Targeted gene correction of α1-antitrypsin deficiency in induced pluripotent stem cells.

Human induced pluripotent stem cells (iPSCs) represent a unique opportunity for regenerative medicine because they offer the prospect of generating unlimited quantities of cells for autologous transplantation, with potential application in treatments for a broad range of disorders. However, the use of human iPSCs in the context of genetically inherited human disease will require the correction of disease-causing mutations in a manner that is fully compatible with clinical applications. The methods currently available, such as homologous recombination, lack the necessary efficiency and also leave residual sequences in the targeted genome. Therefore, the development of new approaches to edit the mammalian genome is a prerequisite to delivering the clinical promise of human iPSCs. Here we show that a combination of zinc finger nucleases (ZFNs) and piggyBac technology in human iPSCs can achieve biallelic correction of a point mutation (Glu342Lys) in the α(1)-antitrypsin (A1AT, also known as SERPINA1) gene that is responsible for α(1)-antitrypsin deficiency. Genetic correction of human iPSCs restored the structure and function of A1AT in subsequently derived liver cells in vitro and in vivo. This approach is significantly more efficient than any other gene-targeting technology that is currently available and crucially prevents contamination of the host genome with residual non-human sequences. Our results provide the first proof of principle, to our knowledge, for the potential of combining human iPSCs with genetic correction to generate clinically relevant cells for autologous cell-based therapies.

Currently available methods for gene targeting rely on positive selection to isolate rare clones that have undergone homologous recombination. To remove the unwanted selection cassettes, Cre/loxP or Flp/FRT recombination systems are used, which leave behind single loxP or FRT sites[10,11]. These small ectopic sequences have the potential to interfere with transcriptional regulatory elements of surrounding genes[12], most of which are not fully characterized in the human genome. An alternative method to remove selection cassettes is to convert them into transposons. The most suitable transposon for this purpose is piggyBac, a moth-derived DNA transposon, which can transpose efficiently in mammalian cells including human embryonic stem cells (hESCs)[9,13]. A remarkable feature of this mobile element is seamless excision, which enables removal of transgenes flanked by piggyBac inverted repeats without leaving any residual sequences[9,14].

To explore the use of piggyBac for the correction of point mutations, we designed a vector to correct an albino mutation (G290T substitution in the Tyr gene) in mouse induced pluripotent stem cells (mIPSCs) isolated from fibroblasts of the C57Bl6-Tyr strain[15]. The targeting vector was constructed, carrying a wild-type 290G sequence and a PGK-puroΔtk cassette flanked by piggyBac repeats into the TTAA site (Fig. 1a). Following isolation of targeted clones, the selection cassette was excised from the mIPSCs genome by transient expression of the piggyBac transposase and subsequent FIAU selection. Genomic modification was verified by Southern blot and PCR analyses (Fig. 1b, c). The correction of the G290T mutation and seamless piggyBac excision were confirmed by sequence analyses (Fig. 1d, e). Two introduced silent mutations were observed, confirming that the T290G substitution was mediated by gene correction, not by spontaneous reversion (Fig. 1e). The function of the reverted allele was tested by injecting the corrected mIPSCs into albino mouse blastocysts. The resulting chimeric mice displayed a black coat color, indicating phenotypic correction of the albino mutation (Fig. 1f). These results collectively demonstrate that the piggyBac transposon can be used as a versatile tool for highly precise modification (e.g. correction or mutation) of the mammalian genome at a single base-pair level.

Figure 1

Correction of the G290T mutation in the Tyr gene in mIPSCs

a, The strategy for precise genome modification using the piggyBac transposon. Top line, structure of the Tyr gene; red line, 5′ external probe for Southern blot analysis; open arrow, piggyBac transposon carrying a PGK-puroΔtk cassette; P1, P2 and P3, PCR primers; B, BamHI; E, EcoNI. b, c, Southern blot (b) and PCR analyses (c) showing insertion (c/PB) and excision (c/Rev) of the piggyBac transposon. ES, mouse ESCs as a control. d, e, Sequence analyses revealed correction of the G290T mutation (d) and seamless excision of the piggyBac transposon (e). Note that two silent mutations (A and T indicated by arrowheads) introduced near the TTAA site were also detected. f, A chimeric mouse generated by injecting corrected Tyr mIPSCs (left) displays black coat color. Right, a non-injected albino mouse.

