Genomic Instability in Pluripotent Stem Cells: Implications for Clinical Applications*

Human pluripotent stem cells (hPSCs) are known to acquire genomic changes as they proliferate and differentiate. Despite concerns that these changes will compromise the safety of hPSC-derived cell therapy, there is currently scant evidence linking the known hPSC genomic abnormalities with malignancy. For the successful use of hPSCs for clinical applications, we will need to learn to distinguish between innocuous genomic aberrations and those that may cause tumors. To minimize any effects of acquired mutations on cell therapy, we strongly recommend that cells destined for transplant be monitored throughout their preparation using a high-resolution method such as SNP genotyping.

Human pluripotent stem cells (hPSCs) are known to acquire genomic changes as they proliferate and differentiate. Despite concerns that these changes will compromise the safety of hPSC-derived cell therapy, there is currently scant evidence linking the known hPSC genomic abnormalities with malignancy. For the successful use of hPSCs for clinical applications, we will need to learn to distinguish between innocuous genomic aberrations and those that may cause tumors. To minimize any effects of acquired mutations on cell therapy, we strongly recommend that cells destined for transplant be monitored throughout their preparation using a high-resolution method such as SNP genotyping.
In the 15 years since the derivation of human pluripotent stem cells (hPSCs) 2 was first reported, the research community has been moving steadily toward the goal of using these remarkable cells as replacements for tissues that are lost as a result of heart disease, diabetes, neurodegenerative disease, and skeletal and muscular disorders and injuries. The progress toward therapies has required incremental improvements in culture and differentiation methods, GMP and industrial-scale cell culture, legal permissions, and considerable investment from both public and private entities. At the same time that new hPSC therapies are being launched, concerns regarding their safety remain a major hurdle. In the worst case scenario, the transplanted cells may harm the patient by becoming cancerous or inducing cancers. For cell types derived from hPSCs, there are two major concerns about tumorigenicity. The first is intrinsic to pluripotency; undifferentiated hPSCs generate teratomas, germ cell tumors that are usually benign, when they are transplanted to immunodeficient mice. Although all of the clinical applications require the cells to be differentiated, there is concern about residual undifferentiated cells in transplanted populations. Second, mutations in transplanted cells are a major concern because genomic mutations are associated with tumorigenicity. In this minireview, we summarize the available information about instability in the genomes of hPSCs and consider the potential impact of genomic instability on the safety of current and future hPSC-derived transplantation therapies.

Clinical Trials Using hPSC-derived Cells
The United States Food and Drug Administration (FDA), as well as regulatory agencies in other countries, requires extensive preclinical safety trials in animals to determine whether hPSCs become cancerous or induce cancers, and the primary goal of a Phase 1 clinical trial is to monitor for adverse effects, including infection and tumorigenicity. The first clinical application of hPSCs, a safety trial using oligodendrocyte progenitor cells derived from human embryonic stem cells (hESCs) to treat spinal cord injury, was initiated in October 2010 and sponsored by the biotechnology company Geron Corp. (ClinicalTrials. gov, study NCT01217008). Geron Corp. had funded the derivation of hESCs in the Thomson laboratory in the late 1990s and held an exclusive license to the patents for hESCs for use in cell therapy for neurological disease, cardiac disease, and diabetes. The Geron trial was discontinued because of financial constraints in November 2011 after transplantation of five patients, with the promise of following those patients for another 15 years (1). No adverse effects have been reported.
