Testes of DAZL null sheep lack spermatogonia and maintain normal somatic cells

Multiplying the germline would increase the number of offspring that can be produced from selected animals, accelerating genetic improvement for livestock breeding. This could be achieved by producing multiple chimaeric animals, each carrying a mix of donor and host germ cells in their gonads. However, such chimaeric germlines would produce offspring from both donor and host genotypes, limiting the rate of genetic improvement. To resolve this problem and produce chimaeras with absolute donor germline transmission, we have disrupted the RNA-binding protein DAZL and generated germ cell-deficient host animals. Using Cas9 mediated homology-directed repair (HDR), we introduced a DAZL loss-of-function mutation in male ovine fetal fibroblasts. Following manual single-cell isolation, 4/48 (8.3%) of donor cell strains were homozygously HDR-edited. Sequence-validated strains were used as nuclear donors for somatic cell cloning to generate three lambs, which died at birth. All DAZL-null male neonatal sheep lacked germ cells. Somatic cells within their testes were morphologically intact and expressed normal levels of somatic cell-specific marker genes, indicating that the germ cell niche remained intact. This extends the DAZL-mutant phenotype beyond mice into agriculturally relevant ruminants, providing a pathway for using absolute transmitters in rapid livestock improvement.


Introduction
Control of the germline in livestock species by assisted reproductive technologies is valuable for animal breeding by driving genetic gain. Simulations have predicted a gain of several years' worth of genetic improvement by multiplying the germline of an elite sire, and therefore the number of its offspring, when compared to conventional breeding approaches 1 .
Multiplication can be achieved by inserting the germline of an elite donor into the gonads of several germ cell-deficient hosts, producing chimaeric sires with cells from distinct zygotic origins. To create such sterile hosts, local testes irradiation has successfully depleted germ cells in sheep 2 and cattle 3 . However, this method requires general anaesthesia and access to expensive specialist equipment 4 . Depending on the radiation dose and animal's age at time of radiation, it may also leave some endogenous germ cells intact 2 , compromise testicular somatic cells [5][6][7] , slow down recovery of spermatogenesis and even reduce libido 7 . The alternative strategy of sterilisation by fetal busulfan treatment resulted in systemic toxicity in pigs and germ cell ablation was again incomplete 5 . Resulting sperm output would be a mixture of donor and host cells, which is unacceptable for breeding applications that require absolute transmission of the elite donor germline. Therefore, production of a suitable host animal by physical or chemical means remains sub-optimal.
Recently, disruption of key germ cell development genes NANOS2 and NANOS3 has ablated germ cells in pigs and cattle, respectively 8,9 . NANOS3 was disrupted in female cattle, but NANOS2 was knocked out in both sexes, with only male germ cells being affected. Both knockouts reproduce the mouse phenotype and provide a suitable genetically sterilised host animal completely devoid of endogenous germ cells 10 . As an alternative target, the RNAbinding protein Deleted in Azoospermia-(DAZ)-Like (DAZL) is highly conserved among vertebrates, including bony fish, chicken, mice and humans 11 . DAZL functions by posttranscriptionally binding mRNA in 3′ untranslated regions, acting either as a transcriptional repressor or activator in a network regulating the cell cycle and the maintenance of germ cell identity 12 . In mice lacking Dazl, primordial germ cells (PGCs) fail to commit to a germ cell lineage and maintain their undifferentiated state 13 . This leads the aberrant germ cells down an apoptotic pathway, producing adult mice of both sexes that entirely lack germ cells 14,15 . In livestock, DAZL has been targeted in pigs, confirming complete ablation of the germline 16,17 .
Although introducing a loss-of-function modification by classical gene targeting is possible in livestock, such as the approach described for producing NANOS3 mutant cattle, the efficiencies are low 18 . Programmable nucleases, like TALENs or Cas9, can introduce precise double strand breaks at desired loci and greatly enhance the efficiency of gene targeting. This was demonstrated in pigs with TALENs targeting DAZL and Cas9 targeting NANOS2 8,16 . To generate animals, introducing genome editors at the zygote stage is a viable approach, but this typically leads to complex genetic mosaics with an unpredictable and difficult-to-define phenotype in the first generation 19,20 . In mosaic animals, the germline would be formed by an unknown proportion of functional wild-type cells, complicating retrospective determination of the extent of germ cell loss due to lack of DAZL. By contrast, animals produced by somatic cell cloning will be non-mosaic as they derive from a single genome-edited donor nucleus.
Therefore, somatic cell cloning enables accurate phenotyping of DAZL-null offspring in the first generation. This allows a strategy for chimaerising DAZL-null host with wild-type donor cloned embryos, generating chimaeric animals which transmit only the donor germline ("absolute transmitters").
Here we used gRNA-Cas9 mediated homology-directed repair (HDR) to produce a DAZL loss-of-function mutation in male ovine fetal fibroblasts. Following manual isolation of edited donor cell strains and somatic cell cloning, we generated DAZL-null male neonate sheep devoid of germ cells. This extends the DAZL-mutant phenotype beyond mice and pigs, providing a pathway for the use of absolute transmitters in accelerated livestock improvement.