We next explored whether this approach could be used to correct a mutation in hIPSCs derived from individuals with α1-antitrypsin deficiency (A1ATD)[16]. A1ATD is an autosomal recessive disorder found in 1 out of 2000 individuals of North European descent and represents the most common inherited metabolic disease of the liver[17,18]. It results from a single point mutation in the A1AT gene (the Z allele; Glu342Lys) that causes the protein to form ordered polymers within the endoplasmic reticulum of hepatocytes[17,18]. The resulting inclusions cause cirrhosis for which the only current therapy is liver transplantation. The increasing shortage of donors and harmful effects of immunosuppressive treatments impose major limitations on organ transplantation, making the potential of hIPSC-based therapy highly attractive. Since homologous recombination is relatively inefficient in hESCs[6], we employed ZFN technology, which stimulates gene targeting in hESCs as well as hIPSCs[7,10,19]. ZFN pairs were designed to specifically cleave the site of the Z mutation (Fig. 2a-c, Supplementary Table 1 and Supplementary Note). A targeting vector was constructed from isogenic DNA with piggyBac repeats flanking the PGK-puroΔtk cassette (Fig. 2a). To minimize the distance between the mutation and the piggyBac transposon, a CTG leucine codon, 10 bp upstream of the mutation, was altered to a TTA leucine codon, generating the TTAA sequence, which would be left in the genome following piggyBac excision (Fig. 2b).

Figure 2

Correction of the Z mutation in A1ATD-hIPSCs

a, The strategy for precise genome modification using ZFNs and the piggyBac transposon. Top line, structure of the A1AT gene; blue lines, Southern blot probes; thin and thick boxes, non-coding and coding exons, respectively; open arrow, piggyBac transposon; B, BamHI; A, AflIII. b, Sequences of wild-type (Reference), Z, PB, and Rev alleles. Amino acid position 342 (blue), recognition sites for ZFNs (green), piggyBac excision site (red) are shown. Sequence changes in Rev allele from Z allele were indicated by asterisks. c, Surveyor nuclease assay showing the cleavage of Z mutation in ZFNs-transfected K562 cells. Non-transfected cells were used as a control. d, Southern blot analysis showing bi-allelic piggyBac insertion (B-16) and bi-allelic excision (B-16-C2, -C3 and -C6) during correction of the A1ATD-hIPSCs line B. Genomic DNA was digested by BamHI (5′ and PB probes) or AlfIII (3′ probe). Genotype: ZZ, homozygous for Z allele; PP, homozygous for insertion of piggyBac; RR, homozygous for reverted allele. e, Sequence analysis showing correction of Z mutation in 3 corrected hIPSC lines. Wild-type sequence (top line) and A1ATD-hIPSC (second line).

Puromycin-resistant hIPSC colonies obtained after co-electroporation of ZFN expression vectors and the targeting vector were screened for targeted clones by PCR. A1ATD-hIPSC lines derived from 3 different patients yielded targeted clones (Table 1). Remarkably, 54% of the puromycin-resistant colonies were targeted on one allele, while 4% were the result of simultaneous targeting of both alleles (Supplementary Fig. 1).

Table 1

Summary of PCR genotyping of ZFN-stimulated gene targeting

A1ATD-iPSC line	Clonesanalyzed	Het.[a]	Homo. /Hemi. [b]	Het. +additonal integrations [c]	Homo. / Hemi. +additonal integrations [c]	Non-targeted [d]
A	84	45	3	23	8	5
B	18	10	2	3	3	0
C [e]	216	112	9	52	21	22
Mean frequency [%]		54	6	23	12	5

Het., clones heterozygous for PB allele.

Homo./Hemi., clones homozygous or hemizygous for PB allele. Cells with one targeted allele and deletion of the other allele are undistinguishable from correctly targeted homozygous clones by PCR. Such cells are designated as hemizygotes.

Vector backbone integration was analyzed by PCR.

Clones showing incorrect PCR bands are included.

A sum of 2 independent experiments.

To remove the piggyBac-flanked selection cassette from these modified clones, we transiently transfected two homozygously targeted clones (B-16 and C-G4) with a hyperactive form of the piggyBac transposase[8] and subjected them to FIAU selection. The genotype of the resulting FIAU-resistant colonies was analyzed by PCR and confirmed by Southern blot (Fig. 2d and Supplementary Fig. 2a). Bi-allelic excision was observed in 11% of FIAU-resistant colonies (Table 2). Sequence analyses demonstrated that the Z mutation was corrected on both alleles and that transposon excision yielded a TTAA sequence as initially planned (Fig. 2b, e and Supplementary Fig. 2b). The resulting corrected A1ATD-hIPSC (c-hIPSC) lines maintained the expression of pluripotency markers for more than 20 passages and their abilities to differentiate into cells expressing markers of the three germ layers (Supplementary Fig. 3), indicating that genome modification did not alter the pluripotency of c-hIPSCs.

