Improving the sensitivity of in vivo CRISPR off-target detection with DISCOVER-Seq+

Discovery of off-target CRISPR–Cas activity in patient-derived cells and animal models is crucial for genome editing applications, but currently exhibits low sensitivity. We demonstrate that inhibition of DNA-dependent protein kinase catalytic subunit accumulates the repair protein MRE11 at CRISPR–Cas-targeted sites, enabling high-sensitivity mapping of off-target sites to positions of MRE11 binding using chromatin immunoprecipitation followed by sequencing. This technique, termed DISCOVER-Seq+, discovered up to fivefold more CRISPR off-target sites in immortalized cell lines, primary human cells and mice compared with previous methods. We demonstrate applicability to ex vivo knock-in of a cancer-directed transgenic T cell receptor in primary human T cells and in vivo adenovirus knock-out of cardiovascular risk gene PCSK9 in mice. Thus, DISCOVER-Seq+ is, to our knowledge, the most sensitive method to-date for discovering off-target genome editing in vivo.

CRISPR-Cas genome editing is a transformative technology with wide-ranging applications, from interrogating basic biological systems to curing genetic diseases in humans 1 . Genome editing by a CRISPR-associated endonuclease such as Streptococcus pyogenes Cas9 relies on the targeted induction of DNA double strand breaks (DSBs), leading to the recruitment of DNA repair factors that repair and potentially modify the genome 2 . However, unintended off-target DNA damage and mutagenesis remain leading concerns for safety and applicability. Therefore, accurate and sensitive methods for discovery of CRISPR-Cas off-target activity are essential 3 .
There are numerous methods for detecting off-target CRISPR-Cas activity, but the majority are limited to purified DNA [4][5][6][7] or restricted cellular systems such as immortalized cell lines or reporter cells [8][9][10] . Measurements in these systems may not translate to in vivo applications. For example, Cas9 behavior such as binding kinetics is very different in vitro 11 , and the epigenome, which is highly divergent between different cell types 12 , strongly influences CRISPR genome editing activity 13 . Therefore, off-target discovery directly in ex vivo and in vivo model systems is highly desired. However, the few methods compatible with these systems may be constrained by limited sensitivity, due to requiring detection of either transient DNA repair protein binding 14 or of mutations that occur at very low frequencies at some off-target sites 15,16 .
In this study, we combined detection of a highly specific DNA repair factor, MRE11 (refs. 14,17,18), with an inhibitor of DNA repair 19 that retains MRE11 residence on genomic DNA, to detect genome-wide CRISPR off-target activity with high sensitivity. Termed DISCOVER-Seq+, this technique enhanced the discovery of CRISPR-Cas-targeted sites in numerous contexts, including in immortalized cell lines, primary human cells and mice at clinically relevant targets. Together, DISCOVER-seq+ represents, to our knowledge, the most Article https://doi.org/10.1038/s41592-023-01840-z Inhibition of Poly (ADP-ribose) polymerase (PARP) and ATM serine/ threonine kinase (ATM) with Olaparib and Ku-55933, respectively, did not exhibit a clear effect, whereas DNA Ligase IV inhibition with Scr7 suppressed MRE11 recruitment (Fig. 1c) 21 . Notably, blocking nonhomologous end joining (NHEJ) by inhibiting DNA-dependent protein kinase catalytic subunit (DNA-PKcs) using 22) or Nu7026 (ref. 23) significantly increased MRE11 recruitment at the target site (P < 1 × 10 −4 ; two-sided Student's t-test) (Fig. 1c). The effect of DNA-PKcs inhibition was consistent across multiple time points (4 h, 12 h, 24 h), three other gRNAs (VEGFA site 2, HEK site 4, FANCF site 2) and/or another cell line (K562) (P < 0.001; two-sided Wilcoxon signed-rank test) (Fig. 1d,e). These results suggest that blocking NHEJ with DNA-PKcs inhibition greatly boosts MRE11 residence at Cas9-targeted sites. Among possible DNA-PKcs inhibitors, Ku-60648 was selected for subsequent experiments due to extensive literature documenting its use in diverse contexts from cell lines to mouse models 19,22 .
We aimed to better characterize the effect of DNA-PKcs inhibition on repair of Cas9-mediated DNA damage. First, we used super-resolution stimulated emission depletion (STED) microscopy [24][25][26] to measure the localization of 53BP1 and BRCA1 foci after Cas9-induced DNA breaks in U2OS cells. 53BP1 corresponds to activation of the NHEJ pathway, whereas BRCA1 is implicated in MRE11-dependent homology-directed repair (HDR) or microhomology-mediated end joining (MMEJ) 27,28 . Using a multi-target gRNA targeting over 100 locations 29 , DNA-PKcs inhibition using Ku-60648 led to a significant reduction in 53BP1 foci relative to BRCA1, consistent with suppression of NHEJ in favor of HDR/MMEJ (P < 1 × 10 −4 ; two-sided Wilcoxon rank-sum test) sensitive method to-date for CRISPR off-target detection that is directly suitable for in vivo applications 20 .