gRNA-Cas9-mediated disruption of DAZL
To validate the genomic sequence predictions of the ovine DAZL gene available from NCBI, we compared the mRNA and protein sequences to the murine and bovine orthologues.
Aligning the sequences, the exon structure was similar, but the annotated sheep translation start site differed considerably compared to mice or cattle ( Supplementary Fig. S1). For both mice and cattle, there was an ATG translation initiation codon at the 3' end of exon 1, whereas for sheep, the entire exon 1 of the sheep gene was annotated to be translated.
Overall, there was high identity between all three species from exon 2-10, but the 5' and 3' untranslated regions were less certain in sheep.
Cas9 was used to introduce a loss-of-function mutation in exon 2, truncating the protein upstream of the predicted RNA recognition motif and DAZ functional domains (Fig. 1a). An HDR template was designed to insert a 6 bp sequence into exon 2 of the DAZL gene, comprising a premature stop codon followed by a TaqI restriction site for easy detection (Fig   1b). Three potential single guide RNAs (gRNA) were selected with minimal off-target sites using the software package CCTop 21 . After transfecting ovine fetal fibroblasts (OFFs) with the gRNA-Cas9 expression plasmid and the single-stranded oligonucleotide repair template, non-transfected cells were removed by transient puromycin selection. Quantification by droplet digital (dd) PCR showed that the average number of HDR events across three transfections was higher for gRNA1 with 6.3 ± 3.1%, compared to 0.26 ± 0.13% for gRNA2, and 0.058 ± 0.058% for gRNA3, even though this was not significant when modelled with an ANOVA (P = 0.11). A mixed cell population from gRNA1 with 12.6% HDR was subcloned by manual selection of mitotic cells to obtain a pure donor strain for subsequent somatic cloning. Of 118 mitotic cells seeded, 48 (41%) proliferated sufficiently for analysis. A PCRbased allelic discrimination assay on expanded strains showed 5/48 (10%) were positive for only the mutant allele ( Supplementary Fig. S2). All putative mutants (5/5=100%) were confirmed as biallelically edited by TaqI restriction digest ( Supplementary Fig. S3). One of these clones (#43) contained the correct 6 bp insertion on one allele, as well as an approximate 200 bp anomalous insertion around the Cas9 cut site on the other allele. The remaining four strains were validated by Sanger sequencing (Supplementary Fig. S4).
To test specificity of the gRNA-Cas9 system, a biased Sanger sequencing screen of the top three potential off-target sites, as identified by CCTop, revealed no mutations ( Supplementary Fig. S5). However, a PCR-based screen for the Cas9 coding region revealed that all four correctly edited clonal strains had integrated the gRNA-Cas9 plasmid into their genome ( Supplementary Fig. S6). Chromosome numbers of the four correctly edited OFF3 strains neither differed significantly from low-passage parental wild-type cells, nor from ovine adult fibroblasts (OAF1) that had previously produced viable cloned lambs in our hands ( Supplementary Fig. S7) 22 .

