Targeted Genome Editing in Human Repopulating Hematopoietic Stem Cells

Abstract

Targeted genome editing by artificial nucleases has brought the goal of site-specific transgene integration and gene correction within the reach of gene therapy. However, its application to long-term repopulating Hematopoietic Stem Cells (HSCs) has remained elusive. Here we show that poor permissiveness to gene transfer and limited proficiency of the homology directed DNA repair pathway constrain gene targeting in human HSCs. By tailoring delivery platforms and culture conditions we overcame these barriers and provide stringent evidence of targeted integration in human HSCs by long-term multilineage repopulation of transplanted mice. We demonstrate the therapeutic potential of our strategy by targeting a corrective cDNA into the IL2RG gene of HSCs from healthy donors and a subject with X-linked Severe Combined Immunodeficiency (SCID-X1). Gene edited HSCs sustained normal hematopoiesis and gave rise to functional lymphoid cells that possess a selective growth advantage over those carrying disruptive IL2RG mutations. These results open new avenues for treating SCID-X1 and other diseases.

Hematopoietic Stem Cells (HSC)-based gene therapy has provided therapeutic benefit in primary immunodeficiencies^1,2, thalassemia³ and leukodystrophies^4,5. Whereas more advanced vectors, such as lentiviral vectors (LV)⁶ have shown improved safety and efficacy, the risk of insertional mutagenesis^7,8 and unregulated transgene expression^9,10 remains a concern when using semi randomly integrating vectors. These adverse effects may trigger oncogenesis, toxicity or elimination of the gene-modified cells.

Artificial endonucleases, such as Zinc Finger Nucleases (ZFNs)¹¹, Transcription Activator-Like Effector Nucleases (TALENs)¹², and RNA-guided nucleases (CRISPR/Cas)¹³ brought the possibility of gene targeting within the reach of gene therapy^14,15. These nucleases are used to efficiently and specifically target a DNA Double Strand Break (DSB) to a preselected genomic site^13,16,17. According to the repair process that seals the break¹⁸, the outcome can be disruption, reconstitution or editing of the original sequence. If the DSB is sealed by the error-prone Non-Homologous End-Joining (NHEJ) pathway, insertions and deletions (indels) are common^19-21. If the DSB is sealed by the high-fidelity Homology Directed Repair (HDR) pathway, which acts preferentially during the S/G2 phase, the targeted sequence can be edited by providing an exogenous DNA template flanked by homologous sequences to the nuclease target site. Targeted editing allows integrating an expression cassette into a safe genomic harbor^22,23, or correcting disease-causing mutations by inserting a functional copy of the affected gene downstream its own promoter^14,24. Gene correction, as opposed to gene replacement, may not only restore the function but also the physiological expression of the gene, a long-sought goal of gene therapy.

Whereas gene disruption by ZFNs has been shown in human hematopoietic stem/progenitor cells (HSPC) assayed by repopulation of SCID mice²⁰, targeted gene editing in these cells has not been reported. Here, we identify and overcome major constraints limiting gene targeting in HSPCs and provide proof-of-efficacy for this approach by functional reconstitution of the IL2RG gene, whose mutations are responsible for X-linked Severe Combined Immunodeficiency (SCID-X1).

Site-specific integration in human HSPC

We developed a protocol for delivery of ZFNs and donor DNA template into human cord blood (CB) CD34+ cells by mRNA electroporation and Integrase-Defective LV (IDLV), and targeted integration of a GFP cassette into the AAVS1 “safe harbor”²² or a mutational hotspot of IL2RG^14,15 (Fig. 1a,b and Extended Data Fig. 1). This protocol yielded on average 5% GFP+ cells in liquid culture and colony-forming cells (CFC) assays and high percentages of indels in the respective ZFNs target sites, although after a transient cell loss (Fig. 1c and Extended Data Fig. 2a-e). PCR analysis and Southern blot showed integration of the GFP cassette at the intended targets in >90% of the GFP+ colonies (n=89) and in induced Pluripotent Stem cells obtained by reprogramming the GFP+ sorted cells (Fig. 1d,e and Extended Data Fig. 2f,g). We then transplanted the CD34+ cells treated for AAVS1 or IL2RG gene targeting into NSG mice and found human cells engraftment in all mice (Fig. 2a). In the first 8 weeks post-transplant, 95% of the mice had circulating GFP+ cells (mean 6.2±1.3%; Fig. 2b), while only 42% of the mice maintained long-term GFP marking. End-point analyses performed on the peripheral blood (PB), spleen and bone marrow (BM) showed that GFP+ cells were present within all human hematopoietic lineages, including lymphoid and myeloid cells, and erythroid precursors (mean 2±0.8%; Fig. 2c). Similar frequencies of GFP+ cells were found among primitive and committed progenitors in the BM (mean 2.2±0.9%; Fig. 2d). PCR analyses confirmed targeted integration in human lymphoid, myeloid and CD34+ cells from the spleen and BM of representative mice (Fig. 2e). CFC assays on CD34+ cells from BM showed GFP+ myeloid and erythroid colonies (Fig. 2f) with targeted integration (Fig. 2e). Analysis of BM cells showed the occurrence of NHEJ-mediated indels in the ZFN target site in the majority of mice (64%; Fig. 2g) at higher levels than observed for GFP marking, indicating that DNA DSB induction and repair by either HDR or NHEJ is compatible with hematopoietic repopulation. Overall, these data show that our gene targeting protocol achieves site-specific integration in human multipotent long-term NSG repopulating cells (SRC), surrogate readouts of HSC.

Figure 1. Targeted integration into AAVS1 or IL2RG in CB CD34+ cells.

a, Schematic of the donor IDLV template containing a GFP cassette driven by the phosphoglycerate kinase promoter (PGK) flanked by sequences homologous to the genomic target locus, the target locus with the ZFNs cleavage site and the locus after HDR showing the PCR primers used to assess 5′ or 3′ HDR-mediated integration junctions (black arrows). b, Flow chart for targeted integration (TI) and cell analyses. c, Representative flow cytometry dot plots (Top) and percentages of GFP+ cells and NHEJ-induced indels at the target locus (Bottom) of CB CD34+ cells treated for TI into AAVS1 or IL2RG. Means ± SEM (AAVS1, n=39 on 19 CB donors; IL2RG, n=10 on 9 CB donors). Unrelated donor: cells treated with IDLV lacking homology to the target site. nd: not detectable, np: not performed. d, Targeting specificity in CFC. Percentage of colonies positive for both (HDR), either (HDR+NHEJ) or none (Unknown) 5′ and 3′ HDR junctions by PCR. Numbers of colonies screened indicated inside the bars. e, Southern blot (top) and PCR (bottom) analyses of iPSC obtained by reprogramming GFP+ or GFP− cells from (c).

Figure 2. Transplantation of gene targeted CD34+ cells in NSG mice.

CD34+ cells treated as in Fig.1b were transplanted into NSG mice. a, (Left) Human cell engraftment (CD45+) 12-23 weeks post-transplant in the indicated organs. (Right) Percentages of the indicated lineages within the human graft. Data from individual mice and mean ± SEM (n=42 mice; 6 independent experiments on 13 CB donors). b, Time course of human GFP+ cells in PB of mice. Dashed lines: mice in which GFP+ cells were no longer detectable (<0.1%) 12 weeks post-transplant. c, GFP+ cells within the human graft in the indicated organs (Left) and lineages (Right). (n=18). d, GFP+ cells within human primitive (CD34+CD38−) or committed (CD34+CD38+) progenitors or differentiated cells (CD34-CD38+) in mouse BM. e, PCR analysis for TI into AAVS1 on human lymphoid (CD19+) and myeloid (CD33+ and CD13+) cells sorted from the mice and on GFP+ CFC from mouse BM. f, Representative images of GFP+ colonies. Scale bars, 0.5 mm. g, NHEJ at AAVS1 or IL2RG ZFNs target sites on total BM cells from (a). n= 25, 3 independent experiments.

