Quantitative evaluation of incomplete preweaning lethality in mice by using the CRISPR/Cas9 system

Various molecular biology techniques implementing genome editing have made it possible to generate mouse mutants for nearly all known genes; as a result, the International Mouse Phenotyping Consortium (IMPC) database listing the phenotypes of genetically modified mice has been established. Among mouse phenotypes, lethality is crucial to evaluate the importance of genes in mouse survival. Although many genes are reported to show “preweaning lethality, incomplete penetrance” in the IMPC database, the survival rates of homozygous knockout mice are highly variable. Here, we propose the lethal allele index (LAI), the ratio of the observed number of mice with homozygous knockout (KO) to the theoretically predicted number of homozygous KO mice, as a simple quantitative indicator of preweaning lethality. Among the mice mutants registered as incompletely lethal in IMPC, the LAI calculated from the genotypes of F1 mice tended to be lower in disease-related genes, and correlated with the frequency of loss-of-function (LOF) alleles in humans. In genome-edited mice using CRISPR/Cas9, the number of mice with homozygous frameshift alleles seemed to be associated with lethality. We edited the Ehd1 gene in cell lines as well as mice using CRISPR/Cas9, and found that the genotype distribution was significantly different. The LAI calculated from these data was similar to the value calculated from the IMPC data. These findings support the potential usefulness of the LAI as an index of preweaning lethality in genome-edited mice.


Results
Possible relationship between lethality and disease-related genes. We defined the Lethal Allele Index (LAI) in a broad sense as the ratio of the observed number of homozygous knockout (KO) mice to the theoretical number of homozygous KO mice predicted using data for other genotypes. When heterozygous KO mice are crossed with each other, the ratio of wild type, heterozygous knockout mice, and homozygous knockout mice of non-lethal genes must be 1:2:1. Thus, we calculated the LAI of genes registered as "preweaning lethality, incomplete penetrance" in IMPC as follows: Lethal Allele Index actual number of homozygous knockout mice theoretical number of homozygous knockout mice actual number of homozygous knockout mice 3 (actual number of wild type and heterozygous knockout mice) We compared the LAI of the genes registered as disease-related genes in the Online Mendelian Inheritance in Man (OMIM) database with the LAI of other genes. The LAI for disease-related genes tended to be lower than that for non-disease related genes ( Fig. 1A and Supplementary Table 1, Mann-Whitney U test, p = 0.082).
Next, we compared the LAI with pLi (probability of loss of function intolerance), which is calculated from the genotypes of humans registered in the ExAC (Exome Aggregation Consortium), and indicates the frequency of loss of function mutations in a human population. The LAI showed a tendency of negative correlation with pLi ( Fig. 1B and Supplementary Table 1, Pearson's correlation coefficient, R = −0.091, p = 0.062, N = 423). These findings show that there is a tendency of the LAI to be correlated with clinical significance and loss of function intolerance in humans.
Genotypes of genome-edited mice generated by CRISPR/Cas9. We generated F 0 mutant mice of Ehd1 (Fig. 2) and determined the genotypes of 55 weaned mice from among 79 pups born. The genotype distribution was compared with other lines of genome-edited mice generated by CRISPR/Cas9: Ntrk1 13 , Chdh 14 , and Nlgn1 15 . The rate of the mice homozygous for frameshift alleles to the sum of frameshift /in-frame and in-frame/ in-frame genotypes was highest (2.22) for Chdh, designated as "viable" in IMPC, lowest (0) for Ntrk1, designated as "lethal" in IMPC, and intermediate (0.26) for Ehd1, designated as "preweaning lethality, incomplete penetrance" in IMPC, suggesting that the rate of obtaining mice homozygous for loss of function (LOF) alleles reflects the degree of lethality ( Table 1).
Establishment of mutant PC12 cell lines by the CRISPR/Cas9 system. To calculate the LAI in the genome-edited mice of Ehd1 generated by CRISPR/Cas9, we additionally established Ehd1 mutant cell lines using the CRISPR/Cas9 system. Using these, we determined the ratio of frameshift alleles that can be regarded as LOF alleles to in-frame alleles, regarded as non-LOF alleles. Following transfection of an sgRNA-SpCas9-GFP all-in-one vector, we cloned single green fluorescent protein (GFP)-positive PC12 cells by fluorescence activated cell sorting (FACS). After colony formation, we established and genotyped Ehd1 mutant PC12 cells (Fig. 3). We performed this cloning step twice, changing the concentration of transfection mixtures in each step. We found that the genotype distribution of frameshift or in-frame indels was significantly different from that in Ehd1 mutant mice (2 × 6 table, p < 5 × 10 −8 by Fisher's exact probability test) ( Table 1). The rate of homozygotes of frameshift alleles was significantly smaller in mice (6/55, 10.9%) than cells (26/55, 47.3%) (2 × 2 table, p < 5 × 10 −5 by Fisher's exact probability test). The distribution of the length of insertion/deletions (indels) is shown in Fig. 4. The rate of frameshift alleles/total cleaved alleles was calculated to be 0.80 (70/87) based on the genotype data of PC12 cells. The expected frequencies of each genotype are shown in Table 2. According to Table 2, the number of homozygous frameshift genotypes (N f/f = 26) and that of homozygous in-frame or heterozygous in-frame/frameshift genotypes (N i/i + N i/f = 2 + 12 = 14) did not significantly deviate from the theoretical ratio of 0.64:0.36 (χ 2 test for goodness-of-fit, p = 0.89). This suggests that the disruption of Ehd1 does not confer a disadvantage in the survival of PC12 cells.  Calculation of LAI of Ehd1 mutant mice. When we define the rate of the frameshift alleles generated by CRISPR/Cas9 as x and denote the frameshift alleles as LOF and in-frame alleles as non-LOF alleles, we can calculate the LAI as follows: The LAI calculated from the Ehd1 genome-edited mice using CRISPR/Cas9 (0.146) was close to the LAI of Ehd1 (0.111) calculated by the data of IMPC.

