Timed inhibition of CDC7 increases CRISPR-Cas9 mediated templated repair

Repair of double strand DNA breaks (DSBs) can result in gene disruption or gene modification via homology directed repair (HDR) from donor DNA. Altering cellular responses to DSBs may rebalance editing outcomes towards HDR and away from other repair outcomes. Here, we utilize a pooled CRISPR screen to define host cell involvement in HDR between a Cas9 DSB and a plasmid double stranded donor DNA (dsDonor). We find that the Fanconi Anemia (FA) pathway is required for dsDonor HDR and that other genes act to repress HDR. Small molecule inhibition of one of these repressors, CDC7, by XL413 and other inhibitors increases the efficiency of HDR by up to 3.5 fold in many contexts, including primary T cells. XL413 stimulates HDR during a reversible slowing of S-phase that is unexplored for Cas9-induced HDR. We anticipate that XL413 and other such rationally developed inhibitors will be useful tools for gene modification. Altering cellular responses to double-strand breaks in DNA could rebalance CRISPRediting outcomes. Here, the authors use a pooled CRISPR screen to identify inhibition of CDC7 as a strategy to improve HDR outcomes.

G enome editing with targeted nucleases, such as CRISPR-Cas9, is a powerful tool for research and a promising approach for therapeutic treatment of human disease. One strategy for efficient genome editing in eukaryotic cells introduces a ribonucleotide protein (RNP) complex comprised of the type II endonuclease Cas9 and a guide RNA (gRNA), which create a double strand DNA break (DSB) at a targeted location in the genome 1,2 . This DSB is repaired by cellular DNA repair pathways to produce two outcomes: error-prone sequence disruption by insertion or deletion (indels) at the DSB, or precise sequence modification via homology directed repair (HDR) that copies donor DNA sequences into the DSB. The targeted incorporation of exogenous DNA sequences enables multiple new research avenues in metazoan cells, including endogenous epitope tagging and the insertion of SNPs to test disease causation, and using these techniques in human tissues could enable therapeutics to correct genetic lesions that drive human disease 3,4 . Strategies to favor precise HDR outcomes over deleterious error-prone repair in human cells are therefore of intense interest both to improve understanding of biological pathways and enable new therapeutic options.
Human cells have multiple overlapping DSB repair pathways that have been implicated in Cas9-mediated gene modification [5][6][7] such as alternative-end joining, synthesis-dependent strand annealing, and HDR, which encompasses multiple different mechanisms of repair. To investigate the various HDR mechanisms in greater detail, we previously developed a reporter 8 that allowed us to interrogate the genetic requirements of Cas9mediated HDR using single-stranded donor DNA (ssDonor) and discovered that single strand template repair (SSTR) requires the Fanconi Anemia (FA) DNA repair pathway 9 . We furthermore found that SSTR does not depend on the classic RAD51 pathway, unlike homologous recombination (HR) repair from a double stranded DNA donor (dsDonor). These distinct requirements for HDR from ssDonor and dsDonor implied that different donors produce molecularly identical gene modifications via different mechanistic routes. To more completely map how different types of donors mediate Cas9-induced HDR, we used genetic screening to reveal the DNA repair factors that are involved in HR using a dsDonor. Here, we describe genes that upregulate or downregulate HR from dsDonor templates, finding pathways that are both shared with and distinct from SSTR. We furthermore investigate factors whose knockdown increases Cas9-induced HDR, discovering that timed administration of a small molecule inhibitor of one of these factors, cycle 7-related protein kinase (CDC7), increases HR and SSTR by up to 3.5-fold in multiple contexts including primary human T cells.

Results
Cas9-induced SSTR and HR overlap, with key distinctions. We adapted a previously described pooled screening platform 9 to define the individual contribution made by each of thousands of genes to Cas9-mediated gene replacement using a dsDonor plasmid template. The basis of this platform is the stable expression of three components in each cell: (1) a dCas9-KRAB CRISPRi construct 10 , (2) a BFP reporter gene 8 , and (3) a gRNA targeting the transcription start site (TSS) of a single gene. We constructed a gRNA library to target genes with Gene Ontology (GO) terms related to DNA, comprising a library of approximately 2000 genes at a density of five guides per TSS 11 . Pooled K562 erythroleukemia cell populations stably expressing BFP and individually downregulating a specific gene were transiently nucleofected with a Cas9 RNP targeted to introduce a DSB in the BFP reporter, together with a dsDonor plasmid with a GFP sequence template that converts BFP to GFP upon successful HR 8 (Fig. 1a). Edited cell populations were separated by fluorescenceactivated cell sorting (FACS) (HR: BFP − GFP + ; gene disruption: BFP − GFP − ) ( Supplementary Fig. 1a), and gRNA frequency in each population was determined by sequencing the stably integrated gRNA cassette. Genes whose upregulation and downregulation altered each repair outcome were determined by comparing the sorted populations to the edited but unsorted cell population. Similarities between the reagents and techniques used in this screening approach permitted direct comparison with our earlier screen editing the same locus but utilizing a ssDonor 9 (Fig. 1a).
Guide RNAs targeting genes that restrict HR were enriched in the BFP − GFP + population (because their knockdown favors HR), while gRNAs targeting genes that are required for HR were depleted in the BFP − GFP + population (Fig. 1b). As expected, classic HR factors such as BRCA1, BARD1, RAD50, and NBN were genetically required for Cas9-mediated repair from a plasmid dsDonor, and these often were found in the same physical complexes with one another in the STRING protein database 12 (Fig. 1c). Our prior work defined the Fanconi Anemia (FA) pathway as necessary for Cas9-mediated SSTR 9 , and we now report that almost the entire FA repair pathway is also required for Cas9-induced HR. 31 of 40 FA and FA-related genes were required for HR, suggesting that this is an activity of the overall FA pathway ( Supplementary Fig. 1b) and indicating its importance for all forms of Cas9-mediated HDR 13 .
