Rapid and precise engineering of the Caenorhabditis elegans genome with lethal mutation co-conversion and inactivation of NHEJ repair.

As in other organisms, CRISPR/Cas9 methods provide a powerful approach for genome editing in the nematode Caenorhabditis elegans. Oligonucleotides are excellent repair templates for introducing substitutions and short insertions, as they are cost effective, require no cloning, and appear in other organisms to target changes by homologous recombination at DNA double-strand breaks (DSBs). Here, I describe a methodology in C. elegans to efficiently knock in epitope tags in 8-9 days, using a temperature-sensitive lethal mutation in the pha-1 gene as a co-conversion marker. I demonstrate that 60mer oligos with 29 bp of homology drive efficient knock-in of point mutations, and that disabling nonhomologous end joining by RNAi inactivation of the cku-80 gene significantly improves knock-in efficiency. Homology arms of 35-80 bp are sufficient for efficient editing and DSBs up to 54 bp away from the insertion site produced knock-ins. These findings will likely be applicable for a range of genome editing approaches in C. elegans, which will improve editing efficiency and minimize screening efforts.

SEQUENCE-SPECIFIC nucleases are a critical tool for manipulation of DNA sequences. The bacterial type II clustered regularly interspaced short palindromic repeats (CRISPR) system, which normally protects against viral DNA and provides a memory of exposure (Jinek ), has recently revolutionized genome editing in multiple organisms (Cong ; DiCarlo ; Gratz ; Hwang ; Li ; Ran ; Gratz ; Nakanishi ). For genome editing, the system has been simplified to two components: the Cas9 nuclease, which generates DNA double-strand breaks (DSBs), and a chimeric small guide RNA (sgRNA) that fills the function of two small RNAs in the native bacterial system (Cong ). Specific genomic sequences are targeted by the 5′-most 15–20 bp of the sgRNA through the formation of an RNA:DNA hybrid (Jinek ; Mali ). An NGG motif (protospacer adjacent motif, PAM) must immediately follow the target sequence in the genome (Jinek ; Ran ). This PAM directs Cas9 to cleave the DNA 3 bp 5′ to the PAM (Jinek ). Depending on the desired experimental outcome, one can select for error-prone repair by pathways such as nonhomologous end joining (NHEJ) to generate insertion–deletion (indel) mutations, or homologous recombination to knock in specific sequences.

Initial genome editing methods in Caenorhabditis elegans harnessed excision of a Tc or Mos transposon to generate a DSB, and a plasmid repair template to knock in (Plasterk and Groenen 1992; Robert and Bessereau 2007; Frøkjaer-Jensen ; Frøkjær-Jensen ), or delete (Frøkjær-Jensen ) desired sequences through homologous recombination. These methods are robust, but the relative rarity of the editing event requires use of a selectable marker, such as rescue or antibiotic resistance (Frøkjaer-Jensen ; Giordano-Santini ), and a transposon site is ideally needed within 1–2 kb of the desired edit (Robert and Bessereau 2007). Zinc finger and transcription activator-like effector nucleases (Wood ) and CRISPR/Cas9 have allowed for similar efficient editing without the constraint of transposon insertions. In particular, the ease and rapidity of generating new sgRNAs for the CRISPR/Cas9 system means that transgenic strains can be created precisely and rapidly and any endogenous NGG sequence can theoretically be targeted. Several CRISPR/Cas9 systems have been described, each with individual strengths and weaknesses (Waaijers and Boxem 2014). Cas9 can be delivered by micro-injection of in vitro transcribed mRNA (Chiu ; Lo ), pure protein (Cho ), or plasmid DNA (Chen ; Dickinson ; Friedland ; Katic and Grosshans 2013; Waaijers ). Similarly, the sgRNAs can be introduced by in vitro transcription, which does not require polyA tailing or 5′ methyl cap addition (Chiu ; Cho ; Lo ), or driven by RNA polymerase III promoters such as U6 (Chen ; Dickinson ; Friedland ; Katic and Grosshans 2013; Waaijers ) or (Chiu ). Most groups use the chimeric sgRNA, though a previous report described higher in vitro nuclease activity using the two separate bacterial small RNAs (Lo ). Multiple groups have developed protocols for both knockouts and knock-ins. Knock-ins have been primarily generated through efficient selection schemes based on the earlier MosI-mediated single-copy transgene insertion methods using genetic markers such as (Dickinson ), drug resistance markers (Chen ), or fluorescence (Tzur ). Typically, plasmid repair templates with 1 kb or more of homology flanking the insert have been used (Chen ; Dickinson ; Tzur ; Kim ).

Recently, several reports have described methods to introduce single-basepair changes, small epitopes, and larger tags such as GFP without the need for selectable markers; these approaches either directly screened all F1 progeny from co-injection marker positive animals (Paix ) or employed a co-CRISPR/co-conversion approach where selection for one editing event resulted in an enrichment for edits at unrelated loci (Arribere ; Kim ). Direct screening of F1’s allows editing without introduction of additional mutations, but is more labor intensive, while co-CRISPR/co-conversion allows for identification of editing events while minimizing hands-on screening, but requires outcrossing or meiotic segregation of the marker allele. Co-CRISPR selects for mutation in the unc-22 gene, with mutant homozygotes identified in the F2 progeny of co-injection marker positive animals, or less frequently in the F1 progeny (Kim ), though haploinsufficiency of the locus in 1% nicotine should allow for identification of mutant heterozygotes in F1 animals (Moerman and Baillie 1979). Co-conversion selects for knock-in of dominant alleles in the , , or genes in the F1 progeny of injected animals (Arribere ).

The goal of this study was to test whether repair of a temperature-sensitive lethal point mutation could be used as an alternate co-conversion marker, as such an approach could in theory provide robust selection, minimal screening, and no requirement for outcrossing or meiotic segregation of marker alleles/mutations. I focused on using single-stranded oligonucleotides (oligos) as a template as they have been successfully used in a range of model organisms (Igoucheva ; Storici ; Chen ; Bedell ; DiCarlo ), are cost effective, and require no cloning. In C. elegans, oligonucleotides have been used to introduce single-base changes (Arribere ; Zhao ), inactivate genes by introducing premature stop codons or deleting sequences (Lo ; Paix ), or insert small protein epitopes (Lo ; Paix ). Introduction of epitopes seamlessly into endogenous loci has numerous experimental uses: chromatin immunoprecipitation, purification of protein complexes followed by mass spectrometry, and detection of proteins by immunofluorescence or immunoblotting. Furthermore, optimizing insertion of epitope tags using oligonucleotide templates is almost certain to be applicable to single-base editing. Here, I describe a robust, cost effective, widely applicable method using co-conversion and inactivation of NHEJ repair to rapidly and precisely engineer the C. elegans genome using oligo-templated repair with the CRISPR/Cas9 system.

Materials and Methods

Genetics

The following strains were used in this study: N2(WT), GE24 III, RB873 III, which were provided by the Caenorhabditis Genetics Center. N2 animals and mutants were propagated at 20°, while mutants were propagated at 15°. Animals were maintained on nematode growth medium seeded with Escherichia coli OP50 (Brenner 1974). For later injection experiments, mutants were maintained on HB101. Growth conditions for each experiment are indicated in figure and table legends. Strains generated for this study are listed in Supporting Information, File S1, Table S1.

Microinjection

Mixtures of plasmids and oligos were microinjected into the gonad of young adult animals. Plasmids were purified using a Qiagen midiprep kit. Oligos were resuspended in TE buffer and working stocks of 2 µg/µl were made with nuclease-free dH20. Repair template oligos are listed in Table S2. For PCR-generated PU6::sgRNA templates, PCR products were purified from 100-µl reactions and concentrated 10-fold using a DNA Clean and Concentrator kit (Zymo Research, no. D4004).

