CRISPR-mediated genotypic and phenotypic correction of a chronic granulomatous disease mutation in human iPS cells

Chronic granulomatous disease (CGD) is a rare genetic disease characterized by severe and persistent childhood infections. It is caused by the lack of an antipathogen oxidative burst, normally performed by phagocytic cells to contain and clear bacterial and fungal growth. Restoration of immune function can be achieved with heterologous bone marrow transplantation; however, autologous bone marrow transplantation would be a preferable option. Thus, a method is required to recapitulate the function of the diseased gene within the patient's own cells. Gene therapy approaches for CGD have employed randomly integrating viruses with concomitant issues of insertional mutagenesis, inaccurate gene dosage, and gene silencing. Here, we explore the potential of the recently described clustered regularly interspaced short palindromic repeat (CRISPR)-Cas9 site-specific nuclease system to encourage repair of the endogenous gene by enhancing the levels of homologous recombination. Using induced pluripotent stem cells derived from a CGD patient containing a single intronic mutation in the CYBB gene, we show that footprintless gene editing is a viable option to correct disease mutations. Gene correction results in restoration of oxidative burst function in iPS-derived phagocytes by reintroduction of a previously skipped exon in the cytochrome b-245 heavy chain (CYBB) protein. This study provides proof-of-principle for a gene therapy approach to CGD treatment using CRISPR-Cas9.

The advent of site-specific nucleases has stimulated much excitement for their potential to spawn a new era of in vitro experimental human genetics, in a similar vein to the impact of transgenic mice in the 1980s. Site-specific nucleases also have great potential as therapeutic tools, in theory capable of elevating homologous recombination in human cells to a level that could truly provide a personalized curative gene therapy option for genetic diseases [1,2]. Here, we investigate the site-specific clustered regularly interspaced short palindromic repeat (CRISPR)-Cas9 system for correction of a point mutation in the CYBB gene that results in chronic granulomatous disease (CGD).
CGD, a disease characterized by recurrent, severe bacterial and fungal infections, results from an inability of phagocytic cells, particularly the innate immune sentinels macrophages and neutrophils, to generate an oxidative burst upon recognition of an invading pathogen [3]. This oxidative burst generates various reactive oxygen species (ROS), such as hydrogen peroxide, that are able to neutralize the pathogen, thereby aiding in clearance and preventing its continued spread. Although antibiotic treatment options exist for CGD, they are not optimal, since there is a lifelong dependency, and the only curative therapy involves heterologous bone marrow transplantation, which has its own inherent risks. Human leukocyte antigen (HLA)-identical donors outside siblings are also extremely rare. An alternative treatment option, gene therapy using autologous bone marrow transplantation of hematopoietic stem cells modified with retroviral vectors to express a wild-type (WT) copy of the mutated gene, has been attempted in clinical trials, with initial curative success [4].
However, the expression of the transgene waned with time, and complications arose due to insertional mutagenesis resulting in myelodysplasia [5]. This demonstrates the potential for success but also the need for a cleaner system to perfectly genetically correct the diseased genome.
Homologous recombination as an experimental tool has historically been an inefficient process, the use of which has been constrained to a limited range of model organisms (notably bacteria, yeast, trypanosomes, and transgenic mice [6][7][8]). The development of site-specific nucleases, such as that based on the bacterial adaptive antiviral immune system, CRISPR-Cas9 [9], have been key in expanding the use of homologous recombination in human cells. Creation of double-strand breaks (DSBs) at the precise location desired for genetic modification can enhance the efficiency of homologous recombination to levels that allow both easy isolation of modified cells and, depending on requirement, the use of the cells as a mixed population of modified and unmodified cells [10].
