In vivo targeted single-nucleotide editing in zebrafish

To date, several genome editing technologies have been developed and are widely utilized in many fields of biology. Most of these technologies, if not all, use nucleases to create DNA double-strand breaks (DSBs), raising the potential risk of cell death and/or oncogenic transformation. The risks hinder their therapeutic applications in humans. Here, we show that in vivo targeted single-nucleotide editing in zebrafish, a vertebrate model organism, can be successfully accomplished with the Target-AID system, which involves deamination of a targeted cytidine to create a nucleotide substitution from cytosine to thymine after replication. Application of the system to two zebrafish genes, chordin (chd) and one-eyed pinhead (oep), successfully introduced premature stop codons (TAG or TAA) in the targeted genomic loci. The modifications were heritable and faithfully produced phenocopies of well-known homozygous mutants of each gene. These results demonstrate for the first time that the Target-AID system can create heritable nucleotide substitutions in vivo in a programmable manner, in vertebrates, namely zebrafish.

Evaluation of the Target-AID system in zebrafish embryos. To investigate whether the Target-AID system induced C > T substitutions at the target site in injected embryos, we injected dCas9-PmCDA1 mRNA or nCas9-PmCDA1 mRNA with chd sgRNA into zebrafish embryos at the 1-cell stage and incubated the embryos until 3 days post fertilization (dpf). We then analyzed sixteen uninjected and sixteen injected embryos, all of which showed wild-type phenotypes, by deep sequencing (Fig. 2A). The uninjected embryos had no mutation around the target site in the chd locus. The dCas9-PmCDA1 and nCas9-PmCDA1 mRNA-injected embryos contained targeted C > T substitutions at the -19 position (c.232 C) in the chd locus, with a frequency of 2.19% and 4.37%, respectively. Moreover, the nCas9-PmCDA1 mRNA-injected embryos contained C > A, G > T, G > C, or G > A substitutions with low frequencies; each of these substitutions was observed less than 1% in all samples sequenced. Of note, insertions and deletions (indels) were detected approximately 9 times less frequently in the dCas9-PmCDA1 mRNA-injected embryos (0.91%) than in the nCas9-PmCDA1 mRNA-injected embryos (8.48%), which is consistent with previous reports in yeast and cultured cells 16 . Heteroduplex mobility assays (HMA) also revealed that dCas9-PmCDA1 induced fewer indel mutations in the chd locus than nCas9-PmCDA1 and Cas9 (Suppl. Fig. 1A-C). These results demonstrate that the Target-AID system is able to induce C > T substitutions predominantly at the target site in zebrafish embryos.
Next, to examine if the Target-AID system produced off-target mutations in zebrafish, we performed deep sequencing for potential off-target sites. The potential off-target sites were predicted by CCtop software (http:// crispr.cos.uni-heidelberg.de/index.html) 19 based on the off-target mismatch score. We selected both of the potential off-target sites with the canonical (5′-NGG-3′) and non-canonical PAM sequences (5′-NAG-3′), as we previously reported 16 . With respect to the chd sgRNA, we analyzed three potential off-target sites. These potential off-target sites have 1-to 4-base mismatches from the true target site. Deep sequencing revealed that off-target#1 and off-target#2 contained no mutations in the uninjected embryos, none in dCas9-PmCDA1 mRNA-injected embryos, and none in nCas9-PmCDA1 mRNA-injected embryos ( Fig. 2A). Off-target#3 had A > G and C > T substitutions at two positions (-10 and -12), both with almost 100% frequencies, which were most likely due to pre-existing single nucleotide polymorphisms rather than induced substitutions ( Fig. 2A).
Next, we evaluated the oep gene locus (Fig. 2B). In each experiment, we analyzed sixteen embryos injected with oep sgRNA and dCas9-PmCDA1 mRNA, or sixteen embryos injected with oep sgRNA and nCas9-PmCDA1 mRNA, by deep sequencing. Collectively, as expected, targeted C > T substitutions were observed in the dCas9-PmCDA1 mRNA-injected embryos (2.27% at the -19 position (c.175 C), and 1.91% at the -16 position (c.178 C)). In the nCas9-PmCDA1 mRNA-injected embryos, targeted C > T substitutions were detected only at the -19 position (1.04%). Indel mutations were detected approximately 2 times less frequently in the dCas9-PmCDA1 mRNA-injected embryos (0.60%) than in the nCas9-PmCDA1 mRNA-injected embryos (1.17%). Similar to what we observed in the chd locus, HMA assays revealed that dCas9-PmCDA1 induced fewer indel mutations in the oep locus than nCas9-PmCDA1 and Cas9 (Suppl. Fig. 1D-F). Thus, the Target-AID system can induce targeted C > T substitutions with few indels in the oep gene locus in zebrafish embryos.