MfeI deletion assay

Wild-type (WT) animals were microinjected with 50 ng/µl of pJW1138 or pJW1236 [ targeting CRISPR/Cas9 plasmids with original and flipped plus extended (F+E) sgRNAs, respectively], 10 ng/µl myo-2::tdTomato co-injection marker, and 40 ng/µl of pBluescript DNA. Marker positive F1 animals were picked into 30 µl of M9 + gelatin in a 96 well. Concentrated OP50 food (30 µl; see the “PCR-based knock-in screening” protocol (File S1) in the for recipe) was then added and F1’s were incubated for 3–4 days at 25° to allow progeny to develop. Worms were lysed and genotyped as described in the PCR-based knock-in screening section (see Supporting Information). To each 10 µl PCR, 5 µl of MfeI-HF (NEB, R3589L) digestion mixture (1 µl 10× CutSmart buffer, 3.5 µl dH20, 0.5 µl MfeI-HF) was added. The reaction was mixed, incubated at 37° for 1 hr, and then resolved on a 1.5% TAE-agarose gel.

Temperature-sensitive pha-1 co-conversion screening

For temperature-sensitive [] co-conversion experiments, animals were microinjected with 60 ng/µl of pJW1285 ( targeting) CRISPR/Cas9 plasmid, 60 ng/µl of either pJW1285 ( PAM no. 1 targeting), or pJW1268 ( PAM no. 2 targeting) Cas9 plasmid and 50 ng/µl of the appropriate repair oligos. In experiments where a co-injection marker was included, myo-2::tdTomato was used at 10 ng/µl. PCR-derived PU6::sgRNA templates were injected at a concentration of 25 ng/µl along with 50 ng/µl of pJW1285 ( targeting CRISPR/Cas9) and 50 ng/µl of the appropriate repair oligos. Injected adults grown at the permissive temperature (15°) were singled into wells of a 24-well plate containing NGM-lite agar and seeded with OP50, shifted to 25°, and incubated for 3–4 days. Rescued F1’s (L3s to adults) were the only animals other than the P0 animals observed in wells. These F1 rescues were singled onto individual plates and incubated for two days to allow progeny development. The parental F1 animals were then genotyped by restriction digestion, as described below, to identify animals with a 2×FLAG insertion. The remaining PCR product was purified using a DNA Clean and Concentrator kit (Zymo Research) and then sequenced using knock-in specific primers (Table S3). To recover homozygotes for sequence-verified knock-ins, 12–24 F2 progeny were singled onto individual plates, incubated at 25° for 2 days, and the parental F2 was genotyped to confirm the 2×FLAG insertion. These progeny were also genotyped for repair by PCR and CEL-1 digestion followed by sequencing of candidate repair homozygotes.

Genotyping PCRs and restriction digestion

Single F1 animals were picked into 10 µl of single-worm lysis buffer [10 mM Tris (pH 8.3), 50 mM KCl, 2.5 mM MgCl2, 0.45% IGEPAL, 0.45% Tween-20] containing 1 mg/ml proteinase K. Tubes were incubated on dry ice for 15 min, 62° for 1 hr, and then heated to 95° for 20 min to inactivate the proteinase K. Genotyping PCRs were performed using Phusion polymerase (NEB, no. M0530S) and High Fidelity buffer; 1/10 volume of lysate was used as template in the PCR. For identification of knock-in events in the F1, 30-µl PCRs were performed. Five microliters of this PCR was removed and BamHI digested by adding 10 µl of digestion mix per PCR (1 µl 10× NEBuffer 3, 0.5 µl BamHI, and 8.5 µl dH20) and digesting for 1 hr at 37° before resolving on a 1.5% TAE-agarose gel. CEL-1 was purified from four heads of nonorganic celery (Safeway) with the celery juice extracted with a BJE510XL 900W juicer (Breville). Purification was performed as described (Yang ) to the end of dialysis in step I, as described by Lo . For CEL-1 digestions, 10 µl of enzyme mix (2 µl CEL-1 + 2 µl 5xPhusion High Fidelity buffer + 6 µl dH20) was added to the 5-µl PCR and incubated at 42° for 1 hr prior to resolving on a 1.5% TAE-agarose gel, as described above. Oligos used for genotyping are listed in Table S3.

Vector generation

pJW1138 ( targeting CRISPR/Cas9) and pJW1185 ( targeting CRISPR/Cas9) were derived from pDD162 using a Q5 Mutagenesis kit (NEB, no. E0554S) as previously described (Dickinson ). sgRNA(F+E) was synthesized as a gene fragment (IDT gBlock; sequence in Table S4) with a Y61A9LA.1 targeting sequence (Friedland ) and introduced into a pDD162-derived vector by Gibson cloning (NEB, no. E5510S) to generate pJW1219. pJW1236 ( targeting CRISPR/Cas9), pJW1254 ( PAM no. 1 targeting CRISPR/Cas9), pJW1268 ( PAM no. 2 targeting CRISPR/Cas9) and pJW1285 ( targeting CRISPR/Cas9) were derived from pJW1219 [CRISPR/Cas9 with sgRNA(F+E)] through Q5 mutagenesis. The , , and PAMs were manually chosen by searching for an NGG sequence in either strand close to the desired insertion site; these sgRNA target sites were then checked for specificity using the http://crispr.mit.edu website. All target sites scored >90 with no off-target sites in genes. The PU6::sgRNA template sequence was deleted from the pJW1219 vector using Q5 mutagenesis to generate pJW1259. All plasmids (standard vector propagation, and those generated by Gibson assembly, Q5 site-directed mutagenesis, or TOPO-blunt cloning) were transformed into PEG/DMSO DH5 alpha competent cells (protocol in File S1) made in house. pJW1219, pJW1259, pJW1285, pJW1310, and pJW1311 are available through AddGene.

Generation of U6 promoter::sgRNA templates by PCR

The U6 promoter and chimeric sgRNA(F+E) template were amplified from pJW1219 and cloned into the pCR-Blunt II-TOPO vector (Invitrogen, no. K2800-20) to generate pJW1310 and pJW1311, respectively. The U6 promoter was then PCR amplified from pJW1310 with oligos 1787 and 1788, while the sgRNA template was amplified from pJW1311 with oligo 1790 and a target-specific 60mer that contained 20 bp of homology to the U6 promoter, 20 bp of new sgRNA target sequence, and 20 bp of homology to the sgRNA template (Table S3). New sgRNA template primers can be made by replacing the N20 in the following sequence with 20 bp of target specific sequence: 5′-cctcctattgcgagatgtcttg(N20)gtttaagagctatgctgg-3′. The U6 promoter and sgRNA template PCRs were mixed (0.5 µl each per 100 µl PCR reaction) and amplified using the external primers (1787 and 1790; Table S3). The cycling parameters were: (i) 98° denaturation; (ii) 35 cycles of 98° for 10 sec, 61° for 30 sec, 72° for 20 sec; and (iii) 72° for 1-min final extension. To generate more PU6::sgRNA templates, the fused product was used in a 100-µl nested reaction with primers 1793 and 1794 (Table S3).

RNAi

Feeding RNAi was performed as described (Kamath ; Ward ). Four gravid adults were placed on 6-cm plates freshly seeded with HT115 bacteria expressing control or dsRNA, obtained from the Ahringer library (Kamath ). Adults were ready to inject on the RNAi plates 4–6 days later. For each experiment mutants were also put on plates seeded with bacteria expressing dsRNA. RNAi efficacy was confirmed by observing molting defects, protruding vulvae, abnormal germlines, and sterility arising from inactivation (Asahina ; Gissendanner and Sluder 2000; Brooks ; Chen ).