The shared reliance of Cas9-induced SSTR and HR on the FA pathway motivated us to systematically explore the overlapping genetic dependencies of HDR from single and double stranded templates. We performed GO term analysis 14 on statistically significant (p < 0.05) genes involved in SSTR and HR to define pathways involved in each process (Supplementary Fig. 1c and Supplementary Data 1 GO Terms). There was substantial overlap between SSTR and HR: both pathways require Fanconi Anemia Repair, nucleotide excision repair (NER), and strand displacement activities, which is driven by mutual reliance on FA proteins, members of the TFIIH complex, and the BLM helicase. Both SSTR and HR may therefore challenge cells to balance NERlike single strand editing activities with templated repair events, as has been suggested for repair of interstrand crosslinks 15 . Despite these similarities, SSTR and HR are distinct in notable ways. SSTR but not HR is associated with Negative regulation of transposition, a GO term including our SSTR hits APOBEC3C, D, F, and G. Originally reported as RNA editing enzymes, these enzymes are known to modify ssDNA during gene editing reactions 16 , and similar proteins have been repurposed as targeted DNA base-editing reagents 17 . HR on the other hand depends on genes involved in Double Strand Break Repair via Homologous Recombination, which includes factors that potentiate meiotic homologous recombination ( Supplementary Fig. 1c). These observations suggest that all HDR requires the FA pathway, but the HR and SSTR mechanisms each require specialized activities to respond to donor topologies or intermediate structures specific to each repair process.
We identified several genes that normally repress HR, such that their knockdown enhances HR efficiency (Fig. 1b). Some of these genes are consistent with a model in which NHEJ and HR compete to repair DSBs, so inhibition of one pathway favors the other [18][19][20] . Examples of these repressor genes include TP53binding protein 1 (53BP1), X-ray repair cross-complementing protein 4 (XRCC4) and non-homologous end joining factor 1 (NHEJ1), which interact at DSBs to promote DNA ligase 4 (LIG4) association during non-homologous end-joining (NHEJ) 21 . Other repressors identified here have not previously been reported to increase HR efficiency, but have roles in processes that have been linked to DNA repair outcomes such as cell cycle progression.  These proteins include CDC7, polo-like kinase 1 (PLK1), E2F transcription factor 1 (E2F1), and polo-like kinase 3 (PLK3).
Inhibiting HR repressors increases both HR and SSTR. HR is quite inefficient in human cells yet desirable for its ability to precisely engineer genomic sequences and even insert long (>500 bp) sequences such as epitope tags or ORF replacements 22 . Our exploration of genes that regulate HR presented us with a number of candidate repressors whose knockdown would increase HR efficiency. However, gene editing is frequently performed in primary cell types or other experimental contexts where transcriptional or genetic repression is unsuitable. Thus, small molecule treatments that increase HR would be more desirable because they can be administered in a variety of settings and easily withdrawn. We selected genes that repress SSTR, HR, or both and performed an extensive literature search to find small molecule inhibitors of these HDR repressors (Fig. 1b). We focused on eight commercially available small molecules that are reported to inhibit CCND1, CDC7, HIPK2, MAPK14, NOX4, PLK3, PLK1, and 53BP1 ( Supplementary Fig. 1d).
We first asked if chemical inhibition of these HR repressors affects HR or SSTR editing outcomes using our K562 cell line stably expressing the BFP-to-GFP HDR reporter system 8 . Reporter cells were nucleofected with a Cas9 RNP targeting the BFP reporter gene and either a ssDonor or plasmid dsDonor. We reasoned that small molecule inhibition of HR repressors would be most effective during gene editing (e.g., post-treatment), so we treated cells with different inhibitors for 24 h and then recovered in inhibitor-free media (Fig. 2a). BFP-to-GFP HDR outcomes were monitored by flow cytometry after four days (Supplementary Fig. 2a). Many compounds resulted in no change or even a reduction of HR, which could be caused by impaired cell fitness. Inhibition of mitogen-activated protein kinase 14 (MAPK14) with SB220025 slightly enhanced SSTR (1.1-fold), and inhibition of PLK3 with GW843682X slightly increased both SSTR and HR from the plasmid dsDonor (1.1-fold and 1.2-fold).
By contrast, treatment with the CDC7 inhibitor XL413 23 resulted in a significant increase in both SSTR and HR (1.4-fold and 1.8-fold, p < 0.0001) (Fig. 2b). siRNA inhibition of CDC7 was less effective at promoting HR (1.4-fold) than small molecule inhibition ( Supplementary Fig. 2b, c), which suggests that inactivating CDC7 kinase activity may be more effective than reducing its overall levels. The effect of XL413 is concentration dependent, as both SSTR and HR increase in a dose-dependent manner, with 33 μM XL413 increasing HDR 1.8-fold to 2.1-fold (Fig. 2c). XL413 concentrations up to 10 μM and exposure for up to 24 h did not result in a notable decrease in viability in K562 cells ( Supplementary Fig. 2d, e), and we therefore used 10 μM XL413 in subsequent experiments unless noted otherwise. To optimize CDC7 inhibition, we explored if different timing of exposure to XL413 (e.g., pre-treatment) altered HDR rates. However, pre-exposure to XL413 and then release during editing resulted in reduced levels of HDR, while pre-exposure and postexposure did not significantly increase HDR over post-treatment alone (Fig. 2d, e). We therefore conclude that the optimal timing for HDR is to incubate edited cells immediately post-treatment in media containing CDC7 inhibitors.
CDC7 inhibition increases HDR at multiple endogenous loci. We next asked if XL413's ability to stimulate HDR is generally applicable to multiple genomic loci and HDR cargo sizes. We used Cas9-induced HR from a plasmid dsDonor template to knock-in a GFP coding sequence at the C-terminus of various genes in K562 cells using editing reagents previously developed as part of a comprehensive cell-tagging effort 24 : Lysosomal-associated membrane protein 1 (LAMP1), fibrillarin (FBL), histone H2BJ (HIST1H2BJ), nucleophosmin (NPM1), Structural maintenance of chromosomes protein 1 A (SMC1A), RNA binding protein FUS (FUS), and translocase of outer mitochondrial membrane 20 (TOMM20). We found that treatment with XL413 for 24 h immediately after nucleofection increased the HR efficiency at all loci, ranging from 1.6-fold to 3.5-fold irrespective of the original frequency of HR ( Fig. 3a and Supplementary Fig. 3a). Importantly, the frequency of HR remained unchanged over two weeks of cell culture, indicating that edits produced during XL413 treatment are stable over long time periods.