Rationale
Direct detection of the DNA repair response as a proxy for CRISPR nuclease activity has shown promise for genome-wide CRISPR off-target detection. A previous method for off-target detection, DISCOVER-Seq 14 , works by detecting the genome-wide localization of MRE11, a DNA repair factor recruited to Cas9 DSB sites, using chromatin immunoprecipitation (ChIP) followed by sequencing (ChIP-seq) 17,18 . However, sensitivity is relatively low, likely because Cas9 editing is not synchronized and MRE11 resides on DNA only transiently during active repair (Fig. 1a). We hypothesized that if DNA repair could be pharmacologically modulated to encourage MRE11 residence, then MRE11 would accumulate at every Cas9-targeted site in all cells, thus enhancing detection sensitivity with ChIP-seq (Fig. 1b).

DNA-PKcs inhibition on the CRISPR-Cas DNA damage response
To identify inhibitors of DNA repair 20 that can modulate MRE11 residence, we first delivered Cas9 with guide RNA (gRNA) targeting VEGFA site 3 into HEK293T cells. VEGFA site 3 and most other gRNAs used in this study were chosen because they have been well validated in earlier off-target detection methods 8-10 . We exposed cells to one of five DNA repair inhibitors, then performed ChIP with quantitative PCR (qPCR) (ChIP-qPCR) after 12 h to measure MRE11 recruitment at the target site.

DNA-PKcs inhibition improves CRISPR off-target detection
Next, we determined whether increased MRE11 residence with DNA-PKcs inhibition can improve the sensitivity of CRISPR off-target discovery. At 12 h after delivery of Cas9 with FANCF site 2 gRNA into K562 cells, we performed ChIP-seq for MRE11 followed by the BLENDER bioinformatics pipeline 14 to detect all Cas9 target sites genome-wide. Sequencing samples with or without DNA-PKcs inhibition were always normalized to the same number of reads for appropriate comparison. Treatment with Ku-60648 significantly increased MRE11 ChIP-seq enrichment at all discovered on-and off-target sites (P < 1 × 10 −3 ; two-sided Wilcoxon signed-rank test), as measured by the number of reads within a 1.5-kilobase (kb) region around the target site per million total reads, that is, reads per million (RPM) (Fig. 3a-d). MRE11 ChIP-seq enrichment at a specific target site can be visualized as a histogram of base pair coverage along the genome; it exhibits two peaks on each side of the cut site because paired-end Illumina sequencing only reads the ends of DNA fragments that are enriched around the cut site (Extended Data Fig. 1a) 14 . MRE11 levels 10 kb away from the target sites did not significantly increase (P ≥ 0.18; two-sided Wilcoxon signed-rank test), further supporting the lack of additional DNA damage caused by the inhibitor itself (Fig. 3e,f).
To reduce the likelihood of reporting false positive sites, MRE11 ChIP-seq was also performed on cells with the same experimental conditions except without Cas9 (Extended Data Fig. 1b). The final set of off-target sites is therefore determined as the set from the sample with Cas9 subtracted by the set from the corresponding sample without Cas9. For the VEGFA site 2 gRNA, 178 sites were discovered with DNA-PKcs inhibition using Ku-60648, which is an over fivefold increase compared with 35 sites discovered without DNA-PKcs inhibition (that is, DISCOVER-Seq) ( Fig. 3g and Extended Data Fig. 2a). Improved performance with Ku-60648 was consistent across different gRNAs and multiple time points (Fig. 3h,i and Extended Data Fig. 2b). The discovered sites with Ku-60648 included almost all the sites identified using DISCOVER-Seq alone (Fig. 3j). Reassuringly, only a small minority (average of 1.7%) of the initial sites were also found in corresponding negative control samples without Cas9, and therefore deemed to be false positives and removed (Extended Data Fig. 3a,b). We therefore use the term DISCOVER-Seq+ to denote CRISPR off-target discovery that combines MRE11 ChIP-seq (that is, DISCOVER-Seq) 14 with DNA-PKcs inhibition to achieve improved detection sensitivity.
Next, we assessed if any of the new sites discovered by DISCOVER-Seq+ harbor evidence of mutagenesis after CRISPR genome editing. For the FANCF site 2 gRNA, DISCOVER-Seq+ identified 15 target sites, compared with only two with DISCOVER-Seq (Fig. 4a). We exposed cells to Cas9 targeting FANCF site 2 for 4 d (without Ku-60648), then measured indel mutations at each discovered target site using deep amplicon sequencing 30 . Of the 13 off-target sites exclusively discovered by DISCOVER-Seq+, five exhibited detectable indels by amplicon sequencing (Fig. 4b). These results demonstrate that DISCOVER-Seq+ identified new off-target sites with evidence of indel mutations, which DISCOVER-Seq alone failed to detect. Although some newly discovered off-target sites lacked detectable indel mutations by amplicon sequencing, they are still essential to identify because DSBs, even in the absence of mutagenesis, are detrimental to the cell 21   The validity of DISCOVER-Seq+ off-target sites was further confirmed by comparing with published results by an independent technique, GUIDE-seq 8 , which is notably not compatible with primary cells or in vivo applications 14 . For both the FANCF site 2 and VEGFA site 2 gRNAs, half or more of target sites found by DISCOVER-Seq+ were also found by GUIDE-seq, and vice versa. (Fig. 4a,c). These results demonstrate robust overlap in the discovered target sites between GUIDE-seq and DISCOVER-Seq+, providing external validity for DISCOVER-Seq+ while confirming the superiority of DISCOVER-Seq+ in identifying off-target sites.

