iPSCs: Unstable Origins?

The advent of induced pluripotent stem cell (iPSC) technology holds great promise to revolutionize personalized cell-based therapies.^1,2,3,4,5 Recent advances in gene transfer technology therefore open up exciting possibilities for efficient and safe genetic modification of autologous cells in this context.⁶ As potential therapeutic applications are being considered, attention has begun to be focused on issues of safety in the clinical arena. This has arisen in part from efforts to understand the biology of reprogramming as well as from an increasing capability to scrutinize the genome and epigenome at high resolution. The derivation of iPSCs is now known to induce changes in DNA sequence, epigenetic profile, and microRNA (miRNA) expression. What remains to be understood is whether these alterations necessarily matter. In this Commentary, we look at the evidence for genomic instability arising through the reprogramming process, and also at the challenges that lie ahead for successful clinical translation.

Retention of reprogramming transgenes in derivative cells, which is clearly undesirable, is potentially associated with mutagenicity arising from persistent gene expression or through insertional mutagenesis. The majority of protocols for generating iPSCs continue to rely on integrating viral systems for delivering the reprogramming factors (Sox2, Oct4, Klf4, c-Myc). These vectors integrate semirandomly throughout the genome, with a preference for certain genomic regions that varies between the groups of viruses from which the vectors have been derived.⁷ In an extensive insertion-site (IS) analysis of eight established iPSC lines derived by lentiviral gene transfer, it was shown that different iPSC clones had no common IS. The number of ISs ranged between 5 to 15 per individual iPSC clone, but none of these could be linked to dominant gain-of-function mutations such as proto-oncogene activation or insertional deactivation of specific genes or gene clusters, which might have been selected for on the basis of enhanced reprogramming or clonal growth.⁸ However, it has also been shown that normal human fibroblasts can be reprogrammed into a pluripotent-like state by using a high dosage of lentiviral vectors devoid of any reprogramming factors. Further analysis of these clones revealed substantial DNA damage, including gross karyotypic abnormalities and alterations in gene and miRNA expression profiles.⁹

Although these findings warrant some caution, insertional activation of genes is not an absolute prerequisite for induction or stability of reprogramming, and it may actually be of minor importance in terms of iPSC genotoxicity. Even so, recently developed self-inactivating polycistronic lentiviral vectors or transposons encoding all four transcription factors have opened up the possibility of selecting and expanding iPSC clones with a single transgene insertion and Cre-mediated excision of all factors, leaving a minimal residual footprint.¹⁰ Furthermore, the IS can be precisely mapped to determine any unexpected changes in DNA structure or gene or miRNA expression in the proximity of the site of vector integration.¹¹ Equally, episomal, synthetic modified RNA-mediated, and protein-mediated delivery of transcription factors have all been shown to generate iPSCs in an integration-free landscape.^12,13,14

For therapeutic purposes, genetic correction of iPSCs before therapeutic application has exciting possibilities, but genotoxicity and transgene phenotoxicity are important considerations. Several clinical gene therapy studies have now demonstrated the potential of integrating vectors to induce mutagenesis in patients. Interestingly, the development of leukemia in X-linked severe combined immunodeficiency patients is dependent on the accumulation of genetic lesions as well as the aberrant proto-oncogene expression induced by the integrated vector. How these additional events occur is still not understood, but there may be interesting parallels with genomic instabilities induced through iPSC culture and expansion ex vivo. Fortunately, whereas gene transfer to a polyclonal hematopoietic stem cell population is relatively poorly controlled in terms of vector localization in the genome, derivation of iPSC clones allows much more precision through targeting of “safe harbors” or correction of specific mutations through homologous recombination. Alternatively, safe locations may be determined retrospectively through bioinformatic analysis of local gene expression following transduction with conventional integrating vectors.¹⁵

Several studies have demonstrated the feasibility of inducing reprogramming in adult somatic cells by excluding the oncogenic transcription factors c-Myc and Klf4 from the reprogramming cocktail, although the efficiency of iPSC generation is significantly reduced (<0.001%). The efficiency of reprogramming can be enhanced by the addition of small-molecule inducers, which include chromatin modifiers (histone deacetylase inhibitors such as valproic acid and DNA methyltransferase inhibitors such as BIX 01294), and modulation of miRNA expression.^16,17,18 Tumor suppressor activity through p53 and p16(INK4a)-retinoblastoma networks is now known to have a critical influence on the efficiency of reprogramming through activation of cell senescence.¹⁹ The successful employment of strategies to override p53 tumor suppressor pathways so as to enhance reprogramming efficiency highlights the fact that genomic instability and the accumulation of genetic lesions may therefore be an inevitable consequence and a significant challenge to overcome.²⁰ For example, a recent study comparing more than 20 different iPSC clones obtained by reprogramming patient keratinocytes revealed a host of point mutations (an average of six mutations affecting protein-coding genes), some of which involved proto-oncogenes.²¹ Copy-number variation (deletions or duplications of entire genes or segments) has recently been explored in human iPSCs generated by either retroviral or transposon-mediated reprogramming.²² Early-passage iPSCs were shown to have higher numbers of copy-number variations compared with human embryonic stem cells (hESCs), but these were selected against through expansion in culture, which is indicative of their detrimental effect on cell growth. Even so, this does not exclude the possibility that dangerous genetic alterations persist. In a high-resolution single-nucleotide polymorphism analysis of large numbers of pluripotent and nonpluripotent cell lines and tissues, the reprogramming process was associated with deletions of tumor-suppressor genes, and continued passage of cells resulted in selection for duplication of oncogenic genes.²³