Generation of cloned DAZL-deficient lambs
Homozygous clonal knockout (KO) strains 9 and 25 were randomly chosen as donor cell strains for zona-free somatic cell cloning. Strains 9 and 25 developed into blastocysts at a rate of 17/77 (22%) and 16/90 (18%), respectively, which was comparable to non-edited parental OFFs at 7/35 (20%) ( Table 1). Cloned blastocysts on day 7 were transferred into hormonally programmed ewes by laparoscopic surgery, either fresh or after warming from vitrification, with a total of 2/13 (15%) embryos surviving to term for strain 9, and 1/13 (8%) for strain 25 ( Table 2). Both ewes carrying lambs from strain 9 developed symptoms of hydroallantois at term. They were induced to deliver at gestational day 147 but did not respond and were therefore euthanised to surgically recover the lambs at day 151. Both recovered lambs died within minutes due to respiratory failure. The lamb from strain 25 was delivered dead on day 146, displaying oedematous and decaying tissues. All lambs that failed at the perinatal stage displayed a variety of anatomical malformations typically observed in somatic cell clones 23,24 , including ankylosis of the forelimbs and hydronephrosis. Developmental rates of cloned concepti in vitro and in vivo were well within the range reported for sheep [25][26][27][28] .

DAZL mutant phenotype
Testes harvested from the two deceased lambs derived from cell strain 9 were histologically analysed. Compared to DAZL +/+ controls, DAZL -/testis cords comprised somatic support cells but lacked spermatogonia (Fig. 2a). Immunostaining on cryosections revealed a cytoplasmic signal for spermatogonial marker DDX4 in the centre of testis cords for DAZL +/+ but not for DAZL -/lambs (Fig. 2b). These observations were confirmed for different wildtype and KO animals across various locations in the testis ( Supplementary Fig. S8).
Lineage-specific gene expression was analysed to quantify germ and somatic cell populations in lamb testes by qPCR. Mutant testis parenchyma showed abolished or significantly reduced germ cell-specific expression of DAZL, NANOS2, DDX4, SALL4, LIN28, and NANOG compared to wild type (Fig. 3a). By contrast, discriminatory markers for Sertoli and peritubular myoid cells (GDNF), Leydig cells (CSF1, HSD3B1) and both Leydig and Sertoli cells (GATA4, NR5a1) were unaffected (Fig. 3b). We conclude that the introduced DAZL loss-of-function mutation led to a germ cell-deficient phenotype without compromising development of somatic support cells.

