Biased genome editing using the local accumulation of DSB repair molecules system

Selective genome editing such as gene knock-in has recently been achieved by administration of chemical enhancer or inhibitor of particular DNA double-strand break (DSB) repair pathways, as well as overexpression of pathway-specific genes. In this study, we attempt to enhance the efficiency further to secure robust gene knock-ins, by using the local accumulation of DSB repair molecules (LoAD) system. We identify CtIP as a strong enhancer of microhomology-mediated end-joining (MMEJ) repair by genetic screening, and show the knock-in-enhancing effect of CtIP LoADing. Next-generation sequencing reveals that CtIP LoADing highly increases the frequency of MMEJ-mediated integration. Selection-free, simultaneous triple gene knock-ins are also achieved with the CtIP-LoADing strategy. Moreover, by replacing the LoADing molecules and targeting strategies, this system can be applied for other specific genome engineering purposes, such as introducing longer deletions for gene disruption, independently introducing multiple mutations without chromosomal deletion, and efficiently incorporating a single-stranded oligodeoxynucleotide donor.

R ecent advances in genome editing technology have enabled functional genomics study in any cell or organism, by disrupting endogenous genes, incorporating polymorphisms or short tags, and integrating exogenous genes, such as fluorescent reporters. Of these techniques, gene cassette knock-in is especially useful for monitoring endogenous gene expression and localization, which is essential for functional annotation of endogenous genes. However, previous studies never achieved simultaneous generation of various patterns or combinations of multiple reporters in multiple genomic loci; although, the capacity of double knock-in was confirmed in mouse androgenetic haploid embryonic stem cells 1 ; thus, gene insertion has been independently conducted at each locus, and serial cloning steps have been required for double or triple knock-in at different gene loci.
Gene-labeling strategies assisted by genome editing tools such as CRISPR-Cas9 have typically been mediated by homologous recombination (HR) 2,3 , and alternative methods utilizing nonhomologous end-joining (NHEJ) 4,5 or microhomology-mediated end-joining (MMEJ) [6][7][8][9] have recently been established and refined for highly efficient knock-in. The HR-mediated method often results in highly accurate knock-in, but the throughput of targeting vector construction and integration efficiency are inferior to those of end-joining pathway-mediated strategies. The NHEJ-mediated homology-independent targeted integration (HITI) system is especially useful for in vivo gene knock-in with high efficiency 5 ; however, this system cannot assign multiple donors with different genomic loci simultaneously because there are no homology arms on the targeting donor vector for the HITI system. To overcome these shortcomings, the MMEJ-mediated precise integration into target chromosome (PITCh) system 6,8 is expected to facilitate high-throughput, simultaneous generation of multiplex knock-in cell libraries because it utilizes short but distinguishable microhomologies (≤40 bp) and results in superior knock-in efficiency compared with the HR-mediated method.
Here we establish the biased genome editing system applicable in PITCh knock-in and beyond, and show its usefulness by enhancing various intended genome editing outcomes including selection-free generation of triple gene knock-in clones by the LoADed PITCh system.

Results
Systems development of biased genome editing. Although, we previously confirmed that the overexpression of exonuclease 1 (Exo1) enhanced the efficiency of PITCh knock-in 9 , we attempted to enhance the efficiency further to secure robust gene knock-ins at various genomic loci. We newly identified CtIP as a strong MMEJ enhancer by genetic screening (Supplementary Fig. 1). CtIP, also known as RB binding protein 8 (RBBP8), is an endresection enzyme of DNA double-strand break (DSB) ends 10 ; thus, its overexpression might induce enhanced generation of single-stranded DNA ends, which is essential for the MMEJ and HR repair pathways 11 . Furthermore, we added MS2 sequences at the stem loops of single-guide RNA (sgRNA), and fused MS2 coat protein with CtIP, repurposing the system previously established for highly active transcriptional control, named synergistic activation mediator (SAM) 12 . Using this local accumulation of DSB repair molecules (LoAD) system, strong pathway choice at the CRISPR target site should be achieved (Fig. 1a).
