Super-Mendelian inheritance mediated by CRISPR-Cas9 in the female mouse germline.

A gene drive biases the transmission of one of the two copies of a gene such that it is inherited more frequently than by random segregation. Highly efficient gene drive systems have recently been developed in insects, which leverage the sequence-targeted DNA cleavage activity of CRISPR-Cas9 and endogenous homology-directed repair mechanisms to convert heterozygous genotypes to homozygosity1-4. If implemented in laboratory rodents, similar systems would enable the rapid assembly of currently impractical genotypes that involve multiple homozygous genes (for example, to model multigenic human diseases). To our knowledge, however, such a system has not yet been demonstrated in mammals. Here we use an active genetic element that encodes a guide RNA, which is embedded in the mouse tyrosinase (Tyr) gene, to evaluate whether targeted gene conversion can occur when CRISPR-Cas9 is active in the early embryo or in the developing germline. Although Cas9 efficiently induces double-stranded DNA breaks in the early embryo and male germline, these breaks are not corrected by homology-directed repair. By contrast, Cas9 expression limited to the female germline induces double-stranded breaks that are corrected by homology-directed repair, which copies the active genetic element from the donor to the receiver chromosome and increases its rate of inheritance in the next generation. These results demonstrate the feasibility of CRISPR-Cas9-mediated systems that bias inheritance of desired alleles in mice and that have the potential to transform the use of rodent models in basic and biomedical research.

A cross of mice that are heterozygous for each of three unlinked genes must produce 146 offspring for a 90% probability that one will be triple homozygous mutant. The likelihood decreases further if any of the three mutations are genetically linked but on opposite homologous chromosomes, since recombination events that combine alleles onto the same chromosome would be very infrequent. The cost, time, and requirement for a large number of mice to obtain a few individuals of the desired genotype are therefore prohibitive for certain complex models of multigenic evolutionary traits or human diseases, like arthritis and cancer.

Recently, CRISPR/Cas9-mediated gene drive systems were developed in Drosophila and Anopheline mosquitoes that increase the frequency of inheritance of desired alleles[1-4]. These utilized genetic elements, which we refer to broadly as “active genetic elements”, that can carry transgenes or orthologous sequences from other species[5]. Critically, an active genetic element includes a guide RNA (gRNA) and is inserted into the genome at the precise location that is targeted for cleavage by the encoded gRNA. In a heterozygous animal that also expresses the Cas9 nuclease, the gRNA targets cleavage of the wild type homologous chromosome. Genomic sequences that flank the active genetic element then resolve the double strand break (DSB) by homology directed repair (HDR), which copies the active genetic element from donor to receiver chromosome and converts the heterozygous genotype to homozygosity. The frequency of transmitting the active genetic element to the next generation is therefore greater than expected by random segregation of heterozygous alleles and is referred to as “super-Mendelian”. In addition to the potential to surmount the obstacles of assembling complex genotypes in laboratory rodents, variations of a CRISPR/Cas9-mediated system have been proposed that might suppress invasive rodent populations and/or reduce the impacts of rodent-borne disease[6,7].

Despite the high efficiency observed in insects, divergence over 790 million years since their last common ancestor with mammals presents two potential obstacles to implementation of active genetics in mice; the frequency of DSB formation using genetically encoded Cas9 and gRNA and/or of HDR may preclude efficient gene conversion. The alternative DSB repair pathway, non-homologous end joining (NHEJ), frequently generates small insertions or deletions, known as “indels”, that make CRISPR/Cas9 an effective means of mutating specific sites in the genome. Although HDR of CRISPR/Cas9 induced DSBs does occur in vitro and in vivo in mammalian cells and embryos, usually from a plasmid or single stranded DNA template, NHEJ is the predominant mechanism of DSB repair in somatic cells[8,9].