Table 2

Frequencies of bi-allelic piggyBac excision

		Bi-allelic excision w/o re-integration		Bi-allelic excision w/ re-integration
Cell line	analyzed	No. of clones	Frequency [%]	No. of clones	Frequency [%]
B-16	88	15	17	33	38
C-G4	94	5	5	19	20
		Mean frequency [%]	11		29

Genomic instability is known to be associated with prolonged culture of hESCs[20,21] and those arising during genome modification would be another concern for clinical application of hIPSCs. Therefore, we analyzed the genomic integrity of the hIPSC lines using comparative genomic hybridization (CGH) (Supplementary Table 2a-c). Two out of three A1ATD-hIPSC primary lines differed from their parental fibroblasts, showing amplifications or deletions ranging from 20 kb to 1.3 Mb, including a gain of 20q11.21, a frequently amplified region in hESCs[22,23] (see Supplementary Analysis and Supplementary Fig. 4). Line A retained a normal genome content compared to its parental fibroblast. Reassuringly, we found that after ZFN-stimulated targeting, four out of six homozygous clones had unaltered genomes compared to their parental hIPSC lines. Sixteen cell lines with bi-allelic piggyBac excision were compared with their corresponding primary hIPSCs and 12 had unaltered genomes. We also analyzed the hIPSC lines by SNP arrays to check for loss of heterozygosity and found that all lines analyzed retained heterozygosity throughout their genome (Supplementary Fig. 5). This observation demonstrates that bi-allelic gene correction was the result of simultaneous homologous recombination followed by simultaneous excision at both alleles and that mitotic recombination was not involved in this process.

ZFN off-target cleavage and imprecise excision after multiple piggyBac transposition might introduce mutations into the genome. In order to investigate these possibilities at a single basepair resolution, we sequenced exomes of the corrected B-16-C2 line and its parental fibroblast. Comparison of these exomes identified 29 mutations (Supplementary Table 3). The genesis of these mutations was determined by analysis of the primary hIPSC line and the homozygously targeted intermediate. Twenty-four point mutations and one 1-bp deletion were detected in the primary hIPSC line and four mutations arose during genetic correction: one during targeting and three during piggyBac excision. These mutations appeared to arise during culture since their genomic signatures were inconsistent with ZFN off-target sites or piggyBac integration sites (Supplementary Analysis). Taken together, we conclude that the combination of ZFNs with piggyBac provides a new method for rapid and clean correction of a point mutation in hIPSCs without affecting their basic characteristics.

To confirm that the genetic correction of hIPSCs resulted in the expected phenotypic correction, hIPSCs were differentiated in vitro into hepatocyte-like cells, the main cell type affected by the disease A1ATD. Differentiation of the corrected lines occurred as expected, resulting in a near homogenous population of hepatocyte-like cells (Supplementary Fig. 6a-c). Remarkably, CGH analysis of differentiated cells showed that hepatic differentiation neither increases the number of genetic abnormalities nor selects for cells with abnormal karyotype (Supplementary Table 2d). The resulting cells shared key functional attributes of their in vivo counterparts including glycogen storage, LDL-cholesterol uptake, albumin secretion and Cytochrome P450 activity (Supplementary Fig. 6d-g). Importantly, immunofluorescence and ELISA both confirmed the absence of mutant polymeric A1AT in c-hIPSCs-derived hepatocyte-like cells that instead efficiently secreted normal endoglycosidase–H-insensitive monomeric A1AT (Fig. 3a-d). In addition, secreted A1AT displayed an enzymatic inhibitory activity that was comparable to that obtained from normal adult hepatocytes (Fig. 3e), thereby suggesting that physiological restoration of enzyme inhibitory activity could be achieved.

Figure 3

Functional analysis of restored A1AT in c-hIPSCs-derived hepatocyte-like cells

a, Immunofluorescence showing the absence of polymeric A1AT protein in hepatocyte-like cells generated from c-hIPSCs. All forms of A1AT (left panels) and misfolded polymeric A1AT (middle panels). b, c, ELISA to assess the intracellular (b) and secreted (c) levels of polymeric A1AT protein in hepatocyte-like cells derived from A1ATD-hIPSCs (ZZ), c-hIPSCs (RR) and control hIPSCs (++). d, Endoglycosidase H (E) and peptide:N-glycosidase (P) digestion of A1AT immunoprecipitated from uncorrected (ZZ), corrected (RR) and control (++) hIPSC-derived hepatocyte-like cells (upper panels) and corresponding culture medium (lower panels). e, Chymotrypsin ELISA showing that corrected cells (RR) have A1AT enzymatic inhibitory activity that is superior to uncorrected cells (ZZ) and close to adult hepatocytes. f, g, Immunofluorescence of transplanted liver sections detecting human albumin (f) and A1AT (g). DNA was counterstained with DAPI. h, ELISA read-out of human albumin in the mouse serum longitudinally followed for each mouse. Asterisk, the mouse was subjected to histology analysis. Scale bars, 100 μm. Data in b, c and e are shown as mean ± s.d. (n=3). Student’s t-test was performed. NS, not significant.

