Single-stranded DNA with internal base modifications mediates highly efficient knock-in in primary cells using CRISPR-Cas9 Authors: Karen L Kanke, Rachael E Rayner, Jack Bozik, Eli Abel, Aparna Venugopalan, Ma Suu, Reza Nouri, Jacob T Stack, Gongbo Guo, Tatyana A Vetter, Estelle Cormet-Boyaka, Mark E Hester, Sriram Vaidyanathan Published: 2024 ABSTRACT AbstractSingle-stranded DNA (ssDNA) templates along with Cas9 have been used for knocking-in exogenous sequences in the genome but suffer from low efficiency. Here, we show that ssDNA with chemical modifications in 12–19% of internal bases, which we denote as enhanced ssDNA (esDNA), improve knock-in (KI) by 2–3-fold compared to end-modified ssDNA in airway basal stem cells (ABCs), CD34 + hematopoietic cells (CD34 + cells), T-cells and endothelial cells. Over 50% of alleles showed KI in three clinically relevant loci (CFTR, HBB and CCR5) in ABCs using esDNA and up to 70% of alleles showed KI in the HBB locus in CD34 + cells in the presence of a DNA-PKcs inhibitor. This level of correction is therapeutically relevant and is comparable to adeno-associated virus-based templates. The esDNA templates did not improve KI in induced pluripotent stem cells (iPSCs). This may be due to the absence of the nuclease TREX1 in iPSCs. Indeed, knocking out TREX1 in other cells improved KI using unmodified ssDNA. esDNA can be used to modify 20–30 bp regions in primary cells for therapeutic applications and biological modeling. The use of this approach for gene length insertions will require new methods to produce long chemically modified ssDNA in scalable quantities. Introduction CRISPR/Cas9-based genome editing is being investigated for therapeutic purposes and is widely used for modeling the biological impact of genetic variants (1,2). Cas9 nuclease directed to a targeted genomic locus by a single guide RNA (sgRNA) can induce targeted double-stranded breaks (DSBs) that are repaired by DNA repair mechanisms non-homologous end joining (NHEJ), microhomology-mediated end joining (MMEJ) and homologous recombination (HR). NHEJ and MMEJ are error-prone pathways that result in insertions and deletions (INDELs) of nucleotides that can knock out (KO) a gene. HR can be used to precisely insert exogenous genetic sequences if a template containing regions homologous to the DSB site is delivered into the cells of interest. Among the different HR template delivery strategies, delivery using adeno-associated virus (AAV) vectors results in the most efficient transgene and mutational knock-in (KI) in multiple cell types including hematopoietic stem and progenitor cells (HSPC), T-cells, airway basal stem cells (ABCs), pluripotent stem cells and mesenchymal cells (3–8). Cas9 and AAV have been used to make gene edited cell therapies for therapeutic applications (9). Unfortunately, a recent trial testing the use of HSPCs gene edited using Cas9/AAV to treat sickle cell disease was paused because of abnormally low blood cell counts following the transplantation of corrected HSPCs (10,11). Indeed, previous studies in HSPCs have reported reduced colony formation and engraftment in mouse models that has been attributed to TP53 activation by the AAV genomes (8,10,12). In addition, the integration of inverted terminal repeats from the AAV genome and concatemerized inserts have also been reported in studies investigating gene editing (13,14). Apart from therapeutic applications, gene editing using Cas9/AAV has been used to model the impact of mutations in genetic disorders such as cystic fibrosis (CF) and cancer (15,16). The generation of AAV based templates requires cloning and vector production which can take weeks to months and thus limit the throughput of such experiments. Thus, adverse outcomes in clinical trials, the observation of aberrant genetic changes and the desire to generate primary cell models quickly have all increased the interest in readily available, alternative HR templates that enable high efficiency mutational KI. Short single-stranded DNA (ssDNA) (≤200 bp) has been used as a readily available non-viral alternative to AAV-based HR templates. However, the efficiency of KI was significantly lower than AAV-based templates in ABCs and HSPCs in previous studies (8,17). This is likely due to the premature degradation of ssDNA within the cells by intracellular nucleases. Studies have reported that chemically modifying the ends of ssDNA improves gene correction in cell lines (18). However, end-modification of ssDNA did not improve KI in hematopoietic stem cells (19). Cas9 target sequences (CTS) that improve the nuclear transport of ssDNA have been reported (20). However, CTS needs to be designed separately for each sgRNA and creates room for errors in the design process. Thus, a simpler approach to obtain highly efficient non-viral KI without requiring any specialized design or optimization can have many applications. The modification of internal bases has been shown to improve sgRNA and mRNA activity (21–24). A similar approach has not been reported for improving KI using ssDNA templates. We reasoned that chemical modification of internal bases may improve the protection of ssDNA from nuclease-mediated degradation. However, it was unclear if the strategy would improve KI because the internal base modifications that inhibit nuclease degradation may also prevent the templates from being recognized by the proteins mediating HR. Furthermore, the approach could be rendered non-viable if the modified bases reduce the fidelity of the HR process and thus introduce incorrect bases resulting in non-synonymous mutations. In this study, we tested the use of ssDNA with different proportions of chemically modified bases for KI using homology-directed repair (HDR) in ABCs, CD34 + cells, human umbilical vein endothelial cells (HUVECs), T-cells and induced pluripotent stem cells (iPSCs). We tested the use of phosphorothioate (PS) and 2′-O-methyl (OM) modifications that have been used to protect sgRNA and antisense oligonucleotide therapeutics (25,26). Since we had no information on how chemically modified nucleotides impact HDR, we began by randomly choosing templates in which every sixth or every third base was modified in addition to the ends. We began our study with a focus on ABCs and attempted to correct mutations associated with CF. CF is caused by mutations in the CF transmembrane conductance regulator (CFTR) gene. Although CF is a systemic disorder affecting many organs, the major cause of mortality is lung failure caused by recurrent lung infections. Gene edited airway stem cell therapies to correct the most common CF-causing mutation (F508del) have been proposed and Cas9/AAV have been used to correct the F508del mutation with high efficiency (5,27). We began by evaluating correction of the F508del mutation in ABCs. To test the widespread applicability of this approach, we further extended these experiments to other therapeutically relevant loci (HBB, CCR5) in ABCs and other cell types such as CD34 + cells, T-cells, HUVECs and iPSCs. Mutations in the HBB locus result in hemoglobinopathies (e.g. sickle cell disease), which are the most common genetic disorders in the world. The CCR5 gene encodes a co-receptor in T-cells that is required for infection by the human immunodeficiency virus (HIV). Cell therapies using genome-engineered CD34 + cells are being developed to treat sickle cell disease, HIV and other hematological disorders (2,3) Engineered T-cells are used to treat cancer (4). Similarly, cell therapies derived from HUVECs have been proposed to treat cardiovascular disorders (28). Recent studies have reported improved HDR by inhibiting NHEJ and MMEJ. This was achieved using small molecules that inhibit DNA-dependent protein kinase catalytic subunit (DNA-PKcs) and polymerase theta (PolT), which are key proteins involved in NHEJ and MMEJ, respectively (29,30). Among the different compounds reported to inhibit DNA-PKcs, AZD-7648 has been reported to be very potent in improving KI (29). Similarly, ART558 is a PolT inhibitor reported to improve KI (31). We also evaluated the impact of using these compounds on KI using ssDNA with chemical modifications in internal bases. Finally, we investigated the role of 3′ repair exonuclease 1 (TREX1) in mediating the degradation of unmodified ssDNA. The results presented below demonstrate that ssDNA templates with chemical modifications in every tenth to every sixth base in addition to the three bases in the end can serve as templates for HDR. Moreover, these templates with internal base modifications show significantly improved gene modification in ABCs, CD34 + cells, T cells and HUVECs compared to unmodified, end-modified and CTS-containing ssDNA. We have named these templates enhanced single-stranded DNA (esDNA). Notably, unmodified ssDNA templates are sufficient for efficient KI in iPSCs because iPSCs do not express TREX1. We further show that knocking out TREX1 results in KI by unmodified ssDNA at levels comparable to esDNA. This indicates that ssDNA degradation is mediated by TREX1 and suggests that esDNA templates resist the activity of this nuclease. Overall, we present a broadly applicable non-viral method to improve genome engineering of multiple clinically relevant cell types. Materials and methods Subject Human ABCs were obtained from the Epithelial Cell Core at Nationwide Children’s Hospital (Columbus OH). CD34 + cells were obtained from StemCell Technologies and the Hematopoietic Cell Procurement and Resource Development Center at The Fred Hutchison Cancer Center. HUVECs were obtained from the American Type Culture Collection (ATCC). Pan-T cells were obtained from StemCell Technologies. Guide RNA spacer sequences and HDR template sequences The sequences corresponding to the guide RNA spacer sequences are presented in the supplementary material (Supplementary Table S1). The last three bases of the sgRNA contained 2′-O-methyl 3′phosphorothioate modified bases. The sequences corresponding to the HDR templates for the CFTR, HBB and CCR5 loci including the bases with chemical modifications are detailed in the supplement. They were ordered as UltramersTM from Integrated DNA Technologies (IDT). Generation of ssDNA templated with CTS CTS templates were designed and assembled as described by Shy et al. (20). Similar to other ssDNA templates, the longer ssDNA template was ordered as Ultramers from IDT. The template and oligos were resuspended in IDT buffer. The short oligos were mixed with long ssDNA at 4:1 molar ratio. and then heated to 95°C and cooled by 5°C per 5 min (1°C per min) in a thermocycler. The annealed product was at a concentration of 8 μM and was used in experiments at a concentration of 800 nM. Cell culture of ABCs Human ABCs were cultured in PneumaCultTM Ex Plus medium (StemCell Technologies #05040) with addition of ROCK inhibitor 10 μM (Y-27632, MedChemExpress # HY-10583) and iMatrix-511 (Nacalai USA #892021) at 5000–10 000 cells/cm2 in tissue culture flasks. Cell culture of HSPCs G-CSF Mobilized Human Peripheral Blood CD34 + Cells were purchased from StemCell Technologies (Catalog: 70060.1) and the Hematopoietic Cell Procurement and Resource Development Center at The Fred Hutchison Cancer Center. Cells were thawed by resuspending in 10 ml of RPMI without FBS. The tube was centrifuged at 300 g for 10 min and the supernatant removed. Cells were resuspended in StemSpan™ SFEM II supplemented with StemSpan™ CD34 + Expansion Supplement (10X) (StemCell Technologies, Catalog #09605 and #02691). Cells were plated on 24-well non-treated plates at a density of 100 000 cells/ml and incubated at 37°C and 5% O2 for 2–3 days. Cell culture of iPSCs Normal human induced pluripotent stem cells (Gibco, A18945) were cultured in StemFlex™ media (Thermo Fisher Scientific, A3349401) at a density of 6.0 × 104 cells/cm2 on plates coated with vitronectin (Thermo Fisher Scientific, A31804) and incubated at 37°C for 2 days. The ROCK inhibitor,Y-27632, was added to the media 1 h before the experiment at a concentration of 10 μM. Cells were electroporated 2–3 days after plating. After electroporation, cells were plated and cultured for 3–4 days after which genomic DNA was extracted. Cell culture of HUVECs Primary Umbilical Vein Endothelial Cells were purchased from ATCC (PCS100010). Cells were cultured in EGM™-2 Endothelial Cell Growth Medium-2 BulletKit™ (Lonza, CC-3162) and plated in T75 culture flask at 5000 cells/cm2 with iMatrix-511 (Nacalai USA #892021). Cell culture of T-cells Primary T-cells were purchased from StemCell Technologies (#70024). Cells were cultured in X-VivoTM 15 media supplemented with 5% human serum and 10 ng/ml of IL-2. T-cells were activated using DynabeadsTM T-cell activator with CD3/CD28 (Thermo Fisher Scientific Inc., #11161D) for 1–2 days using directions provided by the manufacturer (1:1 bead-to-cell ratio). Failure to remove beads resulted in reduced cell proliferation. Cells were edited after 4–6 days in culture. Gene editing of ABCs, iPSCs and HUVECs Electroporation was performed 4–5 days after plating. Cells were suspended by treatment with TrypLE™ Express Enzyme (1X), phenol red (Thermo Fisher Scientific, 12605036). Cells were suspended in Opti-MEM™, Reduced Serum Medium (Thermo Fisher Scientific, 31985088) at 2.5–10 million cells/ml. Cells were electroporated with Lonza 4D-Nucleofector™ in 16-well Nucleocuvette Strips (Lonza, V4XP-3032). Cas9 at 6 μg (IDT) and sgRNA at 3.2 μg (Synthego) were complexed at room temperature for 10 min before adding 100 000 cells (20 μl/well). ssDNA was added such that a final concentration of 800 nM was achieved (unless otherwise specified). Cell mixture was transferred to Lonza strip at 20 μl per well and electroporated using program CA-137 with solution setting P3. Afterward, 80 μl of OptiMEM was added to each well. When editing using AAV, AAV was added after electroporation. ABCs and HUVECs were treated with 105 vg/cell. iPSCs were treated with 104 vg/cell. Cells were then transferred to a 12- or 24-well tissue culture plate. Pneumacult Ex Plus (StemCell Technologies #05040), ROCK inhibitor 10 μM (Y-27632, MedChem Express HY-10583) and iMatrix-511 (Nacalai USA #892021) were added to ABCs and EGM™-2 Endothelial Cell Growth Medium-2 BulletKit™ (Lonza, CC-3162) with iMatrix was added to HUVECs and then incubated in 37°C at 5% O2. iPSCs were plated with media described above and incubated at 37°C. AZD-7648 (MedChem Express, HY-111783) was added for DNA-PKcs inhibition at a concentration of 0.5 μM. ART558 (MedChem Express, HY-141520) was added for PolT inhibition at a concentration of 5 μM. Gene editing of HSPCs and T-cells HSPCs were counted after 2 days and centrifuged at 300 g for 10 min. T-cells were counted after 4–6 days in culture. Supernatant was removed and cells were resuspended in P3 Primary Cell Nucleofector® Solution with Supplement 1 (Lonza, V4XP-3032) at 5–15 million cell/ml. Cell mixture was transferred to Lonza strip at 20 μl per well and electroporated using program DZ100 with solution setting P3. A total of 80 μl of the media used to culture CD34 + cells or T-cells was added to each well. AAV was added after electroporation, if necessary, at 104 vg/cell. CD34 + cells were transferred to 24-well non-treated plates at a density of 100 000 cells/ml and incubated at 37°C for 2–3 days. T-cells were cultured at densities of 500 000–1 000 000 cells/ml. Gene correction measurement Three to four days after electroporation, genomic DNA was extracted using QuickExtract (QE) DNA Extraction Solution (Biosearch Technologies, QE09050). TrypLE Express Enzyme was added to adherent