Efficient and sustained FOXP3 locus editing in hematopoietic stem cells as a therapeutic approach for IPEX syndrome

Immune dysregulation, polyendocrinopathy, enteropathy, X-linked (IPEX) syndrome is a monogenic disorder caused by mutations in the FOXP3 gene, required for generation of regulatory T (Treg) cells. Loss of Treg cells leads to immune dysregulation characterized by multi-organ autoimmunity and early mortality. Hematopoietic stem cell (HSC) transplantation can be curative, but success is limited by autoimmune complications, donor availability and/or graft-vs.-host disease. Correction of FOXP3 in autologous HSC utilizing a homology-directed repair (HDR)-based platform may provide a safer alternative therapy. Here, we demonstrate efficient editing of FOXP3 utilizing co-delivery of Cas9 ribonucleoprotein complexes and adeno-associated viral vectors to achieve HDR rates of >40% in vitro using mobilized CD34+ cells from multiple donors. Using this approach to deliver either a GFP or a FOXP3 cDNA donor cassette, we demonstrate sustained bone marrow engraftment of approximately 10% of HDR-edited cells in immune-deficient recipient mice at 16 weeks post-transplant. Further, we show targeted integration of FOXP3 cDNA in CD34+ cells from an IPEX patient and expression of the introduced FOXP3 transcript in gene-edited primary T cells from both healthy individuals and IPEX patients. Our combined findings suggest that refinement of this approach is likely to provide future clinical benefit in IPEX.


INTRODUCTION
Immune dysregulation, polyendocrinopathy, enteropathy, X-linked (IPEX) syndrome is a rare monogenic primary immunodeficiency, characterized by the loss of functional regulatory T (T reg ) cells crucial for controlling immune responses against self and foreign antigens.The syndrome was first described in 1982 1 in a family with several affected males and the responsible gene, Forkhead box P3 (FOXP3), was identified several years later. 2,3FOXP3 is the lineagedefining transcription factor of thymically derived T reg cells and is essential for both T reg cell development and function.Absent or dysfunctional T reg cell in IPEX patients leads to the failure to maintain peripheral immune tolerance, resulting in the early onset of multisystem autoimmunity with features including, most commonly, severe inflammatory bowel disease, type 1 diabetes mellitus, thyroid disease, and eczema.Supportive immunosuppressive therapies can modulate disease, but are not curative and are associated with multiple complications.Alternatively, allogeneic hematopoietic stem cell transplantation (HSCT) represents a potentially curative approach that can eliminate autoimmune manifestations. 4However, while HSCT can be highly beneficial for IPEX, limitations in donor matching and transplant complications in the setting of severe immunologic dysregulation make implementation of this approach highly challenging. 4,5ior work has demonstrated the efficacy of lentiviruses (LVs) expressing FOXP3 from a constitutively active EF1a promoter to establish a T reg cell phenotype in IPEX CD4 + T cells. 6While potentially advantageous for applications where adoptive transfer of T reg cells might be beneficial, 7,8 adoptively transferred LV-treated T cells are unlikely to provide a long-term cure for IPEX due to the inability to persist in vivo over long periods of time.Gene therapy of murine HSCs using LV with the endogenous FOXP3 promoter driving FOXP3 cDNA expression have also shown promise by rescuing the autoimmune phenotype in scurfy mice, the murine equivalent of IPEX. 9 This approach, however, will require LVs that can precisely replicate the complex endogenous control elements within the FOXP3 locus capable of initiating and sustaining endogenous levels of FOXP3 expression, while limiting both vector silencing and genotoxicity risk due to the random nature of LV integration.CRISPR-Cas9-based gene editing 10,11 of IPEX patient CD34 + hematopoietic stem and progenitor cells (HSPCs) offers an alternative therapeutic option.Homology-directed repair (HDR)-based editing enables insertion of functional FOXP3 cDNA sequence into the endogenous locus while preserving adjacent sequence elements required for expression and ensuring appropriate copy number.Gene editing for IPEX is particularly appealing since locus-and lineagespecific regulation of FOXP3 is critical for stable FOXP3 expression. 12,13Thus, preserving the natural genomic landscape at the locus in HDR-edited HSPCs is paramount to enable appropriate differentiation and acquisition of T reg cell fate.
While in vitro editing has been reported at high rates in HSPCs sourced from umbilical cord blood (CB) or isolated from peripheral blood of granulocyte colony stimulating factor (G-CSF) mobilized adults (mPB), engraftment of HDR-edited HSPCs in immune deficient animals remains challenging due to the difficulty in targeting long-term repopulating HSCs (LT-HSCs); a feature that is particularly evident when using mPB HSPCs. 146][17][18] Thus, engraftment of even a limited proportion of successfully HDR-edited HSC is predicted to provide clinical benefit in the setting of this profound immune disorder.
We have recently demonstrated highly efficient methods to edit the FOXP3 locus in primary T cells enabling conversion to thymus T reg (tT reg )-like cells capable of mediating immunosuppression in the setting of autoimmunity. 19Another group has published HDR-based editing of FOXP3, although the majority of experiments utilize CB-HSPCs rather than more clinically relevant mPB CD34 + HSPCs. 20ere, we present a targeted gene editing strategy to incorporate the codon diverged coding region for FOXP3 gene in mPB HSPCs and test their ability to engraft in immune deficient mice.Further, we demonstrate efficient HDR-editing of IPEX patient CD34 + cells and verify the expression of codon-diverged FOXP3 transcripts in HDR-edited primary CD4 + T cells derived from both healthy controls and IPEX patients.

Optimization of culture conditions for editing CD34 + cells at the FOXP3 locus
The FOXP3 gene comprises 11 coding exons and mutations reported in IPEX patients have been identified throughout the entire gene. 21To develop an editing strategy that can work as a universal cure for IPEX patients, we elected to insert a functional FOXP3 cDNA within the first coding exon, to enable correction of all IPEX patients except those bearing mutations within the promoter region. 22,23This strat-egy was also utilized to disrupt the mutant FOXP3 allele and eliminate the potential for expression of a dominant negative mutant protein.
