Therapeutic gene editing in CD34+ hematopoietic progenitors from Fanconi anemia patients

Abstract Gene targeting constitutes a new step in the development of gene therapy for inherited diseases. Although previous studies have shown the feasibility of editing fibroblasts from Fanconi anemia (FA) patients, here we aimed at conducting therapeutic gene editing in clinically relevant cells, such as hematopoietic stem cells (HSCs). In our first experiments, we showed that zinc finger nuclease (ZFN)‐mediated insertion of a non‐therapeutic EGFP‐reporter donor in the AAVS1 “safe harbor” locus of FA‐A lymphoblastic cell lines (LCLs), indicating that FANCA is not essential for the editing of human cells. When the same approach was conducted with therapeutic FANCA donors, an efficient phenotypic correction of FA‐A LCLs was obtained. Using primary cord blood CD34+ cells from healthy donors, gene targeting was confirmed not only in in vitro cultured cells, but also in hematopoietic precursors responsible for the repopulation of primary and secondary immunodeficient mice. Moreover, when similar experiments were conducted with mobilized peripheral blood CD34+ cells from FA‐A patients, we could demonstrate for the first time that gene targeting in primary hematopoietic precursors from FA patients is feasible and compatible with the phenotypic correction of these clinically relevant cells.


Introduction
Gene editing has recently emerged as a realistic approach for the treatment of both inherited and acquired diseases (Maeder & Gersbach, 2016). Although studies conducted in different cell types, such as fibroblasts or embryonic stem cells, have shown relatively high efficiencies of homology directed repair (HDR) (Reviewed in Urnov et al, 2010), gene editing in hematopoietic stem and progenitor cells (HSPCs) has presented particular difficulties, as compared to other target cells (Lombardo et al, 2007). Significantly, in 2014 Genovese and colleagues  showed significant improvements in the gene editing of human long-term hematopoietic SCID-repopulating cells by tailoring delivery platforms and culture conditions.
In spite of advances in the editing of adult HSPCs Hoban et al, 2015;De Ravin et al, 2016), gene-targeting efficiencies currently obtained in these rare precursor cells are still far from transduction rates already achieved with gamma-retroviral and lentiviral vectors (Cartier et al, 2009;Cavazzana-Calvo et al, 2010;Aiuti et al, 2013;Biffi et al, 2013;Hacein-Bey-Abina et al, 2014) Consequently, clinical applications associated with the positive selection of gene-edited cells are particularly attractive. In this respect, either the specific knockout of the CCR5 receptor of HSPCs from HIV-infected patients (Tebas et al, 2014) or the specific restoration of the IL2RG in HSPCs from X-linked severe combined immunodeficient (SCID-X1) patients  has recently been proposed for clinical application.
In none of the gene-editing studies reported to date is a selection advantage expected in the self-renewing HSCs Hoban et al, 2015;De Ravin et al, 2016). In the current study, we therefore propose a gene-editing approach for the treatment of the bone marrow failure (BMF) that takes place in virtually all Fanconi anemia (FA) patients (Butturini et al, 1994). While the genetic manipulation of HSPCs from FA patients needs to overcome the increased apoptotic predisposition and high sensitivity of these cells to in vitro culture, the selective advantage observed in reverted HSPCs from FA mosaic patients (Waisfisz et al, 1999;Gregory et al, 2001;Gross et al, 2002) and also in induced pluripotent stem cells (iPSCs) corrected by lentiviral vectors (Raya et al, 2009) suggests that the infusion of a very limited number of gene-edited FA HSCs might be sufficient to progressively repopulate the whole hematopoiesis of these patients.
Since FA can be caused by many different mutations in any of the 22 FANC genes discovered so far (Bogliolo & Surralles, 2015;Bluteau et al, 2016;Park et al, 2016;Knies et al, 2017), we decided to develop a versatile gene therapy platform based on the specific integration of a therapeutic transgene into the adeno-associated virus integration site 1 (AAVS1) "safe harbor" locus (Lombardo et al, 2011) using zinc finger nucleases (ZFNs). Although we initially focused our aims in the specific AAVS1 targeting of the FANCA gene, whose mutations account for about 60% of the FA patients (Casado et al, 2007), this approach could be easily applied to the delivery of all the other FANC genes.
In a previous study, we used the same strategy for the integration of FANCA into patient-derived fibroblasts that were subsequently reprogrammed and differentiated in vitro to generate disease-free FA hematopoietic progenitors (Rio et al, 2014). Due to current limitations for the use of iPSC-derived HSCs in human hematopoietic therapies (See review in Daniel et al, 2016), we have now focused our efforts on the gene editing of primary HSPCs directly obtained from FA patients. Our results demonstrate for the first time the feasibility of conducting therapeutic gene editing in primary CD34 + cells from FA-A patients.

