Genome editing allows scientists to change an organism’s genetic information by adding, replacing or removing sections of its DNA sequence. The CRISPR-Cas9 system is a genome-editing tool that has had a large impact on biological research in recent years, and also shows promise for the treatment of patients with genetic disorders.

The tool works as follows: a small piece of RNA (a close cousin to DNA) is used to guide an enzyme called the Cas9 endonuclease to the desired region of the genome. Then, like a pair of molecular scissors, the enzyme cuts the DNA, breaking both strands of its double helix. The cell naturally starts to repair the damaged DNA, and one way to do this is to use another similar piece of intact DNA as a template. Scientists can exploit this repair mechanism (known as homology-directed repair) by giving the cell extra DNA that carries their desired sequence change, with the hope that the cell will use it as a template and edit its own genome in precisely the same way. However, it turns out that mammalian cells rarely use the template DNA to repair the damage. Instead, mammals tend to fix double-stranded breaks in DNA by simply joining the broken ends together, a method that is prone to errors.

To overcome this specific issue, Savic, Ringnalda et al. tested the effect of physically linking the template DNA to the Cas9 enzyme, so that the DNA was already nearby when the enzyme made the cut. Experiments with human cells confirmed that this new approach increased the frequency of homology-directed repair up to 24-fold compared to leaving the enzyme and the template DNA separate. Improving the CRISPR-Cas9 system in this manner makes it more likely that genome editing may one day become a routine treatment for patients with genetic disorders. But first, more preclinical studies are needed to assess the safety of the CRISPR-Cas9 technology for gene editing in patients.

The CRISPR-Cas9 system is a versatile genome-editing tool that enables the introduction of site-specific genetic modifications (Jinek et al., 2012). In its most widespread variant a programmable chimeric single guide RNA (sgRNA) directs the Cas9 nuclease to the genomic region of interest, where it generates a site-specific DNA double-strand break (DSB) (Mali et al., 2013). In mammalian cells the repair of DSBs by different end-joining (EJ) pathways, such as classical non-homologous end joining (c-NHEJ), alternative non-homologous end-joining (a-NHEJ), or single-strand annealing (SSA) often leads to the formation of insertions or deletions (indels) (Shalem et al., 2014; Ceccaldi et al., 2016). Alternatively, when a donor template is provided, mammalian cells can also resolve DSBs via homology-directed repair (HDR) mechanisms, such as the classical homologous recombination (HR) pathway (Mao et al., 2008) and the Fanconi Anemia (FA) repair pathway (Richardson, 2017). While the formation of indels allows the elimination of gene function, repair from an ectopic donor oliogonucleotide (oligo) via HDR mechanisms enables the introduction of DNA modifications with single base pair precision (van den Bosch et al., 2002).

Therapeutic applications of CRISPR-Cas9 generally require the precise correction of pathogenic mutations using donor templates. However, DSBs introduced in mammalian cells are predominantly repaired by error-prone EJ pathways. As the resulting indels inhibit the CRISPR-Cas9 complex from retargeting the locus, error-prone repair indirectly competes with HDR, and therefore reduces the rates of precise correction from donor templates. Furthermore, if the targeted allele is a hypomorph with residual gene function, the generated indels can further worsen the clinical phenotype of the disease (Chu et al., 2015). In recent years, several attempts have therefore been made to enhance HDR-mediated correction of CRISPR-Cas9-induced DSBs from donor oligos. Based on the knowledge that HDR pathways are primarily active during the S/G2 phase of the cell cycle, cells have been synchronized prior to CRISPR-Cas9 delivery (Lin et al., 2014a), and Cas9 expression has been limited to the S/G2/M phase of the cell cycle (Gutschner et al., 2016; Howden et al., 2016). Other studies have increased HDR by chemically modulating the EJ and HDR pathways (Chu et al., 2015; Maruyama et al., 2015; Yu et al., 2015; Song et al., 2016), and by rationally designing DNA repair templates with optimal homology arm lengths (Richardson et al., 2016). In addition, it has been proposed that the availability of the DNA repair template might present a rate-limiting factor for HDR, and that enhancing the local concentration of donor oligos could increase the correction rates (Ruff et al., 2014; Carlson-Stevermer et al., 2017). Based on this hypothesis, we here generated and tested novel CRISPR-Cas9 variants, in which the DNA repair template is covalently conjugated to Cas9 (Figure 1).

Figure 1

Download asset Open asset

To be able to measure HDR efficiencies of novel CRISPR-Cas9 variants in a rapid and high-throughput manner, we first generated a fluorescent reporter system (Figure 2a–b). In brief, the reporter cassette was stably integrated in HEK293T cells, and expresses a green fluorescent protein (GFP) that is preceded by an inactive mutant version of a red fluorescent protein (mutRFP). While precise correction of the mutation via HDR from donor templates leads to re-activation of RFP activity, the generation of frame shifts via error-prone EJ pathways leads to loss of GFP activity (Figure 2a–b). The correction and indel formation events can be visualized by fluorescence imaging and quantified by FACS (Figure 2c–d). To test the functionality of the reporter system, and to determine the optimal length of DNA repair templates, we first transfected Cas9-sgRNA ribonucleoprotein (RNP) complexes and single stranded (ss) oligo repair templates of different lengths (Figure 3—figure supplement 1a). In line with previous studies (Zuo et al., 2017), we found that maximal DSB correction rates are reached with ss-donor oligos of approximately 80 bases. We therefore continued our study with 81-nucleotide (81-mers) ss-donor oligos but also included 65-nucleotide (65-mers) oligos, as we reasoned that if repair templates are brought in proximity to DSBs also shorter homology arms could be sufficient for HDR.

Figure 2

Download asset Open asset

In order to link the donor oligos to the Cas9 protein, we used the SNAP-tag technology, which allows covalent binding of O⁶-benzylguanine (BG)-labeled molecules to SNAP-tag fusion proteins (Keppler et al., 2003). To generate O⁶-benzylguanine (BG)-linked DNA repair templates, we first coupled amine-modified oligos to commercially available amine-reactive BG building blocks (Figure 3a, Figure 3—figure supplement 1c). The BG-linked oligos were further separated from unreacted oligos by HPLC (Figure 3b) and analyzed by liquid chromatography-mass spectrometry (LC-MS) to confirm their purity (Figure 3—figure supplement 1d). Next, we produced recombinant Cas9 proteins with a SNAP-tag fused to the C-terminus (Figure 3c,d). The fusion proteins were then complexed with the BG-coupled oligos, and covalent binding was confirmed by SDS-PAGE (Figure 3e–h). The protein-oligo conjugate was mixed with in vitro transcribed sgRNAs targeting the mutRFP locus (Figure 3—figure supplement 1g,h), finally generating the Cas9 ribonucleoprotein-DNA (RNPD) complex.

Figure 3 with 1 supplement see all

Download asset Open asset

Figure 3—source data 1

Numerical data and the exact p values for all graphs in Figure 3—figure supplement 1.

https://doi.org/10.7554/eLife.33761.007

Download elife-33761-fig3-data1-v4.xlsx

To test if linking the donor oligo to Cas9 changes the ratio between indel formation and correction from the repair template, we used our reporter system to compare Streptococcus pyogenes Cas9 (SpCas9) complexes with linked repair oligos (Figure 4b) to the control SpCas9 complexes with unlinked repair oligos (SpCas9-SNAP with unlabeled oligos). Notably, the correction efficiency (percentage of corrections in edited cells) with bound complexes was significantly enhanced, from 2.1% to 22.5% with the 65-mers and from 8.9% to 25.7% with the 81-mers (Figure 4a, Figure 4—figure supplement 1a,b). In comparison to unbound complexes this represented 11- and 3-fold increases, respectively.

Figure 4 with 2 supplements see all

Download asset Open asset

Figure 4—source data 1

Numerical data and the exact p values for all graphs in Figure 4.

https://doi.org/10.7554/eLife.33761.011

Download elife-33761-fig4-data1-v4.xlsx

Figure 4—source data 2

Numerical data for all graphs in Figure 4—figure supplement 1.

https://doi.org/10.7554/eLife.33761.012

Download elife-33761-fig4-data2-v4.xlsx

Figure 4—source data 3

Numerical data and the exact p values for all graphs in Figure 4—figure supplement 2.

https://doi.org/10.7554/eLife.33761.013

Download elife-33761-fig4-data3-v4.xlsx

To further test our hypothesis that spatial and temporal co-localization of repair templates and Cas9 enhances correction rates, we next developed an independent approach where the donor oligo is not bound to the Cas9 complex that induces the DSB, but to a second catalytically inactive Cas9 complex that binds in close proximity to the DSB (RNP-RNPD system). To avoid the interchange of sgRNAs between both complexes, we designed a two-component system in which the DSB is induced by SpCas9, and the repair template is linked to a catalytically inactive Staphylococcus aureus (Sa)dCas9 (Figure 4c,d). We co-transfected both complexes into the reporter cell line, and quantified correction and indel formation rates. Notably, the correction efficiency increased from 4.1% to 29.9% with 65-mers, and from 11.6% to 32.3% with 81-mers (Figure 4e, Figure 4—figure supplement 1c,d), confirming our previous results with the RNPD system.