We used CRISPR single-guide RNA (sgRNA) T9, 19 and a second sgRNA T3, to target the first coding exon of the FOXP3 gene (Figure 1A).For comparison, the sgRNA utilized by Goodwin et al. 20 binds at the intersection of 5 0 UTR/first coding exon and 47 and 125 nucleotides away from sgRNA T9 and T3, respectively.The sgRNAs were complexed with SpyFi Cas9 24 to form ribonucleic proteins (hereafter referred to as RNPs) then electroporated into HSPCs using the Neon electroporation system.Analysis of gDNA by droplet digital PCR (ddPCR) revealed that T3 RNP induced insertions/deletions (indels) in approximately 93%, while T9 targeted approximately 52% of FOXP3 alleles, respectively (Figure 1B), indicating superior on-target cleavage with T3 RNP in CD34 + cells.
To orchestrate and efficiently track HDR, recombinant adeno-associated virus (rAAV) donor templates were first designed to insert an expression cassette containing a constitutive myeloproliferative sarcoma virus enhancer, negative control region deleted, dl587rev primer-binding site substituted (MND) promoter driven GFP cDNA sequence followed by shortened woodchuck hepatitis posttranscriptional regulatory element (WPRE3) and Simian virus 40 polyadenylation (SV40 polyA) sequences.Flanking the expression cassette, 0.8-kb homology arms centered specifically on the respective sgRNA were designed such that a minimum deletion was created after targeted integration (Figure 1A).Editing with either sgRNA and sgRNA-specific AAV HDR donor combination led to robust HDR rates.As anticipated, we observed 4-fold higher targeted modification using T3-(GFP-T3) or T9-specific (GFP-T9) AAV donors (Figures S1A and S1C) when compared with a common AAV.GFP donor that functioned with either sgRNA, findings that also demonstrated no significant negative impacts on cell viability (Figures S1B  and S1D).For the GFP-T3 vector, no enhancement in HDR was observed as vector dose was increased (multiplicity of infections [MOI] from 250 to 1,000), while a dose-dependent increase was observed with the GFP-T9 AAV donor (Figures S1A and S1C).Since absolute HDR rates were considerably higher at 51% for the T3 RNP and GFP-T3 AAV (vs.8% using T9 RNP), we elected to utilize the T3 RNP for all subsequent experiments.HSCs are largely quiescent, thereby limiting DNA damage due to replication and mitosis.Because HDR pathways are active primarily during G2 and S phases, HSCs are more likely to undergo non-homologous end-joining (NHEJ) upon introduction of a double-strand depicted by dashed lines.pA, SV40 polyadenylation sequence; W, WPRE3 element.(B) Average allelic disruption (% NHEJ) rates observed upon transfection of adult mobilized human CD34 + HSPCs with T3 (n = 3 male donors, 3 independent studies) or T9 (n = 3 male donors, 3 independent studies) RNPs quantified via ddPCR.Significance determined by Mann-Whitney U test.(C) Timeline of procedures for in vitro culturing and editing of adult mobilized CD34 + HSPCs using protocol A (top) and B (bottom).(D) Viability measured by flow cytometry forward and side scatter 1 day after editing with T3 RNP and rAAV6 targeting vector employing either protocol A (n = 6 male donors, 12 independent studies) or protocol B (n = 5 male donors, 1 female donor, 9 independent studies).Significance determined by Kruskal-Wallis test.(E) Representative flow cytometry plots depicting cell viabilities on day 1 and %GFP + cells on day 5 in HSPCs edited with protocol A (left) or B (right).(F) Targeted integration rates measured by GFP-high 5 days after editing with T3 RNP and rAAV6 targeting vector cultured with protocol A (n = 4 male donors, 4 independent studies) or protocol B (n = 5 male donors, 1 female donor, 8 independent studies).Significance determined by Mann-Whitney U test.(G) NHEJ 5 days after editing in cells cultured with protocol A or B and transfected with T3 RNP (n = 3 male donors in 4 independent studies).Wilcoxon matched-pairs signed ranked test.Bar graphs represent mean ± SEM. break (DSB). 25,26To facilitate entry of HSC into cell cycle and repair of DSBs by HDR utilizing the AAV donor, we first pre-stimulated CD34 + cells in culture media with cytokines.Of note, low-density cultures have also been reported to promote HSC expansion and cycling. 27Additionally, SR1 and UM171, compounds reported to support HSC expansion 28 and self-renewal, 29 were included in protocol B. To explore the impact of alternative culture densities and media on the indel and HDR frequencies, CD34 + cells were plated in media using a higher cell density (protocol A) vs. lower density protocol (protocol B), as outlined in Figure 1C.Specific differences between the two protocols are summarized in Table S1.For protocol B, the nucleofection was performed using the optimized program CM-149 (Figure S2).Flow cytometry analysis 1 day after editing revealed slightly higher cell viabilities in both mock and edited cells for protocol B (mock 79%, edited 72%) vs. protocol A (mock 72%, edited 63%) (Figure 1D).To assess potential differences in inducing indel and HDR frequencies, T3 RNPs and GFP-T3 AAV donor vectors were introduced into cells cultured using both protocols followed by flow cytometry to assess HDR rates based on % GFP + cells at day 5 after editing.HDR rates were also subsequently assessed via ddPCR.A significantly higher average GFP expression was observed in the CD34 + cells cultured with protocol B averaging 44% compared with 28% for protocol A (Figures 1E and 1F), despite equivalent percent NHEJ rates (93%) using either protocol (Figure 1G).

