Testicular germ cell tumors arise in the absence of sex-specific differentiation

In mice and humans, defects during embryonic male (XY) germ cell development can lead to the formation of testicular germ cell tumors (TGCTs) (Bustamante-Marin et al., 2013; Oosterhuis and Looijenga, 2019). The spontaneous TGCTs that develop in the 129 family of inbred mice and related strains have proven to be an important model system for studying the developmental origins and inherited genetic risk factors of human type I pediatric teratomas and type II non-seminomas of young men (Litchfield et al., 2016; Lanza and Heaney, 2017). TGCTs in mice are first evident at embryonic day (E)15.5 as foci of proliferative and pluripotent tumor stem cells (embryonal carcinoma cells or ECCs), which ultimately differentiate to form teratomas after birth (Stevens and Hummel, 1957; Stevens and Jackson, 1962). Although TGCTs afflict only 1-10% of 129 males, depending on the substrain, genetic modifiers can be used to increase incidence, and reveal genes and developmental pathways that contribute to risk of germ cell transformation (Lanza and Heaney, 2017). In particular, 80% of 129.Chr19^MOLF/Ei (M19) males, in which both copies of chromosome 19 are derived from the MOLF/Ei strain, develop teratomas (Matin et al., 1999). Together, the 129 and M19 mouse strains provide low and high tumor-risk models that can be used to identify developmental events and genetic risk factors that promote germ cell transformation. Crucially, because most 129 and M19 germ cells develop normally, the characteristics of TGCT-susceptible germ cells that do not transform into EC cells can also be studied.

During their initial specification from the proximal epiblast, primordial germ cells (PGCs) undergo epigenomic and transcriptional reprogramming to establish characteristics of naïve pluripotent cells, including: a hypomethylated genome; expression of core pluripotency factors Nanog, Sox2 and Pou5f1 (Oct4); and expression of naïve pluripotent state-associated genes (Avilion et al., 2003; Pesce and Schöler, 2000; Yamaguchi et al., 2005). Expression of other factors found in germ cells but not in pluripotent cells, including Blimp1 (Prdm1) and Nanos3, restrict differentiation into somatic lineages or reprogramming to a pluripotent state (Ohinata et al., 2005; Tsuda et al., 2003; Saitou and Miyauchi, 2016; Durcova-Hills et al., 2008). PGCs migrate to and colonize the embryonic gonad by E11.5, at which time the expression of the germ cell licensing factor and marker of naïve pluripotency, Dazl, is upregulated in an anterior-to-posterior wave (Hu et al., 2015; Welling et al., 2015; Leitch et al., 2013; Stuart et al., 2019a). Expression of Dazl upon colonization of the embryonic gonad permits XX and XY PGCs to appropriately respond to meiotic and sex differentiation signals from supporting somatic cells (Lin and Page, 2005; Nicholls et al., 2019). From ∼E13.5 to E15.5, both XX (oogonia) and XY (pro-spermatogonia) germ cells respond to extrinsic signals from the soma to initiate expression of intrinsic factors that suppress proliferation, further restrict pluripotent capacity and induce sex-specific differentiation (Adams and McLaren, 2002). As a result of sex-specific differentiation beginning at E13.5, pro-spermatogonia enter G₀ mitotic arrest and maintain quiescence until after birth (Speed, 1982; Western et al., 2008).

Events leading to XY germ cell sex-specific differentiation are initiated when pre-Sertoli cells express and release fibroblast growth factor 9 (FGF9) at E11.5, which suppresses Wnt4 and results in the upregulation of activins and NODAL (Colvin et al., 2001; Wu et al., 2015; Jameson et al., 2012). Activins and NODAL act to establish XY germ cell identity in competent PGCs through the activation of the RNA-binding protein NANOS2 at E12.5 (Wu et al., 2015). NANOS2 acts directly to repress meiosis-associated transcripts to prevent XY germ cell entry into meiosis (Suzuki and Saga, 2008; Kato et al., 2016; Suzuki et al., 2012). Independently, NANOS2 maintains G₀ mitotic arrest, indirectly downregulates expression of pluripotency factors and promotes expression of genes required for XY-specific differentiation events (Saba et al., 2014; Suzuki and Saga, 2008; Sada et al., 2009; Western et al., 2010; Dawson et al., 2018). Expression of NANOS2 peaks around E15.5, at which time XY germ cells are considered to have completed their transition from a proliferative and pluripotent-like state to a mitotically arrested and male germ cell-committed state (Suzuki and Saga, 2008; Western et al., 2010; Yamaguchi et al., 2005).

