Site-specific gene correction of a point mutation in human iPS cells derived from an adult patient with sickle cell disease

Human induced pluripotent stem cells (iPSCs) that are derived from adult somatic cells hold great promise as a renewable cell source for developing novel cell and gene therapies.¹  Since the first report of generating iPSCs from patients' somatic cells in 2008, a wave of studies have demonstrated either undifferentiated patient-specific iPSCs or their differentiated progenies are capable of recapitulating key aspects of disease-specific phenotypes in vitro. So far, most of the successful reports on disease modeling are based on iPSCs bearing monogenic disease mutations. Capable of expanding unlimitedly in culture, iPSCs derived from patients that suffer from a genetic mutation are also promising candidates for cell replacement therapy if the disease-causing mutations can be permanently corrected. Most published studies that corrected mutations in iPSCs relied on adding a copy of a functional transgene randomly into the genome by viruses to restore the wild-type phenotype. For example, disease-corrected Fanconi Anemia iPSCs were established using this approach.²  However, random integrations of a functional copy of the affected gene, especially mediated by viral vectors, could induce oncogenic mutations and thus jeopardize the goal of gene therapy. The precise correction or replacement of mutations at the endogenous loci by homologous recombination (HR) is highly desirable even though the efficiency is extremely low (10⁻⁶) in noncancerous human cells such as human iPSCs. Recently, several studies reported HR-mediated gene addition at a few selective loci in normal or disease-specific human iPSCs.^3-7  However, this gene addition approach is only suitable for phenotypic correction of loss-of-function mutations, and not for correcting dominant or gain-of-function mutations. The ideal solution is targeted and precise gene correction of the underlying disease-causing mutation, which also ensures the corrected gene will be expressed in the appropriate temporal and tissue-specific manner under the regulation of endogenous cis-elements.

Toward this goal, we focused on sickle cell disease (SCD) as a model system to develop methods for precise gene correction of a disease-causing mutation. In the majority of SCD patients, the mutation of A > T (also known as β^A to β^s mutation) in both alleles of the β-globin (HBB) gene that changes codon 6 from Glu (GAG) to Val (GTG) results in a defective form of adult hemoglobin in oxygen-carrying red blood cells. Although SCD was one of the first described molecular diseases, the goal for treating this monogenic disorder using gene therapy approaches has remained elusive.^8,9  Gene correction of β^s in mouse embryonic stem cells (ESCs) by HR has been reported previously.^10,11  A similar strategy was used to correct the β^s mutation in mouse iPSCs derived from a humanized SCD mouse model, followed by successful transplantation of differentiated hematopoietic cells into isogenic mice to cure SCD phenotypes.¹²  However, gene correction of the SCD mutation in human patient-specific iPSCs has not been demonstrated yet. Here, we show that with appropriate gene-targeting vector design and zinc finger nuclease (ZFN) enhancement of HR efficiency, site-specific gene correction of the silent HBB gene in human iPSCs can be achieved. We also demonstrate the expression of the corrected wild-type β^A allele in red blood cells differentiated from thus corrected iPSCs. Some unexpected discoveries and discussions on promise and pitfalls of this site-specific gene correction approach in human cells also are presented.

Human iPSCs were maintained on primary mouse embryonic fibroblasts (PMEFs) as feeder cells using the standard human ESC media. The SCD iPSCs used in this study were generated in a previous study.¹³  The MBP5s1 (S1) iPSC line was derived using piggyBac transposition of reprogramming factors into bone marrow stromal cells (MSCs) from an adult patient with SCD at Johns Hopkins Hospital. To adapt to single-cell passaging, the S1 iPSCs were first passaged from PMEFs feeders to feeder-free culture condition using Matrigel (BD Biosciences) and Stemedia Nutristem XF/FF medium (Stemgent) at 1:1 to 1:2 ratio using 0.05% trypsin (Invitrogen) digestion, and then they were maintained by this trypsin-passaging and feeder-free culture condition. All patient samples were used per approval from Johns Hopkins University internal review board for conducting laboratory research using anonymous human cells. All the iPSC lines described in this report will be provided on request to L.C., under a standard (uniformed) material transfer agreement with the Johns Hopkins University. Human 293T cells that were used to validate ZFNs, and gene targeting vectors were grown in DMEM high glucose supplemented with 10% FBS.