Edited HSPCs engraft long-term in NBSGW mice and undergo multi-lineage differentiation
To determine whether editing protocols influenced the long-term repopulation potential of CD34 + cells, mock or GFP-T3 AAV plus T3 RNP-edited CD34 + cells were transplanted into busulfan-treated, 8to 10-week-old NBSGW recipient mice (Figure 2A).Human cell chimerism was evaluated by sacrificing the mice at 12-16 weeks after transplant and analyzing bone marrow (BM) and spleens for engrafted CD45 + human cells and GFP + edited cells.Similar percentages of human cells were present within the BM for recipients of cells cultured with either protocol (protocol A: mock, 65%; edited 62%; and protocol B: mock 72%; edited 58%) (Figures 2B and S3A).However, the average %GFP + cells observed within the BM were 1.6-fold higher when protocol B was used (average 8%; range, 0.2%-26.0%)compared with protocol A (average 5%; range, 1%-29%) (Figure 2C).The distribution of CD19 + B cells and CD33 + myeloid cells were similar across protocols and groups (Figure 2D).The %GFP + cells within both compartments was slightly higher with protocol B compared with protocol A and CD33 + cells exhibited the highest %GFP + cells (Figures 2E and 2F).
The mean human CD45 + cell engraftment within the spleen was also comparable between groups, ranging from 12% to 19%, with no significant differences between protocols (Figure S4A).Matching the trend observed in the BM, GFP + cells were higher for protocol B compared with protocol A edited cells, averaging 10% and 4%, respectively (Figure S4B).The proportion of CD33 + cells were comparable across all groups.Interestingly, proportion of CD19 + B cells in both mock and edited groups was higher using protocol B (Figure S4C) and significantly higher %GFP + cells were observed for protocol B, 5% compared with 2% for protocol A (Figure S4D).
A key obstacle for developing durable HDR-based gene editing therapies using HSPCs is the inherent quiescence of LT-HSCs that serves as a protective mechanism against endogenous stress. 30To assess the impact of alternative culture conditions on HDR editing in HSCs, we analyzed human cells within the BM of NBSGW mice for expression of surface markers that define a more primitive, engraftment-enriched subset of HSPCs, defined by CD34 + CD38 low (Figure S3B).The proportion CD34 + CD38 low HSPCs was approximately 2-fold higher with use of protocol B (5%-6%), irrespective of editing (Figure 2G).Strikingly, analysis of GFP + cells within CD34 + CD38 low compartment revealed a 3-fold higher proportion of HDR-edited cells for protocol B, 13% compared with 5% for protocol A (Figure 2H).These findings demonstrated that use of UM171 and SR1, in association with the low-density culturing in protocol B, enabled superior long-term engraftment of HSPCs and increased HDR editing efficiency in the CD34 + CD38 low compartment and was selected for use in all ensuing studies.

cDNA-edited HSPCs differentiate into T lymphocytes that maintain targeted integration
Next, we sought to determine whether cDNA edited CD34 + HSPCs retain the potential to develop into naive CD4 T cells, progenitors capable of differentiation into the T reg cell lineage in vivo in response to critical differentiation cues.As the adult NBSGW model used in Figures 2 and 3 does not support robust T cell  differentiation, we utilized two parallel differentiation strategies: OP9-DL1 co-culture 31 and artificial thymic organoids 32 to differentiate cDNA-edited human HSPCs toward T cell lineage in vitro.OP9-DL1 cells are stromal cells (derived from the M-CSF deficient op/op mouse) engineered to express the Notch ligand Delta-like1 (DL1) that can be used to facilitate differentiation of HSPCs into T lymphocytes in response to a combination of key cytokines and notch-dependent signals.As CB-CD34 + cells are more amenable to these differentiation protocols, we cultured and edited healthy donor CB-CD34 + in protocol B conditions.We observed cell viabilities comparable to mPB CD34 + 24 h after editing (Figure 4A, left).Notably, the average HDR editing rates were moderately higher in CB-CD34 + HSPCs than observed in mPB CD34 + HSPCs (Figure 4A, right).One day after editing, mock-treated and AAV.FOXP3.cDNA-editedCD34 + were put into differentiation systems as outlined in Figures 4B and S7A.At the same time, mocktreated and edited cells were analyzed by flow cytometry for surface expression of CD34, CD45, CD19, CD56, CD14, CD5, CD7, CD1a, CD3, and TCRaß to confirm CD34 + purity and the absence of contaminating T cells or other lin + cells (Figure S7B).A small amount of mock-treated and edited cells were kept in culture until 5 days after editing to determine the HDR rate in edited cells that were introduced into the differentiation systems.
Cultures were maintained for 28 days then analyzed by flow cytometry for T lineage differentiation (Figures 4C and S7C).After 28 days in co-culture, differentiated cells predominantly reached CD34 -CD5 + CD7 + pre-T-1 stage (Figures 4D and S7D).Importantly, there was no observable differences between the proportion of pre-T-1 cells in mock-treated compared with edited conditions.Importantly, in the ATO system, we observed a proportion of cells reaching the CD4 + CD8 + double-positive and CD4 + or CD8 + single-positive developmental stages.In addition, we observed a modest amount of CD3 + TCRab + cells in comparable efficiency across mock-treated and AAV+RNP-treated conditions (Figure 4D), consistent with others finding of superior differentiation in the ATO system. 32mportantly, the proportion of HDR-edited cells at termination of OP9-DL1 differentiation closely matched the input HDR frequency (Figure S7E).We observed greater variation in the proportion of HDR-edited cells recovered after differentiation in the ATO system (Figure 4E).We believe the primary cause of high variability observed in the ATO system reflects the much lower number of input cells used (5,000) relative to OP9-DL1 (250,000) and potential skewing due to limited numbers of lymphoid progenitors and HDR-edited lymphoid progenitors in this smaller input sample.
Together, these results demonstrate that mock-edited and HDR-edited CB-CD34 + HSPCs exhibit similar T lineage differentiation capac-ity and that the overall proportion of AAV.FOXP3.cDNAHDR-edited cells remains stable during commitment to the T lineage.