We next examined off-target effects in the oep locus. For the oep sgRNA, we analyzed two potential off-target sites. The potential off-target sites predicted by CCtop contained 4-base mismatches from the true target site. Deep sequencing revealed that both the potential off-target sites contained no mutations in the uninjected embryos, none in the dCas9-PmCDA1 mRNA-injected embryos, and none in the nCas9-PmCDA1 mRNA-injected embryos (Fig. 2B).
In summary, both constructs, dCas9-PmCDA1 and nCas9-PmCDA1, were able to induce targeted C > T substitutions in zebrafish embryos without apparent off-target effects, at least in the putative off-target sites. In particular, dCas9-PmCDA1 induced targeted C > T substitutions in both loci with fewer indel mutations than nCas9-PmCDA1 in zebrafish embryos. Importantly, the ratio of base substitutions to indel mutations 20 with dCas9-PmCDA1 was higher than that with nCas9-PmCDA1 (in the chd locus, 2.41 for dCas9-PmCDA1 and 0.52 for nCas9-PmCDA1; and in the oep locus, 3.78 for dCas9-PmCDA1 and 0.97 for nCas9-PmCDA1). We therefore chose dCas9-PmCDA1 for the following experiments.
Generation of G 0 adult zebrafish carrying the expected nucleotide substitutions. We injected dCas9-PmCDA1 mRNA with the chd sgRNA or oep sgRNA into zebrafish embryos to generate G 0 adult zebrafish (Suppl. Fig. 2). The adult fish are referred to as chd G 0 fish or oep G 0 fish, respectively. To investigate whether the G 0 fish had the expected nucleotide substitutions at the target sites, PCR amplification of the targeted sequences was performed using genomic DNA from the caudal fins of five male chd G 0 fish, and the PCR fragments were sequenced. Nucleotide substitutions were detected as extra peaks below the major peaks, likely due to mosaicism of the G 0 fish (black and red arrowheads in Suppl. Fig. 3). In the analyses of chd G 0 fish, two sequence patterns were obtained. One pattern, seen in three chd G 0 fish, contained no detectable change (pattern (i) in Suppl. Fig. 3A); in the other pattern, two chd G 0 fish had nucleotide substitutions at the target site, c.232 C (pattern (ii) in Suppl. Fig. 3A). No deletions were found around the target nucleotides in the chd gene. There were no detectable changes in non-targeted cytosines in the adjacent region around the target site in all analyzed chd G 0 fish.
In the analyses of five male oep G 0 fish, one fish had no detectable changes (pattern (i) in Suppl. Fig. 3B), and among the other fish, three types of nucleotide substitutions were observed (pattern (ii) to (v) in Suppl. Fig. 3B). The first type had a c.175 C > T substitution (black arrowheads in Suppl. Fig. 3B). The second type had a c.178 C > T substitution (red arrowheads in Suppl. Fig. 3B). In the third type, both c.175 C > T and c.178 C > T substitutions were present. Pattern (v), seen in one oep G 0 fish, likely arose from mosaicism involving the expected nucleotide substitution of c.175 C and c.178 C, as well as at least one small deletion (Suppl. Fig. 3B). According to the experimental design, we expected that c.175 C and c.178 C, but no other nucleotide, would be mutated. Indeed, there were no detectable changes at other cytosines in the adjacent region around the target site in all analyzed oep G 0 fish.
The induced nucleotide substitutions were mostly heritable in the next generation. In order to examine whether the nucleotide substitutions induced by the Target-AID system in the G 0 fish were heritable, genotyping of descendant embryos of the G 0 fish was performed. Wild-type females were bred with the chd G 0 fish or oep G 0 fish (Suppl. Fig. 2), and descendant embryos were obtained from three chd G 0 fish and four oep G 0 fish (chd F 1 embryos and oep F 1 embryos, respectively). We then performed PCR amplification of the targeted sequences using genomic DNA from F 1 embryos and sequenced the PCR products. Nucleotide substitutions were detected as superimposed peaks because the F 1 embryos were mostly heterozygous at the target loci (black and red arrowheads in Fig. 3).