Characterization and chromosome karyotyping (G-banding) of iPSC clones were performed at multiple passages as described previously.^3,13-15  Color and fluorescent images of iPSC markers (in PBS) and teratomas (embedded sections) were acquired using a Nikon Eclipse TE2000-U inverted microscope with 10×ELWD Plan Fluor/0.3 or 20×ELWD Plan Fluor/0.45 objective and a Qimaging Micropublisher 5.0 digital camera with QCapture software (Version 3.1.2). Multichannel fluorescent images were merged using ImageJ software.

The left and right homology arms were amplified from genomic DNA of SCD fibroblasts (GM02340 in Coriell collection) used in a previous study.¹⁵  The primer sets HBBL-F, 5′-ATCGGTACCGTGTGTAAGAAGGTTCCTGAGGCT and HBBL-R, 5′-ATCGCTAGCGGTCTCCTTAAACCTGTCTTGTAACC amplified the 5.9-kb left arm, and HBBR-F, 5′-ACTGCTAGCAATAGAAACTGGGCATGTGGAGACAG and HBBR-R, 5′-GATCTCGAGAAGAAGGGCTCACAGGACAGTCAA amplified the 2-kb right arm. The GTG mutation in the genomic DNA was converted to the wild-type GAG codon using PCR-mediated mutagenesis. A loxP-flanked PGK-hygromycin (Hyg) cassette from a previous targeting donor PHD³  was cloned into NheI site between 2 homology arms in the pCR2.1-TOPO vector. The loxP-flanked PGK-Hyg cassette is to be inserted at nt 179 of the HBB gene, 37 bp downstream of the exon 1 junction (nt 142). The EF1α-TK.GFP gene cassette (for counter selection) obtained from a previously published vector¹⁶  was inserted downstream of the right arm. The BD2 vector plasmid and sequence information is available on request.

The CompoZr custom-designed HBB-ZFNs were obtained from the Sigma-Aldrich. It is now available to the public (CKOZFN1264-1KT). The ZFN pair was designed to target the following DNA sequence in exon 1 of the HBB gene (see Figure 1A underlined). Two different HBB-ZFN subunits that recognize right (12-bp) and left (19-bp) half sequences can only form a dimmer with each other, which is required for endonuclease activity. They are expressed from 2 plasmids under the control of the CMV promoter. For green fluorescent protein (GFP)–based HR reporter assay in 293T cells, we cloned a 63-bp DNA sequence (from codon 1, nt 54 to nt 116) of the HBB gene containing the putative ZFN recognition sites into the middle of the EGIP* vector following a previously published strategy.³  The insertion, together with a stop codon (taa) upstream and a HindIII site downstream aagctt), disrupts the enhanced green fluorescent protein (EGFP) reading frame. A full-length GFP activity could be restored, if the mutated GFP (GFP*) DNA is corrected by HR (see Figure 1B). A plasmid bearing a nonexpressing, truncated GFP (tGFP) that provides a repair template, together with a pair of plasmids expressing 2 ZFNs, was introduced into the stably transfected 293T cells expressing the EGIP*-HBB sequence. The successful (GFP*) gene targeting was measured by the frequency of GFP+ cells 2 days after transient transfection.³  To test the specificity of HBB-ZFNs, the homologous region from other β-locus genes such as HBE, HBG, and HBD also was inserted into the EGIP vector as tested in parallel. A shorter form of the HBB sequence (43 bp) containing the core ZFN recognition site also was tested by the same assay. For testing HBB-ZFNs and the BD2 vector to achieve the HBB endogenous gene targeting, 0.5 million 293T cells were transfected by lipofection with 1 μg of BD2 donor and 0.25 to 2 μg/each HBB-ZFN in one 6-well plate. Five days after transfection, 1 million transfected cells were plated on 150-mm tissue culture dish with 200 μg/mL hygromycin B selection, and then 20μM gancyclovir (GCV) was added 8 days afterward. The total drug selection lasted 15 days, including 7 days for GCV and 15 days for hygromycin B selection. Pooled cell populations were used for genomic DNA extraction and PCR detection assays.