Assessment of potential off-target cleavage sites for FOXP3 sgRNA T3 The off-target sites for T3 guide were predicted in silico by CCtop-CRISPR-Cas9 target online predictor (Table S2).The top five offtarget predicted sites were then interrogated using the Miseq platform.Two of the off-target sites (OT1 and OT4) were within the exons of genes.The top off-target site was Disheveled Binding Antagonist of Beta Catenin 2 (DACT2), a protein involved in intracellular signaling pathways during development.The OT4 site is located within the Exostosin Like Glycosyltransferase 1 gene (EXTL1), a member of the multiple exostoses family of glycosyltransferases involved in the chain polymerization of heparan sulfate and heparin.The off-target NHEJ rates at both OT1 and OT4 were 0.1%, equivalent to that observed for the mock sample that received no RNPs.The highest off-target cleavage (0.9%) was seen at the OT-II in Solute Carrier Family 2 Member 1 gene, which is the major glucose transporter in the mammalian blood-brain barrier (Figure S6A).High rates of NHEJ were observed at the FOXP3 locus (83%), which were equivalent to those determined by the ddPCR assay (94%).The majority of the NHEJ events at the FOXP3 locus were deletions (80%), followed by insertions (15%) and substitutions (10%) (Figures S8A and S8B).A high percentage of NHEJ edits were six nucleotide deletions that constituted 15% of the total NHEJ events (Figures S8C and S8D).Thus, the off-target cleavage rates were significantly lower at <1% for the five off-target sites, while high NHEJ edits were observed at the FOXP3 locus, confirming that our sgRNAs is largely specific for the target locus.

Efficient editing of IPEX patient CD34 + cells
We obtained a small CB specimen from an IPEX patient bearing an I363V missense mutation located in the FOXP3 forkhead domain that renders the protein incapable of establishing the T reg cell transcriptional program.CD34 + cells were isolated from the frozen sample and editing reagents introduced after 2 days of culturing and prestimulation using protocol B. Notably, ddPCR analysis revealed average HDR editing rate of 34%, although significant differences were observed between the two studies performed, likely reflecting the limited cell numbers and relatively lower viability of the CB sample (Figure 5A).
T conv cells, and healthy donor natural T reg (nT reg ) cells following the experimental timeline outlined in Figure 5B and AAV.FOXP3.cDNAvector diagrammed in Figure 3A.Five days after editing, gDNA was extracted and HDR was quantified via ddPCR.HDR rates were similar across groups: T conv cells (healthy) 50%, T conv cells (IPEX) 52%, and nT reg cells (healthy) 56% (Figure 5C).Fourteen days after editing, we extracted RNA, synthesized cDNA, and quantified transcript levels via ddPCR.Because we utilized a codon-diverged FOXP3 sequence in the AAV.FOXP3.cDNAvector, we were able to distinguish between endogenous and codon-diverged FOXP3 expression.As we achieved approximately 50% editing rates, we predicted approximately equal proportions of coFOXP3 and endoFOXP3 transcripts.However, in edited T conv cells (healthy), T conv cells (IPEX), and nT reg cells (healthy) we measured a ratio of coFOXP3:endogenous FOXP3 transcripts of 0.32, 0.42, and 0.37 respectively (Figures 5D-5F).The reduced level of coFOXP3 transcripts in comparison with endogenous transcripts likely reflects differences in RNA processing efficiency of intron-less FOXP3 cDNA donor vs. the endogenous transcripts.
Next, we performed studies designed to identify FOXP3 protein expression mediated by the FOXP3 cDNA cassette following editing of an enriched nT reg cell population.To specifically identify FOXP3 expression in HDR-edited nT reg cells, we created a FOXP3.cDNA.P2A.GFP donor cassette where cis-linked GFP expression would permit precise flow-based identification of HDR-edited cells.Seven days after editing, we quantified GFP and FOXP3 levels in mock-treated, FOXP3 RNP (knockout), and FOXP3.GFP-edited nT reg cells (Figure 5G).As predicted, GFP expression was observed only in the FOXP3.GFP-edited nT reg cell population.We observed efficient ($77%) FOXP3 knockout in the RNP-treated nT reg cells and a low level of FOXP3 restoration in FOXP3.GFP-edited nT reg cells (Figure 5H).Importantly, within the GFP + population in the FOXP3.GFP-edited nT reg cell population, the majority of cells were FOXP3 + ; and this population exhibited approximately 60% the mean fluorescence intensity of wildtype FOXP3 (Figure 5I).We attempted to perform suppression assays using fluorescence-activated cell sorting (FACS)-sorted GFP + populations.However, due to limited cell yields and purity loss during expansion, we did not obtain sufficient viable GFP + T reg cells for functional assays.
Together, these data demonstrate FOXP3 promoter-mediated transcription of coFOXP3 in HDR-edited nT reg cells from healthy control subjects and T conv cells isolated from both IPEX patients and healthy control subjects; we also directly demonstrate expression of exogenous FOXP3 protein in HDR-edited healthy control nT reg cells, albeit at sub-endogenous levels.

