Efficient gene correction of an aberrant splice site in β‐thalassaemia iPSCs by CRISPR/Cas9 and single‐strand oligodeoxynucleotides

Abstract β‐thalassaemia is a prevalent hereditary haematological disease caused by mutations in the human haemoglobin β (HBB) gene. Among them, the HBB IVS2‐654 (C > T) mutation, which is in the intron, creates an aberrant splicing site. Bone marrow transplantation for curing β‐thalassaemia is limited due to the lack of matched donors. The clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR‐associated protein 9 (Cas9), as a widely used tool for gene editing, is able to target specific sequence and create double‐strand break (DSB), which can be combined with the single‐stranded oligodeoxynucleotide (ssODN) to correct mutations. In this study, according to two different strategies, the HBB IVS2‐654 mutation was seamlessly corrected in iPSCs by CRISPR/Cas9 system and ssODN. To reduce the occurrence of secondary cleavage, a more efficient strategy was adopted. The corrected iPSCs kept pluripotency and genome stability. Moreover, they could differentiate normally. Through CRISPR/Cas9 system and ssODN, our study provides improved strategies for gene correction of β‐Thalassaemia, and the expression of the HBB gene can be restored, which can be used for gene therapy in the future.

by mutations, gene therapy provides a potential treatment for the disorder.
With the advent of technological breakthroughs on gene editing, the efficient means of curing the genetic disease is to correct the mutation directly via sequence-specific endonucleases. Endonucleases can create double-strand breaks (DSBs), which then activates DNA repair by two highly conserved competing mechanisms: nonhomologous end joining (NHEJ) or homology-directed repair (HDR). 6,7 NHEJ repairs breaks via ligation of DNA ends throughout the cell cycle, and it causes nearly random insertion and deletion mutations.
Nevertheless, HDR can be exploited to make the desired sequence replacement at the DSB site by homologous recombination with a donor DNA template, which is normally most active during the S or G2 phase of the cell cycle. This allows us to utilize HDR to generate targeted gene deletion, mutagenesis, insertion or gene correction. 8,9 Recently, the clustered regularly interspaced short palindromic repeats (CRISPR)/ CRISPR-associated protein 9 (Cas9), an RNA-guided nuclease from an adaptive immune mechanism present in many bacteria and the majority of characterized Archaea, has been widely used as an endonuclease for gene correction through HDR. The system can bind DNA through 20-bp gRNA which is adjacent to protospacer adjacent motif (PAM) by Watson-Crick base pairing and then generate the cleavage by Cas9 protein. 10,11 Due to the fact that CRISPR/Cas9 system is easier to design and construct, as well as having high efficiency of gene editing, many studies have reported using the system for correcting disease-related mutations in animal somatic 12 and germ line cells, 13,14 as well as in human stem cells 15 and induced pluripotent stem cells. 16,17 The single-stranded oligodeoxynucleotide (ssODN) has been applied as the repair template to generate point mutation. 18 Contrasting to dsDNA, ssODN can be synthesized more easily and quickly, which is not required to excise the section marker. In addition, it was reported that ssODN is more efficient for HDR. 19 In this study, combining CRISPR/Cas9 and ssODN, we successfully repaired the biallelic HBB IVS2-654 mutation in induced pluripotent stem cells (iPSCs) via a seamless approach in one step. Because gRNA was not designed at the mutant site resulting from the PAM region-NGG while the mutation is in intron 2, we could not make a synonymous mutation; a more efficient strategy was adopted to correct the mutation by reducing the occurrence of secondary cleavage. After gene correction, iPSCs still kept pluripotency and genome stability. The corrected iPSCs could be utilized for hematopoietic differentiation normally, and the expression of the HBB gene was restored. Therefore, our study offered improved strategies for gene correction of β-Thalassaemia.

