Introduction

Recent genome-wide screening by next-generation sequencing (NGS) techniques, such as exome sequencing, facilitates analysis of human disease-causing mutations at single-nucleotide resolution. Because more than one candidate disease-causing mutation is often indicated by in silico analysis of genome-wide screening, it is necessary to validate functions of the candidate mutations in vitro and/or in vivo by reproducing such mutations in cultured cells or introducing corresponding candidate mutations in animals. Of mammalian animal models, mice have been the first choice for reproducing human disease-causing mutations since the murine genome can be artificially manipulated by conventional gene targeting technology using embryonic stem (ES) cells [1]. However, there are some issues with the technology used for screening candidates to identify causative mutations. It is costly and time-consuming to produce mice carrying a mutation, inserts unnecessary exogenous sequences such as selection marker cassettes or loxP signals at a manipulated locus, and cannot be applied to repeat-rich regions. These issues have been resolved with the development of genome editing technology. In addition, this enables precise genome modification at the single-nucleotide level.

So far, several genome-editing tools have been developed such as zinc-finger nucleases (ZFNs), transcription activator-like effector nucleases (TALENs), and RNA-guided clustered regularly interspaced short palindrome repeat-associated Cas9 nuclease (CRISPR/Cas9) systems [2,3,4,5]. ZFNs and TALENs consist of a pair of DNA-binding domains fused with a FokI DNA nuclease domain. In contrast, the CRISPR/Cas9 system consists of a single guide RNA (sgRNA) for recognition of a target genomic locus and Cas9 for recognition of the sgRNA and digestion of DNA. sgRNA is a fusion of CRISPR RNA (crRNA) and trans-activating crRNA (tracrRNA) derived from Streptococcus pyogenes [6]. DNA nucleases of genome-editing tools introduce double-strand breaks (DSBs) into specific target genomic sequences that induce non-homologous end-joining (NHEJ)-mediated insertions and deletions (indels) or homology-directed repair (HDR)-mediated knock-in with an oligonucleotide or targeting vector containing exogenous DNA sequence(s) and homologous arms. Of the three genome-editing tools, the CRISPR/Cas9 system is the easiest to establish and has the highest efficiency of genome editing, and its use is now more widespread than that of the others.

In 2013, the first examples of CRISPR/Cas9-mediated genome editing of mammalian genomes were reported by Mali et al. [5] and Cong et al. [4]. Using mammalian cell lines, they demonstrated that indels and knock-ins can be introduced by co-transfection of sgRNA and Cas9 expression vectors with higher efficiency than that of TALENs. The first paper describing CRISPR/Cas9-mediated in vivo genome editing in mouse zygotes was also published in the same year [7, 8]. Shen et al. [7] showed that the EGFP gene in Pou5f1-IRES-EGFP knock-in mice can be mutated by microinjection of in vitro transcribed sgRNA and Cas9 mRNA into fertilized eggs. Using this strategy, Wang et al. [8] demonstrated that NHEJ-mediated indel mutations can be introduced at multiple endogenous genomic loci in vivo. Moreover, they have shown that mice with artificially defined nucleotide substitutions can be efficiently produced by microinjection of a single-stranded oligodeoxynucleotide (ssODN) with sgRNA/Cas9 via the HDR-mediated pathway. These studies clearly show that genetically modified mice, not only carrying simple indels but also nucleotide substitutions, can be obtained within 1 month.

At present, more than 3000 papers on the CRISPR/Cas9 system have been published, including not only methodologies but also functional studies using genome-edited cells or animals. Recently, there has been a gradual increase in the number of functional studies that analyzed phenotypes of mice carrying mutations corresponding to candidate human disease-causing mutations identified by exome sequencing. In this review, we discuss the recent progress of the CRISPR/Cas9 system, focusing on the production of mouse models of diseases.