DISCUSSION
IPEX syndrome, caused by FOXP3 mutations, is a devastating disease that leads to substantial mortality.The requirement for strict epigenetic regulation of FOXP3 necessitates development of a therapeutic approach that preserves the endogenous control elements required to orchestrate both thymus-dependent T reg cell lineage differentiation and maintenance of the T reg cell program in vivo.One approach for the treatment of IPEX is enforced expression of FOXP3 in CD4 + T cells by delivering its coding sequence driven via a robust promoter. 19While such autologous engineered T reg cells may provide temporary clinical benefit, replicating the critical steps in Treg cell lineage programming and selection and generation over time requires thymic repopulation with gene-corrected lymphoid progenitors derived from LT-HSC.Thus, to achieve a cure for IPEX necessitates editing of CD34 + HSC and seamless introduction of the FOXP3 cDNA under transcriptional control of the endogenous promoter.Through optimization of CD34 + culture and HDR editing protocols, we demonstrate efficient insertion of FOXP3 cDNA at the FOXP3 locus and sustained engraftment of cDNA-edited cells in vivo.We show that this approach is feasible using clinically relevant, mobilized CD34 + HSPCs from multiple healthy donors as well as in CD34 + cells derived from an IPEX subject.Further, we show that cDNA-edited CD34 + cells retain the capacity to differentiate into T lineage cells and retain the FOXP3 cDNA cassette in vitro.[17][18] Culturing and editing conditions can have a significant impact on the HDR rates and engraftment potential of HSPCs.Culturing CD34 + cells using a low-density editing protocol and small molecules that support HSC survival (protocol B) enabled higher rates of HDR in the edited cells (>40%), while no differences were observed in the .cDNAediting efficiencies in healthy donor-derived T conv cells (n = 8 male donors, 6 independent studies), healthy donorderived T reg cells (4 male donors, 4 independent studies), and IPEX patient-derived T conv (n = 2 male donors, 2 independent studies) quantified by ddPCR.(D-F) Endogenous (endo) and codon optimized (co) FOXP3 transcript levels in mock-treated and RNP + AAV.FOXP3.cDNAedited.(D) Healthy donor-derived T conv cells (n = 8 male donors, 6 independent studies).(E) IPEX patient T conv cells (n = 2 male donors, 2 independent studies).(F) Healthy donor-derived T reg cells (n = 3 male donors, 2 independent studies) (transcripts quantified by ddPCR and normalized to HPRT control transcript).(G) Representative flow plots of FOXP3 + and GFP + populations in nT reg cells isolated from healthy donors.Comparison of unstained mock treated and FOXP3-stained mock-treated, RNP-treated, and RNP + AAV.FOXP3.cDNA.GFP-treated 7 days after editing.(H) Proportion of FOXP3 knockout (RNP-treated nT reg cells) and FOXP3 restoration (RNP + AAV.FOXP3.cDNA.GFP edited) relative to mock treated.Gated on CD4 + CD25 + , representative plot highlighted in yellow in (G) (n = 3 male donors, 2 independent studies).(I) Mean fluorescence intensity (MFI) of edited FOXP3 (RNP + AAV.FOXP3.cDNA.GFP-treated nT reg cells, gated on CD4 + CD25 + GFP + ) compared with endogenous FOXP3 + MFI (mock-treated, gated on CD4 + CD25 + ).Representative plot highlighted in green in (G).(n = 3 male donors, 2 independent studies).Bar graphs represent mean ± SEM.
rates of NHEJ edits between the two protocols.Upon transplantation of edited cells into humanized mice, a 2-fold higher engraftment of HDR-edited cells was observed with protocol B compared with protocol A edited cells.A similar low-density protocol was previously reported to improve HDR rates within the LT-HSC population by enforcing G2M or S phases of the cell cycle and facilitated higher engraftment of edited cells in vivo. 27We also observed a 2-fold higher percentage of HSC-enriched HSPCs and a 3-fold higher proportion of HDR-edited (GFP + ) cells within the HSC-enriched HSPC gate recovered from mice further corroborating this hypothesis.Normal lineage distribution was observed with cells edited with either protocol compared with mock controls, suggesting that editing did not negatively impact differentiation in vivo.
Recently reported methods to improve editing and engraftment of HDR-edited LT-HSCs include co-transfection of select mRNAs during nuclease delivery.Ferrari et al. 33 have shown improved engraftment of gene-edited LT-HSCs by introduction of dominant negative P53 inhibitor along with adenoviral protein Ad5-E4orf6/7.Transient expression of Ad5-E4orf6/7 triggered an E2F-driven pleiotropic response that facilitated cell-cycle progression and expression of genes encoding for the HDR apparatus leading to increased editing within primitive CD34 + cells.In parallel, the dominant active P53 inhibitor (GSE56), 34 helped to preserve HSC survival or engraftment upon transplantation in immune deficient mice.While not tested in human HSC, another possible approach to enhance HDR is fusion of dominant-negative mutant of 53BP1 to Cas9. 3553BP1 enhances HDR by limiting DNA end resection and hindering recruitment of BRCA1 to the DNA cleavage site. 36,37By fusing Cas9 activity with 53BP1 inhibition, the authors were able to locally retard NHEJ at the site of the introduced DSB without causing global 53BP1 inhibition.In another study, co-delivery of an engineered ubiquitin variant of an inhibitor of 53BP1 (i53) 38 mRNA and GSE56 34 mRNA with editing reagents enhanced long term correction in X-MEN patient BMderived CD34 + cells. 39Robust engraftment of HDR-edited patient cells was observed in the BM of intrahepatic-transplanted, irradiated, newborn NSGS mice in a modulator-dependent manner.While effectively demonstrating modulator efficacy, it remains unclear whether this model accurately predicts HDR-edited HSPC engraftment capability in a clinical setting.