Low targeting efficiency in HSPC

The in vivo studies revealed that only ~40% of mice had on average 2% human GFP+ cells long-term. These figures appear lower than expected from transplanting CD34+ cells with ~5% targeting efficiency in vitro (Suppl. Information) and suggest that either SRCs are targeted less efficiently than the bulk CD34+ cells, or the gene targeted SRCs have a competitive disadvantage in vivo. We thus compared the percentages of GFP+ cells among different subpopulations of cultured CD34+ cells, prospectively identified by surface markers²⁵ as primitive (CD34+133+90+), early (CD34+133+90−) and committed (CD34+133−) progenitors and the differentiated cells (CD34−, Fig. 3a left panel). We found a decreasing frequency of GFP+ cells when moving from the differentiated cells up the progenitor hierarchy. In primitive cells the percentage of GFP+ cells was 20-fold lower than that measured in differentiated cells. We thus investigated the potential rate-limiting steps for gene targeting in primitive cells (Extended Data Fig. 3a-d). Transgene expression upon mRNA electroporation was similar among the subpopulations or slightly lower for the primitive cells. The level of NHEJ induced at the ZFN target site was higher in the primitive cells and progressively lower in committed and differentiated cells. This difference, however, diminished with time in culture, potentially due to the loss of some treated primitive cells. Indeed, induction of apoptosis was higher in this subset. Taken together these data suggest that the primitive cells are more sensitive to our treatment and less permissive to HDR and/or donor template delivery.

Figure 3. Gene targeting in primitive versus committed progenitors.

a, GFP+ cells within the indicated subpopulations 3 days after treatment for TI. The left most panel shows results using the protocol described in Fig.1b. The other panels show the effect of longer prestimulation and/or addition of the indicated drugs, as shown in the schematic in (b). Means ± SEM (n=31,15,14,15,7,5 respectively on 37 total CB donors). *p< 0.05; ***p< 0.001 (one-way Anova). c, Composition of CD34+ cells cultured with or without SR1; subpopulations as in (a) . Means ± SEM (n=4). d, Total (left) and GFP+ (right) colonies from CD34+ cells treated for TI with or without SR1. Means ± SEM (n=20, 14). e, Yield of GFP+ early progenitors relative to that obtained using the original protocol of Fig.1b. Means ± SEM (n=8,7,11,10,3,5) **p<0.01 (one-way Anova). f, Percentage of NSG mice harboring GFP+ cells 14 weeks after transplant of CD34+ cells treated with the indicated TI protocols. g, Time course of human engraftment in PB. Means ± SEM (24h SR1, n=5; 48h SR1, n=6; 48h, n=5) ****p<0.0001, ***p<0.001 (two-way Anova). h, GFP+ cells within CD45+ cells in PB 14 weeks post-transplant. Means ± SEM (n=4). Mice for the 24h SR1- condition are shown for comparison from Fig.2c.

Tailored conditions improved HSC gene targeting

Because cell cycle progression is a requirement of HDR and activation of the primitive progenitors may require longer stimulation, we postponed the gene targeting procedure to the 3^rd day of culture (Fig. 3b). At this time, the cells are also likely to become more permissive to LV transduction. Because increasing time in culture promote differentiation, we added the Aryl Hydrocarbon Receptor Antagonist (StemRegenin 1, SR1)²⁶ and/or 16,16 dimethylprostaglandin E2 (dmPGE2)^27,28 to the culture to better preserve stem and early progenitor cells (Fig. 3c). The delayed protocol increased significantly (≥2-fold) the percentage of GFP+ cells in primitive cells (Figure 3a, right panels). SR1 slightly reduced the percentage of GFP+ cells observed within each subpopulation but increased the yield of GFP+ CFC and early progenitors, consistently with the increased proportion of immature cells in SR1-treated cultures (Fig. 3d,e and Extended Data Fig. 3e). The addition of dmPGE2 increased the percentage of GFP+ cells in all subpopulations when used alone and showed additive effects with SR1. Importantly, both the delayed treatment and the addition of SR1 and dmPGE2 increased the fraction of mice long-term engrafted with GFP+ cells, which reached 100% when used in combination (Fig. 3f). Human cell engraftment significantly increased after SR1 addition to the culture (Fig. 3g) and was stable long-term (Extended Data Fig. 4a,b). Consistently with the increased GFP marking observed in vitro in the primitive cells, the mean percentage of GFP+ cells long-term engrafted in vivo increased with all types of delayed treatments (Fig. 3h). GFP+ cells contributed to multiple lineages and to the progenitor compartment in most mice (Extended Data Fig. 4c,d). Molecular analyses on BM cells showed evidence of targeted integration (Extended Data Fig. 4e). Serial transplant of purified CD34+ cells from the BM of primary mice showed engraftment and differentiation of targeted GFP+ cells in secondary recipients (Extended Data Fig. 4f,g). Overall, these data indicate that by tailoring experimental conditions we could improve the yield and frequency of targeted long-term SRC.

Targeted gene editing of IL2RG in HSPC

In the experiments described in Fig. 3g,h, the gene targeting construct was designed to insert a cDNA comprising exon 5 to 8 of IL2RG together with the GFP cassette into the IL2RG gene and used on CD34+ cells from healthy male donors (Fig.4a). In this way, the cDNA is transcribed from the endogenous IL2RG promoter and spliced to its upstream exons, thus providing a platform for correcting all SCID-X1 causing mutations downstream of exon 4. In order to assess functional reconstitution of the targeted gene, we challenged the repopulated mice with a human tumor cell line (MDA-MB 231) engineered to express human IL-7, IL-15 and GM-CSF (Fig. 4b). We previously reported that this challenge leads to improved reconstitution of functional human T and NK cells that eventually reject the tumor graft²⁹. T and NK cells are strictly dependent on IL2RG expression for survival and activity and are absent in SCID-X1. Upon tumor challenge, we observed a massive (mean 130±40-fold) expansion of the human T and NK lineages in the repopulated mice (Extended Data Fig. 5). GFP+ T and NK cells expressed IL2RG on the cell surface (Fig. 4c) and expanded similarly to their GFP negative counterparts in all mice (Fig. 4d). All repopulated mice effectively rejected the allogeneic tumor, at variance with non-transplanted mice, underscoring the development of a functional human immune system (Fig. 4e). Whereas myeloid cells sorted from the mice showed high levels of NHEJ at the targeted IL2RG site, comparable to those observed in the CD34+ cells pre-transplant, B cells showed very little, and T and NK cells virtually none (Fig. 4f). These findings reflect the dramatic counter selection of lymphoid cells carrying a disrupted IL2RG, as it naturally occurs with inherited SCID-X1 alleles, and confirm the functionality of the reconstituted gene in the expanded GFP+ cells. We then assessed the T-cell receptor repertoire of lymphocytes from the engrafted mice and found substantial diversity with almost overlapping polyclonal pattern between the GFP+ and GFP-sorted cell subsets (Fig. 4g and Extended Data Fig. 6). The GFP+ and GFP− T cells expanded ex vivo after polyclonal stimulation with the same kinetics in the presence of γ-chain dependent cytokines (IL-7 and IL-15), and proliferated to a similar extent in response to the allogeneic MDA-MB 231 cells (Fig. 4h,i and Extended Data Fig. 7a). GFP+ and GFP− T cells were similarly comprised of CD8 and CD4 cell subsets, with a majority of cells showing effector phenotypes (Extended Data Fig. 7b,c). Consistently, both GFP+ and GFP− cells robustly produced γIFN and IL-2 after PMA-ionomycin stimulation or when co-cultured with the allogeneic tumor at different effector/target ratios (Extended Data. Fig. 7d,e). Molecular analyses proved that nearly all GFP+ cells contained targeted integration into IL2RG (Fig. 4j). We then measured the phosphorylation of two downstream effectors in the signaling cascade of γ-chain coupled receptors (Fig. 4k and Extended Data Fig. 8). The targeted T cells displayed similar kinetics and extent of phosphorylation of STAT5 and AKT as their GFP-counterparts after stimulation with increasing IL-15 and IL-2 doses. Overall, these data prove functional reconstitution of the edited IL2RG gene, which supported lymphopoiesis and mature T-cell function indistinguishably from the wild-type allele.