Efficient gene targeting of reporter genes in lymphoblastic cell lines from healthy donors and FA patients
Previously we showed the feasibility of targeting a therapeutic FANCA expression cassette in the AAVS1 site of FA-A fibroblasts (Rio et al, 2014). Due to the fact that in that study a transient expression of FANCA from the non-integrated donor might have facilitated the HDR-mediated insertion of FANCA in FA-A cellswith reported HDR defects (Nakanishi et al, 2005)-in the current study we first investigated the possibility of targeting a non-therapeutic donor (PGK-EGFP) in FANCA-deficient cells, as compared to FANCA-proficient cells. Data in Fig 1A show that the ZFN-mediated targeting of the PGK-EGFP donor was feasible in FA-A LCLs. Although the overall proportion of EGFP + cells obtained after gene editing of FA-A LCLs (mean value: 1.61 AE 1.51%) was apparently lower, either with respect to their complemented FA-A LCLs counterparts (2.53 AE 1.97%; P = 0.02) or HD LCLs (CP1, CP2, C3, and C4; 4.07 AE 3.89%; P = 0.34), no significant differences could be observed among the different experimental groups.
In-out PCR analysis of the 5 0 and the 3 0 ends of the target AAVS1 site confirmed in all instances the integration of the transgene in the AAVS1 locus (Fig 1B), supporting the fact that the proportion of EGFP + cells actually corresponds to the efficiency of gene editing of these cells. Moreover, the sequencing of PCR products evidenced the proper vector-to-genome junctions not only in HD LCLs, but also in FA LCLs, demonstrating the precise HDR-mediated integration of the donor construct in the AAVS1 locus, both in HD and in FA LCLs ( Fig 1C).
Taken together the results obtained in this first set of experiments demonstrate that FANCA is not essential for the precise HDRmediated insertion of a transgene in the genome of human FA-A cells, although we cannot discard the hypothesis that it may enhance the efficiency of this process.
Restored FA pathway in FA-A lymphoblastic cell lines after FANCA insertion in the AAVS1 site To investigate whether the insertion of the FANCA expression cassette in the AAVS1 locus of FA-A LCLs conferred a phenotypic correction in these cells, two therapeutic IDLV donors were used (Fig 2A). In both vectors, the therapeutic FANCA gene was placed under the transcriptional control of the human PGK promoter. In the PGK-FANCA/Puro R construct, a downstream 2A Puro R gene was used, while in the EGFP/PGK-FANCA construct, a promoterless EGFP cDNA was preceded by a splice acceptor (SA) site and a selfcleaving 2A peptide, to facilitate the identification of gene-edited cells (Rio et al, 2014).
When FA-A LCLs were treated both with the ZFNs and the PGK-FANCA/Puro R IDLV, a significant decrease in the hypersensitivity of these FA cells to MMC was observed, showing the complementation by gene editing of these cells (Fig 2B). Consistent with this observation, these ZFN-mediated edited cells were also more resistant to puromycin as compared to non-edited cells (survivals to 0.5 lg/ml puromycin in edited and non-edited FA-56 LCLs cells were 7.52 AE 0.32% and 14.80 AE 1.16%, respectively). Although the ZFN-mediated editing of FA-A LCLs with the EGFP/PGK-FANCA donor did not show the expected expression of EGFP (data not shown), these edited cells also showed the correction of their hypersensitivity to MMC ( Fig 2B).
As the generation of oxidative reactive species (ROS) is increased in FA cells (Joenje et al, 1981), we tested differences in ROS levels both in edited and not edited FA LCLs. As shown in Fig EV1, a significant reduction of ROS was observed when FA-55 LCLs were treated both with the ZFN and the PGK-FANCA/Puro R LV, as compared with the parental cells, or cells only transduced with the therapeutic donor.
These results demonstrate that the targeting of therapeutic FANCA donors in the AAVS1 site of FA-A LCLs mediates the correction of two characteristic phenotypes of FA cells, their hypersensitivity to DNA cross-linking drugs and the spontaneous production of ROS.
Since FANCA is essential for FANCD2 foci formation in DNA damaged sites, the percentage of cells with FANCD2 foci was also analyzed in HD LCLs, as well as in FA LCLs (either untreated or treated with MMC). These cells were analyzed in the absence of any treatment, after treatment with the therapeutic donor IDLV, or after treatment with the therapeutic donor and the ZFNs. In all instances, quantification of FANCD2 foci was determined in MMC-untreated and in MMC-treated samples. While a very low number of cells with FANCD2 foci were observed in FA LCLs, either untreated or only treated with the therapeutic donor, gene-edited FA LCLs efficiently  TGGCTCTGGTTCTGGGTACTTTTATCTGTCCCCTCCACCCCACAGTGGGGCAAGCTTCCACGGGGTTGGGGTTGCGCCTTTTCC  TGGCTCTGGTTCTGGGTACTTTTATCTGTCCCCTCCACCCCACAGTGGGGCAAGCTTCCACGGGGTTGGGGTTGCGCCTTTTCC  TGGCTCTGGTTCTGGGTACTTTTATCTGTCCCCTCCACCCCACAGTGGGGCAAGCTTCCACGGGGTTGGGGTTGCGCCTTTTCC  TGGCTCTGGTTCTGGGTACTTTTATCTGTCCCCTCCACCCCACAGTGGGGCAAGCTTCCACGGGGTTGGGGTTGCGCCTTTTCC  TGGCTCTGGTTCTGGGTACTTTTATCTGTCCCCTCCACCCCACAGTGGGGCAAGCTTCCACGGGGTTGGGGTTGCGCCTTTTCC  TGGCTCTGGTTCTGGGTACTTTTATCTGTCCCCTCCACCCCACAGTGGGGCAAGCTTCCACGGGGTTGGGGTTGCGCCTTTTCC  TGGCTCTGGTTCTGGGTACTTTTATCTGTCCCCTCCACCCCACAGTGGGGCAAGCTTCCACGGGGTTGGGGTTGCGCCTTTTCC  TGGCTCTGGTTCTGGGTACTTTTATCTGTCCCCTCCACCCCACAGTGGGGCAAGCTTCCACGGGGTTGGGGTTGCGCCTTTTCC   CP1  CP2  C3  C4  FA-55  FA-

