ES-mediated chimera analysis revealed requirement of DDX6 for NANOS2 localization and function in mouse germ cells

In embryonic male germ cells, the RNA-binding protein NANOS2 recruits its target RNAs to processing bodies (P-bodies), where they are repressed. This process is necessary to promote male-type germ cell differentiation. However, it remains unclear whether all NANOS2 functions depend on P-bodies. To address this question, we established ES cell lines containing a germ cell-specific inducible Cre and reporter together with the floxed Ddx6 allele. We deleted the Ddx6 gene by administering tamoxifen to chimeric embryos containing germ cells derived from recombinant ES cells. DDX6-null germ cells exhibited both similar and distinct defects from those observed in NANOS2-null germ cells. These results demonstrate that NANOS2 function is carried out via both P-body-dependent and -independent mechanisms. RNA-seq analyses further supported the phenotypic differences between DDX6-null and NANOS2-null germ cells, and indicated distinct molecular cascades involved in NANOS2-mediated gene regulation.

To clarify this issue, we aimed to disrupt P-body formation and analyze the resulting phenotypes. Some previous reports demonstrated that P-body loss can be caused by the depletion of some P-body components 16,17,23,24 . Among these components, we focused on DDX6 (Rck/p54), which is a core component of P-body assembly. DDX6 (also known as Me31b in flies and Dhh1 in yeast) is a DEAD-box protein with ATPase/helicase activity. Although no in vivo knockout study has been reported, Ddx6-KO is expected to be embryonic lethal because of the importance of DDX6-mediated RNA regulation in development. Therefore, we planned to generate conditional KO (cKO) mice. In general, ES-derived chimeric mice are used to generate cKO mouse lines. This line has to be crossed with a germ cell-specific Cre line to establish a male mouse with Cre. Then, the male floxed-Ddx6/ Cre mouse can be crossed with a homozygous floxed female mouse. Therefore, it takes more than 6 months to obtain cKO embryos even using the quickest protocol. Furthermore, in the case of germ cell studies, we only focus on males or females depending on the purpose (male in our case). As such, only 1/8 can be expected to be male cKO embryos. Thus, we need to cross many mice to increase the efficiency. However, we can solve many problems by establishing a good ES cell line for chimera production, and ideally obtain cKO results within 2 weeks.
We demonstrate here that the chimera method is highly effective. Using this method, we generated chimeric embryos that contain DDX6-null germ cells. The results strongly support the recruitment of NANOS2 in P-bodies being a critical step to repress target RNA. On the other hand, NANOS2 may also work in a DDX6-and P-bodyindependent manner.

Establishment of useful mT/mG;Oct4dPE-creER T2 ES line for chimeric analysis of germ cell development. To establish ES lines suitable for chimera analyses in a germ cell-specific manner, we crossed
Oct4dPE-CreER T2 and Rosa-mT/mG mice (Fig. 1a). As Oct4dPE-CreER T2 mice express the tamoxifen-inducible Cre recombinase under the control of the Oct-dPE promoter-enhancer, its expression is restricted to germ cells after E9.5 25 . The mT/mG mice globally express a membrane-targeted version of tdTomato (mT/mTOMATO) and the expression changes to the membrane-targeted version of EGFP (mG/mGFP) after recombination via Cre activity 26 (Fig. 1b). Reporter incorporation is important for chimera analyses because we have to distinguish ES-derived cells (mTOMATO or mGFP) from wild-type host cells. In addition, we can mark Cre-active cells by mGFP expression. Blastocysts obtained from two female mice were separately cultured in 2i-LIF medium and we established 16 ES cell lines; 5 XY and 4 XX mT/mG, and 2 XY and 5 XX mT/mG;Oct4dPE-CreER T2 (we referred to this genotype as TGOC) (Fig. 1c).
