Robust genome editing with short single-stranded and long, partially single-stranded DNA donors in C. elegans

CRISPR-based genome editing using ribonucleoprotein (RNP) complexes and synthetic single stranded oligodeoxynucleotide (ssODN) donors can be highly effective. However, reproducibility can vary, and precise, targeted integration of longer constructs – such as green fluorescent protein (GFP) tags remains challenging in many systems. Here we describe a streamlined and optimized editing protocol for the nematode C. elegans. We demonstrate its efficacy, flexibility, and cost-effectiveness by affinity-tagging all twelve of the Worm-specific Argonaute (WAGO) proteins in C. elegans using ssODN donors. In addition, we describe a novel PCR-based partially single-stranded “hybrid” donor design that yields high efficiency editing with large (kilobase-scale) constructs. We use these hybrid donors to introduce fluorescent protein tags into multiple loci achieving editing efficiencies that approach those previously obtained only with much shorter ssODN donors. The principals and strategies described here are likely to translate to other systems and should allow researchers to reproducibly and efficiently obtain both long and short precision genome edits.


Introduction 37
In theory, CRISPR/Cas9-based genome editing enables researchers to rapidly generate 38 designer alleles of any locus for genetic, cytological, or biochemical analyses. In 39 practice, however, we have found that the technology remains far from routine for many 40 users, especially in applications where long templated insertions are desired. Here we 41 describe a streamlined and optimized protocol for genome editing the nematode C. 42 elegans. We find that seemingly minor details, such as the order of donor DNA addition 43 to the editing mix, can have dramatic consequences for editing efficiency. We 44 demonstrate pronounced toxicity of RNP at high concentrations and provide a strategy 45 for optimizing RNP levels using a co-injected, easily scored reporter. Finally, we show 46 that generating hybrid, partially single stranded long DNA donor molecules dramatically 47 promotes templated repair for the insertion of longer edits such as GFP. Although, we 48 have only tested these strategies in C. elegans, it seems likely that the principals 49 revealed here including the following key features, will be relevant in other systems: 50 • Utilization of a robust, Cas9-independent co-injection marker to control for 51 injection quality, to optimize Cas9 RNP concentration, and to monitor toxicity. 52 • Pre-assembly of the RNP complexes before adding donor DNA to the injection 53 mixture. 54 • Employment of hybrid PCR-based donors with single-stranded homology arms 55 for consistent, high-efficiency insertion of large constructs. 56 Results 58

