- Split View
-
Views
-
Cite
Cite
Jeffrey C Medley, Shilpa Hebbar, Joel T Sydzyik, Anna Y Zinovyeva, Single nucleotide substitutions effectively block Cas9 and allow for scarless genome editing in Caenorhabditis elegans, Genetics, Volume 220, Issue 1, January 2022, iyab199, https://doi.org/10.1093/genetics/iyab199
- Share Icon Share
Abstract
In Caenorhabditis elegans, germline injection of Cas9 complexes is reliably used to achieve genome editing through homology-directed repair of Cas9-generated DNA breaks. To prevent Cas9 from targeting repaired DNA, additional blocking mutations are often incorporated into homologous repair templates. Cas9 can be blocked either by mutating the PAM sequence that is essential for Cas9 activity or by mutating the guide sequence that targets Cas9 to a specific genomic location. However, it is unclear how many nucleotides within the guide sequence should be mutated, since Cas9 can recognize “off-target” sequences that are imperfectly paired to its guide. In this study, we examined whether single-nucleotide substitutions within the guide sequence are sufficient to block Cas9 and allow for efficient genome editing. We show that a single mismatch within the guide sequence effectively blocks Cas9 and allows for recovery of edited animals. Surprisingly, we found that a low rate of edited animals can be recovered without introducing any blocking mutations, suggesting a temporal block to Cas9 activity in C. elegans. Furthermore, we show that the maternal genome of hermaphrodite animals is preferentially edited over the paternal genome. We demonstrate that maternally provided haplotypes can be selected using balancer chromosomes and propose a method of mutant isolation that greatly reduces screening efforts postinjection. Collectively, our findings expand the repertoire of genome editing strategies in C. elegans and demonstrate that extraneous blocking mutations are not required to recover edited animals when the desired mutation is located within the guide sequence.
Introduction
The CRISPR/Cas9 system has become increasingly used to facilitate genome editing in numerous organisms (Ma and Liu 2015; Shrock and Güell 2017; Ma et al. 2018). Cas9 (CRISPR-associated protein 9) is a programmable endonuclease whose specificity is governed by a guide RNA that has sequence complementarity to a specific genomic location (Jinek et al. 2012). The guide RNA comprises two molecules: the CRISPR RNA (crRNA) that contains a 20-nucleotide guide sequence and a trans-acting CRISPR RNA (tracrRNA) that forms a duplex with the crRNA and bridges the guide RNA to Cas9 (Deltcheva et al. 2011; Jinek et al. 2012). A protospacer-adjacent motif (PAM) sequence is located immediately downstream of the RNA guide-complementary genomic sequence and is required for Cas9 to initiate a double-stranded DNA break. In the case of Streptococcus pyogenes Cas9, commonly used for genome editing, the nucleotide PAM sequence is NGG, where N is any nucleotide (Mojica et al. 2009; Marraffini and Sontheimer 2010; Jinek et al. 2012; Sashital et al. 2012). Once a double-stranded DNA break is created, the break is typically repaired through one of two mechanisms: nonhomologous end joining (NHEJ) or homology-directed repair (HDR) (Ceccaldi et al. 2016; Li and Xu 2016; Scully et al. 2019; Han and Huang 2020; Yang et al. 2020). In NHEJ, the broken DNA is repaired through direct ligation of the broken DNA ends. However, this process is error prone as the ligation often requires processing of the broken ends, resulting in additions or deletions of nucleotide bases at the break site (Chang et al. 2017; Zhao et al. 2020). Conversely, HDR uses a donor DNA molecule that has homology surrounding the break site as a template to precisely repair the broken DNA (Haber 2018; Ranjha et al. 2018; Sun et al. 2020). Therefore, HDR has been widely adapted to repair Cas9-generated DNA breaks to introduce precise genome edits in a broad range of organisms. Donor repair templates can be exogenously provided as single-stranded oligodeoxynucleotides (ssODN) or double-stranded DNA (dsDNA) molecules for the purpose of genome editing (Cong et al. 2013; Zhao et al. 2014; Paquet et al. 2016; Yoshimi et al. 2016; Gallagher et al. 2020). During CRISPR/Cas9-mediated genome editing, the process of HDR using an ssODN repair template is referred to as single-stranded template repair (SSTR), and results in higher genome editing efficiencies than HDR pathways that use dsDNA repair templates (Katic et al. 2015; Dokshin et al. 2018; Richardson et al. 2018; Liu et al. 2019; Okamoto et al. 2019; Gallagher et al. 2020; Gallagher and Haber 2021).
Once a Cas9-generated DNA break is repaired through SSTR, Cas9 must be prevented from continuing to target the repaired DNA. To accomplish this, additional blocking mutations are often incorporated into homologous repair templates, disrupting the ability of Cas9 to target the repaired sequence. As the PAM is absolutely required for Cas9 activity (Mojica et al. 2009; Marraffini and Sontheimer 2010; Jinek et al. 2012; Sashital et al. 2012), the most straightforward way to block Cas9 is to introduce silent mutations into the PAM. Alternatively, Cas9 can be blocked by introducing mutations into the guide sequence, which targets Cas9 to a specific genomic location (Deltcheva et al. 2011; Jinek et al. 2012). However, studies in human cells have shown that Cas9 is capable of recognizing off-target sequences that are imperfectly paired to its guide RNA (Jinek et al. 2012; Pattanayak et al. 2013; Jiang and Doudna 2015). Mismatches near the 3′ end of the guide RNA appear to be more effective at blocking Cas9 compared to mismatches toward the 5′ end of the guide RNA (Jinek et al. 2012; Cong et al. 2013; Fu et al. 2013; Hsu et al. 2013; Mali et al. 2013; Pattanayak et al. 2013; Zhang et al. 2015). Increasing the number of mismatches generally leads to increased blocking efficacy (Jinek et al. 2012; Cong et al. 2013; Fu et al. 2013; Hsu et al. 2013; Mali et al. 2013; Pattanayak et al. 2013; Zhang et al. 2015). Nevertheless, Cas9 has been reported to cleave DNA sequences containing up to five mismatches to certain guide RNAs (Hsu et al. 2013), although three mismatches effectively block Cas9 for most guide RNAs (Fu et al. 2013; Hsu et al. 2013; Mali et al. 2013). Therefore, it remains unclear how many nucleotides should be mismatched and where the mismatches should be located within the guide sequence to effectively block Cas9 for genome editing in vivo.