Using one of the XY TGOC ES lines, we produced chimera to check the ability of the ES line to contribute to germ cells in chimera and whether we can induce germ cell-specific Cre activation via tamoxifen injection. We employed the ES aggregation method using 8-cell embryos to produce the chimeras (see the methods section), and orally administered tamoxifen (TM) to the foster mother at 13 days post-aggregation (dpa, which stage corresponds to E14.5) to induce the CreER T2 activity. We dissected the foster mother to recover chimeric embryos at 15 dpa (corresponding to E16.5) (Fig. 1d). We first selected chimeras by black eye color because ES cells originated from colored mice were aggregated with 8 cell embryos from white mice (MCH). We confirmed that most of the embryos with black eyes were males due to the higher contribution of XY-ES cells, as expected (Fig. S1). Most chimeric embryos had strong mTOMATO signal throughout the body, but mGFP signal was observed only in the gonads (Figs S1, 1e). The gonad sections were prepared and immunostained for GFP and the germ cell marker MVH. As a result, only germ cells detected by MVH were positive for mGFP, indicating that Cre recombination was correctly induced only in germ cells (Fig. 1f). Therefore, we decided to use this TGOC ES line for further experiments.
Maintaining ESCs in a good condition is critical for chimera analyses. Once Cre recombinase is activated, it is impossible to perform conditional KO. We found that some ESCs started to express GFP during ES culture (Fig. S2A). To find the appropriate conditions for ESC maintenance, we compared serum-free medium (2i-LIF) with serum-containing medium (serum+). Even when ESCs were cultured in 2i-LIF, approximately 5% of the ESCs exhibited GFP expression, but this population increased to 17% when they were cultured in serum + medium (Fig. S2B). Thus, we maintained the ESCs in serum-free 2i-LIF medium.

Establishment of the floxed-Ddx6 TGOC ES line using CRISPR/Cas9. To conditionally KO the
Ddx6 gene, we introduced the floxed-Ddx6 allele to the TGOC line. For this purpose, we took advantage of CRISPR/Cas9 mediated homologous recombination. We designed two Cas9 vectors for two target sites to remove Ddx6-exon5, and a targeting vector containing floxed exon-5 (Fig. 2a). Removal of exon5 leads to a frame-shift and produces a nonfunctional protein. We transfected these three constructs into TGOC ES cells and obtained ES cell clones after puromycin selection. The genotypes of ES clones were determined by primer pairs that can distinguish WT, ex-5 floxed, and ex-5 deleted alleles (Fig. 2b). Several ES lines had floxed-Ddx6 alleles introduced in both chromosomes (Fig. 2c,d).
Using the established floxed-Ddx6 TGOC ES cell lines, we produced chimeras and evaluated whether Ddx6 was correctly knocked-out in the germ cells. In this experiment, we injected tamoxifen at 11 dpa (corresponding to E12.5) before formation of NANOS2-P-body foci. In chimeric gonads, both mGFP-positive and mGFP-negative germ cells were observed, and mGFP-positive germ cells had lost DDX6 foci on immunostaining  with the anti-DDX6 antibody, whereas mGFP-negative germ cells contained clear DDX6 foci (Fig. 3a). In addition, we found that DDX6-null germ cells lost DCP1A foci, which is a marker of P-bodies (Fig. 3b). Thus, we successfully depleted P-bodies in a germ-cell-specific manner by Ddx6-cKO. Of note, NANOS2 was still expressed in the cytoplasm of DDX6-null cells but was not merged with DCP1A in discrete foci, indicating that NANOS2 foci are not generated without DDX6 (Fig. 3b). In addition to NANOS2 expression, an essential partner of NANOS2 13 , DND1, was also detected in DDX6-null germ cells (Fig. 3c). This suggests that NANOS2 can make a functional complex with DND1 even in the absence of DDX6. An important question is whether NANOS2 can function normally without forming P-bodies. We previously produced Nanos2-KO mice 27 . However, to rule out effects from the chimera analysis and tamoxifen injection, we also established Nanos2-KO TGOC ES lines using CRISPR/Cas9 (Fig. S3A). Similar to the Ddx6-cKO, Nanos2-KO germ cells were also correctly marked by mGFP after TM administration in chimeric embryos (Fig. S3B).

DDX6-dependent P-body formation is required for NANOS2 to repress target gene expression.