EGFP + cells (%)
-   Begoña Diez et al Gene editing in CD34 + cells from FA-A patients EMBO Molecular Medicine restored the formation of FANCD2 foci, mimicking the behavior of HD LCLs ( Figure 2C). As observed in Fig 1A, in-out PCR analyses confirmed the integration of the therapeutic donors, either the PGK-FANCA/Puro R (Fig 2D) or the EGFP/PGK-FANCA ( Fig 2E) donors, in the AAVS1 locus of FA-A cells.
These results demonstrate that the targeted insertion of FANCA in the AAVS1 locus of FA-A LCLs efficiently corrects characteristic phenotypes of these cells.
Efficient and safe targeting in the AAVS1 locus of healthy donor cord blood CD34 + cells To achieve targeted integration into the AAVS1 site of human HSPCs, cord blood (CB) CD34 + cells were investigated first. Cells were prestimulated for 48 h and then transduced with the IDLV donor. Twenty-four hours later, samples were electroporated with the AAVS1-ZFN mRNAs (doses were optimized according to prior analyses; see Fig EV2). To monitor edited cells both in vitro and after transplantation into immunodeficient mice, targeting experiments were first performed with the PGK-EGFP donor. Flow cytometry analyses conducted 3 and 10 days after electroporation showed similar values of EGFP + cells (around 14%) in CD34 + cells treated both with the donor and the ZFNs. In contrast to this data, almost no EGFP + cells were observed (< 0.6%) when cells were only treated with the PGK-EGFP donor ( Fig 3A; representative dot plots are shown in Fig EV3). Analysis of EGFP expression within different CD34 + subpopulations showed the presence of EGFP + cells in all cell types, including the primitive CD34 + / CD133 + /CD90 + precursors. An average of 6.9 AE 4.7% of these cells was EGFP + after 10 days of culture ( Fig 3B). Additionally, clonogenic assays showed that 14.1 AE 7.6% of the colonies were EGFP + . In-out PCR analyses confirmed the integration of the donor cassette in the AAVS1 locus in 90.5% of these colonies. When CB CD34 + cells were only transduced with the PGK-EGFP IDLV, neither EGFP-fluorescent colonies nor integrations in the AAVS1 locus were observed in any of the 50 colonies analyzed corresponding to five independent experiments (see representative in-out PCR in Fig 3C). Taken together, these experiments demonstrate the specific targeting of the PGK-EGFP donor construct in primary human CD34 + cells.
Although AAVS1-ZFNs used in these experiments have been partially modified to improve their specificity (Lombardo et al, 2011) as compared to the ones previously described (Mussolino et al, 2014), the target sequence was maintained. Therefore, the same five potential off-targets previously studied by Mussolino et al (Mussolino et al, 2014), namely ATRNL1, BEGAIN, CHRAC1, H19, and LINC00548 were analyzed by deep sequencing. As shown in Table 1, low off-target activity was detected in three of these target loci: LINC00548, H19 and BEGAIN (0.52, 0.17 and 0.02%, respectively), whereas no off-target activity was observed in ATRNL1 and CHRAC1 (the detection limit of these analyses was 0.01%). The frequencies of detected INDELs were extremely low as compared to the intended AAVS1 target site, which showed an INDEL rate of 57.4% (Fig EV2 and Table 1). Moreover, both the BEGAIN and LINC00548 off-target sites occurred in the middle of an intron, whereas the H19 occurred in an intergenic region. All these observations suggest that no deleterious effects are expected as a consequence of the off-target activity of these AAVS1-ZFNs in human CD34 + cells.