Cas9 RNPs are toxic at high concentrations 59
In the course of adopting Cas9 RNP editing methodologies (PAIX et al. 2015) we 60 decided to monitor injection quality by adding the rol-6(su1006) plasmid to the injection 61 cocktail (MELLO et al. 1991). We were very surprised to find that, despite giving high 62 numbers of edited progeny, the numbers of transgenic Roller (rol-6) animals were 63 greatly reduced. We noted that the few surviving Roller animals obtained were often 64 sick and sterile, suggesting that toxicity, or excessive genome editing might cause the 65 lack of Roller transgenics (data not shown). 66 To address these possibilities, we performed a titration of RNP concentrations 67 while holding the Roller DNA concentration constant. We then examined both the 68 genome editing efficiency and the frequency of Roller transgenics among F1 progeny of 69 the injected animals. Worms expressing the bright fluorescence marker GFP::GLH-1 70 were co-injected with 40ng/µl pRF4::rol-6(su1006) plasmid and dilutions of Cas9 RNPs 71 loaded with an anti-gfp guide ( Figure 1A). In our pilot studies we recovered very few 72 Rollers at 2.5µg/µl of Cas9 used in initial C. elegans Cas9 RNP protocols (CHO et al. down to 0.25µg/µl doubled the frequency of F1 rollers to 33, while a ten-fold dilution to 78 0.025µg/µl resulted in 43 F1 roller progeny per P0 ( Figure 1B). These latter two F1 roller 79 frequencies are comparable to rates reported for pRF4::rol-6(su1006) injected alone (MELLO et al. 1991). Taken together these findings suggest that RNP concentrations 81 below 0.25 µg/µl do not interfere with expression of the co-injected Roller marker gene. 82 We next asked how Cas9 RNP concentrations affected the in-del frequency at 83 the gfp::glh-1 locus. To measure in-del rates in a high throughput fashion we used the 84 TIDE analysis pipeline, which estimates the in-del rates in a mixture of PCR products 85 ( Figure 1A, left) (BRINKMAN et al. 2014). To do this we PCR amplified the gfp::glh-1 86 locus from pools of all F1 rollers segregated by a given injected P0 worm and subjected 87 the mixture to sanger sequencing and TIDE analysis. Using this approach we found that 88 at 0.025 µg/µl ~16% of alleles carried an in-del. The number of edited alleles increased 89 to ~80% at 0.25µg/µl ( Figure 1C), but did not increase further when the Cas9 90 concentration was doubled to 0.5µg/µl, and in fact appeared to decline slightly to ~67% 91 ( Figure 1C). Because GFP::GLH-1 is easily detectable in adult animals under the 92 florescent dissection microscope, we were able to validate the TIDE results directly 93 using microscopy ( Figure 1A, right). For example, we determined that at 0.25µg/µl of 94 Cas9 ~98% of all F1 rollers segregated GFP-negative (successfully edited) progeny 95 ( Figure S1A). Furthermore, ~68% were homozygous, producing only GFP negative 96 progeny, while another ~30% were heterozygous ( Figure S1B). Based on these 97 numbers we can calculate that 83% of all gfp::glh-1 alleles were successfully edited at 98 0.25µg/µl of Cas9 ( Figure S1C). These numbers correlate well with TIDE data (

Efficient editing with ssODN donors using a Roller plasmid co-injection marker 105
The above findings demonstrate that Roller plasmid co-injection identifies animals that 106 are highly likely to undergo CRISPR-induced DNA double strand breaks. We next 107 wished to test this methodology for achieving homology-directed repair (HDR). To do 108 this we decided to introduce a (3X)FLAG-affinity tag into each of the twelve worm-109 specific Argonautes (WAGOs). For each gene we designed guides targeting the PAM 110 site closest to the ATG start codon (without any further optimization or guide testing) 111  wago-1 and wago-2 are highly similar near the ATG and no specific guide could be 113 designed; thus, one guide targeting both loci was used.) Injection mixtures were 114 prepared by simultaneously mixing Cas9 with the guide RNA, plasmid and donor 115 ssODN, without pre-incubating to assemble the RNP ( Figure 2B). (Note: we now, 116 instead, recommend pre-assembling the RNP prior to adding the donor DNA and 117 plasmid (see below), as this simple change in the order of addition dramatically 118 improved repair with longer templates. Each mixture was then injected into ~10 P0 N2 119 worms using standard worm gonadal injection methodology (MELLO AND FIRE 1995). 120 Individual F1 Rollers were singled to plates and after producing broods were genotyped 121 for 3XFLAG insertions ( Figure 2B, See Appendix 2 for detailed protocol). 122 We were able to recover 11 out of 12 WAGO tagged strains among the first 24 123 F1 progeny screened by PCR from each set of injections, and in every case we 124 recovered multiple independent alleles. The average success rate for precise insertion of the 3XFLAG tag ranged from 10-73% and averaged ~34%. Plotting the insertion 126 efficiency versus the distance between the Cas9-induced cut and the desired insertion 127 site (directly after the ATG start codon [ Figure 2A]) we found no strong correlation up to 128 20 base pairs away ( Figure 2C). The wago-6 (sago-2) locus was the only outlier, likely 129 because the nearest available cut site for use with the original donor design was 27 130 bases away from the site of insertion. Although a number of insertions were recovered 131 at this locus they were either out of frame or contained random DNA sequences (data 132 not shown). The wago-6 gene contains a second PAM site located right at the ATG start 133 codon. This site was not used originally because the 3XFLAG donor sequence (which 134 starts with a "G") would not disrupt the PAM. Moreover, the alternative approach to 135 prevent re-cutting of the repaired locus, mutating the guide binding site, would require 136 introducing potentially undesirable mutations into the 5' UTR. To solve this problem, we 137 added an extra CCC, proline codon, to the 3XFLAG donor sequence, immediately 138 downstream of ATG ( Figure S2). Using this donor and guide we recovered flag::wago-6 139 alleles in 52% of the F1 Roller animals analyzed. In all of the edited strains the Roller 140 phenotype was expressed only transiently during the F1. These findings demonstrate 141 the general utility of the Roller marker for identifying edited animals without introducing 142 additional edits or undesired phenotypes into the resulting strains. In addition, these 143 findings indicate that as long as the desired insertion site resides within 20 bp of the cut 144 site, ssODN donors provide for highly efficient editing. 145