DDX6 is implicated in RNA regulation by functioning as a scaffold protein to assemble P-bodies 24   The major function of NANOS2 is proposed to be the degradation of target RNA in P-bodies, and one of the identified target genes is Dazl 28 . Therefore, in the absence of NANOS2, DAZL accumulates due to mRNA stabilization. To evaluate whether NANOS2 functions normally in the absence of P-bodies, we performed immunofluorescence staining using anti-DAZL antibody. Consistent with the previous study, DAZL expression was significantly increased in NANOS2-null germ cells even in chimeras (Fig. 4a). Similarly, DDX6-null germ cells  also had up-regulation of DAZL expression (Fig. 4a). This result strongly suggests that NANOS2 cannot suppress DAZL expression in the absence of P-bodies. Next, we investigated the promotion of male germ cell development by NANOS2 by examining DNMT3L. DNMT3L induces the male-type epigenetic state in cooperation with the DNA-methyltransferases DNMT3A and 3B. Thus, lack of DNMT3L leads to defective spermatogenesis 10,11 . In the chimeras, the majority of WT germ cells highly expressed DNMT3L (Fig. 4b), whereas both NANOS2-null and DDX6-null germ cells lacked DNMT3L expression (Fig. 4b). This indicates that NANOS2 cannot promote proper male germ cell development without P-bodies. To further examine male germ cell development, we confirmed the cell cycle state using pH3 staining. It is known that male germ cells enter cell cycle arrest after E14.5 and the arrested state is maintained until cell proliferation resumes just after birth. In mGFP-positive germ cells in both Nanos2-KO and Ddx6-cKO chimeric embryos, pH3 signals were clearly detected, demonstrating that cell cycle arrest cannot be maintained in DDX6-null or NANOS2-null germ cells 4 (Fig. 4c). Thus, P-bodies may be required for NANOS2 to promote male germ cell development. We concluded that NANOS2 is unable to function efficiently in the absence of P-bodies because germ cells lacking DDX6 and NANOS2-null germ cells had similar phenotypes.
NANOS2 may also function in a P-body-independent manner. In addition to resumption of the cell cycle, some NANOS2-null germ cells entered meiosis 8 . In mice, meiosis is induced by RA, and expression one of the RA responsive genes, STRA8, is used as a marker of meiosis initiation in embryonic ovaries 29 . In male gonads, STRA8 expression is repressed because of low RA levels and the repressive function of NANOS2 4 . STRA8 is expected to work as a transcription factor and localizes to the nucleus to trigger meiosis 4,30,31 . Consistent with the previous study, NANOS2-null germ cells had nuclear STRA8 signals (Fig. 5a). However, such STRA8 expression was not detected in DDX6-null germ cells (Fig. 5a), suggesting that meiosis was not initiated in the DDX6-null germ cells. We also focused on another abnormal phenotype of NANOS2-null germ cells. In male gonads, all germ cells are enclosed within an organized structure, the seminiferous tubules. However, NANOS2-null germ cells are often observed outside of these tubules, suggesting a strange escaping behavior. We performed laminin staining to mark the outer surface of the seminiferous tubules and examined whether DDX6-null germ cells also escaped. Consistent with previous observations, NANOS2-null germ cells escaped from tubules even in the chimeras (Fig. 5b), indicating that this escaping behavior is a cell autonomous defect induced by the lack of NANOS2. On the contrary, we did not observe any escaping DDX6-null germ cells (Fig. 5b). Therefore, some functions of NANOS2 are carried out independent of DDX6-mediated P-body formation.