DISCOVER-Seq+ in editing of primary human cells
We further evaluated the utility of DISCOVER-Seq+ in three applications: patient-derived induced pluripotent stem cell (iPSC) editing, generating engineered T cells for cancer immunotherapy and in vivo characterizations of CRISPR-based therapies in mouse models. First, we used DISCOVER-Seq+ to improve off-target detection in iPSCs. DISCOVER-Seq+ in WTC-11 iPSCs 31 discovered over twofold more off-target sites at VEGFA site 2 compared with DISCOVER-Seq ( Fig. 5a). At all discovered off-target sites, MRE11 ChIP-seq enrichment was also significantly increased (P < 1 × 10 −5 ; two-sided Wilcoxon signed-rank test) (Fig. 5b,c and Extended Data Fig. 3c). For the same VEGFA site 2 gRNA, there were differences in off-target sites between three different cell lines (HEK293T, K562 and WTC-11 iPSC) (Extended Data Fig. 3d,e).
Next, we applied DISCOVER-Seq+ ex vivo to knock-in of a cancer neoantigen-specific transgenic T cell receptor (tgTCR) construct into primary human T cells 32-34 . We electroporated Cas9 targeting TRA (T Cell Receptor Alpha Locus) along with a 4,699-base pair (bp) homology-directed repair template (HDRT) encoding a tgTCR specific for HLA-A*02 loaded with mutant p53 R175H peptide 32 , then performed DISCOVER-Seq+ 12 h later (Fig. 5d). The specific R175H mutation that is targeted by the tgTCR is the most prevalent p53 gain-of-function mutation in human cancers 33 . DISCOVER-Seq+ (with Ku-60648) identified 20 off-target sites genome-wide compared with four with DISCOVER-Seq (Fig. 5e), and led to significantly greater MRE11 enrichment at all discovered sites (P = 1 × 10 −3 ; two-sided Wilcoxon signed-rank test) ( Fig. 5f-h). In contrast, samples without Cas9 exhibited no change in enrichment with Ku-60648, further confirming that the inhibitor alone does not induce damage (P = 0.69; two-sided Wilcoxon signed-rank test) (Fig. 5i).
DISCOVER-Seq+ also has the potential to compare off-target profiles between different types of CRISPR nucleases. As a proof of concept, we compared the performance of Cas9 with Cas12a (Cpf1), targeting the same position in TRA. DISCOVER-Seq+ at 12 h after Cas12a        Fig. 3f-h). Together, our preliminary analysis using DISCOVER-Seq+ revealed that Cas12a maintained adequate tgTCR integration rates while eliminating detectable off-target damage. Importantly, these experiments demonstrated that DISCOVER-Seq+ is directly compatible with CRISPR knock-in using a homology template in primary human T cells.

DISCOVER-Seq+ in vivo
Finally, we evaluated DISCOVER-Seq+ in vivo by targeting the cardiovascular risk gene PCSK9 in mouse liver 36 . We retro-orbitally injected adenovirus encoding Cas9 and PCSK9 gRNA into ten, 8-10-week-old, male C57BL/6J mice, followed by peritoneal injection of either 25 mg kg −1 Ku-60648 (that is, DISCOVER-Seq+) or vehicle (that is, DISCOVER-Seq) twice daily (b.i.d.) (Fig. 6a). We selected the specific PCSK9 gRNA that was also used in the original DISCOVER-Seq study for direct comparison 13 . Ku-60648 has been evaluated as a drug for chemo-sensitization in cancer therapy, exhibits good pharmacokinetics and strongly penetrates tissue including tumors 22 . Mice were killed after 24 h to collect the liver for MRE11 ChIP-seq. DISCOVER-Seq+ mice exhibited increased MRE11 ChIP-seq signal in their liver compared with those without DNA-PKcs inhibition (P < 1 × 10 −4 ; two-sided Wilcoxon rank-sum test) (Fig. 6b-d and Extended Data Fig. 3i). An average of 27 target sites were identified with DISCOVER-Seq+ compared with 18 sites with DISCOVER-Seq across five biologically independent replicates (P < 0.01; two-sided Student's t-test) (Fig. 6e). The identified sites strongly overlap between the two methodologies and with sites identified in the original DISCOVER-Seq study (Fig. 6f) (ref. 14). Pooling sequencing reads across all five replicates identified 98 target sites with DISCOVER-Seq+ versus 49 with DISCOVER-Seq (Fig. 6g). Together, these results demonstrate that DISCOVER-Seq+ is compatible with direct measurement of genome-wide off-target editing in vivo (Supplementary Table 1).