cells, then after 20 min they were centrifuged at 300 g for 10 min. Supernatant was removed, cells were suspended in OptiMEM and counted. A total of 50 000–100 000 cells were taken, re-centrifuged and the pellet was suspended in 50 μl of QE. The QE suspension was heated at 65°C for 6 min then heated at 98°C for 10 min. Amplification was done using primers listed in Supplementary Table S2. PCR was performed using Q5® High Fidelity 2X Master Mix (New England Biolabs, M0492) with the thermocycler settings listed in Supplementary Table S3. The PCR product was sanger sequenced and ICE analysis was used to determine the percentage of corrected alleles (32). Next-generation sequencing ABCs from pwCF were electroporated with Cas9, sgRNA and ssDNA/esDNA templates. Control ABCs were electroporated with esDNA alone without Cas9 and sgRNA. gDNA was extracted 4–5 days post electroporation. The F508del region was PCR amplified as described previously. The PCR product was gel purified using GeneJET Gel Extraction Kit (Thermo Scientific, K0692). The concentration was measured by NanoDrop 2000 (Thermo Scientific, ND-2000) and the samples were prepared as directed by Azenta (GeneWiz, Aplicon-EZ) specifications and sent for next-generation sequencing (NGS). Data received from Azenta was uploaded as directed into the CRISPResso2 (33) site and sequences listed in Supplementary Table S4 were entered for analysis. Air–liquid interface Gene corrected ABCs were plated on 6.5 mm Transwell® plates with 0.4 μm pore polyester membrane insert (StemCell Technologies, 38024) at 40 000–50 000 cells per well. Pnuemacult Ex Plus with ROCK inhibitor (10 μM) was added to both the apical and basal side of insert and iMatrix was added to the apical insert to support expansion. Once cells became confluent, about 3–4 days, media was removed and air–liquid interface (ALI) medium from the University of North Carolina (UNC media) was added to the basal side of the insert (34,35). UNC medium was replaced every other day for 4 weeks. UNC media consists of a base medium that is a 1:1 mixture of LHC basal medium (Thermo Fisher Scientific, Cat. No. 12677019) and Dulbecco’s modified Eagle’s medium high glucose media (Thermo Fisher, Cat. No. 11965092) with bovine serum albumin (BSA) (0.5 mg/ml), bovine pituitary extract (10 μg/ml), insulin (0.87 μM), transferrin (0.125 μm), hydrocortisone (0.21 μm), triiodothyronine (0.01 μm), epinephrine (2.7 μm), epidermal growth factor (0.50 ng/ml), retinoic acid (50 nm), phosphorylethanolamine (0.5 μm), ethanolamine (0.5 μm), zinc sulfate (3.0 μm), penicillin g sulfate (100 u/ml), streptomycin sulfate (100 μg/ml), gentamicin (50 μg/ml), amphotericin (0.25 μg/ml), ferrous sulfate (1.5 μm), magnesium chloride (600 μm), calcium chloride (110 μm), selenium, manganese (30 nm), silicone (1 nm), molybdenum (500 nm), vanadium (5 nm), nickel sulfate (1 nm) and tin (0.5 nm) (35). Immunostaining for epithelial cell characterization CF-donor cells edited with esDNA and differentiated on ALI transwells for a duration of 4 weeks were stained for cytokeratin 5 (KRT5), acetylated tubulin (ACT) and mucin 5B (MUC5B) to identify basal, ciliated and secretory cells, respectively. Phosphate-buffered saline (PBS) was added to the ALI cultures and left overnight to remove mucus. The ALI cultures were washed using PBS and fixed using CytofixTM (BD Bioscience, Catalog No. 554655) for a duration of 30 min at room temperature and then washed with PBS. ALI cultures were permeabilized using CytopermTM (BD Biosciences, BD 554714) for 30 min at room temperature. Fixation and permeabilization reagents were added to both sides of the ALI cultures. The samples were washed using PBS five times and blocked using 2% BSA for 1 h. Mouse anti-ACT (Sigma-Aldrich; T6793) was diluted 1:8000 and Rabbit anti-MUC5B (Sigma-Aldrich, HPA008246) was diluted 1:500 in permeabilization buffer. A total of 200 μl of the diluted primary antibodies were added to the ALI transwells and incubated for 1 h at room temperature. The primary antibodies were removed, ALI transwells were washed with PBS and 200 μl of secondary antibody was added to the ALI transwells. PE Goat anti-mouse IgG (Biolegend, 405307) was diluted 1:200 and AF488 Donkey anti-rabbit (Life Technologies, A21206) was diluted 1:200 in permeabilization buffer. The ALI transwells were incubated in secondary antibody for a duration of 1 h at room temperature. The secondary antibodies were removed, washed with PBS and Recombinant Alexa Fluor 647 anti-Cytokeratin 5 antibody diluted 1:200 (Abcam, ab193895) was added and incubated for 1 h. DAPI (4′,6-diamidino-2-phenylindole) (Sigma-Aldrich, D9542) diluted 1:1000 in permeabilization buffer was added to the ALI transwells and incubated for 30 minutes. The ALI transwells were washed with PBS, placed on slides using Fluoromount-G Mounting Medium (Invitrogen, 00–4958-02), and cover slipped. Microscopy Transwell slides were imaged on a motorized Eclipse Ti2-E inverted microscope (Nikon Instruments) with a SOLA LED engine (Lumencor) and an ORCA Fusion CMOS camera (Hamamatsu). Semrock filter sets were used for individual DAPI, GFP, TRITC and Cy5 channels. Multichannel images were captured as Z-stacks using NIS-Elements AR software (Nikon Instruments, v. 5.30) with 40× Plan Apochromat Lambda objectives (Nikon Instruments) for a final resolution of 0.32 and 0.16 μm/pixel, respectively. The Z-stacks were compressed into 2D images using the Extended Depth of Focus module in NIS-Elements. Ussing chamber functional assay Four weeks after cells were plated in the transwell plates, ion channel function of CFTR was assessed with the Ussing chamber system (VCC MC6 from Physiologic Instruments Inc. San Diego, CA, USA) with the U2500 Self-contained Ussing chambers (Warner Instruments, Hamden CT). To measure Cl− secretion, the fully-differentiated epithelial sheets were placed with mucosal solution on apical side [120 mM NaGluconate, 25 mM NaHCO3, 3.3 mM KH2PO4, 0.8 mM K2HPO4, 4 mM Ca (Gluconate)2, 1.2 mM Mg (Gluconate)2, 10 mM mannitol] and serosal solution on basal side [120 mM NaCl, 25 mM NaHCO3, 3.3 mM KH2PO4, 0.8 mM K2HPO4, 1.2 mM CaCl2, 1.2 mM MgCl2, 10 mM glucose]. Following this, ion channel activators and inhibitors were added as follows: 10 μM amiloride apically to inhibit Epithelial Sodium Channel (ENaC), 10 μM forskolin bilaterally (Abcam #ab120058) to activate CFTR channels via activation of adenylyl cylase concentrations, 10 μM CFTR potentiator drug VX-770 (SelleckChem, #S1144) apically, 10 μM CFTRinh-172(Sigma Millipore, #C2992) apically to inhibit CFTR channel, and 100 μM uridine-5′-triphosphate trisodium salt (Thermo Scientific Chemicals #AAJ63427MC) apically. Staining and fluorescence-activated cell sorting analysis Cells were stained 3–4 days after editing with anti-Cytokeratin 5 antibody [EP1601Y] (Alexa Fluor® 647) (Abcam, ab193895), Rabbit IgG, monoclonal [EPR25A] - Isotype Control (Alexa Fluor® 647) (Abcam, ab199093), Purified Mouse IgG2b k Isotype Ctrl (Biolegend, 401202) and purified anti-TP63 (W15093A) (Biolegend, 687202). Flow cytometry was done on the LSR/Fortessa cytometer and data was analyzed using FlowJo software (FlowJo, LLC). Droplet digital PCR to measure p21 levels CD34 + cells were genome edited as described above. Two days after gene editing, RNA was extracted using the RNAqueous™-4PCR Total RNA Isolation Kit from Thermo Fisher (Catalog: AM1914) according to the manufacturer’s instructions. Reverse transcription droplet digital polymerase chain reaction (RT-ddPCR) was performed using the One-step RT-ddPCR kit from Bio-Rad (Catalog: 1864022). Predesigned primers and probes to detect p21 and TBP were obtained from IDT (Supplementary Table S5). Thermocycler settings used for RT-ddPCR are listed in Supplementary Table S6. The droplets were