To assess the feasibility of active genetic systems in mice, we designed a “CopyCat” element[10] that differs from the genetic element used initially in insects in that it cannot self-propagate because it encodes a gRNA but not the Cas9 protein (Fig. 1a). We designed our strategy to disrupt the Tyrosinase gene (Tyr), because of the obvious albino phenotype of homozygous loss-of-function animals[11] and to make use of a previously characterized Tyr gRNA with high activity[12]. Extended Data Fig. 1 illustrates the insertion of this element precisely into the gRNA cut site in exon 4 of Tyr to obtain the Tyr knock-in allele (Supplementary Methods and Supplementary Fig. 1, 2). Briefly, the Tyr gRNA is transcribed from a constitutive human RNA polymerase III U6 promoter[13]. On the reverse strand of DNA, to minimize possible transcriptional conflict, mCherry is ubiquitously expressed from the human cytomegalovirus (CMV) immediate-early promoter and enhancer[14]. Since the 2.8 kb insert disrupts the Tyr open reading frame, Tyr is a functionally null (albino) allele that is propagated by Mendelian inheritance in absence of Cas9. Crossing mice that carry the Tyr element to transgenic mice that express Cas9 allows us to test whether it is possible to observe super-Mendelian inheritance of the Tyr allele. For this analysis, we assessed eight different genetic strategies using existing tools that provide spatial and temporal control of Cas9 expression in the early embryo and in the male and female germlines.

Figure 1 |

Embryonic Cas9 activity does not copy the Tyrosinase allele from the donor to the receiver chromosome.

(a) The genetically encoded Tyr element, when combined with a transgenic source of Cas9, is expected to induce a DSB in the Tyr-marked receiver chromosome, which could be repaired by inter-homologue HDR. (b) Breeding strategy to unite Tyr with a constitutive Cas9 transgene followed by test cross to Tyr. (c) Sum of F3 test cross offspring from five independent families for each Rosa26-Cas9 and H11-Cas9 genotype. (d) A representative of six Rosa26-Cas9 F2 litters. Black mice did not inherit Tyr. Grey mice inherited Tyr but not Cas9. White mice inherited both transgenes. (e) A representative of five F2 litters in which all offspring inherited H11-Cas9. The mosaic mice also inherited Tyr.

Extended data figure 1.

Knock-in strategy using the Tyr targeting vector.

The U6-Tyr4a-gRNA and CMV-mCherry were inserted into the cut site of the Tyr4a-gRNA by homology directed repair after CRISPR/Cas9 DSB formation targeted by the Tyr4a-gRNA. See Supplemental Methods and Supplementary Fig. 1 and 2 for additional details.

We employed two “constitutive” Cas9 transgenic lines, Rosa26-Cas9[15] and H11-Cas9[16], that reportedly express Cas9 in all tissues that have been assessed. Each is driven by a ubiquitous promoter and is placed in the respective Rosa26 or H11 “safe harbor” locus. In order to track inheritance of the chromosome that is targeted for gene conversion, we bred the chinchilla allele of Tyrosinase (Tyr, here simplified to Tyr) into each Cas9 transgenic line. Tyr encodes a hypomorphic point mutation in exon 5 that is tightly linked to the gRNA target site in exon 4. Tyr homozygotes or heterozygotes complementing a null allele have a grey coat color, and the G to A single nucleotide polymorphism can be scored with certainty by PCR followed by DNA sequencing[17] (Supplementary Fig. 4).

Female Rosa26-Cas9; Tyr and H11-Cas9; Tyr mice were each crossed to Tyr males to unite the gRNA and Cas9 protein in the early embryo (Fig. 1b). In absence of a loss-of-function mutation in exon 4 of the receiver chromosome, Tyr mice should appear grey (see Cas9-; Tyr mice in Fig. 1d). However, we did not observe any grey Cas9+; Tyr mice in the F2 offspring of either cross. Instead, all 17 of the Rosa26-Cas9; Tyr mice were entirely white. Among H11-Cas9; Tyr mice, 21 of the F2 progeny (87.5%) were a mosaic mixture of grey and white fur, and three mice (12.5%) were entirely white (Fig. 1d, e and Extended Data Table 1). The prevalence of mosaicism in the H11-Cas9 strategy, compared to the all-white mice produced by Rosa26-Cas9, suggests that there may be a difference in the level and/or timing of Cas9 expression driven by these two transgenes.