CGD is a monogenic disease and is a prime candidate for gene therapy, particularly since bone marrow transplantation is already a treatment option. Although there are a number of genes involved in the ROS-producing nicotinamide adenine dinucleotide phosphate (NADPH) oxidase complex, the mutation of any of which can result in CGD, the majority of cases (O60%) are due to loss of function of the cytochrome b-245 heavy chain (CYBB) protein (or GP91 PHOX ) [11]. The gene encoding CYBB is located on the X chromosome and, therefore, is only present as a single copy in male sufferers. We [12] and others [13] have previously generated induced pluripotent stem cells from CGD suffers, the differentiated myeloid descendants of which recapitulate the ROS defect characteristic of the disease. Using one of these patient-derived iPS cell lines (CGD2) with a single point mutation (T O G) at the end of intron 1 of CYBB [12], we report high levels of gene correction using CRISPR-Cas9, show recovery of gene function in differentiated phagocytic progeny cells, and demonstrate that the T O G mutation results in exon skipping within the mRNA of CYBB.

Monocyte/macrophage differentiation and ROS activity
Production of monocytes and macrophages from iPS cells has been described previously [15]. Monocytes released from the factories were either used directly, adhered to tissue-culture plates for 24 hours, or differentiated into macrophages on tissue culture plates in XVIVO 15 supplemented with GlutaMAX and 100 ng/mL macrophage-colony stimulating factor (M-CSF), before analysis for ROS production. Nitroblue tetrazolium (NBT), dihydrorhodamine (DHR), and luminol assays were used to identify ROS activity from monocytes and macrophages, as previously described [12], after stimulation with 500 ng/mL phorbol 12-myristate 13-acetate (PMA; 200 ng/mL for the luminol assay) or 500 ng/mL PMA and 0.5 mmol/L formyl-methionyl-leucylphenylalanine (fMLP), respectively.
Results CRISPR-Cas9 system has the potential to perform gene editing at clinically relevant levels for CGD The frequency of cells with a functional NADPH oxidase complex required to relieve CGD sufferers' symptoms is estimated to be only 10% of circulating monocytes and neutrophils [22][23][24]. Thus, we set out to establish if such levels of gene editing were possible using a model system for accurate quantification of homologous recombination rates. The model consisted of cell lines (either HEK293 or OX1-19 iPS cells) transduced by a lentiviral vector expressing a BFP that differs from eGFP by two neighboring amino acid substitutions within the chromophore (S65T and H66Y; Fig. 1A). After transfection of these cells with a plasmid expressing Cas9 and a guide RNA (gRNA-27) targeting a site nearby the BFP/eGFP mutations, along with a repair template containing a partial WT eGFP sequence, the fluorescent marker switch from BFP to eGFP can be easily detected and the gene editing frequency quantified by flow cytometry (Fig. 1B). Using this model, we were able to show that the CRISPR-Cas9 system is able to achieve rates of homologous recombination above 10% in both HEK293 cells and iPS cells (Fig. 1C). This rate of gene editing (17.0% 6 0.35 for iPS cell and 13.6% 6 1.4 for HEK293 cells) was only observed when using a plasmid donor template (Mut plasmid) containing a mutated PAM, a sequence essential for cleavage by Cas9. In contrast, without a mutated PAM (WT plasmid) tenfold lower levels of gene editing were detected (0.77% 6 0.05 for iPS cells and 1.4% 6 0.75 for HEK293 cells). A similar phenomenon was observed when using single-stranded oligonucleotides (ssODNs) as donor templates; an ssODN with a mutated PAM (Mut ssODN) was more efficient than an ssODN with the WT sequence (WT ssODN; 1.16% 6 0.09 versus 0.16% 6 0.1; p ! 0.001; Fig. 1B). Since the ssODN was noncomplementary to the gRNA-27 and, therefore, not a target for cleavage by Cas9, the reduction in gene editing observed with a WT PAM was likely due to Cas9directed cleavage and mutation of the gene after successful homologous recombination, resulting in loss of the eGFP signal. Moreover, in our hands the level of gene editing using plasmid donors far exceeded that of single-stranded oligonucleotides, and, therefore, a plasmid donor was chosen for correction of the CGD-causing mutation.
