The discovery of induced pluripotent stem (iPS) cells has broadened the promises of regenerative medicine through the generation of syngeneic replacement cells or tissues via the differentiation of patient-specific iPS cells. To apply iPS cell-mediated therapy to patients with genetic disorders, however, genome-editing technologies with high efficiency and specificity are needed. Recently, several targeted genome-editing strategies mediated by zinc finger nuclease and transcription activator-like effector nuclease have been applied to human and mouse iPS cells. Furthermore, spontaneous homologous recombination can restore genotype to wild type in mouse iPS cells heterozygous for genetic mutations. Through genome editing, the clinical application of patient-specific genetic mutation-free iPS cells to genetic disorders can finally be realized.
Stem cells can be classified as somatic or pluripotent depending on their developmental potential. Somatic stem cells, which have the ability to self-renew and generate limited kinds of somatic cells, function in maintaining homeostasis of the adult body, whereas pluripotent stem cells have a robust ability to self-renew and can differentiate into all cell types in the adult body (Jaenisch & Young 2008; Evans 2011). Several kinds of pluripotent stem cell lines have been established from different cell sources of developing mouse embryos. The mouse embryonic stem (ES) cell is one kind of pluripotent cell, which is isolated from the inner cell mass cells of blastocysts (Evans & Kaufman 1981). ES cells that maintain a normal karyotype are capable of contributing to the normal development of chimeric mice when microinjected into the blastocoel cavity of blastocysts. Furthermore, a robust potential for cell proliferation in vitro has resulted in the isolation of individual cell colonies, in which rare genetic modifications are introduced into the genome by homologous and nonhomologous gene integration (Capecchi 1989). The generation of genetically mutated mice has greatly contributed to our understanding of gene function in vivo.
The successful generation of human ES cells from blastocysts (Thomson et al. 1998) has significantly advanced the field of regenerative medicine, which involves replacing damaged tissue with stem cell–derived tissue, and paved the way for new therapeutic strategies. However, the generation of ES cells from human embryos has raised ethical concerns (Lo & Parham 2009). To circumvent these, other strategies have been used to obtain pluripotent stem cells. Nuclear reprogramming of somatic cells into pluripotential stem cells has been achieved by the transplantation of somatic nuclei into unfertilized eggs (Wilmut et al. 1997) and cell fusion between mouse ES cells and somatic cells (Tada et al. 2001). Direct reprogramming of somatic cells into induced pluripotent stem (iPS) cells retaining an identity similar to ES cells was achieved by the transient over-expression of defined transcription factors in mice (Takahashi & Yamanaka 2006) and humans (Takahashi et al. 2007; Yu et al. 2007). Besides ethical issues, human iPS cells have key advantages over human ES cells in generating immuno-rejection-free iPS cell derivatives, which are syngeneic to patients. The emergence of iPS cells should facilitate regenerative medicine with patient-specific pluripotent stem cells. iPS cells generated from patients carrying genetic disorders are not, however, applicable to cell therapy, as the iPS cell technology reprograms epigenetic, but not genetic, information in somatic nuclei. To make regenerative medicine applicable to a broader range of disorders, the development of safe and efficient genome-editing technologies is essential in iPS cell-based therapeutic approaches. Here, we introduce recently developed technologies for editing the genome of mouse and human iPS cells carrying pathogenic genetic mutations.
Genome editing through two-step events
The success of genome editing relies on capturing double strand breaks (DSBs) in the genome of selected cells, in which homologous recombination (HR) takes place (Urnov et al. 2010). Therefore, two steps – the generation of DSBs and HR – are required for genome editing, indicating that the efficiency is determined by multiplication of the frequency of DSBs and HR. Spontaneous DSBs often occur, whereas artificial DSBs are introduced by irradiation or site-specific endonucleases (Kasparek & Humphrey 2011). There are several competing repair pathways in response to DSBs (Fig. 1), including nonhomologous end joining (NHEJ), single-strand annealing (SSA) and HR (San Filippo et al. 2008; Hartlerode & Scully 2009). Of them, HR is the only faithful repair pathway, as it uses a homologous DNA segment as a template to guide the synthesis of a new strand (Symington & Gautier 2011). NHEJ is the dominant pathway in higher eukaryotes, whereas HR makes only a minor contribution confined to the G2 phase of the cell cycle, even when a homologous template is available (Heyer et al. 2010).
