The tumorigenic potential of pluripotent stem cells: What can we do to minimize it?

Human pluripotent stem cells (hPSCs) have the potential to fundamentally change the way that we go about treating and understanding human disease. Despite this extraordinary potential, these cells also have an innate capability to form tumors in immunocompromised individuals when they are introduced in their pluripotent state. Although current therapeutic strategies involve transplantation of only differentiated hPSC derivatives, there is still a concern that transplanted cell populations could contain a small percentage of cells that are not fully differentiated. In addition, these cells have been frequently reported to acquire genetic alterations that, in some cases, are associated with certain types of human cancers. Here, we try to separate the panic from reality and rationally evaluate the true tumorigenic potential of these cells. We also discuss a recent study examining the effect of culture conditions on the genetic integrity of hPSCs. Finally, we present a set of sensible guidelines for minimizing the tumorigenic potential of hPSC‐derived cells. © 2016 The Authors. Inside the Cell published by Wiley Periodicals, Inc.


Introduction
Human pluripotent stem cells (hPSCs) can be expanded indefinitely and differentiate into any cell type in the body. These two key features make hPSCs an ideal source of transplantable cells for medical and research purposes, which has raised hope that they could fundamentally change the way that many diseases are treated. To date, three clinical trials have been initiated in the USA for use of differentiated derivatives of hPSCs to treat spinal cord injury, macular degeneration, and Type 1 diabetes [1][2][3][4][5][6]. There have been no adverse reactions reported in people treated in these clinical trials, although the studies are still at an early stage. Although one can imagine multiple ways in which transplanted cells could negatively affect a recipient, the most troubling is the possibility that the transplanted cells could harm recipients by becoming tumorigenic [7][8][9].
This concern arises from a well-known characteristic of undifferentiated pluripotent cells; undifferentiated hPSCs form teratomas, a complex tumor containing a hodgepodge of differentiated cell types, when they are transplanted to immunodeficient mice. However, teratomas are usually benign, and there are no reports of any types of malignant tumors arising from genetically normal undifferentiated hPSCs or their derivatives. In addition, there are no therapies planned that use undifferentiated hPSCs; therapeutic applications are limited to differentiated derivatives of hPSCs, such as retinal pigmented epithelium and pancreatic islet cells. These cell types have been thoroughly tested for tumorigenicity in the preclinical animal safety studies required by the FDA and other regulatory agencies; although preclinical safety studies are not usually published, approval would have been dependent on showing that the cell types currently in clinical trials did not form tumors.
Given this considerable amount of information supporting the safety of hPSCs, why are there continued concerns about tumorigenicity of hPSC-based therapies? The worry was originally based in the fear that undifferentiated hPSCs could be intermingled in differentiated cell populations used for transplantation. Over the last few years, this concern has abated as researchers have developed methods for assuring the absence of undifferentiated cells in their differentiated populations [10][11][12][13]. However, the lessening uneasiness about teratomas has been offset with a growing concern about the effects of genomic aberrations that have been repeatedly reported in hESC and iPSC lines over the last few years [6,[14][15][16][17][18]. Cancers always have genetic aberrations, but do the aberrations discovered in hPSCs indicate that they are potentially cancerous? In this review, we focus on the specific genomic abnormalities that have been noted in hPSCs, assess whether these mutations are associated with cancers, and discuss methods that have been developed to minimize the acquisition of genomic (and epigenetic) changes during culture of hPSCs.
Do culture conditions contribute to genomic abnormalities in human pluripotent stem cells?
Over the last several years, multiple researchers have observed that genetic changes ranging from point mutations to aneuploidies arise as hPSCs are expanded in culture [19][20][21][22][23][24][25][26][27]. Because the expansion of cells is necessary for cell therapy applications, there has recently been a focus on the development of culture methods that minimize genomic aberrations. In the first longterm systematic study of how different culture conditions affected genomic stability over time in hPSCs [28], the most commonly used human embryonic stem cell (hESC) line, WA09, was continuously cultured for an extended time to determine what genomic changes occurred and when, and whether specific culture methods would affect the number and type of mutations that occurred. The cells were taken at an early passage and then split into four different conditions ( Fig. 1): (1) 'MefMech', in which cells were cultured on a mitotically inactivated mouse embryo fibroblast (MEF) feeder layer and passaged by mechanical cutting of colonies into small pieces; (2) 'MefEnz', in which cells were grown on a mitotically inactivated MEF layer and dissociated with an enzymatic solution called Accutase TM ; (3) 'EcmMech', in which cells were cultured in a serum-free defined medium on an extracellular matrix (ECM) substrate and passaged mechanically; and (4) 'EcmEnz', in which cells were cultured in a defined medium on ECM but passaged enzymatically [28,29]. These culture methods have been extensively described in Human Stem Cell Manual: A Laboratory Guide written by the authors [29]. Cells were then  cultured under these conditions continuously for over 2 years, for at least an additional 100 passages (for some conditions as many as 161 passages), and samples were taken for analysis every five passages.
To determine whether findings with hESCs could be extrapolated to induced pluripotent stem cells (iPSCs), the study used a similar experimental design to evaluate the effects of culture conditions on the genomic stability of iPSCs. For this part of the analysis, the researchers used three independent iPSC lines, all derived from the same parental human fibroblast line by retroviral transduction. Based on trends seen with the hESCs, only two conditions were analyzed: (1) 'MefMech' (MEF layer and mechanically passaging) and (2) 'EcmEnz' (serumfree defined medium on ECM with enzymatic passaging). These cells were cultured for an additional 25-35 passages, with samples harvested every five passages.