Extended data table 1

Coat color of F2 individuals that were constitutive Cas9+ and Tyr.

	Rosa26>Cas9	H11>Cas9
White	17	3
Mosaic	0	21

Our next goal was to determine what type of repair events (NHEJ mutations or gene conversions by HDR) were transmitted to the next generation. In order to assess inheritance in many offspring, we crossed each F2 male Rosa26-Cas9; Tyr and H11-Cas9; Tyr mouse to multiple albino CD-1 females (Fig. 1c), which carry a loss-of-function point mutation in Tyr exon 1 (Tyr, here designated as Tyr)[11,17]. We then genotyped F3 offspring of this test cross by PCR and DNA sequencing exon 5 to identify those that inherited the Tyr-marked receiver chromosome (Supplementary Fig. 4). In the absence of gene conversion, effectively none of these chromosomes would be predicted to also carry the Tyr allele due to separation of Tyr exons 4 and 5 by only ~9.1 kb.

Tyr animals should appear grey due to partial activity of the hypomorphic Tyr allele. However, among F3 offspring with this genotype, 100% in the Rosa26-Cas9 lineage and 90.4% in the H11-Cas9 lineage were completely white, indicating frequent transmission of a CRISPR/Cas9 induced loss-of-function mutation on the receiver chromosome and consistent with the primarily albino coat color of the F2 parents (Fig. 1c and Extended Data Table 1, 2). If the induced null alleles resulted from inter-homologue HDR to copy the Tyr allele from the donor to the Tyr-marked receiver chromosome, these white animals should also express the fluorescent mCherry marker. In these experiments, however, none of the F3 offspring that inherited the receiver chromosome in either the Rosa26-Cas9 or H11-Cas9 lineages expressed mCherry. PCR amplification of Tyr exon 4 confirmed that the Tyr element was not present in white Tyr F3 progeny (Extended Data Table 2).

Extended data table 2

Phenotypes and genotypes of all F3 progeny in Fig. 1c that are the offspring of a subset of F2 individuals listed in Extended Data Table 1.

			Chinchilla + F3 offspring
	F2 parent color	Total F3Chinchilla−	Total F3Chinchilla+	White(NHEJ mutation)	Grey(no cut or functionalrepair)	mChemy+(HDR geneconversion)
Rosa26 Family 1	white	30	20	20	0	0
Rosa26 Family 2	white	11	15	15	0	0
Rosa26 Family 3*	white	28*	16	16	0	0
Rosa26 Family 4	white	22	22	22	0	0
Rosa26 Family 5	white	16	16	16	0	0
H11 Family 1	prim. white mosak	14	15	15	0	0
H11 Family 2	print. grey mosak	23	28	25	3	0
H11 Family 3	prim. white mosaic	13	9	9	0	0
H11 Family 4	prim. white mosak	12	15	15	0	0
H11 Family 5	mosaic	23	24	18	6	0

indicates a family with offspring that were Tyr negative and mCherry negative, suggesting a large deletion in the recipient chromosome may encompass the Tyr SNP

“prim.” = “primarily” and refers to the coat color that covers the greatest total area when there was obviously not an equal proportion of white and grey

The different propensities to yield full albino versus mosaic coat color patterns in the Rosa26-Cas9 and H11-Cas9 lineages were paralleled by differences in the number of unique NHEJ mutations that we identified on receiver chromosomes in individuals of each genotype. Sequenced PCR products from Rosa26-Cas9; Tyr F2 tails (somatic tissues comprised of both ectodermal and mesodermal derivatives) and from individual F3 outcross offspring (representing the germline) typically exhibited only two unique NHEJ mutations suggesting that many of these Cas9 induced mutations may have been generated in 2–4 cell stage embryos (average 2.4 alleles among offspring of five families; Extended Data Fig. 2 and Extended Data Table 3). In contrast, H11-Cas9; Tyr F2 tails and F3 offspring harbored significantly more unique NHEJ mutations (average 4.6 alleles in five families; two-tailed Student’s t-test, p=0.041), consistent with the hypothesis that Cas9 is expressed at a later embryonic stage and/or at lower levels in this lineage.