Genome editing by ZFN-mediated artificial DSBs
To increase the frequency of genome editing, a way of generating site-specific DSBs in the genome has been developed. Zinc finger nuclease (ZFN) was first reported as an enzyme for creating DSBs at specific sites in the genome (Kim et al. 1996). ZFN consists of two units (Fig. 2), with each unit having two domains: a DNA-binding domain that is derived from the C2H2 motif of zinc finger protein (ZFP), and a cleavage domain from the restriction endonuclease, FokI. A short linker joins the two domains. The nuclease domains in the two units are identical, and able to produce a DSB anywhere in the genome when dimerized. The binding is specified by the ZFP domain. Each ZFP can specifically bind 3 nucleotides, and therefore an array of 3–4 ZFPs in each unit will impart enough specificity to a ZFN to pick out a single location in the genome (Urnov et al. 2010; Osborn et al. 2011). Another essential part of the ZFN system is the construction of a donor vector, which is similar to that used in conventional gene targeting to provide a template for synthesizing new DNA strands at DSB sites. The donor vectors should be delivered together with ZFNs. To maximize the efficiency of targeting, the ratio of donor vector to ZFN should range between 5 : 1 and 10 : 1.
Genome editing using the ZFN system in iPS cells has been explored in several studies (Hockemeyer et al. 2009; Zou et al. 2009, 2011a,b; Sebastiano et al. 2011; Yao et al. 2011). In the case of anα1-antitrypsin deficiency, a single point mutation at Glu342Lys of the A1AT (SERPINA1) gene was corrected by a combination of ZFN and transposon-based vector, piggyBac technology in human iPS cells (Yusa et al. 2011). The genetic correction restored the structure and function of A1AT in subsequently derived hepatocyte-like cells in vitro and in vivo. Use of ZFN to achieve targeted genome editing at transcription-active and inactive genes in iPS cells has been reported in several cases (Table 1). In theory, ZFN-mediated site-specific DSBs increase the efficiency of gene targeting at a specific locus. However, Southern blot hybridization and DNA sequencing have demonstrated the frequency of gene targeting to range from 1.1 × 10−5 to 1.0 × 10−7 per infected cell, indicating that the frequency of ZFN-mediated gene targeting is not higher than that of spontaneous DBS-mediated gene targeting (Liu et al. 2011; Sebastiano et al. 2011; Soldner et al. 2011; Zou et al. 2011a,b). This implies that increasing site-specific DSBs can improve the efficiency, but not frequency of HR in iPS cells. A novel system or molecule that can raise the frequency of HR will substantially improve genome-editing technology.
Table 1. Summary of genome editing in iPS cells
†Calculated as clones confirmed by Southern hybridization or sequencing (if available) divided by the total colonies picked-up after positive and/or negative selection.
Off-target cleavage activity is a major concern that may limit use of this technology. The off-target effect is defined as a cleavage event at unintended sites within the genome (Pattanayak et al. 2011). This could be minimized by redesigning the ZFP recognition sequences, or by modification of the nuclease domain to prevent homo-dimer formation (Miller et al. 2007). To monitor unintended cleavage(s), whole genome sequence comparison of the targeted and parental cell lines is required to assess the safety of this technology.