analyzed using the QX200 system from Bio-Rad. Results Chemical modification of internal bases in ssDNA templates improves mutational KI in airway cells To test our hypothesis that chemical modification of internal bases in ssDNA may improve KI, we used ssDNA templates with PS modifications in the last three bases and at every sixth or every third base (Figure 1A and B). We used ssDNA with PS modifications in the last three bases (end-modified) and unmodified ssDNA as controls. The modified bases are detailed in the supplementary material. The fraction of bases to be modified was chosen randomly. We targeted exon 11 in the CFTR locus and attempted to incorporate silent mutations in the region associated with the F508del mutation, which is commonly seen in people with CF (pwCF) (Supplementary Figure S1A). We edited non-CF ABCs with ssDNA templates that were either unmodified or chemically modified. Gene editing using ssDNA was inefficient (<10% alleles with HDR) when we used ssDNA templates at a concentration of 800 nM in the absence of the DNA-PKcs inhibitor AZD-7648 (Supplementary Figure S1B) and there was no improvement in KI at higher concentrations of ssDNA (Supplementary Figure S1C). In the presence of AZD-7648, ssDNA templates with PS modification in every sixth base had higher HDR (41 ± 15%) than both unmodified (15 ± 5%) and end-modified (24 ± 6%) ssDNA templates (Figure 1C). Optimal mutational KI was achieved using ssDNA templates modified at every sixth base at a concentration of 800 nM (Supplementary Figure S1C). Subsequent experiments use 800 nM template unless otherwise specified. Interestingly, ssDNA templates with PS modification in every third base showed a significant reduction in HDR, suggesting that templates with excessive modifications are not compatible with the HDR process (Figure 1C). The chemically modified ssDNA did not significantly reduce the proliferation of the edited cells when compared to unedited cells (Supplementary Figure S1D). Chemical modification of internal bases in ssDNA improves mutational KI in ABCs. (A) Graphic showing ssDNA with different levels of chemical modifications. (B) Graphic showing the chemical structure of modified bases. (C) In airway cells at the CFTR locus, ssDNA templates with every sixth base modified (41 ± 15%) show significantly higher HDR than end-modified ssDNA (24 ± 6%), which are both significantly higher than unmodified ssDNA (15 ± 5%). Modifying every third base results in the least mutational KI (11 ± 15%). Templates were used at a concentration of 800 nM. Statistical comparison was made using one-way analysis of variance (ANOVA). (D) Chemical modification of every sixth base in ssDNA enables improved mutational KI compared to ssDNA with CTS. (E) ssDNA modified at every sixth base pair is also more efficient than end-modified ssDNA in the HBB locus. The ssDNA templates modified in every sixth base achieved mutational KI in 51 ± 9% alleles at the 800 nM concentration as opposed to 22 ± 4% for end-modified ssDNA. At 400 nM, ssDNA modified at every sixth base achieved 31 ± 6% mutational KI compared to 7 ± 3% mutational KI achieved by end-modified ssDNA. Statistical comparison was made using two-way ANOVA. (F) Every sixth modified ssDNA improves mutational KI compared to end-modified ssDNA in the CCR5 locus at 200 nM (41 ± 5% versus 17 ± 1%) and 400 nM (55 ± 6% versus 40 ± 4%) but not at 800 nM (53 ± 16% versus 51 ± 10%). Statistical comparison was made using two-way ANOVA. (G) In airway cells, ssDNA with every sixth base modified with 2′O-methyl (OM) also shows higher HDR than unmodified ssDNA (55 ± 19% versus 15 ± 8%) (n = 5 biological replicates for ssDNA modified with PS, OM modifications and unmodified ssDNA. n = 3 biological replicates for AAV). (H) Mutational KI of different lengths was attempted in exon 12 of the CFTR locus using ssDNA templates PS modified at every sixth base pair to introduce the G551D mutation. Editing was performed with DNA-PKcs inhibition. Template 1 (T1) which substitutes bases in a 12 bp region achieved ∼80–90% mutational KI while template 2 (T2) which substitutes bases in a 34 bp region achieved ∼60 bp mutational KI. Statistical comparison was made using T-test. For all panels using one-way ANOVA, multiple comparisons were performed using Tukey’s test. Sidak’s test was used for two-way ANOVA. ****, ***, ** and * represent P < 0.0001, P < 0.001, P< 0.01 and P < 0.05 respectively for all panels. Cas9 target sequences (CTS) were recently reported to improve mutational KI using ssDNA by improving nuclear localization of the ssDNA (20). We designed ssDNA templates with CTS sequences targeting exon 11 of CFTR (Supplementary Figure S1E). Compared to unmodified ssDNA with CTS (7 ± 9%), ssDNA with PS modifications in every sixth base showed significantly improved mutational KI (34 ± 8%) (Figure 1D). Adding CTS sequences to ssDNA with chemical modifications in every sixth base did not improve mutational KI further (Supplementary Figure S1F). To test if these findings can be widely applied in other therapeutically relevant loci, we tested our approach in the HBB and CCR5 loci using ssDNA concentrations ranging from 200–800 nM. We introduced silent mutations in the region corresponding to the sickle cell mutation (E6V) in HBB and introduced a premature stop codon in the CCR5 locus. For HBB, ssDNA modified in every sixth base showed improved HDR at all concentrations (Figure 1E). For CCR5, ssDNA modified in every sixth base improved HDR at 200 and 400 nM but not 800 nM (Figure 1F). Overall, every sixth modified ssDNA showed significantly higher HDR at all loci tested in ABCs with DNA-PKcs inhibition (Figure 1C–F). Mutational KI using every sixth modified ssDNA templates is also up to 2-fold higher than end-modified ssDNA when both DNA-PKcs and PolT are inhibited (Supplementary Figure S1G and H). We named the ssDNA with internal base modifications that enhances HDR, most often in the presence of AZD-7648, enhanced ssDNA (esDNA). For all three loci, template only controls did not show KI without Cas9 (Supplementary Figure S2A and B). This was further confirmed using next-generation sequencing for the esDNA templates targeting the CFTR locus (Supplementary Figure S2B). We then investigated if improved editing can be obtained using other chemical modifications. We used esDNA with OM modifications on approximately every sixth base (Figure 1B). We avoided modifying thymine since the OM bases are RNA. OM modification in esDNA produced even higher HDR (55 ± 19%) than the esDNA with PS modification (38 ± 16%) (Figure 1G). Both esDNA formulations had higher HDR than unmodified ssDNA templates (15 ± 8%). To further understand if higher mutational KI levels can be achieved using esDNA templates by targeting alleles closer to the DSB, we edited exon 12 in the CFTR locus associated with the variant G551D (another common CF-causing mutation) with two different esDNA templates (Figure 1H). Template 1 which replaces bases in a 12 bp region next to the DSB site enabled mutational KI in 80–90% alleles (Figure 1H). Conversely, template 2 which replaces bases in a 34 bp region yielded only ∼60% edited alleles (Figure 1H). Gene correction of the F508del mutation in CF ABCs restores CFTR function in differentiated airway epithelial cells To test the ability to restore CFTR function, we began by gene correcting ABCs homozygous for the F508del mutation. Patient-derived CF ABCs were efficiently corrected using esDNA chemically modified at every sixth base (esDNA-6) with PS and OM modifications (Figure 2A). In contrast to the non-CF ABCs, there was no significant difference between esDNA-6 templates with PS or OM (47 ± 15% versus 49 ± 15%). We further validated that esDNA-6-PS did not result in the introduction of unintended mutations in the