HDR-mediated production of nucleotide substitution in mice

The ssODN-mediated nucleotide substitution in genomes is one of the essential methods for understanding the function of specific amino acids in proteins or for reproducing human mutations in mice for disease modeling. Wang et al. [8] reported the efficient production of mice with nucleotide substitutions, in which two nucleotides were converted at Tet1 (GC>AT substitution) and Tet2 (TA>AT substitution) loci by microinjection of ssODN, sgRNAs, and Cas9 mRNA in fertilized eggs. Inui et al. [9] also showed induction of point mutations in the Sf1 (C>T substitution) gene by this strategy. Reports showing the production of animals with nucleotide substitutions are summarized in Table 1. These reports showed that the efficiencies of creation of substituted alleles using the above strategies are not always high, ranging from 3 to 67% (calculated from the ratio of the number of pups harboring mutations to that of total pups). These low efficiencies could be due to the low frequency of HDR-mediated processes.

Table1 Production of mice carrying nucleotide substitutions using the CRIPSR/Cas9 system

Improvements in HDR-mediated substitution were reported in a knock-in experiment involving EGFP in a genomic locus of interest. Aida et al. [10] have demonstrated that the efficiency was improved by microinjection of recombinant Cas9 protein and crRNA/ tracrRNA, instead of Cas9 mRNA and sgRNA targeting Actb, with a targeting vector that contained an EGFP cassette with homologous arms at the Actb locus, as a donor for HDR-mediated editing, in fertilized eggs. They reported that the microinjection of RNA–protein complexes into pronuclei enables delivery of the complex at earlier stages than that of Cas9 mRNA/sgRNA in the cytoplasm of zygotes. This would be one of the reasons for improvement in the knock-in efficiency.

Other approaches for enhancement of knock-in efficiencies have been reported based on the inhibition of NHEJ using cell lines: Chu et al. [11] demonstrated that inhibition of the NHEJ pathway improved the efficiencies of HDR-mediated knock-in by knock-down of KU70/80 or DNA ligase IV, which are key factors for NHEJ, or by overexpression of E1B55K/E4orf6, which is derived from Adenovirus and known as a suppressor of the Mre11/Rad50/NBS1 complex involved in the NHEJ pathway [12, 13]. In mice, Maruyama et al. [14] have shown that the frequency of HDR is improved by microinjection of a DNA ligase IV inhibitor, SCR7, into mouse zygotes with sgRNA/Cas9/ssODN [15]. Another way to improve the efficiency is enhancement of the HDR pathway. Song et al. [16] showed that overexpression of RAD51 in zygotes, an essential player in the HDR pathway, or treatment with a RAD51 enhancer, RS-1, also improved HDR-mediated in vivo knock-in efficiencies in rabbits [17].

The progress achieved with these methodologies for HDR-mediated nucleotide substitution enables us to generate mice harboring mutations at the single-nucleotide level without inserting any unnecessary sequence. In addition, the required time and cost to obtain model mice reproducing candidates of human disease-causing mutations are less, compared to that for conventional gene targeting technology. These facts enable us to plan functional screening for identification of the responsible gene from a list of candidates by producing mouse models.

Reproduction of diseases in mice carrying nucleotide substitutions

In this section, we introduce several examples of functional studies of murine models carrying mutations recapitulating human disease mutations generated by HDR-mediated genome editing using the above described methods.

The first example shows that generation of mice harboring mutations using genome editing is useful as a screening tool for functional analysis of single nucleotide changes. Singh et al. [18] tried to identify infertility-causing single-nucleotide polymorphisms (SNPs) involving meiosis-associated genes in humans through generation of mice with single-nucleotide substitutions corresponding to the human SNPs by microinjection of sgRNA, Cas9 mRNA, and ssODN into fertilized eggs. The candidate human SNPs, rs3087335 (CDK2), rs63750447 (MLH1), rs61735519 (SMC1B), and rs147871035 (TEX15), were selected based on the criteria of minor alleles in humans, altering evolutionarily conserved amino acids and changes being predicted as functionally deleterious by multiple algorithms. Phenotype analysis of the resulting mutation-bearing mice (Cdk2 p.Y15S, MLH1 p.V384D, Smc1b p.F1055L, and Tex15 p.T218I) showed that Cdk2 p.Y15S/p.Y15S mice exhibited severe spermatogenetic defects due to meiotic arrest at the pachytene stage, while spermatogenesis in homozygous mutants for the other SNPs was indistinguishable from wild type. This paper shows that generation of mice harboring mutations by the CRISPR/Cas9 system can be a screening tool.