To confirm the effect of MS2-CtIP LoADing, we first designed PITCh-mediated mNeonGreen 13 knock-in in front of the stop codon of the endogenous calnexin (CANX) gene, expressing CANX-mNeonGreen fusion protein (Fig. 1b). FACS analysis revealed that the knock-in-enhancing effect of MS2-CtIP LoADing is much greater than those of CtIP, MS2-Exo1, and Exo1 (Fig. 1c, Supplementary Fig. 2). The LoADing of other related proteins such as Lig3 and Nbs1 did not significantly enhance the PITCh knock-in (Supplementary Fig. 3). The effect of MS2-CtIP LoADing was also observed at four other gene loci (ATP5B, PARP1, FBL, and RPL11) with about twofold activation on average (Fig. 1d, Supplementary Fig. 4). Expected patterns of fluorescence localization were also confirmed by microscopic observation (Fig. 1e, Supplementary Fig. 5). The knock-inenhancing effect of MS2-CtIP LoADing was also confirmed in other cell lines, CHO-K1 ( Supplementary Fig. 6a) and HeLa ( Supplementary Fig. 6b) cells. These assays demonstrated the robust, locus-, and cell type-independent knock-in-enhancing effect of MS2-CtIP LoADing.
We next investigated whether the knock-in cells with the LoADed PITCh system were heterozygous or homozygous. Eight and three single-cell clones showing fluorescence in which mNeonGreen cDNA was knocked-in at the ATP5B locus were established with the LoADed PITCh and conventional PITCh strategies, respectively. Subsequently, knock-in and non-knock-in alleles were simultaneously amplified by out-out PCR analysis ( Supplementary Fig. 7a). The results indicated that all three clones obtained with the conventional PITCh system were heterozygous, whereas in two out of eight clones obtained with the LoADed PITCh system, no non-knock-in amplicons were observed, although one of these clones showed another longer amplicon, possibly carrying the plasmid backbone as well as the intended knock-in insert ( Supplementary Fig. 7b). These results suggest that homozygous knock-in cells would be established using the LoADed PITCh system even in aneuploid cells, such as HEK293T. Using the same cell clones, we further performed Southern blot analysis to investigate whether off-target integrants were detected ( Supplementary Fig. 8a). No evidence of off-target integrants was detected among eight clones, except that one minor band showing slightly longer size than expected was observed in one out of eight clones ( Supplementary Fig. 8b). Based on this, high risk of off-target integration was not observed in the LoADed PITCh system, as well as the conventional PITCh system.
Another concern for the LoAD system is cytotoxicity of MS2-CtIP. We co-transfected mCherry-expressing plasmid either with mock, CtIP, MS2-CtIP, or zinc-finger nuclease (ZFN) vector that was previously characterized to have cytotoxicity 9 . FACS analysis revealed that CtIP and MS2-CtIP vectors did not affect cell viability, while the ZFN vector decreased cell viability (Supplementary Fig. 9a). We further analyzed the ability of DSB repair when MS2-CtIP was overexpressed, by performing immunostaining of γ-H2AX staining. The cells transfected with mCherryexpressing vector either with mock-, CtIP-, or MS2-CtIP-vector were administrated with 1 μM etoposide (DSB inducer), and then stained with anti-γ-H2AX antibody ( Supplementary Fig. 9b). Imaging analysis revealed that etoposide-induced DSB repair signal (i.e., the percentage of γ-H2AX-positive cells) in MS2-CtIP-transfected cells was greater than that in mock-transfected cells ( Supplementary Fig. 9c), suggesting that the overexpression of MS2-CtIP resulted in the activation of DSB repair pathway. Collectively, we observed no cytotoxic effect of MS2-CtIP; rather, it activated the DSB repair pathway.