F508del locus using next-generation sequencing. Figure 2B shows that the HDR achieved by esDNA was comparable to the HDR levels achieved using AAV (57 ± 0% versus 65 ± 11% respectively). Unintended mutations (other than INDELs generated by NHEJ) were not observed. A small fraction of alleles (1–3%) had incomplete HDR (Figure 2B and Supplementary Figure S2C). Editing did not induce a reduction in basal cell markers, cytokeratin 5 (KRT5) and tumor protein 63 (P63) (Figure 2C). On average, all conditions showed > 85% cells that were positive for both KRT5 and P63. Correction of F508del mutation restores CFTR function in ABCs from pwCF. (A) Both esDNA-6-PS and esDNA-6-OM (800 nM) corrected ABCs from CF donors (49 ± 14% and 47 ± 14%, corrected alleles respectively) (n = 4 biological replicates). (B) Gene correction of the F508del mutation was confirmed using next-generation sequencing. NGS showed that 57 ± 0% of alleles had perfect HDR in ABCs edited using esDNA-6-PS which was comparable to 65 ± 15% alleles with HDR when editing was performed using AAV. A total of 1–3% alleles had imperfect HDR. INDELs accounted for 27–29% alleles. ABCs treated with esDNA alone only had unmodified alleles. (C) A representative fluorescence-activated cell sorting plot shows that > 98% of ABCs were positive for KRT5 and P63 4 days after editing. There was no reduction in KRT5 and P63 expression in the edited ABCs relative to Mock electroporated controls in experiments from three different donors. (D) Edited samples displayed transepithelial resistances (TEER) similar to unedited controls [wild type (WT) and Mock], indicating that the editing process does not compromise the ability of the ABCs to produce fully differentiated epithelial sheets. (E) Edited ABCs produced epithelial sheets containing basal (KRT5), ciliated (ACT) and secretory cells (MUC5B), further indicating the formation of differentiated epithelial sheets. Scale bar represents 50 μm. (F) Representative traces from Ussing chamber analysis showing a non-CF control, an uncorrected CF sample, esDNA-6-PS- and esDNA-6-OM-corrected CF samples. Uncorrected CF samples show minimal responses to forskolin and CFTRinh-172 which activate and inhibit CFTR, respectively. All edited samples showed restored responses to forskolin and CFTRinh-172 indicating restored CFTR function. (G) Percent restoration of CFTR function in corrected CF samples relative to non-CF controls plotted as a function of allelic correction. Edited samples showed 30–60% restoration of CFTR function on average. This change in current was statistically significant when compared using one-way ANOVA followed by Dunnett’s T2 multiple comparisons test (P < 0.05). (H) Raw CFTRinh-172 short circuit currents observed in CF donor ABCs compared to non-CF controls and uncorrected CF controls (n = 3 biological replicates). Samples corrected using esDNA-6-PS or esDNA-6-OM both show increased response to CFTRinh-172 relative to uncorrected controls. To quantify the function of CFTR following gene correction, corrected CF, uncorrected CF and non-CF samples, were grown at ALI to induce differentiation into polarized airway epithelial sheets and then currents were measured using the Ussing chamber assay. Epithelial barrier integrity was evaluated by measuring the TEER. Edited ABCs produced epithelial sheets with TEER comparable to controls thus indicating that the edited ABCs can form differentiated epithelial sheets (Figure 2D). ALI cultures contained basal cells expressing KRT5, ciliated cells expressing ACT and secretory cells producing mucin 5B (MUC5B) (Figure 2E, merged images, and Supplementary Figure S2D, individual panels). The corrected CF samples showed restored CFTR function indicated by increased forskolin-stimulated currents and increased CFTRinh-172-sensitive short circuit current (Figure 2F–H). On average, uncorrected CF samples showed 1.6 ± 0.9 μA/cm2 change in short circuit currents in response to CFTRinh-172 compared to 16 ± 6 μA/cm2 for non-CF controls. CF samples corrected using both PS and OM esDNA-6 showed 6 ± 2 μA/cm2 change in short circuit currents in response to CFTRinh-172. This change in current was statistically significant when compared using one-way ANOVA followed by Dunnett’s T2 multiple comparisons test (P < 0.05). When we compared CFTR function in gene corrected cells to non-CF controls, epithelial sheets derived from ABCs edited using esDNA-6 with PS and OM showed CFTR function above the therapeutic threshold (20%) for clinical benefit (37 ± 12% for esDNA-6 with PS and 36 ± 15% for esDNA-6 with OM relative to non-CF controls) (Figure 2G). Chemical modification of internal bases in ssDNA templates improves mutational KI in CD34 + cells, T-cells and HUVECs but not iPSCs To further test if esDNA-6-PS templates would be effective in improving mutational KI in other therapeutically relevant primary cells, we tested them in CD34 + cells, HUVECs, T-cells and iPSCs. In CD34 + cells, we tested editing in the HBB, CFTR and CCR5 loci and found that esDNA-6-PS improved HDR in all three loci. In CD34 + cells, esDNA templates improved mutational KI even in the absence of AZD-7648 (Figure 3A). In the absence of AZD-7648, esDNA-6-PS resulted in 29 ± 13% alleles with HDR compared to 10 ± 13% alleles with HDR for end-modified templates, in the HBB locus. In the presence of AZD-7648, esDNA-6-PS (50 ± 21%) improved HDR over end-modified ssDNA (20 ± 13%) in the HBB locus (Figure 3B). CD34 + cells edited using esDNA templates did not show reduced cell viability as measured by trypan blue exclusion (Figure 3C). Similar improvement in mutational KI was seen in CCR5 (38 ± 26% versus 20 ± 23% for esDNA-6-PS versus end-modified) and CFTR (21 ± 7% versus 5 ± 6% for esDNA-6-PS versus end-modified) loci in CD34 + cells (Figure 3D and E). We tested esDNA-6-PS in T-cells and HUVECs by adding silent mutations in the CFTR locus. In T-cells, esDNA-6-PS templates resulted in 30 ± 10% compared to 13 ± 6% obtained with end-modified ssDNA (Figure 3F). In HUVECs, the use of esDNA-6-PS resulted in 12 ± 5% alleles with HDR compared to 4 ± 3% alleles with HDR obtained with end-modified ssDNA (Figure 3G). esDNA templates (800 nM) improve mutational KI in HSPCs, T-cells and HUVECs but not iPSCs. HDR in CD34 + cells significantly increased with esDNA compared to end-modified ssDNA in the (A) HBB locus in the absence of AZD-7648 and (B) in the presence of AZD-7648. (C) The viability of CD34 + cells was not significantly different between edited CD34 + cells and unedited controls. HDR in CD34 + cells was also increased in the (D) CCR5 locus and (E) CFTR locus. (F) In T-cells, esDNA increased HDR in the CFTR locus. (G) In HUVECs, esDNA increased HDR in the CFTR locus from 4 ± 3% obtained with end-modified ssDNA to 12 ± 5% but the difference was not significant. Mutational KI with esDNA was significantly lower in iPSCs in the CFTR locus than unmodified ssDNA both in the (H) absence (not significant) and (I) presence of AZD-7648. 