The second example involves mice carrying CRISPR/Cas9-mediated single-nucleotide substitution used for validation of a disease-causing mutation for a candidate identified from human exome sequencing. Lee et al. [19] carried out exome-sequencing analysis of patients with familial thoracic aortic aneurysms and dissections (FTAAD) to identify a causal mutation of the disease. They finally identified a missense mutation at the LOX gene (p.M298R) as a strong candidate for FTAAD. To reveal the functional relevance of the mutation to the disease, they generated mice harboring a mutation (c.857 T > G encoding p.M292R) at the homologous position to the human mutation by microinjection of sgRNA, Cas9 mRNA and ssODN into fertilized eggs. Lox p.M292R/+ animals had disorganized ultrastructural properties of the aortic wall and Lox p.M292R/p.M292R died from ascending aortic aneurysm and spontaneous hemorrhage shortly after parturition. They concluded that a missense mutation in LOX is associated with aortic disease in humans. It seems quite likely that the amount of literature dealing with the combination of NGS analysis, including exome sequencing and screening or validation of results by generation of CRISPR/Cas9-mediated single nucleotide substituted mice, is increasing enormously.

It should be noted that modeling of human mutations in mice has limitations in that phenotypes of a mutation, which corresponds to the same amino-acid position in orthologous human and mouse genes, are not always similar. Such a case was reported by Miyado et al. [20]. We found two unrelated patients with 46,XX testicular/ovotesticular disorders of sex development (DSD) and a p.R92W variant involving NR5A1/SF1 [21]. We next generated mice carrying a homologous mutation of human p.R92W (Nr5a1 p.R92W) by microinjection of sgRNA, Cas9 mRNA, and ssODN into fertilized murine eggs to reproduce the patients’ phenotype in the mouse; however Nr5a1 p.R92W/p.R92W showed no signs of DSD.

Remedy of diseases in mice harboring mutations by nucleotide substitutions

As mentioned above, the CRISPR/Cas9 system can be used to substitute a wild-type nucleotide with a mutant one. This also suggests that it is applicable to replace a mutation with a wild-type sequence.

Such a case was reported for remedy of a mdx (X chromosome-linked muscular dystrophy) mutant-carrying mouse strain, which is a model of human Duchene muscle dystrophy (DMD) [22]. The responsible mdx mutation is a C>T missense substitution involving exon 23 of the dystrophin (Dmd) gene [23]. Long et al. [24] carried out microinjection of sgRNA, Cas9 mRNA, and ssODN, which was designed to correct the mutated nucleotide of the mdx allele (T) to the wild-type (C), in zygotes collected from mdx mice. The resulting “mutation-corrected” mice showed no signs of muscular dystrophy. Interestingly, a mosaic individual, in which the contribution of the “corrected” allele was only 17%, was also obtained and muscular dystrophy manifestations were only partial [24]. This report implies that the CRISPR/Cas9 system could have potential for radical treatment of genetic diseases.

It could be possible to remedy a genetic disease that develops after birth by in vivo genome editing of somatic cells if a mutation in a causative gene is substituted with wild-type sequence in cells that express the gene. Such a case was reported in a mouse model for the remedy for human hereditary tyrosinemia type I (HTI), which is a fatal recessive genetic disease caused by mutation of an essential gene in the tyrosine catabolic pathway, fumarylacetoacetate hydrolase (FAH) [25]. The mouse model for human HTI, called Fah5981SB mouse strain, carries a G>A point mutation at the boundary of exon 8 and intron 8 of Fah locus [26], which leads to aberrant splicing of mRNA. Fah deficiency causes accumulation of toxic metabolites in hepatocytes and leads to liver damage and body weight loss. It is reasonable that this phenotype might be rescued by correction of the mutation in the liver. Yin et al. [27] introduced a plasmid coding sgRNA and Cas9 together with ssODN carrying wild-type sequence around the mutated “A” as a donor via hydrodynamic tail vein injection, which is a method of in vivo transfection of naked nucleic acids, primarily in hepatocytes of living adult mice through rapid injection of nucleic acids into the tail vein. The results showed that a small population of hepatocytes, but not all hepatic cells, in the mice thus treated carried the corrected allele, and liver damage and body weight loss were substantially rescued. This research demonstrates that the CRISPR/Cas9 system could have potential for in vivo treatment of recessive genetic diseases in individuals after birth.