For correcting β^s mutation in SCD iPSCs, 5 × 10⁶ iPSCs were trypsinized and resuspended in mESC nucleofection buffer (Amaxa) and subsequently electroporated using the A-023 setting as described previously.³  The treated PSCs were plated onto irradiated hygromycin-resistant PMEF feeders (PMEF-HL; Millipore), with Y-27632 (a Rho-associated kinase inhibitor) added at 10μM final concentration into the media for overnight duration to improve survival of the nucleofected cells. From day 3 to day 17, hygromycin B selection at 20 μg/mL was performed. From day 10 to day 17, GCV (2μM) also was supplemented to the media. Surviving colonies were picked on day 17 and expanded on PMEF feeders following standard culture protocols and then screened for targeting events.

Primer set 5′-CAAATGCGAGAGAACGGCCTTAC (in the Hyg cassette) and 5′-CTAGCACTGCAGATTCCGGGTCAC (on HBB locus downstream of 3′-homology arm) was used to amplify a 2.5-kb product of the 3′-junction of a targeted integration (TI; see Figure 2A). Primer set 5′-ACATTTGCTTCTGACACAAC (in HBB exon 1) and 5′-AGCAAGAAAGCGAGCTTAG (in HBB exon 3) were used for RT-PCR detection of mutant or corrected HBB mRNA expression (see Figures 2A and 6B). Primer set 5′-ATTCCGGAAGTGCTTGACATTG and 5′-CACGCCATGTAGTGTATTGACC (both in Hyg coding sequence) were used for RT-PCR detection of Hyg mRNA expression (see Figure 7A).

Primer set 5′-CGATCACGTTGGGAAGCTATAGAG and 5′-AACATCCTGAGGAAGAATGGGAC was used to amplify the 3.5-kb genomic region surrounding HBB gene for both targeted and nontargeted alleles of S1, cre4, and cre16. For c36 (targeted) iPSCs, longer extension time was used to amplify both 3.5-kb nontargeted and 5.8-kb targeted alleles. Mixed alleles from these iPSC clones were cloned into a TOPO vector. Then, DNA in individual bacterial clones was sequenced to identify specific alleles (targeted or nontargeted).

Standard Southern blot protocol using digoxigenin-labeled probes was followed as described previously.³  The 3′-probe was amplified from S1 iPSC genomic DNA using primers 5′-GACTGAGAAGAATTTGAAAGGCG and 5′-TCATCAATTCTGCCATAAATGG. The Hyg probe used was the same as described previously.³ 

The Omni1_Quad BeadArray chip (Illumina) containing more than 10⁶ probes detecting informative single-nucleotide polymorphisms (SNPs) was used. The array analysis was performed by the Johns Hopkins SNP Center as part of Center for Inherited Disease Research (CIDR; www.cidr.org). Based on 1 140 419 SNPs identified, and the array results with 2 control genomic DNA samples (CIDR11993 and CIDR10860) that have been previously sequenced and included in each run, the Johns Hopkins SNP Center reported a 0.27% genotyping error rate in the run, a rate that was within the normal range. Genomic DNA isolated by the DNAeasy kit (QIAGEN) from c36 (after ZFN-mediated HR), S1 iPSCs, and their parental somatic cells (MSCs) were analyzed. Based on the ratio of allele intensities, the disconcordance rate between c36 to S1 iPSCs (2 samples of different passages before and after trypsin-adaptation, p17 and p26) or to MSCs (passages 0 and 6) is below 0.04%, a rate that is much lower than the basal error rate level (0.27%). In comparison, the disconcordance rates between c36/S1 iPSCs to other unrelated iPSCs and between 2 unrelated CIDR11993 or CIDR10860 controls are ∼ 40%. Therefore, our data suggest that the c36 iPSC genome is essentially identical to the S1 iPSCs or their somatic MSCs.