Analysis of the genomic DNA from chd F 1 embryos revealed two patterns in the chd loci (Fig. 3A). Expected c.232 C > T nucleotide substitutions were detected with a frequency of 29.4% (10/34). No detectable change was observed in the rest of the chd F 1 embryos, 70.6% (24/34). Moreover, no other nucleotide substitutions were found within the sequenced regions. Germ line transmission of the modified allele was achieved in two of the three chd G 0 fish (Suppl. Table 1).
From the analyses of oep F 1 embryos from four oep G 0 fish, we observed six patterns of sequence results (  (2/49). Three of the four oep G 0 fish successfully transmitted the modified alleles (Suppl. Table 2). Note that one oep G 0 fish with both the c.175 C > T and c.178 C > T substitutions (pattern (iv) in Fig. 3B) produced only wild-type F 1 embryos ( Fig. 3B and Suppl. Table 2). Taken together, these results indicate that the nucleotide substitutions induced by the Target-AID system are heritable.
We next raised the rest of the F 1 embryos to adulthood (Suppl. Fig. 2) and performed allele-specific PCR ( Fig. 4A-C, E-G) and sequencing (Fig. 4D,H) to identify heterozygous F 1 fish carrying the desired nucleotide substitutions at the target sites. For the identification of promising candidate heterozygous chd F 1 fish carrying the c.232 C > T substitution (chd c.232C>T/+ ), we performed two PCR reactions using allele-specific primers, one for the wild-type allele (w) and the other for the mutated allele (m). Both produced 203 bp fragments (Fig. 4B,C). Another set of primers, which was designed to amplify the chd gene outside of the target site, was also included in the PCR reaction. This produced 635 bp DNA fragments as internal positive controls (Fig. 4B,C). When the 203 bp DNA fragment was detected only in a "w" lane, the result indicated that the fish did not carry the c.232 C > T substitution, and the fish was scored as a "negative" or "wild-type" fish (chd +/+ fish). If the 203 bp DNA fragments were detected in both "w" and "m" lanes, the fish was heterozygous, and was scored as "positive" or "mutated" (chd c.232C>T/+ fish). In total, we identified 39 "negative" (chd +/+ ) and 12 "positive" (chd c.232C>T/+ ) fish (Fig. 4C). Sequencing of the targeted region in the 12 positive fish confirmed the c.232 C > T substitution in all 12 fish (Fig. 4D).
We next examined oep F 1 fish to identify candidate heterozygotes carrying the C > T nucleotide substitution (oep c.178C>T/+ fish) (Fig. 4E-G). For this purpose, we used sets of allele-specific primers for the wild-type allele (w) or for the mutated allele (m) (183 bp fragments in Fig. 4F,G), and another set of control primers (415 bp fragments in Fig. 4F,G), and performed allele-specific PCR. We found 147 (oep +/+ ) and 16 (oep c.178C>T/+ ) fish (Fig. 4G). Sequencing of the targeted region in the 16 (oep c.178C>T/+ ) fish revealed that 12 fish were compound heterozygotes with c.175 C > T and c.178 C > T substitutions, and 4 were heterozygotes for the c.178 C > T substitution only (Fig. 4H).

The nucleotide substitutions at target cytosines recapitulate typical mutant phenotypes.
Next, we bred a pair of chd c.232C>T/+ fish or a pair of oep c.178C>T/+ fish to produce F 2 embryos (Suppl. Fig. 2). As described above, the c.232 C > T and c.178 C > T nucleotide substitutions in the chd and oep genes, respectively, create premature stop codons (Fig. 1). Thus, the F 2 embryos with homozygous mutations should display well-known phenotypes: chd mutants have small heads and expanded blood islands 21 , and oep mutants have single eyes 22 . The breeding experiments revealed that typical chd mutant and oep mutant phenotypes were observed in 28.4% (61/215) and 21.6% (68/315) of the F 2 embryos, respectively. These results fit well with a recessive mode of Mendelian inheritance. Sequence analyses of the genomes of these embryos revealed the expected genotypes, namely F 2 embryos with small heads and expanded blood islands were chd c.232C>T/c.232C>T homozygotes ( Fig. 5B compared with 5A), and F 2 embryos with single eyes were oep c.178C>T/c.178C>T homozygotes (Fig. 5D,F, compared with 5C,E, respectively).