Characterization of DSB repair pathway choice. Next, we analyzed the effect of CtIP overexpression and MS2-CtIP LoADing in terms of the accuracy and repair pathway choice. The short 3× Flag tag was knocked-in in this assay to enable sequence analysis covering the full length (left microhomology, knock-in fragment, and right microhomology) by next-generation sequencing (NGS) (Fig. 2a). For the unbiased analysis of all analyzable alleles, including precise knock-in, imprecise knock-in, and non-knock-in ones, we performed out-out PCR of genomic integration sites at the ATP5B, PARP1, and RPL11 loci, followed by NGS analysis. Consistent with the FACS analysis, the highest frequency of precise knock-in was observed in MS2-CtIP LoADed samples, followed by CtIP-overexpressing and mock ones, at all three loci (Fig. 2b). In contrast, the frequency of the alleles with insertion and/or deletion (indel) was highest in mock samples, followed by CtIP and MS2-CtIP ones. These results suggested that the LoADed MS2-CtIP selectively enhanced the MMEJ pathway while preventing the errorprone NHEJ. To further characterize the mechanisms of donor integration, we sorted the read counts by the length of inserted fragments. The results indicated that the frequency of MMEJmediated integration was much higher than that of NHEJ-mediated integration, and this tendency was particularly noticeable in MS2-CtIP LoADed samples, followed by CtIP-overexpressing and mock samples (Fig. 2c, Supplementary Fig. 11). Similarly, the frequency of precise knock-in mediated by MMEJ was much higher than that of precise knock-in mediated by NHEJ, and the effect of MS2-CtIP LoADing was also confirmed (Fig. 2d, Supplementary Fig. 12). Our finding that suggested superior efficiency of MMEJ compared with NHEJ is somewhat contrary to the previous observation 5 ; however, our data clearly showed that the appropriately structured PITCh donor vector could be incorporated via MMEJ with much higher frequency than via NHEJ. We also analyzed the frequency of base substitutions in and around microhomologous regions utilized for MMEJ ( Supplementary Fig. 13). The base substitution properties were similar among the samples of mock, CtIP, and MS2-CtIP; thus, MS2-CtIP LoADing did not decrease the knock-in accuracy. Interestingly, different frequencies and positional patterns of substitutions were observed at each microhomologous sequence. This might have been affected by base composition or other factors such as CpG methylation status and chromatin accessibility. Further data and analysis are needed to reveal the mechanism involved in this. Detailed NGS data are presented in Supplementary Data 2 and 3.  three kinds of PITCh donor fragments carrying mNeonGreen, mKate2 18 , and TagBFP 19 , targeting three different gene loci (Fig. 3a, Supplementary Fig. 14a). We also constructed an all-inone CRISPR-Cas9 vector based on the previously reported system 20 ( Supplementary Fig. 14b). At 72 h after transfection of the integrated donor and all-in-one CRISPR vector with or without MS2-CtIP LoADing vector, we observed a highly heterologous pattern of three kinds of fluorescence without any drug or fluorescence selection (Fig. 3b). The enhancer effect of MS2-CtIP was also observed (Supplementary Fig. 15). Convinced by this observation, we constructed and transfected six types of triple knock-in donors and CRISPR vectors along with the MS2-CtIP vector, and the successful generation of triple knock-in cells in a selection-free manner was suggested by fluorescence observation in all cases (Fig. 3c). To investigate whether the simultaneous generation of various types of knock-in cell clones can be achieved with this strategy, we performed selection-free singlecell cloning of the cells transfected with two types of vector sets for the triple knock-in, resulting in the production of CANX-mNeonGreen, PARP1-mKate2, and ATP5B-TagBFP, or FBL-mNeonGreen, PARP1-mKate2, and ATP5B-TagBFP. Following single transfection and single 96-well cloning for each type of knock-in, various patterns of single, double, and triple fluorescence were observed, suggesting that they were a variety of single, double, and triple knock-in cell clones (Fig. 3d). Among 75 and 93 clones isolated, the numbers and percentages of single, double, and triple knock-in were 13 (17.3%) and 8 (8.6%), 4 (5.3%) and 8 (8.6%), and 1 (1.3%) and 1 (1.1%), respectively (Fig. 3e). Sequencing analysis of 5′ and 3′ junctions of triple knock-in clones revealed that all 12 junctions were precisely joined with MMEJ, except that one junction contained a mutagenic allele along with the correct allele (Fig. 3f). In addition, analysis of various single and double knock-in clones confirmed that 26 out of 30 junctions had the precise knock-in sequence joined by MMEJ ( Supplementary Fig. 16). On the other hand, we observed no visible evidence of unexpected translocations, with the exception of faint bands detected by only one primer set (Supplementary Fig. 17). Thus, the robustness, high efficiency, and high accuracy of our strategy for selection-free, simultaneous generation of single, double, and triple knock-in clone collections were demonstrated.