Discussion
This study designed and validated DISCOVER-Seq+, the most sensitive method to-date for detecting CRISPR-Cas off-target activity in primary cells and in vivo, to our knowledge. As CRISPR becomes an increasingly a b  All the target sites identified by 'DSeq (this work)' are also found in the other two groups. g, PCSK9 Cas9 target sites detected using DISCOVER-Seq (left) versus DISCOVER-Seq+ (right) in mouse liver. Off-target detection for each condition was performed on sequencing data pooled across five biologically independent mouse replicates. **P < 0.01. Article https://doi.org/10.1038/s41592-023-01840-z feasible approach for therapeutic genome editing 1,20,32,33,36-38 , evaluation of CRISPR off-target activity directly in clinically translatable applications is crucial. Even if a specific CRISPR gRNA does not have detectable off-target activity in one particular cell type, this may not translate to other cell types, tissues or organisms. Directly measuring off-target sites in the system of interest, whether in primary T cells, mice or even nonhuman primates, is therefore essential. By combining unparalleled detection sensitivity with high versatility in ex vivo and in vivo applications, we believe DISCOVER-Seq+ will find widespread use as the state-of-the-art technology for off-target detection in diverse applications of CRISPR editing.
DISCOVER-Seq+ identified off-target sites with evidence of mutagenesis that DISCOVER-Seq alone failed to detect 8,39 . Furthermore, because DISCOVER-Seq+ directly measures off-target DNA damage rather than mutagenesis, it also discovered sites that did not have detectable indel mutations. This is important for three reasons: (1) One study found that less than 15% of DSBs convert to indels in a single damage cycle 17 ; therefore, off-target sites that are not frequently damaged may not result in indels that can appear in amplicon sequencing-based queries. (2) DNA damage outcomes such as DSBs are highly detrimental regardless of mutagenesis, especially to primary cells, by inducing widespread epigenetic changes and perturbing native cellular functions such as cell division and transcription 21,29 . (3) Evaluating indels alone may also miss complex DNA damage outcomes such as large deletions and translocations 8,40 . Furthermore, deep amplicon sequencing may miss rarer off-target sites due to a lack of enrichment of altered DNA molecules and insufficient sequencing depth 41 . For all these reasons, enrichment for sites with evidence of CRISPR-Cas-induced DNA damage is a more holistic readout of off-target activity than indels alone. In summary, DISCOVER-Seq+ directly detects genome-wide CRISPR-induced DNA damage, regardless of whether these sites become mutated, with unprecedented sensitivity and versatility.
DNA-PKcs inhibition could also influence the performance of other methods to detect CRISPR-Cas activity. BLISS/BLESS 9 could be improved with DNA-PKcs inhibition by increasing the quantity of unrepaired DSBs at the time of evaluation. In contrast, GUIDE-seq 8 would likely be impaired because incorporation of its double-stranded oligodeoxynucleotide relies on NHEJ, which is directly inhibited by DNA-PKcs inhibitors. Future studies should explore whether other strategies of inhibiting DNA repair may improve detection of CRISPR-Cas off-target activity.
Limitations of DISCOVER-Seq+ include the need for the DNA-PKcs inhibitor to exert its effect, including in vivo. We believe this is not a major concern because Ku-60648 is a small molecule previously shown to have good bioavailability and tissue penetration, including into tumors 22,23 . In contrast, the main barrier to applicability in other organs in vivo remains the efficiency of CRISPR-Cas delivery to nonliver organs 1,14,36 . In addition, the improvement in the number of discovered off-target sites with DNA-PKcs inhibitor in mice 14 is a factor of two, which is lower than in cell lines. This may be due to reduced inhibitor or Cas9 concentrations in the liver; further optimization of drug dosing and Cas9 delivery may improve performance.
Identifying off-target genome editing is a major barrier to applications of CRISPR-Cas systems. By leveraging MRE11 ChIP-seq with DNA-PKcs inhibition, DISCOVER-Seq+ provides the highest detection sensitivity to-date in systems ranging from ex vivo editing of primary human cells to in vivo editing of mice, setting the standard for genome-wide CRISPR off-target discovery. DISCOVER-Seq+ has the potential to validate the specificity profile of genome editing at numerous stages of the therapeutic development pipeline, from cell lines and primary cells to mice and potentially nonhuman primates 1,18,32,33,36-38 .

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41592-023-01840-z.  (Kreitzer et al., 2013), was used for all iPSC experiments in this study. We followed the guidelines of Johns Hopkins Medical Institute for the use of this hiPSC line. Briefly, frozen WTC-11 cells were first thawed in a 37 °C water bath and washed in Essential 8 Medium (E8 medium; Thermo Fisher Scientific, no. A1517001) by centrifugation. After resuspension, WTC-11 cells were plated onto a 6-cm cell culture dish pre-coated with human embryonic cell-qualified Matrigel (1:100 dilution, Corning, no. 354277). Plate coating should be performed for at least 2 h. Subsequently, 10 μM ROCK inhibitor (Y-27632; STEMCELL, no. 72308) was supplemented into the E8 medium to promote cell growth and survival. For subculture, WTC-11 cells were dissociated from the plate using Accutase (Sigma, no. A6964) and passaged every 2 d. WTC-11 cells were maintained in an incubator at 37 °C with 5% CO 2 .