| Cell culture and hematopoietic differentiation
The iPSCs derived from a patient with β-thalassaemia were provided by the Third Affiliated Hospital of Guangzhou Medical University where the experiments were consented by the ethics committee.
We replaced new medium daily. The iPSCs with 80% confluence in 35 mm dish were treated with 1 mg/mL dispase (Gibco), and then, small scraped clumps were harvested, which were cultivated on Matrigel-coated 12-well plate with 1:60 dilution. According to a fivestep hematopoietic differentiation strategy, the cells were expanded in different mediums containing different cytokines (PeproTech) as previously reported. 20 At days 12 and 22, differentiated cells were collected via fluorescence-activated cell sorting (FACS).

| Differentiation of three germ layers and teratoma formation
The iPSCs were treated with dispase and cultured in ultra-low attach-

| Karyotype analysis and the assay for short tandem repeat
The iPSCs were incubated with the culture medium added 0.25 mg/ mL colcemid (Invitrogen) for 4 hours and then incubated in mixed solution containing 0.4% sodium citrate and 0.4% potassium chloride

| T7 Endonuclease I assay
We designed different primers near the predicted gRNA off-target sites throughout the whole genome from the online software CCTop. When using CCTop, we chose a custom target selection with in vitro transcription while the species was set as Human (Homo sapiens GRCh37/hg19). Other parameters were unchanged. These fragments of Genomic DNA extracted from iPSCs were amplified with the primers by PrimeSTAR GXL DNA Polymerase (TAKARA) and then purified through Universal DNA Purification Kit (Tiangen).
The purified products were used for T7 Endonuclease I (T7E1) assay following the manufacturer's protocol (New England BioLabs) and analysed on 2% agarose gels using Gel Imaging System (Bio-Rad).

| Whole exome sequencing and Sanger sequencing
The

| Statistical analysis
The data were subjected to statistical analysis by unpaired Student's two-tailed t test, which were presented as means ± SEM. A value of P < .05 was considered to be a statistically significant result. The editing efficiency = (the quantity of indel clone, homozygous repairing clone and heterozygous repairing clone/ the quantity of total tested clone) × 100%. The repairing efficiency = (the quantity of homozygous repairing clone and heterozygous repairing clone/ the quantity of total tested clone) × 100%.

| The design of different gene correction strategies for the HBB IVS2-654 mutation and ssODN selection
According to the regulation of PAM, a 20-bp gRNA, which is adjacent to AGG, was designed near the HBB IVS2-654 (C > T) mutation.
It was also proved to have higher efficiency in previous research. 21 Contrasted to dsDNA, ssODN is more efficient for HDR. 19,22 So we used a 127 bp-ssODN in the complementary strand of gRNA as the donor. However, because the IVS2-654 mutation that the designed gRNA did not contain is in intron 2 of the HBB gene and we could not make a synonymous mutation, we made a two-step strategy avoiding the occurrence of secondary cleavage. The first part was that we corrected the HBB IVS2-654 mutation and introduced a new mutation belonging to the region of gRNA at the same time. For the second part, we repaired the introductory mutation. We designed two different introductory mutations. Thus, two different ssODNs were used in the first part and two unlike gRNAs were used in the second part ( Figure 1A,B).
To correct the biallelic HBB IVS2-654 mutation, CRISPR/Cas9 system and ssODN were utilized in iPSCs which derived from the reprogramming of a patient's fibroblasts. To evaluate which ssODN had higher HDR efficiency for the two-step strategy, flow cytometric analysis was performed by gRNA with mCherry reporter, Cas9 with GFP reporter at 48 hours after the electroporation of gRNA vector, cas9 plasmid and ssODN 1 or ssODN 2 ( Figure 1C). Double positive cells were harvested and re-planted with low density. After about a week, the clones were picked up and identified via PCR. We found 29% of clones were indels in the ssODN 1 group with a total of 5% of them were heterozygote at mutant site and 2% of them were homozygote. While 29% of clones were indels, in the ssODN 2 group we found that 16% of them were heterozygote and 4% of them were homozygote ( Figure 1D). The results indicated ssODN 1 and ssODN 2 had similar editing efficiency but ssODN 2's efficiency was more stable. SsODN 2 was also more efficient than ssODN 1 for repairing. Therefore, gRNA2 and ssODN 2 were adopted in the next experiment for the two-step strategy. and GFP (Cas9; Figure 2A) and then cultured them with low density.
Until the formation of clones, we picked up 40-60 clones every time and screened the corrected cell lines at the mutant site via PCR. We could repair the HBB IVS2-654 mutation by both of the two strategies and detected the introductory mutant in the first part but not in the second part for the two-step strategy ( Figure 2D). The results for the one-step group showed 45% of clones were indels, 3% of them were heterozygote for gene editing and 1% of them were homozygote. In the two-step group, while 26% of clones were indels, 12% of them were heterozygote and 9% of them were homozygote. As demonstrated, the editing efficiency for the second strategy did not have much difference compared to the first, whereas the repairing efficiency for the two-step strategy was much higher than the one-step strategy ( Figure 2B,C).