Targeting of reporter and therapeutic genes in the AAVS1 locus of healthy donor hematopoietic stem cells
To study the repopulating ability of CD34 + cells edited with the PGK-EGFP donor, 3-5 × 10 5 CD34 + cells treated as in experiments of Fig 3 were transplanted into NSG (black squares) and also into NSG-SGM3 mice (black circles), that facilitate a preferential reconstitution with myeloid cells (Billerbeck et al, 2011;Miller et al, 2013), which are markedly affected in most FA patients. In these experiments, the efficiency of gene editing determined after 10 days of liquid culture was 11.13 AE 4.87% (not shown). As observed in experiments corresponding to Fig 3C, all EGFP + colonies generated in methylcellulose showed the expected integration in the AAVS1 site ( Fig 4A).
After transplant, the engraftment of human cells was longitudinally monitored by measuring the percentage of hCD45 + cells in serial bone marrow biopsies. Similarly, the percentage of gene editing was determined by analyzing the proportion of hCD45 + cells that were positive for EGFP expression (Fig 4B). The level of human hematopoietic engraftment observed in NSG-SGM3 mice was higher at the earliest time of analysis, 30 days post-transplantation (median value 37.0% of hCD45 + cells), while maximum levels of engraftment in conventional NSG recipients were observed at the last time point of analysis (median value of 69.8% of hCD45 + cells at 90 days post-transplantation; Fig 4C). Analyses of EGFP expression in engrafted human CD45 + cells showed evident, although variable, levels of EGFP expression along the post-transplantation period ( Fig 4D). Overall, a similar proportion of EGFP + cells was observed in NSG-SGM3 and NSG mice at 90 days post-transplantation (median values of 11.9% and 9.8%, respectively, P = 0.66). Interestingly, these results were similar to values obtained in cells that were maintained for only 10 days in culture, suggesting a similar gene-editing efficiency in hematopoietic progenitor cells with respect to more primitive hematopoietic precursors with in vivo repopulation capacity.
Consistent with previous studies (Billerbeck et al, 2011;Miller et al, 2013), NSG-SGM3 mice facilitated the generation of human myeloid cells at the expense of reducing B-and T-cell reconstitution, while human B cells were predominant in NSG mice (Fig 4E). Regardless of the proportion of CD3 + , CD19 + , CD33 + , and CD34 + cells found in recipients' BM and spleen, a significant number of EGFP + cells were found in all lineages, including CD34 + cells (see Figs 4F and EV4).
In-out PCRs conducted with total BM cells from six mice transplanted with EGFP-edited CD34 + cells showed the expected integration of the EGFP cassette in the AAVS1 site ( Fig 4G). To further confirm the targeted insertion in human CFCs generated in the hematopoietic tissues of transplanted mice, BM cells from these animals were seeded in methylcellulose cultures that specifically allowed the growth of human colonies. Once again, in-out PCR analyses conducted in human BM-derived EGFP + colonies confirmed the integration of the donor construct in the AAVS1 locus of engrafted human progenitor cells (Fig 4H).
To confirm that gene editing was targeting long term hematopoietic stem cells (LT-HSCs), BM cells from primary recipients, either A Gene-targeting efficiency in HD CD34 + cells measured by flow cytometry 3 and 10 days after gene editing. The percentage of EGFP + cells is shown. Bars represent mean AE SD, n = 8. P-values were calculated by two-tailed paired t-test. B Gene-targeting efficiency in the different subpopulations of hematopoietic progenitors (differentiated: CD34 À , committed progenitors: CD34 + , early progenitors: CD34 + CD133 + and primitive cells: CD34 + CD133 + CD90 + ). The percentage of EGFP + cells was measured by flow cytometry at day 10 after gene editing. Bars represent mean AE SD, n = 8; ANOVA with Bonferroni's multiple comparison test between subpopulations inside the same group (+ZFN P = 0.0002; **P ≤ 0.01; ***P ≤ 0.001) and two-tailed paired t-test between different groups (+ZFNs versus ÀZFNs) were used. C Integration analysis in hematopoietic colonies obtained from CD34 + cells treated with the EGFP donor IDLV together with ZFNs (EGFP + colonies), or with the EGFP donor IDLV alone. Upper panel: Schematic representation of the PGK-EGFP cassette integrated in the AAVS1 locus with 5 0 and 3 0 primers used for in-out PCR. Bottom panel: PCR for the 5 0 and 3 0 junction. Sizes of the expected bands are indicated. À: negative control.
Source data are available online for this figure. NSG or NSG-SGM3 mice, were re-transplanted into secondary NSG recipients. Bone marrow cells from primary NSG but not from NSG-SGM3 mice repopulated secondary recipients, suggesting a stressed differentiation of human HSCs in NSG-SGM3 mice ( Fig EV5A). The analysis of human hematopoiesis in secondary recipients showed the presence of myeloid, B lymphocytes, and T lymphocytes. Significantly, the presence of EGFP + cells was also detected in human engrafted cells, although at lower levels, around 1% (Fig EV5B-D), with respect to values observed in primary recipients. Taken together, the in vivo experiments conducted with HD CB CD34 + cells that were edited with a non-therapeutic donor demonstrated the targeting of primitive human HSCs characterized by long-term multipotent in vivo repopulating properties.
In the next set of experiments, we moved on to investigate the efficacy of gene editing, now using a therapeutic donor: the EGFP/ PGK-FANCA donor. Consistent with data obtained in FA LCLs, no EGFP-positive cells were detected as early as 10 days after the in vitro culture of cells edited with this donor, in which EGFP expression is driven from the regulatory signals present in the AAVS1 locus. This contrasts with the significant number of CD34 + cells that were positive for EGFP expression after gene editing with the PGK-EGFP donor (Fig EV6A). Because in-out PCR analyses conducted in hematopoietic colonies evidenced the integration of the EGFP/PGK-FANCA donor in the AAVS1 locus (Fig EV6B), these results suggest that the endogenous regulatory sequences of the AAVS1 locus are not sufficiently active to promote detectable EGFP expression in human hematopoietic cells.
To investigate the repopulating ability of HD CD34 + cells edited with this therapeutic cassette, samples were transplanted into NSG-SGM3 mice. As observed after gene editing with the EGFP donor, analysis of the hematopoietic organs of transplanted mice confirmed a multilineage human hematopoietic engraftment of these animals (Fig EV6C-E). qPCR analyses confirmed the integration of the EGFP/PGK-FANCA cassette both in the BM and in the spleen of these mice (data not shown), demonstrating the feasibility of targeting a therapeutic FA donor in the AAVS1 site of HD HSPCs characterized by in vivo repopulating properties.