All of the current ongoing clinical trials have followed the lead of Advanced Cell Technology in using retinal pigmented epithelial (RPE) cells derived from hESCs for macular degeneration or dystrophy (Table 1). In addition to the ongoing trials, a RPE cell trial has been approved in Japan using human induced pluripotent stem cells (hiPSCs) instead of hESCs. 3 There are two logical motivations for the current strategy of using hESC derivatives for eye disease. First, the concern about tumors is lessened because transplantation to the eye allows the cells to be observed; if a tumor arises, the cells can be removed. Second, using FDA-approved cell lines simplifies the clinical application, and the only hPSCs that have so far been approved are the hESC lines derived for the first publication in 1998 (2). It is worth noting that these cell lines were derived before improved culture methods and routine genomic analysis techniques were developed. Although these oldest lines have taken on the mantle as the gold standard, it could also be argued that because of their suboptimal derivation and extended culture, these cell lines are very likely to have genomically evolved over the last 15 years. * This work was supported by National Institutes of Health Grants 5R33 MH087925-04 (to J. F. L.) and 1R21 DA032975-01 (to P. Sanna and J. F. L.). This work was also supported by California Institute for Regenerative Medicine (CIRM) Grants CL1-00502, TR01250, and RM1-01717 (to J. F. L.), TR3-05603 (to J. F. L. and T. Lane), and RB3-05022 (to J. Gottesfeld). This is the fourth article in the Thematic Minireview Series "Development of Human Therapeutics Based on Induced Pluripotent Stem Cell (iPSC) Technology." 1 To whom correspondence should be addressed. E-mail: jloring@scripps. edu. 2 The abbreviations used are: hPSC, human pluripotent stem cell; hESC, human embryonic stem cell; RPE, retinal pigmented epithelial; hiPSC, human induced pluripotent stem cell; CNV, copy number variation; NSC, neural stem cell.
malities that have been observed in hPSCs range from whole chromosome aneuploidies (including mosaic) to subchromosomal aberrations, including gene duplications and deletions and point mutations (10 -13). Some of the duplications are detected repeatedly in hPSCs of diverse origin. Fig. 1 shows a map of duplicated regions detected by SNP genotyping. These karyotypically detectable trisomies of chromosomes 1, 12, 17, and X and amplification of 20q have been detected in as many as 34% of hPSC lines examined (10,13,14). Common Genomic Changes Occur during Generation and Long-term Culture of hPSCs-A recent large-scale study of Ͼ100 hPSC lines from many laboratories demonstrated that late passage hPSCs are twice as likely to have genomic changes than early passage cells (10). Thus, selective pressure exerted over time in culture may be a major contributor to genomic instability in both hESCs and hiPSCs. Not surprisingly, many of the same chromosomal abnormalities found in hESCs are also found in hiPSCs. Although it has been reported that hiPSCs may be more likely to gain chromosome 8 and less likely to gain chromosome 17 than hESCs, both events have been documented in both types of cells (13,15,16). Thus, as has been seen with other studies, the use of larger sample sizes may obviate any perceived differences between hESCs and hiPSCs (11,17,18).
It has been noted that hiPSCs may have higher numbers of subchromosomal copy number variations (CNVs) than hESCs (11,16,19). The source of such variations in hiPSCs is not yet clear, but it was recently shown that somatic mosaicism in cultures of fibroblasts causes much of the variation in CNVs among human and mouse iPSCs (20,21). This is because each iPSC colony is the clonal offspring of a single reprogrammed somatic cell, amplifying whatever variation was present in the parental cell. A minor genomic variation profile may not be detectable in mosaic cultures but is immediately apparent in  the profiles of hiPSCs derived from that culture. In addition, because hiPSC clones need to survive as single cells before they are isolated and expanded, there is strong selective pressure that favors the best adapted cells, and this may contribute to the genomic variation detected in hiPSCs (22,23). Interestingly, although it seems logical that different methods for reprogramming should cause different genomic changes, the reprogramming method does not appear to affect the frequency or types of genomic changes in hiPSCs (13). In many cases, the duplications commonly found in hPSCs include genes that can give cells a growth or survival advantage, and it is generally agreed that as cells divide in culture, there is selection for those that are better adapted to the culture conditions. For example, the common subchromosomal duplication in chromosome 20q was mapped using high-resolution SNP analysis. Fig. 2 shows that the region that is duplicated in multiple cell lines contains at least 25 protein-coding genes and a microRNA. Among the genes is BCL2L1 (20q11.21;, also known as Bcl-x L ), which has been reported to enhance the survival of hESCs (24). Thus, it is easy to imagine how overexpression of BCL2L1 could give cells a selective advantage. However, other genes in this region could also have an impact, as could the microRNA miR-1825. This microRNA has Ͼ400 predicted targets (TargetScan), and it could act by suppressing genes that inhibit growth or enhance apoptosis.