Immunoblotting

Lysates were generated as described in File S1 and resolved on Mini-PROTEAN TGX stain-free 4–15% gradient gels (Bio-Rad, no. 456-8086). Stain-free gels (Bio-Rad) contain a compound evenly distributed in the precast acrylamide gel that reacts with tryptophan following UV exposure and gives a strong fluorescent signal that can be used to stain for total-protein levels in acrylamide gels, monitor transfer in immunoblotting, and serve as a loading control. Following resolution, the stain-free compound was activated (Posch ) and proteins were transferred to an Immobilon FL PVDF membrane (Millipore, no. IPFL00010) at 100 V for 60 min. Total protein, pre- and post-transfer, was monitored using the stain-free fluorophore as described (Posch ). Stain-free imaging of total protein was used to confirm equal loading. The blots were sequentially probed with anti-FLAG (1:1000) (M2 clone, Sigma, no. F1504), and sheep antimouse–HRP conjugate (1:5000) (GE Healthcare, no. RPN4201). WesternBright Sirius HRP substrate (Advansta, no. K-12043-C20) and ECL Prime (GE Healthcare, no. RPN2232) were used to develop the blots in Figures 2 and 5, respectively. All blots were imaged using a ChemiDoc MP imaging system (Bio-Rad).

Figure 2

Detection of NHR-23::2×FLAG in precise knock-ins. Four micrograms of protein from synchronized gravid adults of the indicated strains was analyzed by immunoblotting with anti-FLAG. KRY48 contains a frameshift in the 2×FLAG tag and thus does not express the epitope. Stain-free (Bio-Rad) analysis of total protein on the blot is provided as a loading control. Marker size (in kilodaltons) is provided.

Figure 5

Detection of FLAG epitope expression in nhr-23-, nhr-25-, and smo-1-tagged lines. Anti-FLAG immunoblot analyses of lysates are from mixed stage animals of the indicated genotypes. The 2x and 3xFLAG tagged nhr-23 (A) and nhr-25 (B) lines and a 2×FLAG::smo-1 tagged line (C) were assayed. Stain-free (Bio-Rad) analysis of total protein on each blot is provided as a loading control. Marker size (in kilodaltons) is provided. The same exposure time was used to image all anti-FLAG blots. For the NHR-23 blot (A), the background band observed in Figure 2 was not detected, likely because a more potent ECL substrate was used for that experiment.

Accession codes

The accession codes are as follows: pJW1219 [Cas9-sgRNA(F+E) targeting site in Y61A9LA.1, AddGene plasmid 61250]; pJW1259 (Cas9 plasmid with sgRNA deleted, AddGene plasmid 61251); pJW1285 [Cas9-sgRNA(F+E) targeting site in , AddGene plasmid 61252]; pJW1310 (U6 promoter template vector; AddGene plasmid 61253); and pJW1311 [sgRNA(F+E) template vector, AddGene plasmid 61254].

Additional methods are described in Supporting Information.

Results and Discussion

sgRNA(F+E) displays increased activity relative to the original sgRNA in deleting an MfeI restriction site in the klp-12 gene

Initial reports describing use of oligos to introduce single-base changes reported edits in 0.7–3.5% of the F1 progeny screened (Zhao ). Recent work in mammalian cells reported that a modified sgRNA(F+E), with an extended Cas9 binding structure and removal of a potential PolIII terminator by an A-U basepair flip, exhibits improved activity (Chen ). I therefore tested whether this modified sgRNA(F+E) displayed increased activity in C. elegans, as a potential tool to improve editing efficiency. The modified sgRNA was introduced into the pDD162 CRISPR/Cas9 plasmid (Dickinson ), and to evaluate sgRNA activity, deletion of an MfeI restriction site in the gene (Friedland ) was used as a readout (Figure S1). sgRNA(F+E) produced a significant increase in deletion of the MfeI site in animals positive for the co-injection marker (Table 1), and was used for all subsequent experiments.

Table 1

sgRNA(F+E) has increased activity relative to the original sgRNA in deleting an MfeI restriction site in the klp-12 gene

Genotype	Guide RNA	% deletiona (±95% CI)	N
WT	Original sgRNA	54 (±2.73)	157b
WT	sgRNA(F+E)	83 (±0.31)*	85c

N = number of co-injection marker positive F1 progeny screened.

(Number of animals with loss of MfeI restriction site/total number of co-injection marker positive animals screened) × 100; CI, confidence interval; 95% confidence interval, Zσ/√n = 1.96σ/√n. *Two-tailed t-test comparing original sgRNA to sgRNA(F+E) P = 0.043.

From four independent injections.

From two independent injections.

pha-1(ts) co-conversion increases knock-in efficiency with minimal handling

Having increased sgRNA activity, I next turned to improving the screening process. Two recent reports describe that selection for a visible phenotype produced by a CRISPR/Cas9 triggered editing event [i.e., mutation or knock-in] results in an increase in knock-out and knock-in efficiencies at other genomic loci (Arribere ; Kim ). These approaches can be used in any genetic background, but require outcrossing or meiotic segregation of the selective mutation. My aim was to develop a stringent system with minimal handling and no outcrossing or meiotic segregation of selection markers required. is a temperature-sensitive embryonic lethal mutation; mutants are viable at 15°, but display embryonic and early larval lethality when cultivated at 25° (Schnabel ; Granato ). has been previously used to select and propagate extrachromosomal arrays carrying a marker (Granato ). mutants display a low spontaneous reversion frequency of 2.5 × 10−5 per haploid genome (Schnabel ); the stringency of this selection was confirmed by plating 144 gravid adults and shifting them to 25°; no viable progeny were produced, only dead eggs and arrested larvae.

I selected a CRISPR/Cas9 target site that would produce a DSB 17 bp 3′ from the point mutation and designed an 80mer repair oligo with the PAM silently mutated to prevent recleavage of the site in an edited animal (Figure 1A). Twelve WT animals were injected with the CRISPR/Cas9 plasmid, a co-injection marker, and a commercially purchased, PAGE purified repair 80mer. Eight rescued F1’s were recovered, all heterozygotes for repair (data not shown). Although all eight F1’s were positive for the co-injection marker, in later experiments, I also observed rescue in marker-negative animals. Thus, as previously observed (Arribere ; Zhao ), heritable transgenesis is not a prerequisite for efficient editing. Oligo-mediated repair of allowed recovery of repair heterozygotes, and the complete penetrance of the embryonic lethality meant that only F1 heterozygotes (i.e., rescued animals) would develop, making screening extremely rapid. A homozygote for repair had no significant difference in brood size compared to a WT control (Table S5; two-tailed t-test, P = 0.75).