Targeted mediated gene therapy in CD34 + cells from FA-A patients
Having demonstrated the specific insertion of marker and FA therapeutic cassettes in the AAVS1 locus of primary HD HSPCs, we investigated the feasibility of correcting the phenotype of CD34 + cells from FA-A patients using the same gene-editing approach. To achieve this aim, small aliquots of mobilized peripheral blood (mPB) CD34 + cells from five FA-A patients were treated with the therapeutic donors and the AAVS1 ZFNs, as was done in experiments with HD CB CD34 + cells.
To test the phenotypic correction of FA CD34 + cells after gene targeting, cells were plated in methylcellulose both in the absence and in the presence of MMC, and colonies were scored after 14 days of culture. Mean survivals to MMC corresponding to edited CD34 + cells were 29.91 AE 24.63%, which were significantly increased with respect to the MMC resistance observed in cells only transduced with the therapeutic donor (mean value: 2.98 AE 4.46%; P = 0.016; Fig 5A). These results suggest the phenotypic correction of FA-A primary CD34 + cells by gene editing. To confirm that the increased number of MMC-resistant colonies was due to the targeted integration of the FANCA cassette in the AAVS1 locus, in-out PCRs from a pool of hematopoietic colonies grown in the absence and presence of MMC were performed in samples from two FA-A patients. In the case of patient FA-655, a specific integration of the therapeutic cassette was observed in MMC-selected samples, but not in unselected cells. Notably, in patient FA-712 a faint band could be detected even in the absence of MMC-selected cells, although a marked increase in the intensity of this band was detected after MMC selection. The analysis of individual colonies picked from patient FA-655 showed that 10% of colonies grown in the absence of MMC were positive for the integration of FANCA in the AAVS1 site (See representative in-out PCR in Fig 5C). This final experiment demonstrates for the first time the feasibility of correcting the disease phenotype of primary CD34 + cells from FA-A patients by targeted gene therapy.

Discussion
The development of engineered nucleases and donor constructs capable of inserting DNA sequences in specific sites of the cell genome is opening up new possibilities for therapeutic gene editing of inherited and acquired diseases (reviewed in Corrigan-Curay et al, 2015;Hoban & Bauer, 2016;Osborn et al, 2016). The efficacy and specificity of ZFNs designed for the targeting of the AAVS1 "safe harbor" locus, together with the high number of private mutations characterized in FA-A patients, suggest that the specific insertion of FANCA in the AAVS1 site of FA-A patient's    HSPCs would result in a particularly useful platform for restoring normal hematopoiesis in these patients. Although recent results have shown the efficiency of AAVs to deliver donor cassettes in HSPCs (Wang et al, 2015), the large packaging capacity of IDLVs and the high transduction efficacy of these vectors on human HSPCs (Lombardo et al, 2007) supported the use of IDLVs in our gene-editing studies to facilitate the delivery of the therapeutic FANCA gene together with additional marker genes in these cells.