To further characterize oep c.178C>T/c.178C>T F 2 embryos, we performed whole mount in situ hybridizations with RNA probes for goosecoid (gsc), a molecular marker for the organizer, and hatching gland 1 (hgg1), a molecular marker for anterior dorsal mesoderm (Fig. 5G-J). The expression of gsc was suppressed at the shield stage in oep c.178C>T/c.178C>T embryos ( Fig. 5H compared with 5G). This observation was consistent with a previous study 22 . Moreover, the expression of hgg1 was absent at the tailbud stage in oep c.178C>T/c.178C>T embryos (Fig. 5J compared with 5I). These observations are also consistent with previously described phenotypes of oep mutants 22 , demonstrating that oep c.178C>T/c.178C>T embryos recapitulate mesodermal defects characteristic of oep mutant embryos. These results, as a whole, convincingly show that the Target-AID system can introduce a targeted single-nucleotide substitution from cytosine to thymine in vivo in a vertebrate.

Discussion
It has been shown that the Target-AID system 16 and another nucleotide editing system 20 deliver single-nucleotide alterations in Saccharomyces cerevisiae and mammalian cultured cells. The Target-AID system is based on the CRISPR/Cas9 system 16 which functions as a powerful tool for gene knockouts in various kinds of vertebrates, such as medaka 18 , Xenopus 23,24 , and mouse 25,26 as well as invertebrates 27,28 . Therefore, programmable single-nucleotide editing of genomic DNA in these various organisms should be amenable to the Target-AID system. Indeed, we show in this study that the Target-AID system can induce heritable nucleotide substitutions in zebrafish in vivo.
We evaluated two versions of the Target-AID system, dCas9-PmCDA1 and nCas9-PmCDA1, in zebrafish embryos. In a previous study of the Target-AID system, formation of nicks on the non-deamination strand of target DNA appeared to improve the frequency of C > T substitutions at the target sites 16 . Deep sequencing in injected embryos in the present study revealed that DNA nicking by nCas9-PmCDA1 improved the frequency of targeted C > T substitutions in the chd locus in zebrafish embryos, but not in the oep locus. Moreover, nCas9-PmCDA1 induced C > A, G > T, G > C, or G > A substitutions in addition to C > T substitutions in the chd locus, and increased indel mutations to a level that we consider problematic (Fig. 2 and Suppl. Fig. 1). In fact, indel mutations in zebrafish were observed in another nCas9-based base editing (BE) method, which induced C > T substitutions 29 . In sharp contrast, we observed efficient targeted C > T substitutions and hardly observed indel mutations with dCas9-PmCDA1, as compared to nCas9-PmCDA1. These results suggest that the Target-AID system using dCas9-PmCDA1 provides a safer single-nucleotide-editing method, as compared to nCas9-PmCDA1, for therapeutic applications in animals.
Deep sequencing of G 0 embryos and analyses of F 1 embryos revealed that the nucleotide substitutions at the target sites occurred with different frequencies. These differences may be dependent on the distances from the PAM sequences. With the chd gene, dCas9-PmCDA1 yielded the c.232 C > T mutation at the -19 position from the PAM sequence at a frequency of 2.19% in G 0 embryos ( Fig. 2A). With the oep gene, the c.175 C > T substitution (at position -19 with respect to the PAM) and the c.178 C > T substitution (at position -16 with respect to the PAM) were induced by dCas9-PmCDA1 in G 0 embryos at frequencies of 2.27% and 1.91%, respectively (Fig. 2B). In F 1 embryos, the chd c.232 C > T mutation was obtained at an overall frequency of 29.4% (10/34), and this was the sole mutation that was obtained (Fig. 3A). The c.175 C > T substitution was present in 20.4% of the oep F 1 embryos, and the c.178 C > T substitution was present in 10.2% of the oep F 1 embryos (Fig. 3B) when including compound heterozygotes with both -19 and -16 substitutions. These frequencies, in particular in F 1 embryos, were reminiscent of those observed in a Target-AID system, using nCas9(D10A)-PmCDA1 in yeast 16 ; with the yeast system, nucleotide substitutions of cytosines occurred at the -19 position at a frequency of about 25%, and substitutions at the -16 position occurred with a frequency of about 5%. Our results suggest that, even in zebrafish, the relative distance from the PAM affects the efficiency of cytidine deamination. Our results also indicate that candidate regions for targeted mutations should ideally have only one cytosine, in order to avoid redundant modifications.