Ethics statement and mouse husbandry
All mouse studies were carried out in accordance with guidelines and approval of the Johns Hopkins University Animal Care and Use Committee (Protocol no. MO20M274). The 8-10-week-old male C57BL/6J mice (The Jackson Laboratory) were housed in a facility with 12-h light/12-h dark cycle at 22 ± 1 °C and 40 ± 10% humidity. Teklad Global 18% protein rodent diet and tap water were provided ad libitum.

Immunofluorescence and imaging by STED microscopy
U2OS cells stably expressing Cas9-EGFP cells were seeded onto 35-mm, glass bottom dishes and transfected with multi-target guide RNAs (mtgRNAs) for 12-24 h. Cleavage was activated by ultraviolet light for 1 min. To fix cells, 4% pre-warmed paraformaldehyde in 1 × PBS was used for 10 min. After rinsing three times with 1 × PBS, cell membrane permeabilization was performed with Triton-X used for 10 min. Then, 2% w/v BSA in 1 × PBS was used for blocking for 1 h and at room temperature. The primary antibodies, mouse anti-BRCA1 (sc-6954 D9, Santa Cruz Biotechnology), rabbit anti-53BP1 (ab172580, Abcam), were diluted (1:500) in 1 × PBS and directly added into the imaging dish. After 1 h of incubation, the primary antibody was removed, and the sample was washed with 1 × PBS three times. The samples were then incubated for 30 min with the secondary antibodies, goat anti-Mouse Alexa-594 (A-21235, ThermoFisher) and goat anti-Rabbit Atto-647N (40839, Sigma), diluted (1:1,000) in 1 × PBS. Finally, the sample was rinsed three times and mounted with Prolong Diamond mounting medium (Thermo Fisher Scientific) overnight.
All STED images were obtained using a home-built two-color STED microscope (Han and Ha 24 ; Ma and Ha 25 ). In short, a femtosecond laser beam with a repetition rate of 80 MHz from a Ti:Sapphire laser head (Mai Tai HP, Spectra-Physics) is split into two parts: one part produces an excitation beam coupled into a photonic crystal fiber (Newport) for wide-spectrum light generation. The beam is further filtered by a frequency-tunable acoustic optical filter (AA Opto-Electronic) for multi-color excitation. The other part of the laser pulse is temporally stretched to ~300 ps (with two 15-cm-long glass rods and a 100-m-long polarization-maintaining single-mode fiber, OZ optics), collimated and expanded, and wave-front modulated with a vortex phase plate (VPP-1, RPC photonics). This modulation produces a hollow STED spot generation to de-excite the fluorophores at the periphery of the excitation focus, thus improving the lateral resolution. The STED beam is set at 765 nm with a power of 120 mW at the back focal plane of the objective (NA = 1.4 HCX PL APO 100X, Leica), and the excitation wavelengths are set as 594 nm and 650 nm for imaging Alexa-594-and Atto-647N-labeled targets, respectively. Two avalanche photodiodes detect the fluorescent photons (SPCM-AQR-14-FC, Perkin Elmer). The images are obtained by scanning a piezo-controlled stage (Max311D, Thorlabs) controlled with the Imspector data acquisition program.

Electroporation of Cas9 RNP and DNA-PKcs inhibitor delivery into cell lines and iPSCs
CRISPR RNA (crRNA) and trans-activating crRNA (tracrRNA) sequences are listed in Supplementary Table 2. First, 2 μl of 100 μM crRNA was mixed with 2 μl of 100 μM tracrRNA (Integrated DNA Technologies) and heated to 95 °C for 5 min in a thermocycler, then allowed to cool on the benchtop for 5 min. To form the ribonucleoprotein (RNP) complex, 3 μl of 10 μg μl −1 (~66 μM) purified Cas9 was mixed with the annealed 4 μl of 50 μM cr:tracrRNA, then 8 μl of dialysis buffer (20 mM HEPES pH 7.5 and 500 mM KCl, 20% glycerol) was mixed in for a total of 15 μl. This solution was incubated for 20 min at room temperature to allow for RNP formation.
HEK293T cells were maintained to a confluency of ~90% before electroporation. Twelve million cells were trypsinized with 5 min of incubation in the incubator, then 1:1 of DMEM complete was added to inactivate trypsin. This mixture was centrifuged (3 min, 200g) and supernatant removed, followed by resuspension of the cell pellet in 1 ml of PBS, centrifugation (3 min, 200g) and finally complete removal of supernatant. Then, 90 μl of nucleofection solution (16.2 μl of Supplement solution mixed with 73.8 μl of SF solution from SF Cell Line 4D-Nucleofector X Kit L, Lonza) was mixed thoroughly with the cell pellet. The 15 μl of RNP solution was mixed in along with 2 μl of Cas9 Electroporation Enhancer (Integrated DNA Technologies). The entirety of the final solution (approximately 125 μl) was transferred to one well of a provided cuvette rated for 100 μl. Electroporation was then performed according to the manufacturer's instructions on the 4D-Nucleofector Core Unit (Lonza) using code CA-189. Some white residue may appear in the cell mixture after electroporation, but that is completely normal. A total of 400 μl of DMEM complete was used to completely transfer the cells out of the cuvette, before plating to culture wells pre-coated with 1:100 collagen. A minimum of 4 million cells are used for each ChIP. For time-resolved experiments, this means one electroporation equates to three samples.
For WTC-11 iPSCs, cells were dissociated from the plate using accutase (Sigma, no. A6964). Electroporation was performed using the Lonza P3 Primary Cell 4D-Nucleofector X Kit L using code CA-137, on 10 million cells, and using 65 μl of the P3 solution mixture with EP enhancer per electroporation cuvette (compared with 90 μl of comparable SF solution mixture for HEK293T cells). After electroporation, cells were resuspended in E8 medium supplemented with 10 μM ROCK inhibitor (Y-27632; STEMCELL, no. 72308), and plated onto a 10-cm cell culture dish pre-coated with human embryonic cell-qualified Matrigel (1:100 dilution, Corning, no. 354277) for at least 2 h.