PGK-FANCA/Puro R + ZFNs PGK-FANCA/Puro R + ZFNs
Previous studies have demonstrated that gene editing is feasible in CD34 + cells from patients with hematological disorders such as SCID-X1  or chronic granulomatous disease (CGD) (De Ravin et al, 2016). Nevertheless, the editing of CD34 + cells from FA patients, where the number and function of these cells is highly compromised (Ceccaldi et al, 2012), has not been attempted so far. Additionally, gene editing in FA cells is challenging owing to previous observations showing the involvement of the FA pathway in HDR (Nakanishi et al, 2005). Despite these limitations, the proliferation advantage expected as a result of the therapeutic editing of FA HSPCs may constitute a characteristic advantage of FA compared to many other inherited disorders in which the HSPCs do not reveal any phenotypic defect.
Although our previous study demonstrated the possibility of conducting gene editing in FA-A fibroblasts, this could be a consequence of the potential transient expression of FANCA driven by the therapeutic donor prior to integration in the cell genome (Rio et al, 2014). Our results in this new study confirm in LCLs from four different FA-A patients the feasibility of conducting gene editing, even with a non-therapeutic donor that only carries a reporter EGFP transgene, demonstrating for the first time that FANCA is not essential for the HDR-mediated editing of the human genome. Nevertheless, our data also showed that the efficacy of gene editing in not-complemented FA-A cells was two-to threefold lower as compared to either their complemented cell counterparts or to HD LCLs. Although these differences did not reach statistical significance, our results are consistent with previous data obtained in different experimental models showing that, in contrast to other FA proteins such as BRCA2, FANCA is only partially involved in HDR (Nakanishi et al, 2005(Nakanishi et al, , 2011. Significantly, when donors carrying the therapeutic FANCA gene (PGK-FANCA/Puro R or EGFP/PGK-FANCA-IDLVs) were used for the editing of FA-A LCLs, an efficient phenotypic correction of these cells was deduced from their decreased MMC-hypersensitivity and production of ROS, and also their restored ability to translocate FANCD2 to DNA damage repair foci. In spite of the phenotypic correction of FA cells edited with the EGFP/PGK-FANCA donor, these cells did not show the expected EGFP-expression, suggesting the low activity of the AAVS1 regulatory sequences in FA LCLs, which contrasts with our results in FA-A fibroblasts edited with the same construct (Rio et al, 2014). This observation reinforced the relevance of including an exogenous promoter in therapeutic donors to be inserted in this locus, at least in FA hematopoietic cells. In the other therapeutic donor (the PGK-FANCA/Puro R -IDLV), the Puro R gene was included in the cassette to facilitate the selection of edited cells. Although puromycin resistance could be confirmed in edited FA cells, because of the inherent hypersensitivity of parental FA cells to MMC, edited FA cells were very efficiently selected with this drug and thus was routinely used in our experiments.
In our studies conducted with primary HD CD34 + cells, geneediting efficacies around 14% were achieved using AAVS1 mRNA ZFNs and EGFP IDLV donors. Moreover, off-target analyses confirmed the high specificity of the AAVS1-ZFNs in these cells, as demonstrated by the 100-fold increase in the targeting of the AAVS1 site, compared to the off-target sites. Analyses of mice transplanted with gene-edited CD34 + cells also demonstrated the in vivo repopulating capacity of these samples. In this respect, while rapid engraftments were noted in NSG-SGM3 mice, higher long-term engraftment levels were generally achieved in NSG recipients. Remarkably, regardless of the strain of immunodeficient mouse used in our experiments, most of the primary transplanted recipients showed the presence of human hematopoietic cells that expressed significant levels of the EGFP transgene, both in myeloid and T and B lymphoid lineages, and even in CD34 + cells. Additionally, the expected insertion of the transgene in the AAVS1 locus of these cells was demonstrated in our experiments. Interestingly, secondary transplants in NSG mice demonstrated the long-term repopulating capacity of gene-edited cells. Nevertheless, consistent with previous studies , a decreased efficacy of gene editing was observed in the very primitive HSCs capable of repopulating secondary recipients.
In our final set of experiments, we demonstrated for the first time the gene editing of mobilized PB CD34 + cells directly obtained from FA-A patients, the target population currently considered for the gene therapy of several inherited diseases, including FA (see review in Tolar et al, 2012). Moreover, our data showed that the integration of the therapeutic transgene in the AAVS1 site of FA-A CD34 + cells resulted in the correction of the characteristic MMC-hypersensitivity of these cells, indicating the possibility of conducting therapeutic gene editing in CD34 + cells from FA-A patients.
Although efficacies of gene targeting achieved in FA CD34 + cells are still significantly lower compared to the transduction efficacy achieved by lentiviral vectors (Jacome et al, 2009;Becker et al, 2010;Gonzalez-Murillo et al, 2010), our results suggest that gene editing could be applicable to BMF syndromes such as FA, in which a selection advantage of edited cells could take place, as we already demonstrated in primary fibroblasts and iPSCs from FA patients (Rio et al, 2014).
The possibility of harvesting significant numbers of HSPCs from FA-A patients with improved mobilizing drugs (See review in Adair et al, 2016) together with the development of efficient gene-editing tools (Tebas et al, 2014) suggests that gene editing may constitute a realistic approach for the treatment of FA in the future. Indeed, the use of drugs such as aryl hydrocarbon receptor antagonist (StemRegenin, SR1) or 16,16-dimethyl-prostaglandin E2 (dmPGE2) capable of improving the engraftment of gene-edited HSCs in transplanted recipients , or molecules capable of enhancing the efficacy of HDR (Zhang et al, 2016) may further aid in the development of therapeutic gene editing for the treatment of inherited diseases affecting the hematopoietic system, including FA.

Cell lines
293T (ATCC: CRL-11268) were used for the production and titration of the integrase-defective lentiviral vectors (IDLVs). Cells were grown in Iscove's modified Dulbecco's medium GlutaMAX TM (IMDM; Gibco) supplemented with 10% Hyclone (GE Healthcare) and 1% penicillin/streptomycin solution (Gibco). Lymphoblastic cell lines (LCLs) were generated by Epstein-Barr virus transformation of peripheral blood B cells from healthy donors or FA patients. FA-A patients were classified by retroviral cell complementation as previously described (Casado et al, 2007). FA-A patients and healthy donors were encoded to protect their confidentiality and informed consent was obtained in all cases. Mutations described in FA-A LCLs (Castella et al, 2011) have been included in Table EV1.

Hematopoietic progenitor cells from healthy donors and FA-A patients
Healthy donor CD34 + cells were provided by the Centro de Transfusiones de la Comunidad de Madrid from umbilical cord blood samples (CB) scheduled for discard, after written informed consent from the mothers. Mononuclear cells from pooled CBs were obtained by fractionation in Ficoll-Hypaque according to manufacturer's instructions (GE Healthcare). Purified CD34 + cells were obtained using the MACS CD34 Micro-Bead kit (Miltenyi Biotec). Cells were grown in StemSpam (StemCell Technologies) supplemented with 1% GlutaMAX TM (Gibco), 1% penicillin/streptomycin solution (Gibco), 100 ng/ml SCF and Flt3-ligand, and 20 ng/ml TPO and IL6 (all from EuroBiosciences).

