Robust and efficient knock-in in embryonic stem cells and early-stage embryos of the common marmoset using the CRISPR-Cas9 system

Genome editing technology greatly facilitates the genetic modification of various cells and animals. The common marmoset (Callithrix jacchus), a small non-human primate which exhibits high reproductive efficiency, is a widely used animal model in biomedical research. Developing genome editing techniques in the common marmoset will further enhance its utility. Here, we report the successful establishment of a knock-in (KI) method for marmoset embryonic stem cells (ESCs), which is based on the CRISPR-Cas9 system. The use of CRISPR-Cas9, mediated by homologous recombination (HR), enhanced the KI efficiency in marmoset ESCs. Furthermore, we succeeded in performing KI in early-stage marmoset embryos. In the course of the experiments, we found that HR in the marmoset ESCs is innately highly efficient. This suggested that the marmoset possesses a repair mechanism for DNA double-strand breaks. The current study will facilitate the generation of genetically modified marmosets and gene function analysis in the marmoset.

The common marmoset (Callithrix jacchus) is highly useful as a non-human primate biomedical model, because of its high fertility, short gestation period (144 days), and high physiological similarity to humans 9,10 . Our group has previously reported the successful generation of transgenic and KO marmosets 11,12 and demonstrated the utility of marmoset models for studying diseases. However, the methods to generate such models are still under development and should be improved to enable more efficient generation. Here, we present a HR-mediated gene targeting method combined with the CRISPR-Cas9 genome editing technique in marmoset ESCs (cjESCs). Using this method, we obtained high KI efficiencies in several different genetic loci of cjESCs. Furthermore, we used this method to successfully perform KI in early-stage marmoset embryos. Surprisingly, not only did we establish a method to obtain high KI efficiency in cjESCs, but also, we found that cjESCs possess innate high HR activity. Especially, in the case of targeting exon 1 in the proteolipid protein 1 (PLP1) gene, almost all of the clones that survived positive selection were homologous recombinants, even without using CRISPR-Cas9. This unique feature will be useful for future studies in disease modelling and gene function analysis in marmosets, and possibly other non-human primates.

Results
Evaluation of KI efficiency in cjESCs using the CRISPR-Cas9 system. To test whether the CRISPR-Cas9 system works in marmoset cells, we introduced marmoset-specific gRNA sequences (summarized in Supplementary Table 2) into pSpCas9-2A-Puro (Cas9-gRNA vector; PX459) and evaluated the genomic cleavage activity (GCA) of Cas9 and the gRNAs in cjESCs by transfecting each Cas9-gRNA vector and transiently selecting the transfected cjESCs with puromycin (see Experimental procedures). We confirmed the GCA of Cas9 and all of the gRNAs to be used in the current study ( Supplementary Fig. S1a-h).
Next, we decided to target the ACTB gene locus to test whether CRISPR-Cas9 enhances KI efficiency in cjESCs. Previously, by using a promoter-trapping ACTB-EGFP targeting vector (TV) carrying a G418 resistance gene (Fig. 1a) 13 , we confirmed that most of the G418-resistant and EGFP-positive colonies are homologous recombinants. Therefore, the number of G418-resistant and EGFP-positive colonies are considered to be indexes for KI efficiency.
We next evaluated the KI efficiency using three newly constructed ACTB-EGFP TVs with shortened homology arms (Fig. 1f). As expected, using the shortened TVs resulted in the reduction of KI efficiency in the control group that was not transfected with Cas9-gRNA. However, we did not see a decrease in KI efficiency when Cas9-gRNA (gRNA2) was transfected (Fig. 1g).