Figure 4. Functional reconstitution of IL2RG in the lymphoid progeny of HSC.

a, Schematic of the IL2RG donor: a promoter-less IL2RG cDNA, comprising exons 5-8 plus 3′ untranslated region (UTR), and a PGK-GFP cassette are flanked by homologous sequences to those surrounding the IL2RG ZFNs target site. b, Flow chart of cell transplantation, tumor challenge and analyses. c, Density plots of γ-chain expressing T (Top) and NK (Bottom) cells showing GFP marking. (n=7,11). d, Expansion of GFP− and GFP+ T and NK cells after tumor challeng. e, Tumor weight 3 weeks after challenge, in mice transplanted (n=16) or not (n=3). ****p<0.0001 (unpaired t test). f, NHEJ in the IL2RG gene on CD34+ cells cultured in vitro and on their progeny sorted from the transplanted mice. g, TCR complexity score calculated on GFP+ or GFP− T cells from transplanted mice. Human PB mononuclear cells (PBMCs) were used as positive control. ****p<0.0001 (one-way Anova). h, Ex vivo growth of GFP+ and GFP− T cells from the spleen of transplanted mice upon stimulation (n= 4). i, Division index of GFP+ or GFP− T cells 7 days after PHA stimulation or co-culture with tumor cells at the indicated effector-to-target (E/T) ratios. p=ns (unpaired t test). T cells from healthy donor (HD) were used as controls. j, Southern blot (top), PCR (middle) and GFP q-PCR (bottom) analyses showing TI of the corrective IL2RG cDNA in sorted GFP+ T cells from (h). UT: untreated cells. k, Heatmap showing changes in phosphorylation levels of STAT5 after the indicated time of exposure to decreasing amounts of IL-2 or IL-15, on T cells from (h). p=ns (two-way Anova).

Specificity of IL2RG ZFNs on the HSC genome

We previously performed a genome-wide screening in K562 cells to identify potential off target sites of the IL2RG ZFNs used in this study¹⁶ and found a low rate of indels in few genomic loci bearing homology to the intended target site. We then determined whether these sites were also affected in the HSPC treated here with ZFNs containing the same IL2RG DNA binding domains but coupled to improved obligate heterodimeric FokI domains. We deep sequenced the genomic regions encompassing the identified potential target sites on treated CD34+ cells cultured in vitro and on human cells from the BM of long-term engrafted NSG mice (Table 1 and Extended Data Fig. 9). The intended IL2RG target site showed 45 to 61% indels rate in the in vitro cultured cells and 20 to 43% in the in vivo engrafted cells. However, we detected indels only in 2 in vitro samples (at 0.17 to 0.7%) for the top 2 previously identified off target sites, while from the in vivo samples we found just one site with evidence of NHEJ (at 0.02%). Deep sequencing of all other putative off-target sites gave results not statistically different from the background error rate, which limits detection at 0.01% in our analysis (see Suppl. Information). The elimination of detectable off-target activity at some previously identified sites is consistent with the adoption of obligate heterodimeric FokI variants in this study, which would detarget activity from sites requiring ZFN homodimers. This analysis demonstrates a high specificity for the ZFNs used, evidenced by the 100 fold ratio between activities at the intended target site vs. the top identified off target site. It is possible, however, that additional off-target sites exist, which have not been identified by our previous screening.

Table 1. Indels quantification on the target and candidate off target sites of IL2RG ZFNs in the progeny of treated CD34+ cells.

In/Ex indicates whether the ZFN target site is within an exon (Ex), intron (In) or intergenic (Outside) near to the RefSeg gene indicated in the first column.Homology: percent of sequence identity to the IL2RG-ZFNs binding sites. ZFN-Dimer: the site can be bound by a homodimeric (LL/RR) or heterodimeric (LR/RL) ZFN pair. Number indicates the spacer length in bpbetween the ZFN binding sites. NHEJ (%): percent of indels on treated CD34+ cells cultured in vitro (sample A, B and G) and on human cells harvested from the BM of long-term engrafted NSG mice from Fig. 3 (mouse B2, C0 and E2). For quantification see Supplementary Information and Extended Data Fig. 9. Ns: not significant (Fisher exact test for contingency data).

Nearest RefSeq gene	In/Ex	Homology (%)	ZFN-Dimer	NHEJ (%)
				In vitro			Engrafted in Mouse
				A	B	G	B2	C0	E2
IL2RG	Ex	100	L_5_R	54.60	61.18	45.60	26.08	43.51	20.07
SCARB1	Outside	70.8	L_5_R	0.17	0.70	ns	ns	ns	ns
SLC31A1	In	75	R_5_L	0.61	ns	ns	ns	0.02	ns
FAM133B	In	66.7	R_6_R	ns	ns	ns	ns	ns	ns
KIAA0528	In	87.5	L_5_L	ns	ns	ns	ns	ns	ns
SF3B1	Outside	66.7	L_5_L	ns	ns	ns	ns	ns	ns
A2BP1	Outside	75	L_5_R	ns	ns	ns	ns	ns	ns
ANKFY1	Ex	87.5	L_3_R	ns	ns	ns	ns	ns	ns
TRIM43	Outside	91.7	L_4_L	ns	ns	ns	ns	ns	ns
SEC16A	Ex	70.8	R_6_L	ns	ns	ns	ns	ns	ns

IL2RG gene correction in SCID-X1 HSPC

We applied the optimized protocols developed for CB to CD34+ cells from adult BM and obtained an overall gene targeting efficiency of 6±0.5% (n=4 donors) and a high rate of indels induced at either the AAVS1 or IL2RG ZFNs target sites (Fig. 5a). Targeting was less efficient in the more primitive populations although reaching similar values as those observed for CB cells. Xenotransplantation proved the long-term multilineage repopulation capacity of the targeted cells, with all transplanted NSG mice bearing GFP+ cells at frequencies comparable to those observed with CB cells (Fig. 5b). Based on these results, we then tested our gene correction strategy on BM CD34+ cells from a symptomatic four-month old SCIDX1 patient bearing a missense mutation in IL2RG Exon 7 (c.865C>T; R289X). As expected for this mutation, blood sampling or BM harvest from the patient did not show any T or NK cells (Extended Data Fig. 10a,b). From 3 to 11% of the treated cell progeny became GFP+, depending on primitive vs. committed progenitor status (Fig. 5c). CFC assays yielded 3 GFP+ colonies out of ~100 scored (Extended Data Fig. 10c). Flow cytometry showed normal expression of the γ-chain protein in the myeloid progeny of the GFP+ CFCs (Fig. 5d). PCR analyses of these colonies proved targeted integration into IL2RG leading to expression of the expected fusion transcript between the corrective cDNA and the upstream endogenous exons (Fig. 5e,f and Suppl. Information). Overall, this data shows reconstitution of a functional IL2RG gene upon targeted editing of a SCID-X1 allele in HSPC.