We then investigated if SSTR is similarly increased by CDC7 inhibition at endogenous loci. We designed two ssDonor editing strategies: insertions of a 2 × FLAG-tag at the C-terminus of TOMM20 ( Supplementary Fig. 3b), and SNP modifications at five different genomic loci. Using amplicon PCR and next-generation sequencing, we found that XL413-treated K562 cells had up to a 2.5-fold increase in SSTR-based FLAG tagging and introduction of SNPs relative to untreated cells (Fig. 3b, c and Supplementary Data 2). Genomic increases in tagging corresponded with increased ability to detect FLAG-tagged TOMM20 by Western blotting ( Supplementary Fig. 3c, d). These findings suggest that CDC7 inhibition robustly increases SSTR and HR, and can be used to increase the frequency of both single nucleotide substitutions and larger endogenous gene tagging.
CDC7 inhibition enhances HDR in primary cells. HDR in primary cells is a long-standing goal of gene editing, both for its ability to correct disease-causing SNPs and to deliver large payloads such as chimeric antigen receptors 22,25 . We therefore investigated the ability of XL413 to increase HR in human T cells derived from peripheral blood mononuclear cells (PBMCs) of healthy donors. We performed gene editing using RNPs targeting the RAB11A, TUBA1B, and CLTA loci, together with linear dsDonor templates for each locus to generate fluorescently-tagged protein fusions (GFP, mCherry, or BFP) 22 . XL413 treatment after editing produced a dose-dependent increase in HR efficiency at each locus in T cells (Fig. 4a), without evidence of decreased viability ( Supplementary Fig. 4a). Using a ssDonor template and targeting five additional genomic loci, we found that XL413 also boosts SSTR in human T cells while NHEJ frequencies decreased ( Fig. 4b and Supplementary Data 2). This shows that CDC7 inhibition does not affect the total amount of editing but rather shifts the ratio of edited alleles from NHEJ to HDR. Furthermore, treatment with XL413 increased SSTR beyond 60% efficiency when introducing a naturally occurring SNP that alters recognition of CD4 by the OKT4 antibody without affecting recognition by the SK3 antibody 26 (Fig. 4c and Supplementary Fig. 4b-d).
Reversal of pathogenic mutations in patient-derived stem cells is a promising therapeutic application of HDR 27,28 . We therefore tested XL413 in primary human hematopoietic stem and progenitor cells (HSPCs). We found that XL413 addition during SSTR editing in HSPCs slightly elevated SNP conversion at two different loci, including the causative allele of sickle cell disease (Fig. 4d, e and Supplementary Data 2). Short-term HSPC viability was not grossly decreased by treatment with XL413 (Supplementary Fig. 4e). Taken together, we find that XL413 can increase multiple forms of HDR in two primary cell types. Additional work will be needed to determine the long-term effects of XL413 treatment to increase the edited fraction of cells if deployed in a pre-clinical context.  **** **** **** **** **** ****   Supplementary Fig. 3a. b XL413 increases SSTR at endogenous loci. K562 cells were nucleofected with RNP targeting TOMM20 and an ssDonor encoding 2xFLAG ( Supplementary Fig. 3b) in the presence or absence of 10 μM XL413 for 24 h, gDNA was extracted after 4 days, and SSTR frequencies were determined by amplicon sequencing. c XL413 increases the frequency of SNP conversion. RNPs targeting five loci and ssDonors encoding SNPs were introduced into cells and editing outcomes quantified as described in b. All values are shown as mean ± SD (n = 3 biological replicates). Statistical significances were calculated by unpaired two-tailed t-test using the Holm-Sidak method (p-values are reported as *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001) Source data are available in the Source Data file.
frequency of multi-allelic editing on a per-cell basis. We performed GFP-tagging in K562 cells at multiple loci as previously described with or without XL413, fluorescently sorted single GFPexpressing clones, and quantified knock-in alleles in each clone by junction PCR (Supplementary Fig. 5a, b). We found that XL413 dramatically increased the frequency of homozygous HR even in triploid K562 cells from undetectable to approximately 17% of cells edited at the FUS and HIST1H2BJ loci. HDR at the SMC1A locus was increased 4.9-fold such that almost all clones tested were homozygous (Fig. 5a). Homozygous editing was also increased up to 3.2-fold in T cells from healthy donor PBMCs, as measured by a fluorescent reporter assay that mixes two linear Our T-cell homozygous HDR assay only measures integration of the two different fluorophores at a given locus (e.g., BFP + /mCherry + ), and so these values likely undercount the true frequency of biallelic editing by missing dual integration of the same fluorophore (e.g., BFP + /BFP + ). We therefore conclude that CDC7 inhibition increases the rate of homozygous HDR, which could be very useful for both research and therapeutic purposes.
CDC7 inhibition boosts HDR in multiple cell types. We profiled XL413 in additional cell types at multiple loci, and found that cells of different origins vary in the ability of XL413 to increase HDR ( Supplementary Fig. 6a). K562 cells, T cells, and HSPCs are responsive to XL413, boosting all tested forms of HDR up to 3.5-fold. Using XL413 in HEK293T, U-251, HeLa and induced pluripotent stem cells (iPSCs) increased HDR at some loci and with some donors, but not with others ( Supplementary  Fig. 6a, b and Supplementary Data 2). XL413 did not increase HDR in primary mouse glial cultures, primary human dermal lymph endothelial cells (HDLECs) and primary human dermal fibroblasts (HDFs). However, these cell types proliferate particularly slowly and CDC7 inhibition may not work in this context ( Supplementary Fig. 6a, c). Because XL413 is known to be nonfunctional in a cell-type specific manner 29 , we tested other CDC7 inhibitors for their ability to increase HDR in contexts where XL413 was inactive. The dual CDC7/CDK9 inhibitor PHA-767491 30 increased HR in K562 cells 2.3-fold and at two additional loci (1.5-fold and 1.7-fold) in HCT116 cells, which do not respond consistently to XL413 ( Supplementary Fig. 7a, b). However, we found that treatment with PHA-767491 is more toxic to cells than XL413 and that it results in a distinctly different cell cycle profile than XL413 ( Supplementary Fig. 7c-e). Overall, we find that transcriptional inhibition of CDC7 ( Supplementary  Fig. 2b, c) and two small molecule inhibitors of CDC7 can all increase levels of HDR during Cas9 genome editing.