In addition, in order to estimate the KI efficiency without having to perform positive selection, we also evaluated the transfection efficiency and colony formation efficiency immediately following transfection. Transfection with a mVenus expression vector (pCXN2-mVenus) revealed that the transfection efficiency was 32.0 ± 6.3% (n = 5), and colonies were formed from 1.97 ± 0.26% (n = 4) of passaged cjESCs. Thus, from 1 × 10 6 cjESCs, approximately 6300 colonies were transfected and expected to form colonies before positive selection. Accordingly, in the gRNA2 and control group, the targeting efficiency of transfected colony-forming cjESCs was calculated to be approximately 1.43% (gRNA2) and 0.13% (control). To experimentally validate this approximate calculation, we performed fluorescent-activated cell sorting (FACS) analysis. In short, we transfected cjESCs with the ACTB-EGFP TV and Cas9-gRNA vector (gRNA2), and transiently selected the cells with puromycin. These cjESCs were further expanded, and the proportion of EGFP-positive (EGFP(+)) cells was analyzed by FACS. The PX459 alone was used as the control. In the control group (gRNA(−)), there were few EGFP(+) cells, calculated to be around 0.18 ± 0.05% ( Supplementary Fig. S3a). In the gRNA2 group (gRNA(+)), the percentage of EGFP(+) cells were increased to 1.75 ± 0.17% ( Supplementary Fig. S3b), which was a significant increase when compared to the control (P < 0.001; Supplementary Fig. S3c). This result indicates that the number of counted cjESC colonies which underwent positive selection reflects KI efficiency, and the approximate calculation obtained using the mVenus expression vector helps to translate the number of counted colonies into KI efficiency to some extent.
Furthermore, to further verify that a sampling or differentiation bias did not influence the KI efficiency of cjESCs, we performed quantitative reverse transcription-PCR (qRT-PCR) analysis of pluripotency-related genes OCT4 and NANOG. Results showed that the cjESCs maintained a similar undifferentiated state either with or without G418 selection, which indicates that the high KI ratio was not due to a differentiation bias of the cjESCs during G418 selection ( Supplementary Fig. S6b). In addition, the efficient KI of PLP1 exon 1 without using Cas9-gRNA was confirmed in another cjESC line ( Supplementary Fig. S6c). These observations suggested an innate high HR activity in cjESCs.
Humanization of the FOXP2 gene. Since PLP1 is an X-chromosomal gene, we next tested our method in FOXP2, an autosomal gene. Forkhead box protein P2 (FOXP2) gene is a transcription factor related to lung and corticobasal development 17,18 . In addition, several evolutional and pathogenic mutations in the gene are associated with the language ability of humans 19,20 .
In conclusion, by using the FOXP2 TV and Cas9-gRNA (gRNA: FOXP2-4) vector, we were able to successfully generate homozygous KI clones harbouring human-specific mutations at the correct site.

Evaluation of KI efficiency in early-stage marmoset embryo.
We have demonstrated that the developed KI method for cjESCs resulted in high KI efficiency. Next, we attempted to apply this KI method to early-stage marmoset embryos (Fig. 4a). For this experiment, we selected the PLP1 exon 2 KI construct (Fig. 4b), since this construct enabled homologous recombination in cjESCs the most efficiently (Supplementary Figs S7eh, S8c, S9, and Table 5) among the experiments described above. In addition, the PLP1-CDS2-2 gRNA induced KI the most efficiently among the four gRNAs targeting PLP1 exon 2 ( Supplementary Fig. S7f) with the least concern of mosaicism in cjESCs (Supplementary Table 5  early-stage embryos. We used the PLP1-P15L TV encoding the pathogenic P15L substitution 22 for PLP1 exon 2 and harbouring a SacI restriction enzyme site instead of an ApaI restriction enzyme site in the WT allele (Fig. 4b).