Discussion
Our data provide another proof-of-principle for using HDR to precisely modify the sheep genome 29 . This has been shown to be useful for producing human disease models in areas where sheep more physiologically relevant 29 . The HDR approach was convenient as the appearance of a known allele simplifies the screening of strains. It allows direct insertion of a premature termination codon, which prevents the formation of novel protein sequences with unknown function after NEHJ 30 . However, caution is required when selecting strains to ensure the expression plasmid has not integrated into the genome, which has been shown to happen with the backbone of HDR constructs for introducing the hornlessness (polled) trait in cattle 31,32 . Plasmid integration may have been favoured by the transient puromycin selection we applied to enhance editing efficiency. These issues can be avoided by screening lines before somatic cell cloning, or by transfecting CRSIPR-Cas9 as a ribonucleoprotein complex.
Even though low efficiency limits the large-scale production of cloned farm animal models, somatic cell-mediated genome editing provides an effective tool for generating defined genotypes. This is in contrast to embryo-mediated editing that often requires time-consuming breeding strategies to obtain non-mosaic offspring from mosaic founder animals 19 .
Developmental malformations, a known side effect of the cloning procedure, are a potential confounding variable when analysing the DAZL -/phenotype. However, the developmental defects associated with somatic clones have not been reported to include germ cells 23 . This provides confidence in the specificity of the DAZL -/phenotype to the germline in sheep.
Similarly, the death of the three DAZL -/lambs perinatally can be attributed to somatic cell cloning rather than genome editing. This poor viability at the neonatal stage has been observed in most published reports involving sheep clones, with at least half of the term animals perishing within the first day 26,28 . The main developmental problems in sheep clones have affected the pulmonary, nephrotic, and musculoskeletal systems 24 . These abnormalities have been observed in most other cloned animals, which has been collectively termed 'cloning syndrome' 23 . In total, DAZL -/lambs displayed developmental malformations beyond the germ cell loss which resemble the cloning phenotype syndrome.
We have shown that testes from DAZL -/male neonatal sheep are devoid of germ cells. Dazl is initially expressed around 11.5 days post-conception in inbreed C57BL/6 mice, where it is crucial for the correct acquisition of germ cell fate in both sexes 13 . This specification role is likely conserved in sheep, since DAZL -/neonatal male lambs lack germ cells. In female sheep, the germ cells begin to enter meiosis around 55 days post-conception 33 . If ovine DAZL provides a permissive state for germ cells to sexually differentiate, similar to the role in murine germ cell development, its expression likely starts before day 55 in both sexes.
Therefore, ovine DAZL -/germ cells will have approximately 100 days of development before birth, compared to approximately 8 days for mice. This gives a considerably longer time in sheep for germ cell quality control by apoptosis, the mechanism known to destroy aberrant male germ cell clones 34  We selected germ cell markers based on gene expression in neonate mice, which identified the target gene DAZL, along with NANOS2 and DDX4, as germ cell markers 43,44 . This was confirmed by in situ hybridisation colocalising DAZL, NANOS2 and DDX4 mRNA to germ cells in mice 10,45,46 . Dazl-deficient germ cells retain gene expression characteristic of migrating PGCs, suggesting that DAZL is involved in committing pluripotent PGCs to the germ cell fate 13,14,17 . Thus, we also surveyed expression of pluripotency-related genes LIN28, SALL4 and NANOG in DAZL -/germ cells.
All germ cell markers were reduced by 1-2 orders of magnitude or below the detection limit relative to the wild type, concurrent with the morphological disappearance of spermatogonia from the testis cords. Since we did not observe any germ cells by histological and immunofluorescent analysis, we cannot attribute the remaining low-level DAZL, NANOS2, DDX4, and SALL4 transcripts to any germ cells in DAZL -/testes. These data indicate that the somatic cells of the testis may produce a low-level background germ cell transcript signature or that there may be some rare germ cells remaining at the neonatal stage. The notion that tumorigenic germ cells may remain in DAZL -/gonads is supported by the observation that DAZL -/female pigs frequently form teratomas, which are absent from DAZL +/+ wild-type controls 17 . The rate of teratoma formation was also significantly elevated in DAZL -/-129S mice 17 , which naturally show incidences of spontaneous testicular teratomas between 1-3% 47 . Likewise, DAZL mutations in humans were associated with an increased testicular teratoma risk 48 . In contrast to pigs, humans and 129 mice, no teratomas were observed in several other Dazl -/mouse strains 17 55 . This is a considerable limitation that has so far not been resolved in livestock. The latter approach would generate chimaeras by embryo complementation, a mechanism that has already been demonstrated as feasible in sheep 56 , cattle 9,57-59 and pigs [60][61][62] . NANOS3 -/germline-deficient female cattle embryos have been used as hosts in chimaeras 9 , but further work would be required to unequivocally attribute embryo-derived germ cells to the donor germline. This approach would also need to be progressed in males, which are more relevant for livestock breeding since they generate far more offspring, either by natural mating or artificial insemination. The generation of 'absolute transmitters' from complementing genetically sterilised livestock hosts with genomically-selected elite embryonic donor cells provides an exciting opportunity for accelerating genetic gain 63 .

Materials and methods
Investigations complied with the New Zealand Animal Welfare Act 1999 and were approved by the Ruakura Animal Ethics Committee.

Donor cells
Ovine fetal fibroblasts (OFFs) were established from a male slaughterhouse fetus (crownrump length: 230 mm, estimated age 12-14 weeks of gestation) following our standard operating procedures in cattle 64 . Briefly, the skin was dissected, washed briefly in 70% ethanol, minced, and cultured in hanging drops. Genotyping using an in-house ovine highdensity SNP chip indicated that the resulting line ('OFF3') was a composite of Perendale, Texel, Coopworth and Romney breeds, in descending order of relative contribution (K. Dodds, personal communication). For somatic cell cloning, G0 cells were obtained by culture in medium containing 0.5% FCS for 5 days and harvested by trypsinization 64 .