Human iPSCs were directly differentiated into hematopoietic progenitor cells by a modified embryoid body (EB) formation method as described previously.¹⁴  Two weeks after the EB-mediated hematopoietic differentiation, the EB-derived cells were collected and dissociated with accutase treatment. The dissociated cells were cultured in a serum-free medium containing SCF (100 ng/mL), IL-3 (10 ng/mL), insulin-like growth factor II (40 ng/mL), erythropoietin (2 U/mL), and dexamethasone (1μM) for erythroid cell expansion and maturation as described previously.¹⁴  Cells were harvested on day 8 of erythroid expansion and differentiation, and the differentiated cells were analyzed by fluorescence-activated cell sorting analysis, cytospin and histology staining, and RT-PCR for globin expression. Giemsa stain of cytospin was captured using the same camera and software used for iPSCs except 40×ELWD Plan Fluor/0.6 objective. Quantitative (q)RT-PCR primers and probes are from Applied Biosystems (identification numbers are Hs00361131_g1 for HBG1/2 and Hs00747223_g1 for HBB).

Adult hemoglobin (HbA) antibody is from Santa Cruz Biotechnology (sc-21757PE). Fetal hemoglobin (HbF) antibody is from Invitrogen (catalog HFH-01).

Based on our previous studies of gene targeting in human iPSCs and ESCs,³  we attempted to correct the mutated β^s allele in SCD iPSCs. To achieve specific HR-mediated gene replacement at the HBB gene, but not at nearby HBD (δ), HBG1 (Aγ), HBG2 (Gγ), and HBE (ϵ) globin genes (Figure 1A), we decided to use ZFNs that make a specific double-strand break to stimulate HR near the β^s mutation (codon 6) in the HBB gene. After initial screening, a pair of HBB-ZFNs was identified that targets a bipartite 31-bp sequence (with a 6-bp spacer) in exon 1 (21 bp downstream of the β^s mutation) of the HBB gene (Figure 1A). We first tested the activity and specificity of this newly designed HBB-ZFN pair using a GFP reporter system³ : DNA sequences from HBB or highly homologous β-locus genes HBE, HBD, and HBG were tested as candidate alternative targets of this ZFN pair (Figure 1A). Successful HR between the EGIP* reporter (containing a ZFN target sequence) and a tGFP donor will restore a full-length GFP gene (Figure 1B). Using flow cytometry to measure HR-mediated gene correction in stably transfected 293T cells expressing EGIP*-HBB target, we monitored the number of GFP+ cells after tGFP donor transfection, in the absence or presence of the HBB-ZFN stimulation. Without cotransfection of HBB-ZFN expression vectors, the basal gene correction rate was ∼ 4 to 5 per million cell events as observed previously ³  (Figure 1C). In the presence of the HBB-ZFNs, numbers of GFP+ cells jumped to 0.16% from cells expressing the EGIP* reporter that contains the HBB target sequence: ∼ 350-fold stimulation (Figure 1C). In contrast, the HBB-ZFNs showed no detectable stimulatory activity on EGIP* containing the homologous target DNA from the HBD, HBE, and HBG genes, indicating the HBB-ZFNs are highly specific (Figure 1D).

To achieve targeted correction of β^s mutation at the HBB locus, we designed a plasmid donor (BD2) to provide wild-type GAG for codon 6 (β^A) and 2 constitutively expressed drug-selectable genes: PGK promoter-driven hygromycin-resistant gene (PGK-Hyg) for positive selection and EF1α-promoter–driven thymidine kinase-GFP fusion gene for counterselection. The PGK-Hyg selection gene cassette to be inserted into HBB intron 1 also was flanked by loxP sequences for future Cre-mediated excision (Figure 2A). We first evaluated the BD2 vector for gene targeting of the endogenous HBB gene in human 293T cells, in the presence or absence of the HBB-ZFNs delivered as 2 expression plasmids. After both hygromycin B and GCV selection (Hyg^R-GCV^R), pooled 293T cells were used in genomic PCR assays to detect whether there was any cell with a TI of the PGK-Hyg in the intron cocurrent with the SCD mutation replacement after HR by the same donor template. Two PCR primers, one primer within the inserted Hyg gene and the other primer downstream of the right homology arm, were used to detect the 3′-TI (2.5 kb). Our data showed that despite the high transfection efficiency in 293T cells, the BD2 donor vector alone did not generate TI mediated by HR in the silent HBB locus. However, adding HBB-ZFNs greatly stimulated gene targeting at the endogenous HBB locus (Figure 2B). Thus, we validated that the BD2 donor is capable of targeting HBB locus, and a pair of 3′-TI primers can be used to screen targeted iPSCs after cotransfection of the BD2 donor vector and 2 HBB-ZFNs followed by hygromycin and GCV double selections.