Additionally, recently, CRISPR/Cas9-based in vivo treatment of dominant genetic diseases has been reported. Xie et al. [28] identified a c.1589 A>G missense mutation (H530R) of the PRKAG2 gene encoding the γ2 subunit of AMP-activated protein kinase in a familial case of Wolff–Parkinson–White syndrome, which is a dominant genetic disease and which usually exhibits cardiomyopathy characterized by fatal ventricular tachyarrhythmia and progressive heart failure. The association of the mutation and the phenotype was examined by generation of mice with mutations through microinjection of sgRNA, Cas9 mRNA, and ssODN carrying mouse sequence around the corresponding human mutation with A>G missense mutation (H527R) into fertilized eggs. The resulting mutant (Prkag2 +/p.H527R), which recapitulated human symptoms, indicated that the mutation is responsible, and is useful as a model for the disease. To treat the mutants, they intraventricularly injected Prkag2 +/p.H527R mice with adeno-associated virus-9 (AAV9) carrying expression cassettes for Cas9 and sgRNA, which recognized only the mutant allele. Expression levels of Prkag2 were reduced in the injected mutant hearts, suggesting that the mutant allele was cleaved by Cas9/sgRNA and transcribed mutant Prkag2 mRNA was degraded by nonsense-mediated mRNA decay. The morphology and function of the heart were substantially restored after the effects of the gene product derived from mutant alleles were reduced.

The great potential of the CRISPR/Cas9 system for the remedy of recessive and dominant genetic diseases is evident, but efficient delivery of Cas9/sgRNA to causative cells remains challenging for most tissue types. One such tissue is the skin. Wu et al. [29] constructed a method of delivery and treated model mice for postnatal recessive dystrophic epidermolysis bullosa (RDEB), which causes skin fragility and blistering. They generated a model by microinjection of Cas9/sgRNA and a 1.5 kb DNA fragment carrying a part of the Col7a1 gene containing c.6485 G>A involving exon 80, a homologous mutation of which is a causative mutation found in human RDEB patients, and homozygous mice with mutations exhibited typical pathologic features. As a therapy model, they intradermally injected Cas9/sgRNA ribonucleoproteins targeting exon 80 of the Col7a1 locus into postnatal homozygous mutants. Genome editing of the target locus caused skipping of exon 80, resulting in a slightly shortened form of collagen VII protein, which retained its normal function, and the skin blistering phenotype of treated mutants was significantly diminished. It may be that development of various delivery systems for Cas9/gRNA will expand the CRISPR/Cas9 system-based therapy for genetic diseases.

Reproduction of diseases in mice carrying chromosome rearrangements

Reproducing human chromosomal rearrangements such as deletions, inversions, and duplications in mice is also an important approach to examine relationships between mutations and disease symptoms, and to establish animal disease models. To date, CRISPR/Cas9 system-based methods for producing mice with chromosomal rearrangements have been developed. In principle, chromosomal rearrangements can be induced by introduction of Cas9 and a pair of sgRNAs targeting chromosomal breakpoints. Using this strategy, deletions can be generated using sgRNAs directed towards sequences flanking the deletion target fragment, with lesions flanking the deletion target being connected intra- or inter-chromosomally. Where the same reaction occurs inter-chromosomally, duplications can be generated. Additionally, inversions can be introduced when lesions generated with different sgRNAs are connected between the inside and outside of a target fragment. Translocations can be generated when lesions induced on different chromosomes are connected; however, the efficiency of this reaction is empirically low in mouse zygotes.

Successful generations of chromosomally rearranged mice have been reported using the above strategy. Wang et al. [30] reported a 95 kb deletion in mice. Recently, we demonstrated generation of mice carrying deletions of 0.5, 2, and 5 Mb and an inversion of 5 Mb [31, 32]. To induce the 2 and the 5 Mb deletions, we microinjected Cas9/two sgRNAs and a donor plasmid carrying the breakpoint sequences so that the breakpoint sequence could be artificially defined. Over the course of this experiment, we also obtained mice carrying a 5 Mb inversion but no duplications were obtained [32]. Boroviak et al. [33] also reported that they microinjected Cas9/two sgRNAs and ssODN as a donor to guide deletion in fertilized eggs and detected 1.1 Mb deletions and inversions in 30 and 21% of embryos genotyped, respectively, although no duplications were observed. When they tried to generate 155 and 545 kb rearrangements using the same strategy, only 1–2% of mice analyzed had the duplication, while 18–24 and 18–30% of mice contained deletions and inversions, respectively. It seems likely that the frequency of generation of duplications of the target locus using this strategy is much lower than that of deletions and inversions. However, recently, Birling et al. [34] reported efficient generation of large duplications in mice and rats by microinjection of Cas9 and two pairs of sgRNAs: one pair targeted upstream and the other pair downstream of the region of interest and where the distance between sgRNAs of a pair was less than 150 bp.