Finally, the in vivo function of c-hIPSCs-derived hepatocyte-like cells (B-C16-2 line) was assessed following transplantation into the liver of Alb-uPA mice via intra-splenic injection. Livers harvested 14 days after injection were colonized by human cells identified using antibodies specific to human albumin and A1AT (Fig. 3f, g). These human hepatocyte-like cells were distributed throughout the liver lobes and were seen to be integrated into the existing mouse parenchyma (Fig. 3f, g). In addition, human albumin was detected in the serum of transplanted animals for at least 5 weeks (Fig. 3h), while no tumor formation was detected in any mice. Therefore, c-hIPSCs-derived hepatocyte-like cells were able to colonize the liver in vivo and display functional activities characteristic of their human ESC-derived counterparts[24]. Collectively these analyses demonstrate that genetic correction of the Z mutation resulted in functional restoration of A1AT in patient-derived cells.

All experimental evidence above strongly support the applicability of genetic correction in patient-specific iPSCs for cell-based therapy of A1ATD. We therefore repeated the genetic correction in more clinically relevant cells using patient-specific iPSCs reprogrammed from fibroblasts with Sendaiviral vectors, an integration-free method[25] (Supplementary Fig. 7a-f). One primary hIPSC line with an intact genome by CGH analysis (Supplementary Fig. 7e and Supplementary Table 4) was corrected by the method described above. The final product, iPSC-3-G5-A7, had the corrected A1AT, had an intact genome compared to the parental fibroblast, and expressed normal A1AT protein when differentiated to hepatocyte-like cells (Supplementary Fig. 8 and Supplementary Table 4). This is the first demonstration of the generation of mutation-corrected patient-specific iPSCs, which could realize the therapeutic promise of hIPSCs.

In the present study, we demonstrate that ZFNs and piggyBac transposon enable simultaneous bi-allelic correction of diseased hIPSCs. No residual ectopic sequences remain at the site of correction and the genome appears to be undisturbed elsewhere. Although we could readily obtain cell lines without large genomic alterations during genetic modification, the resulting corrected hIPSCs carry 29 mutations in protein coding exons, of which 22 were non-synonymous or splice site mutations. The likely impact of this mutation load needs to be considered in the context of their likely functional impact, taking into account the normal germ-line load, accumulated somatic variation, the presence of compensating normal gene copies and the requirement for the gene product in the derived differentiated cells. From this point of view, only eight mutations might affect gene functions in hepatocyte-like cells (Supplementary Table 3). Nevertheless, the corrected iPSCs could efficiently differentiate to hepatocyte-like cells and engraft into the animal model for liver injury without tumor formation. Therefore, limited genomic abnormalities might have restricted biological consequences. Careful screening of primary and corrected hiPSCs using deep sequencing analyses would contribute to the safe use of hIPSCs in clinical applications.

hIPSCs derived from different patients were effectively corrected, demonstrating that this method could be applied to a large number of A1ATD-hIPSC lines. Since the bi-allelic correction could be carried out in less than 4 months, our approach may be compatible with large-scale production of corrected patient-specific hIPSCs not only for A1ATD but also for other monogenic disorders.

Methods summary

A1ATD-hIPSCs were described previously[16]. 2 × 106 hIPSCs were co-transfected with ZFN expression vectors and the donor template, and subjected to puromycin selection (1 μg ml−1) initiated 4 days after transfection. For transposon excision, targeted cells were transfected with pCMV-hyPBase[8], cultured for 4 days, replated and selected in 250 nM 1-(2-Deoxy-2-fluoro-beta-D-arabinofuranosyl)-5-indouracil (FIAU). To increase clonogenicity, cells were treated with ROCK inhibitor[26], Y-27632 (10 μM) 4 hours prior to dissociation and 24 hours post plating. Resulting colonies were picked 2 weeks later, analyzed by PCR and further verified by Southern blot analysis. Primer sequences are listed in Supplementary Table 5.