Adenovirus and DNA-PKcs inhibitor delivery into mice
For in vivo gene delivery, 8-10-week-old mice were anesthetized with 2.5% isofluorane/oxygen mixture. Mice received a single retro-orbital injection of 1 × 10 9 infectious adenoviral particles (Ad-Cas9-U6-mPCSK9-sgRNA) in 100 μl of sterile saline. Immediately following, mice received intraperitoneal delivery of KU-60648 dosed at 25 mg kg −1 (or vehicle only) in 100 μl of citrate buffer, or 100 μl of citrate buffer vehicle. Mice received a dose of KU-60648 or vehicle every 12 h via intraperitoneal injection.

Extraction of mouse liver into cell suspension
At the experimental endpoint of 12 h, mice were anesthetized with isofluorane and euthanized via cervical dislocation. Liver tissue was collected, washed three times in 2 ml of PBS with 1 × protease inhibitor (Halt Protease Inhibitor Cocktail, Thermo), then the tissue disrupted in 1 ml of PBS with 1 × protease inhibitor using a loose-fitting Dounce homogenizer. For MRE11 ChIP-seq, homogenized tissue was placed on ice and used immediately.

DISCOVER-Seq+, DISCOVER-Seq and MRE11 ChIP-seq
The protocol was adapted from previous literature (Wienert et al. 14 ) and describes the reagents for one MRE11 ChIP-seq experiment. No animals or data points were excluded from the analysis.
For adherent cells, approximately 10 million cells were gently rinsed with room temperature PBS, washed off the plate using 10 ml of DMEM with assistance from pipette squirts and a cell scraper, then transferred to a 15-ml Falcon tube. For suspension cells, approximately 10 million cells were transferred to a 15-ml Falcon tube, spun down at 200g for 1 min, decanted, then resuspended with 10 ml of DMEM. Then, 721 μl of 16% formaldehyde (methanol-free) was added and the tube was mixed by inversion at room temperature: 7 min for WTC-11 iPSCs, 12 min for HEK293T cells or 15 min for K562 cells. For mouse liver, 300 μl of Dounce-homogenized mouse liver was diluted into 10 ml of PBS. Then, 721 μl of 16% formaldehyde (methanol-free) was added and mixed by inversion at room temperature for 10 min.
Afterwards, 750 μl of 2 M glycine was added to quench the formaldehyde. Cells were spun down at 1,200g and 4 °C for 3 min, then washed with ice-cold PBS twice, spinning down with the same centrifugation conditions. Pellets can be decanted, flash-frozen, then stored at −80 °C for later use. Cells were then resuspended in 4 ml of lysis buffer LB1 (50 mM HEPES, 140 mM NaCl, 1 mM EDTA, 10% glycerol, 0.5% Igepal CA-630, 0.25% Triton X-100, pH to 7.5 using KOH, then add 1 × protease inhibitor right before use) for 10 min at 4 °C, then spun down at 2,000g and 4 °C for 3 min. The supernatant was decanted. Cells were then resuspended in 4 ml of LB2 (10 mM Tris-HCl pH 8, 200 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, pH to 8.0 using HCl, then add 1 × protease inhibitor right before use) for 5 min at 4 °C, spun down with the same protocol and the supernatant decanted. Cells were then resuspended in 1.5 ml of LB3 (10 mM Tris-HCl pH 8, 100 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 0.1% Na-deoxycholate, 0.5% N-lauroylsarcosine, pH to 8.0 using HCl, then add 1 × protease inhibitor right before use) and transferred to 2-ml Eppendorf tubes for sonication with 50% amplitude, 30 s ON, 30 s OFF for 12 min total time (Fisher 150E Sonic Dismembrator). Sample was spun down at 20,000g and 4 °C for 10 min, and supernatant was transferred to 1.5 ml of LB3 in a 15-ml falcon tube. Then, 300 μl of 10% Triton X-100 was added, and the entire solution was well mixed by gentle inversion.
Beads pre-loaded with antibodies were prepared before cell collection. First, 50 μl of Protein A beads (ThermoFisher) were used per immunoprecipitation and transferred to a 2-ml Eppendorf tube on a magnetic stand. Beads were washed twice with blocking buffer BB (0.5% BSA in PBS), then resuspended in 100 μl of BB per immunoprecipitation. Then, 4 μl of MRE11 antibody (Novus NB100-142) per immunoprecipitation was added and placed on a rotator for 1-2 h. Right before immunoprecipitation, the 2-ml tube was placed on a magnetic rack and washed three times with BB, before resuspending in 50 μl of BB per electroporation (EP). Next, 50 μl of beads in BB were transferred to each 3-ml immunoprecipitation (effective 1:750 dilution) and placed at 4 °C on a rotator for 6+ hours.
Samples were transferred to 2-ml Eppendorf tubes on a magnetic stand, washed six times with 1 ml of RIPA buffer (50 mM HEPES, 500 mM LiCl, 1 mM EDTA, 1% Igepal CA-630, 0.7% Na-deoxycholate, pH to 7.5 using KOH), then washed once with 1 ml of TBE buffer (20 mM Tris-HCl pH 7.5, 150 mM NaCl) before decanting. Beads containing DNA from ChIP were mixed with 70 μl of elution buffer EB (50 mM Tris-HCl pH 8.0, 10 mM EDTA, 1% SDS) and incubated at 65 °C for 6+ hours. Then, 40 μl of TE buffer was mixed in to dilute the SDS, followed by 2 μl of 20 mg ml −1 RNaseA (New England BioLabs) for 30 min at 37 °C. Next, 4 μl of 20 mg ml −1 Proteinase K (New England BioLabs) was added and incubated for 1 h at 55 °C. The genomic DNA was column purified (Qiagen) and eluted in 35 μl of nuclease-free water.