Gene editing of the endogenous CYBB gene locus using the CRISPR-Cas9 system can reach clinically relevant levels With the aim being establishment of protocols for gene therapy of CGD using the CRISPR-Cas9 system, we chose to make use of the D10A-mutated version of Cas9, which is only capable of cleaving a single strand of DNA [25]. By providing two neighboring gRNAs, the Cas9 D10A protein can create a DSB, resulting in similar levels of homologous recombination to the WT Cas9, but with the added advantage of increasing the specificity of DSB formation and, therefore, reducing off-target mutagenic events [25]. The location of the gRNAs selected to target the CYBB gene of iPS cells from patient CGD2 are shown in Figure 2A. Note that one of the gRNA target sites contains the disease-causing mutation within its PAM, which is not present within the WT CYBB sequence; our intention was to improve the efficiency of gene editing as seen in the fluorescent reporter model. Using these conditions, gene editing at the CYBB locus was detected at frequency of 11% Single cell clones, generated from CGD2.GC16A cells, were sequenced for their CYBB identity (Supplementary Figure E2, online only, available at www.exphem.org). Of the 60 isolated clones, 14 (23%) were correctly modified, containing the WT sequence at the intron/exon boundary, 21 (35%) contained insertion or deletion mutations at the site of CRISPR-Cas9 cleavage, and 18 (30%) were unaffected by the treatment. Thus, in agreement with our fluorescent model, levels of gene conversion that have the potential to be clinically relevant were obtained using CRISPR-Cas9 gene editing at the CYBB locus. Two clones (C4 and E4) were expanded and assayed for pluripotency (Supplementary Figures E3A and E3B, online only, available at www.exphem.org) and karyotypic abnormalities (Supplementary Figure E3C, online only, available at www.exphem.org) to ensure that the procedure had not adversely affected their potential to generate phagocytic cells and to serve as isogenic cell lines for the parental unmodified CGD2 cell line.

Gene correction results in recovery of ROS activity
To demonstrate phenotypic correction at the CYBB locus, as well as genotypic correction, the iPS cells were differentiated into monocytes using an embryoid body-based protocol developed in our laboratory [15,26]. As was observed previously [12], NADPH-oxidase-positive and -negative lines all differentiated with similar efficiency, and monocyte factories from all cell lines produced nonadherent Induced pluripotent stem cells generated from patient CGD2 were grown under feeder free conditions, electroporated with CRISPR-gRNA constructs and repair template, and selected for transfected cells using puromycin. The residual heterogeneous population of cells was assayed for gene-editing frequency by sequencing and was subsequently passaged onto MEF feeder cells. Single-cell clones were picked, grown on MEFs, and assayed individually for their sequence at the CYBB gene. Successfully modified clones were expanded, converted to feeder-free culture, assayed for pluripotency and intact karyotype, and differentiated into monocytes/macrophages using an embryoid-body-based protocol [15]. Monocytes harvested from the factory supernatant, and subsequently M-CSF-differentiated macrophages, were then assayed for phenotypic correction of the CGD mutation. single cells that were over 95% CD14 positive, as determined by flow cytometry (Supplementary Figure E3D, online only, available at www.exphem.org). To act as controls for subsequent experiments, alongside the CGD2 cell lines and its derivatives, monocytes were differentiated from WT iPS cells as well as from a CGD patient (CGD1) with a mutation in the P47 Phox subunit of the NADPH oxidase complex, which completely abolishes ROS production [12].