Discussion
In this study, we proposed the LAI as a quantitative index for "preweaning lethality, incomplete penetrance" as the rate of the number of mice with homozygous LOF alleles to the number predicted from the genotypes of other mice. The significant difference in genotypes between Ehd1 mutant mice and Ehd1 mutant PC12 cell lines generated by the CRISPR/Cas9 system suggests that the lower rate of mice homozygous for frameshift alleles is due to the lethality of homozygous LOF alleles.
The present result that the homozygous frameshift genotypes of PC12 cells did not show apparent lethality is consistent with a previous report showing that Ehd1 homozygous knockout murine embryonic fibroblasts (MEFs) can survive 16 . A recent study reported essential genes in humans by identifying genes whose homozygous knockouts were lethal in cell lines 17 . EHD1 was not included in the list of essential genes, which is also consistent with the present results.
Although we attempted to analyze the genotype distribution of the genome-edited mice by CRISPR/Cas9 for genes designated as "preweaning lethality, incomplete penetrance", based on previous studies, none of these

In-frame Frameshift
In-frame  papers reported on genotype distribution. We believe it would be useful to report on the genotype distribution of F 0 mice generated by CRISPR/Cas9 as a phenotype of the mutant mice. In order to calculate the LAI in genome-edited mice by CRISPR/Cas9, we quantified the cleavage pattern, the rate of frameshift alleles to total cleaved alleles, using genome-edited cell lines. Although we performed single cell cloning to establish these cell lines, which took time out of our study and was very costly, it would be enough to perform a transfection of the TA-cloned PCR product from genome-edited cells into competent cells followed by colony PCR and direct sequencing to calculate the rate of frameshift alleles generated by CRISPR/Cas9.
Recently, however, it was reported that CRISPR/Cas9 often causes large deletions near the cut sites 18 , and we thus may have missed large deletions in otherwise apparently "homozygous" genotypes such as WT/WT, in-frame/in-frame, and frameshift/frameshift. Actually, 7 out of the 14 in-frame/in-frame genotypes, and all frameshift/frameshift genotypes, showed single electropherograms, and a possibility that the other allele has a large deletion cannot be ruled out. This fact would not influence the significance of the rate of homozygous frameshift alleles to the sum of frameshift /in-frame and in-frame/in-frame genotypes. In the calculation of LAI, frameshift (LOF)/frameshift (LOF) might be contaminated by frameshift (LOF)/large deletion. If the large deletion is predicted to disrupt genes, that allele should also be considered as an LOF allele. The other value used is "frameshift (LOF)/in-frame + in-frame/in-frame". In-frame/in-frame may be contaminated by in-frame/large deletion. If large deletion is a LOF mutation, the sum of LOF/in-frame and in-frame/in-frame is not affected by the presence of the large deletion. However, the presence of large deletions affects the ratio of the LOF alleles generated by CRISPR/Cas9 and hampers accurate calculation of LAI. Thus, for the accurate calculation of LAI in genome edited mice using CRIPSR/Cas9, accurate genotyping by long-read sequencing using PacBio platform among others 18 .
The major limitation of this study was the inaccuracy of the LAI of Ehd1 mutant mice because we did not genotype large deletions using long read sequencing. Nevertheless, the concept of LAI would help to discuss the quantitative evaluation of lethality of genome-edited mice.