Global transcriptome changes between NANOS2-null and DDX6-null germ cells. To more comprehensively examine the difference between DDX6-null and NANOS2-null germ cells, we compared their transcriptomes. As KO germ cells are labeled by EGFP, we used fluorescence-activated cell sorting (FACS) to collect them from chimeric embryos (Fig. S4). We also collected EGFP-positive germ cells from chimeras generated by parental ES cells as a control. We first compared Nanos2-KO and Ddx6-KO data with control data, and found that the expression of many genes changed in both genotypes (Fig. 6a). Consistent with our immunostaining data, the expression of Nanos2, Dnd1 and Cnot1 was not downregulated in Ddx6-cKO germ cells. This confirmed that in DDX6-null germ cells, NANOS2 can make a complex with its known interactors. Moreover, in accordance with previous microarray and RNA-seq data using whole gonads 28,32 , NANOS2 target genes, Dazl and Sohlh2, and other genes, Stra8, Nanos3 and Sox2, were upregulated, and male marker genes, Dnmt3l and Tdrd9, were down-regulated in Nanos2-KO germ cells (Fig. 6b). The expression of Ddx6 and known P-body components, Dcp1a and 2, was not significantly different in Nanos2-KO germ cells, but Cnot1 was significantly upregulated in germ cells from both mutants (Fig. 6b). Although most gene expression changes were also observed in Ddx6-cKO germ cells, Stra8 was not significantly upregulated (Fig. 6b), consistent with our IF data, as shown in Fig. 5a. In addition to Stra8, the RA receptor genes Rarα and Rarγ were highly up-regulated only in Nanos2-KO germ cells (Fig. 6b), suggesting that a DDX6-independent mechanism is involved in the RA-related gene regulation by NANOS2.
To search for possible genetic cascades responsible for the DDX6-or P-body-dependent or independent NANOS2 functions, we classified gene sets common to Nanos2-KO and Ddx6-cKO (common DEGs), and those only in Nanos2-KO (Nanos2-KO specific) or Ddx6-cKO (Ddx6-cKO specific), and genes that were up or down-regulated in each case were plotted (Fig. 6c-f). To further characterize these gene sets, we performed Gene Ontology (GO) analyses. The results were visualized using R package GOplot 33 , and the top 10 GO terms with representative genes are described (Figs 6g-i and S5A-C). Common GO terms comprised genes involved in muscle development and many biological processes involved in embryogenesis, many of which were downregulated (Fig. S5A). We also noted distinct GO terms for Nanos2-KO and Ddx6-cKO germ cells, respectively. Specific terms observed for Ddx6-cKO germ cells were gene expression related, such as transcription and RNA metabolism, and most included genes were up-regulated, suggesting that DDX6 suppresses global transcription and translation in germ cells (Fig. S5C). This is reasonable because P-body loss results in the de-repression of global gene expression. In addition, genes enriched in PGCs or early male germ cells, such as Prdm14 and Otx2, were listed. This suggests that DDX6-null germ cells fail to initiate the male pathway and remain in a pluripotent state compared with normal and NANOS2-null germ cells. On the other hand, in the case of Nanos2-KO, many of the categories were related to ion transport and signaling, and the included genes were often downregulated, indicating structural and functional changes in the cell membrane (Fig. S5B). These results clearly indicate that NANOS2 can function independently of DDX6 in germ cells.

Discussion
In this study, we introduced a method of chimera analysis to facilitate germ-cell-specific gene knockout studies. Although it is limited to genes functioning in a cell-autonomous manner, it can markedly shorten the experimental time. Currently, several gene editing techniques are available to quickly modify genes in ESCs, which makes chimera analyses more attractive. As we used the CRISPR/Cas9 technology for gene modification, any off-target modifications that were introduced in ES cells may have contributed to the observed phenotypes. However, we consider this unlikely because we also established a Ddx6-cKO mouse line using the same ES cells and confirmed all phenotypes observed in the chimera analyses. One problem of chimera analysis for germ cell studies is the possible disagreement between the sex of the host embryos and that of the ES cells. As we used XY-ES cells, embryos with a higher ES cell contribution became male. Therefore, this chimera method works well for the analysis of male germ cells in testes. However, we cannot use XY-ES cells for female germ cell analyses. Although we expected XX-ES cells to be replaced during the formation of ovaries, because somatic sex is determined by the Y chromosome and XY embryos cannot easily become female, we recommend XX host embryos for analyses of ovaries. Another matter of discussion is the choice of the driver for the inducible-Cre recombinase. We chose Oct4-dPE to drive CreER T2 because we have previously reported that this mouse line exhibited strong and specific Cre activity in germ cells upon TM injection 13 . Although the Oct4 enhancer is active in ES cells, we did not expect Cre proteins to translocate to the nucleus in the absence of TM. However, we noted non-specific Cre activity during ES cell culture. We speculate that this is due to weak estrogen-like activity in the serum. This problem may be avoided by selecting a germ cell-specific enhancer lacking any activity in ES cells. We are currently searching for a more suitable enhancer.
