progress in overcoming these challenges.

Clinical applications of human stem cells
Since the successful transplantation of hematopoietic stem cells (HSCs) from the bone marrow or cord blood for the treatment of various bloodrelated diseases, stem cellbased therapy has been vigorously pursued to treat various human diseases. Because of their immuno modu latory activity, multipotency (the ability to differentiate into several cell types) and ability to produce trophic factors that promote tissue regeneration, mesenchymal stem cells are being tested in over 100 clinical trials to determine their efficacy to treat a large panel of human diseases, such as autoimmune diseases, spinal cord injury and myocardial infarction [12]. Some of these clinical trials have shown promising results. In addition, fetal derived neural stem cells have been tested in a clinical trial to treat amyotrophic lateral sclerosis [13]. Although these stemcellbased therapies are promising, they are limited by the technical difficulty in retrieving these adult and fetusderived stem cells, their inability to under go extensive selfrenewal, and the oncogenic poten tial of the transplanted cells [14]. Human pluripotent stem cells could overcome these hurdles by becoming a renewable source of mesenchymal stem cells and neural stem cells.
The feasibility of the clinical development of hESC based therapy has been improved by the scalability of hESCs as well as the formulation of chemically defined medium containing material only from human ('xenofree' medium) for hESC culture ( Table 1). Some of the hESC lines have been banked (cryopreserved) under good laboratory practice procedures with minimal involve ment of animal components [1517]. For example, to eliminate the need for mouse feeder layer cells, human feeder layer cells have been established to support the longterm selfrenewal of hESCs [1821]. Recent studies have also shown that the feeder layer can be replaced with purified human extracellular matrix proteins, such as human recombinant laminin511 and vitronectin, to support the longterm culture of hESCs [2225]. In addition, the recent development of chemically defined medium with the addition of small molecules that can promote the selfrenewal of hESCs greatly facilitates their clinical development [2629]. The development of syn thetic peptideacrylate surfaces and microcarriers, which can support longterm selfrenewal of hESCs in chemi cally defined and xenofree medium, further addresses the scalability issue [2932]. Finally, the successful validation in animal models and initiation of clinical trials of hESCbased therapy described earlier highlights the significant recent advances in their clinical development.

Genome-scale studies of pluripotent stem cells
Genomic stability of human pluripotent stem cells is a prerequisite for their application in human therapy. The routine method to evaluate the genomic stability of human cells is chromosome karyotyping, which has been used extensively to identify aneuploidy and large chromo somal deletion or duplication. However, this method cannot identify subtle genomic abnormalities, such as missense mutations in coding sequence. Recent advances in genome sequencing have allowed the identification of such subtle genomic aberrations. As a complementary approach, the recent establishment of highresolution single nucleotide polymorphism (SNP) comparative genome hybridization (CGH) arrays has also enabled the characterization of subtle chromosomal deletions and duplications in pluripotent stem cells. These genomic approaches are becoming increasingly important in the evaluation of the genomic stability of pluripotent stem cells. The recent development of chromatin immuno preci pitation sequencing (ChIPSeq) and wholegenome bisulfite sequencing has helped to reveal the global Table 1

The genomic and functional stability of pluripotent stem cells
Advances in the genomescale functional characteriza tion of pluripotent stem cells have revealed challenges for the clinical development of pluripotent stem cells. Here, we discuss recent findings related to the genetic stability, risk of teratoma formation and immunogenicity of hESCs and human iPSCs (hiPSCs).