| Characterization of pluripotency in the genecorrected iPSCs
To identify whether the iPSCs after gene repair retain normal pluripotency, two gene-corrected iPS cell lines (corrected C1-iPS and corrected C2-iPS) from the two-step strategy were chosen for further detecting. The iPSCs before (pre-iPS) or after gene correction displayed typical morphology and the AP staining of them was positive ( Figure 3B). Quantitative PCR analysis showed the iPSCs before or after gene correction had higher expression of pluripotency-related genes, such as OCT4, SOX2, NANOG, GDF3 and DPPA4, comparing with the patient's fibroblasts ( Figure 3A). Immunofluorescence results also revealed the typical pluripotency markers: OCT4, SSEA4, SOX2 and TRA-1-81 were expressed in these iPSCs ( Figure 3C, 3). Moreover, pre-iPS, corrected C1-iPS and corrected C2-iPS cell lines could differentiate into three different germ layers in vitro after the formation of EBs, which were showed via immunofluorescence ( Figure 3E). We also acquired teratomas that had three different germ layers in vivo  Figure 3F). All the above results indicated that the gene-corrected iPSCs kept pluripotency.

| The stability of genome in the genecorrected iPSCs
At first, we made sure that pre-iPS, corrected C1-iPS and corrected C2-iPS derived from the same patient though STR assay ( Figure 4A).
These iPSCs had normal karyotype ( Figure 4B). We obtained these predicted off-target sites using the online software CCTop. Then, T7E1 assay was performed after we did PCR with the extracted DNA of iPSCs. The T7E1 assay and Sanger sequencing revealed there was no off-target mutagenesis at top 9 sites in corrected C1-iPS and corrected C2-iPS cells ( Figure 4C, Table S1). To further confirm whether  Figure 4D, Table S2-S4). All the sites of SNVs and indels were not in accord with the predicted off-target site. The sequences at these sites had too many mismatches with gRNA which made it not easy for gRNA to target these sites. Thus, that suggested CRISPR/Cas9 system was not the direct reason for the mutagenesis.

| The restoration of HBB gene expression in the gene-corrected iPSCs
The patients with β-thalassaemia show issues related to the β-globin