We confirmed that Ddx6-KO led to P-body loss and the simultaneous loss of NANOS2 foci in germ cells even though NANOS2 was still abundant in the cytoplasm, indicating that the formation of NANOS2 foci is dependent on the DDX6. According to several previous reports on DDX6 function, DDX6 may not only be involved in P-body formation, but also in gene regulatory functions such as the storage of mRNA, nuclear exportation and promotion of translation 34  fold changes (log2FCs) of selected genes are presented. Red and yellow bars indicate log2FCs of each gene in Nanos2-KO and Ddx6-cKO, respectively. *P < 0.05, **P < 0.01 and ***P < 0.001. (c) Adjusted P-value comparison between Nanos2-KO and Ddx6-cKO. Red dots indicate genes whose adjusted P-value (padj) was below 0.1 in both mutants (common DEGs), green dots indicate genes whose padj was below 0.1 only in Nanos2-KO (Nanos2-KO specific DEGs) and yellow dots indicate genes whose padj was below 0.1 only in Ddx6-cKO (Ddx6-cKO specific DEGs). functions. To clarify whether disrupted P-body formation caused the loss of NANOS2 function, KO analyses of other P-body components will be useful. If these mutant germ cells, DDX6-null germ cells and NANOS2-null germ cells all have a similar phenotype, it will confirm that P-body formation is the major role of DDX6 for NANOS2 function. P-body mediated RNA regulation is a critical step for regulating gene expression. As such, the P-body-mediated machinery regulates not only NANOS2 functions but also other developmental events, and differences between Nanos2-KO and Ddx6-cKO may be explained by global effects of P-body loss. We expect that the genes affected only in Nanos2-KO germ cells are involved in the phenotypes observed only in Nanos2-KO germ cells such as escaping from tubules. This phenotype may be due to the dedifferentiation of germ cells to migrating PGCs. Indeed, Nanos2-KO germ cells re-express a set of genes that are normally expressed in migrating PGCs, including Nanos3, Sox2 and Nanog. However, these genes were up-regulated in both genotypes, ruling them out as the cause of germ cell escape. We found that many genes involved in membrane anchored proteins, membrane biogenesis and ion transports, are included for Nanos2-KO specific GO terms. We speculate that these genes are important to maintain cell membrane characteristics as male gonocytes.
This study also gave rise to an important question regarding the molecular mechanisms underlying NANOS2 functions. We previously reported that NANOS2 interacts with the CNOT-deadenylation complex, leading to the degradation of mRNA. Therefore, we speculated that most NANOS2 functions can be explained by P-body-dependent RNA destabilization. However, we observed DDX6-independent NANOS2 phenotypes, such as STRA8 up-regulation and escape from tubules, indicating that NANOS2 is still active without P-bodies. One explanation for this is the existence of a P-body-independent RNA destabilization mechanism because other P-body components involved in RNA degradation were still expressed in the Ddx6-cKO germ cells (Fig. 6b). In addition, NANOS2 can still interact with CNOT1 for RNA deadenylation and with DND1 for RNA interaction in the absence of DDX6 13 . It is also possible that NANOS2 has a role other than in RNA-decay because Drosophila Nanos was recently reported to stabilize target RNA in a 3′-UTR-dependent manner. A similar mechanism by which some targets are stabilized by NANOS2 may be conserved in mouse germ cells, and it may be independent of DDX6 and P-bodies. Further studies are required to address this possibility.

Experimental Procedures
Mice. Mice were housed in a specific pathogen-free animal care facility at the National Institute of Genetics (NIG). mT/mG mice (provided by Jackson Laboratory), Oct4dPE-CreER T2 (provided by A. Suzuki, Yokohama National University) and MCH (provided by CLEA Japan, Inc., Tokyo, Japan) mice were used in this study.