Interestingly, base substitution rates in germ lines were higher than those of whole embryos in deep sequencing (Fig. 3A,B compared with Fig. 2A,B). Our observation suggests that base substitutions by dCas9-PmCDA1 might occur more effectively in germline cells than in somatic cells.
We also found A > C substitutions in the oep F 1 embryos (Fig. 3B). The previous study reported that the A > C substitutions induced by the Target-AID system are rare (less than 1%) 16 . A deaminated adenine (hypoxanthine) is capable of base pairing with cytosine in E. coli, resulting in an A > G mutation 30 . However, this mechanism cannot explain the A > C substitution that we observed in the present study. A deletion event was also observed, even though nuclease-dead Cas9 was used for targeting (Figs 2, 3B, Suppl. Figs 1, and 3B). Deletion events with the Target-AID system using dCas9-PmCDA1 were also detected in the previous study at a frequency of less than 1% 16 . This deletion might be due to the sequence context in the targeted region and/or epigenetic modifications of the target loci. Further studies are needed for a detailed understanding of the rare A > C substitutions and deletions induced by the Target-AID system.
Off-target assessments in G 0 embryos revealed that the Target-AID system did not appear to create off-target mutations, at least in the putative off-target sites we examined (Fig. 2). These results indicate that strict base-pair matching is required for the induction of C > T substitutions in zebrafish. Of course, there is a slight possibility that some of the injected embryos might die by 3 dpf because of unpredictable off-target effects. Further analysis about off-target effects of the Target-AID system is required to more precisely evaluate the Target-AID system in vivo.
G 0 fish whose caudal fins possessed the desired nucleotide substitutions did not always transmit the nucleotide substitutions to their F 1 descendants. The results are consistent with the possibility of mosaicism, where different nucleotide substitutions occurred in each daughter cell after the cell division of the zygote co-injected with the dCas9-PmCDA1 mRNA and sgRNA. The results also indicate that the genotype of cells in caudal fins is not a perfect indicator of germline transmission.
In addition to the creation of premature stop codons, as shown in this study, the Target-AID system has a potential to regulate transcription and splicing in vivo. It is difficult to make suitable precise modifications at promoter/enhancer regions or splicing sites by the regular CRISPR/Cas9 system because the CRISPR/Cas9 system induces uncontrolled DSBs. In contrast, the Target-AID system may allow researchers to modulate the promoter/ enhancer activities and splicing pattern/efficiency by the substitutions of appropriate target cytosines with thymines in the promoter/enhancer and splicing sites, respectively. In this regard, the Target-AID system may provide an opportunity for therapeutic treatment by regulating gene expression and/or splicing. Future efforts will be directed toward overcoming the limitation of the PAM recognition, etc., to achieve nucleotide substitutions at any desired position of the genome.
Single guide RNA synthesis. For synthesis of sgRNAs with customizable 18-nucleotide targeting sequences, we used the oligonucleotide pairs listed in Suppl. Table 3. Each oligonucleotide pair was annealed at 25 °C for 1 h after incubation at 95 °C for 2 min. The annealed oligonucleotides were ligated into BsaI-digested pDR274 vector 18,32 with Ligation high Ver.2 (Toyobo, Osaka, Japan). After the ligated products were linearized with DraI, sgRNAs were synthesized in vitro using MEGAshortscript T7 Transcription Kit (Thermo Fisher Scientific, Waltham, USA) according to the manufacturer's protocol.