In vivo, the delivery efficiency of RNPs and oligos is generally lower than in vitro. Thus, if the repair template is not bound to Cas9, there is a substantial probability that only one of the two components would be delivered into the cell. In addition, at lower transfection efficiencies fewer repair templates are present in the nucleus, potentially decreasing HDR rates. As we presumed that linking the donor oligo to Cas9 should largely alleviate these limitations, we investigated whether the repair efficiency with template-conjugated Cas9 is affected when complexes are transfected at 5-fold lower concentrations. Importantly, although under these conditions the correction efficiencies were generally lower with both coupled and uncoupled Cas9 complexes, the difference between the two systems was however even more pronounced. Compared to the uncoupled RNP complex, the RNPD system yielded 20-fold and a 4-fold increases in repair efficiency with 65-mer and 81-mer repair template oligos, respectively (Figure 4f, Figure 4—figure supplement 1e,f). Similarly, the two-component RNP-RNPD system led to a 21-fold increase with 65-mers and a 6-fold increase with 81-mers (Figure 4f, Figure 4—figure supplement 1e,f). Taken together, our results suggest that linking the repair template to the Cas9 complex leads to improved correction efficiency compared to the unlinked control CRISPR-Cas9 system, and that this effect is even more pronounced when CRISPR-Cas9 components are delivered at lower concentrations.

Next, we aimed to gain mechanistic insight into the processes that lead to enhanced correction rates when the donor oligo is linked to the Cas9 complex. We first assessed if the BG-labelling of the donor oligo itself already influences the correction efficiencies, and compared the correction rates of wild-type SpCas9 lacking the SNAP-tag together with either unlabelled oligo or BG-labelled oligo. While the correction rates were enhanced when the donor oligo was labelled with BG, the increase was several fold lower compared to the system where the oligo was conjugated to the Cas9 RNP complex (Figure 4—figure supplement 2a–c). We then investigated if the observed improvement in correction efficiency is due to the donor oligo being brought in close proximity to the DSB, or if it is sufficient to direct the donor oligo to the nucleoplasm. We therefore again employed the two-component system with DSB inducing SpCas9 and catalytically inactive SadCas9 conjugated to the donor oligo. While in one group SadCas9 was complexed with a sgRNA that directs it in close proximity to the DSB, in the other group the sgRNA was omitted and SadCas9 was therefore only directed into the nucleus. Importantly, our results demonstrated that adding the sgRNA did not further enhance correction rates, suggesting that the increase of donor oligo concentration in the nucleoplasm was sufficient to fully account for the positive effect of the Cas9-donor oligo conjugation (Figure 4g, Figure 4—figure supplement 2d,e). In line with these observations, a number of previous studies demonstrated that exogenous DNA transport into the nucleus is one of the major barriers to effective gene delivery (Subramanian et al., 1999; Zanta et al., 1999; Ludtke et al., 1999; Aronsohn and Hughes, 1998).

To validate our results from the HEK293T reporter cells, we next tested our approach at different endogenous genomic loci and in different cell types. We first targeted the human beta globin (HBB) locus in the K562 cell line, and analyzed correction and editing frequencies using next generation sequencing (NGS). The mean correction efficiency with the RNPD system was 19.6%, which represented a 17-fold increase compared to the control RNP system (Figure 5a, Supplementary file 2). Next we targeted the Rosa26 and proprotein convertase subtilisin/kexin type 9 (Pcsk9) locus in mouse embryonic stem cells (mESCs). Again, the mean correction efficiencies of RNPD systems were significantly increased, to 18.6% at the Rosa26 locus and 23.2% at the Pcsk9 locus (Figure 5b,c, Supplementary file 2). In comparison to the uncoupled RNP complexes, this represented 2- and 6-fold increases, respectively.

Figure 5

Download asset Open asset

Figure 5—source data 1

Numerical data and the exact p values for all graphs in Figure 5.

https://doi.org/10.7554/eLife.33761.015

Download elife-33761-fig5-data1-v4.xlsx

In the previous experiments the RNPD system was always compared to Cas9 SNAP-tag fusion proteins with uncoupled donor oligos. To also directly compare the engineered RNPD system to the classical CRISPR-Cas9 system, we performed experiments where we used wild-type Cas9 with the uncoupled donor oligos as a control. We first targeted the fluorescent reporter locus and analyzed it by NGS. We found that while the mean percentage of corrected loci increased from 0.8% with the classical Cas9 system to 4.9% with the RNPD system, the number of incorrectly edited loci slightly decreased from 12.6% to 9.3% (Figure 6a, Supplementary file 2,3,4). This corresponds to a 7-fold increase in correction efficiency (Figure 6b). In addition, the analysis of three computationally predicted off-target sites (Lin et al., 2014b; Cradick et al., 2014) of the reporter locus, suggests that the risk for generating off-target mutations is not enhanced with the RNPD system (Figure 6c, Supplementary file 2,3,4). In the next step we also targeted and analyzed the endogenous loci HBB, empty spiracles homeobox 1 (EMX1), and C-X-C chemokine receptor type 4 (CXCR4) in HEK293T cells. NGS analysis revealed that in all three loci the mean correction efficiency of the RNPD system was markedly increased to: 34,4% at the HBB locus, 28.6% at the EMX1 locus and 33.1% at the CXCR4 locus (Figure 6d,e,f, Supplementary file 2,3,4). Compared to the classical CRISPR-Cas9 system this represents a 20-fold, a 10-fold, and a 24-fold increase, respectively (Figure 6d,e,f).

Figure 6

Download asset Open asset

Figure 6—source data 1

Numerical data and the exact p values for all graphs in Figure 6.

https://doi.org/10.7554/eLife.33761.017

Download elife-33761-fig6-data1-v4.xlsx

Direct delivery of Cas9 RNP complexes into tissues promises great potential for therapeutic applications. Compared to genetically encoded systems, RNPs avoid the danger of genomic integration, and due to their limited lifetime, the risk of off-target activities is low (Kim et al., 2014). In addition, procedures for large-scale production of recombinant proteins for clinical use are well established, and several recently developed protocols enable in vivo delivery of Cas9 RNP complexes in animal models (Wang et al., 2016; Zuris et al., 2015; Staahl et al., 2017; Lee et al., 2017). Here, we present a method where we enhance correction efficiency of Cas9-induced DSBs by conjugating the donor oligo to the Cas9 complex. Our data suggests that the increase in HDR efficiency is caused by enhanced nuclear concentration of the repair template. Unlike previous approaches that increase HDR rates by chemically modulating DNA repair pathways, our approach does not alter endogenous cellular processes, thus reducing risk of potential negative side effects. In addition, covalent linkage of the repair template to the Cas9 RNP complex also addresses another central challenge of in vivo gene editing therapies – namely that simultaneous delivery of the RNP complex and the repair template needs to be ensured. Taken together, we suggest that covalently linking the DNA repair template to the Cas9 RNP complex is poised to further drive the CRISPR/Cas technology towards clinical translation.

Please see Supplementary file 1-Supplementary Tables 1–6 for a list of the DNA sequences used in this manuscript.

The data that support the findings of this study are available within the paper and its Supplementary files. Source data files have been provided for Figure 4, Figure 5, Figure 6, Figure 3-Figure Supplement 1, Figure 4-Figure Supplement 1 and Figure 4-Figure Supplement 2. Scripts for mapping sequencing data, counting mutations and generating plots are available at https://github.com/HLindsay/Savic_CRISPR_HDR (copy archived at https://github.com/elifesciences-publications/Savic_CRISPR_HDR). Fastq files have been uploaded to ArrayExpress, and accession number is E-MTAB-6808.

1. 1. Helen Lindsay
   2. Natasa Savic

   (2018) Fastq file from Covalent linkage of the DNA repair template to the CRISPR-Cas9 nuclease enhances homology-directed repair

   Publicly available at the EMBL-EBI Array Express (accession no. E-MTAB-6808).

   https://www.ebi.ac.uk/arrayexpress/experiments/E-MTAB-6808/

1. 1. Aronsohn AI
   2. Hughes JA

   (1998) Nuclear localization signal peptides enhance cationic liposome-mediated gene therapy

   Journal of Drug Targeting 5:163–169.

   https://doi.org/10.3109/10611869808995871

   • PubMed
   • Google Scholar
2. 1. Brazma A
   2. Parkinson H
   3. Sarkans U
   4. Shojatalab M
   5. Vilo J
   6. Abeygunawardena N
   7. Holloway E
   8. Kapushesky M
   9. Kemmeren P
   10. Lara GG
   11. Oezcimen A
   12. Rocca-Serra P
   13. Sansone SA

   (2003) ArrayExpress--a public repository for microarray gene expression data at the EBI

   Nucleic Acids Research 31:68–71.

   https://doi.org/10.1093/nar/gkg091

   • PubMed
   • Google Scholar
3. 1. Carlson-Stevermer J
   2. Abdeen AA
   3. Kohlenberg L
   4. Goedland M
   5. Molugu K
   6. Lou M
   7. Saha K