We performed genotyping PCR using the entire genomic DNA extracted from each embryo which developed to the 8-cell stage or beyond (Fig. 4a, right). The PCR products were subjected to restriction fragment-length polymorphism (RFLP) analysis and DNA sequencing after subcloning the product into a blunt vector. The RFLP analysis of the KI using the TV was first validated in cjESCs ( Fig. S11a and b). Based on preliminary experiments (data not shown), we optimized the concentration of each component in the microinjection solution. The optimized solution consisted of the Cas9 protein (100 ng/μl), annealed crRNA and tracrRNA (50 ng/μl), and TV (100 ng/μl) in nuclease-free water, which were injected into the cytoplasm of two pronuclear-stage (2PN) zygotes (Fig. 4a, left). For the microinjection, 38 embryos were used; 18/38 (48.6%) embryos developed to the 8-cell stage or beyond; and 17 of them were genotyped. Surprisingly, under this condition, all the genotyped embryos were genetically modified (Fig. S12a and c). In 4/17 (24%) embryos (#2, #9, #12, #14), the PCR fragment was digested by SacI, which confirmed KI (Fig. 4c, Supplementary S12b). KI was further validated with DNA sequencing of the subcloned PCR fragments (Fig. 4d and Supplementary Fig. S12c). At least 7 subclones per embryo were analyzed. However, more than two sequence variations were found in 7/17 (41%) embryos, which showed that mosaicism occurred in these embryos (Supplementary Fig S12c). These results are summarized in Supplementary Fig. S12a. Furthermore, we utilized a zygote-electroporation method for early-stage marmoset embryos. Referring to previous studies performed in mice embryos 23,24 , we set an electroporation condition for marmoset embryos. Components of the electroporation solution included the Cas9 protein (100 ng/μl), annealed crRNA and tracrRNA (50 ng/μl), and the TV (100 ng/μl) in 1x OPTI-MEM. Additionally, since successful KI using a double-stranded DNA vector has never been achieved by electroporation in rodent embryos 23-25 , we utilized a 200-bp single-stranded oligonucleotide (ssODN) which overlaps the DSB site of the PLP1-CDS2-2 gRNA, and encodes the P15L substitution, SacI site, and silent mutations, included for identifying the type of KI template transfected in each embryo. The ssODN-mediated KI was validated in cjESCs by RFLP analysis. The ssODN showed lower KI efficiency (2.8% and 4.1%) compared to that with the TV (3.9% and 8.3%) in our condition (n = 2; Supplementary Fig. S11b), different from a previous report showing an increased efficiency in KI for ssODN relative to TV 26 but consistent with a previous study using murine ESCs 27 , However, further experiments would be required to make a definitive conclusion on this subject. The sequence of the ssODN is appended in the Supplementary Information. For these experiments, 7 embryos were used, 6 (86%) of them developed to the 8-cell stage or beyond, and subsequently genotyped. We found that 4/6 (66%) embryos carried genetically modified alleles, and 2/6 (33%) embryos had KI alleles resulting from transfection of either the ssODN or the TV, which were detected by DNA sequencing and RFLP analysis ( Supplementary Fig. S13b and c). The data are summarized in Supplementary Fig. S13a.
Thus, by using the microinjection and electroporation conditions we developed and optimized, we succeeded in efficiently introducing gene modifications in marmoset embryos, resulting in either KI or KO.

Discussion
In the current study, we established a gene targeting method for cjESCs and early-stage embryos, using CRISPR-Cas9 for directed DSB. This method dramatically increased the number of colonies that survived positive selection, and enabled bi-allelic homologous recombination. Furthermore, this method is robust, since KI of several genetic loci, including genes that are normally not expressed in undifferentiated cjESCs, was obtained with efficiencies similar or even higher than that of ACTB. Although the KI efficiency among total transfected cjESCs were low, experimentally shown to be between 1.8-8.3% with Cas9-gRNA (Supplementary Figs S3b and S11b), the percentage of KI clones among cjESC clones that survived positive selection was over 80% in most cases (Figs 2e, 3f and Supplementary Fig. S7h), which is higher than that of previous studies recently reported using a similar strategy in human ESCs and induced pluripotent stem cells (iPSCs) [28][29][30][31][32][33] and in macaque monkey ESCs (less than 50%) 34 . Additionally, in our method, we succeeded in obtaining over 10% of homozygous-KI clones in most cases (Figs 2e, 3f and Supplementary Fig. S7h). This was surprising, as homozygous KIs has been considered to be difficult to obtain using the conventional KI system in human iPSCs 33 .
We also observed an innate high HR activity of cjESCs, which may have also contributed to the extremely high KI efficiency revealed in this study. In the course of the current study, we observed a high KI ratio when targeting PLP1 exon 1, even in the absence of site-directed DSB with Cas9-gRNA. We were able to obtain KI clones with 88.6% (31/35 clones) efficiency without negative selection, one of which was a homozygous KI clone ( Fig. 2e and Supplementary Fig. S4). Since gene targeting efficiency of PLP1 exon 1 in mouse ESCs was 2% using a similar construct 13 , we consider our result in cjESCs to be extraordinary. Although this high efficiency was not observed at other loci without Cas9-gRNA, an HR bias may have occurred in cjESCs. Further studies to explore the mechanism of HR in cjESCs will be useful for improving the KI efficiency in other species.