To validate expression and assess the levels of the introduced codon optimized transcript, we edited primary T lymphocytes and nT reg cells from healthy donors.Expression of codon optimized transcripts was readily detected in both HDR-edited T cells and T reg cells.Exogenous cDNA expression was highest in HDR edited T reg cells likely due to the open chromatin landscape in the region compared with conventional CD4 + T cells. 40,41Our HDR editing methodology also performed similarly using CD4 + T lymphocytes derived from two independent IPEX subjects including CB T cells (isolated from a subject with a I363V mutation) and peripheral blood CD4 + T (from a subject with a polyA region mutation).Despite efficient editing in both healthy control and IPEX patient-derived T cells, exogenous cDNA expression levels comprised between 32% and 42% of wildtype FOXP3.Consistent with our findings, in a separate study by Goodwin et al., 20 introduction of a FOXP3.cDNA.LNGFR cassette into T reg cells, resulted in sub-endogenous levels of FOXP3 protein expression in edited T reg cells compared with control T reg cells.Lower or absent protein expression has been reported when fully spliced cDNA is introduced into the first coding exon of a target gene. 20,42This approach precludes the splicing process, often required for optimal transcription and translation.Intronic sequences harbor regulatory elements and their interaction with the splicing machinery can play a critical role in modulating initiation/processivity by RNA polymerase II, pre-mRNA processing, and/or mRNA export. 43We speculate that, by redesigning AAV HDR cDNA donors to include alternate post-transcriptional elements such as the full WPRE element, a stronger polyadenylation signal and/or candidate intronic elements, that codon optimized FOXP3 transcript and protein expression will reach endogenous expression levels.Such modifications in donor design are likely to be required to achieve consistent T reg cell function.
Investigation of the FOXP3 T3 sgRNA cut site using NGS revealed the indel spectrum in edited cells.Larger deletions (>3 nucleotides) were favored over fewer than three nucleotide deletions which accounted for only 6% of the total NHEJ events.The indel signature of sgRNAs has been utilized by Tatiossian et al. 44 to predict the outcome of HDR frequency and further corroborated using donor templates demonstrating that a larger proportion of nucleotide deletions of more than three nucleotides favor HDR upon template introduction, as was seen with this specific sgRNA.Analysis of the top five predicted off-target sites revealed less than 1% off-target DSBs.Moving the FOXP3 T3 sgRNA toward therapeutic application, however, will necessitate additional unbiased off-target assessments such as GUIDE-seq or alternative methodologies. 45 HDR editing strategy similar to that described in our study was utilized to target FOXP3 locus in T cells (from healthy and IPEX donors) and healthy donor CB-derived HSPCs using CRISPR sgRNAs and rAAV6 vectors. 20Compared with our findings, Goodwin et al. 20 demonstrated a lower level of engraftment of HDR edited CB progenitors.Consistent with this limited engraftment, purified edited vs. unedited T cells derived from engrafted animals failed to demonstrate suppressive activity in vitro.In contrast, we focused primarily on editing and transplantation of HSPCs derived from apheresis of G-CSF-mobilized healthy donors-the HSC cell source anticipated to be utilized for clinical application.In parallel, as in Goodwin et al., 20 we demonstrate successful HDR editing in T cells from both healthy and IPEX donors.Importantly, here we also demonstrate successful editing of CD34 + cells from IPEX patients.As an alternative therapeutic approach, LV-mediated gene delivery of a FOXP3 expression cassette (utilizing the proximal FOXP3 promoter and conserved non-coding sequences and FOXP3 cDNA) into murine HSCs, followed by transplantation of purified T cells into neonatal scurfy mice (the murine equivalent of IPEX) was shown to limit disease. 9owever, heterogeneous expression correlating with viral copy number (VCN) was observed and high VCNs (>3) were required to reach a therapeutic threshold of transgene expression.Further, the critical conserved non-coding sequence 2 within the LV sequences did not retain endogenous methylation dynamics.Finally, LV gene therapy has other potential disadvantages including inability to control VCN or integration site, position-effect variegation, and the potential of insertional mutagenesis.
In summary, we demonstrate efficient HDR-based editing of the FOXP3 locus in control and IPEX patient CD34 + HSPCs.Further, we show that control HDR-edited HSPCs are capable of sustained engraftment in vivo in humanized mice.This editing methodology sets the foundation for developing a definitive therapy for IPEX patients.Incorporation of recent advances in the field and additional HDR donor design changes will likely assist in improving outcomes and paving the way for clinical translation.

MATERIALS AND METHODS
Reagent source and category numbers listed in Tables S3-S10 and rAAV6 sequences are listed in Table S11.
MS5-hDLL4 cells.MS5 murine stromal cells transduced with a lentiviral vector encoding human DLL4 were a provided by Dr. Gay Crooks (UCLA).Stable expression of DLL4 was confirmed by flow cytometry after multiple weeks in culture.MS5-DLL4 cells were cultured in DMEM (Gibco, Thermo Fisher Scientific, Waltham, MA, USA) + 10% FBS (Omega Scientific).

Primary cells Mobilized peripheral HSPCs and CB CD34 + cells from healthy donors.
Human CD34 + HSPCs enriched from mobilized PBMCs were obtained from the Cooperative Centers of Excellence in Hematology, Fred Hutchinson Cancer Research Center (supported by NIDDK Grant DK106829).CB from healthy donors was purchased from Bloodworks Northwest (Seattle, WA); CD34 + cells were isolated from CB using human CD34 MicroBead Kit (Miltenyi Biotec, Bergisch Gladback, Germany).
PBMC-derived CD4 + T lymphocytes and tT reg cells from healthy donors.PBMCs were obtained from the Cooperative Centers of Excellence in Hematology, Fred Hutchinson Cancer Research Center.Human primary CD4 + T cells and tT reg cells were isolated from thawed PBMCs using negative selection for CD4 and positive selection for CD4 + CD127 low CD25 + enrichment, respectively (both from STEMCELL Technologies).T cells were cultured in T cell media (RPMI 1640; Gibco) with 20% FBS (Omega Scientific), 10 mM HEPES (Gibco), 2 mM Glutamax (Gibco), 55 mM b-mercaptoethanol (Sigma-Aldrich, St. Louis, MO, USA) supplemented with IL-2 (50 ng/mL, Peprotech, Thermo Fisher Scientific).

IPEX patient samples
CB samples from IPEX patient with I363V mutation and PBMCs from IPEX patient with a polyA mutation (AAUAAA>AAUGAA within the endogenous poly A sequence) were obtained after informed consent using protocols approved by the institutional Review Board of Seattle Children's Research Institute.