   (2017) Assembly of CRISPR ribonucleoproteins with biotinylated oligonucleotides via an RNA aptamer for precise gene editing

   Nature Communications 8:1711.

   https://doi.org/10.1038/s41467-017-01875-9

   • PubMed
   • Google Scholar
4. 1. Ceccaldi R
   2. Rondinelli B
   3. D'Andrea AD

   (2016) Repair pathway choices and consequences at the Double-Strand break

   Trends in Cell Biology 26:52–64.

   https://doi.org/10.1016/j.tcb.2015.07.009

   • PubMed
   • Google Scholar
5. 1. Chu VT
   2. Weber T
   3. Wefers B
   4. Wurst W
   5. Sander S
   6. Rajewsky K
   7. Kühn R

   (2015) Increasing the efficiency of homology-directed repair for CRISPR-Cas9-induced precise gene editing in mammalian cells

   Nature Biotechnology 33:543–548.

   https://doi.org/10.1038/nbt.3198

   • PubMed
   • Google Scholar
6. 1. Cradick TJ
   2. Qiu P
   3. Lee CM
   4. Fine EJ
   5. Bao G

   (2014) COSMID: a Web-based tool for identifying and validating CRISPR/Cas Off-target sites

   Molecular Therapy - Nucleic Acids 3:e214.

   https://doi.org/10.1038/mtna.2014.64

   • PubMed
   • Google Scholar
7. 1. Gutschner T
   2. Haemmerle M
   3. Genovese G
   4. Draetta GF
   5. Chin L

   (2016) Post-translational regulation of Cas9 during G1 enhances Homology-Directed repair

   Cell Reports 14:1555–1566.

   https://doi.org/10.1016/j.celrep.2016.01.019

   • PubMed
   • Google Scholar
8. 1. Howden SE
   2. McColl B
   3. Glaser A
   4. Vadolas J
   5. Petrou S
   6. Little MH
   7. Elefanty AG
   8. Stanley EG

   (2016) A Cas9 variant for efficient generation of Indel-Free knockin or Gene-Corrected human pluripotent stem cells

   Stem Cell Reports 7:508–517.

   https://doi.org/10.1016/j.stemcr.2016.07.001

   • PubMed
   • Google Scholar
9. 1. Jinek M
   2. Chylinski K
   3. Fonfara I
   4. Hauer M
   5. Doudna JA
   6. Charpentier E

   (2012) A programmable dual-RNA-guided DNA endonuclease in adaptive bacterial immunity

   Science 337:816–821.

   https://doi.org/10.1126/science.1225829

   • PubMed
   • Google Scholar
10. 1. Keppler A
    2. Gendreizig S
    3. Gronemeyer T
    4. Pick H
    5. Vogel H
    6. Johnsson K

    (2003) A general method for the covalent labeling of fusion proteins with small molecules in vivo

    Nature Biotechnology 21:86–89.

    https://doi.org/10.1038/nbt765

    • PubMed
    • Google Scholar
11. 1. Kim JH
    2. Lee SR
    3. Li LH
    4. Park HJ
    5. Park JH
    6. Lee KY
    7. Kim MK
    8. Shin BA
    9. Choi SY

    (2011) High cleavage efficiency of a 2A peptide derived from porcine teschovirus-1 in human cell lines, zebrafish and mice

    PLoS One 6:e18556.

    https://doi.org/10.1371/journal.pone.0018556

    • PubMed
    • Google Scholar
12. 1. Kim S
    2. Kim D
    3. Cho SW
    4. Kim J
    5. Kim JS

    (2014) Highly efficient RNA-guided genome editing in human cells via delivery of purified Cas9 ribonucleoproteins

    Genome Research 24:1012–1019.

    https://doi.org/10.1101/gr.171322.113

    • PubMed
    • Google Scholar
13. 1. Lee K
    2. Conboy M
    3. Park HM
    4. Jiang F
    5. Kim HJ
    6. Dewitt MA
    7. Mackley VA
    8. Chang K
    9. Rao A
    10. Skinner C
    11. Shobha T
    12. Mehdipour M
    13. Liu H
    14. Huang W-chin
    15. Lan F
    16. Bray NL
    17. Li S
    18. Corn JE
    19. Kataoka K
    20. Doudna JA
    21. Conboy I
    22. Murthy N

    (2017) Nanoparticle delivery of Cas9 ribonucleoprotein and donor DNA in vivo induces homology-directed DNA repair

    Nature Biomedical Engineering 1:889–901.

    https://doi.org/10.1038/s41551-017-0137-2

    • Google Scholar
14. 1. Li H

    (2010)

    Aligning New-Sequencing Reads by BWA

    Aligning New-Sequencing Reads by BWA.

    • Google Scholar
15. 1. Lin S
    2. Staahl BT
    3. Alla RK
    4. Doudna JA

    (2014a) Enhanced homology-directed human genome engineering by controlled timing of CRISPR/Cas9 delivery

    eLife 3:e04766.

    https://doi.org/10.7554/eLife.04766

    • PubMed
    • Google Scholar
16. 1. Lin Y
    2. Cradick TJ
    3. Brown MT
    4. Deshmukh H
    5. Ranjan P
    6. Sarode N
    7. Wile BM
    8. Vertino PM
    9. Stewart FJ
    10. Bao G

    (2014b) CRISPR/Cas9 systems have off-target activity with insertions or deletions between target DNA and guide RNA sequences

    Nucleic Acids Research 42:7473–7485.

    https://doi.org/10.1093/nar/gku402

    • PubMed
    • Google Scholar
17. 1. Lindsay H
    2. Burger A
    3. Biyong B
    4. Felker A
    5. Hess C
    6. Zaugg J
    7. Chiavacci E
    8. Anders C
    9. Jinek M
    10. Mosimann C
    11. Robinson MD

    (2016) CrispRVariants charts the mutation spectrum of genome engineering experiments

    Nature Biotechnology 34:701–702.

    https://doi.org/10.1038/nbt.3628

    • PubMed
    • Google Scholar
18. 1. Lindsay H

    (2018) Github

    Github, https://github.com/HLindsay/Savic_CRISPR_HDR.

    https://github.com/HLindsay/Savic_CRISPR_HDR

    • Google Scholar
19. 1. Ludtke JJ
    2. Zhang G
    3. Sebestyén MG
    4. Wolff JA

    (1999)

    A nuclear localization signal can enhance both the nuclear transport and expression of 1 kb DNA

    Journal of cell Science 112:2033–2041.

    • PubMed
    • Google Scholar
20. 1. Mali P
    2. Yang L
    3. Esvelt KM
    4. Aach J
    5. Guell M
    6. DiCarlo JE
    7. Norville JE
    8. Church GM

    (2013) RNA-guided human genome engineering via Cas9

    Science 339:823–826.

    https://doi.org/10.1126/science.1232033

    • PubMed
    • Google Scholar
21. 1. Mao Z
    2. Bozzella M
    3. Seluanov A
    4. Gorbunova V

    (2008) Comparison of nonhomologous end joining and homologous recombination in human cells

    DNA Repair 7:1765–1771.

    https://doi.org/10.1016/j.dnarep.2008.06.018

    • PubMed
    • Google Scholar
22. 1. Maruyama T
    2. Dougan SK
    3. Truttmann MC
    4. Bilate AM
    5. Ingram JR
    6. Ploegh HL

    (2015) Increasing the efficiency of precise genome editing with CRISPR-Cas9 by inhibition of nonhomologous end joining

    Nature Biotechnology 33:538–542.

    https://doi.org/10.1038/nbt.3190

    • PubMed
    • Google Scholar
23. 1. Richardson CD
    2. Ray GJ
    3. DeWitt MA
    4. Curie GL
    5. Corn JE

    (2016) Enhancing homology-directed genome editing by catalytically active and inactive CRISPR-Cas9 using asymmetric donor DNA

    Nature Biotechnology 34:339–344.

    https://doi.org/10.1038/nbt.3481

    • PubMed
    • Google Scholar
24. Preprint

    1. Richardson CD

    (2017) CRISPR-Cas9 genome editing in human cells works via the fanconi Anemia pathway

    bioRxiv.

    https://doi.org/10.1101/136028

    • Google Scholar
25. 1. Ruff P
    2. Koh KD
    3. Keskin H
    4. Pai RB
    5. Storici F

    (2014) Aptamer-guided gene targeting in yeast and human cells

    Nucleic Acids Research 42:e61.

    https://doi.org/10.1093/nar/gku101

    • PubMed
    • Google Scholar
26. 1. Shalem O
    2. Sanjana NE
    3. Hartenian E
    4. Shi X
    5. Scott DA
    6. Mikkelson T
    7. Heckl D
    8. Ebert BL
    9. Root DE
    10. Doench JG
    11. Zhang F

    (2014) Genome-scale CRISPR-Cas9 knockout screening in human cells

    Science 343:84–87.

    https://doi.org/10.1126/science.1247005

    • PubMed
    • Google Scholar
27. 1. Song J
    2. Yang D
    3. Xu J
    4. Zhu T
    5. Chen YE
    6. Zhang J