XL413 is more effective than other chemical HDR enhancers.
To contextualize the effectiveness of CDC7 inhibition, we compared XL413 to other approaches previously reported to boost HR. SCR7 inhibits NHEJ to rebalance DNA repair outcomes 20,31 , RS-1 is a RAD51 agonist that activates recombination 32,33 , L755507 is a β-3 adrenergic receptor agonist that works through an unknown mechanism 34 , and i53 inhibits 53BP1 to reduce NHEJ and favor recombination 35 . In our editing workflow, SCR7, RS-1, and L755507 had little effect on HR (Fig. 6a, b). Combining SCR7 or RS-1 with XL413 for 24 h also did not increase HR beyond XL413 alone ( Supplementary Fig. 8a) i53 peptide treatment increased HR 1.5-fold when considering only the cells expressing the plasmid encoding for i53 (Fig. 6c), which is comparable to the 1.7-fold seen with XL413 in this experiment. However, when determining the efficiency of i53 to increase HR in the whole population of cells ( Supplementary Fig. 8b) it is less effective than XL413, and does not further increase editing in combination with XL413 ( Supplementary Fig. 8c).
XL413 treatment produces a transient cell cycle arrest. Exposure of cells to CDC7 inhibitors, like XL413, has additional effects on cells beyond increasing HDR. CDC7 is the catalytically active subunit of DBF4-dependent kinase (DDK), which phosphorylates and activates the MCM helicase to initiate the G1/S transition 36,37 . XL413 has been reported to rapidly inhibit phosphorylation of MCM2 and arrest cells in early S-phase 23,38 .
As timing of XL413 administration is important for its activity (Fig. 2d, e), we investigated the overlap between XL413 stimulation of HDR increase and cell cycle arrest. 24 h of XL413 treatment causes inhibition of MCM2 phosphorylation as measured by Western blot (Supplementary Fig. 9a) and cell cycle arrest phenotypes as measured using a FUCCI live cell reporter system 39 and Hoechst33342 staining (Supplementary Fig. 9b-g). Importantly, these effects are reversible and return to baseline 24 h after XL413 is removed. Post-editing treatment with XL413 for 24 h therefore synchronizes cells in S phase, and we speculate that this extends the time cells spend in the HDR-permissive S/G2/M phases of the cell cycle. Such a model would also explain why dosing with XL413 before editing does not increase HDR, as the cell cycle delay occurs before editing reagents are introduced. Interestingly, treatment with other S-phase inhibitors such as Aphidicolin or Hydroxyurea did not increase HR consistently ( Supplementary Fig. 9h) suggesting that cell cycle arrest with XL413 is distinct from arrest with other S phase inhibitors. The microtubule poison nocodazole has been previously described to boost HDR by placing cells at a permissive point in the cell cycle 7 , and we next compared nocodazole treatment to XL413. In our hands, nocodazole increased HDR to a comparable level as XL413 when cells were arrested with nocodazole prior to editing (Fig. 6d). However, nocodazole has also been reported to increase the frequency of aneuploidy and failed mitoses [40][41][42] . We measured the frequency of cells with >4 N DNA content produced by failed mitoses in compound-treated and untreated cell cultures. We observed no significant increase in >4 N DNA staining in untreated or XL413 treated cells after four days, but we Fig. 4 CDC7 inhibition increases HR and SSTR in primary human cells. a XL413 increases HR at endogenous loci in T cells. Primary CD3 + T cells from two healthy blood donors were nucleofected with RNPs targeting three different genomic loci (RAB11A, TUBA1B, and CLTA) with a linear dsDonor template to knock-in a fluorescent reporter protein fused to the N-terminal end of the target protein. Nucleofected cells were treated with the indicated doses of XL413 for 24 h, then the percentage of reporter-positive T cells was quantified by flow cytometry 4 days after nucleofection. Gating strategy is depicted in (Supplementary Fig. 4a). b XL413 increases SSTR at endogenous loci. Primary CD3 + T cells were nucleofected with RNPs targeting the indicated loci with ssDonors to insert point mutations, treated with the indicated concentration of XL413 for 24 h. After 4 days, gDNA was extracted and SSTR and NHEJ frequencies determined by amplicon sequencing. c XL413 increases the frequency of SNP conversion. RNPs targeting the CD4 locus and an ssDonor encoding a naturally occurring SNP ( Supplementary Fig. 4b-d) were introduced into T cells from three donors, and editing outcomes quantified by flow cytometry after 4 days. Statistical significances were calculated by ordinary one-way ANOVA and Dunnet's multiple comparison test (adjusted p-values are reported as *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). d XL413 increases the frequency of SNP conversion in the HBB gene in primary human CD34 + HSPCs. RNPs targeting the HBB locus and an ssDonor encoding the E6V mutation 25 were nucleofected into HSPCs in the presence or absence of 30 μM XL413, gDNA was extracted after 4 days, and mutation frequencies determined by amplicon sequencing. Statistical significances were calculated by unpaired two-tailed t-test using the Holm-Sidak method (p-values are reported as *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). e XL413 increases the frequency of SSTR in primary human CD34 + HSPCs. Cells were nucleofected with RNPs targeting TOMM20 with a ssDonor that contains three synonymous SNPs in the presence or absence of 30 μM XL413 for 24 h, gDNA was extracted after 4 days, and SSTR and NHEJ frequencies were determined by amplicon sequencing. Values are shown as mean ± SD of the indicated number of samples. Source data are available in the Source Data file.
found a significant increase in these events with nocodazole treated samples (p < 0.001) (Fig. 6e, f).