Extended data figure 2

Rosa26-Cas9 and H11-Cas9 constitutive lineages have different numbers of unique NHEJ indels.

Sanger sequencing of the Tyr4a-gRNA target exon amplified from tail tip genomic DNA using TyrHALF2 and TyrHARR2 primers as specified in Supplementary Table 3. Above: A single representative Sanger sequence trace of the bulk PCR product amplified from a Rosa26-Cas9, Tyr F2 animal (Rosa26 Family 1 in Extended Data Table 3) with either major or minor peaks called revealing two distinct alleles. Five Tyr F3 offspring of this F2 individual each match one of the two alleles [marked 1 (insertion) and 2 (deletion)]. Below: A single representative sequence trace of the bulk PCR product amplified from an H11-Cas9, Tyr F2 animal (H11 Family 1 in Extended Data Table 3). Alternate alleles cannot be called because of the complexity of overlapping peaks. Five Tyr F3 offspring each have one of four different alleles (marked 1, 2, 3, 4). Sequence trace data is representative of all 90 individuals total in five families of each constitutive strategy detailed in Extended Data Table 3.

Extended data table 3

Allelic complexity of the constitutive Rosa26- and H11-Cas9 families.

	Number of distinct NHEJ alleles in‘n’ Tyrchinchilla+ offspring
Rosa26 Family 1	2 (n=9)
Rosa26 Family 2	3 (n=10)
Rosa26 family 3*	2 (n=9)
Rosa26 Family 4	2 (n=9)
Rosa26 Family 5	3 (n=13)
H11 Family 1	4 (n=10)
H11 Family 2	6 (n=9)
H11 Family 3	4 (n=6)
H11 Family 4	2 (n=8)
H11 Family 5	7 (n=9)

indicates a family with offspring that were Tyr negative and mCherry negative, suggesting a large deletion in the recipient chromosome may encompass the Tyr SNP. This was counted as one of the two unique NHEJ indels.

We considered two explanations for the observation that Tyr was not copied to the receiver chromosome in the early embryo. The first possibility is that homologous chromosomes are not aligned for inter-homologue HDR to repair DSBs. Second, the DNA repair machinery in somatic cells typically favors NHEJ over HDR[8,9]. A possible solution to overcome both potential obstacles is to restrict CRISPR/Cas9 activity to coincide with meiosis in the developing germline. During meiosis, recombination of the maternal and paternal genomes is initiated by the formation of DSBs that are repaired by exchanging regions of homologous chromosomes that are physically paired during Meiosis I[18]. Indeed, the molecular mechanisms of NHEJ are actively repressed to favor HDR during meiosis in many species, including mice[19].

In order to test the hypothesis that CRISPR/Cas9 activity will convert a heterozygous active genetic element to homozygosity during meiosis, we designed a crossing scheme to initiate Cas9 expression during germline development in Tyr mice. Since no currently available transgenic mice express Cas9 under direct control of a germline-specific promoter, we combined conditional Rosa26- or H11-LSL-Cas9 transgenes, each with a Lox-Stop-Lox preceding the Cas9 translation start site[15,16], with available Vasa-Cre or Stra8-Cre germline transgenic mice. Vasa-Cre is expressed later than the endogenous Vasa transcript in both male and female germ cells[20] while Stra8-Cre is limited to the male germline and is initiated in early stage spermatogonia[21]. Although oogonia and spermatogonia are pre-meiotic, and spermatogonia are in fact mitotic, we reasoned that Cre protein must first accumulate before Cas9 can be expressed from the LSL-Cas9 conditional allele. The possible time delay may require initiation of Cre expression before the onset of meiosis so that Cas9-induced DSBs can be resolved by inter-homologue HDR prior to segregation of homologous chromosomes at the end of Meiosis I. We generated each combination of these Cre and conditional Cas9 lines in case the timing or levels of Cas9 expression are critical variables in these crosses. We also assessed males and females of the Vasa strategies in case there are sex-dependent differences in animals that inherit the same genotype.