Genetic stability of pluripotent stem cells Genetic stability of hESCs
To develop hESCs for cell therapy it is important to scale up their production. Because DNA damage occurs during normal cellular proliferation, the accumulation of un repaired DNA damage in hESCs during extended proliferation could lead to genetic instability and thus promote tumorigenesis in cells differentiated from hESCs. In support of this notion, highresolution SNP CGH studies have shown that centrosomal amplification has been associated with extended culture of hESCs [33]. Therefore, it is crucial to elucidate the mechanisms by which ESCs maintain genetic stability. However, the pathways for maintaining the genetic stability of ESCs and somatic cells are not completely conserved [34]. For example, in somatic cells the ataxiatelangiectasia mutated (ATM) protein, a Ser/Thr protein kinase, is the master regulator of cellular responses to DNA doublestrand break (DSB) damage and is required for maintaining genomic stability by phosphorylating and activating many DNA repair and cell cycle checkpoint proteins [35]; however, ATM seems not to be as important for main taining genomic stability in hESCs [36]. Instead, recent studies suggest that the tumor suppressor p53 is impor tant in maintaining hESC genomic stability. In somatic cells, p53 is a critical transcription factor that directly regulates the expression of hundreds of genes that mediate cell cycle arrest, apoptosis, and senes cence [37]. In addition, p53 has transcription indepen dent roles in regulating microRNA (miRNA) processing and metabolism [38,39]. As another example indicating that DNA damage responses in hESCs differ from those in somatic cells, hESCs lack the cell cycle G1/S check point and do not show senescence after DNA damage [40,41]. Instead, p53 has a new role in the DNA damage response in ESCs in suppressing the expression of NANOG (a transcription factor required for hESC selfrenewal) after DNA damage, leading to the differentiation of DNAdamaged ESCs [41,42]. Therefore, an important mechanism for maintaining genetic stability in selfrenewing ESCs is to eliminate ESCs with DNA damage from the selfrenewing pool through differentiation ( Figure 1). In addition, p53 is required for the G 2 /M checkpoint and apoptosis in hESCs after DNA damage [40,41]. In support of the crucial roles of p53 in maintaining genomic stability in hESCs, p53 / hESCs, generated by sequentially targeting both p53 alleles of hESCs by homologous recombination, show extensive genomic instability during extended proliferation [41]. In summary, although ESCs lack the typical p53dependent functions, such as cell cycle G1/S checkpoint and senescence, p53 maintains the genomic stability of ESCs by coordinating their DNA damage responses and selfrenewal capability ( Figure 1). There fore, it is crucial that the p53 pathway is not adversely affected during longterm expansion of hESCs, for example by the formulation of the tissue culture medium.

Genetic instability of hiPSCs
The reprogramming factors that have been used to generate iPSCs, including octamerbinding transcription factor 4 (OCT4, also known as POU5F1), sex determining region Ybox 2 (SOX2), NANOG, LIN28, Krüppellike family of transcription factor 4 (KLF4) and cMYC, have some oncogenic potential [43]. For example, cMYC and KLF4 are oncoproteins important for cellular prolifera tion [44,45]. In addition, although NANOG, OCT4 and SOX2 are critical pluripotency factors required for the selfrenewal of ESCs, they are also overexpressed in many types of cancers, and their expression has been correlated with poor prognosis of cancer patients [4649]. LIN28 regulates the expression of miRNA and is highly onco genic, inducing the expression of oncogenes such as cMYC [50,51]. Therefore, the expression of these repro gramming factors in somatic cells could activate tumor suppressor pathways, including p53 and an alternative reading frame product of the CDKN2A locus (ARF) that is important for activating p53 after oncogenic stress, leading to p53dependent cell cycle arrest, apoptosis or senescence [43]. In support of this, recent studies have shown that p53 and its transcriptional target p21 (also known as CIP1/WAF1) suppress induced pluripotency [5257]. These studies also indicate that reprogramming induces DNA DSBs that can also activate p53. Given that p53 is required to maintain genomic stability in both somatic and pluripotent stem cells, these findings raise concerns regarding the genomic stability of iPSCs ( Figure 2). Furthermore, although one study has identified very few genetic abnormalities in mouse iPSCs [58], recent genomic sequencing studies have identified coding sequence mutations in human iPSCs [5961]. However, the origin of these genetic mutations, whether from preexisting rare mutations or acquired during reprogramming, remains unclear. Recent SNPCGH and genome sequencing studies have also identified chromosomal duplications and deletions in iPSCs, all of which are potentially oncogenic [33,62,63]. Such genetic abnormalities could cause other functional changes, such as immunogenicity of iPSCderived cells. Therefore, to ensure the genetic stability of iPSCs, it is important to elucidate which p53 dependent responses, such as cell cycle arrest, apoptosis and senescence, are important for protecting against re pro gramminginduced DNA damage. The p21 protein is required for p53dependent cell cycle arrest, and p53 upregulated modulator of apoptosis (PUMA) is required for p53dependent apoptosis [64]. Depletion of PUMA and p21 greatly promotes reprogramming efficiency without increasing reprogrammingassociated DNA damage by activating the senescence pathway [65]. There fore, with improved understanding of the mecha nisms involved in induced pluripotency and reprogram minginduced DNA damage responses, it may be possible to optimize reprogramming strategies to minimize the genetic instability in iPSCs.