Since the PLP1 gene is crucial for the stability of myelin which are formed by OLs 35 , its mutation, deletion, or duplication leads to a functional impairment of the central nervous system 14 . Some PLP1 missense mutations cause severe phenotypes, yet null mutations or genetic deletions of the gene causes mild phenotypes 36 . Furthermore, the PLP1 gene is thought to be associated with schizophrenia in mice and human 37,38 . Therefore, enabling genetic modification such as KO or introduction of missense mutation(s) in the PLP1 gene in non-human primates, as performed in the current study, would help to generate new disease models for analysing motor functions and higher brain functions, which could be used for testing drug candidates and cell implantation in a preclinical setting.
FOXP2 was originally identified as the gene responsible for hereditary language disorder 19 . As such, the gene is likely associated with the language acquisition of human beings during evolution, since two amino acid residues (N303 and S325) are specific to human FOXP2. Humanized Foxp2 mice exhibit neuroanatomical changes of the striata and enhanced memory learning 39,40 . Since some groups suggested the existence of homologues of the Broca's area in non-human primates, including marmosets [41][42][43] , FOXP2 would be an interesting candidate for gene modification in the marmoset.
In addition, we also established KI technology in early-stage embryos of the common marmoset, which are based on both microinjection and electroporation methods. We succeeded in obtaining KI embryos with around 30% efficiency. In the future, this technology could be utilized for the generation of KI marmosets.
In conclusion, we developed an efficient KI method for marmoset ESCs and early-stage embryos. The method may be used to generate KI animals and for analyzing gene functions in vitro using a non-human primate model. In addition, the HR bias, which occurs after DSB in cjESCs and was observed in several experiments presented herein, should be studied thoroughly for future development of the KI technology. The findings in the current study would facilitate the use of non-human primate species (marmosets) as bridging models to fill the gap between mouse and human.

Experimental Procedures
Animals. All protocols for animal experiments were performed in accordance with the guidelines for laboratory animals set forth by the National Institutes of Health, and the Ministry of Education, Culture, Sports, Science and Technology (MEXT) of Japan, and were approved by the Institutional Animal Care and Use Committee of the RIKEN (approval No. H27-2-306(4)). Animal care was conducted in accordance with the National Research Council (NRC) Guide for the Care and Use of Laboratory Animals (2011). Marmosets used in the current study were 2-6-years-old (average weight from 250 to 450 g). The marmosets were pair/family-housed in a warm and humid condition (25 °C, 65%). In total, thirty female marmosets were used as oocyte donors and eleven male marmosets were used as sperm donors. The marmosets were obtained from the in-house breeding colony at RIKEN Institute. Oocyte donors were kept pairwise with vasoligated males. Oocyte and sperm collection, and in vitro fertilization (IVF) was performed as previously described 12 . In brief, for oocyte collection, the oocyte donors whose plasma progesterone levels were monitored were intramuscularly injected with recombinant human follicle-stimulating hormone (FSH, 25 IU; Fuji Pharma) for 9 days, followed by the intramuscular injection of human chorionic gonadotropin (hCG, 75 IU; ASKA Pharmaceutical) on day 10. On day 11 (16-20 hours after the hCG injection), the hormone-treated female marmosets were pre-anesthetized with 0.04 mg/kg medetomidine . These cells were cultured as described previously 45 . In brief, cjESCs were cultured on 30 Gy-irradiated mouse embryonic fibroblasts (MEFs) in ES medium (ESM) consisting of 1x KnockOut DMEM (Thermo Fisher) supplemented with 20% KnockOut Serum Replacement (Thermo Fisher), 1mM L-glutamine (Nakalai Tesque), 1% non-essential amino acids (Sigma), 0.2 mM 2-mercaptoethanol (Sigma) and 10 ng/ml fibroblast growth factor 2 (Peprotech). The KI cjESC lines generated in the current study will be distributed by the corresponding authors upon request.