Oligo sequences for library preparation are listed in Supplementary Table 4. End-repair/A-tailing was performed on 17 μl of DNA from ChIP using NEBNext Ultra II End Repair/dA-Tailing Module (New England BioLabs), followed by ligation (MNase_F/MNase_R) with T4 DNA Ligase (New England BioLabs). Thirteen cycles of PCR using PE_i5 and PE_i7XX primer pairs were performed for MRE11 ChIP samples to amplify sequencing libraries. Samples were pooled and quantified with QuBit (Thermo), Bioanalyzer (Agilent) and qPCR (BioRad).
Cell line samples were sequenced on a NextSeq 500 (Illumina) using paired 2 × 36-bp reads. Mouse liver samples were sequenced on a DNBSEQ PE100 (BGI) using paired 2 × 50-bp reads. All ChIP-seq raw reads in FASTQ format and processed alignments in BAM format are uploaded to the Sequence Read Archive under BioProject accession PRJNA801688.
Reads were demultiplexed after sequencing using bcl2fastq. Paired-end reads were aligned to hg38, hg19 or mm10 using bowtie2. To ensure fair comparison between DISCOVER-Seq+ (with DNA-PKcs inhibitor) and DISCOVER-Seq (without inhibitor), equal numbers of sequencing reads were obtained by subsetting for each set of samples. Samtools was used to filter for mapping quality ≥ 25, remove singleton reads, convert to BAM format, remove potential PCR duplicates and index BAM-formatted output files. The software that coordinates these steps as well as performs subsequent analyses is open source (https:// github.com/rogerzou/DSeqPlus).
BLENDER (Wienert et al. 14 ) (https://github.com/staciawyman/ blender) was used to determine Cas9 off-target sites, outputting a curated list of all off-target sites with corresponding visualization. A more sensitive cutoff threshold of 2 (-c 2) was used for all samples, except a threshold of 3 (-c 3) for the merged PCSK9 samples.

CRISPR-Cas9 or Cas12a editing of primary human T cells
Engineered T cells expressing a TP53 R175H:HLA-A*02:01-specific T cell receptor (TCR) under control of an EF1-alpha promoter were generated via CRISPR-Cas-mediated HDR electroporation as follows. Nucleotide sequences of the TCR of interest, promoter and homology arms for the TRAC gene locus were generated by de novo gene synthesis (GeneArt). A 4,699-bp HDRT double-stranded DNA was generated by amplification from a plasmid template using the Q5 High-Fidelity 2X Master Mix (New England BioLabs) with primers containing truncated Cas9 target sequences (IDT) (PMID: 31819258). Amplicon DNA was purified with AMPure beads (Beckman Coulter), eluted in water and quantified. Purified PCR products were analyzed by agarose gel electrophoresis to assess correct amplicon size and purity. T cells were isolated by negative selection using immunomagnetic cell separation (EasySep Human T Cell Isolation Kit) from cryopreserved healthy donor peripheral blood mononuclear cells collected via leukapheresis. Purified CD3 + T cells were activated with Dynabeads Human T-Activator CD3/CD28 (Ther-moFisher) at a 1:2 bead-to-cell ratio in RPMI-1640 (ATCC) supplemented with 10% FBS (HyClone Defined), 100 units per ml of Penicillin (Gibco), 100 μg ml −1 Streptomycin (Gibco), 100 IU ml −1 recombinant human IL-2 (Proleukin, Prometheus Laboratories) and 5 ng ml −1 recombinant human IL-7 (BioLegend) at 37 °C, 5% CO 2 . After 48 h and before electroporation, beads were removed with a magnet. Cas9 RNP targeting TRAC (AGAGTCTCTCAGCTGGTACA) or Cpf1 (Cas12a) RNP targeting a juxtaposed nucleotide sequence in TRAC (GAGTCTCTCAGCTGG-TACAC) was assembled by mixing the appropriate sgRNA (IDT) with either Alt-R S.p. Cas9 nuclease V3 (IDT) or Alt-R A.s. Cas12a (Cpf1) Ultra nuclease (IDT) and matching ssDNA Electroporation Enhancer (IDT) and incubating the mixture at room temperature for 15 min. RNPs were Article https://doi.org/10.1038/s41592-023-01840-z mixed with 0.5 μg of the same HDRT and incubated for 5 min at room temperature. To edit activated T cells, 20 μl of T cells was resuspended in P3 buffer at 5 × 10 7 cells per ml (Lonza) and added to the electroporation mixture. Electroporation was performed with a 4D-Nucleofector X Unit (Lonza) in 16-well cuvettes using pulse code EH115. After electroporation, T cells were recovered by immediately adding 80 μl of warm, cytokine-free T cell medium to the cuvettes and incubation at 37 °C for 15 min. Then, T cells were diluted in T cell growth medium containing 100 IU ml −1 recombinant human IL-2 and 5 ng ml −1 recombinant human IL-7 in the presence of 1 μM Ku-60648 or vehicle (dimethylsulfoxide) and incubated for 12 h at 37 °C, 5% CO 2 . A fraction of electroporated T cells for each condition was maintained in T cell growth medium with human IL-2 and IL-7 for 7 d before analysis for gene editing rates and surface expression of TP53 R175H:HLA-A*02:01-specific TCR by flow cytometry.