Deep sequencing of target and off-target sites in zebrafish embryos. Genomic  was carried out as previously described 16 . The primers used for deep sequencing are listed in Suppl. Table 3. After nested PCR, index sequences were added to the amplicons. The index sequences were matched to samples as described in Suppl. Table 4. Sequencing reactions were performed with a MiniSeq sequencing system (Illumina, CA, USA) to obtain paired 151 bp read lengths. The obtained reads were mapped to each reference sequence from the zebrafish genome database (DanRer10) by the following setting: Masking mode = no masking; Mismatch cost = 2; Insertion cost = 3; Deletion cost = 3; Length fraction = 0.5; Similarity fraction = 0.8; Global alignment = No; Auto detect paired distances = Yes; Nonspecific match handling = Map randomly. The variant calling was performed with the following settings: Ignore positions with coverage = 150,000; Ignore broken pairs = Yes; Ignore Nonspecific matches = Reads; Minimum coverage = 10; Minimum count = 2; Minimum frequency = 0.5%; Base quality filter = No; Read detection filter = No; Relative read direction filter = 0.5%; Read position filter = No; Remove pyro-error variants = No. Rearrangement of the output file was made using Excel (Microsoft, WA, USA).
Heteroduplex mobility assay. Heteroduplex mobility assays (HMA) were performed essentially as described in a previous study 18 . Targeting sequences in the chd and oep loci were amplified with BIOTAQ DNA Polymerase (Bioline, London, UK) and with primer sets listed in Suppl. Table 3. The primer sets produce 117 bp DNA fragments for the chd locus, and 132 bp DNA fragments for the oep locus. The DNA fragments were separated by electrophoresis in an 8.0% acrylamide gel.
Genotyping by sequencing. PCR amplification of the targeting sequences in the chd and oep genes was performed with BIOTAQ DNA Polymerase (Bioline, London, UK) and with primer sets listed in Suppl. Table 3. PCR was carried out with the following parameters: for the chd gene, a pre-denaturation of 94 °C for 5 min, Genotyping by allele-specific PCR. Primers used in allele-specific PCR 33 are listed in Suppl. Table 3. The chd_wF and the chd_mF primers were designed to anneal to the antisense strand of intron 2 and exon 3 in the chd gene, while the chd_wR primer to the sense strand of intron 3 in the chd gene. The 3′ ends of both of the forward primers were designed to anneal to the target nucleotide, c.232 C, in the chd gene. The chd_wF primer has a C residue at the 3′ end, while the chd_mF primer has a T residue at the 3′ end. Moreover, a T residue two nucleotides from the 3′ ends of both of the forward primers generates a template/primer C/T mismatch. The oep_wF primer was designed to anneal to the antisense strand of intron 2 in the oep gene, while the oep_wR and oep_mR primers to the sense strand of exon 3 in the oep gene. The 3′ ends of both of the reverse primers were designed to anneal to the target nucleotide, c.178 C, in the oep gene. The oep_wR primer has a G residue at the 3′ end, while the oep mR primer has an A residue at the 3′ end. The primer sets of chd_bF and chd_bR, and oep_bF and oep_bR were used for internal controls to amplify DNA sequences 800-1400 bp downstream of the target nucleotides.
Alelle-specific PCR for the target nucleotides in the chd and oep genes was performed with BIOTAQ DNA Polymerase (Bioline, London, UK) with the following PCR parameters: for the chd gene, a pre-denaturation of 94 °C for 5 min, 30 cycles of amplification (94 °C for 20 s, 60 °C for 20 s and 72 °C for 20 s) and a final extension at 72 °C for 1 min; for the oep gene, a pre-denaturation of 94 °C for 5 min, 25 cycles of amplification (94 °C for 20 s, 55 °C for 20 s and 72 °C for 20 s) and a final extension at 72 °C for 1 min. The PCR products were separated by electrophoresis in a 2.0% agarose gel.
Phenotypic observation of zebrafish embryos. The phenotypes of F 2 embryos were observed at 24 hours post fertilization (hpf) for the chd mutant embryos and at 27 hpf for the oep mutant embryos. Images of the embryos were taken under a SZX16 stereo microscope (Olympus, Tokyo, Japan) equipped with a MicroPublisher 5.0 camera (QImaging, Surrey, Canada).
Whole mount in situ hybridization. Each RNA probe was synthesized to detect gsc, or hgg1 endogenous mRNA, as described previously 34 . Whole mount in situ hybridization was performed by a protocol described previously 35 . Genomic DNA from each embryo was extracted essentially as described 36 . The genotype of each embryo was determined by allele-specific PCR.
Data availability. Data obtained by deep sequencing have been deposited in the NCBI Sequence Read Archive (SRA), and the accession code is SRP140583. The remaining data are available from the corresponding author upon request.