Risk of teratoma formation
Pluripotent stem cells can form teratomas when trans planted in vivo [1,66]. The lineagespecific differentiation of hESCs or iPSCs is rarely complete, so a major hurdle that hinders the clinical application of hESC or iPSC based therapy is the risk of teratoma formation associated with the presence of undifferentiated hESCs or hiPSCs following transplantation [66]. This risk is further complicated by the intrinsic variability associated with lineagespecific differentiation of hESCs or hiPSCs. The risk of teratoma formation for hESC or hiPSCbased cell therapy would thus depend on the batchtobatch varia tion in lineagespecific differentiation, and this could make evaluation of this risk a lengthy and costly process. An efficient and scalable approach to eliminate this risk would greatly facilitate the development of hESC or hiPSCbased cell therapy.
Significant effort has been devoted to the development of antibodybased strategies for the removal of un differ entiated hESCs or iPSCs remaining in the differentiating culture before transplantation. For example, recent studies have shown that a cytotoxic antibody recognizing podocalyxinlike protein1 (PODXL) can kill hESCs [67,68]. Recently, it has also been reported that immune depletion of hESCs using a combination of antibodies against multiple hESC surface markers (SSEA5, CD9, CD90 and CD200) can remove hESCs from differentiat ing cultures [69]. Although the expression of SSEA5 might be relatively specific for hESCs, other surface markers such as CD9, CD30, CD90 and CD200 are broadly expressed in differentiated tissues. Therefore, the limitation of these antibodybased strategies is the lack of specificity of the targeted hESC surface markers. In addition, it has been reported that the use of antibody based approaches can be limited by the potential of some progenitor cells to spontaneously dedifferentiate into the pluripotent state after transplantation, leading to tera toma formation [70].
The pluripotent state of hESCs and hiPSCs is required for teratoma formation, so another approach to reduce the risk of teratoma formation by hESCs or hiPSCs is to Figure 2. p53 suppresses induced pluripotency. p53 is activated by multiple stimuli in cells undergoing reprogramming, including the oncogenic potential of the reprogramming factors, reprogramminginduced oxidative stress and DNA double-stranded breaks (DSBs), and telomere shortening, which can ultimately block successful iPSC production. Therefore, the transient or partial inactivation of p53 might be required for cells undergoing reprogramming to complete their dedifferentiation into iPSCs, but this might also provide a window of opportunity for the accumulation of genetic abnormalities.  [71,72]. Strategies to suppress the expression of NANOG could therefore reduce the risk of teratoma formation. Accordingly, disruption of the interaction between NANOG and PIN1 results in destabilization of NANOG in ESCs and partially suppresses their potential for teratoma formation in vivo [73]. This finding suggests that, by targeting multiple pluripotency pathways, it could be possible to eliminate the risk of teratoma formation by hESCs or iPSCs.

Immunogenicity of pluripotent stem cells
Recent studies have investigated the immunogenicity of ESC and iPSCderived cells, as it is particularly impor tant to avoid immune responses to these cells in clinical applications.

Immunogenicity of ESC-derived cells
ESCs express undetectable or very low levels of major histocompatibility complex (MHC) molecules, indicating that these cells are not immunogenic, although the expres sion levels of MHC molecules increase during the differentiation of ESCs [74]. Therefore, allogeneic immune responses would not be able to eliminate undifferentiated ESCs in transplants. Despite significant progress in the development of hESCbased cell therapies, a major issue that remains to be resolved is the immune rejection of hESCderived cells by recipients because their derivatives express MHC molecules that are different from those of the recipients (Figure 3). Although allograft rejection can be delayed if the recipient's immune system is persistently suppressed, a significant fraction of allografts are still rejected within 5 years of transplantation [75]. In addition, longterm immunosuppression has serious adverse effects, such as increasing the risk of cancer and infection, especially in people chronically infected by cyto megalo virus, which make up 50 to 80% of the urban adult population in the US [76,77]. Therefore, in the course of developing hESCbased cell therapies, it is important to determine the risk/benefit ratio of therapies that require longterm immune suppression. It will greatly facilitate the development of hESCbased therapy if new effective approaches can be developed to induce immune tolerance to hESCderived allogeneic cells without long term systemic immune suppression. Allogeneic immune rejection is primarily mediated by Tcelldependent immune responses [74]. Several studies have suggested new strategies to induce immune toler ance of allografts by modulating the costimulatory path ways required for T cell activation, including the pathways involving cluster of differentiation 28 (CD28) and B7.1/2 (B7), CD40 and its ligand (CD40L, also known as CD154), or lymphocyte functionassociated antigen 1 (LFA1) and intercellular adhesion molecule 1 (ICAM1) [78,79]. These costimulatory pathways are crucial for Tcell activation, and in their absence T cells tolerate the antigen instead of being activated. For example, cytotoxic Tlymphocyte antigen 4 (CTLA4) disrupts the inter action between the CD28 and B7.1/2 proteins that is impor tant for Tcell activation [80]. CTLA4Ig, a secreted fusion protein that blocks the binding of B7 to CD28, can suppress autoimmunity and allograft rejection in experimental mouse models [81]. Recombinant CTLA4Ig has been approved for treating patients with rheumatic arthritis and juvenile idiopathic arthritis [79]. Although data on the impact of costimulatory inhibition on allogeneic immune rejection in patients are lacking, xenograft transplantation experiments have shown that blocking costimulatory pathways with CTLA4Ig, anti CD40L or antiLFA1 prolongs the survival of hESC derived xenografts in immune competent mice [82].
In addition to the disruption of costimulatory path ways, recent clinical studies have shown that infusion of bone marrow or HSCs from a donor into a recipient before the transplantation of allografts from the same donor can induce immune tolerance of allografts, probably by transiently establishing hematopoietic chimer ism (in which hematopoietic cells come from both the donor and recipient) in the recipients and thus the immune toler ance of allografts from the donor [83]. Recent success in using mixed chimerism to induce longterm immune tolerance to renal transplantation in patients further supports the feasibility of this approach in inducing