Versatility of the LoAD system. To further expand the application of the LoAD system, we additionally examined the enhancement of other types of genome editing by accumulating various effector proteins. First, we focused on erroneous endjoining-mediated short deletion without a targeting donor (Fig. 4a). Since a previous study revealed that the overexpression of Trex2 resulted in higher mutagenic efficiency 21 , we observed the effects of Trex2 and MS2-Trex2, as well as Exo1, MS2-Exo1, CtIP, and MS2-CtIP, at the LMNB1 locus. Restriction fragment length polymorphism (RFLP) analysis 22 revealed that the LoADing of MS2-Trex2 resulted in higher efficiency of short deletions compared with mock or CtIP-overexpressing samples (Fig. 4b,  Supplementary Fig. 18a). A similar effect was also observed at the RPL11 locus (Fig. 4c, Supplementary Fig. 18b). Consistent with the results of the RFLP analyses, subcloned sequencing analysis revealed that the rates of short deletion among the total number of indel alleles were 25%, 66.7%, and 90.9% in mock, Trex2, and MS2-Trex2 samples, respectively (Fig. 4d, Supplementary  Fig. 18c). Next, we investigated whether the frequency of chromosomal deletions induced by double cutting 23 was affected by the LoADing of certain molecules (Fig. 4e, Supplementary  Fig. 19a). Consistent with the previous report 24 , we found that LoADed MS2-Trex2 strongly inhibited chromosomal deletions in two different sgRNA pairs (Fig. 4f, g, Supplementary Fig. 19c, d); whereas, the independent mutation efficiency was enhanced (Fig. 4h, Supplementary Fig. 19b,     single-strand template repair 25 (Fig. 4i, j, Supplementary Fig. 20) and HR repair pathways (Fig. 4k, l, Supplementary Fig. 21).

Discussion
In this study, we established a system of biased genome editing by recruiting specific repair molecules such as CtIP at the DSB site, and then demonstrated the simultaneous generation of multicolored cell collections, as well as showing other applications including the efficient independent mutagenesis with the inhibition of chromosomal deletions.
In particular, consistent with the recent report describing the knock-in-enhancing effect by Cas9 fused with CtIP (Cas9-CtIP) 14 , the enhancement of MMEJ-, SSTR-, and HR-mediated knock-in was achieved by LoADing MS2-CtIP. The enhancement of MMEJ-mediated knock-in by MS2-CtIP LoADing should become popular because the MMEJ-dependent strategy has been utilized in various applications including scarless disease modeling in human-induced pluripotent stem cells 26 , in vivo gene knock-in 27 , gene knock-in in neurons 28 , epitope tagging for chromatin immunoprecipitation (ChIP)-seq analysis 29 , and generation of knock-in mice containing exogenous gene cassette 9 , floxed allele 9 , or multiple SNPs spanning multiple exons 30 . In addition, a recent screening-based approach for the investigation of genetic interactions 31 revealed the need for a high-throughput strategy for generating cell clone libraries harboring various combinations of gene knock-in alleles. Our system will be an adequate technology    The knock-in rate was calculated as the percentages of knock-in clones among the total number of clones for such high-throughput cell engineering. On another front, however, potential risk of increasing off-target mutations by MS2-CtIP LoADing should be carefully examined in the future study, although, we showed the risk of off-target integration was minimum even when MS2-CtIP was LoADed.