Figure 1

Selection for pha-1(ts) oligo-mediated repair enriches for nhr-23::2×FLAG knock-in. (A) Sequence of the pha-1(e2123) genomic locus targeted, with the PAM (red text), e2123 G-to-A mutation (boldface text and underline), sgRNA target sequence, and position of the DSB indicated. The 80mer repair oligo also contains a silent C-to-A mutation to inactivate the PAM. (B) Sequence of the nhr-23 genomic locus targeted. The stop codon (blue text), PAM (no. 1, red text) and sgRNA target, the DSB position, and an alternate PAM (no. 2, red text) are indicated. The 200mer repair oligo was designed to insert a 2×FLAG epitope with a flexible GSGGGG linker sequence, which also contains a BamHI site. The oligo contains silent C-to-A and G-to-A mutations to inactivate the two indicated PAMs. (C) (i) pha-1(ts) mutants propagated at the permissive temperature were injected with 60 ng/µl each of CRISPR/Cas9 plasmids targeting the PAM in pha-1(ts) and PAM no. 1 in nhr-23, 50 ng/µl of an oligo designed to correct the pha-1(ts) allele, and 50 ng/µl of the 200mer nhr-23::2×FLAG repair oligo. (ii) P0 animals were singled onto individual plates and shifted to the restrictive temperature (25°). (iii) Three to four days later, the plates were screened for the presence of viable progeny (L3s to adults); the e2123 embryonic lethality phenotype is completely penetrant at 25°, and only rescued animals develop. Rescued F1 progeny were singled onto individual plates, allowed to lay eggs (2–3 days), and the parental F1 was genotyped by PCR followed by BamHI digestion. Correct insertion of the epitope is confirmed by sequencing-purified PCR products with knock-in-specific primers. (iv) Homozygotes were recovered by plating 12–24 progeny from candidate nhr-23:2×FLAG knock-in F1’s, allowing them to lay eggs (2–3 days), and genotyping the parental F2 animal by PCR and BamHI digestion. Marker size in kilobases is provided. (D) Summary of pha-1(ts) co-conversion experiments. Viable P0 are the number of injected animals that produced eggs; a variable number of animals are sterile in each experiment. The length and polarity (with respect to the coding strand) of the pha-1(ts) repair and nhr-23::2×FLAG oligos are provided. P0 were propagated on OP50 E. coli for these injections.

I next tested whether repair could be used to enrich for knock-in of a 2×FLAG oligo into the 3′ end of the gene (Figure 1B), a nuclear hormone receptor involved in molting and embryonic development (Kostrouchova , 2001). I identified two potential PAMs near the desired insertion site that could be silently mutated (Figure 1B). I designed a 200mer oligo to insert a 2×FLAG epitope just before the stop codon. As half of sgRNAs have been reported to fail in C. elegans (Kim ), I mutated two PAMs in the oligo to allow knock-in attempts with different sgRNAs. An 18-bp linker sequence encoding the flexible linker peptide glycine-serine-4xglycine (GSGGGG; Figure 1B) was included to spatially separate the 2×FLAG tag from the NHR-23 C terminus, potentially facilitating accessibility of the tag for immunoprecipitation. Additionally, the tag encoded a BamHI restriction site, which was used for diagnostic restriction digestion in PCR-based screening (Figure 1B). I injected a CRISPR/Cas9 plasmid targeting , an CRISPR/Cas9 construct targeting the PAM nearest the stop codon (PAM no. 1; Figure 1B), the 80mer repair oligo (Figure 1A), and the 200mer oligo (Figure 1B); this oligo was not PAGE purified to test whether PAGE purification was a necessary cost. From 16 viable injected P0 animals, nine rescued F1 progeny were obtained (Figure 1D). These animals were plated, allowed to self-fertilize, and then single-worm genotyping was performed (Figure 1C). Two of these nine animals were heterozygous for a potential 2×FLAG insertion by diagnostic digest, of which one was a correct insertion; the other candidate had a frameshift in the 2×FLAG epitope. F2 progeny of the animal carrying the precise 2×FLAG knock-in were singled and a homozygote for both the insertion and the repaired allele was obtained; both knock-ins were confirmed by sequencing. Brood-size analysis of a representative knock-in confirmed viability with no phenotype, indicative of loss of function (i.e., molting defects or high embryonic lethality) (Table S5).

To distinguish between repair homozygotes and heterozygotes without sequencing, I used CEL-1 digestion of PCR products (Figure S2). CEL-1 is a celery endonuclease that recognizes and cleaves mismatches in double-stranded DNA (dsDNA) formed from single nucleotide polymorphisms or small insertions or deletions (Yang ; Wood ). In this assay, PCRs from repair heterozygotes will lead to a digestion product (Figure S2). Normally, the absence of digestion could indicate that the animals were homozygous for either the ts allele or the repaired allele. However, growth at the restrictive temperature eliminates all ts homozygotes; therefore, all undigested PCR products are repair homozygotes. The genotypes predicted by CEL-1 digestion were confirmed by sequencing. CEL-1 or other mismatch cutting nucleases such as mung bean nuclease or T7E1 represent an efficient method to monitor single-basepair changes and help reduce the number of animals that need to be sequenced to identify homozygous knocked-in point mutations.

Short oligos are effective templates for gene conversion

The stringency of co-conversion offered a powerful tool to optimize oligo editing parameters. As much experimental effort was spent isolating homozygotes to confirm correct epitope insertion, I designed primers that allowed sequencing of the 2×FLAG tag in heterozygotes. These primers bound the insertion junction, with the two 3′-most bases binding to the inserted sequence (Figure S3). In some cases, poor sequencing quality made it difficult to confirm the sequence in the center of the epitope; in these cases, a separate PCR was performed using the epitope-specific primer and an external primer and sequencing was performed on this purified PCR product using the external primer. Screening in heterozygotes greatly reduced hands-on effort required to identify correct insertions.

I next tested whether oligo length (200mer, 80mer, and 60mer) affected repair efficiency. The 80mer was PAGE purified, but the 200mer and 60mers were not. These oligos were all sense in relation to the coding strand. The 80mer and 200mer produced similar numbers of rescued F1’s (9 and 12, respectively) and with the 80mer producing one insertion and the 200mer producing four insertions (Figure 1D). Interestingly, I observed a high percentage of males in the repaired F1, generated using the 200mer (Table S6). Despite the 60mer only producing three rescues, two of those F1 also had knock-ins at the locus. One knock-in was incomplete and produced a frameshift; the other had an extra insertion after the 2×FLAG tag, but frame was maintained. I confirmed the expression of the 2×FLAG tag by immunoblotting in five knock-in lines carrying a precise insertion of the 2×FLAG tag, as well as a line that contained an imprecise insertion. All five precise knock-ins expressed the FLAG tag, whereas no band was seen for the strain carrying the frameshifted FLAG tag (Figure 2). co-conversion allowed isolation of sequence-verified homozygotes for a knock-in event within 8–9 days.

Additional repair template considerations

I then examined whether oligo polarity had an impact on editing, using a PAGE purified, antisense version of the 80mer repair oligo. The sense oligo was homologous to the coding DNA strand, whereas the antisense oligo was homologous to the template DNA strand. The antisense oligo produced the fewest knock-ins per viable P0 animal of all conditions tested, but still had a high rate of ::2×FLAG knock-in per rescued F1 (Figure 1D). Thus, it appeared to be a poor repair template for repair, but once this repair event was selected could still lead to enrichment in ::2×FLAG knock-in. To further explore this variable, I tested whether ::2×FLAG could be knocked in using an antisense 200mer. Using a sense 200mer, five rescued F1’s were obtained from 57 viable P0 animals, but no ::2×FLAG knock-ins were detected (Figure 1D). Including the antisense 200mer with the sense 200mer did result in a decrease in knock-in efficiency (1/5 vs. 1/9), but more animals need to be screened to determine if this effect is significant (Figure S4). Interestingly, when annealed ::2×FLAG sense and antisense oligos were injected, no knock-ins were recovered, indicating that dsDNA is not a better template than single-stranded DNA (ssDNA) and may be less effective (Figure S4). These data suggest that the strand to which oligo homology is derived could be an important parameter for editing efficiency and that ssDNA may be more effective than dsDNA for epitope knock-in. However, these inferences must be tested with many additional combinations of sgRNAs, loci, and repair templates to assess their generality.