Culture conditions modulate genomic stability over time
To evaluate the integrity of the genome at each 5-passage interval, cells were analyzed with single nucleotide polymorphism (SNP) genotyping using the Illumina OmniQuad version 1 platform with CNVPARTITION 3.2.0 used for CNV calling. The results from this analysis were very clear (Fig. 2). First, both hESCs and iPSCs accumulated genomic alterations over time in culture, which was not surprising and has been reported previously [19,20,41]. This finding supports the idea that the culture dish acts as a selective environment in which genomic alterations that afford increased survival are positively selected and become dominant within the cell population. It is interesting to note, however, that in one of the conditions (MefMech), the cells could be passaged for long periods of time without detectable alterations.
Secondly, it is clear that the most labor-intensive conditions, in which cells are physically cut into small pieces for passaging, keep the cells more genomically stable than techniques in which cells are dissociated enzymatically. Not only were there fewer duplications and deletions in mechanically passaged cells, but the duplications and deletions that arose were smaller (Fig. 2). Although fewer alterations were seen in the iPSCs because they were cultured for less time than the hESCs, they showed a similar trend toward more genomic instability when passaged enzymatically. These findings, wherein more labor-intensive mechanical passaging techniques preserve the genome better, may be disappointing to many hPSC culturists, but they are important findings, especially considering that many labs have abandoned mechanical passaging all together, to save time, or because this method requires greater technical expertise. Based on these observations, it is clear that, compared with current methods, manual passaging on feeder layers would be preferable for preserving genomic integrity when hPSCs are being expanded for differentiation into cells to be used for cell therapy. The development of new, simpler culture methods should use the more laborintensive methods as a comparative benchmark to assure quality control.

Duplications and deletions: familiar foes and new enemies
Another finding gleaned from the CNV analysis was that the specific regions of the chromosomes that were altered with time in culture were not unfamiliar regions. In particular, duplication of a small segment of the long arm of chromosome 20 (20q11) was seen in all of the conditions examined (Fig. 3A). This segment has been frequently reported as duplicated in prior reports [20,41,42]. Interestingly, this region contains the BCL2L1 gene (also known as BCLXL), and previous studies have shown that hESCs with this duplication express twofold to threefold higher levels of BCL2L1 mRNA and threefold more protein [43,44]. Consistent with this gene's anti-apoptotic function, hESCs with the duplication were shown to be significantly less prone to apoptosis in situations in which they were placed under stress by being dissociated into single cells [43]. Thus, it is easy to understand why cells with a duplicated 20q11 region of the genome would have a survival advantage over other cells.  Another chromosomal region that was frequently duplicated in almost all of the conditions studied was chromosome 12. Chromosome 12 was one of the first chromosomes found to be recurrently gained in hESCs [19,41]. The long-term study found both duplication of the whole chromosome 12 as well as duplication of the short arm of chromosome 12 (Fig. 3B). Although it is not known which gene(s) on chromosome 12 provide a selective advantage when duplicated, previous studies have shown that cells with an extra copy of chromosome 12 may have a reduction in pluripotency [45]. Curiously, one study showed that a region containing an expressed NANOG pseudogene was the smallest common region to be duplicated in hPSCs, suggesting that NANOGP1 may not only be functional but also provides a survival advantage to hPSCs in culture [20].
In addition to the duplications of chromosome 12 and 20q11, one consistent deletion was observed in many but not all of the conditions analyzed. This deletion was in a region on the short arm of chromosome 17 encompassing a region containing the TP53 (tumor protein p53) gene (Fig. 3C). The deletion was seen at late passage in all conditions except the cells grown on a MEF feeder layer and passaged mechanically (MefMech). All examples of this deletion overlapped in a 158 kb region containing 17 genes. Because this study was paired with gene expression analysis, it was possible to determine that only 3 of the 17 genes in the commonly deleted region showed a consistent reduction in mRNA levels. These three genes were SENP3 (SUMO1/sentrin/SMT3 specific peptidase 3), SOX15 (SRY- BOX 15), and TP53. Interestingly, this was the first report documenting the deletion of a TP53-containing region in hPSCs; but given TP53's wellknown role as a tumor suppressor gene, its deletion would be expected to provide a survival advantage. Later, we will discuss the implications of this deletion on the safety of cell therapies.