In summary, we mined our HR and SSTR genetic dependency datasets to identify CDC7 inhibition as a potent strategy to boost HDR in human cell lines. XL413, a CDC7 inhibitor, improves the efficiency of gene editing workflows for basic research or therapeutic applications, increases gene replacement in multiple cell types, and has utility in challenging contexts with therapeutic potential such as primary T cells and HSPCs. CDC7 inhibition is most effective when administered while editing is underway and is thus maximally compatible with RNP-mediated gene editing workflows.

Discussion
Our focused CRISPRi screen defined a network of genes that contribute to Cas9-mediated HR from dsDonor templates. Comparing networks of genes involved in HDR from different donor templates (ssDonor vs. plasmid dsDonor) revealed that many DNA repair factors are shared between SSTR and HR. The most striking commonality is the shared dependence on the FA pathway for both forms of repair. Despite this shared reliance on the FA pathway, we also observed genetic differences between the different types of HDR. Overall, this suggests that the early steps of HDR may be similar for templated repair of a Cas9 break with ssDonor or dsDonor, but the downstream stability and incorporation of different donor templates requires different factors. Future work could expand this platform to genome-wide screens or focus more narrowly on the FA pathway to determine if different FA sub-complexes act specifically during SSTR or HR.
By focusing on genes that restrict HDR and can also be targeted by small molecule inhibitors, we identified CDC7 inhibition as a strategy to boost HDR in human cells. We find that the small molecule inhibitor, XL413, occupies the sweet spot among HDR adjuvants of high efficacy, ease of use, and low toxicity. However, XL413 is ineffective in some cell types 29 . Thus, XL413-resistant cell lines could instead be treated with PHA-767491, if toxicity is not a limiting factor. Future development may yield CDC7 inhibitors that combine the consistency of PHA-767491 and the low toxicity of XL413.
As CDC7 plays a key role in DNA replication initiation, this fits well into the intimate link between DNA repair and cell cycle control 43 . CDC7 inhibition by XL413 both increases HDR and delays cells in S phase. The most parsimonious explanation is that XL413 promotes HDR by increasing the amount of time that cells spend in the HDR-permissive S phase. However, this model is partially at odds with observed stimulation of HDR with PHA-767491 treatment, which arrests cells in G1 phase prohibiting cells to move into S-phase 38 (Supplementary Fig. 7e). This differential cell cycle profile between XL413 and PHA-767491 could potentially be explained by PHA-767491's promiscuous activity against other kinases such as Cdk9, which is required for S-phase entry 44,45 . Another possibility is that CDC7 directly regulates DNA repair, which would be consistent with reports that CDC7null yeast cells are sensitive to UV damage 46 . Mechanistic links between CDC7 and DNA repair remain unexplored, but might include direct phosphorylation of DNA repair proteins by CDC7. For example, CDC7 has been reported to phosphorylate the RAD18 ubiquitin ligase 47 , mutant alleles of which support increased HDR rates 48 . REV7 is also a target of CDC7 and forms a complex with shieldin to regulate DNA repair outcomes, and thus inhibition of CDC7 could have an indirect effect on DNA repair through REV7 49,50 . Another possible mode of interaction is scaffolding of DNA repair proteins by CDC7, which has been reported for RAD18 51 . Better understanding of the links between DNA replication control and DNA repair control may define these mechanisms and suggest more targeted interventions, for example preventing phosphorylation or degrading the key HDRregulating substrate of CDC7.
The general link between HDR stimulation and cell cycle illustrates that HDR-boosting treatments can be a balancing act between desired and undesired outcomes. It is tempting to wish for a treatment that increases HDR to 100% efficiency, no matter the mechanism. The microtubule poison nocodazole boosts HDR in some contexts via G2/M arrest 7 , but also increases aneuploidy and failed mitoses (Fig. 6d, e). Similarly, inappropriately activating DNA repair during mitosis leads to chromosome fusions 52 . In our hands, CDC7 inhibition reversibly slows S-phase and effectively promotes stable genetic changes without increasing the rate of genomic instability. However, we cannot rule out that  Fig. 5a, b). The total number of clones analyzed is noted above each column with the proportion of heterozygous, homozygous, or no knock-in (GFP + but no detection of GFP sequence at on-target site) genotype shown per condition. b XL413 increases biallelic editing in T cells. CD3 + T cells were nucleofected with RNPs targeting the RAB11A locus and two linear dsDonors encoding N-terminal fusions of either mCherry or BFP and treated with the indicated concentrations of XL413 for 24 h. Single-fluorescent or doublefluorescent reporter populations were quantified by flow cytometry (Supplementary Fig. 5c, d). Shown is the frequency of biallelic dual-positive integration (BFP + /mCherry + ) events displayed as mean ± SD (n = 2 biological replicates). Source data are available in the Source Data file. CDC7 inhibition has undesirable side-effects that remain to be discovered. Thus, CDC7 inhibition to increase HDR in the context of gene editing therapeutics warrants further investigation. Here we demonstrate that understanding the genetics of gene editing reactions can be used to design interventions that favor certain types of DNA repair. While genetic CRISPR screens are a powerful tool to uncover phenotypes directly related to the expression levels of targeted genes, they fail to determine phenotypic outcomes that are caused by post-translational modifications. We anticipate that further work to map fundamental DNA repair pathways including studying the effects of posttranslational changes of DNA repair proteins will suggest new strategies and targetable regulators to increase the precision and the efficacy of gene editing workflows.