It is important to note that non-pluripotent stem cells, including neural stem cells (NSCs), hematopoietic stem cells, and mesenchymal stem cells, also show frequent chromosomal abnormalities that are typical for each cell type. For example, independent cultures of NSCs showed duplication of chromosome 19, and mesenchymal stem cells showed a deletion of chromosome 13 (15).
Genomic Changes Occur during Differentiation of hPSCs-Besides the genomic alterations that take place while hPSCs are cultured in their pluripotent state, it is also possible for such changes to occur during their differentiation. As culture conditions are changed to direct differentiation, new selective pressures are applied to the cells, potentially selecting for new genomic variants. This type of change is typically difficult to detect because differentiation is often associated with decreases in proliferation, and standard karyotyping procedures require dividing cells. The earliest reported example of genomic changes arising during differentiation was found during a largescale SNP genotyping study (11). This report showed that a genomically abnormal subpopulation present in cultures of undifferentiated WA07 hESCs was selected for in a cardiac differentiation experiment. In this example, after only 5 days, the differentiated population was shown to be greatly enriched for cells with multiple duplications in chromosome 20. In a more recent study (25), researchers found that hPSCs differentiated into NSCs lost their ability to senesce, and this loss of senescence was associated with amplification of a region of chromosome 1q. Importantly, this occurred in several lines (both hESCs and hiPSCs) tested and was not detected in the undifferentiated lines. Interestingly, however, when the NSCs with the 1q amplification were injected into immunocompromised rat brains, they did not form tumors (25). There are still too few examples to know how often differentiation-driven selection occurs, and the functional consequences remain to be investigated. However, it is important to note that even when the undifferentiated hPSC population is free of detectable genomic abnormalities, selection during differentiation can amplify abnormal cells. Because these cells are the ones designated for cell therapy, it is critical that the last step before transplantation be a final check of the genomic state of the cells.
Other Genomic Changes-There are several examples of dramatic genomic changes that appear when cells are reprogrammed. Studies of some trinucleotide repeat diseases have reported changes in the repeat length following reprogramming (26,27). Specifically, in Friedreich ataxia, the GAA/TTC triplet repeat length in the FXN (frataxin) gene appeared to change following reprogramming of patient fibroblasts (26). In a second example, reprogramming of cells from fragile X syndrome patients, in which the CGG/CCG repeat is located in the 5Ј-untranslated region of the FMR1 (fragile X mental retardation 1) gene, resulted in hiPSC clones with varying repeat length (27). Interestingly, there is evidence that a similar phenomenon may occur in vivo during early embryogenesis; there have been reports of monozygotic twins born with different FXN repeat lengths (28,29). Additionally, recent studies have shown that when Down syndrome patient fibroblasts are reprogrammed, they frequently lose their extra copy of chromosome 21 (30, 31). This chromosome loss is seen in as many as 25% of resulting hiPSC clones. Again, this type of chromosomal instability may be present during the early embryo in vivo, as many Down syndrome individuals have been shown to be mosaics (32).

Does Genomic Instability in hPSCs Increase the Likelihood of Tumorigenesis?
Many authors have pointed out the similarities between pluripotent stem cells and cancer cells and wondered if hPSCs are intrinsically cancerous. Much of this speculation rests on the fact that hPSCs can divide indefinitely, a property shared only with cancer cell lines. However, there is very little evidence to support this idea. In their undifferentiated state, hPSCs do form teratomas, which are multilineage complex tumors; usually teratomas are fully differentiated and benign and do not metastasize. Teratomas can be produced from both genomically normal and abnormal hPSCs, and existing evidence suggests that some genomic abnormalities can lead to the more dangerous form of teratoma that retains undifferentiated cells.