Three different assays for the generation of ROS activity were performed on the iPS-derived myeloid cells. Firstly, a qualitative assay for ROS, the NBT assay, was carried out on monocytes adhered to tissue culture-treated plastic for , and the two gene-edited single-cell clones from CGD2.GC16A (C4 and E4) were stimulated with PMA or with PMA and fMLP, respectively, in the presence of NBT for 1 hour. Following exposure, the cells were washed in phosphate-buffered saline and fixed. Brightfield images were taken of cells not exposed to the stimulant or the NBT (Neg), cells exposed to the NBT but without stimulation (UnStim), and cells exposed to both NBT and stimulant (Stim). Images are representative of two independent experiments. (C) Macrophages were incubated for an additional 1 hour with the NBT and stimulant to show the very low residual level of ROS production within the parental CGD2 cell line. Active ROS production can be seen by the precipitations of dark ROS-mediated reduced NBT. Images were taken on an EVOS inverted microscope; scale bar 5 100 mm.
24 hours (Fig. 3A), and also on monocytes differentiated into macrophages for 7 days in M-CSF-containing medium (Fig. 3B). The NBT assay relies on the reduction of the soluble yellow NBT substrate into a colored precipitate by the action of ROS. Wild-type cells were able to generate ROS upon stimulation, whereas the negative control line, CGD1, was not (Figs. 3A and 3B). As expected, phagocytic cells from the CGD2 cell line appeared negative for ROS by the NBT assay. The mixed population of gene-edited cells, CGD2.GC16A, showed cells staining positive for ROS, and the single-cell clones (CGD2.GC16A.C4 and CGD2. GC16A.E4) derived from CGD2.GC16A all stained positive, showing highly effective phenotypic correction of the ROS defect in cells derived from the CGD patient. As noticed previously [12], a very low level of residual NADPH oxidase activity is present in CGD2 cells, which only becomes apparent upon extended incubation of the cells with the NBT reagent for an additional hour (Fig. 3C).
For a quantitative assessment of the restoration of ROS activity, monocytes and macrophages were assayed using a DHR assay. This fluorescence-based assay relies on ROS converting the nonfluorescent DHR reagent into fluorescent rhodamine123b, providing a more quantitative measure of ROS production. As with the NBT assay, CGD1 and CGD2 cells showed little to no ROS production, whereas WT cells showed a dramatic shift in fluorescence upon stimulation (Fig. 4). The corrected single-cell clones, CGD2.GC16A.C4 and CGD2.GC16A.E4, phenocopy the WT cells both as monocytes and macrophages, and, as expected, the mixed population CGD2.GC16A has interme-diate ROS levels. The lack of definition between positive and negative cells in CGD2.GC16A cells results from the very low levels of ROS activity in CGD2 cells (seen in both the NBT and DHR assays) and possible transfer of the ROS hydrogen peroxide from gene-edited cells to parental cells; addition of catalase to breakdown hydrogen peroxide minimizes, but does not completely remove, this effect [27].
Finally, to provide kinetic information about ROS generation, a real-time luminol assay was performed on the monocytes. Upon oxidation of the luminol reagent by ROS, light is released, which can be measured within a minute after addition of the cell stimulant PMA. As with the previous results, CGD1 and CGD2 showed no production of ROS upon stimulation, whereas clonal cells CGD2.GC16A.C4 and CGD2.GC16A.E4 showed similar quantities and kinetic properties to WT cells (max: WT 5 169.5, C4 5 112.4, E4 5 147.4; slope: WT 5 13.7, C4 5 12.96, E4 5 11.85; and time to reach 50% total signal: WT 5 62.7 sec, C4 5 59.6 sec, E4 5 53.6 sec; sigmoidal curve fit) after the addition of PMA, and the mixed population CGD2.GC16A had lower levels but similar kinetics (max 5 29.6, slope 5 13.3, time to reach 50% total signal 5 67.6 sec; Fig. 5).