Methods
Pooled screen. Replicate cultures of K562 cells stably expressing a dCas9-KRAB construct and a cassette containing a BFP reporter and a guide RNA targeting a library of DNA repair factors (Supplementary Data 1 GUIDES and previously described 9 ) were thawed, cultured, and puromycin treated. 10e+06 cells from each replicate were subcultured (UNZAP). 25e+06 cells from each replicate were harvested for nucleofection. Each nucleofection aliquot was spun down, washed in PBS, and resuspended in 825 ml of nucleofection buffer SF (Lonza, Basel, Switzerland). Two hundred and seventy-five microliter of RNP editing mixture was added and mixed by pipetting. RNP for each replicate contained 2000 pmoles of sgRNA, 1,650 pmoles of Cas9, and 220 mg of plasmid pCR1075 donor DNA in Cas9 buffer (20 nM HEPES [pH 7.5]), 150 mM KCl, 1 mM MgCl 2 , 10% glycerol, and 1 mM TCEP). Each replicate of the RNP cell slurry was split and nucleofected in parallel in a Lonza 96-well Shuttle nucleofector (code FF-120), re-pooled, and cultured for two (replicate 1) or three (replicate 2) days. Nucleofected replicates were sorted into GFP + (GFP) and non-fluorescent (NON) populations on a Sony SH800S sorter. In parallel, an 85e+06 cell aliquot was harvested from each nonelectroporated population the day of the sort (UNZAP), and an 85e+06 cell aliquot was harvested from each nucleofected cell library on the day of sorting (PRE-SORT). Harvested and sorted populations were spun down, rinsed in PBS, and frozen at −80°C.
DNA from each cell population: PREZAP, UNZAP, PRESORT, GFP, and NON (non-treated) was purified using Machery-Nagel Blood Purification kits and the total amount of DNA quantified. One microgram of genomic DNA was amplified per Phusion HiFi PCR reaction using primers specific to the gRNA cassette as described 45 . Up to 24 PCR reactions were set up for each cell population to obtain desired coverage of the cell library. The thermocycler was set for one cycle of 98°C for 30 s, 25 cycles of 98°C for 15 s, 56°C for 15 s, and 72°C for 15 s, and one cycle of 72°C for 10 min and held at 4°C. PCR reactions were pooled, purified using SPRI beads, and sized on an agarose gel. Amplified DNA from each cell population was normalized to input cell numbers, purified a second time using SPRI beads, and sequenced on a HiSeq2500 (Illumina).
Pooled screen analysis. Data analysis was performed as described 11 . Briefly, sequence reads were trimmed, aligned to DNA Repair Guide Sequences (Supplementary Data 1 GUIDES) and quantified. Read counts for each gRNA were normalized and compared to the distribution of untargeted control guides to determine significance and log2 magnitude of change. The top three guide-level phenotypes were collapsed to produce gene-level phenotype score. Results for the GFPvPRE comparison are available in (Supplementary Data 1).
Pooled screen GO-term comparison. Data from SSTR 9 and HR (this manuscript) screens was filtered for statistical significance (p > 0.05) and separated into two categories: genes involved in SSTR or genes involved in HR. The gene list for each category was compared to the starting guide pool used in our screens (Supplementary Data 1 GUIDES) using DAVID v6.8 14  Cell lines and culture. HEK293, HCT116, HeLa, U251, and K562 cells were acquired from the UC Berkeley Tissue Culture Facility. HEK293, HCT116, and HeLa cells were maintained in DMEM medium supplemented with 10% fetal bovine serum and Penicillin/Streptomycin. K562 cells were maintained in RPMI medium supplemented with 10% fetal bovine serum and Penicillin/Streptomycin. U251 cells were maintained in DMEM-F12 medium supplemented with 10% fetal bovine serum and Penicillin/Streptomycin. Cell lines were tested regularly for mycoplasma contamination using enzymatic (Lonza, Basel, Switzerland) and PCRbased assays (Bulldog Bio, Portsmouth, New Hampshire). Primary mouse glial cells were extracted as described 53 and maintained in DMEM medium supplemented with only 10% fetal bovine serum. Human dermal lymphatic endothelial cells (HDLECs) were maintained in EBM-2 MV media from Lonza. Human iPSCs were generated from dermal fibroblasts from a healthy male subject (WTc) using the episomal reprogramming method 54 . iPSCs were cultured in Stemfit (Ajinomoto) on matrigel coated plates at 37°C. After passaging, iPSCs were plated in Stemfit with Rock inhibitor Y-27632 (SelleckChem). Human dermal fibroblasts (HDFs) were derived from a healthy male subject and cultured in DMEM supplemented with 1% sodium pyruvate (Thermo Fisher), 10% fetal bovine serum and Penicillin/ Streptomycin. Derivation and use of human iPSCs and HDFs was approved by the UCSF Committee on Human Research, San Francisco, CA (IRB 10-02521). All subjects provided written informed consent prior to participation. Cryopreserved wildtype human mobilized peripheral blood CD34 + HSPCs from volunteer donors were purchased from Allcells, and were cultured in SFEMII + CC110 (StemCell Technologies). Fig. 6 Comparisons between XL413 and other small molecule HDR-boosting strategies in K562 cells. a Direct comparison between XL413 and RS1 or SCR7. Cells were nucleofected with RNP targeting the LAMP1 or FBL loci with dsDonors encoding C-terminal fusions to GFP, cultured in XL413 (33 μM), SCR7 (1 μM) or RS-1 (10 μM) for 24 h, and editing rates were measured 4 days post nucleofection by flow cytometry. b Comparison between XL413 and L755507. Cells were nucleofected with RNPs targeting the RAB11A, LAMP1, or TOMM20 loci with dsDonors encoding N-terminal (RAB11A) or C-terminal (LAMP1, TOMM20) fusions to GFP, cultured in XL413 (10 μM) or L755507 (5 μM) for 24 h, and editing rates were measured 4 days post nucleofection by flow cytometry. c Comparison between XL413 and i53. Cells were nucleofected with RNP targeting the RAB11A locus with dsDonor encoding an N-terminal fusion to GFP. i53-treated samples were lipofected with (3 μg) plasmid encoding i53 and BFP 24 h prior to nucleofection with editing reagents. XL413treated samples were cultured in 10 μM XL413 for 24 h after nucleofection, then editing outcomes were measured 4 days post-nucleofection by flow cytometry. i53-transfected cells were gated using a BFP fluorescent marker expressed from the i53 expression plasmid. d Direct comparison between XL413 and nocodazole. Cells were nucleofected with RNP targeting the RAB11A, LAMP1, TOMM20, or FBL loci with dsDonors encoding N-terminal (RAB11A) or C-terminal (LAMP1, TOMM20, and FBL) fusions to GFP. Samples were treated with nocodazole (33 ng mL −1 ) for 24 h before nucleofection. XL413treated samples were cultured in 10 μM XL413 for 24 h after nucleofection. Editing rates were measured 4 days post-nucleofection by flow cytometry. e 1Cells were treated with nocodazole or XL413 for 24 h and then released into normal media for four days prior to propidium iodide staining and flow cytometry. Example plots are shown for untreated, nocodazole, and XL413 treated samples. f Quantification of frequency of cells with a > 4 N DNA content (indicative of failed mitoses) from e. All values are shown as mean ± SD (n = 3 biological replicates). Statistical significances were calculated by ordinary one-way ANOVA and Dunnet's multiple comparison test (adjusted p-values are reported as *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). Source data are available in the Source Data file.