We next used a SCD patient-specific iPSC line MBP5s1 (S1) that was established previously by a virus-free method.¹³  We adapted S1 iPSCs to single-cell passaging so that they can survive better after nucleofection. After adaption for 10 continuous passages, the resulting S1 iPSCs at p26 retained iPSC characteristics, pluripotency, and a normal karyotype (data not shown). We then transfected S1 iPSCs cells with the BD2 donor and 2 HBB-ZFN plasmids by nucleofection, and we selected for clones conferring hygromycin and GCV resistance. Proceeding thus, we screened 300 Hyg^R-GCV^R S1 iPSC clones from multiple HR experiments. First using PCR for detecting the 3′-junction of TI, we identified 4 candidate clones (Figure 2C). One clone, c36, was further expanded and confirmed by Southern blot analysis to bear a TI of the PGK-Hyg cassette into intron 1 on one HBB allele and no additional random integration (Figure 2D-E). The other 3 clones either failed to proliferate or could not be confirmed positive by Southern blot (data not shown). DNA sequencing of c36 genomic DNA surrounding the HBB exon 1 also confirmed a GTG-to-GAG correction in the targeted allele (Figure 7B).

Although the inclusion of the PGK-Hyg selection gene cassette in intron 1 enables selection of rare HR clones critical to target genes such as HBB that are silent in undifferentiated iPSCs and ESCs, it may potentially interfere with expression of the corrected HBB allele.¹⁰  To alleviate this potential problem, we next excised the loxP-flanked PGK-Hyg cassette using transient transfection of a Cre-expressing plasmid. Four of 24 PCR-screened clones were shown to be negative of 3′-TI (Figure 2F) and 3 of those clones (cre4, cre16, and cre19) were selected and confirmed by Southern blot to be free of the integrated PGK-Hyg gene cassette (Figure 2G-H). Sequencing of genomic DNA PCR products using primers in exon 1 and 2 further confirmed the presence of 1 unmodified β^s allele and 1 corrected (β^A) allele that also bears the leftover loxP site (46 bp including linker DNA) in intron 1, as expected (Figure 3).

Through 2 steps of genetic manipulation, we obtained multiple human iPSC clones (c36, cre4, cre16, and cre19) with precise gene correction of the SCD mutation. These clones showed human ESC-like morphology, maintained normal karyotypes and pluripotency under ESC/iPSC culture conditions (Figure 4), similar the parental s1 iPSCs. Our genome-wide SNP concordance analysis did not detect overt difference between the c36 (after HR) and its S1 parental iPSC line (see “Methods”), although a detailed copy number variation or whole-genome sequencing analysis may provide more information about subtle genetic changes.

Because the HBB gene is not expressed in the undifferentiated iPSCs, we used differentiated iPSCs to evaluate the restoration of normal β-globin expression. Human iPSC lines S1, c36, cre4, and cre16 were differentiated into erythroid cells using a 2-step serum-free differentiation protocol (Figure 5A). During the first step of EB formation, all the iPSC lines form cystic EBs efficiently. At the end of the second step (erythroid expansion and maturation for 8 days), the differentiated cells expressed CD71⁺ (94%-98%) and CD235a (glycophrin A, 24%-29%) at similar levels (Figure 5B) and comparable to normal iPSC lines we tested previously.¹⁴  Giemsa stain confirmed most of the cells in culture were nucleated erythroblasts (Figure 5C).