| D ISCUSS I ON
Thalassaemia is one of the most common genetic disease resulting from the imbalance of globin chain production mostly caused by gene mutation. 23 Due to the limitations of different clinical treatment at present, patients cannot have effective recovery, which emphasizes the importance of seeking new ways for therapy that targets thalassaemia. Hope is offered and found with engineered nucleases. Zinc-finger nucleases (ZFNs) and transcription activator-like effector nucleases (TALENs) are produced via proteins fused with the nuclease domain of the restriction enzyme FokI and they work through protein-DNA interaction. 24 Previous reports have shown HDR rates varied at 33% and 68% when we used TALEN and dsDNA to edit. They are not used widely because they need to engineer and clone a new protein when targeting a new site, which is a complicated process. The difference of the repairing efficiency is large as well. 21,25 However, CRISPR/Cas9, an RNA-guided system, is easier to be designed and constructed with lower costs.
Gene correction is based on the sequence-specific targeting of gRNA, the DSB caused by Cas9 protein and the repair of gene via the donor template for HDR. In previous report, the HBB IVS2-654 mutation in β-thalassaemia iPSCs had been corrected by CRISPR/ Cas9 with dsDNA. 21 Another team produced the disruption of genomic elements to make indels for removing the mutation using LbCas12a RNP, and the efficiency was up to 76.6%. Whether there are potential risks for changing genome sequences is unknown. 26 Nevertheless, contrasted to dsDNA, ssODN is more efficient for HDR and has lower cytotoxicity. 19 In this study, we demonstrated an efficient approach for correcting the biallelic HBB IVS2-654 mutation in β-thalassaemia iPSCs combining CRISPR/Cas9 system with ssODN which were electroporated into iPSCs. Because antibiotic selection may interfere with the expression of corrected gene, 27,28 the positive cells with fluorescent reporter were harvested by FACs.
We finally acquired the corrected iPS cell line after Sanger sequencing for expanding clones. Next, many assays revealed the corrected iPSCs remained pluripotency, genome stability and differentiation ability. Most importantly, we examined the expression of HBB gene and concluded the function of HBB gene was successfully restored in the gene-corrected iPSCs.
Because most patients who have the heterozygous HBB IVS2-654 mutation do not show to have symptoms, repairing the monoallelic mutation can cure them. In the one-step group, the repairing efficiency was 4% on average including both heterozygote and homozygote. However, for that strategy, we could not design gRNA including HBB IVS2-654 mutant site, which should be adjacent to NGG due to spCas9. Therefore, gRNA was designed near the mutation. Then, there is the problem that gRNA can still F I G U R E 5 Hematopoietic differentiation of gene-corrected iPSCs. A, Experimental scheme for a five-step hematopoietic differentiation strategy from iPSCs. B, Images of representative morphology changes in different hematopoietic differentiation stages. Scale bar, 500 μm (Day 0, day 2, day 4, day 6, day 12); Scale bar, 200 μm (Day 22). C, Flow cytometric analysis of CD34 + expression at day 6 during the hematopoietic differentiation of n-iPS, pre-iPS, corrected C1-iPS and corrected C2-iPS. D, Representative images of colony morphologies for CFU assay after another 14 days differentiation using the CD34 + cells at day 12 during the hematopoietic differentiation. Scale bar, 100 μm (CFU-E); Scale bar, 500 μm (CFU-G, CFU-M, CFU-GM, CFU-MIX). E, Flow cytometric analysis of CD235a + expression at day 22 during the hematopoietic differentiation of n-iPS, pre-iPS, corrected C1-iPS and corrected C2-iPS. F, The agarose gel images of RT-PCR product by amplifying HBB cDNA of CD235a + cells derived from the hematopoietic differentiation of n-iPS, pre-iPS, corrected C1-iPS and corrected C2-iPS cell lines. CD34 + cells from cord blood were used as a positive control (Cord blood-E), and normal iPSCs were used as a negative control. G, Quantitative PCR analysis of HBB gene expression (normalized to β-actin) in CD235a + cells derived from the hematopoietic differentiation of cell lines before or after gene correction. Results were represented as mean ± SEM for n = 3 individual experiments; **, P < .01, ***<0.001; t test target the sequence after gene correction. The mutation is from an intron and we cannot make a synonymous mutation. To reduce the occurrence of secondary cleavage, we adopted a two-step strategy. In the first part, we corrected the HBB IVS2-654 mutation and introduced a new mutation at the same time which belong to the region of gRNA. Next, we repaired the introductory mutation in the second part. Two strategies had similar editing efficiency, whereas the repairing efficiency of the two-step strategy could reach 21% on average, which was about 5 times that of the one-step strategy. The HDR rate was also higher than it was with CRISPR/Cas9 and dsDNA in the previous report (12.3%). 