Genome editing
The sequence of the DAZL genes obtained from NCBI databases from sheep, mouse and cattle were compared. The predicted mRNA (sheep: XM_027964130.1, mouse:  Table S1) was also delivered with PX459 using the Neon® Transfection System (Thermo Fisher, New Zealand) at 1,500 V with one 20 ms pulse. After selection with puromycin (2 µg/mL) for 48 h, genomic DNA was isolated. The QX200 (Bio-Rad) ddPCR system was used to quantify HDR using hydrolysis probes for both the wildtype and mutant variants with VIC and 6-FAM fluorophores, respectively (Supplementary Table S1).

Manual isolation of clonal cell strains
The population with the highest proportion of HDR events was used for manual isolation of clonal cell strains. We modified our mitotic shake-off procedure for synchronising cells in the G1 cell cycle stage to obtain sufficient starting material for isolating genome-edited clones 66,67 . Briefly, 2.5x10 4 cells/cm 2 were seeded on 4-well plate and cultured for 18-20 h. Cells were washed once with PBS and cultured for another 2-3 h before mitotic shake-off by gently tapping the dish with moderate horizontal force. Individual mitotic cells or couplets in ana-or telophase were selected with a finely pulled glass Pasteur pipette and seeded into single wells of a 96-well plate. Cells were expanded for 6-7 days until a confluent colony formed, passaged onto a 48 well plate, and cryopreserved once confluency was reached.

Validation of genome editing
Genomic DNA was isolated from cell lines in lysis buffer (100 mM Tris pH 8, 1 mM EDTA, 0.5% (v/v) Tween-20, 0.5% (v/v) Triton X-100) and 1 mg/ml Proteinase K (QIAGEN, Germany). The solution was incubated at 55°C for 15 min, and heat inactivated by incubating for 5 min at 95 °C. We used the Genotyping ToughMix (Quantabio, USA), 0.9 µM primers, 0.25 µM of the hydrolysis probes and 1 µl of template up to a final volume of 10 µl. For RT-qPCR, samples were run on the Rotor-Gene 6000 (Corbett Life Science, Australia) with the following program: initial denaturation at 95°C for 5 min; then 40 cycles of 95°C for 15 seconds, 60°C for 60 seconds. For the resulting Cq value, copies were calculated from a standard curve that was prepared by diluting genomic DNA of known concentration for the wild-type allele, and a synthetic gBlock (Integrated DNA Technologies, USA) template for the mutant allele.
For identified mutant clones, a larger genomic region was PCR-amplified (Table S1) and digested with 1 unit of TaqI at 65°C for 4 h. Insertion was confirmed by Sanger sequencing (Massey University Genome Service, New Zealand) of PCR-amplified fragments, using above primers (Supplementary Table S1).

Somatic cell cloning, pregnancy monitoring and parturition
Sheep somatic cell cloning was carried out based on a modified protocol for cattle zona-free embryo reconstruction 64

Testes histology
Neonate testes from edited and wild-type animals were collected and either fixed in Davidson's fixative for paraffin embedding, fixed in 4% paraformaldehyde (PFA) with 4% (w/v) sucrose for frozen sections, or the testis parenchyma was snap-frozen in liquid nitrogen for gene expression analysis. Davidson's-fixed testes were embedded in paraffin, sectioned, haematoxylin-eosin stained, analysed by a pathologist service (Veterinary Pathology Ltd., Hamilton, New Zealand) and photographed. Cryosections we made by washing PFA-fixed testes through 30% (w/v) sucrose, freezing in Optimal Cutting Temperature (OCT) compound, then sectioned at 8 µm using a Leica CM1850 cryostat (Leica Biosystems). The OCT sections were immuno-stained by permeabilising in 0.1% Triton X-100 for 10 min,  Table S2).
For reference genes, we relied on a previous study that identified GAPDH, HPRT and ACTB as stably expressed in testes tissue (PMID: 24952483). For markers of spermatogonial and somatic cell populations, we relied on a single cell qPCR screen at the neonatal stage in the mouse 44 . After designing primers for 17 genes using NCBI Primer-BLAST, spanning introns when possible, we validated their expression in wild-type testis by confirming i) amplification of a single product through gel electrophoresis, and ii) sufficient dynamic range over four orders of magnitude by running a 4-fold dilution series. Only the primer pairs that passed these quality controls were used for analysis, which resulted in DAZL, NANOS2, DDX4, SALL4, NANOG, and LIN28 as spermatogonial markers and GATA4, HSD3B1, NR5a1, CSF1, and GDNF as somatic markers (Supplementary Table S2). Assays were optimised to ensure a single melting peak corresponded to the correct PCR product size and absence of primer-dimer formation.
Relative quantification was carried out as described 72 , taking reaction efficiency into account by amplification curve analysis. The mean reaction efficiencies for each assay is listed in Supplementary Table S2. When amplicons were not detected (as indicated by a single specific melting peak) an arbitrary baseline was set for that sample with a Cq of 35, which was then normalised on reference genes.