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15845-1 ARTICLE T Cell isolation and culture. Primary human T cells were isolated from two deidentified healthy human donors from residuals from leukoreduction chambers after Trima Apheresis (Vitalant), with written informed consent under a protocol approved by the UCSF IRB (BU101283). Peripheral blood mononuclear cells (PBMCs) were isolated by Ficoll centrifugation using SepMate tubes (STEMCELL, per manufacturer's instructions), then T cells were further isolated from PBMCs by magnetic negative selection using an EasySep Human T Cell Isolation Kit (STEMCELL, per manufacturer's instructions). Isolated T cells were cultured at 1 million cells mL −1 in XVivo15 medium (STEMCELL) with 5% fetal bovine serum, 50 μM 2-mercaptoethanol, and 10 mM N-Acetyl L-Cystine, and stimulated for 2 days prior to electroporation with anti-human CD3/CD28 magnetic dynabeads (ThermoFisher) at a beads to cells concentration of 1:1, along with a cytokine cocktail of IL-2 at 200 U mL −1 (UCSF Pharmacy), IL-7 at 5 ng mL −1 (Thermo-Fisher), and IL-15 at 5 ng mL −1 (Life Tech). T cells were harvested from their culture vessels and magnetic dynabeads were removed by placing cells on an EasySep cell separation magnet for 5 min. Immediately prior to electroporation, debeaded cells were centrifuged for 10 min at 90 g, media aspirated, and resuspended in the Lonza electroporation buffer P3 using 20 µL buffer per one million cells. Cas9 RNP assembly and nucleofection. Fifty pmoles of sgRNA was diluted using Cas9 buffer (20 nM HEPES [pH 7.5]), 150 mM KCl, 1 mM MgCl 2 , 10% glycerol, and 1 mM TCEP) or water. 1.25 μL of 40 mM Cas9-2xNLS (50 pmoles) was slowly mixed in, and the resulting mixture was incubated for 5 min at room temperature to allow for RNP formation. After incubation, either 0.5 μL of 100 μM ssDonor or 1.5 or 2 μg of plasmid DNA was introduced and mixed by pipetting. The total volume of RNP solution was 5 ml, where the volume of Cas9 buffer was adjusted to account for volume differences between ssDonor and plasmid DNA. Between 1e +05 and 2e+05 cells were harvested, washed once in PBS, and resuspended in 15 μL of nucleofection buffer (Lonza, Basel, Switzerland). Five microliter of RNP mixture was added to 15 μL of cell suspensions. Reaction mixtures were electroporated in Lonza 4D nucleocuvettes, incubated in the nucleocuvette at room temperature for five minutes, and transferred to culture dishes containing prewarmed media. Large-scale nucleofections were performed by splitting cultures and conducting multiple parallel nucleofections.
Editing outcomes were measured four days post-nucleofection by flow cytometry or by amplicon sequencing (see below). Resuspension buffer and electroporation conditions were the following from each cell line: K562 in buffer SF with FF-120, HEK293 in buffer SF with DS-150, T cells in buffer P3 with EH-115, primary mouse glial cells in buffer P3 with DS-112, HDLECs in buffer P3 with CA-137 HCT116s in buffer SE with EN-113, and HeLa cells in buffer SE with CN-114, U251 cells in buffer SE with DS130, iPSCs in buffer P3 with DS-138, HDFs in buffer P3 with DS-150 and HSPCs in buffer P3 with ER-100.
Unless otherwise indicated, cells were separated into three stocks, then nucleofected, recovered, and analyzed separately (biological triplicate).
Fifty picomolar of RNP and 0.25 μg of dsDonor, or 100 pmol of ssDonor, were mixed for 5-10 min, then added to cells 3-5 min before electroporation. One million T cells per well with RNP and donor template were electroporated using the Lonza 4D 96-well electroporation system with pulse code EH115, in biological replicate of n = 2-3. Immediately post-electroporation, prewarmed media was added to rescue the cells then each electroporation condition was split into 5 wells of a 96-well U-bottom tissue culture plate. Electroporated cells were incubated at 1 million cells mL −1 in final volume 200 μL media with IL-2 at 500 U mL −1 and increasing concentrations of XL413. After 24 h, all cells were washed in PBS, and fresh media was added containing only IL-2 at 500 U mL −1 . Approximately 3 days post-electroporation, cells were collected by centrifugation at 300 × g, media discarded, and antibody stains were added (UCHT1-CD3-PE, OKT4-CD4-PE-Cy7, RPA-T8-APC (all from BioLegend), and GhostDye780 (Tonbo)) for 20 min (Supplementary Data 3). To assess conversion to the CD4 SNP sequence, SK3-CD4-FITC (Biolegend) was also added to the antibody staining. Cells were washed, resuspended in PBS + 1% serum (120 μL per well), and then an equal volume (80 μL) of each well was sampled using an Attune NxT Focusing Flow Cytometer with Autosampler attachment (ThermoFisher).