Figure 5. Targeted integration and IL2RG gene correction in BM-derived CD34+ cells from healthy donors and a subject with SCID-X1.

a, (Top) GFP+ cells within the indicated subpopulations derived from BM CD34+ cells of adult healthy donors, treated for TI according to the best performing protocol from Fig. 3. (Bottom) NHEJ at the ZFN target site on total cells. Means ± SEM, (n=10,3 from 4,3 donors for AAVS1 or IL2RG, respectively) b, (Top left) Human cells in PB of NSG mice transplanted with cells from (a). (Top right) Percentages of the indicated lineages within human cells 15 weeks post-transplant. (Bottom) GFP+ cells within the indicated populations. c, GFP+ cells measured as in (a) in BM CD34+ cells from a subject with SCID-X1 treated for IL2RG gene correction. d, IL2RG expression on myeloid (CD33+) cells from a GFP+ colony from the cells treated in (c) or from pooled wild type colonies. e, PCR analysis for TI into IL2RG of the corrective cDNA on cells from (c) and (d). f, Expression of the fusion transcript bearing the corrective IL2RG cDNA measured by Q-PCR (top) or RT-PCR (bottom) on cDNA from a GFP+ SCID-X1 myeloid colony. IL2RG targeted T cells from engrafted mice analyzed in Fig. 4j were used as positive control, while a myeloid colony from wild-type BM cells (WT) and PBMCs were used as negative controls.

Discussion

Here we developed a strategy for targeted genome editing in human long-term repopulating HSCs and exploited it to insert transgenes into a genomic safe harbor or downstream of the promoter of an endogenous gene to reconstitute its functional expression. By the latter approach, we demonstrate correction of the defective IL2RG gene in HSPC from a subject with SCID-X1. As we obtained consistent results when targeting two different loci, we expect that our genome editing strategy can be used to target a variety of genomic sites. As the procedure employs a combination of IDLV infection and mRNA electroporation to deliver donor template and ZFNs, it has the potential to be applicable to a range of other genome editing tools.

We found that primitive cells are more sensitive than committed progenitors to the cytotoxicity of the gene targeting procedure and less proficient at performing HDR, likely because of their quiescence or slow cycling. These findings are consistent with reports showing delayed DNA repair and enhanced apoptosis after γ-radiation in human HSC as compared to progenitors³⁰ and preferential repair of DNA DSBs by NHEJ in quiescent murine HSC³¹. By delaying the time of treatment and exploiting compounds reported to support ex vivo maintenance and expansion of HSCs^26-28, we were able to partially relieve the block to HDR. This effect is likely due to an increased transit through the S/G2 phases of the cell cycle, when HDR can occur, and, possibly, up-regulation of its endogenous machinery. Other beneficial effects might be increased permissiveness to gene delivery, more efficient mRNA translation and reduced growth arrest and apoptosis in response to the gene targeting procedure. As improved procedures for ex vivo HSC expansion become available, they might increase the yield of gene targeted cells and allow their selection before in vivo administration. This would enable wide application of safe harbor sites, such as AAVS1, for robust expression of therapeutic transgenes^22,23.

When applied to the functional correction of IL2RG in HSCs, we show that our strategy is compatible with normal development of the lymphoid lineages. IL2RG edited lymphoid cells repopulated the mice and responded to γ-chain-dependent cytokines indistinguishably from their unedited counterparts. On the contrary, lymphoid cells carrying NHEJ-mediated IL2RG inactivation were counter-selected in the mice, phenocopying the SCID-X1 disease¹.

This disease may be suitable for the clinical translation of targeted gene correction since prior clinical studies have demonstrated the potential efficacy but also the risks of HSC gene therapy using early generation vectors^7,8. Whereas our strategy should abrogate the risk of insertional mutagenesis and the concern for unregulated transgene expression, it remains to be established whether a potentially limiting amount of targeted HSPCs would enable effective and safe correction of SCID-X1 in humans (even though both HSC and progenitors, which are more efficiently targeted, would help repopulating the absent lymphoid lineages). Another concern is with the potential ZFNs off target activity, although our analysis showed a high specificity for the IL2RG ZFNs used in this study. Overall, we envisage a fairly straightforward path to clinical translation of our strategy, considering that GMP manufacturing of LV and mRNA electroporation are already established. Moreover, infusion of autologous gene targeted cells could be performed without pharmacological conditioning, taking advantage of the selective advantage of the corrected cells.

Methods

Vectors and zinc finger nucleases

Homology-directed repair donor templates were generated from HIV-derived, third-generation self-inactivating transfer constructs. IDLV stocks were prepared as previously described¹⁴ and titered by a Q-PCR designed to discriminate the reverse transcribed vector genome from plasmid carried over from transient transfection³². Sequence and maps of AAVS1-PGK.GFP were previously reported²², while IL2RG-cDNA.PGK.GFP is described in detail elsewhere (Firrito et al., manuscript in preparation). ZFNs that target intron 1 of PPP1R12C or exon 5 of IL2RG were previously described^14,16,22,33. The latter pair was modified to contain high-fidelity obligate heterodimeric FokI variants³⁴. Both pairs of ZFNs were transiently expressed as mRNAs. Plasmid templates for ZFNs mRNA production (described in Extended Data Fig. 1b) were linearized and purified by phenol/chloroform extraction followed by ethanol DNA precipitation. 2 μg/reaction of linearized plasmid template was in vitro transcribed at 37 °C for 2 hr using T7 RNA polymerase and 7.5 mM nucleotide triphosphates (MEGAscript Kit; Ambion). Cap0 mRNAs were generated by supplementing the reactions with 6 mM m7(3′-O-methyl)-G(5′)ppp(5′)G, a nonreversible cap analog (ARCA, New England Biolabs) and lowering the concentration of GTP to 1.5 mM. After TURBO DNase treatment (4U/reaction, 1 hr at 37°C), mRNAs were poly(A) tailed with E. Coli Poly(A) Polymerase (8U/reaction) for 1 hr at 37°C (PolyA tailing kit; Ambion), yielding ≥ 150 nt polyA. Transcripts were purified by RNeasy Plus Mini Kit (Qiagen). All RNA samples were analyzed by denaturing agarose gel electrophoresis for quality assurance.

In vitro culture and assays on human CB or BM-derived CD34+ cells

CD34+ cells were either freshly purified from human CB after obtaining informed consent and upon approval by the San Raffaele Hospital Bioethical Committee, or purchased frozen from Lonza. 10⁶ CD34+ cells/ml were stimulated in serum-free StemSpan medium (StemCell Technologies) supplemented with penicillin, streptomycin and human early-acting cytokines (for CB-derived cells: stem cell factor (SCF) 100 ng/ml, Flt3 ligand (Flt3-L) 100 ng/ml, thrombopoietin (TPO) 20 ng/ml, and interleukin 6 (IL-6) 20 ng/ml; for BM-derived cells: SCF 300ng/ml, Flt3-L 300 ng/ml, TPO 100 ng/ml, and IL-6 60 ng/ml; all purchased from Peprotech) for 24 or 48 hr and then infected with IDLVs at multiplicity of infection (MOI) 100-500. The following day the cells were electroporated with 175 μg/ml ZFNs encoding mRNAs (P3 Primary Cell 4D-Nucleofector X Kit, program EO-100; Lonza). For some experiments, the following drugs were supplemented to the culture media: 1 μM SR1 (kindly provided by T. Boitano and M. Cooke, GNF) added at every medium change, and 10 μM dmPGE2 (Cayman) added at the beginning of the culture, 1 hour before and just after electroporation. For CFC assays, 800 cells/plate were seeded one day after electroporation in methylcellulose-based medium (MethoCult H4434, StemCell Technologies). Two weeks after plating, colonies were counted and identified according to morphological criteria.