In the analysis of CtIP mutants, we could not obtain any superior results compared with wild-type MS2-CtIP in HEK293T cells. The 3KR mutant, circumventing the requirement of SIRT6 17 , did not alter the knock-in-enhancing effect. The S327E/T847E mutations, which mimicked phosphorylation 15,16 , decreased the activity. However, these mutants might increase the activity in other cells with low SIRT6 expression and/or poor phosphorylation of CtIP. On the other hand, all the truncated HE mutants 14 with or without additional amino acid changes showed comparable knock-in efficiency to mock-transfected control, although Charpentier et al. showed that Cas9 fused with HE mutant (Cas9-HE) resulted in higher HDR efficiency compared to that of Cas9 14 . In addition, full-length MS2-CtIP harboring either phosphorylation-blocking (3A) or phosphorylationmimicked (3E) mutation positioned within the HE region did not show any change in knock-in efficiency. These findings suggest that HE region is not functional in our LoADed PITCh system.
Collectively, our results demonstrate the high-throughput generation of gene knock-in cell collections, which can contribute not only to the production of multiplexed reporter cells, but also to the simultaneous generation and evaluation of multiple polymorphisms spanning several gene loci for analyzing phenotypic variations or disease severity caused by multifactorial inheritance. Although, the utility of our LoAD system was demonstrated, the examples of its application presented in this paper were based on the screening of only a small number of DSB repair molecules. Further possibilities of the LoAD system will be revealed by large-scale effector screening, contributing to much more biased, controllable, and precise genome engineering.

Methods
Plasmid construction and oligonucleotide preparation. The all-in-one CRISPR-Cas9 vectors expressing Cas9 nuclease and single or multiple sgRNAs were constructed using the Multiplex CRISPR/Cas9 Assembly System Kit (Kit #1000000055, Addgene) 20 with some modifications. Briefly, oligonucleotides for sgRNA templates were synthesized, annealed, and inserted into the modified pX330A and pX330S vectors. Golden Gate assembly was used to assemble the constructed vectors into the all-in-one CRISPR/Cas9 vectors. For the addition of MS2 stem loops to sgRNA scaffold, the sgRNA expression cassette of the pX330A and pX330S vectors was replaced with that of the sgRNA(MS2) cloning backbone vector (Plasmid #61424, Addgene). The oligonucleotides used for the sgRNA template are listed in Supplementary Data 1.
The effector vectors without MS2 coat protein were constructed using an RT-PCR and In-Fusion cloning method (Takara) 9 . For the CtIP cloning, the CtIPvector-F/R and CtIP-insert-F/R primers listed in Supplementary Data 1 were used. For the effector vectors with MS2 coat protein, we first constructed MS2-Exo1 vector by the addition of MS2-NLS obtained from the MS2-P65-HSF1_GFP vector (Plasmid #61423, Addgene) to the Exo1 overexpression vector by the In-Fusion cloning method using the primers listed in Supplementary Data 1. Then, the coding sequence of Exo1 was replaced with other effector cDNAs using the primers listed in Supplementary Data 1.
The PITCh donor vectors carrying mNeonGreen were constructed by PCR and TA-cloning with TArget Clone™ -Plus-(Toyobo) using the primers listed in Supplementary Data 1, except the donor for the FBL locus. For the construction of the FBL-mNeonGreen donor vector, the T2A-Puro R sequence was removed from the PITCh(gRNA-s1)-FBL 9 . The PITCh donor vectors carrying 3× Flag tag, donor vectors for triple knock-in, and the HR donor vector were constructed by PCR and In-Fusion cloning (Takara) using the primers listed in Supplementary Data 1. The PITCh donor vectors for triple knock-in were constructed by PCR and In-Fusion cloning using the primers listed in Supplementary Data 1, as illustrated in Supplementary Fig. 14a.