In the nematode Caenorhabditis elegans, injection of Cas9 ribonucleotide protein (RNP) complexes and ssODN repair templates into the germline of hermaphrodite animals has been reliably used to facilitate heritable genome editing (Paix et al. 2015, 2017; Farboud et al. 2019). However, certain types of genome edits remain challenging to design due to the need to block Cas9 from targeting the repaired DNA. Protein coding sequences are highly amenable for genome editing experiments, as codon redundancy frequently allows silent blocking mutations to be introduced without changing the amino acid identity. However, genome editing of regulatory and nonprotein coding portions of the genome remain a challenge. It is often difficult to predict how extraneous blocking mutations may affect the function of noncoding regulatory sequences such as noncoding RNAs, untranslated regions, and other regulatory elements. Extraneous blocking mutations can be avoided when the intended edit also alters a PAM site and eliminates Cas9 ability to recut a repaired genome site. In such cases, genome editing is performed in a “scarless” fashion. However, the dinucleotide GG of the PAM sequence (NGG) is only expected to occur, on average, every 16 bases and must overlap with the desired edit to generate a scarless edit. This frequency is likely reduced in noncoding regions that are often AT-rich. It has been suggested that single nucleotide substitutions located within three nucleotides of the PAM are sufficient to allow for genome editing in C. elegans (Paix et al. 2017; Farboud et al. 2019). This is of particular interest for scarless genome editing, as non-PAM mutations could block Cas9 and thereby bypass the need for additional blocking mutations. We reasoned that single nucleotide substitutions beyond the three PAM-adjacent nucleotides, located in the guide-binding region could effectively block Cas9, further facilitating scarless genome editing of noncoding sequences. Toward this end, we have performed a systematic analysis of the blocking efficacy of single nucleotide mismatches throughout the guide sequence in C. elegans. We demonstrate that single nucleotide substitutions throughout the guide-binding sequence are sufficient to block Cas9 and allow for effective recovery of edited animals. Furthermore, we were able to recover heritable genome edited strains without introducing any blocking mutations, suggesting that a temporal block to Cas9 activity limits the ability of Cas9 to target repaired DNA. We also show that editing of the maternal genome of self-fertile hermaphrodite animals occurs at much greater frequency compared to editing of the paternal genome. Finally, we propose a new method of mutant isolation that selects for maternally provided haplotypes and greatly reduces screening efforts postinjection. As a proof-of-principle, we use this method to generate otherwise scarless genome edits in the let-7 microRNA. Our collective findings expand the repertoire of possible genome edits in C. elegans and will facilitate scarless editing of noncoding sequences.
Materials and methods
Caenorhabditis elegans strains and genetics
All C. elegans strains were derived from the wild-type N2 strain and maintained at 20°C unless otherwise noted. Strains were grown under standard conditions using nematode growth medium (NGM) plates seeded with Escherichia coli OP50 (Brenner 1974). A full list of strains used in this study is provided in Supplementary Table S1.
For experiments examining maternal vs paternal editing, matings were performed by adding several wild-type males to a plate containing L4-staged, tra-2 mutant hermaphrodite animals generated in this study. For let-7 genome editing experiments, several males containing the tmc24 balancer were mated to L4-staged, wild-type hermaphrodite animals. Animals were allowed to mate 16–24 h prior to injection. Successful mating was verified by genotyping F1 animals through RsaI digestion to confirm tra-2 heterozygosity, or presence of pharyngeal Venus to confirm presence of the tmc24 balancer.
CRISPR/Cas9 genome editing
Commercially available S. pyogenes Cas9 (IDT, Alt-R® S.p. Cas9 Nuclease V3) was injected at a final concentration of 2.65 µM and was included in all injection mixes. All injection mixtures also contained 200 µM KCI and 7.5 µM HEPES [pH 7.4]. Each injection mix contained one or two crRNAs targeting dpy-10 (5 µM), tra-2 (50 µM) or let-7 (50 µM), and equimolar tracrRNA was included (5, 50, or 55 µM). Single-stranded DNA oligonucleotides for dpy-10 (3 µM), tra-2 (6 µM), and let-7 (6 µM) were used to facilitate HDR (Supplementary Table S2).
To generate nondumpy genome edits at the tra-2 locus, which we used to examine how effectively single nucleotide substitutions blocked Cas9, we replaced dpy-10 coconversion with the Pmyo-2::mCherry coinjection marker (5 ng/µl). The Pmyo2::mCherry coinjection marker was included to mark broods that were successfully injected and to enrich for genome-edits as previously described (Prior et al. 2017). A full list of oligonucleotides (IDT) used in this study is provided in Supplementary Table S2.
Screening and genotyping
For experiments testing the efficacy of tra-2 editing, we blindly sequenced (i.e., without Rsal digestion) the F2 generation Dumpy and nondumpy animals originating from each F1 Roller animal. F1 Rollers were picked from jackpot broods (Paix et al. 2015) that were defined as having >20 F1 Rollers from a single P0 injected animal. F1 rollers that did not produce Dumpy, Roller and nondumpy progeny were excluded from our analysis. The tra-2 genomic locus was PCR amplified using the primers 5′-CTGCTAAAGGTTAGTTGTT-3′ and 5′-ATAATGTATTCTTCATTGTTCG-3′ and sequenced using the primer 5′-ATTTTAGGAATAATTGGAGCC-3′. All Rsal digestions were performed at 37°C for at least 1 h. A Rsal-positive control was included on all gels used for quantification to confirm successful Rsal digestion occurred.
To examine genome editing of let-7, we singled F1 generation Roller animals from jackpot broods that showed pharyngeal Venus signal that indicated successful mating to the tmc24 balancer preinjection. We then blindly sequenced F2 generation animals lacking pharyngeal Venus that were therefore homozygous for the maternally provided X-chromosome haplotype. The let-7 genomic locus was PCR amplified using the primers 5′-GTTTGCGTATGTGTATGTAG-3′ and 5′-TCCCCTGAAAATAAAACATGA-3′ and sequenced using the primer 5′-TATTCTAGATGAGTAGCCCA-3′. All genome edits were verified through Sanger sequencing.
Statistical analysis
All P-values were calculated using two-tailed t-tests assuming equal variance. All statistics are presented as mean ± 1 standard deviation.
Results
Coconversion of tightly linked genes to test genome editing efficiency
In C. elegans, standard genome-editing practices involve injection of Cas9, guide RNAs and homologous repair templates into the germline of self-fertile hermaphrodite adult animals (Xu 2015; Chen et al. 2016; Dickinson and Goldstein 2016; Farboud 2017; Iyer et al. 2018; Kim and Colaiácovo 2019; Nance and Frøkjær-Jensen 2019; Ghanta and Mello 2020). Due to the syncytial nature of the distal gonad, a single injection can be distributed among numerous germ cells (Evans 2006; Kadandale et al. 2008). Although injection of Cas9 into the distal germline is expected to affect the genomes of maternal oocytes, homozygously edited animals can be recovered from the F1 generation postinjection, suggesting that editing of both maternal and paternal germ cells can occur from a single injection (Friedland et al. 2013; Kim et al. 2014; Zhao et al. 2014; Paix et al. 2015; Wang et al. 2018). In addition, PCR amplification of heterozygous animals may not amplify large deletions if the deletions affect primer binding sites (Katic and Großhans 2013; Kim et al. 2014; Dokshin et al. 2018; Wang et al. 2018), further complicating quantification of genome editing rates by providing an inaccurate picture of editing nature (HDR or indel) and frequency. Overall, edited F1 animals can carry maternal genome edits, paternal genome edits, or both. Unless both maternal and paternal haplotypes can be analyzed separately, it can be difficult to determine whether one or two independent genome editing events may have occurred, complicating quantification of genome editing rates.