    (2016) RS-1 enhances CRISPR/Cas9- and TALEN-mediated knock-in efficiency

    Nature Communications 7:10548.

    https://doi.org/10.1038/ncomms10548

    • PubMed
    • Google Scholar
28. 1. Staahl BT
    2. Benekareddy M
    3. Coulon-Bainier C
    4. Banfal AA
    5. Floor SN
    6. Sabo JK
    7. Urnes C
    8. Munares GA
    9. Ghosh A
    10. Doudna JA

    (2017) Efficient genome editing in the mouse brain by local delivery of engineered Cas9 ribonucleoprotein complexes

    Nature Biotechnology 35:431–434.

    https://doi.org/10.1038/nbt.3806

    • PubMed
    • Google Scholar
29. 1. Subramanian A
    2. Ranganathan P
    3. Diamond SL

    (1999) Nuclear targeting peptide scaffolds for lipofection of nondividing mammalian cells

    Nature Biotechnology 17:873–877.

    https://doi.org/10.1038/12860

    • PubMed
    • Google Scholar
30. 1. van den Bosch M
    2. Lohman PH
    3. Pastink A

    (2002) DNA double-strand break repair by homologous recombination

    Biological Chemistry 383:873–892.

    https://doi.org/10.1515/BC.2002.095

    • PubMed
    • Google Scholar
31. 1. Wang M
    2. Zuris JA
    3. Meng F
    4. Rees H
    5. Sun S
    6. Deng P
    7. Han Y
    8. Gao X
    9. Pouli D
    10. Wu Q
    11. Georgakoudi I
    12. Liu DR
    13. Xu Q

    (2016) Efficient delivery of genome-editing proteins using bioreducible lipid nanoparticles

    PNAS 113:2868–2873.

    https://doi.org/10.1073/pnas.1520244113

    • PubMed
    • Google Scholar
32. 1. Yu C
    2. Liu Y
    3. Ma T
    4. Liu K
    5. Xu S
    6. Zhang Y
    7. Liu H
    8. La Russa M
    9. Xie M
    10. Ding S
    11. Qi LS

    (2015) Small molecules enhance CRISPR genome editing in pluripotent stem cells

    Cell Stem Cell 16:142–147.

    https://doi.org/10.1016/j.stem.2015.01.003

    • PubMed
    • Google Scholar
33. 1. Zanta MA
    2. Belguise-Valladier P
    3. Behr JP

    (1999) Gene delivery: a single nuclear localization signal peptide is sufficient to carry DNA to the cell nucleus

    PNAS 96:91–96.

    https://doi.org/10.1073/pnas.96.1.91

    • PubMed
    • Google Scholar
34. 1. Zhang J
    2. Kobert K
    3. Flouri T
    4. Stamatakis A

    (2014) PEAR: a fast and accurate Illumina Paired-End reAd mergeR

    Bioinformatics 30:614–620.

    https://doi.org/10.1093/bioinformatics/btt593

    • PubMed
    • Google Scholar
35. 1. Zuo E
    2. Cai YJ
    3. Li K
    4. Wei Y
    5. Wang BA
    6. Sun Y
    7. Liu Z
    8. Liu J
    9. Hu X
    10. Wei W
    11. Huo X
    12. Shi L
    13. Tang C
    14. Liang D
    15. Wang Y
    16. Nie YH
    17. Zhang CC
    18. Yao X
    19. Wang X
    20. Zhou C
    21. Ying W
    22. Wang Q
    23. Chen RC
    24. Shen Q
    25. Xu GL
    26. Li J
    27. Sun Q
    28. Xiong ZQ
    29. Yang H

    (2017) One-step generation of complete gene knockout mice and monkeys by CRISPR/Cas9-mediated gene editing with multiple sgRNAs

    Cell Research 27:933–945.

    https://doi.org/10.1038/cr.2017.81

    • PubMed
    • Google Scholar
36. 1. Zuris JA
    2. Thompson DB
    3. Shu Y
    4. Guilinger JP
    5. Bessen JL
    6. Hu JH
    7. Maeder ML
    8. Joung JK
    9. Chen ZY
    10. Liu DR

    (2015) Cationic lipid-mediated delivery of proteins enables efficient protein-based genome editing in vitro and in vivo

    Nature Biotechnology 33:73–80.

    https://doi.org/10.1038/nbt.3081

    • PubMed
    • Google Scholar

[Editors’ note: this article was originally rejected after discussions between the reviewers, but the authors were invited to resubmit after an appeal against the decision.]

Thank you for submitting your work entitled "Covalent linkage of the DNA repair template to the CRISPR/Cas9 complex enhances homology-directed repair" for consideration by eLife. Your article has been evaluated by a Senior Editor and three reviewers, one of whom is a member of our Board of Reviewing Editors. The reviewers have opted to remain anonymous.

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife.

The paper by Savic et al., "Covalent linkage of the DNA repair template to the CRISPR/Cas9 complex enhances homology directed repair" reports a novel approach to improve the efficiency of HR relative to NHEJ upon Cas9 induced DSB. The strategy relies on the assumption that enhancing the spatial proximity or local concentration of DNA used as template for HR would increase the use of this pathway. This is achieved by generating covalently coupled Cas9-oligonucleotide. An increased efficiency of HR is observed which could be interesting for some applications if confirmed.

However all reviewers have identified several problems in the manuscript that could only be solved pending designing a whole new set of experiments. This additional work extends beyond a revision and the manuscript cannot therefore be accepted for publication in eLife.

The major issues raised by the reviewers are:

- The effect observed with the coupled substrate does not demonstrate that it is due to bringing the oligonucleotide in proximity to the DSB.

- The effect observed should be validated on several genomic target sites, first to extend the observation to a genomic site and second to show the reproducibility of the effect on different targets.

- The assay used does not allow distinguishing the NHEJ and HR pathways as claimed by the authors.

Reviewer #1:

This paper reports a novel approach to improve the efficiency of HR relative to NHEJ upon Cas9 induced DSB. The strategy relies on the assumption that enhancing the spatial proximity or local concentration of DNA used as template for HR would increase the use of this pathway. This is achieved by generating covalently coupled Cas9-oligonucleotide. The increased efficiency of HR observed could be interesting for some applications.

The experiments are well designed and presented. However several aspects require clarifications as indicated below. Although the increased efficiency of HR is convincing the interpretation is open to several alternatives (i.e. proximity of oligo or stability of oligo). Two additional experiments could clarify this important point.

1) Figure 2A. The figure should be understandable with help of the legend and several pieces of information are missing:

What is 2A ? What is c.190_191delinsCT? (Use a more generic term; specific construct name can be provided in Materials and methods).

I assume the “X” labels the mutation and/or the Cas9 DSB?

I assume the mutant RFP is a substitution?

One needs to see exactly where the substitution is, where the guide maps and where the DSB is introduced and whether the guide would still be able to induce Cas9 cleavage after HR (is the PAM mutated?). One reason is that depending on where the DSB is, more or less end processing will be needed for HR.

2) Figure 2B lower panel (IF): indicate the channels used. Most of the cells seem to be GFP negative in the central panel. How could this be as the FACS indicate 16%? Single channels and overlays should be provided.

What is the percentage of GFP positive among the RFP positive cells? (Close to 100% expected).

It should be noted that the% of edited cells is underestimated since in frame indels (one third ?) can generate a GFP positive cell.

3) Figure 3A. Explain the lane no BG coupling?

4) Figure 4. The same data should not be plotted twice in panel B and C: Panel C should be removed (same comment for Figure 4F and G).

The main question about the use of SadCas9 is to determine whether the difference with SpCas9 is statistically significant. The tests should be performed (Figure 4E, between RNPD coup and RNP-RNPD coup) and if not significant the conclusions should be revised.

A map showing the positions of the different SadCas9 guides used (1 to 4) should be shown.

5) Two experiments are required to describe the effect observed: In order to know if the effect observed is due to coupling the oligo to Cas9, a control should be performed with the oligo coupled to BG but not to Cas9. Clearly at least part of the effect observed could be due to stabilization of the oligo rather than proximity to DSB per say. To distinguish these possibilities the authors should test an oligo coupled to SadCas9 but without the corresponding guide.

I assume S. aureus was used such as to design a distinct guide specific for S. aureus and not bound by S. pyogenes, if so, this should be explained in the text.

6) Figure 4—figure supplement 1.

What is the interpretation for the decrease of percentage of edited cells with oligo coupled to SadCas9 ? (Figure 4—figure supplement 1D).

Legend of Figure 4—figure supplement 1C and D: RNP-RNP unco is grey but should be hatched grey box.

7) Two-tailed tests should be used not one-tailed since there is no reason to assume a priori that the coupled protein would be more efficient.

8) Subsection “Expression and purification of Cas9-SNAP”, missing word: "The was further purified…"

Reviewer #2:

The authors show that covalent linkage of the DNA repair template to the CRISPR/Cas9 complex enhances HDR efficiency. They conclude this based on the use of a traffic light HDR reporter system, transiently transfected into HEK293 cells.

Although this is interesting and aspects of the study are really clever (e.g. the use of the sa-dCAS9 linked template to prove that physical proximity is the key), the work strikes me as quite preliminary as the increases are only shown in this transient reporter system using only one target site.