Although typical cell therapy strategies involve the transplantation of cells that are at least partially differentiated, purification of these cells from residual undifferentiated cells is not straightforward, and even small numbers of undifferentiated cells can lead to tumor formation (33). Several studies have shown that transplanted populations of mouse or human PSC derivatives containing undifferentiated cells resulted in benign teratoma generation (34,35). Also, residual undifferentiated cells in cultures of hESC-derived dopamine neurons can cause the development of benign teratomas (36) or growth of immature mitotic neuroepithelial cells (37) when transplanted into rat models of Parkinson disease. These examples highlight the need to optimize the differentiation of cell populations destined for transplant.
Disturbingly, the most common types of genomic alterations found in hPSCs are also found in embryonal carcinomas. These malignant cells, which are usually pluripotent like hPSCs, show gains of chromosome 1, 12, 17, and X, the same chromosomal abnormalities that accumulate in hPSCs (38 -41). In addition, another common chromosomal aberration seen in hPSCs, amplification of part of the long arm of chromosome 20, has been seen in yolk sac carcinoma and non-seminomatous germ cell tumors (42)(43)(44). Like cancer cells, hPSCs with genomic alterations have been shown to divide more quickly than their normal counterparts (45,46). Also, hESC-like gene expression signatures have been found in some aggressive human tumors (47). Interestingly, there are few studies that directly show that genomically aberrant hPSCs actually cause tumors; however, there is a report that use of a karyotypically abnormal line in a teratoma assay caused immature teratoma formation, whereas injection of the karyotypically normal parental line caused only benign teratomas (48). This suggests that, in certain circumstances, genomic aberrations may alter the tumorigenic potential of hPSCs.
In addition to the possibility that genomic abnormalities may increase the tumorigenicity of teratomas, it is also possible that such abnormalities could cause the transformation of more differentiated hPSC derivatives. Recent studies examined a WA09 variant line and its tumorigenic properties (46,49). The variant line was originally characterized as karyotypically normal. However, more in-depth analyses using comparative genomic hybridization indicated that it had an amplification of the long arm of chromosome 20, a deletion on the long arm of chromosome 5, and a mosaic gain of chromosome 12. The variant cells and the parental cell line were differentiated into neural precursor cells and injected into the brains of immunocompromised mice. The mice injected with neural precursor cells from the variant line developed tumors with molecular features associated with medulloblastoma, whereas the mice injected with neural precursors from the parental line did not (49). This suggests that, in certain circumstances, hPSC lines harboring genomic abnormalities may be able to generate non-teratoma tumors in vivo after they have been differentiated.

Genomic Mosaicism: What Is Normal?
With the exception of the few examples described above, there is very little evidence linking genomic abnormalities in hPSCs with tumorigenesis. It seems likely that only a few types of abnormalities are actually dangerous and that the majority of genomic aberrations in hPSCs are harmless. In fact, a growing body of evidence is revealing that there is considerable variation in the genomes of normal cells within our bodies (50 -53). Since the development of in vitro fertilization and preimplantation diagnostic screening, it has become clear that cells from the developing embryo are chromosomally mosaic, meaning that the cells show random non-clonal chromosomal gains and losses (54 -57). In addition, this type of mosaic aneuploidy has been seen in normal developing and mature brains (58 -63). These chromosomal changes in the cells lead to changes in gene expression that may contribute to neural diversity (64). In addition, a study designed to determine the source of genomic differences between hiPSCs and the skin fibroblasts from which they were derived estimated that ϳ50% of the CNVs found in hiPSCs were present in the parental fibroblasts (20). This study, along with the Down syndrome and Friedreich ataxia examples mentioned above, indicates that widespread somatic mosaicism is normal.