3 or more biological replicates were tested for each cell type. Groups in A–B and D–G were compared using a paired T-test. Statistical comparisons for panels (C), (H) and (I) were made using one-way ANOVA followed by Tukey’s test. ** and * represent P < 0.01 and P < 0.05, respectively, for all panels. We also tried adding silent mutations in the CFTR locus in iPSCs using esDNA-6-PS. We observed alleles with HDR even in the absence of DNA-PKcs inhibition (Figure 3H). Surprisingly, we found that in iPSCs, unmodified ssDNA was more effective in enabling HDR-mediated mutational KI over end-modified ssDNA (71 ± 18% versus 56 ± 19%). In addition, esDNA-6-PS showed reduced HDR (30 ± 12%) compared to all other conditions (Figure 3I). Knocking out TREX1 exonuclease improves HDR using unmodified ssDNA templates Since iPSCs were edited efficiently using unmodified ssDNA, we reasoned they may lack nucleases that degrade ssDNA. It has been reported that the nuclease TREX1 may be involved in degrading ssDNA in cell lines and primary cells (36). Indeed, a western blot probing TREX1 expression in iPSCs showed that iPSCs do not express it (Figure 4A). To further investigate the effect of TREX1 on ssDNA in other cells, we genome edited ABCs to KO TREX1 (TREX1 KO) by using Cas9 to create insertions and deletions in the gene and verified KO using western blot (Figure 4A). HDR using unmodified ssDNA was dramatically improved in TREX1 KO ABCs compared to WT (4 ± 4% for WT and 44 ± 14% for TREX1-KO). However, HDR was not improved in TREX1 KO ABCs when esDNA-6-PS template was used (46 ± 4% for WT and 41 ± 11% for TREX1 KO) (Figure 4B). HUVECs and T-cells with TREX1 KO also showed improved mutational KI using unmodified ssDNA (Supplementary Figure S3A and B). These studies suggest that extensive chemical modifications protect esDNA from TREX1. Knocking out TREX1 improved mutational KI without chemically modifying ssDNA. (A) Western blot shows TREX1 expression in ABCs and HUVECs but not iPSCs. ABCs with TREX1 KO show reduced TREX1 expression. (B) HDR using unmodified ssDNA templates (800 nM) was improved when TREX1 was knocked out in non-CF ABCs. Knocking out TREX1 did not improve HDR by esDNA templates (n = 4 biological replicates). (C) Mutational KI with ssDNA using combinations of chemical modifications in ABCs. Both esDNA OM (40 ± 21%) and esDNA PS (34 ± 16%) showed improved mutational KI relative to unmodified (3 ± 3%), end-modified (11 ± 6%) and Alt-R HDR modified ssDNA (23 ± 9%). All other combinations did not improve on esDNA OM (40 ± 21%) and esDNA PS (34 (± 16%) (n = 4–6 biological replicates). (D) Modifying the 3′ half of the template alone with PS improved mutational KI compared to end modification (n = 3 biological replicates). Including PS chemical modifications throughout the template improved editing slightly but the improvement was not statistically significant compared to templates in which only the 3′ half was modified. (E) Templates with PS modified internal bases spread throughout the template resulted in more consistent HDR compared to template with PS modified bases in just the last 20 bases (n = 4 biological replicates). (F) Mutational KI in the HBB locus was significantly higher in ABCs using esDNA-6-PS templates (53 ± 14%) compared with Alt-R HDR templates (13 ± 12%) (n = 4 biological replicates). G) Mutational KI in the HBB locus was also significantly higher in CD34 + cells using esDNA-6-PS templates (66 ± 2%) compared with Alt-R HDR templates (17 ± 1%) (n = 4 biological replicates). (H) Modification of every 10th base with PS groups also improved mutational KI compared to unmodified or end-modified ssDNA. Mutational KI was reduced when every fourth base was modified. Statistical comparison was made using one-way ANOVA. Multiple comparisons were performed using Tukey’s test. ***, ** and * represent P < 0.005, P < 0.01, P < and P < 0.05 respectively in all panels. All experiments used 800 nM template. Since modification of the 3′ end has been reported to protect ssDNA from TREX1 (36), we tested if protection of 3′ end by modifying the OH group in esDNA-6-PS would improve mutational KI further (Supplementary Figure S3C). The addition of a 3′ C3 spacer to esDNA-6-PS did not improve mutational KI (Figure 4C). We also investigated if modifying every sixth base in the 3′ half of the ssDNA template would improve mutational KI. Modifying every sixth base in the 3′ half (in addition to the three bases in the ends) resulted in HDR comparable to esDNA-6-PS (39 ± 19% versus 47 ± 22%) (Figure 4D) but it was also not statistically significantly better than the end-modified template (29 ± 18% alleles with HDR). Modifying the last 20 bases of the template on either end with PS did not result in mutational KI comparable to esDNA-6-PS (25 ± 19% versus 39 ± 26%) (Figure 4E). We also investigated if proprietary end protecting modifications available from IDT (Alt-R HDR modifications) may enhance HDR further. Alt-RTM HDR modifications added to the ends of the ssDNA improved mutational KI relative to end modification with PS alone in the CFTR locus in ABCs (23 ± 9% versus 11 ± 7%) (Figure 4C). The percent of alleles with HDR was still higher using esDNA-6-PS (34 ± 15%) and esDNA-6-OM (40 ± 21%) than Alt-R HDR modifications alone. However, there was no further improvement in mutational KI when Alt-R HDR modifications were added to esDNA-6-PS (Figure 4C). In contrast to the CFTR locus, Alt-R HDR templates showed substantially lower HDR than esDNA-6-PS in the HBB locus in both ABCs (53 ± 14% versus 13 ± 11%) and CD34 + cells (66 ± 2% versus 17 ± 1%) (Figure 4F and G). Using higher concentrations of Alt-R HDR templates may improve its efficiency at least in CD34 + cells (Supplementary Figure S4). We reasoned that increased protection of ssDNA by combining modifications may slow degradation by ssDNA further. Therefore, we investigated if using templates with combinations of PS and OM modifications improved mutational KI further (Supplementary Figure S3C). However, this was not the case (Figure 4C). Since modification of every third base reduced mutational KI, we wanted to assess if modifying fewer internal bases than every sixth would further improve mutational KI. Modifying every tenth base in ssDNA resulted in 37 ± 14% alleles with HDR which was similar to esDNA-6-PS (31 ± 14%) (Figure 4H). Modifying every fourth base reduced alleles with HDR to 19 ± 8%. We wanted to further validate that the activity of TREX1 was specific to naked ssDNA templates. Therefore, we tested the impact of TREX1 on AAV-based HDR templates targeting the CFTR locus (5). These AAV templates have significantly longer homology arms (>800 bp) compared to the ssDNA and esDNA templates (∼90 bp) (Supplementary Figure S3D). AAV-based HDR templates can be used to reliably edit the CFTR locus in ABCs, CD34 + cells, HUVECs and iPSCs (Supplementary Figure S3E). When using AZD-7648, we obtained ∼90% modified alleles in airway cells and 60–80% modification in CD34 + cells using AAV-based templates (Supplementary Figure S3E). ABCs with TREX1 KO did not show any change in mutational KI of the CFTR locus when edited using Cas9 and AAV-based templates in the presence of AZD-7648 (Supplementary Figure S3F). Thus, the activity of TREX1 is specific to naked ssDNA templates. Mutational KI using esDNA templates induces less p21 activation in CD34 + cells KI using AAV-based templates has been reported to reduce the regenerative potential of CD34 + cells by the activation of the DNA damage response pathway (12). The protein p53 plays a critical role in this pathway and it acts to increase the expression of p21 (37). To test if esDNA templates activate p21, we edited CD34 + cells in the CFTR locus using unmodified, end-modified (PS and Alt-R HDR), esDNA-6-PS and AAV-based templates. AAV-based templates resulted in 61 ± 13% alleles with HDR compared to 29 ± 7% and 19 ± 4% alleles with HDR using esDNA-6-PS and Alt-R HDR templates (Figure 5A). There was no significant difference in the cell viability between the cells edited using the different templates when measured using the trypan blue dye exclusion assay 2 days after editing (Figure 5B). The activation of p21 was quantified in cells lysed 2 days after editing by measuring the level of p21 mRNA using RT-ddPCR. TATA box binding protein (TBP) was used as a housekeeping control (38). The p21 to TBP ratio in CD34 + cells edited using AAV was 7 ± 2 compared to 3 ± 2 for cells edited using esDNA-6-PS. Both values were higher than the ratio observed in unedited controls (1 ± 1). Notably, exposure to AAV alone resulted in an