Without more thorough testing of the improvement in a real experimental system, I can't recommend acceptance of this study in its current form at eLife. My recommendations to improve the impact of this publication would be to demonstrate real efficiency improvements using a targeted edit within a genomic target site. In this respect, the existing work using transient reporter system would be excellent preliminary data for a real proof-of-concept – namely the increased efficiency in a real experimental genome engineering project e.g. to create a point mutation.

In addition, the CRISPR field has been populated by numerous claims showing positive perturbations to shift the balance of repair pathway from NHEJ to HDR (alterations in experimental design, NHEJ inhibitors, etc.). Frequently these claims have been made on the basis of data from manipulation of a single target site. Reproducibility of these studies has proven to be low, as when applied by other labs at different target sites, the reported improvements are frequently not replicated. To avoid this occurring, I would recommend that the authors address several different genomic target sites and report the level of improvement in HDR seen in these various experimental settings.

Figure 4E presents data which suggests that the positioning of the sgRNA (and hence the DSB) away from the target nucleotide to be mutated (0 bp, 52 bp, 61 bp, 83 bp and 128bp) has little influence on the correction efficiency. This is significantly at odds with experimental data from genomic target sites, and should be discussed. In addition, I wonder whether this is a curious artefact of using transient plasmids and thus provides support for my recommendation of addressing a single copy genomic target site or sites to validate their approach in a real experimental setting.

The work is very similar to a recent bioRxiv report by Janet Rossant's group (https://doi.org/10.1101/204339) who demonstrated that covalent attachment of the repair template to the CRISPR/Cas9 complex enhanced HDR rates, reporting data from fluorescent cassette insertions at 5 genomic target sites. Although this paper hasn't been peer reviewed, it's tempting to speculate that the authors of this manuscript have submitted what is essentially a preliminary study to eLife to compete with this more thorough study, which is presumably in review at another journal.

Reviewer #3:

Gene targeting constitutes a promising approach for the generation of novel biological models and for future gene therapy strategies. The capacity to generate targeted cleavage by the CRISPR-CAS9 system raised many hopes. However, targeted gene replacement still remains to too low levels. In the present work the authors design a strategy aiming at delivering the correcting DNA in the vicinity of the cleaved site. They covalently bound the correction oligonucleotide to the CAS9 itself via "click chemistry", using the SNAP-tag technology. Then they assemble in vitro the complex with the RNA-guide and transfected cells (HEK293). This strategy is elegant and promising. However, many concerns should be addressed.

The main problem is the fact that the authors used only on system to monitor HR; and that the designing of this substrate is based on wrong considerations on HR and NHEJ.

It is claimed that NHEJ is error-prone. This is wrong. There are two kinds of end-joining processes, the canonical NHEJ (which is not error-prone, but conservative) and the Alternative end-joining (alt-NHEJ, MMEJ, B-NHEJ), which is mutagenic and error-prone. Therefore mutagenic repair does not automatically imply NHEJ. Second it is said in the Introduction that HR and NHEJ directly compete. This is also wrong; in fact things happen in two phases: first competition between cNHEJ and single-strand DNA resection, second on resected DNA extremities, competition between HR and alternative end-joining. Finally, HR can also generate mutagenesis. Therefore many concerns exist on the strategy used here (which is the sole assay used) because mutagenic repair can arise by many other processes than NHEJ.

Moreover, is correction with oligonucleotides (65, 81 b) an actual HR mechanism? This is not consistent with concept of MEPS.

The authors should first genetically validate their reporter system, in cells mutated for HR or NHEJ.

The authors should also verify their strategy with natural endogenous target sequences, instead of the reporter.

There are no data on the transfection efficiency. Especially, comparing the CAS9 with the engineered one.

Similarly, does the two CAS9 cleave with similar efficiency?

Does the modification of the CAS9 affect its cleavage specificity?

Are there any off-target effects? (off-target cleavage, off-target integration of the oligo).

The authors should test different RNA guides for a common target.

The author should test different cell lines.

It is not clear what is actually measured. Where are the mutagenic repair (pseudo-NHEJ) measurements? Is it the frequency of HDR or the ratio HDR/pseudo-NHEJ?

[Editors’ note: what now follows is the decision letter after the authors submitted for further consideration.]

Thank you for resubmitting your work entitled "Covalent linkage of the DNA repair template to the CRISPR-Cas9 nuclease enhances homology-directed repair" for further consideration at eLife. Your revised article has been favorably evaluated by Diethard Tautz (Senior Editor) and the Reviewing Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

The authors provide convincing answers to all reviewers' comments. Importantly, the authors show that the main improvement achieved by their experimental approach is due to enhanced targeting of the oligonucleotides to the nucleus rather that its targeting to the specific target genomic site. This could also be highlighted in the Abstract.

A few points need to be clarified:

1) Source data 6 refers to Figure 4—figure supplement 2 (not 3?) should be registered as Figure 4—source data 3.

2) Source data 5 refers to Figure 4—figure supplement 1 (not 2?) should be registered as Figure 4—source data 2.

3) Figure 2B requires clarification (and in the legend as well). Legend says "the mutation substitutes… CT.. to TA". This is ambiguous because TA is wild type. It would clarify to draw in Figure 2, the wild type and mutated genomic sequences (with the codon substitution) and the sequence of the guide (also can be confusing as drawn because the guide has CT not TA).

4) Figure 5: In Figure 5, only the reference is shown. Mutated variant should be indicated. For instance, on 5A, at HBB, the DNA sequence shown should be the one after correction (to be consistent with Figure.2B) ? However it is identical to guide sequence?

Please explain the correction efficiency values, obtained after correction for transfection efficiency? If so, provide transfection efficiencies, otherwise read data from Supplementary file 2 cannot be understood: i.e. 27% at Rosa 26, how was this obtained? In addition, since no HR reads is detected at Rosa26 (why is this?), the authors should comment in the main text on the difference uncoupled/coupled from Figure 5B for this locus.

5) Figure 6A requires clarifications both in the main text and in the presentation of the figure. The authors summarize in one sentence all the NGS data (Results and Discussion, eighth paragraph). There is a lot of data in this analysis and it would be better to present the analysis step by step and not to refer to all panels 6A, B, D, E, F at once. Reporter locus could be presented first, with comments about percentage of corrected reads and other events (legend should explain the nomenclature, for instance -7:7D, and also what are I, II and III: triplicates I assume). Then results at other loci should be briefly discussed (and referred to Supplementary file 2). Also indicate in Supplementary file 2 that the variants called "SNV" are the ones predicted after repair by HR. An interesting information is also the relative proportion of HR versus non HR events. In Figure 6, it seems that this proportion is about 30% in experiment I (5.63/(100-85.19)). Is this correct? It could be discussed in the main text.

[Editors’ note: the author responses to the first round of peer review follow.]

The major issues raised by the reviewers are:

- The effect observed with the coupled substrate does not demonstrate that it is due to bringing the oligonucleotide in proximity to the DSB.

To gain mechanistic insight into the process that leads to an increased correction rate when the donor oligo is coupled to the Cas9 complex, we performed the experiments reviewer 1 suggested. First, we tested if the increased repair rate is caused by BG-labelling of the donor oligo, and compared the editing rates of SpCas9 WT together with either unlabelled oligo or BG-labelled oligo (note that SpCas9 WT cannot bind BG-oligos). Although we observed a higher repair rate with the BG-labelled donor oligo, the increase was several fold lower compared to the experiment where the oligo was linked to the Cas9 complex (Figure 4—figure supplement 2A-C). Next, we tested if the observed improvement in correction efficiency is due to the donor oligo being brought in close proximity to the DSB, or if it is sufficient to direct the donor oligo to the nucleus. We therefore used our 2-component system (SadCas9 binds the oligo and SpCas9 generates the DSB) with and without the SadCas9 guide RNA. Our results demonstrate that targeting the donor oligo to the nucleus via SadCas9 fully explains the observed increase in HDR rates (Figure 4G, Figure 4—figure supplement 2D, E). In line with this hypothesis, previous studies have demonstrated that promoting active transport of plasmid DNA to the nucleus (e.g. by binding plasmids to proteins with a nuclear localisation signal (NLS)) enhances expression of genes located on these plasmids (doi: 10.1517/17425247.1.1.127).

- The effect observed should be validated on several genomic target sites, first to extend the observation to a genomic site and second to show the reproducibility of the effect on different targets.

We fully agree with the reviewers that confirming the results from the initial manuscript by targeting endogenous genomic loci in different cell lines/types is required to proof reproducibility and robustness of the system. For the revised manuscript we therefore targeted five endogenous loci in three different cell types (HEK cells, K562 cells, mESCs). Importantly, next generation sequencing revealed that linking the donor oligo to Cas9 robustly enhanced correction rates in all tested loci and cell types (Figure 5A-C, Figure 6D-F).

- The assay used does not allow distinguishing the NHEJ and HR pathways as claimed by the authors.