Gene-targeting experiments
Lymphoblastic cell lines from healthy donors or FA-A patients were transduced with the corresponding IDLV donors [multiplicity of infection (MOI): 50 transducing units/cell (TU/cell)] in 96well culture plates in a final volume of 100 ll. One day after transduction, cells were collected, washed, and nucleofected with 1.5-6 lg of total in vitro synthetized mRNA (0.75-3 lg of each ZFN pair), using the SF Cell Line 4D-Nucleofector X Kit (Lonza).
Purified CD34 + cells from healthy donor CBs or from FA-A patients were prestimulated for 48 or 24 h, respectively, as described above, at a density of 10 6 cells/ml. Thereafter, 10 5 cells were transduced with the corresponding IDLV donor at a MOI of 100 TU/cell in a 96-well culture plate in a final volume of 100 ll. One day after transduction, cells were collected, washed, and nucleofected with 6 lg of total in vitro synthetized mRNA (3 lg of each ZFN pair) using the P3 Primary Cell 4D-Nucleofector X Kit (Lonza). One day after nucleofection, gene-edited cells were assessed either in clonogenic assays or in transplantation experiments.
In some experiments, cells nucleofected with the PGK-FANCA/ Puro R donor IDLV were selected with 0.5 lg/ml of puromycin for 48 h. Cell viability was analyzed by flow cytometry as previously established for MMC survival assay.
Cell sorting was also used to select gene-edited LCLs by the selection of EGFP + cells in a BD Influx cell sorter (BD Biosciences).

Surveyor nuclease assay
DNA was extracted using NucleoSpin â Tissue kit (Macherey-Nagel). Then, a PCR was performed to amplify the region of the AAVS1 locus in which the ZFNs cut using the primers: NHEJ Fw (5 0 CTTCAGGACAGCATGTTTGC 3 0 ) and NHEJ Rv (5 0 ACAG-GAGGTGGGGGTTAGAC 3 0 ). The PCR was performed as follows: 10 min 95°C, 40 cycles of 45 s at 95°C, 45 s at 62°C, and 45 s at 72°C, and finally one step for 10 min at 72°C. The 224-bp PCR product was then dehybridized and rehybridized in order to obtain the heteroduplexes. If an INDEL (integration or deletion) has occurred, the corresponding hairpin can be recognized by the Surveyor â nuclease (Survor â mutation detection kit; IDT), generating a band pattern that was visualized on 10% TBE gels 1.0 mm (Invitrogen) and analyzed using ImageJ software. Percentages of cleavage were determined using the equation: In-out PCR For PCR analysis, genomic DNA from bulk populations was extracted using the NucleoSpin Tissue kit (Macherey-Nagel). Single colonies were pelleted and resuspended in 10 ll of PBS, and genomic DNA was extracted as previously described (Charrier et al, 2011).
To detect the targeted integration of the HDR cassette in the AAVS1 locus, two different pairs of primers for the 3 0 or the 5 0 integration junction were used (Table EV2).

Deep sequencing of putative AAVS1 off-targets
Among the different AAVS1 off-targets that were identified by Mussolino et al (2014) using the bioinformatic tool PROGNOS, we studied five that have also been experimentally validated: ATRNL1, BEGAIN, CHRAC1, H19, and LINC00548.
Genomic DNA from CD34 + cells targeted with the PGK-EGFP IDLV, together with AAVS1-ZFNs, was extracted using the NucleoSpin â Tissue kit, and the putative off-target sites were amplified by PCR, generating amplicons of 270 bp surrounding the potential ZFN binding site (Table EV3).
PCR products were purified using the AxyPrep PCR Clean-Up (Axygen), quantified using a Qubit fluorometer (Thermo Fisher Scientific), and used for library construction using the KAPA Library preparation Kit (Kapa Biosystems) for Ion Torrent platforms. The generated DNA fragments (DNA libraries) were sequenced on the Ion Torrent PGM platform, using 400-bp single-end sequencing reads.

Analysis of AAVS1 off-target sites
The analysis of the raw sequence data obtained by Ion Torrent PGM platform sequencing was carried out using CLC Genomics Workbench 9.0.1, and those high-quality sequencing reads were mapped against the reference sequences using a length and a similarity fraction of 0.80.
A low-frequency variant calling was performed, obtaining a list of variants with a low representation within each sample following two criteria: a minimum frequency of 0.01% (and a probability of 0.0001) and a forward/reverse sequence reads balancement ≥ 30% combined with a localization in homopolymeric regions ≤ 2 nucleotides.
Those sequences containing INDELs of ≥ 1 bp located within a region encompassing the spacer were considered as ZFN-induced genome modifications.
To evaluate the level of human hematopoietic engraftment, cells were collected by femoral bone marrow aspiration at days 30 and 60 post-transplantation. Mice were sacrificed at 90 days posttransplantation, and cells were collected from their femora, tibiae, and spleen. Secondary transplants were conducted by transplanting 80% of femoral BM cells from each primary recipient into one secondary recipient mouse. In all instances, NSG mice irradiated with 1.5 Gy were used as recipients for secondary transplants. At the final time point of analysis, the presence of cells corresponding to the different hematopoietic lineages was determined. Cells obtained from BM aspirates were stained with hCD45 APC (eBioscience) and CD34 PE (BD Biosciences) antibodies according to the manufacturer's instructions. Analyses of multilineage engraftment were performed using CD45 APC-Cy7 (BioLegend), CD34 APC (BD Biosciences), CD33 PE-Cy7 (BD Biosciences), CD19 PE-Cy7, and CD3 APC (BioLegend), according to the manufacturer's instructions.
Analysis of FA pathway functionality FANCD2 foci were analyzed by immunofluorescence of LCLs treated for 16 h with 40 nM of MMC. After MMC treatment, cells were counterstained with rabbit polyclonal anti-FANCD2 (Abcam, ab187-50) and DAPI as previously described (Hotta & Ellis, 2008;Raya et al, 2009). Cells with more than ten foci per cell were scored as positive.
To test the MMC sensitivity of FA LCLs, 5 × 10 4 LCLs were seeded in 24-well plates in a final volume of 1 ml and increasing concentrations of MMC were added (from 0 to 333 nM). Cell viability was determined 10 days afterward by flow cytometry using the DAPI vital marker in a LSR Fortessa. Off-line analyses were performed with the FlowJo software v10.