Therefore, we first aimed to develop a new method for quantifying genome editing rates that would allow separate analysis of each parental haplotype (Figure 1A). We used a coediting (co-CRISPR) approach, using two tightly linked genes that were simultaneously targeted using two different guide RNAs (Figure 1A). An advantage of coediting is that the editing of one locus ensures that Cas9 was active and available to target the second locus, which can at least partially normalize injection efficiencies across different injections (Kim et al. 2014; Paix et al. 2015). We chose to edit the tra-2 gene, which is located on chromosome II, 0.16 map units away from the commonly used co-CRISPR gene dpy-10. The dpy-10(cn64) variation results in a semi-dominant phenotype that is easily visualized on a stereomicroscope (Arribere et al. 2014; Paix et al. 2015). Animals homozygous for the dpy-10(cn64) mutation have a dumpy phenotype marked by a reduced body length whereas heterozygous animals have a normal body length but display an abnormal rolling behavior (Figure 1A, Arribere et al. 2014; Paix et al. 2015).
Due to their close proximity, meiotic recombination between tra-2 and dpy-10 is only expected in 1/625 haplotypes. Therefore, the haplotype arrangement of tra-2 and dpy-10 alleles will be stably maintained across generations (Figure 1A). As F1 generation roller animals contain a single dpy-10(cn64)-marked (Dpy-marked) chromosome and one chromosome that is not Dpy-marked, we were able to distinguish the two parental haplotypes and determine whether either or both haplotypes carried an edited allele of tra-2 (Figure 1A). Segregation of dumpy and nondumpy animals in the F2 generation homozygoses for each F1 generation haplotype, which allowed us to definitively determine whether one or two genome editing events had taken place in the F1 generation by sequencing the tra-2 genomic locus (Figure 1A).
Single nucleotide blocking mutations in the guide-binding region allow for effective genome editing
To determine whether single nucleotide substitutions in the guide-binding region effectively block Cas9 and allow for recovery of genome-edited animals, we designed a series of blocking mutations within a guide-binding region located in the 5′ UTR of tra-2 (Figure 2A). Because previous reports have suggested that substitutions proximal to the 3′ end of the guide are more effective at blocking Cas9 (Jinek et al. 2012; Cong et al. 2013; Fu et al. 2013; Hsu et al. 2013; Mali et al. 2013; Pattanayak et al. 2013; Zhang et al. 2015), we introduced substitutions every three nucleotides to test the positional effects of single nucleotide blocking mutations (Figure 2A). We named each mutation according to its position (“P”) relative to the 3′ end of the guide sequence (Figure 2A). For example, we refer to the mutation affecting the second nucleotide from the 3′ end of the guide sequence as “P2” and the mutation affecting the twentieth nucleotide from the 3′ end as “P20” (Figure 2A). As a control, we designed a mutation within the PAM domain, which is expected to completely block Cas9 activity (Mojica et al. 2009; Marraffini and Sontheimer 2010; Jinek et al. 2012; Sashital et al. 2012). Each of the repair templates were designed to also include a nonblocking, single nucleotide substitution downstream of the PAM sequence that introduces an RsaI restriction enzyme cutting site (Figure 2A). As an additional control, we designed a repair template that only included the RsaI cutting site, which should not block Cas9 activity. This repair template would therefore not be expected to allow for HDR editing since Cas9 should continue targeting the repaired DNA.
We performed co-CRISPR of dpy-10 and tra-2 for each of the blocking conditions that we designed (Figure 2A). We used ssODN repair templates, which themselves are not subject to Cas9 cleavage and effectively promote genome editing when injected alongside preassembled Cas9 RNP complexes in C. elegans (Paix et al. 2015). Following injection, we singled F1 generation roller animals from “jackpot” broods containing the highest percentage of F1 generation dpy-10-edited animals (Paix et al. 2015). As rollers are heterozygous for the dpy-10(cn64) variation, F2 progeny are a mixture of homozygous dumpy animals, heterozygous roller animals, and homozygous nondumpy animals (Figure 1). Because tra-2 and dpy-10 are genetically linked, F2 generation dumpy and nondumpy animals will each be homozygous for different alleles of tra-2 (Figure 1). We then sequenced the tra-2 genomic locus of the F2 animals to determine whether a single haplotype or both haplotypes had been edited. We scored the frequency of HDR editing and the frequencies of insertion or deletion (indel) mutations (Figure 2, B and C, Supplementary Table S3). In some cases, we observed alleles containing both indel mutations and HDR edits, which could result if Cas9 recuts an HDR-edited allele thereby preventing HDR. Because accurate HDR was not achieved, we classified alleles containing both indels and HDR edits as indels and not HDR-edited. We also quantified the number of animals that did not show any apparent editing of tra-2 (unedited), which could represent alleles that were never targeted by Cas9 or alleles that were targeted by Cas9 but repaired back to the wild-type genomic sequence without incorporating any of the designed edits (Figure 2, B and C, Supplementary Table S3). We defined HDR-edited animals as any animal that incorporated any of the designed mutations, regardless of whether partial or complete repair had occurred. As the tra-2 guide RNA targets the 5′ UTR of tra-2, relatively small indel mutations might lead to loss of tra-2 function. Loss of tra-2 is not lethal but results in masculinization of hermaphrodite animals (Hodgkin and Brenner 1977; Doniach 1986), which allowed us to recover deleterious mutations such as indels.
In F2 generation dumpy animals, we observed nearly complete HDR editing of tra-2 when the PAM was mutated (95.4% HDR-edited, Figure 2B, Supplementary Table S3), which is consistent with PAM mutations blocking Cas9 activity (Mojica et al. 2009; Marraffini and Sontheimer 2010; Jinek et al. 2012; Sashital et al. 2012). We observed similarly high HDR-editing rates in F2 generation dumpy animals when the PAM was mutated alongside an additional single nucleotide mismatch in the guide-binding region such as P2+PAM (93.8% HDR-edited), P11+PAM (88.6% HDR-edited) or P20+PAM (89.5% HDR-edited, Figure 2B, Supplementary Table S3). Although the efficiency was slightly reduced compared to the PAM blocking conditions, we found that the P2 (80.0% HDR-edited), P5 (75.8% HDR-edited), P8 (67.6% HDR-edited), P11 (75.7% HDR-edited), P14 (59.1% HDR-edited), P17 (54.8% HDR-edited), and P20 (66.6% HDR-edited) single nucleotide guide substitutions allowed for effective HDR editing in F2 generation dumpy animals (Figure 2B, Supplementary Table S3). While we used dpy-10 coconversion to at least partially normalize injection efficiencies (Kim et al. 2014; Paix et al. 2015), we cannot rule out the possibility that small differences in editing efficiencies observed under different blocking conditions could result from variation of injection efficiencies across each individual injection. Nevertheless, the position of the single nucleotide substitutions within the guide-binding region may influence the frequency of recovering HDR-edited animals, as substitutions closer to the 3′ end of the guide sequence tended to be more effective compared to substitutions near the 5′ end of the guide sequence (Figure 2B, Supplementary Table S3). For example, the P2 (80.0% HDR-edited), P5 (75.8% HDR-edited), and P8 (67.6% HDR-edited) substitutions located within the 3′ half of the guide sequence averaged significantly higher HDR-editing rates (74.5 ± 6.3% average HDR-edited) compared to the P14 (59.1% HDR-edited), P17 (54.8% HDR-edited), and P20 (66.6% HDR-edited) substitutions that are located in the 5′ half of the guide sequence (60.2% average HDR-edited, P < 0.05). This positional effect of guide substitutions was not surprising given that previous studies have demonstrated that mismatches near the 3′ end of the guide sequence are more effective at blocking Cas9 than mismatches near the 5′ end of the guide sequence (Jinek et al. 2012; Cong et al. 2013; Fu et al. 2013; Hsu et al. 2013; Mali et al. 2013; Pattanayak et al. 2013; Zhang et al. 2015).