Flow cytometry

For immunophenotypic analysis of CD34+ cells and their progeny (performed on FACSCanto II; BD Pharmingen), we used the antibodies reported in Supplementary Information. Single stained and FMO stained cells were used as controls. For quantitative flow cytometry we used Flow-count Fluorospheres (Beckman Coulter) according to the manufacturers’ instructions. Apoptosis analysis was performed on CD34+ cells one day after electroporation using PB-conjugated Annexin V (Biolegend) and Apoptosis Detection Kit with 7-Aminoactinomycin D (7AAD, BD Pharmingen) according to the manufacturers’ instructions. Percentages of live (7AAD−, AnnexinV−), early apoptotic (7AAD−, AnnexinV+), late apoptotic (7AAD+, AnnexinV+) and necrotic (7AAD+, AnnexinV−) cells are reported. Cell sorting was performed using MoFlo XDP Cell Sorter (Beckman Coulter).

Molecular analyses

For molecular analyses, genomic DNA was isolated with DNeasy Blood & Tissue Kit or QIAamp DNA Micro Kit (QIAGEN). Extraction of genomic DNA from colonies in CFC assays was performed with Lysis Buffer as previously described⁵. NHEJ in AAVS1 locus or IL2RG gene was detected by the mismatch selective Cel1 assay as previously described¹⁴. Primers for PCR amplifications to detect targeted integration or for the Cel1 assay are indicated in Supplementary Information. PCR amplicons were resolved on agarose gel and visualized by ethydium bromide staining. For Southern Blot analyses, genomic DNA was extracted with Blood & Cell Culture DNA Midi Kit (QIAGEN) and digested using restriction enzymes (BglI for AAVS1 locus and BspHI for IL2RG) and probes (see Supplementary Information) both located outside of the homology regions included in the vectors. Matched DNA amounts were separated on 1% agarose, transferred to a nylon membrane and probed with ³²P-radiolabeled sequences indicated in Supplementary Information. Membranes were exposed in a Storage Phosphor Screen. For Q-PCR analysis, 200 ng of genomic DNA were analyzed using primers and probes complementary to a vector backbone sequence (Primer Binding Site), the GFP sequence and human TERT, the latter amplification used as normalizer, as previously described¹⁴. For gene expression analysis on the SCID-X1 gene corrected colony, mRNA was extracted using RNeasy Micro Kit (QIAGEN) and cDNA was synthetized using SuperScript VILO cDNA Synthesis Kit (Invitrogen). The resulting cDNA was amplified before Q-PCR by Taqman PreAmp Master Mix Kit (Applied Biosystems) according to the manifacturers’ instructions. Gene expression was performed in triplicate with a TaqMan Gene Expression assay specific for the recoded exon 7 of the IL2RG gene (Applied Biosystems; see Supplementary Information) in a 7900HT real-time PCR thermal cycler. The relative expression level of the recoded IL2RG gene was calculated by the ΔΔCt method and represented as fold change relative to the housekeeping gene control (HPRT), as previously described²².

Mice transplantation and analysis

For the in vivo studies, 8- to 11-week-old NOD-SCID-IL2Rg^−/− (NSG) mice were purchased by Jackson laboratory. The experimental protocol was approved by the Institutional Animal Care and Use Committee of the San Raffaele Scientific Institute. At day 4 of culture, 3×10⁵ gene targeted CB-derived CD34+ cells (or 7.5×10⁵ BM-derived cells) were infused intravenously into the mice after sub-lethal irradiation (200 cGy). Sample size was determined by the total number of available treated cells. Mice were attributed to each experimental group randomly. MDA3 human mammary carcinoma cell line was obtained by stable transduction of MDA-MB 231 cells with lentiviral vectors expressing the human cytokines GM-CSF, IL-7 and IL-15 from the PGK promoter, as previously described²⁹. 4×10⁶ MDA3 cells were implanted orthotopically in the mammary fat pad of NSG mice 14 weeks after CD34+ cells transplantation or in age-matched untransplanted NSG mice. Human CD45+ engraftment was followed by serial collections of blood from the mouse tail and, at the end of the experiment (12-23 weeks after transplantation), BM and spleen were harvested and analyzed for lineages composition and GFP content (see Supplementary Information for gating strategy).

T lymphocytes analyses

Human T cells were enriched from splenocytes harvested from the transplanted mice using magnetic beads conjugated to anti human CD3 and CD28 antibodies (Dynabeads human T-activator CD3/CD28; Invitrogen), following the manufacturers’ instructions, and grown in Iscove’s Modified Dulbecco’s Media (IMDM) (GIBCO-BRL) supplemented with penicillin, streptomycin, 10% FBS and 5 ng/ml each of IL-7 and IL-15 (PeproTech)²¹.

For TCR V-β repertoire analysis, mRNA was extracted from the expanded T cells using RNeasy Mini Kit (QIAGEN) and cDNA was synthetized using SuperScript VILO cDNA Synthesis Kit (Invitrogen). Multiplex PCRs optimized from a previous work ³⁵ were carried out on cDNA using V-β primers specific for 4 or 5 different families and a single FAM-labeled C-β primer. PCR products were fractionated on 6% polyacrylamide gel, visualized on Molecular Dynamics Typhoon 9410 (Amersham Biosciences) and analyzed using ImageQuant TL 7.0 (Amersham Biosciences). V-β complexity was determined by counting the number of distinct peaks and graded on a score of 0–8 ³⁶. The overall TCR complexity score was determined by summing up all 23 individual TCR V-β family specific scores.

To analyze phosphorylation of downstream effectors of IL2RG pathway, T cells were starved overnight at 37°C in IMDM without cytokines and then stimulated with IL-2 (1000 IU/ml, 100 IU/ml, 10 IU/ml; purchased from Novartis) or IL-15 (10 ng/ml, 5 ng/ml, 1 ng/ml) at 37°C for increasing times. Cells were then fixed in PBS 2% paraformaldehyde (PFA) for 10′ at 37°C, and after washing in PBS 0.1% BSA (3 times), they were permeabilized with ice-cold absolute methanol for 7′ on ice. After 60′ incubation of each time point of cytokine stimulation with different dilutions of Pacific Blue Succinimidyl Ester (PBSE) (Life Technologies), cells were washed, pooled and stained for flow cytometry.

For proliferation assay, 10⁵ T cells were labeled with Cell Proliferation Dye eFluor 670 (eBioscience) according to the manufacturer’s instructions. Labeled T cells were co-cultured in IMDM supplemented with penicillin, streptomycin, 10% FBS and 5 ng/ml each of IL-7 and IL-15, with different dilutions of MDA-MB 231 cells previously irradiated at 10000 rad or stimulated for 3 days with PHA (2μg/ml). After 7 days of culture, cells were analyzed by flow cytometry. Division index was calculated according to FlowJo software rules.

For IFN-γ release assay, T cells were stimulated at 37°C for 6 hours with PMA (50ng/ml) and Ionomycin (1μg/ml) in presence of 2μl per ml of culture of BD Golgi Plug (BD Pharmingen). Cells were then fixed and permeabilized using BD Cytofix/Cytoperm Kit (BD Pharmingen) and stained for flow cytometry.

Elispot assay for IFNγ release was performed as previously described^21,29. When the number of measured spots was above the detection limit of the plate reader (Eli.Expert, A.EL.VIS.), it was arbitrarly set to 500.