Mouse strains
NBSGW mice (NOD.Cg-Kit WÀ41J Tyr + Prkdc scid Il2rg tm1Wjl /ThomJ, Stock 026622, Jackson Laboratory, Bar Harbor, ME, USA) used for the experiments were either purchased from Jackson Laboratory or inbred and maintained in the specific pathogen-free animal facility of the Seattle Children's Research Institute according to Institutional Animal Care and Use Committee and approved protocols.

AAV6 donor templates and vector production
In-Fusion HD cloning kit (Takara, Kusatsu, Japan) was used to insert PCR amplified fragments into pAAV.GFP (a gift from John T. Gray, Addgene plasmid #32395) replacing the GFP and a-globin polyadenylation site in the original vector.T3 and T9 FOXP3.MND.GFP-targeting vectors contain the MND promoter 46 upstream of a GFP cDNA, followed by WPRE3 47 and SV40 polyadenylation signal elements; this expression cassette was flanked 5 0 and 3 0 by 0.6 or 0.8 kb FOXP3 homology arms.The FOXP3.cDNA vector contains the codon-optimized FOXP3 cDNA cassette followed by WPRE3 and SV40pA, flanked by 0.8 kb FOXP3 homology arms on either side.The FOXP3 cDNA.P2A.GFP vector contains the codon-optimized FOXP3 cDNA cassette followed by a P2A ribosomal skip sequence followed by a promoter-less GFP cDNA sequence with WPRE3 and SV40pA elements.The FOXP3 cDNA.P2A.GFP contains the same 0.8-kb homology arms as the FOXP3 cDNA construct.AAV6 stocks were produced by transient transfection of HgT1-Adeno, Repcap6, 48 and vector plasmid into HEK 293T cells as previously described. 49Briefly, 48 h after transfection of the vector and helper plasmids, the cells were harvested, pelleted, and frozen thawed three times.The lysate was then treated with benzonase nuclease, loaded onto an iodixanol density gradient and subjected to ultracentrifugation at 67,000 g in Ti70 rotor (Beckman Coulter, Brea, CA, USA).The virus was extracted from the 60%-40% iodixanol interface, aliquoted, and stored at À80 C. The titers of the AAV stocks were determined by qPCR using primers and probes specific for the viral inverted terminal repeats. 50
For protocol B, cells were seeded at a density of 0.25 Â 10 6 cells/mL in SFEM-6 + media (SFEMII as basal media supplemented with the same cytokines as protocol A plus 1 mM StemRegenin1 [STEMCELL Technologies] and 35 nM UM171 [ApexBio, Houston, TX, USA]).Fortyeight hours later, RNPs were nucleofected into 2 Â 10 5 cell using Lonza 4-D nucleofector (Lonza) at the same concentration as protocol A. The cells were plated at 1 Â 10 6 cells/mL concentration post nucleofection and transduced with AAV at MOIs ranging from 0.1 to 2K vg/cell.Sixteen hours after transfection, cells were diluted to a density of 2.5 Â 10 5 cells/mL.The cells were cultured for 5 days after dual delivery of RNPs and AAV.SpyFi Cas9 (Aldevron, Madison, WI, USA) nuclease was used in both protocols.Flow cytometry analysis was performed 1, 2, and 5 days after editing, following which the cells were pelleted and gDNA extraction performed using Qiagen Dneasy Blood and tissue Kit (Qiagen, Hilden, Germany).Editing was performed using the same protocol as above for IPEX I363V cord-derived CD34 + cells.
Primary and patient-derived CD4 + T and nT reg cell culture and editing Human primary CD4 + T cells and nT reg cells isolated from thawed PBMCs as described above were activated with Dynabeads Human T-expander beads CD3/CD28 (Gibco) at a 3:1 bead to cell ratio for 72 h.After beads were removed, the cells further rested overnight in T cell media followed by nucleofection of 20 pmol Cas9 and 50 pmol sgRNA complexes using Lonza 4D-nucleofector.Donor AAVs were added to the cultures immediately post nucleofection at 15%-20% of the culture volume.After an approximately 24-h incubation at 37 C, fresh media was added to cultures to dilute the AAV to 7.5%-10% of the culture volume.Cells were then split every 2-3 days until day 14 after editing.CD4 + T cells from a FOXP3 poly A mutation and IPEX I363V IPEX patients were also edited as described above for healthy donors.
For nT reg cell phenotyping, cells were surface stained for flow cytometry with the following antibodies: CD4-BV605, CD25-PECy7, and CD127-BV510.Intracellular FOXP3 staining with FOXP3-PE antibody was performed after fixation and permeabilization with True-Nuclear Transcription Factor buffer set (BioLegend, San Diego, CA, USA).
ddPCR analysis for determination of NHEJ rates PCR amplicons spanning the guide cleavage site were generated with the NHEJ probe binding to the guide cleavage site.A control amplicon of similar size was generated from another region of the FOXP3 gene.The PCR reactions were partitioned into droplets using a QX200 Droplet Generator (Bio-Rad, Hercules, CA, USA).Amplification was performed using ddPCR Supermix for Probes without UTP (Bio-Rad), 900 nM of primers (IDT, Coralville, IA, USA), 250 nM probe (IDT), and 50 ng genomic DNA.Droplets were analyzed using the QX200 ddPCR System (Bio-Rad) and analyzed using QuantaSoft software (Bio-Rad).The NHEJ rates were calculated using the formula: À ðsignal from NHEJ probe=signal from control probeÞ mock sample À ðsignal from NHEJ probe=signal from control probeÞ RNP treated sample Á Ã100: ddPCR analysis for determination of targeted integration Genomic DNA was extracted from cultured cells and an HDR amplicon was generated by in-out ddPCR using one primer within the AAV construct and another outside the region of homology.An amplicon for either ActB (1.3 kb) or CCR5 (1.5 kb) was generated to serve as the control.Probes for both amplicons were labeled with FAM and the reactions were performed in separate wells.The ddPCR was performed as described in the section above.