For the revised manuscript we analysed the sequence of the targeted HEK reporter locus as well as other endogenous loci by next generation sequencing (Figure 5A-C, Figure 6A-F). These data allowed us to directly measure the events in which the DNA break was precisely repaired from the donor oligo (e.g. in the reporter locus exchange of two base pairs at the DSB in the RFP fluorophore region), and events in which the DSBs led to indel mutations.

Nevertheless, as reviewer 3 correctly stated, we always measure DNA repair outcomes (corrections and indels) and not the activities of repair pathways. Since different pathways can trigger indel formation (e.g. c-NHEJ, A-NHEJ, or SSA) or correction from repair templates (e.g. HR or FA repair), it is indeed important to precisely define our wording. In the revised version of the manuscript we discuss this issue in the Introduction, and define homology-directed repair (HDR) as all mechanisms that can repair the locus from the donor oligo sequence, and end-joining (EJ) as all mechanisms that do not use the oligo template for DSB repair.

Of note, the main goal of our study was to develop a novel method for therapeutic gene editing, in which precise repair of DNA breaks from donor oligos is enhanced. Even though it would be very interesting to decipher which molecular pathways are involved in that process, we believe that this question would go beyond the scope of this study, and rather plan to address these topics in future investigations.

Reviewer #1:

[…] The experiments are well designed and presented. However several aspects require clarifications as indicated below.

Although the increased efficiency of HR is convincing the interpretation is open to several alternatives (i.e. proximity of oligo or stability of oligo). Two additional experiments could clarify this important point. In order to know if the effect observed is due to coupling the oligo to Cas9, a control should be performed with the oligo coupled to BG but not to Cas9. Clearly at least part of the effect observed could be due to stabilization of the oligo rather than proximity to DSB per say. To distinguish these possibilities the authors should test an oligo coupled to SadCas9 but without the corresponding guide.

Thank you for suggesting these elegant experiments. As explained above (major issue (i)) we performed both experiments, and found that targeting the donor oligo to the nucleus via SadCas9 fully explains the observed increase in HDR rates (Figure 4G, Figure 4—figure supplement 2D, E).

1) Figure 2A. The figure should be understandable with help of the legend and several pieces of information are missing:

What is 2A?

2A stands for 2A “self-cleaving” peptide: a viral oligopeptide widely used for co-expression of multiple genes. The peptide employs ribosome skipping to mediate “cleavage” of polypeptides during translation in eukaryotic cells. In the revised manuscript we included this explanation in the figure legend.

What is c.190_191delinsCT? (Use a more generic term; specific construct name can be provided in Materials and methods).

According to the recommended mutation nomenclature c.190_191: denotes the mutation in the coding DNA in nucleotides in positions 190 and 191; delins stands for deletion / insertions (substitution of two nucleotides); TA: denote nucleotides that are present in the mutated sequence. In the revised version we clarified the term in the figure legend.

I assume the “X” labels the mutation and/or the Cas9 DSB?

“X” labels the mutation in RFP. In the revised manuscript we added an explanation to the figure legend.

I assume the mutant RFP is a substitution?

Yes, the mutated RFP has a substitution of two nucleotides in the fluorophore. In the revised manuscript we included a more detailed explanation in the figure legend.

One needs to see exactly where the substitution is, where the guide maps and where the DSB is introduced and whether the guide would still be able to induce Cas9 cleavage after HR (is the PAM mutated?). One reason is that depending on where the DSB is, more or less end processing will be needed for HR.

Thank you for this useful comment. In the revised version we included a map of the locus, which indicates the mutated turboRFP fluorophore sequence as well as the protospacer and PAM motif of the guide RNA (Figure 2B). In short, the mutated region is in the fluorophore domain of turboRFP, and upon correction the protospacer sequence is modified (2 bases at position 1 and 2). Thus, if repaired from the donor oligo, the locus cannot be retargeted.

2) Figure 2B lower panel (IF): indicate the channels used. Most of the cells seem to be GFP negative in the central panel. How could this be as the FACS indicate 16%? Single channels and overlays should be provided.

In the revised version we indicated the channels used for the IF (Figure 2D), and included the single channels next to the overlay images. The IF image is just a representation of how well one can distinguish the cells that have lost GFP and gained RFP in our reporter system. The percentage of the cells that lost GFP cannot be determined from one fluorescent image, since it contains only approximately 100 cells that are grown from a few clones. In FACS analysis we always quantify at least 200.000 cells, thus the data is more accurate, and as seen in our biological replicates very robust. In the revised version we show an IF image that is more representative to the FACS quantification.

What is the percentage of GFP positive among the RFP positive cells? (Close to 100% expected).

On average 99.7% of the cells that are RFP positive are also GFP positive (see Author response image 1).

Author response image 1

Download asset Open asset

It should be noted that the% of edited cells is underestimated since in frame indels (one third ?) can generate a GFP positive cell.

Thank you for this important comment. In the revised manuscript we illustrated in Figure 2A and described in the figure legend that the percentage of edited cells is underestimated when using FACS analysis as a readout. Importantly, the NGS experiments (Figure 6) allowed us to accurately quantify in-frame and out-of-frame indels. For the mutRFP reporter locus we found that on average 34% of indels are in-frame mutations. Although this leads to an underestimation of the absolute number of edited cells in our FACS reporter assay, it still allows to accurately compare the editing- and repair efficiencies of different Cas systems.

3) Figure 3A. Explain the lane no BG coupling?

Thank you for suggesting this clarification. The amino modified oligo BG coupling reaction was controlled in two ways:

1) uncoupled: where amino modified oligonucleotide was mixed with amine-reactive BG building block, but the sample was taken before the reaction has started.

2) no BG coupling: no amine-reactive BG building block was added to the sample.

In the revised version we included the explanation in the figure legend.

4) Figure 4. The same data should not be plotted twice in panel B and C: Panel C should be removed (same comment for Figure 4F and G).

Indeed, the same data were plotted twice; as percentage of the correction efficiency and as fold change. In the revised version we removed the plots for the fold change.

The main question about the use of SadCas9 is to determine whether the difference with SpCas9 is statistically significant. The tests should be performed (Figure 4E, between RNPD coup and RNP-RNPD coup) and if not significant the conclusions should be revised.

Thank you for the suggestion. We performed statistical analysis, and indeed we found that the differences were not significant in all groups (see Author response image 2). Since the main goal of using the two-component RNP-RNPD system was to show in a second and independent manner that the correction efficiency is enhanced when the donor oligo is locally concentrated, we decided to show the results from the two-component system without direct comparison to the one component system.

Author response image 2

Download asset Open asset

A map showing the positions of the different SadCas9 guides used (1 to 4) should be shown.

In the revised manuscript the map showing the positions of SadCas9 sgNAs have been added (Figure 4C, Supplementary file 1). The sequences and the positions of SaCas9 guides are given in respect to the turboRFP fluorophore mutation and the SpCas9 guide cut site.

5) Two experiments are required to describe the effect observed: In order to know if the effect observed is due to coupling the oligo to Cas9, a control should be performed with the oligo coupled to BG but not to Cas9. Clearly at least part of the effect observed could be due to stabilization of the oligo rather than proximity to DSB per say. To distinguish these possibilities the authors should test an oligo coupled to SadCas9 but without the corresponding guide.

Please see our first response to reviewer #1.

I assume S. aureus was used such as to design a distinct guide specific for S. aureus and not bound by S. pyogenes, if so, this should be explained in the text.

The reviewer is correct, the S. aureus Cas9 was used in combination with S. pyogenes Cas9 in order to make sure that guides are not interchanged between the Cas9 that is cleaving and the catalytically inactive Cas9 that holds the template. We added this explanation in the revised version of the manuscript.

6) Figure 4—figure supplement 1.

What is the interpretation for the decrease of percentage of edited cells with oligo coupled to SadCas9 ? (Figure 4—figure supplement 1D).

Our hypothesis is that the constant presence of the donor oligo coupled to SadCas9 at the target locus could lead to interference of the ss-oligo with the DNA double-helix and disturb the cutting efficiency of the SpCas9.

Legend of Figure 4—figure supplement 1C and D: RNP-RNP unco is grey but should be hatched grey box.

Thank you for spotting this mistake. We corrected it in the revised manuscript.

7) Two-tailed tests should be used not one-tailed since there is no reason to assume a priori that the coupled protein would be more efficient.

We used the one-tailed t-test since we specifically want to test (the alternative hypothesis of) whether increasing the local oligo concentration at the DSB increased the correction efficiency (not whether the correction efficiency decreased). Notably, even when using the two-tailed t-test, the conclusions remain the same, namely that the difference in correction efficiency is statistically significant.

8) Subsection “Expression and purification of Cas9-SNAP”, missing word: "The was further purified…"

Thank you for pointing out this mistake. We corrected it in the revised manuscript.

Reviewer #2:

[…] Although this is interesting and aspects of the study are really clever (e.g. the use of the sa-dCAS9 linked template to prove that physical proximity is the key), the work strikes me as quite preliminary as the increases are only shown in this transient reporter system using only one target site.

In the revised manuscript we described our reporter system in more detail to avoid any misunderstandings (Figure 2A). The fluorescent reporter system was not transiently transfected, but generated by a single stable integration of the reporter cassette into the genome of HEK293T cells.