Generation of large duplications is useful for modeling human mutations and for understanding functions of largely “scattered” regulatory regions; however, while it is not impossible, it is not very efficient. Another method to generate large duplications may be introduction of Cas9/a pair of sgRNAs into ES cells instead of fertilized eggs, since the number of clones to be screened can be easily increased when ES cells are used [35]. A disadvantage of this approach is that it takes a longer time for the generation of mice with mutations from ES cells compared to that of zygote injection. Another method could be insertion of bacterial artificial chromosomes (BACs) containing a fragment of the duplication of interest at the target genomic locus by a method named the “two-hit two-oligo with plasmid” (2H2OP), in which Cas9, two sgRNAs, BAC DNA and two ssODNs are microinjected into fertilized eggs [36]. One of the pairs of sgRNAs recognizes a target in the genome and in the BAC DNA. One of two ssODNs contains terminal sequences generated by the Cas9/sgRNA of one end of the genomic target and one end of the BAC, and the other contains those of the other ends of the genome. The BAC is integrated in the genome at the target locus. With this method, the authors confirmed that 18% of offspring contained BAC DNA at the locus of interest. The method was originally developed to insert human BAC sequences into rat genomic loci of interest; however, it should also be applicable for the generation of duplications, at least in principle. So far, this is the most efficient means to insert large fragments into a target locus, but there are some limitations: the size to be inserted can be up to several hundred kb, available only when BAC clones containing entire target regions can be obtained and the backbone BAC vector can be introduced along with the insert. Further studies are needed to develop efficient methods for generation of disease model mice with large duplications.

Closing remarks

In this review, we discussed the current status of methods for creation of murine disease models and some examples of phenotypic analyses and remedies of genetic defects in such mice using the CRISPR/Cas9 system.

In the field of medical genomics, NGS is a powerful tool for identification of disease-causing mutations, especially small indels, and minor nucleotide substitutions and deletions. These mutations can be easily mimicked in living mouse genomes by the CRISPR/Cas9 system; therefore, the combination of NGS and the CRISPR/Cas9 system has been inextricably linked for recent medical genetic studies. Further advances in NGS and genome editing should strongly promote medical genomics.

As for other mutations, such as inversions and duplications, NGS is not very suitable, if not impossible. Instead, array comparative genomic hybridization (aCGH) is useful for detection of duplications, although resolution is not at the single nucleotide level. Identification of inversions and duplications at high resolution from patients’ genomic DNA is not facile. Currently, requirements for modeling inversions and duplications do not seem very high compared to that of indels and local nucleotide substitutions. However, this requirement will be increased when such mutations can be easily identified at high resolution in the near future. It will be necessary to construct efficient methods to reproduce inversions and duplications of various sizes in mice using genome editing technology as soon as a method for identification of inversions or duplications at high resolution is made available.

Before the emergence of genome editing technology, reproduction of human mutations in mice was carried out by ES cell-based technology, which required excessive cost, time and labor. This limited the number of mutant strains that could be generated. Genome editing requires much fewer of these, so it can be used even for screening purposes, whereas generation of mutant lines (up to dozens) is practical for screening. NGS analysis of patient sequences often resulted in listing of a large number of candidate mutations (typically more than one hundred). Further acceleration of the generation of mice harboring mutations by genome editing would be required.

Lastly, as we introduced in this review, it should be noted that reproduction of candidates of disease-causing mutations in mice is becoming more important to establish, since this enables validation of the functional importance of the candidates in vivo. However, it should be kept in mind that, when one analyzes model animals that carry the corresponding mutations as human disease-causing ones, the phenotype in animal models and human symptoms are sometimes different.