Investigation in OMIM database and acquisition of pLi.
To investigate whether the "preweaning lethality, incomplete penetrance" genes were registered in the OMIM database, we simply searched the gene name registered in IMPC. The list of pLi was obtained from a previous report 19 . Whenever the gene name registered in IMPC was not found in the list of pLi, alias names registered in the National Center for Biotechnology Information (NCBI) database were verified.

Establishment of Ehd1 mutant mice. All animal experiment protocols were approved by the Wako
Animal Experiment Committee, RIKEN, and all experiments were performed in accordance with the approved guidelines and regulations. All other experimental procedures were approved by the RIKEN Wako Safety Center and were performed in accordance with the approved guidelines.
We designed a single guide RNA (sgRNA) and a single-strand oligo deoxynucleotide (ssODN) to simulate a single-nucleotide deletion in exon 5 of EHD1, identified as a de novo mutation in a patient 20 , in mice ( Fig. 2A). The mutation introduces a stop codon in the last exon, and is predicted to produce a truncated protein, escaping nonsense-mediated mRNA decay. We generated an sgRNA-Cas9-GFP all-in-one vector and detected its cleavage activity of Ehd1 in Neuro2A cells in a Surveyor assay (Fig. 2B). We injected a mixture of sgRNA, Cas9 mRNA, and ssODN into fertilized eggs and obtained Ehd1 mutant mice (F 0 ). We performed the injection twice at different concentrations. No knock-in alleles were detected, likely because of the low concentration of ssODN used. Thus, knock-in alleles were not considered in genotype analysis. We excluded undetermined genotypes showing unrecognizable electropherograms.
Plasmid construction. We generated the sgRNA-SpCas9-GFP all-in-one vector as previously described by Ran et al. 2 . The candidate sequences of an sgRNA for Ehd1 were predicted using CRISPR direct (https://crispr. dbcls.jp/). For the sgRNA oligo insert, 100 pmol of primer set A (Supplemental Table 2A Proteinase K (Roche, Basel, Switzerland), and cell lysates were incubated overnight. The incubated cell lysates were treated with phenol-chloroform extraction followed by ethanol precipitation. Genomic DNA fragments for the surveyor assay were amplified with primer set B (Supplemental Table 2A) and Ex Taq polymerase (Takara, Shiga, Japan) under the following reaction conditions: 96 °C for 2 min, 30 cycles at 94 °C for 30 s and 68 °C for 1 min, and 72 °C for 5 min. The surveyor assay was performed using the Surveyor Mutation Detection Kit (Integrated DNA Technologies).
Generation of Ehd1 mutant mice. sgRNA for fertilized eggs injection was transcribed using the MEGAshortscript Kit (Life Technologies, Carlsbad, CA, USA) and the sgRNA-SpCas9-GFP all-in-one vector as a template. A mixture of sgRNA (50 ng/μL), Cas9 mRNA (100 ng/μL, Sigma), and ssODN (first set; 7.9 ng/μL of ssODN_A and second set 50 ng/μL of ssODN_B. Supplemental Table 2B) was microinjected into the cytoplasm of fertilized eggs obtained from C57BL6/N mice. The injected eggs were transferred into the uterus of pseudopregnant ICR females.
Genotyping of mutant mice. All mutant mice were genotyped by using genomic DNA extracted from tail clips. Tail biopsies were performed on postnatal day 14. Tail clips were incubated in 100 μL of lysis buffer (25 mM NaOH and 0.2 mM EDTA) at 95 °C for 30 min. An equal volume of 40 mM TRIZMA hydrochloride was added to the lysate after incubation, and the mixture was vortexed briefly. Next, 1 μL of tail samples diluted by 2.5-fold using water was amplified by primer set B (Supplemental Table 2A) with Tks Gflex polymerase (Takara) in a 10-μL reaction mix. PCR conditions were as follows: 94 °C for 1 min, 30 cycles at 98 °C for 10 s and 68 °C for 1 min. The PCR products were sequenced with BigDye Terminator V3.1 and an ABI 3730xl sequencer (Life Technologies) using sequencing primers (Supplemental Table 2A).