Growth on HB101 suppresses pha-1(ts) low brood size and sterility

A number of injected P0s in these experiments were sterile. This sterility could have been a consequence of injection trauma or general sickness of mutants. To discriminate between these possibilities, I shifted adult mutants to 25° to mimic the selection protocol. Interestingly, these animals displayed sterility (1/24 P0) and low brood sizes (<10 eggs) in 6 of 24 P0s. Growth on HB101 has been shown to suppress the slow growth rate of and mutants, which have defective pharyngeal pumping (Shtonda and Avery 2006). Moreover, strains are frequently grown on HB101 in genome editing protocols to ameliorate sickness of that strain (Frøkjær-Jensen ; Dickinson ). I therefore tested the effect of HB101 growth on mutants by propagating animals for several generations on HB101 at 15°, then picking adult animals and shifting them to 25°. Growth on HB101 suppressed the low brood size and sterility observed during growth on OP50. mutants grown on HB101 and then shifted to OP50 at L4 had brood sizes of 127 ± 34 (Table S5) with 0% viable progeny. Subsequently, the strain was maintained on HB101 at 15° prior to micro-injection.

A PCR fusion method to rapidly generate new sgRNA templates

Having demonstrated that oligo-mediated repair of the mutation can be used to efficiently enrich for knock-ins at other loci, I next turned to optimizing sgRNA delivery. Generation of new sgRNA templates for plasmid-based CRISPR/Cas9 systems typically involves site-directed mutagenesis (Dickinson ) or cloning using oligonucleotides (Waaijers ) or PCR products (Friedland ; Kim ). A major impediment to the success of a CRISPR/Cas9 experiment is the efficiency of the sgRNA; Kim report that half of their tested sgRNAs fail. Given that efficient promoter::GFP reporters can be generated by fusion PCR (Hobert 2002), I tested whether the same approach could be used to express sgRNAs from the U6 promoter. This approach would allow rapid screening of sgRNAs, as PCR templates could be injected on the same day of amplification without need for cloning or transformation. For cost efficiency, four PCR primers were used, of which only one is changed to generate a new PU6::sgRNA template (Figure 3A). As a proof of principle, I generated PU6::sgRNA template fusions targeting and PAM no. 1, and deleted the sgRNA template sequence from the pJW1219 Cas9/CRISPR plasmid. Injecting this Cas9 plasmid (pJW1259), the PCR-generated and PU6::sgRNA templates, the sense 80mer repair template, and the sense 200mer, efficient knock-in at both loci with comparable frequency to the corresponding plasmid-based PU6::sgRNA templates was observed (Figure 3B). Fusion PCR thus allows rapid, cost-effective generation of sgRNA templates driven by the U6 promoter.

Figure 3

PCR generated sgRNA templates can be used for pha-1(ts) co-conversion. (A) Method for generating PU6::sgRNA templates by PCR. The U6 promoter and sgRNA template were separately subcloned to generate pJW1310 and 1311, respectively. The U6 promoter is amplified with oligos 1 and 2, and the sgRNA template is amplified with oligos 3 and 4. Oligo 3 contains 20 bp of homology to the sgRNA template, the new 20 bp sgRNA targeting sequence, and 20 bp of homology to the U6 promoter. The resulting PU6 and sgRNA template PCR products are then mixed and used as template in a second PCR reaction using oligos 1 + 4. (B) Animals were injected with 60 ng/µl of the Cas9 plasmid, 25 ng/µl each of PCR generated PU6::sgRNA templates targeting pha-1 and nhr-23 (same targeting sequence as in Figure 2), and 50 ng/µl each of the 200mer sense pha-1(ts) repair oligo and nhr-23::2×FLAG oligo. Viable P0 are the number of injected animals that produced eggs; a variable number of animals are sterile in each experiment. P0 were propagated on HB101 E. coli for these injections. Oligo polarity (sense) is with respect to the coding strand.

DSBs up to 54 bp from the insertion site can generate nhr-23::2×FLAG knock-ins

During my initial design of the ::2×FLAG repair oligo, I mutated two potential PAMs, as it has been reported that half of all sgRNAs fail; this single oligo would allow testing of two separate sgRNAs (Kim ). While analyzing the sequence of the 22 candidate ::2×FLAG knock-ins generated using sgRNAs targeting PAM no.1 and identified by BamHI digest (Figure 1D, Figure 3B, Table 2), I observed co-conversion of the PAM no. 2 silent mutation in six lines. These co-conversion events suggested that selection could be used to explore gene conversion track lengths and effectiveness of DSB position relative to the insertion site.

Table 2

cku-80 RNAi results in increased knock-in efficiency

P0 strain	pha-1 oligo	Repair oligo	Viable injected P0	pha-1 rescued F1	P0 with rescued F1	PCR hits	Knock-ins	Knock-ins/F1 rescue (%)	Knock-ins/P0 (%)
pha-1(ts); control(RNAi)a	80mer	nhr-23:::2×FLAG	10	1	1	1	1	100.0	10.0
pha-1(ts); cku-80(RNAi)a	80mer	nhr-23:::2×FLAG	16	10	6	6	5	50.0	31.3
pha-1(ts); control(RNAi)b	200mer	nhr-25:::2×FLAG	21	7	4	0	0	0.0	0.0
pha-1(ts); cku-80(RNAi)b	200mer	nhr-25:::2×FLAG	22	36	12	5	4	11.1	18.2
pha-1(ts); control(RNAi)b	200mer	nhr-23::3xFLAG	34	7	4	1	1	14.3	3.0
		nhr-25::3xFLAG				1	1	14.3	3.0
pha-1(ts); cku-80(RNAi)b	200mer	nhr-23::3xFLAG	13	5	3	2	2	40.0	15.4
		nhr-25::3xFLAG				2	1c	20.0	7.7
pha-1(ts); cku-80(RNAi)b	200mer	2×FLAG::smo-1	15	29	7	14	11	38.0	73.3
		lig-4 stop				0	0	0.0	0.0

Summary of pha-1(ts) coselection experiments testing NHEJ inactivation. The RNAi treatment of the injected P0 animals is indicated. Viable P0 are the number of injected animals that produced eggs; a variable number of animals are sterile in each experiment. The length of the pha-1(ts) sense repair oligo is provided. For nhr-23::2×FLAG experiments, animals were injected with 60 ng/µl each of the pha-1 and nhr-23 CRISPR/Cas9 plasmids, and 50 ng/µl each of the pha-1(ts) repair and nhr-23::2×FLAG (sense) oligos. For the remaining experiments, animals were injected with 50 ng/µl of the pha-1 CRISPR/Cas9 plasmid, 25 ng/µl of appropriate PU6::sgRNA template PCR product, and 50 ng/µl each of the pha-1(ts) repair oligo and knock-in oligo.

Pre-RNAi diet = OP50 E. coli.

Pre-RNAi diet = HB101 E. coli.