Interestingly, for all blocking conditions examined, we noticed a substantial reduction in HDR editing efficiency of the tra-2 locus in F2 generation nondumpy animals (Figure 2C, Supplementary Table S3) compared to their dumpy siblings (Figure 2B, Supplementary Table S3). For example, whereas the PAM blocking condition resulted in very high HDR-editing rates in F2 generation dumpy animals (95.4% HDR-edited, Figure 2B, Supplementary Table S3), the rate of HDR editing was much lower in their nondumpy siblings (6.8% HDR-edited, Figure 2C, Supplementary Table S3). Although all single nucleotide substitutions in the guide-binding region had lower HDR editing rates in F2 generation nondumpy animals compared to dumpy animals, we did not observe a strong correlation between the position of substitutions within the guide sequence and the efficiency of HDR editing for nondumpy animals (Figure 2C, Supplementary Table S3), which is in contrast to the positional effects that we observed in their dumpy siblings (Figure 2B, Supplementary Table S3). While the substitutions in the 3′ half of the guide sequence (P2, P5, and P8) resulted in recovery of more HDR-edited animals than substitutions in the 5′ half of the guide sequence (P14, P17, and P20) in dumpy animals (Figure 2B, Supplementary Table S3), the same substitutions in the 3′ half of the guide sequence resulted in a similar HDR-editing rate (8.67 ± 4.62% average HDR-edited) as substitutions in the 5′ half of the guide sequence (7.67 ± 6.41% average HDR-edited, P = 0.83) for the F2 generation nondumpy siblings (Figure 2C, Supplementary Table S3). Importantly, we were able to recover HDR-edited animals for all blocking mutations that we tested in F2 generation dumpy and nondumpy animals, showing that single nucleotide guide substitutions are sufficient to allow for effective HDR-editing in C. elegans and can be used as an alternative to PAM mutations when silently mutating the PAM is not possible.
We also quantified the frequency of indel mutations that occurred in the tra-2 locus for each blocking condition that we tested (Figure 2, B and C, Supplementary Table S3). Because indels often result from NHEJ repair pathways, the presence of indels might suggest that NHEJ had been favored over HDR, which might be expected under conditions where Cas9 was not completely blocked. Consistent with this idea, we observed low indel rates in F2 generation dumpy animals when the PAM was mutated (2.3% indels) or when the PAM was mutated alongside an additional mutation in the guide sequence: P2+PAM (0.0% indels), P11+PAM (5.7% indels), or P20+PAM (0.0% indels, Figure 2B, Supplementary Table S3). By comparison, we observed slightly increased indel rates when using single nucleotide blocking mutations within the guide sequence in F2 generation dumpy animals (8.1 ± 4.6% indels, P < 0.05, Figure 2B, Supplementary Table S3). Furthermore, the position of the substitutions within the guide sequence appears to influence the frequency of indels observed under each blocking condition (Figure 2B, Supplementary Table S3). For example, we observed a low indel rate for the P2 (2.0% indels) and P5 (1.5% indels) substitutions, located near the 3′ end of the guide sequence, whereas we observed increased indel frequency P17 (11.8% indels) and P20 (9.6% indels) substitutions that are closest to the 5′ end of the guide sequence (Figure 2B, Supplementary Table S3). Thus, the expected blocking efficiency of each mutation appears to inversely correlate with the frequency of indel mutations under that blocking condition. In most cases, the indel rate in F2 generation nondumpy animals was increased compared to the respective blocking mutations in their dumpy siblings (Figure 2, Supplementary Table S3) For example, while the P2 mutation led to a 2.0% indel rate in dumpy animals, this rate increased to 18.0% in nondumpy animals (Figure 2, Supplementary Table S3). This suggests that nondumpy animals are biased toward NHEJ-dependent repair pathways compared to their dumpy siblings. This observation suggests that different mechanisms might influence repair of the dpy-10-edited haplotypes compared to the haplotypes that are not edited at the dpy-10 locus.
We were also able to recover HDR-edited animals when a blocking mutation was not incorporated into the repair template. We found that 21.8% of dumpy and 2.3% of nondumpy F2 generation animals carried HDR-edited mutations (Figure 2, Supplementary Table S3), with HDR-editing assessed by the presence of the nonblocking RsaI restriction site. Although the HDR-editing rates under the no-blocking condition appeared to be substantially reduced compared to conditions that introduced a blocking mutation (Figure 2, Supplementary Table S3), it was unexpected that any HDR-edited animals could be recovered given that Cas9 was not blocked from targeting the repaired DNA. We defined HDR editing for the no-blocking condition as the incorporation of the RsaI restriction site that is located on the 3′ side of the PAM sequence, although it is possible that HDR had occurred without introducing the RsaI restriction site. In support of this idea, HDR using ssODN donor molecules has been shown to favor repair in one direction of the PAM depending on which strand the ssODN donor is complementary to Farboud et al. (2019). It is therefore possible that our analysis underestimated the frequency of HDR repair when a blocking mutation was not included. These findings suggest that blocking mutations do not appear to be absolutely required to recover HDR-edited animals in C. elegans, suggesting that a temporal block might restrict Cas9 from continuing to target the genome after HDR.
We next examined whether there was a correlation between the tra-2 genotypes of dumpy and nondumpy F2 generation sibling animals that originated from the same F1 hermaphrodite animal (Figure 2D, Supplementary Table S4). Under all blocking conditions that we examined, the majority of F2 generation dumpy animals were HDR-edited, whereas their nondumpy siblings were usually not edited (Figure 2D, Supplementary Table S4). We did not observe any other strong correlations between the two alleles, suggesting that editing of one allele does not affect the probability of editing for the other allele (Figure 2D, Supplementary Table S4). When no blocking mutation was introduced into the repair template, we found that the most frequently observed combination of genotypes was that both dumpy and nondumpy sisters were not edited (Figure 2D, Supplementary Table S4). Thus, editing of each tra-2 haplotype, one each of maternal and paternal origin, appears to occur independently.