Hybrid dsDNA donors promote the integration of large constructs 146
High rates of HDR have been reported using PCR-generated double stranded DNA 147 (dsDNA) ~1kb-sized donors with ~35bp homology arms (PAIX et al. 2015). However, we have struggled to reproduce these successes using the original or optimized protocols 149 (data not shown). Extending the homology arms from 35bp to 120bp resulted in low, 150 ~1%, but reproducible levels of GFP integration at 3 different loci (Table 1). We 151 speculated that perhaps the additional donor dsDNA (which was 5X greater in 152 concentration than the Roller plasmid) was interfering with Cas9 RNP assembly. We 153 therefore decided to pre-assemble the Cas9 RNPs for 10 minutes at 37°C prior to  Our findings suggest that concentrations of Cas9 RNPs recommended in some 211 protocols are quite toxic and eliminate F1 progeny that receive the largest amounts of 212 co-injected long dsDNA. Since RNP activity and toxicity will vary depending on the 213 specific target or guide sequence, or due to variations in protein preparation across 214 vendors or home-made sources, we recommend that Cas9 RNPs be tested routinely for 215 optimal concentration using the simple and inexpensive rol-6/TIDE approach (Figure  216 1A).
We do not yet know how or why hybrid dsDNA PCR donors stimulate HDR. It 218 seems likely that other modifications, such as chemical modifications to the ends of the 219 donor molecule may drive even greater efficiencies. However, we decided to report 220 these findings now prior to fully exploring these issues, since the procedure for 221 generating hybrid donors is extremely easy to implement and has worked efficiently on 222 every one of over a dozen loci attempted in our group thus far. We anticipate that hybrid 223 donors will also stimulate precise editing in other systems. In summary, it is now as 224 easy to precisely edit the worm genome as it is to generate the iconic Roller transgenics

For generating point mutations:
Pick 35bp homology upstream and 35bp homology downstream of your guide cut site, which should ideally be within 20bp of the desired mutation site. Introduce the desired mutations in the donor and the PAM/ guide binding sequence.

For large deletions with 2 guides:
Pick 35bp homology upstream of the left-guide cut site and 35bp homology downstream of the right-guide cut site and put them together. Everything in between will be removed. Deletions up to 1kb can be easily achieved through this strategy. In principle, this should work for larger deletions as well.
dsDNA asymmetric-hybrid donors: 1. Order 140bp oligos from IDT as Ultramers; 120bp as homology arms and 20bp complementary to GFP (or any other desired insert). Also, order standard oligos just complementary to your insert (no homology arms). 2. Generate two PCR products as shown below in figure; one with 120bp homology arms and the other without any homology arms (only insert sequence) using an insert containing plasmid as the template for PCR; perform 4 to 8 50 µl reactions for each product.
(mount several F2 animals onto 2% agarose pads) or by using a fluorescence dissecting scope.