The silent point mutations in PAMs no. 1 and no. 2 were 29 bp apart, which agrees with my observation that a 60mer with 29 bp of homology can be used to efficiently knock-in point mutations (Figures 1D and Figure 4A). For these next experiments, I sequenced all rescued F1’s. I first tested whether an sgRNA targeting PAM no. 2 could be used to knock in the 2×FLAG tag into the 3′ end; 3 of 12 -rescued F1 animals had the PAM no. 2 mutation knocked in, with one animal also containing the PAM no. 1 mutation and correct insertion of the 2×FLAG tag (Figure 4B, Table S7). I next examined whether a DSB 54 bp away from the desired insertion site could be used to knock in the 2×FLAG epitope (Figure 4A). I designed a modified version of the 200mer used for the above experiment that also carried three silent mutations to disrupt the sgRNA binding to the PAM no. 3 target site, as well as the mutations in PAMs no. 1 and no. 2 (Figure 4A). A fourth potential PAM (no. 4) was left intact in the repair template, as inactivation would leave only 10 bp of 5′ homology; this sgRNA was inactive (Figure 4, A and B). Using an sgRNA targeting PAM no. 3 to generate a DSB, two of eight -rescued F1’s carried the silent mutation in PAM no. 3; the others all had wild-type sequence, suggesting that this may be an inefficient sgRNA (Figure 4B, Table S7). Of the two animals with mutations in PAM no. 3, both had the PAM no. 2 inactivating mutation, and one had the PAM no. 1 mutation and insertion of the 2×FLAG tag (Table S7). Though there was a 1-bp deletion within the tag, this experiment demonstrates that a DSB 54 bp from an insertion site can be used for introduction of epitopes, and that the presence of mutations in a stretch of homology does not prevent gene conversion; five point mutations in a 55-bp stretch were introduced into the genome along with the 65 bp of 2×FLAG tag (with the 1-bp deletion). Animals were not tracked to identify originating P0s; however, for the PAM no. 1 knock-in experiment, the six animals with PAM no. 2 mutation co-conversion were isolated from six different injections, and thus represent independent events (Figure 4B). Knock-in efficiency appeared higher the closer the DSB was to the insertion site, but sgRNA efficiency could also play a role in this observation.

Figure 4

DSBs up to 54 bp from an insertion site, and 35-bp homology arms can be used for oligo-templated repair in nhr-23. (A) Schematic of the nhr-23 3′ end, indicating the stop codon (blue text), four PAMs tested (red text), and position of the DSBs (scissor). Mutations used to inactivate PAMs in repair templates are provided in the nhr-23(PAM MUT) sequence and indicated by the vertical lines between the (+) and (PAM MUT) sequence. (B) Testing effect of DSB position on nhr-23::2×FLAG epitope knock-in efficiency. The PAMs for the sgRNAs used to generate the DSB and mutations used to inactivate the PAMs in the repair templates are provided in A. Animals were co-injected with 50 ng/µl of pha-1 targeting CRISPR/Cas9 plasmid, 50 ng/µl each of the pha-1(ts) repair oligo and a 200mer sense nhr-23::2×FLAG repair oligo, and either 50 ng/µl of nhr-23 targeting CRISPR/Cas9 plasmid (PAMs no. 1 or no. 2), or 25 ng/µl of PU6::sgRNA template PCR product (PAMs no. 3 and no. 4). For the PAM no. 1 and no. 2 sgRNA experiments, the repair template carried mutations in both PAMs. For the PAM no. 3 and no. 4 experiments, the repair template carried mutations in PAMs no. 1, no. 2, and no. 3. PAM no. 3 had a single PCR hit with a 2×FLAG insertion carrying a 1-bp deletion. For the PAM no. 1 row, all experiments using nhr-23::2×FLAG 200mers were pooled (Figure 1, Figure 3, Table 2); only animals that displayed a knock-in signature in the diagnostic BamHI digest from these experiments were sequenced. For the PAM no. 2, no. 3, and no. 4 experiments, all pha-1(ts) F1 rescued animals were sequenced. (C) Testing homology arm length on nhr-23::2×FLAG knock-in efficiency. Homology lengths in basepairs for 5′ and 3′ arms are provided. 5′ arm homology numbering starts from the middle G in PAM no. 1; 3′ homology numbering starts from the first basepair of the stop codon. Animals were co-injected with 50 ng/µl each of pha-1 targeting and nhr-23 PAM no. 1 targeting CRISPR/Cas9 plasmids, 50 ng/µl each of the pha-1(ts) repair oligo and nhr-23::2×FLAG repair oligo. For the 76/54-bp homology arm row in the table, data were pooled from all non-RNAi experiments using pha-1 sense repair oligos and nhr-23::2×FLAG sense 200mers (Figure 1 and Figure 3). With the exception of the pooled data, all injected P0 animals were grown on HB101 E. coli in B and C. Oligo polarity (sense) is with respect to the coding strand.

Homology arms of 35 bp are sufficient for oligo-templated insertion of a 2×FLAG tag

Exploring DSB position relative to the insertion site demonstrated that the stringency of co-conversion combined with the speed and ease in recovering editing events allowed rapid testing of editing parameters. Interestingly, the 200mer used for the PAM no. 3 knock-in experiments only had 29 bp of homology 5′ to the DSB (Figure 4, A and B), suggesting that homology arms could be relatively short. Efficient repair of the allele was observed using a 60mer with 29 bp of homology (Figure 1D). Furthermore, recent data using PCR-derived dsDNA repair templates demonstrated that 30- to 60-bp homology arms were optimal, with knock-in efficiency actually decreasing with longer homology arms (Paix ). To test the ideal length of homology for oligo-mediated insertion of epitopes, I designed ::2×FLAG oligos with 35 bp, 25 bp, or 15 bp of homology and tested their ability to introduce the 2×FLAG epitope using a DSB generated by an sgRNA targeting PAM no. 1. Knock-in efficiency using these homology arms was compared to pooled data from the and ::2×FLAG sense oligo experiments (Figure 1D and Figure 3B). This 200mer sense ::2×FLAG oligo (Figure 1B) contained 54 bp of homology 3′ to the insertion site and 76 bp of homology 5′ to PAM no. 1 (76/54). With the 35-bp homology arms, two animals from 9 rescued F1’s had correct insertion of the 2×FLAG epitope (Figure 4C). This 22% knock-in per F1 screened (2 of 9) is comparable to the 26% efficiency (7 of 27) observed using the 76/54 oligo (Figure 4C). The knock-in rate per successfully injected P0 was also comparable for the 35-bp and the 76/54-bp homology arms [4.76% (2/42) vs. 6.86% (7/102)]. The 15- and 25-bp homology arm produced fewer rescues per successfully injected P0, and only the 25-bp homology arms could produce a 2×FLAG insertion, though this knock-in carried a point mutation (Figure 4C). These data suggest that, similar to dsDNA templates, homology arms of 35–80 bp are ideal for oligo-templated editing.

Inactivation of NHEJ repair increases knock-in efficiency

In Drosophila, increased homologous recombination efficiency can come from inactivation of NHEJ (Beumer ; Bottcher ). I therefore used the system to test whether NHEJ inactivation impacted knock-in by homologous recombination. NHEJ mutants may have additional background mutations due to compromised repair of endogenous DSBs and would require additional outcrossing. I therefore tested the effect on knock-in frequency of temporary inactivation of the C. elegans homolog of Ku80 (), part of a heterodimer that binds the end of DSBs in NHEJ, and which had reported RNAi phenotypes (Dmitrieva ; McColl ). and CRISPR/Cas9 plasmids, the sense 80mer repair oligo, and the sense 200mer repair oligo were injected into mutants treated with control or RNAi. Control RNAi produced one rescued F1, which also carried a sequence-confirmed knock-in event (Table 2). However, inactivation of by RNAi produced an increase in both rescue (10 F1’s) and co-knock-ins recovered (n = 5) (Table 2). These experiments suggested that NHEJ inactivation boosts oligo-mediated knock-in efficiency.