Statistical Analysis
An Analysis of Variance Model (ANOVA) was used for analysing ddPCR quantification of genome editing efficiency 73 . For qPCR gene expression analysis, log transformed relative expression was analysed by linear mixed-effects model fit by REML 74 . A Cq value of 35 was given for missing values for computation of the baseline relative expression.     3 Cleavage normalised on nIVC. 4 Blastocyst development normalised on ≥ 2-cell embryos. 5 Proportion of grade 1-2 blastocysts normalised on total number of blastocysts.  TTTTAACTTGATGGCCGCGATTAGTGTCCACAGTAATTTTGGAGCCCAAG  AAAATAAAGTCTGTTGCTGTTTCCATTTTTCCTTGTTGGTTTGCCATTAA  GTGATGGGACCAGATGCCAGGATCTTAGTTTTTTGAATGTTGAGTTTCAA  GCTAGCTTTTTCACTCTCTTTCACCTTCATCAAGAGGCTCTTTAGTTCCT  CTTCACTTTCTTCCATTAGGGTGGTGACATCTGCATATCTGAGGCTGGTG  ATATTTCTCCCAACAATCTTTGATTCCAGCTTGTGATTCATCCATCCCAG  CATTTTGCATGATGTACTCTGTGTAGAAGTTAAATGGGCAGGGTGACAAT  ATGCATCCTTACAAGCAAACTATATGTCACAGCATCTATGATTTTAGCCT  GCAAATCCAGTGATGACATCTGCTTTCCAAATTACAGTCTGCAGCAAATC  CTGAGACTCCAAACTCAGCTGTCTCCAGAGAGGCCAGCACCCAGTCTTCT  TCAGCTGCCACCAGCCAAGGCTATGTTCTGCCAGAAGGCAAAATCATGCC  GAACTGATCGACTGTTTTCGTTGGTGGAATTGATGTTAGGGTAATGTATT  CATGTTTTCATTTATTGAAAGTATTTTGTGAATAGTGGGATCTGGGTAAT  GTTGTACACATCAGACGTTGTTCTTCTAAGTACTTCATGCAGGTTTGGAA  TTGTTTTTATGTATAACTTTTAGGGGCTAAAAATATCTTTGATACAGATA  TGTAGTTTGGGGATGCTTGGAGGAGATATTTCCACATCACCAACTCTTCT  GTATAGTATCACTGAAATCTGTCAATGTTGAATCTGTACTCTTAAGACTC  ACATTCGTCTGCATAGATGGATGAAACAGAAATTAGAAGTTTCTTTGCTA  GATATGGTTCAGTAAAAGAAGTGAAGATAATCACGGATCGAACTGGTGTG  TCCAAAGGGTGAGTACAACATATGTTCAATATCAAGTATCATCAATACCT  GAACTATACCATATATAGAATTTCAGAGAAGCAGAATTGAACACTGGTAT  TCTAACATCTGTATAAGTGAAACTTGTTGCTTTGTAACTTCTTTCATGTA  GAGGATAAGTTTTGCTGCCAGCTCTTTTATTTAATCATTTGCCTAGTGTC  TGTATGTACATATTGTGTGTCTTCATCTCACTTAAGAAGAGTAGTCTGTC  TTGATCAAAGAATTGTTTTCTTCATGCATTTCACAGTAGCGCGC