Deep sequencing of potential IL2RG ZFN off-target loci

Genomic DNA from ZFN-treated CD34+ cells or their progeny harvested from transplanted mice was amplified using REPLI-g Mini Kit (QIAGEN) and the top ranking candidate off-target genomic loci from our previous study¹⁶ amplified by PCR generating amplicons of 389±20 bp surrounding the potential ZFN binding site. PCR products were purified using Agencourt AMPure XP beads (Beckman Coulter) and adaptors were added by TruSeq DNA LT Sample Prep Kit (Illumina). In order to build an equimolar library, PCR products were quantified with KAPA Library Quantification Kit for Illumina sequencing platforms (KAPABIOSYSTEMS) on C1000 Thermal Cycler (BIO-RAD) and sequenced on MiSeq Illumina Platform using MiSeq Reagent v.3 (Illumina). Raw paired-end reads were joined with Fastq-Join program from the EA-Utils NGS suite (http://code.google.com/p/ea-utils/) and aligned to the specific genomic target sequences using Burrows-Wheeler Alignment Tool with maximal exact match version, BWA-MEM ³⁷. Alignments were evaluated and filtered using SAMtools ³⁸, Picard (http://picard.sourceforge.net) and BAMtools³⁹. Sequences with only primary alignments with quality > 15 were kept for further analysis. Deletions and insertions (indels) were quantified by a custom pipeline based on Python (http://www.python.org, version 2.7.6) and the PySAM library (https://code.google.com/p/pysam, version 0.7.5). Sequences with indels of ≥1 bp located within a region encompassing the spacer + 5 bp on each side were considered as ZFN-induced genome modifications. Coverage statistics were computed by the binomial distribution online calculator (http://www.vassarstats.net/binomialX.html). Fisher exact test were computed with the SciPy Python package (http://www.scipy.org, version 0.9.0) within the “stats” library. Multiple sequence alignment for indels visualization and plot was performed with ClustalW2⁴⁰ and MView (http://bio-mview.sourceforge.net/).

SCID-X1 cells

Peripheral blood and bone marrow samples from a subject with SCID-X1 were obtained according to the guidelines of the Medical Ethics Committee of the Erasmus MC, University Medical Center Rotterdam, the Netherlands.

Statistical analyses

Statistical analyses were performed by unpaired Student’s t test for pairwise comparison or one-way or two-way analysis of variance (ANOVA) with Bonferroni’s multiple comparison post-test for three or more groups, as indicated. Values are expressed as Mean ± standard error of the mean (SEM). Percent values were transformed into a log-odds scale (log (%x/(100-%x)) to perform statistical analyses.

Supplementary Material

Extended data Figure 1. Optimization of gene targeting protocol in CD34+ cells.

We optimized the delivery platform, dose and timing of ZFNs and HDR donor template administration. a, Performance of different gene delivery platforms. Cord blood (CB) CD34+ cells were pre-stimulated with early acting cytokines for 24 hr and transduced with GFP-encoding IDLV (MOI: 5×10²) or Adenoviral Vector serotype 5/35 (MOI: 5×10³) or electroporated with GFP-expressing mRNA (500μg/ml) or plasmid DNA (25μg/ml). The cells were analyzed by flow cytometry at the indicated days after the procedure. (Top) Representative density plots of GFP expression 24 hr post treatment. SSC: Side Scatter. (Bottom) Kinetics of transgene expression measured as percentage of GFP+ cells (Left) and relative GFP fluorescence intensity (RFI, measured as the ratio between the mean fluorescence intensity of the treated cells at each time point to the untreated cells) in arbitrary units (Right). UT: untreated cells. mRNA electroporation outperformed all approaches tested in terms of frequency of transfected cells and protein expression level. Note that while IDLV infects the majority of cells in these conditions its expression is constrained by the unintegrated nature³². Because mRNA transfection drives a robust but short-lived spike of expression, it appeared best suited for ZFN delivery, allowing proficient activity of the nucleases at the genomic target site while avoiding prolonged exposure. b, (Top) Schematic representation (not in scale) of a plasmid DNA template used for in vitro mRNA transcription with the T7 promoter (T7 prom.), the Kozak sequence and the XbaI restriction enzyme used for the plasmid linearization depicted. The protein domains of a ZFN are shown within the open reading frame (ORF). NLS: nuclear localization signal; ZFP: Zinc Finger Protein; FokI: FokI nuclease domain. (Middle) Representative denaturing gel electrophoresis of in vitro transcribed mRNAs encoding for the pair of ZFNs specific for AAVS1, before (-) and after enzymatic polyadenylation (pA). The ZFNs mRNAs were produced either as two separated transcripts (ZFN-L and ZFN-R) or as a single construct encoding for both ZFNs linked by a Tav.2A self-cleavage peptide sequence (ZFN-L.2A.ZFN-R; Bottom left). (Bottom Right) CB CD34+ cells were electroporated either with the two separate transcripts or with the single mRNA co-expressing both ZFNs. ZFN activity was measured on treated cells as percentage of NHEJ detected at the ZFN target site by Cel1 assay 10 days after electroporation. c, dose-response for AAVS1 targeting ZFNs mRNAs in CD34+ cells. CB CD34+ cells were transduced with Integrase Defective Lentiviral Vector (IDLV)¹⁴ bearing homology to the AAVS1 locus and expressing GFP, and then electroporated with the indicated escalating doses of AAVS1 ZFN mRNAs. ZFN activity was scored by measuring the extent of NHEJ-mediated repair at their genomic target site, and HDR was scored by the frequency of GFP+ cells obtained in liquid culture. (Left) NHEJ measured by Cel1 assay at day 10 post electroporation for the indicated dose of mRNA. Means ± SEM (n=3). (Right) Percentages of GFP+ cells by flow cytometry 3 days after treatment. The percentages of viable cells (indicated on top of the histogram) were calculated as percentages of 7AAD negative cells gated on singlets. A dose dependent increase in the percentage of NHEJ and GFP+ cells was observed for the first three mRNA doses, whereas the highest dose caused a significant reduction in the number of viable cells, and a reduction in the efficiency of gene targeting. Based on these data, we selected the dose of 175 μg/ml RNA to perform all further experiments. d, Choice of delivery platform for the HDR donor template. CB CD34+ cells were either transduced with the AAVS1 donor IDLV and electroporated with the cognate ZFNs mRNAs, or co-electroporated with AAVS1 donor plasmid DNA and ZFNs mRNAs. (Left) Cell viability measured by flow cytometry 3 days after electroporation, comparing untreated cells (UT) and gene targeted cells using IDLV or plasmid as donor templates. ****p<0.0001 (one-way Anova with Bonferroni’s multiple comparison post-test). (Right) Percentage of GFP+ cells using either donor templates. Means ± SEM (UT, n=3; IDLV, n=18; Plasmid, n=10). *p<0.05 (unpaired t-test). IDLV infection outperformed plasmid DNA electroporation in terms of the frequency of GFP+ cells and cell viability, consistent with our previous findings in other primary cell types^14,22. e, Schedule optimization for ZFNs and donor template delivery. After one day of prestimulation, CB CD34+ cells were first transduced with the AAVS1 donor IDLV and then electroporated at the indicated hours post-infection with ZFNs mRNAs (Left) or, on the contrary, first electroporated with ZFNs mRNAs and then transduced with IDLV (Right). The time lines of the experiments are shown on Top of the histograms. The percentages of GFP+ cells measured by flow cytometry three days after treatment and NHEJ measured by Cel1 assay ten days after treatment are shown on bottom left. On bottom right, the percentage of GFP+ cells is expressed as fold to the percentage achieved in the same experiment with the best strategy on the left. The highest frequency of GFP+ cells was obtained by combining IDLV-based donor template delivery 24 hours before ZFNs mRNA electroporation. Sequential exposure to the two delivery platforms avoids competition for cell entry and minimizes mutual interference likely due to activation of innate responses to exogenous nucleic acids or the timing of peak ZFN expression relative to IDLV reverse transcription and nuclear import.