Genome editing by TALEN-mediated artificial DSBs
Transcription activator-like effector nuclease (TALEN) is an alternative for creating DSBs in the genome (Miller et al. 2011). Similar to ZFN, TALEN has a cleavage domain containing FoK1 endonuclease, which can produce DSBs in the genome. However, the TALEN-specified location is guided by the transcription activator-like effector (TALE), originally discovered in a plant pathogen to subvert the genome regulatory network. The DNA-binding domain of TALE is composed of ∼34 amino acids (TALE repeats) arranged in tandem, and among them there are two critical and adjacent variable amino acids called the ‘repeat variable diresidue’ (RVD), which determines the base-recognition specificity. The relationship between the preferred binding site of a TALE and its successive RVDs appears to constitute a simple code, with each repeat independently specifying its targeted base. Consequently, TALEN can be engineered to induce DSBs at a predetermined position in the genome with great precision (Christian et al. 2010).
TALEN-mediated genome editing has been successfully accomplished in human ES cells and iPS cells (Hockemeyer et al. 2011). The TALEN system was used to target three genes (PPP1R12C (AAVS1), OCT4 (POU5F1) and PITX3) in human pluripotent cell lines including one iPS cell line. Southern blot hybridization analyses demonstrated that the targeting frequency with TALEN was similar to that with ZFNs (Table 1). This result furthermore confirms that the rate-limiting step is HR. A TALEN-mediated system is yet to be fully used to engineer DNA information through genome editing. In the near future, a comparison between the ZFN and TALEN approaches could show advantages and disadvantages in terms of targeting efficiency and off-target effects for clinical applications of human iPS cells to correct mutated genes.
Genome editing by spontaneous DSBs and vector-mediated gene delivery
The poor efficiency of gene targeting by spontaneous DSB-mediated conventional HR with an exogenously provided DNA template has impeded the use of human ESCs and iPSCs in disease models. In fact, genome editing with a plasmid, virus and bacterial artificial chromosome (BAC) has succeeded at only a relatively low frequency. The targeting efficiency varies widely in human and mouse iPS cells, because of the gene delivery system and size of the DNA delivered.
Adeno-associated virus (AAV) is a nonpathogenic, nonenveloped virus containing a 4.7-kb single-stranded DNA genome, flanked by short inverted terminal repeats. AAV vectors lacking an integrase are safe and less cytotoxic, although their packaging capacity is limited to 9 kb. Exogenous DNA inserted into the AAV genome can be efficiently delivered into nuclei of dividing and nondividing cells (Asuri et al. 2011). The loci of HPRT1 and NANOG were targeted at a frequency of 6.2 × 10−6 ∼ 5.3 × 10−7 and 1.6 × 10−6 per infected cell, respectively (Mitsui et al. 2009). Furthermore, the loci of HPRT1 and HMGA1 were targeted at 6.9 × 10−5 per infected cell (Khan et al. 2010). Interestingly, the targeting efficiency is highly varied among loci and researchers.
Helper-dependent adenoviral vectors (HDAdVs) are a platform for genome editing in human iPS cells. HDAdVs, in which viral genes are completely removed, have less cytotoxicity and a packaging capacity of up to 37 kb. At the locus of Lamin A (LMNA), more than two hundred different mutations were genetically corrected by gene targeting at a frequency of around 1.5 × 10−5 per infected cell in human iPS cells (Liu et al. 2011). Gene knockout and knock-in were carried out by HDAdV transduction at the loci of HPRT1, KU80, LIG1, LIG3 and LB9 in human ES cells and iPS cells at a targeting frequency of 2.0 × 10−7 to 5.6 × 10−5 per infected cell (Aizawa et al. 2012). Notably, the genome was edited not only at transcriptionally active genes but also at inactive genes without artificial DSBs.
BACs have the highest packing capacity of any vector, between 30 and 300 kb. A point mutation in exon 7 of the ornithine-aminotransferase (OAT) gene was genetically corrected by electroporation-mediated delivery of a BAC vector into human iPS cells established from a patient with gyrate atrophy (Howden et al. 2011). The targeting frequency can vary widely depending on the size of packed DNA in the BAC vector. In the case of OAT, the targeting frequency was estimated at 4.0 × 10−7 per infected cell. Therefore, the BAC system improves the packing capacity of insert DNA, but not the targeting frequency in human iPS cells.