Without more thorough testing of the improvement in a real experimental system, I can't recommend acceptance of this study in its current form at eLife. My recommendations to improve the impact of this publication would be to demonstrate real efficiency improvements using a targeted edit within a genomic target site. In this respect, the existing work using transient reporter system would be excellent preliminary data for a real proof-of-concept – namely the increased efficiency in a real experimental genome engineering project e.g. to create a point mutation.

As stated above, the reporter system was generated by a single stable integration into the genome of HEK293T cells. We nevertheless fully agree with reviewer 2 that analysis of different endogenous loci is necessary to validate our findings (see also our second response to major issues). In the revised manuscript we therefore tested our system in three different cell types on five different genomic loci (Figure 5A-C, Figure 6D-F). Importantly these data fully support our hypothesis that coupling the donor oligo to the Cas9 complex increases repair rates.

In addition, the CRISPR field has been populated by numerous claims showing positive perturbations to shift the balance of repair pathway from NHEJ to HDR (alterations in experimental design, NHEJ inhibitors, etc.). Frequently these claims have been made on the basis of data from manipulation of a single target site. Reproducibility of these studies has proven to be low, as when applied by other labs at different target sites, the reported improvements are frequently not replicated. To avoid this occurring, I would recommend that the authors address several different genomic target sites and report the level of improvement in HDR seen in these various experimental settings.

As mentioned above, in the revised manuscript we reproduced our findings in different cell types and endogenous loci by NGS (Figure 5A-C, Figure 6D-F).

Figure 4E presents data which suggests that the positioning of the sgRNA (and hence the DSB) away from the target nucleotide to be mutated (0 bp, 52 bp, 61 bp, 83 bp and 128bp) has little influence on the correction efficiency. This is significantly at odds with experimental data from genomic target sites, and should be discussed. In addition, I wonder whether this is a curious artefact of using transient plasmids and thus provides support for my recommendation of addressing a single copy genomic target site or sites to validate their approach in a real experimental setting.

Here has been a misunderstanding, and we therefore described the experiment in more detail in the revised version (Figure 4C). The DSB is always generated by the same SpCas9 sgRNA at position 0 from the mutation. 0 bp, 55 bp, 78 bp, 86 bp and 145bp indicates the distances of the different SadCas9 sgRNAs binding sites from the mutation in the RFP fluorophore (position 0).

The work is very similar to a recent bioRxiv report by Janet Rossant's group (https://doi.org/10.1101/204339) who demonstrated that covalent attachment of the repair template to the CRISPR/Cas9 complex enhanced HDR rates, reporting data from fluorescent cassette insertions at 5 genomic target sites. Although this paper hasn't been peer reviewed, it's tempting to speculate that the authors of this manuscript have submitted what is essentially a preliminary study to eLife to compete with this more thorough study, which is presumably in review at another journal.

In the mentioned report the authors use a Cas9-streptavidin fusion in combination with a biotinylated repair template to recruit the repair template to the editing site. This improves their targeting efficiencies. Since this is a non-covalent attachment system, it has certain disadvantages to the SNAP-tag system, and thus we consider our work complementary to their work. Of note, in our revised manuscript we also analysed several genomic loci in different cell types.

Reviewer #3:

[…] The main problem is the fact that the authors used only on system to monitor HR; and that the designing of this substrate is based on wrong considerations on HR and NHEJ.

To clarify, using our reporter system we were able to detect Cas9-induced frame-shift indels (loss of eGFP fluorescence) and corrections from donor templates (activation of RFP). As explained in our third response to major issues, we fully agree with reviewer 3 that different repair pathways could be responsible for indel formation and correction from oligo templates, and we therefore rephrased the text in the revised manuscript accordingly. In the revised manuscript we also used NGS to analyse the repair outcomes in a more direct and precise manner. Of note, our main goal in this study is the enhance precise repair from donor oligos (potentially for future clinical application of CRISPR/Cas), and we are aware that we cannot draw conclusions about the activity of specific molecular repair pathways. Since ss-oligos are more efficient for HDR correction of DSBs than ds-DNA, we also solely focused on ss donor oligos in our study.

It is claimed that NHEJ is error-prone. This is wrong. There are two kinds of end-joining processes, the canonical NHEJ (which is not error-prone, but conservative) and the Alternative end-joining (alt-NHEJ, MMEJ, B-NHEJ), which is mutagenic and error-prone. Therefore mutagenic repair does not automatically imply NHEJ.

We fully agree, and want to thank the reviewer for pointing out this issue. In the revised manuscript we rephrased the text accordingly (see Introduction).

Second it is said in the Introduction that HR and NHEJ directly compete. This is also wrong; in fact things happen in two phases: first competition between cNHEJ and single-strand DNA resection, second on resected DNA extremities, competition between HR and alternative end-joining. Finally, HR can also generate mutagenesis. Therefore many concerns exist on the strategy used here (which is the sole assay used) because mutagenic repair can arise by many other processes than NHEJ.

We fully agree with reviewer 3; it is not a direct competition between HDR and NHEJ pathways. We were referring to an indirect effect: Indel mutations disable the CRISPR-Cas9 complex from re-targeting the same locus, and thus precise repair from the donor oligo is no longer possible. We rephrased the text in the revised manuscript.

Moreover, is correction with oligonucleotides (65, 81 b) an actual HR mechanism? This is not consistent with concept of MEPS.

Again we agree with reviewer 3. In the revised manuscript we define homology-directed repair (HDR) as any repair pathway that uses a donor oligo as a template for the repair. Interestingly a recent study suggests that ss-donor oligos are predominantly used by the Fanconi Anemia (FA) repair pathway (doi: https://doi.org/10.1101/136028) rather than the classical homologous recombination (HR) pathway.

The authors should first genetically validate their reporter system, in cells mutated for HR or NHEJ.

We initially validated the reporter system by quantifying the increase in correction efficiencies between the classical Cas9 system and the Cas9-Geminin system. Cas9-Geminin expression is limited to the S/G2 phase of the cell cycle, and therefore HDR rates are enhanced (doi: 10.1016/j.celrep.2016.01.019). Importantly, when using plasmid donors as well as ss-oligo donors we could reproduce these results (see Author response image 3), validating that our reporter is able to pick up increases in correction rates.

Author response image 3

Download asset Open asset

The authors should also verify their strategy with natural endogenous target sequences, instead of the reporter.

In the revised manuscript we targeted 5 endogenous loci (Figure 5A-C, Figure 6D=F) (see also our second response to major points).

There are no data on the transfection efficiency. Especially, comparing the CAS9 with the engineered one.

Similarly, does the two CAS9 cleave with similar efficiency?

Images of an experiment where a SpCas9-mCherry fusion protein was transfected using lipofectamine 2000 is shown in Author response image 4 (the upper panel). In the lower panel, we included the comparison of the cleavage efficiencies between classical SpCas9 WT and the engineered Cas9-SNAP. Furthermore, in the revised version of the manuscript the comparison can be seen in Figure 6C and Figure 4—figure supplement 2B.

Author response image 4

Download asset Open asset

Does the modification of the CAS9 affect its cleavage specificity?

Are there any off-target effects? (off-target cleavage, off-target integration of the oligo).

This is indeed a very important question. In the revised manuscript we compared DNA editing rates at the top 3 predicted off-target loci between the classical Cas9 WT (with the uncoupled oligo) and our engineered Cas9-SNAP with the coupled oligo (Figure 6C). Importantly, no increase in off-target editing was observed.

The authors should test different RNA guides for a common target.

We have initially tested 7 different guide RNAs that bind in the mutated RFP locus (using the system from Mashiko et al., 2013) and from there we pre-selected 2 guides that were the most efficient (and were binding over the mutRFP fluorophore). Next, we tested those 2 guides for cleavage in the stably integrated reporter system (see FACS plots in Author response image 5) and selected sgRNA^SpCas9(mutRFP)-1, which was 1.3 fold more efficient for all further experiments. Additionally, sgRNA^SpCas9(mutRFP)-1 was the only sgRNA that allowed introduction of the DSB directly in the fluorophore of mutated RFP. This is important for two reasons: (1) positioning the DSB away from the targeted nucleotide (fluorophore of mutated RFP) would reduce the correction efficiencies; (2) introducing the mutation in the protospacer sequence of the guide prevents the guide from reintroducing the DSB.

Author response image 5

Download asset Open asset

The author should test different cell lines.

We fully agree with reviewer 3 that these experiments would allow to validate the robustness or our system. As stated above, in the revised manuscript we analysed three different cell types (HEK cells, K562 cells, mouse ES cells) (Figure 5A-C, Figure 6D-F).

It is not clear what is actually measured. Where are the mutagenic repair (pseudo-NHEJ) measurements? Is it the frequency of HDR or the ratio HDR/pseudo-NHEJ?

We thank the reviewer for pointing that the definition was not clear enough. In the revised manuscript we included the more detailed explanation (Materials and methods; FACS analysis). The correction efficiency (percentage of corrections in edited cells) = 100 * (turboRFP positive population / (eGFP negative population + turboRFP positive population)).