elevated p21 to TBP ratio (5 ± 2). CD34 + cells exposed to esDNA-6-PS alone showed a p21 to TBP ratio of 1 ± 1 which is similar to the unedited controls. Editing using esDNA-6-PS elicits less p21 activation in CD34 + cells. (A) In CD34 + cells, mutational KI by esDNA-6-PS was less than the mutational KI achieved using AAV templates (n = 3–4 biological replicates). (B) There was no significant difference in the viability of edited CD34 + cells when measured using the trypan blue assay on day 2 after editing (n = 3 biological replicates). Statistical comparison was made using ANOVA followed by Tukey’s multiple comparisons test. (C) CD34 + cells edited using AAV templates showed a greater increase in p21 expression when measured by ddPCR. Treatment using AAV alone also resulted in increased p21 expression. Editing with esDNA-6-PS resulted in less p21 activation (n = 3 biological replicates with 2 technical replicates). Statistical comparison was made using one-way ANOVA. Multiple comparisons were performed using Tukey’s test. ** and * represent P < 0.01 and P < 0.05, respectively in all panels. Discussion Previous studies investigating the use of ssDNA to perform HDR-mediated mutational KI have explored the protection of the last three bases of the ssDNA templates but not internal bases (8,18–20). Although chemical modification of internal bases has been used to protect sgRNA and mRNA used in gene editing, the approach may not have been an obvious choice for HDR templates due to differences in the biological processes involved (21,22,25). In the case of sgRNA, the modified bases do not interact with host proteins. Thus, modification of sgRNA has a lower chance of compromising Cas9 nuclease activity as long as the modifications do not hinder base pairing or interactions with Cas9. In the case of mRNA, many of the most successful modifications are already present in nature and thus compatible with the natural translational apparatus in cells. By way of contrast, chemical modifications in HDR templates must be compatible with a variety of proteins involved in the HDR pathway. Our results show that the HDR machinery can indeed process chemically modified bases. Our results also show that this compatibility with the HDR machinery can be lost if too many bases are modified. We only tested two types of chemical modifications. It is possible that other types of chemical modifications may yield similar results. The combination of at least the OM and PS modifications did not improve editing further. It is unclear if combining other base modifications with PS or OM bases will improve mutational KI further. Surprisingly, the modified bases do not appear to compromise the fidelity of HDR. The next-generation sequencing results show that the chemical modifications are processed similarly to normal bases by the HDR machinery since significant unintended mutations were not present in the edited cells. Furthermore, CF airway cells corrected using esDNA templates show restored CFTR function (Figure 2F–H). Further studies may be needed to assess the ability of esDNA templates to correct disease-associated phenotypes in other cell types and to investigate if other modifications affect the fidelity of HDR. The use of internal base modifications to stabilize ssDNA enables the use of 3–10-fold lower concentrations of ssDNA templates (800 nM) compared to other studies (3–10 μM) (8,17,19,39). Although prior studies have reported reduced cell viability in CD34 + cells edited using ssDNA templates with end modifications, we do not observe such a reduction while using esDNA templates (8). Further studies using more sensitive assays may be needed to better characterize the impact of esDNA templates on cell viability. We hypothesize that the editing process described here preserves cell viability due to the 3–10-fold lower concentrations of esDNA templates used. AAV-based templates, which have significantly longer homology arms (>400 bp), can consistently achieve HDR in up to 90% of alleles in the presence of DNA-PKcs inhibitors (Figure 1G and Supplementary Figure S3D). We used homology arms of 80–90 bp because chemically modified ssDNA templates longer than 200 bases are not available commercially. It has been reported that the incorporation of mutations using HDR is more efficient closer to the DSB (40). It has previously been shown that transgene KI in the CFTR locus in airway cells can be increased by ∼2-fold by increasing homology arm length to > 150 bp (41). Therefore, we posit that long ssDNA (>400 bp) with chemical modifications in every tenth to every sixth base may show even better mutational KI. In the absence of long ssDNA with chemical modifications, the placement of DSB close to the target site (Figure 1H) allows the use of esDNA to modify a similar proportion of alleles as AAV templates (80–90% of alleles). The use of esDNA templates to achieve consistent mutational KI relies heavily on the availability of effective DNA-PKcs inhibitors that improve HDR by blocking NHEJ (29). We observed efficient mutational KI in ABCs using ssDNA templates only in the presence of the DNA-PKcs inhibitor, AZD-7648. We observed measurable HDR in iPSCs and CD34 + cells without DNA-PKcs inhibition and the use of esDNA still improved mutational KI relative to ssDNA in CD34 + cells (Figure 3A and B). Thus, although extensive chemical modification of ssDNA improves gene editing in multiple cell types, the need for DNA-PKcs inhibition to obtain detectable mutational KI using ssDNA or esDNA appears to be cell-type dependent. It is possible that the differences in the efficiency of mutational KI correlate with the degradation of ssDNA in these cells. Indeed, it has been reported that TREX1 expression levels in different cell types may predict the efficiency of mutational KI using ssDNA templates (36). This is also consistent with the fact that AAV-based templates enable similar gene modification in all the cell types tested (Supplementary Figure S3E). In attempting to identify nucleases that may be involved in ssDNA degradation, we investigated TREX1 primarily because TREX1 KO was recently reported to improve mutational KI using ssDNA templates in multiple immortalized cell lines and primary cells (36). Notably, iPSCs, which can be edited efficiently using unmodified ssDNA, do not express TREX1. Our study validates the role of TREX1 in three primary cell types and shows that knocking out TREX1 can improve mutational KI using unmodified ssDNA templates. Moreover, since knocking out TREX1 did not improve mutational KI using AAV (Supplementary Figure S3F), our results further show that the role of TREX1 in genome editing is specific to naked ssDNA. It has already been proposed that protecting the ends increases HDR by preventing TREX1 from degrading ssDNA (36). However, our data shows that modifying internal bases along with the ends somehow further improves mutational KI. The mechanism for this protection from TREX1 is unclear. It is possible that TREX1 is blocked or slowed down each time it encounters a modified base. However, the modification of 20 consecutive bases at either end (40 bases in total) is insufficient to block TREX1 despite the fact that a similar number of bases are modified in esDNA-6-PS and esDNA-10-PS templates (25–40 modified bases). Thus, further studies are needed to determine how chemical modifications may alter the interaction of TREX1 with ssDNA. We further compared esDNA templates described here against Alt-R HDR templates from IDT which contain proprietary end modifications in two different loci in two different cell types. Although Alt-R HDR templates improve mutational KI compared to templates modified by PS modifications at the ends in the CFTR locus, the esDNA templates result in further improvements in mutational