Genomic DNA extraction (for amplicon sequencing). Approximately 1e+05 cells were harvested and resuspended in 50 μL of QuickExtract DNA extract solution (Lucigen). Reactions were incubated for 20 min at 65°C and 5 min at 95°C. Extractions were diluted 1:4 with dH2O, spun for 5 min at max speed in a microcentrifuge, and the supernatants retained for downstream analysis.
PCR amplification of edited regions. PCR reactions were generated from 2× Q5 master mix (NEB), primers (Supplementary Data 3) at 500 nM, and 5 μL of genomic DNA (see above). Unless otherwise noted, PCR primers have a 5′ sequence tag (GCTCTTCCGATCT) that allows re-amplification for Illumina sequencing (amplify-on PCR). The thermocycler was set for one cycle of 98°C for 1 min, 35 cycles of 98°C for 10 s, 63°C for 15 s, 72°C for 60 s, and one cycle of 72°C for 4 min, and held at 4°C. PCR amplicons were purified using SPRI beads, run on a 1.5% agarose gel to verify size and purity, and quantified by Qubit (Thermo Fisher, Waltham, MA).
NGS library generation and sequencing. Illumina adapters and index sequences were added to 100 ng of purified PCR amplicons by amplify-on PCR. Amplify-on was performed using 100 ng of template DNA, 0.5 μM of forward/reverse primers, and 2× Q5 Master Mix (NEB). The thermocycler was set for one cycle of 98°C for 30 s, 16 cycles of 98°C for 10 s, 62°C for 20 s, 72°C for 30 s, and one cycle of 72°C for 1 min, and held at 4°C. Each adapter-conjugated amplicon was quantified by qubit, normalized, and pooled at equimolar amounts. Pooled samples were purified using SPRI beads. Library size and purity was verified by Bioanalyzer trace prior to sequencing on an Illumina MiSeq using reagent kit v3 (2 × 300 bp).
NGS analysis of amplicons. Samples were deep sequenced on an Illumina MiSeq at 300 bp paired-end reads to a depth of at least 10,000 reads per sample. Cortado (https://github.com/staciawyman/cortado v1.0) was used to analyze editing outcomes. Briefly, reads were adapter trimmed then joined before performing a global alignment between reads and the reference and donor sequences using NEEDLE 56 . Rates of HDR are calculated as total number of reads that are successfully converted to the donor sequence (and have no insertions or deletions at the cut site) divided by the total number of aligned reads. NHEJ rates are calculated as any reads where an insertion or deletion overlaps the cut site or occurs within a six base pair window around the cut site divided by the total number of aligned reads. SSTR/HDR rates were calculated at specific regions by counting total number of reads with flag occurrence divided by the number of aligned reads. Amplicon sequencing results can be found in (Supplementary Data 2).
i53 treated cells were transfected using Lipofectamine 2000 from Invitrogen. Three microgram of each of three plasmids containing a broken i53 gene (Addgene #94939), an intact i53 gene (Addgene #74940), and both an intact i53 gene and a BFP reporter (Addgene #111145, respectively, were incubated in OPTI-MEM with Lipofectamine 2000 for 30 min. The cells were then incubated with the Lipofectamine plasmid solution for 24 h, pre-nucleofection. The BFP reporter was used to analyze the transfection efficiency of the i53 plasmid into the cells by flow cytometry. qPCR. Between 1e+05 and 2e+05 cells were harvested and RNA extracted using Qiagen RNeasy kits (Qiagen, Venlo, Netherlands). cDNA was produced from 1 mg of purified RNA using the iScript Reverse Transcription Supermix (Bio-Rad). qPCR reactions were performed using the SsoAdvanced Universal SYBR Green Supermix (Bio-Rad) in a total volume of 10 mL with primers at final concentrations of 500 nM. The thermocycler was set for 95°C for 2 min and 40 cycles of 95°C for 2 s and 55°C for 8 s. Fold enrichment of the assayed genes over the control ACT1B and/or GAPDH loci were calculated using the 2 −ΔΔCt method essentially as described 57 . Primer sequences can be found in Supplementary Data 3.
Biallelic editing assays. To check for frequencies of multi-allelic knock-ins in K562 cells, cells were nucleofected with RNPs targeting FUS, HIST1H2BJ or SMC1A and corresponding dsDonors, respectively, and were kept in complete media or media treated with 10 μM XL413 for 24 h. After 24 h the drug was removed and cells cultured for an additional 3 days. Four days of post nucleofection cells were analyzed for GFP fluorescence using an Aria II (BD Biosciences) cell sorter. GFP positive cells in both treated and untreated samples were sorted as single cells into 96-well culture plates containing complete media. Single-cell clones were grown for 10 days and DNA was extracted using a QuickExtract DNA extraction buffer (Lucigen). PCR across the C-terminal insertion site was performed using Q5 polymerase (New England Biolabs). Primers for PCRs can be found in Supplementary Data 3. PCR products were run on a 1% agarose gel and genotypes were determined. Clones classified as heterozygotes showed both the wild-type and knockin PCR band and clones classified as homozygotes only showed the knock-in PCR band (higher molecular weight). Clones that only showed the wildtype band are named no knock-in but could be considered as clones that show integration of the GFP sequence somewhere else in the genome (as they were in the GFP positive cell pool but do not show integration at the target site). In some clones we observed deletional bands. These clones were excluded in our analysis.
To quantify biallelic editing rates in T cells, cells were nucleofected with RNPs targeting the RAB11A locus as described above but with two different dsDonor constructs in a 1:1 ratio, one of which contains a BFP reporter sequence and one of which contains an mCherry reporter. Cells were treated with increasing amounts (0, 3.3, 10, and 33 μM) of XL413. After 4 days cells were analyzed for successful knock-in by flow cytometry and frequencies of double positive cell populations (BFP + and mCherry + ) were determined that represent the biallelic knock-in rates.