The position of single nucleotide blocking mutations influences the completeness of HDR
Although we observed high HDR-editing rates among all of the blocking conditions that we tested, we found that partial HDR-dependent repair often occurred, where only a subset of the designed mutations was incorporated into the genome (Figure 3A, Supplementary Table S5). For example, in some cases, the blocking mutation was incorporated while the RsaI restriction site had not been edited and vice versa (Figure 3A, Supplementary Table S5). We next asked whether the position of blocking mutations influenced the efficacy of HDR (Figure 3A, Supplementary Table S5). For the purpose of this analysis, we only considered genotypes where partial or complete HDR-editing had occurred (Figure 3A, Supplementary Table S5). We quantified the percentage of HDR-edited genotypes that contained a designed blocking mutation, the RsaI restriction site, or both blocking and RsaI mutations (Figure 3A, Supplementary Table S5). We found that blocking conditions where the PAM was mutated led to increased incorporation of both the blocking mutation and RsaI restriction site among the HDR-edited animals (Figure 3A, Supplementary Table S5). When only the PAM was mutated, the majority of HDR-edited chromosomes contained both the PAM blocking mutation and the RsaI restriction site (97.8% both mutations), while a small percentage contained only the RsaI restriction site (2.2% RsaI only) (Figure 3A, Supplementary Table S5). Interestingly, all HDR-edited chromosomes that we examined for the P2+PAM blocking condition were edited for both the blocking mutation and the RsaI restriction site (100% both mutations), while the P2 blocking condition itself resulted in less frequent incorporation of both blocking mutations and the RsaI restriction site (36.2% both mutations) and commonly resulted in partial repair of only the blocking mutation (31.9% P2 only) or the RsaI restriction site (31.9% RsaI only) (Figure 3A, Supplementary Table S5). We observed a similar trend for the P11+PAM or P20+PAM blocking conditions compared to the P11 or P20 blocking conditions respectively, where complete HDR of both blocking mutations and the RsaI restriction site was more frequent when the PAM was also mutated (Figure 3A, Supplementary Table S5). While we cannot be certain why the presence of PAM blocking mutations led to increased frequency of complete HDR repair compared to blocking mutations in the guide sequence, we hypothesize that partial HDR repair could reflect incomplete blocking of Cas9. This, in turn, could lead to Cas9 recutting of repaired DNA and increased frequency of inaccurate HDR as repair is attempted multiple times. We also observed that HDR appeared to occur in an asymmetric fashion, favoring blocking mutations over RsaI site incorporation. For example, when P20 was used as a blocking mutation we observed more frequent incorporation of the P20 mutation (87.5% P20 edited) than the RsaI restriction site (46.9% RsaI edited), despite RsaI being located closer to the expected Cas9-generated break site (Figure 3A, Supplementary Table S5). This observation suggests that edits that do not block Cas9 from recutting (such as introduction of the RsaI site) may result in multiple cleavage and repairs, thus ultimately favoring repair events that incorporate the blocking mutations. However, the presence of RsaI edits that do not include blocking mutations further supports the existence of a temporal block that prevents Cas9 from recutting the repaired template.
Since previous reports have shown that the efficiency of HDR editing is inversely correlated with distance to the Cas9-generated break site (Arribere et al. 2014; Inui et al. 2014; Paix et al. 2014, 2015; Ward 2015), we asked whether the distance of blocking mutations from the break site might be confounded with their actual ability to block Cas9. In other words, the P2 blocking mutation might increase the frequency for recovery of HDR-edited animals compared to the P20 blocking mutation because it is closer to the dsDNA break site and is therefore more likely to incorporate during repair, rather than the P2 mutation blocking Cas9 more effectively than P20. To address this, we examined how frequently partial or complete HDR occurred when using repair templates that mutated one nucleotide within the guide sequence as well as the PAM domain to ensure that Cas9 was equally blocked under each condition (P2+PAM, P11+PAM, and P20+PAM) (Figure 3B, Supplementary Table S6). When both the PAM and P2 mutations were introduced, we found that all of the edited chromosomes that we examined contained both PAM and P2 mutations (100% both edited) (Figure 3B, Supplementary Table S6). However, when the PAM was mutated alongside the P11 or P20 mutations, we found that partial HDR repair had often occurred. In particular, we found that the majority of HDR-edited chromosomes contained only the PAM blocking mutation when either P11+PAM (72.2% PAM only) or the P20+PAM (71.8% PAM only) repair templates had been used (Figure 3B, Supplementary Table S6). Only a small fraction of HDR-edited chromosomes contained both the guide blocking mutation and the PAM mutation for the P11+PAM (16.6% both edited) or P20+PAM (28.2% both edited) mutations (Figure 3B, Supplementary Table S6). Therefore, the distance of blocking mutations from the dsDNA break site, which is located near the 3′ end of the guide sequence, appears to influence the rate at which they are incorporated via HDR. Blocking mutations located farther away from the dsDNA break site may not always be incorporated through HDR, leading to an overall decrease in HDR-editing efficiency since Cas9 can re-target the partially repaired chromosome. Despite the reduced incorporation of the more distant blocking mutations, the overall high efficiency of obtaining desired edits (Figure 2) strongly supports practical use of single nucleotide blocking mutations when PAM mutations are not possible.
Single nucleotide guide substitutions effectively, although not completely, block Cas9
Having determined that single nucleotide blocking mutations allow for recovery of HDR-edited animals (Figure 2), we next asked how effective each of blocking mutations might be in preventing Cas9 from targeting the tra-2 genomic locus after HDR occurred. To address this question, we took advantage of the tra-2 HDR-edited animals that we generated (Figure 2), containing blocking mutations as well as the RsaI restriction enzyme cutting site. We reasoned that if the blocking mutations built into the tra-2 locus prevent Cas9 from targeting the mutated sequences, then re-injection of Cas9 and the same guide RNA, perfectly matched to the wild-type tra-2 genomic sequence, should not lead to editing of the mutated tra-2 loci (Figure 5A). To examine whether re-editing of tra-2 could occur under any of the blocking conditions, we designed a repair template that would revert tra-2 back to the wild-type genomic sequence (Figure 4A). Because the HDR-edited strains contain blocking mutations and the RsaI restriction enzyme site, reversion of tra-2 back to the wild-type sequence will result in removal of the RsaI restriction enzyme site (Figure 4A). Furthermore, as the RsaI restriction enzyme site is located immediately downstream of the tra-2 PAM (Figure 2A), indel mutations might also be expected to disrupt the RsaI site. On the other hand, if Cas9 is completely blocked from generating a dsDNA break at the mutated tra-2 locus, then all of the F1 progeny postinjection should still contain the RsaI restriction enzyme site. Importantly, any animal now lacking the RsaI site must have been targeted by Cas9 for genome editing to have occurred, which would indicate that Cas9 was not completely blocked under that condition. To test whether the position of blocking mutations within the guide sequence influenced their blocking efficacy, we examined the RsaI reversion efficiencies of the P2, P11, and P20 mutations (Figure 4A). As a positive control, we reverted the RsaI site in a strain that did not contain any Cas9 blocking mutations (“no-block”), which would still be expected to be targeted by Cas9 loaded with the wild-type guide (Figure 4A). Because the PAM is absolutely required for Cas9 activity (Mojica et al. 2009; Marraffini and Sontheimer 2010; Jinek et al. 2012; Sashital et al. 2012), we attempted to revert the RsaI site in PAM-edited animals as a negative control (Figure 4A).