Genome editing by spontaneous DSBs and mitotic recombination
Mitotic recombination, which occurs at a low frequency in somatic cells, has been shown to function in DNA repair through cell divisions (Moynahan & Jasin 2010). However, experiments to drive targeted chromosome elimination from pluripotent ES-somatic hybrid cells (Matsumura et al. 2007) suggest that the frequency of genetic repair events through spontaneous mitotic recombination to be higher in pluripotent stem cells than in somatic cells (Kipps & Herzenberg 1986). This raises the possibility that genome editing could be applied to human and mouse iPS cells to correct gene information in heterozygous disorders through mitotic recombination. Furthermore, chromosome-specific loss of heterozygosity was created in a knockout allele by a high-dose of G418 selection to the Neo gene in mouse ES cells (Lefebvre et al. 2001). Circumstance evidence provided an idea that genome editing could be possible in human and mouse iPS cells generated from patients carrying genetic disorders. To proof the hypothesis, an animal model of autosomal dominant polycystic kidney disease (ADPKD) (Muto et al. 2002), which is the most common of all the hereditary cystic diseases with an incidence of 1–1000 live births, and caused by genetic mutations in the Polycystic kidney disease 1 (Pkd1) (85% of ADPKD) and Pkd2 (15% of ADPKD) genes, was examined. Clonally expanded subclones of mouse iPS cells generated from embryos heterozygous for Pkd1 deficiency (Pkd1(+/−)) were PCR-genotyped without any drug-selection. Notably, genome-edited iPS clones carrying the Pkd1(+/+) genotype were isolated at a frequency of 1.9 × 10−4 per cell clone through spontaneous mitotic recombination as shown by Southern blot hybridization and DNA fluorescence in situ hybridization analyses (Fig. 3) (Cheng et al. 2012). In contrast to the frequent generation of polycysts detected in chimeric adults with Pkd1(+/−) iPS cells, those with genome-edited Pkd1(+/+) iPS cells showed no cysts, similar to the wild-type iPS cells. Artificial genetic manipulation-free genome editing mediated by mitotic recombination is safer than ZFN and TALEN-mediated genome editing for patients carrying genetic diseases. The ability of generating iPS cells from various tissues of adult body realizes the usage of iPS cells generated from non-nidus tissues without the secondary mutation for developing disease phenotype. Mitotic recombination–mediated genome editing is applicable in iPS cells carrying the genetic heterozygosity for causing diseases.
Several technologies have been developed for genome editing using disease-specific iPS cell lines. The incorporation of genome editing in patient-derived iPS cells is a promising step toward personalized cell therapy. For safer clinical applications, genetic modification generated by integration of the exogenous genes into the host genome can be minimized by excision using the Cre-LoxP and transposon (sleeping beauty, piggyBac) systems (Kos 2004; Zhou et al. 2010; Ivics & Izsvák 2011; Tsukiyama et al. 2011). The Cre-LoxP and sleeping beauty systems usually leave a short residual sequence in the genome, whereas piggyBac does not leave any footprints. Thus, the piggyBac system may be safer and more suitable for generating therapeutic iPS cells. In addition, the targeted generation of DSBs created by the ZFN and TALEN systems is not sufficient to increase the frequency of genome editing, and the rate-limiting step lies in the low frequency of HR. To achieve practical genome editing, a novel strategy to increase the frequency of HR and innovational developments to increase HR efficiency are required. Genome editing in personalized iPS cells raises the possibility of curing genetic diseases. Although challenges remain before the full realization of iPS-based therapy, the preclinical application of genome editing for genetic disorders is close at hand.
We thank our colleagues for discussions. LTC and LTS are postdoctoral fellows supported by the Takeda Science Foundation JAPAN, and the Whitaker Foundation USA, respectively.