Most genomic changes do not cause dramatic phenotypic changes
It is well established that the DNA sequences in the cells of a single person are not all the same; thus all of us are genomic mosaics, having cells in our bodies that contain genomic variations, some of which would be considered to be deleterious [46][47][48][49][50][51][52][53][54][55][56][57][58][59][60][61][62][63]. In most cases, these genomic differences do not lead to blatant phenotypic differences. To determine whether the genomic aberrations observed in the long-term culture study would lead to detectable phenotypic differences in hPSCs, cells from each of the four culture conditions were analyzed based on a number of different phenotypic parameters. Some of the aspects examined were proliferation, cell cycle distribution, telomerase activity, apoptosis, pluripotency, and embryoid body (EB)-based differentiation. Each phenotypic assay was assessed after at least 50 passages in a particular culture condition. In contrast to the findings observed with genomic aberrations, the correlation between culture conditions and phenotype was much less overt. Cells passaged enzymatically and/or in feeder-free conditions had a small but significantly higher rate of proliferation and a higher percentage of cells in S-phase of the cell cycle ( Fig. 4A-C) compared with the cells on feeder layers. This may suggest that an underlying cause in the increase of genomic alterations could be an increase in the growth rate. In the analysis of telomerase activity, telomere length, and apoptotic rate, the differences among the culture conditions were almost negligible (Fig. 4D-F). All cells remained pluripotent as assessed by the differentiation of all three germ layers in EBs.
Whole genome gene expression and DNA methylation were analyzed to assess epigenetic changes in the cells over time in culture. Cells in all culture conditions retained a pluripotent gene expression signature throughout the culture period [64], and there were no large changes in DNA methylation. When all conditions were compared, it was apparent that the cells cultured on an MEF feeder layer and passaged mechanically (MefMech) were more similar to early passage hPSCs than cells cultured in other conditions, again suggesting that this culture method is best among those tested for preserving the genomic and epigenetic integrity of hPSCs.
Although most of the assays used in this study did not expose striking changes in phenotype caused by different hPSC culture conditions, phenotypic differences were observed in a teratoma assay. Cells under all of the conditions were able to form teratomas in immunodeficient mice, consistent with the gene expression and EB assays for pluripotency. However, teratomas from all of the culture conditions except for the MefMech condition contained OCT4 (also known as POU5F1) positive foci in the teratomas (Fig. 4G-H). The extent of OCT4positive tissue was higher in cells with more severe genomic abnormalities. Retention of OCT4-positive cells, which are presumably still undifferentiated, in teratomas has been reported previously in hPSCs with genomic aberrations [65]. Cells with a large number of genomic aberrations are unlikely to be used for transplantation, so the retention of OCT4-positive cells will probably never be relevant to cell therapy. However, this may suggest that the cells have a decreased differentiation capacity.