RNA extraction and transcript analysis in edited CD4 + T lymphocytes RNA was extracted using RNeasy mini kit (Qiagen) from cultured T lymphocytes 14 days after editing.Complementary DNA was synthesized using Maxima First Strand cDNA Synthesis Kit (Thermo Fisher Scientific) utilizing 10 ng input RNA.Two microliters cDNA was used in three separate ddPCR reactions to detect codon optimized FOXP3, endogenous FOXP3, and control HPRT transcripts using in-house designed or Taqman gene expression assays (Thermo Fisher Scientific).All reactions were performed in duplicates.Transcript concentration was quantified with the following formula: Research Institute according to Institutional Animal Care and Use Committee and approved protocols.Mock-treated or edited HSPCs treated with either protocol A or B were transplanted to NBSGW recipient mice one day after editing.The recipient mice were treated with 12.5 mg/kg clinical grade Busulfan (Otsuka America Pharmaceutical, Rockville, MD, USA) intraperitoneally 24 h prior to human stem cell transfer followed by retro-orbital injections of 1-2 Â 10 6 CD34 + cells per animal.The transplanted mice were sacrificed 12-16 weeks after transfer, and cells harvested from BM and spleens were analyzed using flow cytometry on LSR II flow cytometer (BD Biosciences, San Jose, CA, USA).To assess engraftment of edited cells in various hematopoietic lineages within the BM and spleen, cells were stained with the following fluorophore-conjugated antibodies: human and mouse CD45, CD33, and CD19.To assess the HSC phenotype, cells were stained with the following fluorophore-conjugated antibodies: CD34, CD38, CD90, and CD133.
In vitro differentiation of cord-derived CD34 + cells in ATO system CB CD34 + cells were thawed and edited as described previously for adult mobilized CD34 + cells.One day after editing, 5E4 CD34 + were differentiated in artificial thymic organoid as previously described. 32nput cells were simultaneously analyzed for surface expression of hCD45, CD34, CD14, CD56, CD19, CD1a, CD7, CD3, TCRaß, CD4, snf CD8.CD34 + cells were also kept in stem cell media and gDNA was extracted 5 days after editing to quantify input cell HDR by ddPCR.After four weeks, ATO cultures were harvested and analyzed by FACs for T lineage differentiation and ddPCR for HDR.
Off-and on-target cleavage validation using Miseq Off-target cleavage sites for guide T3 were determined using CCTop-CRISPR-Cas9 target online predictor.The top five predicted off target sites (Table S2) along with the target FOXP3 site were amplified using 200 ng input DNA from two edited donor CD34 + cells using MiSeq oligos and PrimeSTAR GXL DNA polymerase (Clontech, Takara).The above-mentioned amplifications were also performed on donor CD34 + in parallel without delivering any editing reagents to serve as a control.The samples were purified using Agencourt AMPure XP (Beckman Coulter) and analyzed on PAGE gel.The samples were quantified on Qubit (Thermo Fisher Scientific), pooled and analyzed on MiSeq 500 CycleV2 kit (Illumina, San Diego, CA, USA).Data mining was performed with Crispresso2 algorithm. 51antification and statistical analysis Statistical analysis was performed using GraphPad Prism software (GraphPad).

Figure 1 .
Figure 1.Targeted gene integration in primary human CD34 + HSPCs treated with CRISPR-Cas RNPs and AAV6 vectors (A) Schematic of the FOXP3 genomic locus and the AAV6 targeting vectors specific for sgRNAs T9 or T3 designed to insert the MND promoter-driven GFP expression cassette into exon 1 of FOXP3.Black and red rectangles represent exons and UTR elements, respectively; the location of homology arms with respect to the FOXP3 locus is (legend continued on next page)

Figure 3 .
Figure 3.Long-term engraftment and differentiation of cDNA-edited cells in the BM of NBSGW mice after transplantation of HDR-edited HSPCs (A) Schematic of the FOXP3 genomic locus and the rAAV6 targeting vector utilized to insert codon optimized FOXP3 cDNA at the endogenous start site of the FOXP3 gene.FOXP3 cDNA, codon optimized cDNA sequence for FOXP3; pA, SV40 polyadenylation sequence; W, WPRE3 element.(B) Viability of HSPCs 1 day after editing with T3 RNP + AAV.FOXP3.cDNAusing protocol B (n = 3 male donors, 4 independent studies).Significance determined by Kruskal-Wallis test.(C) HDR frequency determined by ddPCR of gDNA extracted from cells 5 days after editing with T3 RNP + AAV.FOXP3.cDNA(n = 3 male donors, 3 independent studies).(D) Engraftment (%hCD45 + ) in the BM of NBSGW recipient mice 16 weeks after transplant of mock (n = 8, 1 male donor, 1 female donor), AAV (n = 6, 1 male donor, 1 female donor), or AAV+RNP (n = 15, 1 male donor, 1 female donor) HSPCs cultured with protocol B from two independent studies.Significance determined by Kruskal-Wallis test.(E) HDR frequency determined by ddPCR of gDNA extracted from the BM of NBSGW recipient mice from (D). (F) NHEJ frequency determined by ddPCR of gDNA extracted from the BM of NBSGW recipient mice from (D). (G) Proportion of B cells (CD19 + ) and myeloid cells (CD33 + ) in the BM of NBSGW recipient mice from (D). (H) Representative flow cytometry data from BM of NBSGW recipient mice from (D).Bar graphs represent mean ± SEM.
Xenotransplantation of edited CD34 + cells into NBSGW miceNBSGW mice (NOD.Cg-Kit WÀ41J Tyr + Prkdc scid Il2rg tm1Wjl /ThomJ, Stock 026622, Jackson Laboratory) used for the experiments were either purchased from Jackson Laboratory or inbred and maintained in the specific pathogen-free animal facility of Seattle Children's