We first used qRT-PCR to measure the total expression levels of the HBB gene (the β^s and a corrected allele) as well as the (fetal-type) HBG genes in all SCD iPSCs (S1, cre4, and cre16) and differentiated iPSCs at the EB stage and at the erythroblast stage (Figure 6A). Although the level of HBG expression in the SCD iPSC-derived erythroblasts (S1-EryB, cre4-EryB, and cre16-EryB) is comparable to that in cord blood mononuclear cells (CB-MNCs), the level of HBB expression in the erythroblasts is > 100-fold lower than that in CB-MNCs, even though they are ∼ 100-fold higher than those in the EBs (Figure 6A). Using a pair of primers located in exon 1 and 3 (Figure 2A blue arrows), we amplified the HBB transcripts from the erythroblasts of gene-corrected SCD iPSCs (c36, cre4, and cre16) by conventional RT-PCR (Figure 6B). Then, we cloned each PCR product that contains mixed DNA of corrected/wild-type and uncorrected/mutant alleles into TOPO vector and sequenced individual clones. This approach will tell us the frequencies of both gene-corrected (GAG) and uncorrected β^s (GTG) transcripts in erythrocytes from each iPSC line. We found that only the uncorrected β^s (GTG) allele was expressed in c36-derived cells because no sequenced HBB cDNA contained corrected β^A (GAG). In contrast, erythroid cells from both cre4 and cre16 iPSC clones (that had undergone Cre-mediated excision of the selection cassette) expressed the corrected GAG HBB allele, although at the level 25% to 40% compared with the uncorrected (and mutated) allele (Figure 6C). Our flow cytometry assay using antibodies for adult or fetal hemoglobin proteins confirmed the vast majority of hemoglobin expression in the erythroblasts is fetal-type (HbF), consistent with our qRT-PCR results (Figure 6D). Our result is also consistent with 2 recent reports on erythropoiesis of human iPSCs,^17,18  indicating that a better culture system is needed to increase erythrocyte maturation and survival, increase the HBB gene expression, and improve translation from existing HBB transcripts.

Our data showed that although HBB transcripts can be detected at a low level in the iPSC-derived erythroid cells, expression of gene-targeting corrected β^A allele was somehow totally or partially lost in differentiated erythroblasts at the same stage. We then took several approaches to investigate the potential epigenetic or genetic factors that may be responsible for the repressed β^A expression in c36, cre4, and cre16 cells. For c36, there was a 2.2-kb loxP-flanked PGK-Hyg gene cassette inserted in the first exon, excision of which resulted in partial expression of the gene-targeted allele (Figure 6C). The PGK-Hyg insertion at the first intron (37 bp away from the exon–intron junction) may affect the HBB gene expression in 2 ways. (1) The PGK-Hyg (intronless) gene cassette in the same orientation may introduce a cryptic/alternative splicing acceptor. We used 3 bioinformatics programs (NNSPLICE, http://www.fruitfly.org/seq_tools/splice.html; ASSP, http://www.es.embnet.org/∼mwang/assp.html; and NetGene2, http://www.cbs.dtu.dk/services/NetGene2/) to predict potential cryptic/alternative splicing acceptors in the insertion fragment. We identified several candidates and performed RT-PCR analysis, but we did not detect any alternative transcript around the region (data not shown). (2). The PGK promoter may make an antisense transcript that blocks HBB expression. However, no RT-PCR products can be found using several primers covering both strands of HBB exon 1 and the Hyg gene sequences. The silencing of the Hyg gene expression (driven by the PGK promoter) in expanded iPSCs is a surprise but consistent with the observation that c36 iPSCs lost their resistance to hygromycin selection (that was used in the initial clone screening) after expansion in the absence of selection. When we compared the hygromycin expression levels among several iPSCs where the same PGK-Hyg gene cassette is inserted either in the silent HBB gene (such as c36) or in the constitutively active PIG-A gene (such as FPHR iPSCs we published previously³ ), we found both expanded c36-iPSCs and c36-EryB expressed much lower hygromycin transgene expression than FPHR gene-targeted iPSCs (Figure 7A). This suggests epigenetic factors may play a role in silencing the gene-targeted allele including both the inserted PGK-Hyg transgene cassette and adjacent HBB gene.