[Editors’ note: the author responses to the re-review follow.]

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

The authors provide convincing answers to all reviewers' comments. Importantly, the authors show that the main improvement achieved by their experimental approach is due to enhanced targeting of the oligonucleotides to the nucleus rather that its targeting to the specific target genomic site. This could also be highlighted in the Abstract.

In the revised version of the manuscript we have now added the information that increased correction is due to targeting the repair template to the nucleus in the Abstract.

A few points need to be clarified:

1) Source data 6 refers to Figure 4—figure supplement 2 (not 3?) should be registered as Figure 4—source data 3.

2) Source data 5 refers to Figure 4—figure supplement 1 (not 2?) should be registered as Figure 4—source data 2.

Thank you for suggesting these clarifications. We have now renamed the figure supplements and source data files in order to be consistent.

3) Figure 2B requires clarification (and in the legend as well). Legend says "the mutation substitutes… CT.. to TA". This is ambiguous because TA is wild type. It would clarify to draw in Figure 2, the wild type and mutated genomic sequences (with the codon substitution) and the sequence of the guide (also can be confusing as drawn because the guide has CT not TA).

In the revised manuscript we now illustrated the RFP reporter sequence before and after the correction (see Figure 2A, B). In addition we also marked the nucleotides and amino acids that are exchanged upon repair.

4) Figure 5: In Figure 5, only the reference is shown. Mutated variant should be indicated. For instance, on 5A, at HBB, the DNA sequence shown should be the one after correction (to be consistent with Figure 2B) ? However it is identical to guide sequence?

Thank you for this important suggestion. As for the reporter system, we have now drawn the schemes for all other endogenous sequences before and after the correction (see Figure 5A, B, C and 6D, E, F).

Please explain the correction efficiency values, obtained after correction for transfection efficiency? If so, provide transfection efficiencies, otherwise read data from Supplementary file 2 cannot be understood: i.e. 27% at Rosa26, how was this obtained?

Thank you for spotting that we forgot to precisely explain how we calculated the correction efficiency values from NGS data. We have now included the formula for this calculation in the Materials and methods section: Correction efficiency (%) = 100* Perfectly corrected/Edited reads.

Using this formula and the values from the Supplementary file 4 (categorized variant count tables), one can calculate the correction efficiencies at the Rosa26 locus: 11.78%, 10.16%, and 7.8% for the uncoupled control and 19.29%, 20.51% and 15.86% for our coupled system. This on average gives a correction efficiency of 9.9% and 18.6%, respectively (presented in the Figure 5B and Figure 5—source data 1).

Of note, the correction efficiency represents the ratio between precise correction and total edits, and therefore does not change with different transfection rates (if we would correct for transfection efficiency in the formula, we would correct the denominator and the numerator by the same value, and thus the result would remain the same).

In addition, since no HR reads is detected at Rosa26 (why is this?), the authors should comment in the main text on the difference uncoupled/coupled from Figure 5B for this locus.

We are very thankful for spotting this mistake in the allele plot for the Rosa26 locus; the corrected sequence (SNV:-2C,-1T) was by mistake plotted close to the bottom of the graph. In the revised version we placed the corrected allele in the 3^rd row, as described in the figure legend and as shown in the other plots.

By having a close look at the count table for the Rosa26 locus, we also realized that we made an error when categorizing alleles in the count table (some indels were counted as reference – Supplementary file 4). In the revised version we fixed this error, and adapted the values accordingly. Importantly, these changes did not substantially affect our results, and did not change any of the conclusions of the manuscript.

5) Figure 6A requires clarifications both in the main text and in the presentation of the figure. The authors summarize in one sentence all the NGS data (Results and Discussion, eighth paragraph). There is a lot of data in this analysis and it would be better to present the analysis step by step and not to refer to all panels 6A, B, D, E, F at once. Reporter locus could be presented first, with comments about percentage of corrected reads and other events (legend should explain the nomenclature, for instance -7:7D, and also what are I, II and III: triplicates I assume). Then results at other loci should be briefly discussed (and referred to Supplementary file 2).

Thank you very much for this important comment. As suggested, in the revised manuscript we discuss the data shown in Figure 6 in more detail in the main text (Results and Discussion, eighth paragraph), and refer to Supplementary file 2. In addition, in the figure legends we now explain the nomenclature for SNVs and indels, and indicate that the different colors in the x-axis labels represent the three different replicates.

Also indicate in Supplementary file 2 that the variants called "SNV" are the ones predicted after repair by HR.

In the revised version we always specify the precisely corrected variant in the allele plots.

An interesting information is also the relative proportion of HR versus non HR events. In Figure 6, it seems that this proportion is about 30% in experiment I (5.63/(100-85.19)). Is this correct? It could be discussed in the main text.

In the revised version we explain the editing results of the experiment shown in Figure 6A in more detail: We found that while the mean percentage of corrected loci increased from 0.8% with the classical Cas9 system to 4.9% with the RNPD system, the number of incorrectly edited loci slightly decreased from 12.6% to 9.3%, respectively (Figure 6A, Supplementary files 2, 3, 4). This corresponds to a 7-fold increase in correction efficiency from 5.3% to 36.4% (see Figure 6B).

Author details

1. Natasa Savic

   The Institute of Molecular Health Sciences, ETH Zurich, Zurich, Switzerland

   Contribution

   Conceptualization, Data curation, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing, Conceived and designed the study, performed the experiments, Wrote the paper

   Contributed equally with

   Femke CAS Ringnalda

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-3110-5780

2. Femke CAS Ringnalda

   The Institute of Molecular Health Sciences, ETH Zurich, Zurich, Switzerland

   Contribution

   Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing—original draft, Writing—review and editing, Performed the experiments, Wrote the paper

   Contributed equally with

   Natasa Savic

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-0684-4613

3. Helen Lindsay

   1. The Institute of Molecular Life Sciences, University of Zurich, Zurich, Switzerland
   2. SIB Swiss Institute of Bioinformatics, Zurich, Switzerland

   Contribution

   Data curation, Software, Formal analysis, Investigation, Visualization, Writing—review and editing

   Competing interests

   No competing interests declared

4. Christian Berk

   Institute for Pharmaceutical Sciences, ETH Zurich, Zurich, Switzerland

   Contribution

   Investigation, Writing—review and editing, Performed HPLC and MS analysis

   For correspondence

   christian.berk@pharma.ethz.ch

   Competing interests

   No competing interests declared

5. Katja Bargsten

   Department of Biochemistry, University of Zurich, Zurich, Switzerland

   Contribution

   Investigation, Methodology, Writing—review and editing, Produced the recombinant proteins

   For correspondence

   k.bargsten@bioc.uzh.ch

   Competing interests

   No competing interests declared

6. Yizhou Li

   Institute for Pharmaceutical Sciences, ETH Zurich, Zurich, Switzerland

   Contribution

   Investigation, Performed HPLC and MS analysis

   For correspondence

   yizhou.li@pharma.ethz.ch

   Competing interests

   No competing interests declared

7. Dario Neri

   Institute for Pharmaceutical Sciences, ETH Zurich, Zurich, Switzerland

   Contribution

   Methodology, Provided technical advice

   For correspondence

   dario.neri@pharma.ethz.ch

   Competing interests

   No competing interests declared

8. Mark D Robinson

   1. The Institute of Molecular Life Sciences, University of Zurich, Zurich, Switzerland
   2. SIB Swiss Institute of Bioinformatics, Zurich, Switzerland

   Contribution

   Software, Supervision, Writing—review and editing

   For correspondence

   mark.robinson@imls.uzh.ch

   Competing interests

   No competing interests declared

9. Constance Ciaudo

   The Institute of Molecular Health Sciences, ETH Zurich, Zurich, Switzerland

   Contribution

   Investigation, Writing—review and editing

   For correspondence

   cciaudo@ethz.ch

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-0857-4506

10. Jonathan Hall

    Institute for Pharmaceutical Sciences, ETH Zurich, Zurich, Switzerland

    Contribution

    Methodology, Writing—review and editing, Provided technical advice

    For correspondence

    jonathan.hall@pharma.ethz.ch

    Competing interests

    No competing interests declared

    "This ORCID iD identifies the author of this article:" 0000-0003-4160-7135

11. Martin Jinek

    Department of Biochemistry, University of Zurich, Zurich, Switzerland

    Contribution

    Supervision, Methodology, Writing—review and editing, Provided technical advice

    For correspondence

    jinek@bioc.uzh.ch

    Competing interests

    No competing interests declared

    "This ORCID iD identifies the author of this article:" 0000-0002-7601-210X

12. Gerald Schwank

    The Institute of Molecular Health Sciences, ETH Zurich, Zurich, Switzerland

    Contribution

    Conceptualization, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing, Conceived and designed the study, Supervised the experiments, Wrote the paper

    For correspondence

    schwankg@ethz.ch

    Competing interests

    No competing interests declared

    "This ORCID iD identifies the author of this article:" 0000-0003-0767-2953

• 12,391

  Page views

• 1,998

  Downloads

• 96

  Citations

Article citation count generated by polling the highest count across the following sources: Scopus, Crossref, PubMed Central.