It has also been shown that retrotransposons such as LINE-1 elements are highly expressed in developing neural cells, where they generate genomic alterations (65). These LINE-1 elements propagate through RNA sequences that are reverse-transcribed into DNA sequences that integrate into the genome. These elements can alter the genome by insertional mutagenesis or by deletion of intervening loci between homologous transposons through recombination. As such, they are thought to increase neural diversity within the developing brain (65)(66)(67). Interestingly, these LINE-1 retrotransposons may be activated during reprogramming (68). Thus, many different normal cells within our bodies contain genomic variations that are well tolerated and may even be beneficial.

Reality Check: What Is the Likely Impact of Genomic Instability of hPSCs on Regenerative Medicine?
All of the available literature indicates that human pluripotent stem cells are often genomically unstable. Selective pressures in the culture dish lead to shifting populations, in which certain advantageous duplications or deletions will eventually dominate. Selection begins when certain clones thrive during initial hPSC derivation and continues when the cells are expanded in culture and when culture conditions are changed to drive specific forms of differentiation. The fact that cells must be expanded and differentiated if they are to be clinically useful means that it will be very difficult to develop culture conditions that maintain homogeneous genomically stable populations. However, the mounting evidence of normal somatic mosaicism in the human body indicates that variation itself is not necessarily a harmful characteristic and may in fact be vital to normal human embryonic development.
The sum of the reviewed literature indicates that despite the fears about genomic instability, there is little evidence that many of the genomic changes detected in hPSCs are likely to be harmful. However, it is also clear that certain mutations will prove to be dangerous, and we do not yet know which mutations fall into this category. Fig. 3 shows an overview of the possible ways in which transplanted hPSC derivatives could lead to adverse effects in patients. For most of these possibilities, there is a means to minimize the potential danger. Making sure that transplant-ready populations contain no undifferentiated cells should eliminate the main source of transplant-induced teratoma tumors or other uncontrolled growth of immature cells. Checking for genomic abnormalities at all stages of cell processing will identify cells that are unfit for transplantation; importantly, we know that populations can become aberrant under differentiation conditions, so every transplant-ready preparation should be tested as late in the process as possible. Although post-mitotic differentiated cells are difficult to karyotype, SNP genotyping methods will reveal both aneuploidies and CNVs. The only genomic mutations that we cannot now detect in advance are those that arise very late in the cell processing or occur after the cells are transplanted.
The reports so far indicate that if precautions like those described above are strictly followed, there is little chance that cells with malignant potential will be transplanted. So far, the record is good; none of the 15 patients who have been transplanted with hPSC derivatives has developed tumors (69). However, this should not be interpreted as a green light for all hPSC-derived transplantations because the Phase 1 trials use low doses of cells; when the cell numbers are increased to therapeutic levels (for cardiac disease, many millions of cells), the probability of a tumor arising in a patient will increase.
Although the probability of an FDA-approved hPSC-derived cell therapy causing harm to a patient appears to be low, the consequences of adverse events are enormous. There is an important lesson from the failures in early gene therapy trials (70,71). If even one patient is harmed in an FDA-approved trial using hPSC derivatives, all further trials will be in serious jeopardy, and the promise of stem cell therapy will be put on indefinite hold. However, heterogeneous populations containing abnormal cells can shift because of selective pressures arising during differentiation protocols, resulting in genomically abnormal differentiated cells (F), which also occur if the population has already become abnormal (G). Lower panel, post-transplantation. H, teratomas can form from subpopulations of undifferentiated cells. Also, if partially differentiated cells proliferate without completing differentiation, they can form tumors. These tumors are generally benign but can be harmful. When genomically normal cells are transplanted, it is possible for these cells to become spontaneously transformed after transplantation (I). If genomically abnormal cells are not detected before transplant, they may give rise to tumors (J).