Lack of ROS in CGD2 cells is due to exon skipping, resulting in the absence of a functional protein The cause of the ROS defect within CGD2 cells has not, to our knowledge, been previously investigated, so we took advantage of the isogenic properties of clones C4 and E4, which are genetically and phenotypically WT at the CYBB locus but within the genetic background of CGD2 cells, and we measured the levels of CYBB mRNA and protein in monocytes. Quantification of mRNA using primers located within exon 1 and exon 3 showed similar levels of CYBB mRNA across all cell lines, with CGD2 cells having a slight, yet significant, reduction (Fig. 6A). However, when using primers specific to the splice junctions of exons 1, 2, and 3, a more dramatic, hundredfold reduction in CYBB message in CGD2 cells was observed (Fig. 6A). These data indicate that, although CYBB mRNA is produced within CGD2 cells, it is incorrectly spliced. A splicing defect was confirmed by running the product of the exon 1 and exon 3 primers on a gel to observe the amplicon length compared with its predicted size (Fig. 6B). Correct splicing of the CYBB mRNA produces a product of 254 bp as observed (and confirmed by sequencing) in WT, CGD1, and the fully corrected CGD2.GC16A.C4 and CGD2.GC16A.E4 clones (Fig. 6C). In CGD2 cells, the WT product was not detected; instead, a smaller product of 158 bp was obtained, corresponding to an exon-2skipped variant, the identity of which was confirmed by sequencing (Fig. 6C). The mixed population CGD2.GC16A showed both the exon-2-skipped variant band as well as a faint WT band from the corrected mRNA. An additional larger species, detected in CGD2, more pronounced in CGD2.GC16A, but absent in all other reactions, was found to be an artefact of the PCR reaction of a mixed population (in CGD2, 1% of the CYBB mRNA are correctly spliced according to the qRT-PCR results with splicing-specific primers in Fig. 6A). The CYBB protein is an integral membrane protein with six transmembrane domains, and the deletion of exon 2 results in the removal of half of the first transmembrane domain and the neighboring loop (Fig. 7), which is likely to lead to an incorrectly folded protein susceptible to endoplasmic reticulum (ER)-associated degradation or a nonfunctional, possibly topologically-altered protein.

Discussion
Even in the era of potent antibiotics and fungicides, the diagnosis of CGD implies lifelong health complications and a reduced life expectancy [22]. In this article, we have demonstrated the ability of the CRISPR-Cas9 system to provide a footprintless strategy to correct a CGD-causing mutation at potentially therapeutically useful levels. This serves as a proof-of-principle in the development of genetically clean gene therapy approaches to cure this monogenic inherited disease.
We have shown here that the CRISPR-Cas9 system has the potential to provide highly efficient gene editing at the CYBB locus. This type of phenotypic correction is preferable to previous attempts to introduce the WT copy of the gene into iPS cells using lentiviral vectors [28] because it is completely clean; no residual foreign exogenous DNA remains to contaminate the genome and cause insertional mutagenesis, a hallmark problem with viral insertions. Moreover, the newly corrected gene is endogenously and, therefore, correctly controlled. This has been a major limitation to lentiviral introduction of exogenous transgenes, which either quickly become silenced or are not truly expressed in a cell-specific manner, despite numerous attempts to obtain specificity [29][30][31]. The limitation of nonendogenous control is also true of the recently published gene therapy strategies for CGD using zinc finger nucleases targeting the AAVS1 safe harbor site for transgene insertion in iPS cells, in which WT copies of the mutant genes are constitutively expressed and, therefore, not myeloid-specific [28,32,33]. Targeting the causative mutation within the endogenous locus using scarless, or foot-printless, gene editing is, therefore, the ideal gene therapy approach. Since high gene-targeting efficiencies are now possible with the help of site-specific nucleases, it just remains to transition this technology into clinically relevant cells for its potential to be fully realized. Although here we show correction of a single-point mutation that causes CGD, there are numerous mutations within the CYBB gene, as well as within the CYBA, NCF1, NCF2 and NCF4 genes, that are also responsible for CGD in different patients. Therefore, the strategy outlined here will require tailoring specific gRNA pairs and donor templates for each patient. It is worth noting that some genetic alterations within these genes may not lend themselves to CRISPR-Cas9 targeting (e.g., LINE1-retrotransposition) [34].