Extended data Figure 2. Impact on cell viability and specificity of integration in CD34+ cells treated for TI.

a, Percentage of GFP+ cells measured in liquid culture 3 days after treatment as in Fig.1b for TI into AAVS1 or IL2RG and on the corresponding GFP+ colonies counted in CFC assays 2 weeks after plating. Means ± SEM (n=7 CB donors). ns: not significant (unpaired t test). b, Representative bright field and fluorescence microscopy images of GFP+ erythroid and myeloid colonies. Scale bar, 0.5 mm. c-e, The impact of the gene targeting procedure on the viability, proliferation and clonogenic output of the CD34+ cells was analyzed. c, Representative growth curves of CD34+ cells treated for TI or transduced with IDLV only or untreated (UT). d, Apoptosis analysis performed 24 hr after treatment on cells in liquid culture and e, number of CFCs plated one day after the indicated treatments. Means ± SEM; n=2 or 3, respectively. Overall, there was a transient reduction in viable cell number 24 hr after electroporation, also observed in CFC yield, which resulted from the combined exposure to electroporation, ZFNs mRNA and IDLV. However, the surviving cells grew with similar kinetics as the untreated controls in liquid culture and gave rise to similar proportions of myeloid and erythroid colonies. f,g, Targeting specificity of integration. f, Southern blot (top) and PCR (bottom) analyses for TI into IL2RG on iPSC obtained by reprogramming GFP+ cells from Fig. 1c. UT: untreated cells. g, Genomic DNA from GFP+ colonies was analyzed by PCR for TI into AAVS1 or IL2RG. The gels show the PCR amplicons for either the 5′ or 3′ HDR integration junction at the genomic target site and a control locus (CCR5). The percentages of colonies positive for TI by PCR are reported in Fig.1d.

Extended data Figure 3. Investigating lower gene targeting efficiency in the more primitive cells.

a, Gating strategy used to identify subpopulations of CB cells according to expression of CD90, CD133 and CD34 surface markers. b, After 24 hr of prestimulation, CD34+ cells were electroporated with 175 μg of GFP mRNA (the selected dose of a ZFN mRNA from Extended data Fig.1). Flow cytometry analysis was performed 2 days later using the gating strategy shown in Fig.3a. Bars represent the percentage of GFP+ cells (plotted on left axis) while the line shows the level of transgene expression (plotted on the right axis as MFI, measured in arbitrary units). Means ± SEM (n=16 on 6 CB donors). c, CD34+ cells treated for TI were sorted by FACS one day after electroporation according to the gating strategy showed in Fig.3a. The sorted populations were sampled at the indicated times and levels of NHEJ at the ZFN target site (AAVS1) were determined by Cel1 assay (n=3). d, Apoptosis analysis performed one day after electroporation on CD34+ cells transduced with IDLV and electroporated with ZFNs mRNAs. Percentages of live (7AAD−, AnnexinV−), early apoptotic (7AAD−, AnnexinV+), late apoptotic (7AAD+, AnnexinV+) and necrotic (7AAD+, AnnexinV−) cells. Means ± SEM (n=5 on 4 CB donors). e, Representative growth curves of CD34+ cells cultured in the presence or not of SR1 and treated (TI) or not (UT) with the protocol described in Fig.1b. Note that the TI and UT growth curves are reproduced from Extended Data Fig. 1c. Addition of SR1 to the culture did not change the proliferation rate of the cells. Because SR1 did not increase the total number of cells but the percentage of more primitive cells in culture (Fig. 3c), the absolute number of primitive cells is larger in cultures containing SR1.

Extended data Figure 4. Long-term multilineage engraftment of gene targeted CD34+ cells in primary and secondary NSG mice.

a, Percentages of human cells in the indicated organs of NSG mice 15-23 weeks after transplantation with CD34+ cells treated with the improved protocols from Fig. 3a with or without SR1. b, Percentage of the indicated lineages within human cells in PB of mice 14 weeks post-transplant. Means ± SEM (48h, n= 8; 48h SR1, n=11; 48h PGE2, n=3; 48h PGE2 SR1, n=6). Overall, the addition of SR1 and PGE2 to the in vitro culture did not significantly affect the in vivo differentiation of treated cells. Notably, the increased human engraftment achieved with the optimized culture conditions (as illustrated in Fig.3g) correlates with increased T cell output. c, Multi-lineage GFP marking in individual NSG mice transplanted with CD34+ cells treated with the indicated protocols for TI. Percentages of GFP+ cells were calculated within the CD45+ Lin+ populations (represented with different shapes of dots) in different organs (represented by different dot colors). The analysis was performed on PB 14 weeks post transplantation and on spleen and BM at the end of the experiments. Only mice displaying ≥0.1% GFP+ cells are represented in the graph (n=2 independent experiments). Note that when using the improved protocols for TI GFP+ cells are found in multiple lineages in all mice. d, Analysis of the primitive human compartment in the BM of transplanted mice from (c). (Top) Gating strategy used to define progenitors (CD34+ CD38+), multilymphoid progenitor (MLPs, CD34+ CD38− CD90low/− CD45RA+), multipotent progenitors (MPPs, CD34+ CD38− CD90− CD45RA−) and HSCs (CD34+ CD38− CD90+ CD45RA−). (Bottom) Percentages of GFP+ cells within the defined populations. Means ± SEM (48h SR1, n=4; 48h PGE2, n=3; 48h PGE2 SR1, n=5). e, DNA from total BM cells of transplanted mice was analyzed by PCR to determine TI into IL2RG. Each column represents one mouse. The schematics show the different sets of primers used to detect on target insertions mediated by HDR or NHEJ (the latter shown with the vector in sense or reverse orientation with respect to IL2RG). While evidence of HDR-mediated insertion of the cassette was retrieved from all mice assayed, there was also indication of some NHEJ-mediated integration of the donor IDLV. Trapping of IDLV at sites of NHEJ has been previously reported¹⁶. We note that our strategy could be adjusted to also exploit this type of insertion to drive transgene expression²⁴ and potentially increase the overall efficiency of gene correction (see schematic in Fig.4a). In such case, one should target insertion within an intron of the gene so that the splice acceptor site of the corrective cDNA is next in line for processing with the splice donor site of the upstream endogenous exon and any intervening sequence can be spliced out from the chimeric transcript leading to reconstitution of a functional open reading frame (provided that the insertion occurred in the same orientation as gene transcription). An additional benefit of targeting an intronic sequence would be to spare exons from disruption by NHEJ, although this is of minor concern when dealing with already defective alleles. f, 15-23 weeks after the primary transplant, human CD34+ cells were purified from the BM of 11 mice from (c) (mice were chosen among those best engrafted with GFP+ cells in the different CD34+ cells treatment groups) and transplanted (one mouse to one mouse) into 7-11 weeks old NSG mice. Secondary recipient mice were monitored for engraftment of human CD45+ and GFP+ cells at 8 and 12 weeks post-transplant in PB, and on BM at the end of the experiments. (Top left) Percentages of human cells in the indicated organs. (Top right) Percentages of the indicated lineages within the human cells. B cells were defined by expression of CD19 and myeloid cells were defined by expression of CD13 in PB or CD33 in BM. Dots represent individual mice. (Bottom left) Percentages of GFP+ cells within the human graft in the indicated organs and (bottom right) within the indicated lineages, as on top. g, Human cells from the BM of the top 2 engrafted secondary recipient mice from (f) were analyzed by PCR for TI into IL2RG.

Extended data Figure 5. Expansion of lymphoid cells in the transplanted mice after tumor challenge.

a, (Left) Percentage of human cells in the PB of mice transplanted with male CD34+ cells treated as indicated for TI into IL2RG. Analyses performed at the time of tumor injection (top panel) and 3 weeks later (bottom panel). (Right) Percentages of T and NK cells (CD3+ and CD16/56+ cells, respectively) measured within the human CD45+ cells in PB. b, Fold change in the absolute number of the indicated lineages in PB 3 weeks after tumor challenge. Means ± SEM (24h SR1, n=5; 48h SR1, n=6; 48h, n=5). c, Counts of GFP+ and GFP− T and NK cells in the PB of transplanted NSG mice before (top) and 3 weeks after (bottom) injection of the MDA-MB 231 tumor cell line engineered to express human IL-7, IL-15 and GM-CSF. Fold changes calculated from these values are plotted in Fig. 4e. d, Tumor growth in mice transplanted (n=16) or not (n=3) with treated CD34+ cells. ****p<0.0001 (two-way Anova).