To confirm that co-conversion, PCR-generated sgRNAs templates, and NHEJ inactivation were effective on other loci, I attempted to introduce a 2×FLAG epitope in , a broadly expressed nuclear hormone receptor that regulates several developmental and physiological programs (Asahina ; Gissendanner and Sluder 2000; Chen ; Asahina ; Mullaney ; Ward , 2014). I selected the 3′ end of as there are two isoforms that share a common 3′ end, but differ in promoter use; thus, targeting the 3′ end of with the 2×FLAG should in principle allow for labeling of all known isoforms. I designed a 175mer repair template with six silent point mutations in the 20-bp sequence preceding the PAM, as the PAM could not be silently mutated (Figure S5, A and B). As with the ::2×FLAG construct, an 18-bp spacer encoding with a BamHI site was also included (Figure S5, A and B). This oligo contained 41 bp of homology 5′ to the mutated sgRNA target sequence and 43 bp of 3′ homologous sequence. mutant animals grown on either control or RNAi were injected with a CRISPR/Cas9 plasmid, an PU6::sgRNA template PCR product, and and ::2×FLAG repair oligos. Seven rescues were recovered from the control RNAi-treated animals and 36 from the RNAi-treated animals (Table 2). No knock-in candidates were recovered from the control RNAi injection, but five were recovered from the RNAi-treated animals, of which four were correct insertions of the 2×FLAG tag (Table 2). Strikingly, one of these animals was an F1 homozygous knock-in. These strains were viable and did not display defects associated with inactivation (protruding vulvae, molting defects, embryonic lethality; see Table S5 for a representative brood size). I confirmed expression of the 2×FLAG tag in an outcrossed, representative line (Figure 5B).

Multiplexed editing using pha-1(ts) co-conversion and cku-80 RNAi

Given the efficient editing observed in RNAi-treated animals and the 35–80 bp of homology sufficient for oligo-templated repair, I next tested whether I could introduce larger epitopes using co-conversion. As Kim had reported generating mutations in both and in one co-CRISPR experiment, I attempted to knock in 84-bp 3xFLAG tags containing the GSGGGG spacer into both and in one injection experiment (Figure S5A). From 13 successfully injected P0 animals treated with RNAi, two precise ::3xFLAG lines were obtained (Table 2). A viable ::3xFLAG line that expressed the 3xFLAG tag, but had a 51-bp duplication in the 3′ UTR was also obtained; as the tag was expressed and the strain was viable, this line was scored as a correct knock-in (Table 2, Table S5). A single knock-in animal was obtained from 34 successfully injected P0 animals grown on control RNAi (Table 2). This F1 animal was homozygous for an ::3xFLAG insertion and heterozygous for an ::3xFLAG insertion (Table 2). Viability was confirmed for representative ::3xFLAG and ::3xFLAG lines and for the ::3xFLAG; ::3xFLAG double knock-in line (Table S5). No defects consistent with or inactivation were observed, though the double knock-in did have a lower brood size (Table S5). I confirmed FLAG epitope expression in outcrossed ::3xFLAG lines and ::3xFLAG lines (Figure 5, A and B). The 3xFLAG epitope lines displayed a marked increase in band intensity in immunoblots compared to 2×FLAG lines, for both NHR-23 and NHR-25 (Figure 5, A and B). Notably, the ::2×FLAG epitope was not detectable in this experiment (Figure 5A), as a more potent ECL substrate was required for detection by immunoblotting (Figure 2); this more potent ECL substrate may explain the nonspecific band observed in Figure 2, but not in Figure 5A. I also observed extra bands in the 3xFLAG lines for NHR-23 and NHR-25 compared to the corresponding 2×FLAG lines, which would be consistent with NHR-23 and NHR-25 isoforms, though they could also represent degradation products (Figure 5, A and B). Comparing all NHEJ inactivation experiments, growing P0 animals on RNAi produced a significant increase in the number of knock-ins recovered per injection experiment (two tailed t-test P = 0.01) and percentage of knock-ins per viable P0 (two-tailed t-test P = 0.016). No significant difference was observed in the number of P0 animals with rescued F1 progeny (two tailed t-test P = 0.16) or in the percentage of knock-ins per rescued F1 (two tailed t-test P = 0.85). Recent manuscripts have described a “jackpot” phenomenon (Arribere ; Paix ), where the majority of edits come from a small number of P0 animals. There may be a trend of “richer jackpots” in RNAi-treated animals with 2.6 rescued F1’s/P0 producing F1 rescues (±0.44; 95% confidence interval) vs. 1.6 (±0.50; 95% confidence interval) in control RNAi-treated animals. This difference was not significant in a two-tailed t-test (P = 0.19) potentially due to the large variation in number of F1 rescues and knock-ins produced across assays, and this observation will require further exploration.

PAGE purification of oligos is not necessary, but increases editing efficiency

The ::3xFLAG oligo appeared to be a poor repair template, as it yielded lower knock-in efficiency in comparison to the ::2×FLAG template, and complex insertions were observed in two ::3xFLAG candidate knock-in lines. Although experiments demonstrated that PAGE purification was not necessary, the ::3xFLAG template was ideal to test whether PAGE purification nonetheless increased knock-in efficiency or fidelity. Resolving unpurified and purified oligos on a denaturing TBE-Urea gel revealed that the unpurified oligo preparation contained large amounts of incorrectly sized product (Figure S6A). animals grown on control or RNAi were injected with the targeting CRISPR/Cas9 vector, an sgRNA template PCR product, a repair oligo, and either purified or unpurified ::3xFLAG 175mers. Interestingly, this experiment demonstrated that PAGE purification resulted in higher rates of knock-ins per F1 and per viable P0 (Figure S6B). Comparing all experiments using the ::3xFLAG oligos (Table 2, Figure S6C), both knock-in candidates obtained using the purified oligo were precise insertions, whereas two of the four candidates generated with the unpurified oligo contained additional sequence inserted (Figure S6C). These experiments suggested that one can either opt for an increased knock-in rate with PAGE purified oligos or avoid this cost and inject/screen more animals.

Efficient editing at other genomic loci and rapid testing of sgRNA activity using pha-1(ts) selection and NHEJ inactivation

Finally, I wished to test these editing parameters on other genes to test their general applicability and robustness. Based on my previous work on NHR-25 sumoylation (Ward ), I chose a previously described weakly efficient sgRNA (Kim ) that cleaved 19 bp 3′ to the start codon of the single C. elegans SUMO gene, (Broday ). I designed a 175mer repair oligo to create a SMO-1 N-terminal 2×FLAG fusion (Figure S5, C and D). I also designed two overlapping sgRNAs to target the NHEJ gene, , and a 60mer oligo as the homologous recombination template to insert a stop codon (oligos were unpurified). From nine successfully injected P0 animals grown on RNAi and injected with PCR-derived sgRNA templates, 27 rescued F1 animals were obtained. Fourteen of these F1’s carried knock-ins in , of which 11 were precise (Table 2). I confirmed the expression of the 2×FLAG epitope in a representative line (KRY82) by immunoblotting; signal was observed over a large range of molecular weights, as expected from SUMO conjugation to hundreds of substrates in C. elegans (Kaminsky ) (Figure 5C). Although viable, 2×FLAG::smo-1 homozygotes displayed partially penetrant phenotypes consistent with reduction of function, such as small body size and protruding vulvae. Thus, epitope placement may need to be optimized in , though the creation of a viable hypomorph will be a useful reagent for the community. The sgRNAs failed, as no knock-ins or mutations were obtained in , demonstrating that this approach also allows rapid testing of sgRNA efficacy, similar to Kim (Table 2). Together, these results demonstrate that a co-conversion approach using a temperature-sensitive mutant allele, PCR-generated sgRNA templates, and NHEJ inactivation by RNAi provide a flexible, robust platform to recover genome editing events.