We used dpy-10 coconversion to enrich for genome-edited animals and examined whether F1 generation roller and dumpy animals contained the tra-2 RsaI restriction site (Figure 4B, Supplementary Table S7). We calculated the haplotype RsaI reversion rate by dividing the number of reverted haplotypes (one for heterozygous reverted animals and two for homozygous reverted animals) by the total number of haplotypes that were examined. As expected, we observed robust reversion of the RsaI restriction site in the no-blocking control and never observed reversion of the RsaI site in PAM-edited animals (Figure 4B). For the no-blocking control, we found that the RsaI reversion rates of F1 generation dumpy (32% haplotypes reverted) and roller (25% haplotypes reverted) animals were similar, although we only observed homozygous-reverted animals in F1 generation dumpy animals (Figure 4B, Supplementary Table S7). The increased prevalence of homozygous-reverted animals in F1 generation dumpy animals might not be surprising, since the dumpy phenotype indicates that homozygous editing of the dpy-10 genomic locus had also occurred. We found that the P2 blocking condition was highly effective at blocking Cas9, as only a small percentage of F1 generation dumpy (2.7% haplotypes reverted) or roller (0.9% haplotypes reverted) animals had reverted the RsaI site (Figure 4B, Supplementary Table S7). Similarly, the P11 mutation was highly effective at blocking Cas9 and resulted in only a low frequency of RsaI reversion in F1 generation dumpy (2.7% haplotypes reverted) and roller (2.7% haplotypes reverted) animals (Figure 4B, Supplementary Table S7). Despite increased RsaI reversion frequency in P20 blocking mutants for both F1 generation dumpy (12.8% haplotypes reverted) and roller (6.4% haplotypes reverted) animals (Figure 4B, Supplementary Table S7), the P20 blocking mutation blocked Cas9, as the RsaI reversion rates were much lower in P20 mutants compared to no-block controls (Figure 4B, Supplementary Table S7). Therefore, consistent with previous reports (Jinek et al. 2012; Cong et al. 2013; Fu et al. 2013; Hsu et al. 2013; Mali et al. 2013; Pattanayak et al. 2013; Zhang et al. 2015), the position of single nucleotide guide substitutions appears to influence their blocking efficacy where substitutions located proximal to the 3′ end of the guide are more effective at blocking Cas9 (Figure 4B, Supplementary Table S7). Interestingly, although we observed similar HDR editing rates for the P11 (75.7% HDR-edited in Dumpy animals, Figure 2, Supplementary Table S3) and P20 mutations (66.6% HDR-edited in Dumpy animals, Figure 2, Supplementary Table S3), P11 was more effective at blocking Cas9 than P20 (Figure 4B, Supplementary Table S7). This discrepancy between blocking efficacy and HDR editing rates further supports the idea that perhaps a temporal effect contributes toward allowing for robust genome editing when Cas9 is not completely blocked. Collectively, these findings show that single nucleotide substitutions in the guide sequence effectively, albeit not completely, block Cas9 after HDR occurs, and that conditions where Cas9 is not completely blocked still allow for efficient genome editing in C. elegans.
Editing of maternal haplotypes occurs at greater frequency than editing of paternal haplotypes
Given that we observed significantly higher HDR-editing rates of the tra-2 locus in dpy-10-edited haplotypes compared to non-dpy-10 haplotypes (Figure 2, B–D), we asked whether there could be differences in the HDR-editing efficiencies of each parentally contributed haplotype. To determine the HDR editing rates of each parental haplotype, we used a mating-based approach to differentiate between maternal and paternal haplotypes (Figure 5A). We crossed wild-type males to tra-2 mutant hermaphrodites that carried the RsaI restriction site just upstream of the tra-2 coding sequence (Figure 4A). We then injected the mated animals with Cas9 RNP complex to generate the dpy-10(cn64) variation (Figure 5A). Because dpy-10 and tra-2 are genetically linked and not expected to independently assort, the resulting F1 generation roller animals will have two possible haplotype arrangements for the dpy-10 and tra-2 mutations (Figure 5A). One possible F1 genotype will have the maternally contributed tra-2 mutation (RsaI site) on the same chromosome as the newly introduced dpy-10(cn64) mutation and another possible F1 genotype would have the tra-2 and dpy-10 mutations on different chromosomes (Figure 5A). Both of these possible haplotype arrangements can be differentiated in the F2 generation by examining whether the dumpy or nondumpy progeny contain the tra-2 RsaI restriction site (Figure 5A). If maternal editing of dpy-10 occurs, then the F2 generation dumpy animals would be homozygous for the tra-2 RsaI allele whereas the nondumpy F2 generation animals would be homozygous for the paternally provided wild-type tra-2 allele (Figure 5A). If editing occurred in the paternal genome, then the F2 generation dumpy animals would be homozygous for the wild-type tra-2 allele that lacks the RsaI cut site (Figure 5A). We genotyped one F2 generation dumpy and one F2 generation nondumpy animal per F1 generation roller animal (n = 44 animals). In all F2 generation animals that we examined, RsaI digestion occurred only in dumpy animals, supporting the idea that editing of the maternal genome is strongly preferred over paternal genome editing (Figure 5, B and C). Thus, we conclude that the distinct tra-2 HDR rates we previously observed in dumpy vs nondumpy F2 generation animals (Figure 2, B and C) are likely due to differences in the editing efficiencies of each paternal genome, where dumpy animals are likely homozygous for the maternally provided haplotype and nondumpy animals are likely homozygous for the paternally provided haplotype.
Selection of maternally provided haplotypes using balancer chromosomes to mutate the let-7 miRNA in a scarless fashion
As we observed increased HDR editing efficiency of maternally provided genomes compared to paternally provided genomes (Figure 5), we devised a strategy to select for the maternally provided haplotypes postinjection (Figure 6A). Our approach was to label paternally provided chromosomes prior to injection by mating wild-type hermaphrodite animals to males expressing fluorescently labeled balancer chromosomes (Figure 6A). Importantly, the use of balancer chromosomes restricts recombination between the maternally provided wild-type chromosome and the paternally provided balancer chromosome. This allows for segregation of each parental haplotype in subsequent generations and easily homozygoses for the desired edit, thereby reducing screening efforts postinjection (Figure 6A). We used this strategy, along with single nucleotide blocking mutations to mutate the let-7 miRNA in a scarless fashion. As loss of let-7 function is lethal (Reinhart et al. 2000), this strategy also allowed us to immediately maintain deleterious let-7 mutations in a balanced, heterozygous genetic background, using the tmc24 balancer chromosome (Dejima et al. 2018). Because the tmc24 balancer contains a pmyo-2::Venus fluorescent marker that is pharyngeal expressed (Dejima et al. 2018), F1 generation cross-progeny should be fluorescently labeled whereas self-progeny would not be labeled (Figure 6A). In the subsequent F2 generation, animals without expression of pharyngeal Venus should be homozygous for the maternally provided let-7 haplotype as the non-Venus animals lack the paternally provided tmc24 chromosome (Figure 6A).
We targeted let-7 using a guide sequence overlapping with the mature let-7 miRNA sequence and used ssODN repair templates to introduce single nucleotide blocking mutations into the endogenous let-7 locus. As a proof of principle, to demonstrate our ability to create scarless edits within nonprotein-coding portions of the genome using a single nucleotide mismatch within the guide region, we aimed to generate a P6 blocking mutation, which recapitulates the classical let-7(n2853) hypomorphic mutation. let-7(n2853) disrupts the let-7 seed sequence and leads to dysregulation of let-7 mRNA targets in a temperature sensitive manner (Reinhart et al. 2000; Vella et al. 2004) (Figure 6B). We also designed two nonseed mutations as controls: one located within the PAM domain that is expected to completely block Cas9 and a second nonblocking mutation located downstream of the PAM (Figure 6B). We used dpy-10 coconversion to enrich for genome-edited animals and singled both F1 generation roller and dumpy animals (Figure 6A). We then sequenced non-Venus F2 generation animals to examine how each blocking condition affected HDR efficiency (Figure 6A).