Analysis of radical oxygen species (ROS)
The production of ROS was measured by flow cytometry using a non-fluorogenic probe (CellROX â Deep Red Reagent; Life Technologies) that upon oxidation exhibits strong fluorogenic signal (640nm excitation and 665-nm emission) following manufacturer's instructions. Routinely, 5 × 10 4 cells were incubated with CellROX â Deep Red Reagent at a final concentration of 5 lM in PBS for 20 min at 37°C. After incubation, cells were washed with PBA and DAPI was added. Quantification of ROS production was conducted in LSR Fortessa and data were analyzed with FlowJo software v7.6.5.

Clonogenic assays
Clonogenic assays were established using 900 CD34 + HD CB cells or 10,000 mPB CD34 + cells from FA-A patients in enriched methylcellulose medium (StemMACS TM HSC-CFU complete with Epo, Miltenyi Biotech). To test the MMC sensitivity of hematopoietic progenitors from FA-A patients, 3-10 nM of MMC was added to these cultures.

Statistical analysis
The statistical analysis was performed using GraphPad Prism version 5.0 for Windows. For the analyses of experiments in which n < 5, a nonparametric two-tailed Mann-Whitney test was performed when two variables were compared, or Kruskal-Wallis with Dunn's multiple comparison test when more than two variables were compared. In the experiments in which n ≥ 5, a Kolmogorov-Smirnov test was done to test the normal distribution of the samples. If samples showed a normal distribution, a parametric two-tailed paired t-test was performed when two variables were compared or an ANOVA with Bonferroni's multiple comparisons test or Tukey 0 s multiple comparisons test when more than two variables were compared. If samples did not follow normal distribution, the previously mentioned nonparametric tests were used. Significances are indicated in the figures and legends.
Expanded View for this article is available online.
expressing NOD-SCID IL2Rgamma(null) humanized mice. Blood 117: 3076 -3086 The paper explained Problem Thanks to the development of designed nucleases, gene editing has become a good alternative to conventional gene therapy in both inherited and acquired diseases. Nevertheless, the possibility of conducting gene editing in HSPCs is still a major challenge, particularly in diseases such as Fanconi anemia (FA), which are characterized by evident phenotypic defects at the HSPC level. Despite these difficulties, the possibility that FA HSPCs treated by therapeutic gene editing may develop in vivo proliferation advantage over uncorrected cells is of particular relevance in the field of gene editing.

Results
We initially showed specific ZFN-mediated targeting of donor constructs in the AAVS1 "safe harbor" locus in lymphoblastic cell lines (LCLs) from FA-A patients. In this study, we demonstrate that this process is feasible, not only using a therapeutic donor, but also when non-therapeutic donors are used, indicating that FANCA is not essential for the homologous directed repair (HDR)-mediated gene editing of human cells. The specific insertion of FANCA in the AAVS1 site of FA-A LCLs conferred an evident phenotypic correction of these cells, demonstrating therapeutic gene editing in FA-A hematopoietic cells. Prior to developing therapeutic gene editing in CD34 + cells from FA patients, an efficient protocol aiming at facilitating the insertion of donor constructs in HSPCs from healthy donors was developed. Efficiencies of HDR around 14% were achieved in in vitro cultured cells and around 10% and 1%, respectively, in HSCs capable of repopulating the hematopoiesis of primary and secondary immunodeficient recipients. In our final set of experiments, we demonstrated that the proposed gene-editing approach was capable of correcting the phenotype of hematopoietic CD34 + cells from FA-A patients, showing for the first time the possibility of conducting therapeutic gene editing in HSPCs from patients with a syndrome characterized by bone marrow failure and DNA repair defects.

Impact
Our results show for the first time the phenotypic correction of primary HSPCs from FA-A patients by means of the specific insertion of a therapeutic FA gene into a "safe harbor" locus, thus minimizing risks of insertional oncogenesis. Our study constitutes the first proof of concept showing the feasibility of conducting gene editing in primary FA CD34 + cells as a future therapeutic approach for the treatment of bone marrow failure syndromes such as FA, characterized by evident functional defects at the HSPC level.