There are two methods that can be envisioned to exploit this technology therapeutically. The most tractable approach would be to convert the experimental protocol outlined in this article to hematopoietic stem cells (HSCs) isolated from bone marrow [35]. The technical limitations for such an approach would be the efficiency of gene editing in the HSCs and maintaining the bone-marrow-reconstituting potential of these fragile cells ex vivo throughout the transfection, expansion, and selection procedures. Alternatively, a technological advancement could allow for the creation of authentic bone-marrow-repopulating HSCs from iPS cells [36][37][38]. This would complete the circle of personalized medicine: patient-derived iPS cells for gene editing, expansion, and selection, differentiated into HSCs for reintroduction into the patient to repopulate the hematopoietic system with disease-free cells. Such a procedure, although theoretically possible, currently has multiple practical, safety, and ethical issues. Most importantly, the karyotypic stability of the cells would need to be very closely monitored. This becomes acutely obvious as more publications demonstrate , and the two gene-edited single-cell clones from CGD2.GC16A (C4 and E4) were measured by qRT-PCR with two primer sets: one pair in which the primers bind the spliced junctions of exon 1-2 and exon 2-3 (Exon 2) and one splicing independent pair that binds exon 1 and exon 3 (Exon 1-3). Results are calculated relative to b-actin internal control primer pair, normalized to the WT cell line, and represent the average 6 SEM of three independent cell harvests. Statistical significance was calculated by two-way analysis of variance with Sidak's multiple comparison test. ***p ! 0.001; ****p ! 0.0001. (B) Schematic representation of the CYBB mRNA pre-and postsplicing events producing either the correct splicing pattern (WT) or potential exon-2-skipped variant (Exon2 -); the sizes of amplicon expected with primers binding exon 1 and exon 3 are shown. (C) qRT-PCR products using primers in exon 1 and exon 3 on cDNA from monocytes were separated on an agarose gel. Major bands corresponding to correctly spliced (WT) and the exon-skipped variant (Exon2 -) are marked with an arrow. Larger bands in CGD2 and CGD2.GC16A (marked with an asterisk) were PCR artefacts. Figure 7. Location of exon 2 of CYBB when inserted into the membrane. Schematic representation of the CYBB protein within a membrane; the region encoded by exon 2 that would be lost due to exon skipping is highlighted in black. Adapted from Marques et al. [49], based on Taylor et al. [50].
the potential for stem cells in culture to accumulate mutations and karyotypic abnormalities over time [39][40][41][42][43]. Indeed, although the two single-cell clones isolated in this study were grossly karyotypically normal at the resolution of single nucleotide polymorphism (SNP) densities, this does not provide genome-wide sequencing levels of coverage [40,44], and a third clone from CGD2.GC16A that was karyotyped during the course of this study had an isochromosome 12p, a chromosome that has previously been seen to enhance the proliferation of stem cells in culture when duplicated [43]. Additionally, off-target effects of the CRISPR-Cas9 system are a potential issue, particularly if the cells are intended for clinical use. Although Cas9 uses complementarity between the gRNA and the target DNA to determine cleavage site selection, specificity is not absolute [45,46]. It was for this reason that we opted to use the nicking version of Cas9, which minimizes this issue [25,47]; however, clinical use may still require full genome sequencing to ensure the genetic integrity of the cells after gene correction.
Although neither of the potential clinical approaches outlined above is currently practicable, it is worth noting that site-specific nucleases have already been used ex vivo to modify T cells in clinical trials for human immunodeficiency virus (HIV), with no negative side effects [48]. Thus, further work is merited to transfer this technology into primary hematopoietic stem cells. Finally, a greater understanding of human hematopoiesis is necessary to generate repopulating HSCs ex vivo and thereby make personalized gene therapy a reality.