. The interaction between the ESC/iPSC-derived cells and T cells in recipients.
In addition to the engagement of the T-cell receptor (TCR) and the allogeneic or syngeneic major histocompatibility complex (MHC) containing self-or foreign peptides on the surface of ESC-or iPSC-derived cells, secondary activation pathways such as those involving the interaction of CD28 with B7 and CD40 with CD40L are also critical for T-cell activation. CTLA4 has higher binding affinity for CD28 and can effectively block the interaction between B7 and CD28, leading to the inhibition of T-cell activation. immune tolerance to allograft transplantation [84]. There fore, it is possible that immune tolerance to hESC derived allogeneic cells could be achieved by the induction of hematopoietic chimerism in the recipient using HSCs derived from the same hESCs. However, although several studies have reported protocols to differentiate hESCs into potential multipotent HSCs [85,86], these potential HSCs have failed to repopulate the human immune system in vivo. The feasibility of inducing immune tolerance of hESCderived allografts through hematopoietic chimerism depends on the successful derivation of authentic HSCs from hESCs. The identification of surface protein CD49f as a marker for human HSCs could facilitate the purification of hESC derived HSCs [87], and the recent discovery of the important roles of Sallike protein 4 (SALL4) in promot ing the selfrenewal of HSCs might also help to expand hESCderived HSCs [88]. Ectopic expression of SALL4 in human HSCs can lead to significant expansion of human HSCs in defined culture conditions.

Immunogenicity of iPSC-derived cells
iPSCs that have been reprogrammed from somatic cells of patients with defined factors could offer a key advantage in regenerative medicine as a renewable source of autologous cells for human cell therapy [8992]. The initial success of iPSC production was achieved with retroviral vectors, but these vectors pose safety concerns for clinical application because the random integration of the retroviral vectors into the genome increases genomic instability [93]. In addition, the spontaneous reexpres sion of randomly integrated reprogramming factors, most of which have oncogenic potential, can pose a serious cancer risk [43]. Significant progress has been achieved in developing various approaches to produce integrationfree hiPSCs, including the use of an episomal vector that can be maintained extrachromosomally in mammalian cells, and proteinbased modified mRNA, miRNA and chemical biology approaches [94100]. In addition, chemical biology approaches have been developed to improve reprogramming efficiency when combined with the standard reprogramming factors [99,101105]. These approaches will improve the feasibility for clinical development of hiPSCs. In one proofofprinciple study to demonstrate the use of iPSCs in autologous cell therapy, mouse iPSCs were reprogrammed from adult fibroblasts of a sickle cell anemia mouse model, and corrected for the causative genetic mutation by homologous recombination [106]. The corrected mouse iPSCs were then differentiated into HSCs that were transplanted back into the lethally irradiated mouse model to cure the disease [106]. More recent studies have also indicated the feasibility of using iPSCderived cells to treat mouse models of hemophilia A and liver degeneration in an immunodeficient back ground [107,108]. Although these studies demonstrate the functionality of iPSCderived cells in vivo, their experi mental design has not allowed the evaluation of the immunogenicity of iPSCderived cells because the immune system of the recipients is either deficient or ablated by high dose irradiation.
To examine the immunogenicity of iPSCderived cells, it would be ideal to differentiate iPSCs into various lineages of cells with therapeutic potential, which are then transplanted into syngeneic recipients. However, owing to the lack of physiologically relevant differen tia tion protocols and chemically defined conditions, it is possible that factors involved in the in vitro differentiation process itself might induce immune responses in syn geneic recipients [109,110]. It would therefore be difficult to determine whether any observed immunogenicity was intrinsic to the iPSCderived cells or induced by the differ en tiation process. To resolve this issue, a recent study took advantage of the ability of ESCs and iPSCs to form teratomas in mice, enabling the simultaneous evaluation of the immunogenicity of various differen tiated cell types, including terminally differentiated cell types such as muscle and pancreatic β cells [111]. The feasibility of this approach was supported by the finding that C57BL/6 (B6) mice show immune tolerance for teratomas formed by B6 ESCs without any evidence of immune rejection, whereas teratomas formed by allo geneic ESCs undergo robust immune rejection in B6 mice. However, it was found that cells derived from B6 iPSCs can become immunogenic in syngeneic recipients through activation of antigenspecific Tcellmediated immune responses [111]. Because there are few un differ entiated iPSCs in the formed teratomas, it appears that the observed Tcell response was directed against the cells differentiated from iPSCs. Therefore, the Tcell dependent immune responses directed against the iPSC derived grafts appeared to have been activated as a result of the abnormal expression of immunogenic proteins during iPSC differentiation. Furthermore, despite the great similarity in global gene expression between ESCs and iPSCs [8992], recent wholegenome bisulfite sequen cing studies have identified abnormal epigenetics in iPSCs [112114]. Therefore, although remaining to be confirmed, the abnormal epigenetics of iPSCs could contribute to abnormal gene expression and immuno genicity during differentiation of iPSCs. However, the immunogenicity of lineagespecific cells differentiated from iPSCs in vitro remains to be evaluated.