Extended data Figure 6. Generation of T cells with substantial Vβ TCR diversity from the engrafted gene-targeted HSPC.

a, Analysis of TCR Vβ repertoire performed on PBMCs from a healthy donor. The histogram shows the frequency distribution of the different complementarity-determining region 3 (CDR3) lengths identified within the indicated Vβ families. As expected from a highly polyclonal TCR repertoire, all Vβ families display a Gaussian distribution of the CDR3 lengths. b, Frequency distribution of the different CDR3 regions measured as in (a) for the GFP+ (left) and GFP− (right) T cells harvested from the transplanted C1, C5 and A4 mice from Extended Data Fig. 5c. c, Complexity score³⁶ assigned to each Vβ family for the samples showed in (b). The sum of the scores for all the family is plotted in Fig. 4g. Note that all the samples analyzed display similar TCR Vβ repertoire distributions, constrained for some families and more polyclonal for others, as might be expected for human T cells developed in hematochimeric mice⁴¹, and that no significant differences are observed between the GFP+ and GFP− cells. This finding indicates that the rate of gene targeting achieved in the transplanted stem/progenitor cells does not detectably limit the generation of a polyclonal and functional T cell repertoire in vivo.

Extended data Figure 7. Functional and phenotypic characterization of IL2RG edited T cells harvested from transplanted NSG mice.

a, Graph showing the viability of GFP+ and GFP− T cells harvested from the transplanted NSG mice (from Extended Data Fig. 6) or of T cells from PB of healthy donor (HD T-cells), cultured in presence (+ cyto) or absence (no cyto) of human IL-7 and IL-15. b, (Left) Representative density plots of GFP+ and GFP− T cells harvested from mice, stained for CD8 (left) and CD4 (right). (Right) GFP+ and GFP− T cells harvested from mice were activated ex vivo with beads coated with anti CD3 and CD28 specific antibodies, and cultured with IL-7 and IL-15. CD4 and CD8 composition of GFP+ and GFP− cells, measured during ex vivo culture is shown (n=3). c, Surface phenotype of CD4 and CD8 T cells from (b) and HD T-cells at day 19 after stimulation. A representative plot (left) and histograms with medians + SEM (right) are shown. T stem memory cells (TSCM) are defined as CD62L+ CD45RA+ ^42,43, T central memory (TCM) as CD62L+ CD45RA−, T effector memory (TEM) as CD62L− CD45RA− and terminal effectors (TEMRA) as CD62L− CD45RA+. d, Production of IL-2 and IFNγ by GFP + or GFP− T cells as in (b), and by HD-T cells after 6 hr stimulation with PMA+Ionomycin. A representative plot (left) and percentages of IL2+ and IFNγ+ cells are shown. p=ns (unpaired t test). e, IFNγ Elispot assay showing the frequencies of IFNγ producing cells from (b) challenged with the MDA-MB 231 tumor cell line at different effector to target ratios. PHA stimulation was used as positive control.

Extended data Figure 8. Functionality of γ-chain dependent signaling pathway in IL2RG gene edited T cells.

GFP+ or GFP− T cells from the transplanted mice (as in Extended data Fig. 6 and 7) or T cells from the PB of healthy donor (HD T-cells) were exposed to the indicated doses of γ-chain related cytokines. The phosphorylation levels of STAT5 on Y694 (pSTAT5), STAT3 on Y705 (pSTAT3) and AKT on S473 (pAKT) were measured at the indicated time points by flow cytometry analyses. a, (Top) Representative plots showing pSTAT5, pSTAT3 and pAKT. Each time point of analysis is labeled by a different concentration of the intracellular dye PBSE. (Bottom) Fold changes in the levels of pSTAT5, pSTAT3 and pAKT relative to the time 0 of stimulation with the indicated maximal doses of IL-15 and IL-2 (n=3). Phosphorylation of STAT3 was used as a specificity control. p=ns (two-way Anova). b, Table showing the fold changes in the levels of pSTAT5 which are graphically represented in Fig. 4k. Statistics in the first column indicate significant change during time. ****p<0.0001, ***p<0.001 (two-way Anova). c, Heatmap representing fold changes in pAKT and pSTAT3 levels after the indicated time of exposure (min) to increasing amounts of IL-2 or IL-15. p=ns between GFP+ and GFP− cells (two-way Anova).

Extended data Figure 9. Indels quantification on the intended “on” target and candidate “off” target genomic sites of IL2RG ZFNs.

a, Indels quantification expressed as percentage of the total number of reads for each sample and target. The untreated (UT) sample F is shown in the rightmost column. 0-values underlie undetected events due to low sequencing reads coverage (see Supplementary Information). b, Table of p-values obtained comparing the number of indels for each target and its untreated control sample (Fisher exact test for contingency data). p-values allowed distinguishing real accumulation of indels (labeled with a green arrow close to the p-value) from background noise (marked with a yellow horizontal bar). Only SCARB1 and SLC31A1 show indels in samples A, B and D at a frequency significantly higher than the untreated sample. c, Indels frequency distribution along the amplified genomic sequences of the positive control IL2RG and the two loci that showed a low but significant off target activity, SCARB1 and SLC31A1. The x-axis represents the amplified sequence in base pairs while the y-axis shows for each base the percentage of reads that reported indels for each sample after noise subtraction (see Supplementary Informations). Note that indels mainly occur in the central region of the amplicon, corresponding to the spacer between the genomic sequences expected to be bound by the ZFNs. d, Representative sequence alignments of retrieved indels in IL2RG (sample A), SCARB1 (from samples A and B) and SLC31A1 (from samples A and D) focusing the analysis on the region bearing identity or homology to the intended ZFNs target sequence. For each sequence type the relative frequency of retrieval in the sample is reported as percentage. As showed in (c) indels mostly occur in the central spacer region between the 2 ZFN binding sites (underlined).

Extended data Figure 10. IL2RG gene correction in CD34+ cells from the BM of a genotyped subject with SCID-X1.

a, Blood cell counts in the PB and BM of a SCID-X1 male child carrying the R289X mutation in the IL2RG gene (see Supplementary Informations) and showing virtual absence of T and NK cells. * in the leftmost table indicate values calculated within the leukocyte gate. b, (Left) Representative density plots showing the cellular composition of a BM harvest from a healthy donor (top) and the subject with SCID-X1 (bottom) after purification of nucleated cells (bone-marrow-derived mononuclear cells, BMMC). Myeloid cells are stained for CD15 and CD33, B cells for CD19, T cells for CD3 and NK cells for CD16. (Right) γ-chain expression within the indicated cell populations gated from the plots shown on left. As expected from the missense R289X mutation, the γ-chain protein is expressed on the cell surface but it is not functional, as indicated by the absence of T and NK cells in the patient. In Fig. 5 we show normal expression of the γ-chain protein in the myeloid cell progeny of 3 gene corrected CFC from the patient cells (gene correction was proven by evidence of targeted integration in the only IL2RG allele of this male individual in the clonal CFC progeny and by the expression of the fusion transcript bearing the corrective cDNA). These data indicate normal γ-chain expression from the reconstituted allele in the edited patient cells. c, Bright field and fluorescence microscopy images of the 3 GFP+ myeloid colonies obtained after IL2RG gene correction of SCID-X1 cells.