Males heterozygous for Tyr and the Vasa-Cre transgene were crossed to females homozygous for both the Tyr allele and one of the two conditional Cas9 transgenes (Fig. 2). We avoided the reverse cross using female Vasa-Cre mice, since Cre protein maternally deposited in the egg[20] might induce recombination of the conditional Cas9 allele and induce mutations in the early embryo similar to what we observed in the experiments using constitutive Cas9 transgenes. Introducing the Vasa-Cre transgene by inheritance from the male instead resulted in most offspring that were entirely grey, due to the Tyr genotype, and a few mosaic animals (Supplementary Table 4). The presence of any mosaicism suggests that this conditional approach to restrict Cas9 expression to the germline resulted in some spurious cleavage of the Tyr locus in somatic tissues.

Figure 2 |

Breeding strategy to produce Tyr mice with a conditional Cas9 transgene and a germline restricted Cre transgene.

F3 offspring were test crossed to Tyr animals to assess F4 phenotypes and genotypes. Quantification of observed outcomes in F4 test cross offspring are presented in Table 1.

We first tested whether Cas9 in the female germline could promote copying of the Tyr element onto the receiver chromosome by crossing F3 female mice of each Vasa-Cre lineage to CD-1 (Tyr) males. In each cross, we identified F4 offspring that inherited the Tyr-marked chromosome (Fig. 2). As in the crosses to assess the effects of embryonic Cas9 expression, we expected Tyr mice without a loss-of-function mutation in exon 4 of the receiver chromosome would be grey. Mice with a CRISPR/Cas9 induced NHEJ mutation in exon 4 should be white. Similarly, mice carrying a CRISPR/Cas9 induced mutation that was repaired by inter-homologue HDR should also be white, but should additionally fluoresce red due to transmission of the mCherry-marked Tyr active genetic element.

Table 1 summarizes the results of these crosses that demonstrate gene conversion upon Cas9 expression in the female germline. In contrast with early embryonic expression of Cas9, we observed that the Tyr transgene was copied to the Tyr-marked receiver chromosome in both Vasa-Cre; Rosa26-LSL-Cas9 and Vasa-Cre; H11-LSL-Cas9 lineages. However, the observed efficiency differed between genotypes; three out of five females of the Vasa-Cre; Rosa26-LSL-Cas9 lineage and all five of five females of the Vasa-Cre; H11-LSL-Cas9 lineage transmitted a Tyr-marked chromosome containing a Tyr insertion to at least one offspring. Although there was considerable variation among females with the same genotype, the highest observed frequency of gene conversion (72.2%) within a germline produced 13 out of 18 Tyr offspring with a Tyr insertion in the Vasa-Cre; H11-LSL-Cas9 lineage (Table 1 and Extended Data Table 4). The probability of obtaining an animal with this genotype by natural meiotic recombination mechanisms is very low (4.7×10−5) due to ultra-tight linkage between the Tyr and Tyr alleles. Although it seems likely that inter-homologue HDR of Cas9-induced DSBs utilizes the same DNA repair machinery that is active during meiotic recombination, these copying events cannot be explained by an increased incidence of chromosomal crossover, since all animals that inherited the donor chromosome lacking the Tyr marker expressed mCherry (Extended Data Table 4).

Table 1:

Observed F4 outcomes of germline Cas9 strategies.

Cre	Cas9	F3 parent(F)emaleor (M)ale		No cut orfunctionalrepair	NHEJdisruption	HDRconversion	%HDRconversionobserved
Vasa	Rosa26	F	1	-	15	-	-
			2	1	2	-	-
			3	-	3	1	25%
			4	4	5	3	25%
			5	6	6	1	8%
		M	1	-	25	-	-
			2	-	17	-	-
	H11	F	1	4	1	13	72%
			2	10	7	4	19%
			3	4	-	5	56%
			4	8	4	4	25%
			5	8	3	10	48%
		M	1	-	15	-	-
			2	-	5	-	-
			3	-	2	-	-
			4	-	3	-	-
Stra8	Rosa26	F	1	-	19	-	-
			2	-	3	-	-
	H11	M	1	3	21	-	-

Extended data table 4

Phenotypes and genotypes of all F4 progeny that are the offspring of a subset of F3 individuals listed in Supplementary Table 4.