The oligonucleotide for SSTR-mediated knock-in was purchased from IDT. The sequence is shown in Supplementary Data 1.
The plasmid sequences of CtIP vector, MS2-CtIP vector, and the PITCh donor vectors for Fig. 2d-f are shown in Supplementary Data 4.
Cell culture and transfection. HEK293T and HeLa cells, obtained from ATCC, were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (FBS). CHO-K1 cells, obtained from Riken BRC Cell Bank, were maintained in Ham's F12 medium supplemented with 10% FBS. All cell lines were tested negatively for mycoplasma contamination using e-Myco Mycoplasma PCR Detection Kit (iNtRON Biotechnology). Human cells (HEK293T and HeLa cells) were authenticated by short tandem repeat analysis (Takara).
Lipofectamine LTX (Life Technologies) and serum-free DMEM were used to transfect the plasmids in accordance with the supplier's protocol, except the SSTRmediated knock-in. For the SSTR-mediated knock-in, the 4D-Nucleofector system was used with the solution SF under the program CM-130, in accordance with the supplier's protocol. The amounts of plasmids and oligonucleotides, cell numbers, and plates used were as follows: for the FACS analysis and fluorescence observation of mNeonGreen knock-in HEK293T cells at the CANX locus, 200 ng of plasmids in total (PITCh donor vector, CRISPR vector, and mock/overexpression/LoADing vector) into 4.5 × 10 4 cells using a 96-well plate; for the FACS analysis and fluorescence observation of mNeonGreen knock-in HEK293T cells at the ATP5B, PARP1, FBL, and RPL11 loci, 200 ng of plasmids in total (PITCh donor vector, CRISPR vector, and mock/CtIP/MS2-CtIP vector) into 7.5 × 10 4 cells using a 96well plate; for the mNeonGreen knock-in at the ATP5B locus, followed by singlecell cloning, genomic out-out PCR, and Southern blotting, 200 ng of plasmids in total (PITCh donor vector, CRISPR vector, and mock/MS2-CtIP vector) into 5.0 × 10 4 cells using a 96-well plate; for the cytotoxicity analysis, 1, 3. filtered with a Flowmi Tip Strainer (Bel-Art Products). The number of cells with green and red fluorescence was counted using a BD FACSCalibur 4A (BD Biosciences) with a 488-nm laser and the corresponding fluorescence filters (FL1 for green, FL2 for red). After the preliminary FSC/SSC gating, up to 10,000 cells were analyzed for each sample. The data were analyzed using BD CellQuest Pro ver. 5.2 (BD Biosciences). For the cytotoxicity analysis, cell survival rates were determined as the percentages of the number of mCherry-positive cells in total cell counts 9,33 .
Single-cell cloning. For the triple knock-in and HR knock-in, cells were collected at 72 h post transfection, suspended in PBS, and filtered with a Flowmi Tip Strainer. Single cells were sorted using a PERFLOW® Sort (Furukawa Electric) without fluorescence filters, collected in 96-well plates containing DMEM medium supplemented with 10% FBS, and cultured for 1-3 weeks until the colonies formed. For the mNeonGreen knock-in, cells were collected at 72 h post transfection, adjusted to 1.5 cells per 100 µl by adding DMEM, and moved 100 µl of adjusted cells to each well of a 96-well plate. Genomic PCR and direct sequencing of knock-in junctions. Genomic DNA was extracted using a DNeasy Blood and Tissue kit (Qiagen) from the knock-in cells collected at 72 h post transfection. Genomic PCR was performed using PrimeSTAR Max DNA Polymerase (Takara) or KOD FX Neo (Toyobo) with the primers listed in Supplementary Data 1. For the allele number analysis, the PCR products were run on a 2% (wt/vol) agarose gel. Direct sequencing analysis of knock-in junctions was performed by FASMAC or using an ABI 3130xl Genetic Analyzer (Life Technologies) with a BigDye Terminator v3.1 Cycle Sequencing Kit (Life Technologies). The sequence data were collected using the software 3130 Series Data Collection Software 4 (Life Technologies).