Targeting strategies to produce Ddx6-cKO and Nanos2-KO ES cells. For construction of the Ddx6 targeting vector, homology arms (LA; left arm and RA; right arm) and Ddx6 exon5 were prepared by PCR from the mouse genome, and integrated into a vector containing two lox sequences via the sequential infusion method (The targeting vector sequence is available upon request). Cas9 target sites were selected using CRISPR direct (http://crispr.dbcls.jp/), and each target sequence was integrated into a modified px330 Cas9 vector (addgene) containing the pgk-puromycin cassette. These vectors were transfected into TGOC (XY) ESCs using Lipofectamine2000 ® (Invitrogen) and incubated with 1.5 µg/ml puromycin-containing medium for 2 days.
Production of chimeric mice. Eight-cell embryos collected by natural mating of MCH mice were treated with Tyrode's solution to remove the zona pellucida. The embryos were aggregated with 4~10 ES cells and cultured in KSOM (Millipore, Germany) overnight. On the next day, embryos that developed into blastocysts or morulae were transplanted into the uterus of pseudopregnant foster mothers. To induce the CreER T2 activity, 500 µl of tamoxifen (Sigma-Aldrich, T5648) (10 mg/ml) was orally administered.
Immunostaining. Embryonic gonads were fixed in 4% paraformaldehyde for 30 min at 4 °C. Gonads were next submerged in 10 and 20% sucrose in PBS for 1 h each at 4 °C, and in 30% sucrose in PBS overnight at 4 °C. The gonads were then embedded in Tissue-Tek O.C.T. compound (Sakura Finetek, Tokyo, Japan), and frozen in liquid nitrogen. Six-micrometer thick sections of each gonad were applied to glass slides, and autoclaved in citrate buffer. After pre-incubation with 3% skim milk in PBS-T (PBS with 0.1% Tween20 (Sigma-Aldrich)) at room-temperature for 30 min, the sections were reacted with primary antibodies overnight at 4 °C at the following dilutions: (1:1000, Jackson ImmunoResearch) were used. DNA was counter-stained with DAPI (100 ng/ml). Fluorescence microscopy was performed using Olympus FV1200, and images were processed with ImageJ.
Germ cell collection. To collect mutant germ cells, single-cell suspensions were prepared from 3-9 gonads from 15 dpa chimera embryos by incubation with 0.15% trypsin-EDTA at 37 °C for 5-10 min, and filtered through a 35-µm sieve (BD BioScience, San Jose, CA). The mutant germ cells were sorted by a JSAN Desktop Cell Sorter using the GFP fluorescein signal.

RNA-Seq analysis.
For RNA-Seq analysis, we prepared RNA from 2,500-20,000 sorted germ cells using the RNeasy Micro Kit (Qiagen). Twenty percent of the purified RNA was used for amplification with the TargetAmp 1-Round aRNA Amplification Kit 103 (epicenter, WI, USA). The amplified RNA was used for cDNA library construction using the KAPA Stranded mRNA-Seq Kit (KAPA, NIPPON genetics, JAPAN). Two independent cDNA libraries were prepared for each group and sequenced on the Illumina HiSeq 3000 platform to obtain 36-base single reads. Reads were aligned to the mouse genome (mm10) using Bowtie 34 , and gene expression changes were analyzed by DESeq 2 35 . During the sequence read quality check, the quality of one Ddx6-cKO sample was found to be too low. Thus, we excluded this sample from further analyses.
GO and pathway analysis. To characterize the major differences among the samples, we utilized the Gene Ontology (GO) and pathway analyses via the DAVID tool (https://david.ncifcrf.gov/) 36 . For visualization of GO results, we used the R package GOplot 33 .

Statistical analysis. Statistical significance between genotypes was assessed by the Student's t-test. Error
bars represent the s.d.
Ethical approval. The research reported involves mice. All mice were handled and propagated in accordance with National Institute of Genetics (NIG) guidelines, and all experimental procedures were approved by the Committee for Animal Care and Use in National Institute of Genetics.

Data Availability
All data generated during this study except for RNA-seq are included in this article (and its Supplementary Information Files). Sequence data have been submitted to DNA Data Bank of Japan under accession number DRA007505.