Human pluripotent stem cell-mediated tumorigenesis: is it likely?
The results of tumorigenicity studies performed as a prerequisite for approval for clinical trials are not usually made public, so we do not have access to the data supporting the current clinical trials using derivatives of hPSCs. We also do not have information about the genomic integrity of the cells used for transplantation. The FDA generally requires karyotyping of the cells, while the Japanese regulators require that the cells be completely sequenced. While it is safe to assume that the cells currently in trials passed these tests, it would be useful to have published information available on experiments that were deliberately designed to directly ask the question about the effects of genomic abnormalities on transplanted cells. The requisite experiments would require that hPSCs with specific genomic changes be differentiated and then transplanted into animal models of a disease, and these experiments have not been done in a systematic manner. Step 1: avoid undifferentiated cells However, even without this knowledge, we can at least begin to identify common sense guidelines that can be used to minimize the potential threat of tumorigenesis. First, despite the fact that we do not know how genomic alterations affect the tumorigenicity of hPSCs, we do know that genomically unaltered hPSCs make teratomas when injected into an immunocompromised host. This is not a concern for cell therapy as long as the differentiated cells used for transplantation are free of undifferentiated cells. Thus, thorough differentiation is the best line of defense against hPSC-derived tumors. Differentiation will, depending on the differentiation scheme used, eliminate many, if not all, of the undifferentiated cells in the preparation. The application of differentiation protocols in which undifferentiated cells are completely eliminated is of paramount importance for minimizing the risk of hPSCbased tumorigenicity.
It is important to note, however, that even with sorting steps and thorough optimization, differentiation protocols are not always 100% efficient. Thus, some undifferentiated cells could be transplanted along with differentiated hPSC derivatives. These undifferentiated cells would have the ability to form tumors in immune compromised individuals regardless of whether or not they have genomic aberrations. Importantly, however, studies suggest that the number of undifferentiated cells that must be transplanted in order for a tumor to form is actually quite high [66]. Specifically, one study showed that, depending on the site of injection, between 10 4 and 10 5 undifferentiated hESCs need to be transplanted in order for a tumor to form in an immunocompromised mouse [66]. Assays now in use for quality control of cell populations for transplantation generally use a PCR-based assay that can detect one undifferentiated cell in 10 5 differentiated cells [10]. The number of transplanted cells will be different for treatment of different conditions, but the existing assays are theoretically adequate for transplants of 10 9 cells, which cover most of the current cell therapies under development.
Step 2: use low passage cells cultured in a responsible manner Another consideration is that this study, as well as many others, show that time in culture is the single biggest factor contributing to genomic abnormalities in hPSCs [19,20,41]. This finding is consistent with the idea that the culture dish is a highly selective environment in which cells that are able to gain a survival advantage through genomic alterations are selected for and go on to proliferate and potentially take over the population. Even though we do not definitively know that genomic alterations make hPSCs more tumorigenic, it makes sense to minimize this possibility. To do so, we propose that hPSCs to be used for cell therapy approaches should be as low passage as possible.
Lastly, we should consider the nature of the genomic alterations and, based on available data, determine which mutations are likely to be benign and which are likely to cause problems. Several studies have identified chromosome 12 and 20q11 as hot spots for duplication in long-term culture. These two regions have been reported to be duplicated in a small number of tumor types, including chromosome 12 duplications in embryonal carcinoma and chronic lymphocytic leukemia [67][68][69], and skin cancer, colorectal cancer, and chronic lymphocytic leukemia for chromosome 20 [70][71][72]. Although these regions are duplicated in these cancers, it is not believed that they are intrinsically oncogenic on their own. Thus, these two chromosomal alterations should be avoided, but may not present a universal threat in terms of tumorigenicity. In contrast, deletion of the tumor suppressor TP53, which occurred after very long-term culture under some conditions, raises a red flag. Loss of TP53 is estimated to occur in up to 50% of cancers and has been called the 'guardian of the genome' because of its role in preventing DNA damage and oncogenic mutations. Because the long-term study showed that cells with TP53 deletions are positively selected in hPSC populations after very long times in culture, it now becomes crucial to not only limit the expansion of hPSCs and choose culture conditions carefully, but also to test for loss of TP53 in populations of differentiated cells used for transplantation.
Step 3: in vivo tumorigenicity testing Of significant importance is the fact that immature progenitor cells can also form tumors or ectopic tissue, as has been previously shown with immature neural progenitor cells [73]. This is why preclinical in vivo tumorigenicity testing is critical to assess the safety of the final clinical product to account for not only residual undifferentiated cells, but also immature

Conclusions and outlook
While there is no evidence that hPSC-derived cells become tumorigenic in human transplants, there have been few clinical trials involving hPSC derivatives. However, given the devastation that would be done to the field of regenerative medicine if a cell therapy patient were to get an hPSC-derived tumor, it makes sense to err on the side of caution. Because of this, and the fact that undifferentiated hPSCs are tumorigenic even when they do not have genomic anomalies, we propose the following strategy for minimizing tumorigenic potential (Table 1). First, differentiation strategies should be optimized, following a well-established, stringent protocol that includes a step to identify and eliminate cells that have not fully differentiated from the cell preparation for transplantation. Second, the time that cells destined for transplant spend in culture should be minimized, as studies show that time in culture is the greatest contributor to genomic alterations. Third, if at all possible, hPSCs that will be used for cell therapy should be expanded under culture conditions that minimize selection for genomically aberrant cells. And finally, it would be wise to screen for genomic alterations that are highly significant for cancer risk, such as the TP53 deletion. We think that by following these common sense precautions it should be possible to avoid transplantation of cells that have tumorigenic potential.