High-throughput sequencing of genomic DNA samples
Genomic DNA was extracted using the DNeasy Blood & Tissue Kit (Qiagen, 69504) following manufacturer instructions. Approximately 1 million cells were used from cell lines and iPSCs. Approximately 10-20 μl of mouse liver cell suspension was used out of 1.5 ml total, and the genome extraction protocol included the Buffer ATL step for tissue lysis.
Genomic DNA samples were amplified with PCR using Q5 Hot-Start High-Fidelity 2X Master Mix (New England BioLabs, M0494). Primer pairs for all sequences are listed in Supplementary Table 3. For example, the primer set for amplifying around the FANCF site 2 on-target site is NGS_Fs2_ON_F and NGS_ Fs2_ON_R. After amplicon PCR, cleanup was performed using 1.4 × AMPure XP (Beckman Coulter A63881 https://www.ncbi.nlm.nih.gov/nuccore/A63881) following the manufacturer's instructions. Dual-indexing PCR was performed using KAPA HiFi HotStart ReadyMix (Roche, 07958935001) and PCR cleanup was performed using 1 × AMPure XP. Samples were quantified using QuBit (Thermo Fisher Scientific), pooled, diluted and loaded onto a MiSeq (Illumina). Sequencing was performed with the following number of cycles, '151|8|8|151', with the paired-end Nextera sequencing protocol.
Sequencing reads were either demultiplexed automatically using MiSeq Reporter (Illumina) or with a custom Python script to individual FASTQ files. For indel calling, sequencing reads were scanned for exact matches to two 20-bp sequences that flank +/−20 bp from the ends of the target sequence. If no exact matches were found, the read was excluded from analysis. After additional filtering for an average quality score > 20, an indel is defined as a sequence that differs in length from the reference length.

Flow cytometry and analysis of primary human T cells
Surface staining for flow cytometry was performed by washing T cells in PBS, pH 7.4, followed by staining with LIVE/DEAD Fixable Violet Stain (ThermoFisher) at recommended concentration for 30 min on ice in the dark. T cells were then resuspended in Cell Staining Buffer (BioLegend) plus relevant antibodies (APC anti-NGFR (BioLegend)) and PE-conjugated HLA-A*02:p53 R175H tetramer (NIH Tetramer Facility) for 30 min at room temperature in the dark. Cells were washed twice in Cell Staining Buffer before resuspension for analysis. Flow cytometric analysis was performed on an IntelliCyt iQue Screener PLUS VBR (Sartorius). FlowJo v.10 was used for flow cytometry data analysis.
The gating strategy used for primary human T cells is shown in Extended Data Fig. 2. For all experiments, debris was first excluded by a morphology gate based on FSC-H and SSC-H. Nonsinglets were excluded from analysis by a single-cell gate based on FSC-H and FSC-A. Live cells were selected by gating on cells negative for staining with LIVE/DEAD Fixable Violet dye (405-nm excitation, Invitrogen). Anti-NGFR APC + /tetramer PE + cells were gated to define successfully edited live, single T cells using appropriate compensation with single-stained controls.

Statistical analyses
Student's t-test was used to compare samples whose distributions approximate the normal distribution, with the unpaired t-test for independent samples and paired t-test for paired samples. Wilcoxon rank-sum and signed-rank tests were used to compare independent and paired samples, respectively, whose distributions do not approximate the normal distribution. Two-sided tests were used in all cases. The P values for each statistical test are presented in the corresponding figure legend.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability
All sequencing data are uploaded to the Sequence Read Archive under BioProject accession PRJNA801688. All other data are available within the paper and its Supplementary Information files. Source data are provided with this paper.