Conclusions and future directions
Tremendous progress has led to the initiation of clinic trials of two hESCbased cell therapies for spinal cord injury and macular degeneration. However, several major hurdles remain for the clinical development of hESC based therapies, including those currently undergoing clinical trial. For example, it is crucial to elucidate the pathways that are important to maintain the genomic stability of hESCs during their longterm expansion and lineagespecific differentiation. In addition, efficient and scalable approaches to remove undifferentiated hESCs from their derivatives are needed to eliminate the risk of teratoma formation associated with pluripotent stem cells. Considering the severe adverse side effects of chronic systemic immune suppression, the risk/benefit ratio of hESCbased cell therapy must be carefully evaluated. Therefore, the development of novel approaches to protect hESCderived cells from allogeneic immune rejection will greatly improve the feasibility to develop hESCbased cell therapies. In this context, the optimi zation of humanized mouse models with a functional human immune system will be important for developing and evaluating new approaches to induce immune tolerance to hESCderived allografts.
Although hiPSCs have become a valuable tool for modeling human diseases and for drug discovery, the develop ment of hiPSCs for cell therapy has been challenged by the recent discovery of epigenetic and genetic abnormalities associated with induced pluri potency. The immunogenicity of cells derived from human iPSCs remains to be vigorously examined in the context of an autologous human immune system. There fore, a key challenge for the clinical development of iPSCs is to understand the mechanisms underlying induced pluripotency so as to optimize reprogramming condi tions and minimize genetic and epigenetic abnormalities. Recent findings have supported the feasibility of achieving this goal. Hypoxic conditions and small mole cules that promote anaerobic metabolism can also promote reprogramming [99,115]. Reactive oxidative stress, a byproduct of oxidative phosphorylation during aerobic metabolism, has been considered to be a major inducer of genomic mutations [116], so hypoxic condi tions or compounds that suppress oxidative phosphory lation might reduce the genomic instability of iPSCs. In addition, elucidating the roles of tumor suppressors in induced pluripotency could also facilitate the develop ment of strategies to reduce the genomic instability associated with iPSCs.
The epigenetic state of pluripotent stem cells derived by somatic nuclear transfer appears to be more similar to that of ESCs than the epigenetic state of iPSCs is to that of ESCs [117]. This raises the possibility that the epi genetic differences between ESCs and iPSCs could be minimized by including additional oocytespecific factors in the reprogramming cocktail. In addition, several small molecules that are known to regulate the activities of enzymes involved in epigenetics can also increase reprogramming efficiency. It is therefore possible that epigenetic differences between ESCs and iPSCs could be minimized by the addition of these small molecules in reprogramming cocktails [118].
Genomic approaches will continue to be crucial in evaluating the genomic and epigenomic stability of pluripotent stem cells. These approaches will need to be performed alongside functional analyses to facilitate a better understanding of the functional stability of these cells and their clinical utility. Rapid progress in genomic approaches and in understanding human pluripotent stem cell biology suggests that the challenges discussed here that might hinder the development of pluripotent stem cellbased therapy can be overcome.

Competing interests
The authors declare that they have no competing interests.