Even though we had expected the insertion of drug-selection cassette might interfere with the transcription of gene-targeted allele, we were surprised to see partially repressed β^A expression in the cre4 and cre16 after Cre-mediated excision (Figure 5C). One possibility is that the remnant loxP sequence that is 37 bp away from the exon–intron junction interferes with the HBB gene expression. To examine other possibilities, we sequenced 3.5-kb sequence covering the whole HBB gene and its flanking genomic regions of all 3 gene-targeted lines (c36, cre4, and cre16) that had T > A replacement in exon 1 (Figure 7B left panels). Two single nucleotide variants that differ from the parental S1 sequence were found in the targeted allele of all 3 iPSC lines. The first variant, located downstream of exon 3 in the gene-targeted HBB allele (Figure 7B middle panels), is a polymorphic nucleotide change (G > T) in the 3′-homology arm and is probably introduced by the gene-targeting vector containing nonisogenic sequence. This SNP, however, has not been associated with any cis-regulatory elements or SNPs affecting HBB expression. The second variant (A > G) is found only in the targeted allele located downstream of the 3′-end of the right homology arm (Figure 7B right panels); therefore, it was not contained in the gene-targeting donor. This A > G mutation modified the core GATA binding site in the 3′-enhancer of the HBB gene expression.¹⁹  Deletion of the GATA-containing 3′-enhancer element has been shown to reduce HBB expression in transgenic mice.^20,21  Because this mutation was only found on the targeted allele of gene-corrected SCD iPSCs (Figure 7B), it was probably introduced during the HR-mediated repair and passed from c36 through cre4 and cre16. At present, it is unclear whether this mutation in the GATA-containing 3′ enhancer, or the remaining loxP site in the first intron, or both, are responsible for the reduced expression from the targeted HBB allele.

Various methods have been explored to increase gene-targeting efficiencies in human iPSCs and ESCs, with or without the use of ZFNs. Recombinant adenovirus-associated viruses,^5,6  helper-dependent adenoviruses,²²  and bacterial artificial chromosome vectors²³  have been used in gene targeting/correction without ZFNs in several cell types, including human ESCs and iPSCs. However, these previous studies have been focusing on constitutively transcribed genes in human iPSCs. Regardless, if the underlying target allele is expressed or silent in undifferentiated iPSCs, the method we described here should apply to correction or replacement in human iPSCs of any specific mutation or even an SNP locus. During the revision of this report, 2 other studies reported the correction of single mutations in human iPSCs.^24,25  Together with our study reported here on gene correction of the SCD mutation, these studies show the feasibility regardless the underlying gene is expressed or silent in human iPSCs.

Numerous studies have reported that homology-directed repair is often much less efficient when a silent gene is targeted.^26,27  This also explains our observed low efficiency of targeting the HBB gene even in the presence of a pair of ZFNs, compared with that of a constitutive gene (PIG-A) we published previously.³  The HBB-ZFNs used here is 50% less efficient than the PIG-A ZFNs measured by the similar GFP* reporter assays (Figure 1B), which also could contribute to a lower gene targeting of the HBB gene in human iPSCs. With an improved ZFN technology²⁸  or other emerging methods such as TALE nuclease,²⁹  we will probably find a high-affinity ZFN or TALE nuclease at or near the SCD mutation in HBB, without cutting the nearby HBD, HBG, and HBE genes.

For early passage human iPSCs that typically have less than 1% colony-forming efficiency, successful gene targeting of a silent gene such as HBB in human iPSCs relies more on a co-integrated drug-selection gene driven by a constitutively active promoter. Even with a floxed drug-resistant gene, gene targeting of a silent gene may still face obstacles because of silencing of the targeted locus in undifferentiated iPSCs. Indeed, we noticed that the HR-targeted c36 clone (c36) obtained after 2 weeks of Hyg selection became sensitive again to hygromycin on subsequent culture. Our RT-PCR results confirmed very low expression of PGK-Hyg transgene cassette in both undifferentiated c36 iPSCs and differentiated erythroid cells (Figure 7A). We hypothesized that epigenetic mechanisms such as DNA methylation in pluripotent cells may cause expression silencing of a transgene inserted at unfavorable loci, which in turn induce repression in gene expression of nearby targeted endogenous gene such as HBB even after differentiation. Similar reduction of gene-targeted allele expression also was observed in the previous report of correcting a humanized SCD mouse ESCs harboring the human β^S allele.¹⁰  However, the transgene expression cassette inserted into an intron can be easily excised out via the Cre-loxP system, and the endogenous gene expression can be largely restored after excision.