We observed high HDR-editing rates (60.7% HDR-edited) and low indel rates (10.7% indels) when the PAM was mutated (Figure 6C). We observed comparable rates of HDR-editing (56.7% HDR-edited) when P6, which recapitulates the let-7(n2853) mutation (Figure 6B), was used as the blocking mutation (Figure 6C). Although the HDR-editing rates were similar for the PAM and P6 blocking conditions, the indel rate was twice as high for the P6 blocking mutation (23.3% indels) compared to the PAM mutation (10.7% indels) (Figure 6C). The increased indel rate observed under P6 blocking conditions might indicate that P6 does not completely block Cas9. Consistent with this idea, we found that the no-blocking condition led to high frequency of indel mutations (40.7% indels) and low rate of HDR-editing (3.7% HDR-edited) (Figure 6C).
Interestingly, we did not observe a similar increase in indel rates for the tra-2 no-blocking condition (Figure 2, B and C), suggesting that there may be gene-specific or guide-specific differences in indel rates vs HDR-editing rates. Importantly, we were able to isolate a small percentage of let-7 HDR-edited animals, even when the ssODN repair template did not contain a blocking mutation (Figure 6C). This observation supports the idea that in situations where no blocking mutations can be designed, desired edits can nonetheless be obtained, albeit at a low frequency.
Many microRNAs, including let-7, are members of microRNA families that share the same seed sequence and are therefore expected to target similar mRNA sequences (Lewis et al. 2003; Lim et al. 2003). As a result of sharing the same seed sequence, many members of microRNA families often exhibit functional redundancy with other family members (Abbott et al. 2005; Miska et al. 2007; Alvarez-Saavedra and Horvitz 2010). As a further proof of principle, to demonstrate the utility of selecting for maternal chromosomes via paternally provided balancers, we aimed to recapitulate the let-7(n2853) variation in a genetic background devoid of the other three major let-7 family members: miR-48, miR-84, and miR-241. Such strain has been difficult to generate using conventional genetic methods, since the let-7 and miR-84 microRNAs are genetically linked and complete loss of let-7 activity is lethal (Reinhart et al. 2000). To generate a strain containing mutations in all four let-7 family members, we crossed tmc24 balancer males to hermaphrodite animals containing deletions of mir-48, mir-84, and mir-241 (Figure 6D). We then performed co-CRISPR of let-7 and dpy-10 to recapitulate the let-7(n2853) mutation (Figure 6D). As let-7 and mir-84 are both located on the X chromosome, F1 generation cross-progeny will have a paternally provided tmc24 chromosome and a maternally provided chromosome that contains a mir-84 deletion, which we targeted for the let-7 editing (Figure 6D). We then sequenced F2 generation non-Venus animals and identified animals containing maternal let-7(n2853) mutations. Using this strategy, we were able to create a stable strain that contains homozygous deletions in mir-48 and mir-241 (nDf51) and the mir-84(n4037) deletion and let-7(n2853-equivalent) mutation balanced by tmc24 in a heterozygous state.
Collectively, these findings demonstrate that single nucleotide substitutions within the guide RNA targeting sequence can be used to effectively mutate miRNAs through HDR in an otherwise scarless fashion. Furthermore, we propose that balancer chromosomes can be used to select for maternally provided haplotypes and introduce deleterious mutations directly into a balanced heterozygous background in a single injection step.
The use of dpy-10(cn64) coinjection may further facilitate the single-step injection approach, since F1 generation rollers containing a single dpy-10(cn64) edited chromosome are more likely to only contain a single edit at a second locus. Previous strategies used in C. elegans to introduce potentially lethal genome edits directly into balanced genetic backgrounds have relied on two-step editing approaches, where the first edit introduces a nascent PAM to a gene of interest that can be specifically edited in a second injection after being crossed to a balancer chromosome that lacks the nascent PAM (Dejima et al. 2018; Duan et al. 2020). As nearly 90% of the C. elegans genome is covered by balancer mutations (Dejima et al. 2018), our strategy can be used to target most C. elegans genes in a single step editing approach, eliminating the need for extensive postinjection screening.
Discussion
Strategies for designing Cas9 blocking mutations in Caenorhabditis elegans
In this study, we performed a detailed analysis of a single guide-target pair to determine the blocking efficacy of single nucleotide substitutions within the guide region of the donor molecule. As we only tested a single guide-target pair, it is possible that some of our conclusions do not apply to all guide-target interactions. However, our findings that single nucleotide substitutions are sufficient to allow recovery of HDR-edited animals, and that blocking mutations are not strictly required for recovery of HDR-edited animals are also supported by our analysis of the let-7 locus. Furthermore, we have routinely used single nucleotide substitutions in the guide-binding region to generate additional genome-edited strains, supporting that this approach is broadly effective across several genetic loci (Supplementary Figure S1). Mutating the PAM domain remains the most effective way to block Cas9 and leads to the highest HDR editing rates (Figure 7A). However, when the desired mutation overlaps with the guide sequence, additional blocking mutations are not necessary to recover HDR edited animals, which can facilitate scarless HDR editing (Figure 7B). For protein-coding genes, it is also possible to introduce silent blocking mutations into the guide sequence, ideally close to the 3′ end of the guide, when silently mutating the PAM is not possible (Figure 7C). For edits that do not overlap with the guide sequence or PAM, it is possible to forego the use of blocking mutations in order to recover animals in an otherwise scarless genetic background (Figure 7D). The ability to recover HDR edited animals without including a blocking mutation suggests that a temporal block to Cas9 activity exists in C. elegans, preventing Cas9 recutting of the repaired genomic region (Figure 7E). Cas9 activity might be highest in the distal end of the maternal germline, since this is where Cas9 RNP complexes are injected (Figure 7E). The temporal block to Cas9 activity might result from degradation of the injected Cas9 RNP complexes or dilution of Cas9 activity as germ cells passage through the maternal germline from distal end to proximal end (Figure 7E). Collectively our findings expand the repertoire of possible genome edits in C. elegans and should facilitate analysis of noncoding regulatory sequences without the need for extraneous Cas9 blocking mutations.
Differences in maternal vs paternal genome editing rates in C. elegans
We found that the HDR editing rates of tra-2 were much higher for the haplotypes that contained the dpy-10(cn64) allele compared to the haplotypes that were not edited for dpy-10 (Figure 2). Given that the Cas9 RNPs are injected into the maternal germline of hermaphrodite animals, it seemed likely that editing of the maternally provided haplotype is preferred over editing of the paternally provided haplotype. Indeed, several others have speculated that editing of maternal haplotypes is preferred in C. elegans (Arribere et al. 2014; Kim et al. 2014; Paix et al. 2014, 2016), although both paternal and maternal germ cells are competent for HDR (Clejan et al. 2006). We used a mating-based approach that allowed us to quantify the editing rates of maternal and paternal haplotypes and demonstrated that maternal editing was preferred over paternal editing (Figure 5). This is in contrast to a recent study, which suggested that paternal genome (embryonic) editing is preferred over maternal editing (Farboud et al. 2019). A key difference in our experimental design was that we were able to assess maternal vs paternal genome editing in a single injection step, whereas previous studies have assessed maternal and paternal editing in separate injections (Farboud et al. 2019). Farboud and colleagues performed two injections, where each injection was designed to specifically target one parental haplotype. However, the allele-specific editing was predicated on the assumption that a single nucleotide mismatch in the guide-binding region (equivalent to “P2”) was sufficient to completely block Cas9, which our findings suggest is not accurate. Furthermore, Farboud et al. used a much higher concentration of Cas9 (15 µM) than we used in this study (2.65 μM), which could have allowed Cas9 to persist longer in the C. elegans germline and lead to more effective editing of paternally provided haplotypes. Nevertheless, it is also possible that there could be gene-specific differences in haploid genome editing efficiency.