Transfection and genotyping. Double-stranded DNA targeting vectors (TVs) used for KI experiments in
cjESCs were linearized by one-cut restriction enzyme before transfection. Each Cas9-gRNA vector and linearized TV was prepared at 1 μg/μl in Tris-HCl-EDTA buffer (pH 8.0). For transfection, a total of 10 μg of DNA was transfected, which consisted of the TV and Cas9-gRNA vector at a 4:1 molar ratio. For cjESC transfection, DNA vector(s) (total 10 μg), lipofectamine-LTX PLUS reagent (2.5 μl; Thermo Fisher) and LTX reagent (25 μl; Thermo Fisher) were diluted in 500 μl OPTI-MEM (Thermo Fisher), and added into sub-confluent cjESCs cultured in one well of a 6-well plate. 24 hours after transfection, the cells were dissociated and counted. 1 × 10 6 cjESCs were suspended in ESM containing Y-27632 (10 μM; Merck Millipore), and re-seeded onto new feeder cells resistant to G418 or puromycin (day 1). On day 3, the medium was changed to ESM containing G418 (50 μg/ml; Thermo Fisher), or puromycin (1 μg/ml; Thermo Fisher) and Y-27632 (10 μM). Y-27632 was removed from the medium from day 5. After 2 weeks, the drug-resistant colonies were counted and picked for further cloning. For genotyping by PCR and Southern blotting, the cloned ESCs were lysed overnight at 55 °C in cell lysis buffer consisting of Tris-HCl (0.2 M), EDTA (10 mM), SDS (0.2%) and NaCl (0.2 M) in nuclease-free water with proteinase K (10 μg/ml). Genomic DNA was purified using a standard method with phenol-chloroform and ethanol. PrimeSTAR Max DNA polymerase (Takara) was used for genotyping PCR, according to the manufacturer's instructions. PCR was performed as follows: 30 s at 94 °C; 35 cycles of 10 s at 98 °C and 8 min at 68 °C; then 10 min at 68 °C; and a final incubation at 4 °C until gel electrophoresis. The primers are listed in Supplementary  Table 1. Southern blotting was performed as described previously 45 . For digestion of genomic DNA, we used XbaI (ACTB-EGFP), BglII (PLP1-EGFP) and EcoRV (FOXP2, PLP1-P216S, S253T and A39T) (purchased from Takara or NEB). The entire images of the gels in Southern blotting analysis were appended in Supplementary Fig. S14. Transient selection. The procedure for the transfection of cjESCs is described above. After transfecting each TV or ssODN (8 μg) with the Cas9-gRNA vector (2 μg), the cjESCs were re-seeded onto new feeder cells (day 1). On day 2, the medium was changed to ESM. On day 3, the medium was changed to ESM containing puromycin (0.2 μg/ml) and Y-27632 (10 μM). On day 5, Y-27632 was removed from the medium. From day 6, the medium was changed to ESM every other day for further expansion.  . To construct the PLP1-EGFP vector, a 3.2-kb fragment containing a region spanning upstream of the PLP1 gene to the PLP1 initiation codon, and a 5.0-kb fragment containing PLP1 intron 1 were ligated with the vector pHNEO-EGFP, which harbours promoter-less EGFP, bovine growth hormone polyadenylation signal (pA), and floxed G418-resistance gene (Neo) under the mouse Pgk-1 promoter (PGK).
To construct the PLP1-P216S and PLP1-S253T vectors, a 2.7-kb fragment containing PLP1 introns 2 to 5, and a 4.2-kb fragment containing PLP1 intron 5 and the downstream region of the PLP1 gene, were ligated with pSINTK. The 2.7-kb fragment was mutagenized by PCR using specific primers to obtain the sequence encoding the Pro216Ser substitution (CCT > TCT), or the 4.2-kb fragment was mutagenized to obtain the sequence encoding the Ser253Thr substitution (TCC > ACC).
To construct the PLP1-A39T and PLP1-P15L vectors, a 2-kb fragment containing a region from PLP1 intron 1 to intron 2, and a 1.4-kb fragment spanning PLP1 intron 2 to intron 4, were subcloned into pDONR vectors, and introduced into pDEST-R3R4(R) with pENTR-L1-PGK-PuroTK-pA-L2 47 . The 2-kb fragment was mutagenized by PCR to generate A39T or P15L substitutions, and silent mutations to render the vectors to become undetectable by the gRNAs. For gene targeting experiments in the embryos, the PGK-PuroTK cassette was removed from the PLP1-P15L vector by Cre recombinase (NEB).