In contrast to the 41 copying events we observed out of 132 total Tyr offspring of female mice, we observed no copying in a total of 113 offspring of males where conditional Cas9 expression was induced by either Vasa-Cre or Stra8-Cre (Table 1 and Extended Data Table 4). It is possible, however, that the number of families and of offspring in each family, which was limited by unexplained low male fertility, was insufficient to detect low efficiency copying in each of four genetic strategies. If there is indeed a difference between males and females in the efficiency of Tyr copying, we can conceive of two potential explanations. First, despite equivalent genotypes in the male and female Vasa-Cre lineages, Cre, Cas9, and/or gRNA may not be well expressed in the male germline. However, the high frequency of white Tyr offspring suggests that DSB formation is very efficient in males. Second, spermatogonia continually undergo mitosis and produce new primary spermatocytes throughout the life of a male in mammals[22]. In contrast, oogonia directly enlarge without further mitosis to form all of the primary oocytes in the embryo[23]. The difference in the observed efficiency of inter-homologue HDR between females and males at this locus may therefore reflect a requirement for the precise timing of CRISPR/Cas9 activity to coincide with meiosis (Fig. 3). NHEJ indels in males may result from DSB repair that occurs prior to the alignment of homologous chromosomes in Meiosis I. Similarly, the higher observed efficiency of inter-homologue HDR in females of the H11-LSL-Cas9 conditional strategy may relate to the lower or delayed Cas9 expression from the H11 locus, compared to Rosa26, that was evident in the constitutive crosses (Fig. 1d, e and Extended Data Table 1 and 3). Thus, in the Vasa-Cre; H11-LSL-Cas9 mice, DSB formation may have been fortuitously delayed to fall within a more optimal window during female meiosis.

Figure 3 |

Gene conversion by an active genetic element was observed in the female germline and not in the male germline or in the early embryo.

Schematic representation of early embryonic and male and female germline development overlaid with the presence or absence of observed HDR. [PGCs: primordial germ cells, n: number of homologous chromosomes, c: chromosome copy number. Asterisk indicates the difference between male sperm (n, 1c) and female ovum, which remains (n, 2c) until second polar body extrusion after fertilization.]

In summary, we demonstrate that the fundamental mechanism of a CRISPR/Cas9-mediated gene drive is feasible in mice. However, our comparison of eight different genetic strategies indicates that precise timing of Cas9 expression may present a greater challenge in rodents than in insects in order to restrict DSB formation to a window when breaks can be efficiently repaired by the endogenous meiotic recombination machinery. These data are therefore critical to the ongoing discussion about whether CRISPR/Cas9-mediated gene drives could be used to reduce invasive rodent populations, since it appears that both the optimism and concern are likely premature. Further optimization to increase the frequency of gene conversion in both males and females and to reduce the prevalence of drive-resistant alleles (NHEJ indels that alter the gRNA target site) would be necessary to achieve rapid and sustained suppression of wild populations[24-29].

Nevertheless, the copying efficiencies we observed here would be more than sufficient for a broad range of laboratory applications. For example, the average observed copying rate of 44% using the most efficient genetic strategy in females (Vasa-Cre; H11-LSL-Cas9) combined ultra-tightly linked Tyrosinase mutations such that 22.5% of all offspring inherited a chromosome with both alleles, an impossibility by Mendelian inheritance. This observed average copying rate would also be expected to increase the inheritance of a single desired allele from 50% to 72%, and the highest rate of gene conversion we observed (72.2%) would result in an 86% frequency of transmitting a desired allele. If multiple genes could be simultaneously converted to homozygosity, such high transmission frequencies that bypass the onerous constraint of genetic linkage stand to greatly accelerate the production of rodent models for a variety of complex genetic traits.