In another study, mutational load was investigated in one iPSC clone that was genetically corrected by homologous recombination after episomal plasmid-based reprogramming, then after bacterial artificial chromosome-mediated gene targeting, and finally after removal of the antibiotic selection cassette.²⁴ Although no gross G-band metaphase abnormalities were detected, array comparative genomic hybridization and exome sequencing revealed subchromosomal alterations (two deletions, one amplification, and nine protein-coding-region mutations) in both pretargeted and post-targeted iPSCs that were absent from the original patient fibroblasts. There was, however, no increase in mutational load through further culture and clonal selection for gene correction, suggesting that a particularly dangerous time is likely to be during initial reprogramming, when cell senescence pathways are relatively suppressed, rather than during subsequent genetic modification.

Genomic adaptation to cell culture situations has been previously recognized in hESC lines through karyotypic changes, increased cloning efficiency when plated as single cells, and concomitant resistance to apoptosis.^25,26 One hESC line was found to develop a chromosomal homogeneous staining region highly characteristic of cancer cells, further supporting the link between culture adaptation and oncogenic potential. It is therefore quite likely that iPSCs generated from adult somatic cells in a virus-free, integration-free scenario would exhibit similar aberrations over extended periods of in vitro culture. In one study of a large number of hESC and iPSC lines, a substantial number were shown to carry partial chromosomal aberrations, several of which were thought to derive from culture adaptation, and also a high incidence of aneuploidy enriching for cell cycle–related genes.²⁷ Clearly, therefore, genomic changes may vary both dynamically and anatomically throughout the reprogramming and cell culture process, indicating a need for high-resolution genomic monitoring at each stage.

The source of cells used to generate iPSCs may have an important impact on safety. For example, skin keratinocytes, although utilized by several groups for obtaining disease- and patient-specific iPSC lines,^5,28 may have potential disadvantages (aside from those of persistent epigenetic memory). First, they have a considerably higher probability of harboring silent genetic aberrations as a result of exposure to ultraviolet radiation. Second, the establishment of keratinocyte or fibroblast cultures from patient skin biopsy specimens is a relatively lengthy procedure that could allow the accumulation and enrichment of cellular subpopulations harboring mutations that may either hinder subsequent reprogramming or encourage clonal dominance. Alternatively, peripheral blood mononuclear cells obtained by routine venipuncture have been successfully used by several groups as a starting source for iPSC generation because they are amenable to reprogramming within a few days of ex vivo culture and expansion.^29,30,31,32 Other sources of starting cells may be equally amenable and may also be selected on the basis of the desired final therapeutic lineage.

Considerable interest has been focused on the epigenetic signature and transcriptional memory of iPSCs derived from different cell types. For example, iPSCs derived from nonhematopoietic cells may have persistent DNA methylation at loci important for hematopoietic development. Again, this signature may be dynamic in that it has been shown to change over time in culture, but the consequences of aberrant methylation are as yet unknown.³³

Overall, studies conducted using existing methodologies point toward an unavoidable mutation load associated with reprogramming human somatic cells and their subsequent long-term culture and differentiation. Ironically, integrating vectors and other vector-mediated gene correction strategies are likely to be the lesser evils during this process, and also the easiest to screen and circumvent. There are thus several significant challenges to the field before widespread clinical testing can be contemplated. The technologies to derive iPSCs, genetically modify them, and differentiate them into engraftable cells require fundamental development with clinical applicability in mind. Although high-resolution analysis of genomic and epigenomic changes has provided considerable information on the biology and cellular consequences of reprogramming, there remains a big gap in our understanding of the significance of many of these. As investigators in the gene therapy field have found, preclinical toxicology studies may be difficult and poorly predictive of cell behavior in human subjects. Of course, it may be possible to draw up a framework of genomic changes that are allowable based on existing knowledge of gene networks that contribute to events such as oncogenesis. Where risks are judged to be significant, strategies to protect patients from direct toxicities may be useful – for example, employing suicide genes to provide a mechanism for elimination of cells that behave in an unexpected way. Pragmatically, there may be clinical applications that are more ready than others. Learning from the field of gene and cell therapy, enucleated blood products – and even terminally differentiated nondividing cell types – are likely to tolerate relatively high levels of genetic mutation in parental iPSCs without significant oncogenic risk for a patient. Regeneration of tissues through transplantation of lineage-directed stem and progenitor cells may take a little longer to establish.