All vectors were purified using plasmid DNA purification kit (Qiagen). The vectors used in the current study are listed in Supplementary Table 3. The listed vectors will be provided by Addgene (https://www.addgene.org) or the corresponding authors upon request.
Neuronal induction. The cjESCs were induced into neuronal cells, including OLs, using a previously described method with slight modifications 48 . Briefly, dorsomorphin (3 μM; Sigma), SB431542 (3 μM; Tocris Bioscience) and CHIR99021 (3 μM; Wako) were added at days 1-3 of embryoid body formation. Immunochemical analysis was performed using Hoechst 33258 (Sigma) and the antibodies listed in Supplementary Table 4. The detailed experimental protocols will be provided by the corresponding authors upon request.
Microinjection of the embryos. Marmoset embryos at the 2PN stage were prepared as described previously 12,49 . The microinjection solution consisted of annealed crRNA and tracrRNA (50 ng/μl; IDT), Cas9 protein (100 ng/μl; IDT) and TV (PLP1-P15L; 100 ng/μl), which were suspended in nuclease-free water. The TV was purified using QIAquick PCR purification kit (Qiagen). Approximately 5-10 pl of the injection solution was injected into the cytoplasm of the pronuclear stage embryos in M2 medium (Sigma). Following the microinjection, the embryos were cultured in ORIGIO sequential cleavage medium (Origio). The embryos that developed normally past the 8-cell stage were used for genotyping. The KAPA mouse genotyping kit (Kapa Biosystems) and PrimeSTAR Max DNA polymerase were used for embryo genotyping. Briefly, each developed embryo was washed in PBS drop once and then transferred into the extraction solution (3 μl) consisting of extraction buffer (1×) and extraction enzyme (2%), and was placed in the following conditions: 75 °C for 10 min, 95 °C for 5 min, and a final incubation at 4 °C. After extraction, PCR solution (22 μl) was added to the extract solution and centrifuged briefly before PCR reaction. The final PCR mixture (25 μl) contained the PrimeSTAR Max premix (1×) and primers (1.6 μM each). PCR was performed by temperature cycling as follows: 30 s at 94 °C; 40 cycle of 10 s at 98 °C and 150 s at 68 °C; 10 min at 68 °C; and a final incubation at 4 °C. A portion (4 μl) of the PCR solution was used for electrophoresis on 1% agarose gel, and the 1.5-2.2-kb DNA product band was extracted from the gel. The extracted PCR fragments were purified using a phenol-chloroform and ethanol method, and subcloned into pCR-BluntII-TOPO (Thermo Fisher) utilizing DH5αcompetent cells (Takara). Each cloned vector was sequenced using the BigDye Terminator v1.1 Cycle sequencing kit (Thermo Fisher) with the 3130xl DNA Analyzer (Applied Biosystems). For the RFLP analysis, a portion (4 μl) of the PCR mixture was used. The RFLP solution (20 μl) contained buffer L (1×), and ApaI or SacI (1 μl; Takara). The RFLP reaction was performed as follows: 4 hours at 37 °C, followed by 20 min at 80 °C, and a final incubation at 4 °C until gel electrophoresis.
RT-PCR and qRT-PCR. RNA was isolated using the RNeasy mini kit (Qiagen) according to the manufacturer's protocol. Total RNA (1.0 μg) was reverse-transcribed in the ReverTra Ace qPCR RT master mix (Toyobo). The Scientific RepoRts | (2019) 9:1528 | https://doi.org/10.1038/s41598-018-37990-w resultant cDNAs were diluted in nuclease-free water (to 4 ng/μl). RT-PCR was performed using the PrimeSTAR Max DNA polymerase according to the manufacturer's instructions. qRT-PCR was performed using the TB Green Premix Ex Taq II (Takara) on Viia 7 (Applied Biosystems) according to the manufacturer's instructions. The primers used are listed in Supplementary Table 1. DNA sequencing. DNA sequencing analysis was performed using the BigDye Terminator v1.1 cycle sequencing kit (Thermo Fisher) with the 3130xl Genetic Analyzer (Applied Biosystems). The sequence data presented in the figures were illustrated using the Snap Gene software (GSL Biotech).
Statistical analysis. All data are expressed as mean ± s.e.m. Differences between means were compared using the Student's t-test. Differences were considered significant at P < 0.05.