21 That revealed the two-step strategy about inducing a new mutation can indeed reduce the occurrence of secondary cleavage and improve the repairing efficiency ( Figure 2). In addition, we found that the repairing efficiency, not the editing efficiency, also has significant difference if inducing a new mutation from different nucleotide ( Figure 1). Nevertheless, comparatively speaking, the two-step strategy is not simple. Hence, we tried another method in the one-step method. We electroporated all the gRNAs and ssODNs as well as Cas9 used in the two-step strategy into iPSCs together in a single time. We assumed that they could work in cells twice because plasmids can stay in cells for several days. But the final result showed the editing efficiency from the mixed gRNAs and ssODNs was lower than that of the one-step strategy mentioned above ( Figure S1). We think the most likely reason is that the mass of gRNA and gRNA2 were comparing half to the mass of gRNA used in the one-step strategy because the same total mass of gRNA should be kept in the experiment. For this aspect, we can solve the problem by utilizing a vector which can carry multiple gRNAs in the further study. Moreover, there are many other ways to improve the efficiency of gene repair, for example, adding small molecules, synchronizing the cell cycle, and adjusting delivery timing and methods. [29][30][31] Besides, using ribonucleoprotein (RNP) delivery of Cas9 with gRNAs consistently can increase activity in cells and then enhance the efficiency. 19,31 Since gRNA can target similar sequences in a genome, it is possible to have off-targets when we use SpCas9 system. 32,33 Thus, we did T7E1 assay for 9 potential off-target site analysed by the online software CCTop and confirmed them by Sanger sequencing.
This revealed that the corrected iPS cell line had no indel. Next, the whole exome sequencing was performed to assess off-targeting effects, and then, we found some SNVs and indels from the corrected iPS cell lines contrasting to the iPS cell line before repair. But all the sites from these SNVs and indels were not in the potential off-target regions according to gRNA targeting sequence ( Figure 4, Tables S2-S4). We considered the fact that sometimes high-throughput sequencing can have false-positive results, and it is possible for cells to produce some mutations after more and more passage. To solve these problems, we can do experiments with cells at the early passages and perform Sanger sequencing to identify the results from the whole exome sequencing. To improve SpCas9 system's targeting specificity, we also can adopt the following strategies: shorter gRNAs design or gRNAs with two unpaired Gs on the 5′ end which are more sensitive to mismatches, 34,35 paired nCas9s, 36 have the ability to differentiate into all cell types. Thus, we did hematopoietic differentiation and found that the function of HBB gene was successfully restored in the gene-corrected iPSCs ( Figure 5).

| CON CLUS IONS
From this study, we describe a one-step approach to correct the biallelic HBB IVS2-654 mutation in β-thalassaemia iPSCs through CRISPR/Cas9 and ssODN-mediated HDR. For the mutation in an intron and no appropriate gRNA containing the mutant site, a twostep strategy can be adopted to reduce the occurrence of secondary cleavage for improving the repairing efficiency. The corrected iPSCs keep pluripotency and genome stability. Moreover, the expression of HBB gene can be restored in vitro after hematopoietic differentiation. Therefore, our findings demonstrate that the strategies of gene correction we report here will facilitate the development of cell therapy for genetic disease in iPSCs.

ACK N OWLED G EM ENTS
The authors would like to thank Yi Liang for Karyotype and STR

CO N FLI C T O F I NTE R E S T
The authors declare no competing interests.

DATA AVA I L A B I L I T Y S TAT E M E N T
The raw data of whole exome sequencing reported in this paper have been deposited in the Genome Sequence Archive in BIG Data Center, Beijing Institute of Genomics (BIG), Chinese Academy of Sciences, under accession numbers CRA001893, CRA001893 that are publicly accessible at https ://bigd.big.ac.cn/gsa.