Southern blot analysis. A total of 5-μg aliquots of genomic DNA were digested with EcoRV, PvuII, and SacI, and resolved on a 0.7% agarose gel. 32 P-labeled DNA probes were made by PCR using Ex Taq polymerase (Takara) and [alpha-32 P] dATP (Perkin Elmer) with primers listed in Supplementary Data 1. Membrane transfer with Hybond-N+ (GE Healthcare), ultraviolet cross-linking (120 mJ cm −2 ), pre-hybridization and hybridization were performed according to the instructions for ExpressHyb Hybridization Solution (Takara). The radio-isotope signal was detected using a bioimaging analyzer BAS-2500 (Fujifilm).
NGS and data analysis. Genomic DNA was extracted using a DNeasy Blood and Tissue kit from the knock-in cells collected at 72 h post transfection. Preparation of libraries and NGS were performed by FASMAC. Briefly, NGS libraries were prepared via two-step PCR using the primers listed in Supplementary Data 1. The PCR amplification was performed using PrimeSTAR GXL DNA Polymerase (TakaRa) and the PCR products were purified using Agencourt AMPure XP (Beckman Coulter). Note that the PCR products amplified using PrimeSTAR Max DNA Polymerase were used as templates in the first-step PCR of PARP1 samples instead of genomic DNA because genomic PCR amplification failed in the PARP1 samples. NGS was performed using a MiSeq (Illumina). The PCR products were also analyzed by microchip electrophoresis using MultiNA ™ (hereinafter referred to as "MultiNA") (Shimadzu) with the software MultiNA ver. 1.11 (Shimadzu).
The NGS data were analyzed using CRISPResso ver. 1.0.0 34 (Fig. 2b) or Cas-Analyzer 35 (Fig. 2c, d, Supplementary Fig. 11-13). The parameters used for CRISPResso analysis are described below. The parameters used for Cas-Analyzer were as follows: comparison range, 70; minimum frequency, 0; and WT marker, 5. The exported results of CRISPResso and Cas-Analyzer are described in Supplementary Data 2  Genotyping of the cells other than PITCh and HR samples. For the genotyping of the cells other than PITCh and HR knock-in samples, the cells were collected at 72 h post transfection. The templates for PCR amplification were prepared using the Cell Lysis Buffer in a GeneArt Genomic Cleavage Detection Kit (Life Technologies), in accordance with the manufacturer's instructions. The PCR amplification was performed using the primers listed in Supplementary Data 1. The pseudo-gel images were obtained and quantification of each amplicon was performed using MultiNA with the software MultiNA ver. 1.11. The RFLP analyses were performed using AluI, HpyAV, and AflII at the LMNB1, RPL11, and HPRT1 loci, respectively. The genomic cleavage detection assay was performed using a GeneArt Genomic Cleavage Detection Kit, in accordance with the manufacturer's instructions. Bacterial cloning of the PCR products was performed using TArget Clone™ -Plus-. Sanger sequencing was performed by FASMAC or using an ABI 3130xl Genetic Analyzer with a BigDye Terminator v3.1 Cycle Sequencing Kit. The sequence data were collected using the software 3130 Series Data Collection Software 4.
Statistical analysis. Statistical analysis was performed using the R software ver. 3.4.1. All data are presented as the mean ± s.e.m. All measurements were taken from distinct samples. Statistical significance was determined by Student's t-test.
Data availability. The authors declare that all data supporting the findings of this study are available within the paper and its Supplementary Information files. The plasmid sequences shown in Supplementary Data 4 were also deposited in NCBI via DDBJ. The accession codes are LC307119-LC307122. The NGS data were submitted to NCBI via DDBJ and are accessible with BioProject ID: PRJDB7211.