Taking advantage of the preference for maternal editing, we were able to select for the maternally provided chromosomes using balancer chromosomes that restrict recombination between the maternal and paternal chromosomes. By mating hermaphrodite animals to males containing a balancer chromosome before injection, the maternally provided nonbalancer chromosome of coedited F1 roller animals is more likely to be edited. Because typical balancer chromosomes contain a fluorescent or physical marker to identify animals harboring the balancer, nonmarked animals can be easily identified and should be homozygous for the maternally provided haplotype. Homozygosing for the edited, maternally provided chromosome would homozygose for the mutation of interest and therefore reduce molecular method-based screening. It would also be possible to recover edited animals by injecting directly into balanced, heterozygous animals without mating prior to injection. F1 generation animals could be screened for heterozygosity and would carry a mutation on either the balancer chromosome or the nonbalanced chromosome. If the edit was not on the balancer chromosome, F2 generation animals lacking the balancer chromosome would be homozygous for the desired edit. As most of the C. elegans genome is covered by balancer chromosomes (Dejima et al. 2018), this approach can be broadly applied to most C. elegans genes and has the added advantage of introducing potentially deleterious mutations directly into a balanced genetic background.
Single nucleotide substitutions are sufficient to allow recovery of HDR-edited animals in C. elegans
In this study, we demonstrate that a single nucleotide mismatch at any point in the guide sequence can block Cas9 and can be reliably used to recover HDR-edited animals. However, despite the fact that these single nucleotide blocking mutations are sufficient to allow for recovery of HDR-edited animals, we show that Cas9 was still able to target genomic sequences containing a single mismatch to its guide RNA, including the P2 substitution (Figure 4). These data provide direct evidence that off-target cutting can occur in C. elegans, in contrast to previous speculation that off-target cutting may not readily occur in C. elegans (Schwartz and Sternberg 2014; Xu 2015). Previous targeted approaches to identify potential off-target cutting by Cas9 did not identify bona fide off-target events (Chiu et al. 2013; Dickinson et al. 2013; Silva-García et al. 2019). Similarly, whole-genome sequencing-based approaches did not identify variants at predicted off-target sites and the overall rate of variant formation was not significantly higher than the spontaneous mutation rate (Paix et al. 2014; Waaijers and Boxem 2014; Au et al. 2018). Why have not previous approaches identified off-target cutting events in C. elegans? There are several contributing factors that might make identification of off-target cutting events difficult. We found that the efficiency of off-target cutting was much lower than on-target cutting (Figure 4), suggesting that off-target cutting events might be rare. Furthermore, off-target editing likely occurs in a heterozygous fashion, which can complicate detection in sequencing reactions since heterozygosity is rapidly lost in hermaphroditic organisms (Brenner 1974). Another factor could be the software used to design guide RNAs, many (or all) of which might not allow guides that carry single mismatches to other regions of the genome. Although we did not test whether multiple guide substitutions were more effective at blocking Cas9, previous studies have suggested that increasing the number of mismatches in the guide sequence leads to increased blocking of Cas9 (Jinek et al. 2012; Cong et al. 2013; Fu et al. 2013; Hsu et al. 2013; Mali et al. 2013; Pattanayak et al. 2013; Zhang et al. 2015). Thus, effective guide RNA design might essentially eliminate off-target cutting by Cas9. Our observation that off-target cutting can occur in C. elegans emphasizes the importance of careful guide design and backcrossing of genome-edited strains to remove potential unwanted mutations when off-target effects are suspected.
A temporal block to Cas9 activity appears to limit recutting of repaired DNA
Although we observe significantly reduced HDR editing rates when no blocking mutation was included in the ssODN repair template, the editing efficiency was high enough that we were able to reliably recover HDR-edited animals. Given the ability of Cas9 to target double-stranded repair templates, blocking conditions may remain critical for studies that use dsDNA repair templates. It is worth noting that many of the cases that might necessitate foregoing a blocking mutation, such as the editing of noncoding RNAs, would typically be small genomic changes that can be accomplished using ssODN repair templates.
The fact that we are able to recover HDR-edited animals without including a blocking mutation suggests that a temporal block to Cas9 activity exists in C. elegans, where the repaired region escapes repeated targeting by Cas9 (Figure 7E). What might lead to a temporal block to Cas9 activity? One possibility is that the reagents used for genome editing are not stable and/or targeted for degradation in the C. elegans germline, which would lead to reduced Cas9 activity over time. Consistent with this idea, RNP complexes are rapidly degraded (Kim et al. 2014; Liang et al. 2015; DeWitt et al. 2017; Prior et al. 2017; Farboud et al. 2019). Plasmid-expressed Cas9 may persist longer and might not result in the same temporal block that we observed for Cas9 RNP injection in this study. A second explanation for a temporal block might be that germ cells could become less receptive to Cas9 as they passage through the C. elegans germ line. Cas9 RNP complexes are injected into the syncytial maternal germline, where immature germ cells share a common cytoplasm and can all be targeted by a single injection (Evans 2006; Kadandale et al. 2008; Pazdernik and Schedl 2012; Hubbard and Schedl 2019). Many of the syncytial germ cells are in pachytene stage, during which time germ cells are receptive to homology directed repair pathways (Woglar and Jantsch 2014, McClendon et al. 2016). As germ cells mature, they could become less amenable to genome editing. Finally, it is also possible that genome editing reagents are diluted as they passage through the tubular-shaped maternal germline. This could explain the differences in editing efficiencies for maternal and paternal haplotypes, as the injection mix could become less available in the proximal germline. In any case, our findings suggest that Cas9 activity is attenuated over time, leading to a temporal block to Cas9 activity. This temporal block to Cas9 activity may then allow for effective HDR editing, even under conditions where Cas9 is not completely blocked. Additional work will be required to fully understand how a temporal block to Cas9 activity is established. It will also be interesting to see if a similar temporal block to Cas9 activity exists in other organisms, or if the unique germline architecture of C. elegans leads to reduced Cas9 activity over time.
Data availability
All strains and reagents used in this study are available upon request. The authors affirm that all necessary data to support the conclusions of this study are included within the manuscript.
Supplementary material is available at GENETICS online.
Acknowledgments
The authors thank current and former members of the Zinovyeva lab for valuable discussions and experimental assistance. They thank David Greenstein for providing helpful comments and suggestions. Some strains were provided by the CGC (Caenorhabditis Genetics Center), which is funded by the National Institutes of Health Office of Research Infrastructure Programs P40-OD010440.
Funding
This work was supported by a grant from the National Institute of General Medical Sciences (R35GM124828 to A.Y.Z.) and K-INBRE funds (P20 103GM418 to J.C.M. and J.T.S.).
Conflicts of interest
The authors declare that there is no conflict of interest.