Although in most applications we do not need to completely restore expression level of the corrected allele to the same level of the uncorrected alleles, it is better for us to understand why only a partial restoration (25%-40%) of the corrected allele in cre4 and cre16 iPSCs after the loxP excision (Figure 6B). Currently, we could not distinguish the following 2 possibilities and more studies are needed. (1) The presence of 1 copy of loxP left in the first intron interfered with the endogenous HBB gene expression. (2) A genetic mutation in the GATA site of HBB 3′-enhancer (probably introduced in the iPSC HR selection process, the clonal selection process, or both) is responsible for the reduction of erythroid-specific HBB expression. For the former possibility, we may use piggyBac DNA transposition to achieve zero-footprint excision of the selection gene cassette.³⁰  For the latter possibility, our study highlights the importance to conduct high-resolution whole-genome sequence analysis. Although the HR-gene targeting did not show increased mutation rates compared with other proliferating cell populations in studies published by us and others,^3,4,24  previously used methods of chromosomal karyotyping, SNP arrays, and even exome sequencing may not be sufficient to detect small changes across the whole genome. With recent cost reduction of whole-genome sequencing that is now affordable to an average laboratory, it is now possible to sequence the whole genome of selected iPSC clones after ZFN-mediated gene targeting.

Although in this study we focused on site-specific gene correction of a defined single mutation in the HBB gene, our functional analysis and relevance to future gene therapy is hampered by the fact that existing culture methods only allow us to generate nucleated erythrocytes that express γ-globin instead of β-globin protein. Nevertheless, HBB transcripts can be detected, albeit at a low level, by RT-PCR in differentiated erythroblasts derived from parental S1 and corrected cre iPSC lines (Figure 6). This allowed us to show that the corrected HBB allele after gene targeting can be expressed. An ideal iPSC-based gene therapy for SCD will require both precise correction of disease-causing mutation while preserving cis-regulatory elements regulating HBB expression and complete switching from fetal-type globin to adult-type globin. Therefore, further genetic and epigenetic studies are required to better understand globin gene switching in the iPSC system and to improve hematopoietic differentiation into more mature/enucleated erythroid cells from gene-corrected iPSCs, if they are going to be used for cell therapy. In addition, reproducible in vivo models to generate transplantable hematopoietic progenitor cells and red blood cells from human iPSCs and ESCs are needed,^31,32  before we can assess the preclinical feasibility of the combined cell and gene therapy approach using gene corrected iPSCs derived from SCD patients. Despite these hurdles to be overcome in future years, our current study demonstrates that it is now possible to correct a specific point or small mutation at its endogenous locus in patient-specific iPSCs and also to restore gene expression of the corrected form on cell differentiation. The gene targeting methodology we reported here would extend the use of various iPSC lines containing disease related mutations in genes that are either expressed or silent in human iPSCs.

The authors thank the laboratory of Prof Y. W. Kan for providing the PGK-Cre expression vector, and a Sigma-Aldrich team for designing and providing the plasmids expressing HBB-ZFNs.

This research was supported by a Siebel scholarship (P.M.), Fellowships and an Exploratory grant from the Maryland Stem Cell Research Fund (2008-MSCRFF-009 and 2010-MSCRFE-0044, J.Z. and 2010-MSCRFF-0095, X.H.), and the National Institutes of Health (grants R01 HL073781 and HL073781-05S1; L.C.).

National Institutes of Health

Contribution: J.Z. and P.M. designed and performed experiments, collected and analyzed data, and wrote the manuscript; X.H. and S.N.D. performed experiments and collected and analyzed data; and L.C. designed the experiments and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Linzhao Cheng, The Johns Hopkins University School of Medicine, Broadway Research Bldg, Rm 747, 733 N Broadway, Baltimore, MD 21205; e-mail: lcheng2@jhmi.edu; or Jizhong Zou, The Johns Hopkins University School of Medicine, Broadway Research Bldg, Rm 780, 733 N Broadway, Baltimore, MD 21205; e-mail: jzou2@jhmi.edu.