pha-1(ts) coselection increases screening efficiency

Based on recent co-CRIPSR/co-conversion reports (Arribere ; Kim ), and the frequency at which relatively rare repair events were associated with edits at other loci, it was highly probable that coselection increases screening efficiency. Prior to developing co-conversion, I had attempted to introduce 2×FLAG tags onto the 3′ end of and and recover knock-ins through direct screening of F1 progeny, similar to the approach described by Paix (see Supporting Information for detailed methods). No knock-ins were recovered from 380 screened WT F1’s (Table S8). I also tested the effect of NHEJ inactivation by injecting into mutants; DNA ligase 4 () encodes the enzyme that seals DSBs in canonical NHEJ, and is an out-of-frame deletion that removes the catalytic ligase domain and is predicted to result in a premature stop codon (Clejan ). A single knock-in was recovered from 768 screened mutant F1’s (Table S8); after outcrossing the mutation, this ::2×FLAG line had a brood size equivalent to a knock-in produced by coselection (Table S5). In comparable coselection experiments, no knock-ins were recovered from 7 F1 laid by control RNAi treated P0s and four knock-ins were recovered from 36 F1’s laid by RNAi-treated P0 (Table 2).

For ::2×FLAG knock-in experiments, I used a knock-in-specific PCR screening approach; as in optimization experiments, it detected knock-in DNA diluted 1:1280 with WT DNA, whereas diagnostic restriction digestion of PCR products could only detect up to a 1:20 dilution of knock-in DNA with WT DNA (Figure S7). No knock-ins were recovered from 200 WT F1 animals or 840 mutant F1’s (Table S8). In comparison, coselection using a similar sense ::2×FLAG repair template yielded seven knock-ins from 27 screened rescued F1’s (Figure 4C). Comparing all similar ::2×FLAG and ::2×FLAG direct selection and co-conversion experiments (i.e., excluding antisense oligo and NHEJ inactivation experiments) demonstrated a significant increase in screening efficiency compared to direct F1 screening (two-tailed t-test, P = 0.01).

Conclusions

Here, I demonstrate that selection for repair of a temperature-sensitive point mutation using an oligonucleotide template can be used to efficiently select for knock-in at other loci in as quickly as 8–9 days. Several findings of this work should be applicable for a range of co-conversion/co-CRISPR approaches.

First, the increased sgRNA(F+E) activity (Table 1) may improve knock-in and knock-out efficiencies and reduce the number of sgRNAs that one must test to identify an active sgRNA. Second, for introduction of single-basepair changes, efficient editing was observed using 60mers, 80mers, and 200mers. There may be a slight improvement in knock-in efficiency with longer repair templates, but using shorter oligos still produces repair events at a high rate. These data are consistent with the 29-bp repair track length observed using the second silent PAM mutation in the ::2×FLAG repair oligo and the efficient ::2×FLAG knock-ins obtained using 35 bp of homology (Figure 4). This optimal homology for oligo-mediated editing is similar to that reported for dsDNA templates (Paix ). The 84-bp 3xFLAG and knock-ins is a larger insertion than described by several recent reports (Lo ; Arribere ; Paix ; Zhao ); a 66-bp 3xFLAG insertion in the gene is the most comparable edit (Paix ). Based on the minimal homology length of 35 bp and current oligo synthesis limit of 200 bp, it may be possible to knock in sequences of up to 130 bp using oligo templates. Third, I demonstrated that oligos do not need to be PAGE purified, as efficient editing was observed using unpurified 60mers and 200mers to edit the allele and to introduce the ::2×FLAG epitope (Figure 1D). However, PAGE purified oligos did result in increased knock-in efficiency and fewer imprecise knock-ins (Figure S6). Thus, an investigator can choose between the additional cost for the increased editing efficiency or simply inject more P0 animals and screen more F1’s. It is very important to note that there is likely wide variation in quality of unpurified oligos between different suppliers. Different preparations may contain different inhibitors or cytotoxic compounds and one should confirm the efficiency of unpurified oligos from a new supplier by testing a control such as rescue or knock-in (Arribere ). Fourth, PCR-generated PU6::sgRNA templates allow rapid production of new sgRNAs without need for cloning. Moreover, the robust sgRNA activity observed when injecting linear dsDNA templates enables other technologies for sgRNA template production such as gene synthesis or oligonucleotide arrays (Bassik ; Gilbert ), which could be used to create pooled sgRNA template libraries. Injecting multiple PCR-generated PU6::sgRNA templates allows several editing experiments to be performed simultaneously. From single-injection experiments, I was able to knock 3xFLAG tags into both the and loci or generate 2×FLAG::smo-1 lines while also determining that the sgRNAs were inactive (Table 2). Finally, the observation that inactivation of NHEJ can lead to improved knock-in rates can be adapted for any HR-based editing system, though it would reduce the efficiency of systems that rely on coselection of CRISPR-generated mutations as a marker. The increased knock-in efficiency observed following NHEJ inactivation was surprising and suggests that at least some knock-ins must be occurring outside in the germline; previous work has shown that NHEJ is actively suppressed in meiosis to ensure faithful repair of DSBs by homologous recombination (Adamo ). Determining the cell types in which knock-ins are occurring and the DNA repair pathways involved could lead to improvements in editing efficiency.

It was intriguing that, for and , higher editing efficiency was observed using oligos with homology to the coding strand. In these experiments, the sgRNA for recognized the coding strand, while the PAM no. 1 sgRNA recognized the template strand (Figure 1D, Table S9). Although numerous explanations could be invoked for a polarity bias (sgRNA sequestration of the oligo, oligo:mRNA hybridization, different secondary structures in a given oligo and its reverse complement, etc.), it is unclear whether my data reflect a biologically meaningful trend or a chance observation. Best practice should be to clearly report the strand to which the sgRNA binds and the strand to which the oligo contains homologous sequence (Table S9). A metaanalysis of many more combinations of sgRNAs and oligos of differing polarities at a large number of genes is required to definitively determine whether oligo polarity is an important parameter. If one fails to obtain oligo-templated knock-ins when using a sgRNA with confirmed activity, then it may be worth testing the complement of the oligo.

The approach is distinct from reported co-CRISPR and co-conversion methods (Arribere ; Kim ) in that it starts with a mutant animal and results in restoration of a wild-type animal. In contrast, the other two approaches can be used with any strain, but require outcrossing or meiotic segregation of the selection marker. co-conversion may be advantageous as a marker in cases where the desired insertion site is linked to the selectable marker site. Arribere demonstrated iterative editing using co-conversion. co-conversion should allow for iterative editing events, though not as elegantly as the Arribere approach. F1 rescues have all been heterozygous for the repair event. By shifting to the permissive temperature (15°) after generating homozygous animals of a desired knock-in, one could reisolate the allele and perform another round of editing; this reisolation would require an additional 5 days. Alternatively, one could simply cross the allele back into an edited strain.

The optimizations I report make oligo-mediated editing efficient, cost effective, and can be applicable to any CRISPR-mediated editing system. The development of three distinct variations of coediting selection ( mutation; Kim ), knock-in (Arribere ), and repair (this article) highlight the robustness of this method to select for genome editing events. The recent description of PCR-derived dsDNA templates (Paix ) will make these approaches even more powerful; the abundance of potential genetic markers in other model organisms should make these widely applicable approaches. A current challenge in cell-culture-based systems has been the laborious recovery of rare knock-in events. Coselection markers such as oligo-mediated repair of mutated GFP or drug resistance cassettes, or inactivation of hypoxanthine phosphoribosyl transferase, could yield similar improvements in editing event recovery in these systems.