Generation of Chdh mutant mice.
The method for the generation of Chdh mutant mice is described elsewhere (Ohnishi et al., in submission). In brief, we utilized Cas9 nickase mRNA (CAS504A-1). We injected the Cas9n mRNA and two sgRNAs into fertilized eggs. The sequences of the targets of sgRNAs were as follows: 5′-AGCATGTGGCAGGTCCTTAG-3′ and 5′-CAGTTCCTCTATCCACCGAC-3′.

Establishement of Ehd1 mutant PC12 cells by CRISPR/Cas9 System
Transfection of sgRNA-Cas9-GFP all-in-one vector. sgRNA and ssODN for PC12 cells were also designed to simulate the mutation found in a patient. Because no knock-in alleles were detected, knock-in alleles were not considered in the analysis. Forty-eight hours before transfection, PC12 cells were plated at 12 × 10 5 cells per well in 6-well plates. Next, 4.0 μg of sgRNA-Cas9-GFP all-in-one vector and ssODNs (1st set: 10 ng of ssODN_C, D and 2nd set: 100 ng of ssODN_D; Supplemental Table 2B) for mutation knock-in were transfected with 6.0 μL of Lipofectamine 2000 (Invitrogen). ssODN_C was used for mutation knock-in and ssODN_D was used for knock-in of a silent mutation to inhibit additional cutting by Cas9 21 . Preparation of conditioned culture medium. The supernatant of the culture medium used to grow confluent PC12 cells was collected. The supernatant was centrifuged for 5 min at 500 × g to exclude contaminants. The supernatant was sterilized through a 0.22-μm filter (Merck Millipore, Billerica, MA, USA). The sterilized supernatant was diluted by 10-fold using the normal culture medium (conditioned culture medium).
Scale-up from single PC12 cell. After single-cell sorting (day 0), the plates were incubated (37 °C and 5% CO 2 ) under stationary conditions until day 7. On day 7, a few cells were detected, and 100 μL of conditioned medium were added to each well which had contained 200 μL of conditioned medium. After that, a part of the medium (150 μL) was replaced with new conditioned medium every 4-5 days until the colonies grew to a sufficient size. On day 22, the colonies were split by pipetting and transferred to new 96-well plates. For the passage from the 96-well plate, 60 μL of 0.25% Trypsin-EDTA (Gibco) was added to PC12 cells for 3 min at 37 °C. After incubation, 120 μL of conditioned culture medium was added. Next, 120 and 60 μL of the cell suspension were transferred to a 24-well plate for scale-up and a 96-well plate for genotyping, respectively. In the second independent trial, all the cell suspension was transferred to a 96-well plate for genotyping. Confluent cells cultured on the 24-well plate were trypsinized with 300 μL of 0.25% Trypsin-EDTA (Gibco) for 3 min at 37 °C. After this treatment, trypsin was inactivated by adding 700 μL of conditioned culture medium, and the cells were collected by centrifugation at 200 × g for 3 min and transferred into a 6-well plate. Confluent cells in this plate were frozen with Cell Banker 1 plus (Nippon Zenyaku Kogyo, Koriyama, Japan).

Genotyping of mutant PC12 cells by crude PCR.
Genotyping of PC12 cells was performed using crude cell lysates. PC12 cells cultured on 96-well plates were treated with 50 μL of lysis buffer for crude PCR (10 mM Tris-HCl (pH 8.0), 0.1% Triton-X, 10 mM EDTA, and 1/100 diluted proteinase K) at 37 °C overnight. The crude cell lysates were diluted by 10-fold and 1-μL aliquots of the diluted lysates were added to the PCR mix with Ex Taq polymerase (Takara) using primer set C in a total volume of 10 μL (Supplemental Table 2A). Reaction conditions were as follows: 96 °C for 2 min, 30 cycles at 94 °C for 30 s, 63 °C for 30 s, and 72 °C for 30 s, followed