KI (Figures 4C and 5A). Alt-R HDR templates show considerably lower levels of mutational KI in the HBB locus compared to esDNA templates in both airway cells and CD34 + cells (Figure 4F and G). Thus, esDNA templates show consistently high KI across different loci at least in these two cell types. The Alt-R HDR modifications are also available in double-stranded DNA (dsDNA) templates. We did not evaluate dsDNA templates because they often result in strong innate immune responses in primary cells, which results in increased cell death (42). The esDNA templates described here may be useful in characterizing the biological impact of small INDELS and other point mutations. Previous studies have attempted to model the biology of CFTR variants by introducing them into cell lines in vitro (15). Such models are valuable since they are isogenic and account for the impact of modifier genes that results in widely varying disease phenotypes (43). Similarly, Cas9 and ssDNA templates have been used to model the role of mutations in TP53 in myeloid malignancies in cell lines such as K562 (44). Attempts have also been made to model oncogenic mutations in other genes in HSPCs and epithelial stem cells directly (16,45). The use of esDNA templates with DNA-PKcs inhibition may thus enable efficient generation of primary cells with such variants and aid such studies significantly. Further characterization of aberrant changes in the genome (e.g. translocations) caused by DNA-PKcs inhibition is needed, especially if this approach is used for therapeutic purposes. The use of esDNA templates to generate engineered primary cells is easier to implement and more versatile when compared to alternative methods such as prime editing and base editing, respectively, and does not disrupt the reading frame unlike homology independent targeted integration (HITI) (46–48). Unlike base editing, which is limited to making single base changes, HDR using esDNA templates can create precise insertions, base changes and deletions (not demonstrated in this study). Although prime editing can be used to make all the same changes that are possible using HDR mediated by esDNA templates, optimization of prime editing to obtain high editing efficiencies requires significant effort (49,50). Among competing methods, HITI is just as easy to execute as HDR-based mutational KI and is also effective in non-dividing cells. However, mutational or transgene KI using HITI is not scar free since the process inserts additional sequences that are partial copies of the sgRNA protospacer target on the two ends of the insert. Therefore, KI using HITI will disrupt the reading frame if performed within exons (48,51). In addition, HITI results in the integration of the template in the reverse orientation at almost the same frequency as correctly modified alleles (51). By contrast, HDR-based insertion is scar-free when completed successfully and can be used to modify regions within exons without disrupting the reading frame. Thus, the HDR-based approach presented here is a readily implemented method that enables precise, scar-free mutational KI at high efficiencies. Apart from modeling the impact of mutations, cell therapies have been proposed to treat genetic disorders caused by different mutations and small INDELs such as sickle cell disease and CF (3,5). The ∼40% allelic correction of the F508del mutation reported here enables the partial restoration of CFTR function (∼35% of non-CF) in airway cells obtained from pwCF. This is therapeutically relevant because restoration of ∼20% CFTR function relative to non-CF controls is thought to be sufficient to provide therapeutic benefit (52,53). Similarly, correction of 20–60% alleles in HSPCs from donors with sickle cell disease has been shown to restore mature hemoglobin production and prevent sickling (54). One advantage of short esDNA templates is that they can be readily produced in a few days as opposed to AAV template construction and packaging, which can take a few weeks to months. In addition, our results show that esDNA templates do not activate the DNA damage response pathway which has been reported to compromise the regenerative potential of HSPCs edited using Cas9 and AAV (8,12). A recent study reported that HSPCs edited using end-modified ssDNA (in the presence of a 53BP1 inhibitor) to correct the sickle cell causing mutation in the HBB locus retained their regenerative potential (39). Although they achieved 40–50% alleles with HDR, they report using 10 μM ssDNA. In contrast, esDNA templates allow us to achieve the same level of editing in the HBB locus with > 10× less ssDNA. Further studies are necessary to evaluate the impact of esDNA templates on the regenerative potential of CD34 + cells in vitro and in vivo. Apart from biological reasons, non-viral templates that can be mass produced readily are also of interest for improving access to genetic therapies globally (55). Personalized ex-vivo genetic therapies require complex manufacturing of cells and AAV. In addition, they also require complex medical procedures that are not feasible in low-resource environments. In vivo gene editing using off-the-shelf reagents may enable widespread gene correction to treat genetic diseases globally (55). Notably, in vivo gene editing using lipid nanoparticles (LNP) packaged with Cas9 mRNA, sgRNA and ssDNA has been proposed to correct CFTR mutations in vivo (56). Future studies may test the use of esDNA packaged in LNP for correcting pathogenic mutations using in vivo gene editing. Non-viral in vivo editing approaches are less likely to be limited by immune responses which limit the efficacy of viral vectors such as AAV. LNPs targeting organs such as muscles which were traditionally only targetable using AAV have also been recently reported (57). Thus, the findings presented here can have broad therapeutic implications. Other non-viral templates have been reported for transgene KI including ssDNA with CTS and dsDNA with interstrand cross-linking (20,58). We did not see improved mutational KI using CTS in airway cells. It is possible that the observation may be due to the cell type or locus targeted. Compared to the CTS approach, esDNA does not require the design of special sequences and can thus be applied readily without additional effort or room for errors. Interstrand crosslinked dsDNA has not been reported for use in stem cells such as HSPCs and may be limited by innate immune responses that have been previously shown to limit transgene KI using dsDNA templates in primary cells (59). Although further studies are needed to validate the impact of esDNA templates on the regenerative potential of ABCs and HSPCs, the results presented here represent a starting point for further improving the efficiency of ssDNA templates using internal chemical modifications. In conclusion, our study shows that HDR in primary cells can tolerate modified bases in the template. The results show that chemical modification of internal bases in ssDNA templates at every tenth to every sixth base significantly improves mutational KI using CRISPR-Cas9 in multiple clinically relevant human donor-derived primary cell types (CD34 + cells, T-cells, ABCs and HUVECs). Upon optimization, mutational KI can be achieved in 80–90% of alleles which is comparable to AAV mediated mutational KI. In addition, our study shows that the chemical modifications work by protecting the ssDNA from degradation by TREX1. Overall, we present a broadly applicable method to improve the insertion of short sequences in primary cells using CRISPR-Cas9 and chemically modified ssDNA templates. These tools can enable improved production of cell therapies and enable studies characterizing the biological impact of different genetic variants in primary cells. Supplementary Material