In mice, TGCT initiation coincides with the developmental window during which XY germ cells undergo male sex-specific differentiation. We have previously shown that, from E13.5 to E15.5, a subpopulation of tumor-susceptible XY germ cells concurrently remain mitotically active, retain expression of core pluripotency factors, express genes normally restricted to pre-meiotic XX germ cells, delay expression of factors and events crucial to male germ cell sex-specific differentiation, and maintain expression of genes critical to establishing primed pluripotency (Dawson et al., 2018; Heaney et al., 2012; Lanza et al., 2016). From E13.5 to E15.5, these abnormalities become restricted to a continually smaller subpopulation of germ cells in 129 and M19 embryos. However, high-TGCT incidence M19 embryos display an increased proportion of abnormal germ cells at E15.5, the time point at which ECCs first appear, leading to the hypotheses that: (1) overall TGCT incidence (i.e. tumor risk) is proportional to the size of the population of abnormal XY germ cells capable of transforming; and (2) ECCs arise from mitotically active and pluripotent XY germ cells disrupted in male sex-specific differentiation (Dawson et al., 2018; Heaney et al., 2012).

Previous data from our laboratory and others indicate that dysregulation of Nanos2 in TGCT-susceptible mouse germ cells plays a crucial role in facilitating the events that lead to transformation (Dawson et al., 2018; Imai et al., 2020). We have previously found delayed expression of Nanos2 mRNA in the germ cell population of M19 mice at the onset of XY germ cell sex-specific differentiation; Nanos2 deficiency in 129 germ cells increases TGCT incidence; and Nanos2 deficiency in TGCT-resistant C57BL/6J germ cells promotes characteristics of TGCT-susceptible germ cells (Dawson et al., 2018). However, crucial questions remain about the developmental events that lead to germ cell transformation into ECCs and the role that Nanos2 expression plays. (1) Is the delayed expression of NANOS2 in TGCT-susceptible germ cells due to a population-wide reduction in expression or to a subpopulation of germ cells failing to express NANOS2? (2) Do ECCs in TGCT-susceptible mice arise from germ cells that fail to express NANOS2 (i.e. fail to differentiate) or express NANOS2 but fail to complete differentiation (i.e. de-differentiation), or both? (3) What molecular pathways contribute to transformation of germ cells delayed or disrupted in XY sex-specific differentiation?

In the present study, we demonstrate that failure to initiate XY sex-specific differentiation precipitates germ cell retention of proliferative and pluripotent capacity, and transformation into ECCs. We show that a subpopulation of XY germ cells in TGCT-susceptible embryos do not express NANOS2 through E15.5. Crucially, using both targeted RNA and protein analyses, and single-cell RNA sequencing (scRNA-seq), we reveal that the subpopulation of germ cells that fails to express NANOS2 is enriched with cells that have not initiated male sex-specific differentiation, remain mitotically active and maintain pluripotent capacity. Crucially, lineage-tracing studies show that mouse ECCs originate exclusively from germ cells that have not expressed NANOS2. Finally, using scRNA-seq, we identify a population of NANOS2-negative cells with an expression profile that is consistent with ECCs and enriches for genes and pathways crucial for tumorigenesis and primed pluripotency, including targets of MYC signaling, glycolysis and NODAL signaling. Together with our previous findings that Nanos2 deficiency in mice increases TGCT incidence, the results presented here demonstrate that failure to initiate the male sex-specific differentiation program is a prerequisite for TGCT initiation. We propose that, in the absence of NANOS2 expression, a subpopulation of XY germ cells inappropriately respond to extrinsic signals in the developing gonad to transition from naïve pluripotent-like PGCs to primed pluripotent ECCs.

To test whether embryonic germ cells fail to express NANOS2 in TGCT-susceptible embryos, we generated a Nanos2-P2A-tagRFP knock-in allele (Nanos2^RFP) on the 129-SvIm/J background (Fig. S1A). From this allele, expression of a Nanos2-P2A-tagRFP transcript is driven by the endogenous Nanos2 promoter and post-transcriptionally regulated by the Nanos2 5′- and 3′-untranslated regions, which together ensure proper spatiotemporal protein expression (Tsuda et al., 2006).

We first used confocal microscopy to verify XY germ cell-specific expression of RFP from the Nanos2^RFP allele in 129 embryos harboring a germ cell-specific GFP transgene (OCT4::GFP) (Youngren et al., 2005). We found that RFP expression was detectable exclusively in XY germ cells beginning at E14.5 (Fig. S2A,B), which is consistent with previous studies of NANOS2 protein expression (Tsuda et al., 2003, 2006; Youngren et al., 2005). Nanos2^RFP was subsequently backcrossed to the high TGCT-risk M19 background and bred to homozygosity with males being fertile, confirming that the modified allele does not affect NANOS2 function.

We next used fluorescence-activated cell sorting (FACS) of 129.OCT4::GFP, Nanos2^RFP and M19.OCT4::GFP, Nanos2^RFP embryonic testes to quantify the proportion of GFP-positive germ cells that express Nanos2^RFP from E13.5 to E15.5 (Fig. 1A, Fig. S3). At E13.5, we found that <1% of XY germ cells are Nanos2^RFP positive in both strains. As expected, the proportion of Nanos2^RFP-positive germ cells increased in both strains at E14.5, with 41% (±9.6) of 129 and 23% (±5.2) of M19 germ cells being Nanos2^RFP positive. Although there is almost a twofold increase in the mean percentage of Nanos2^RFP-positive cells in 129 embryos compared with M19 embryos, the observed difference is not significant due to variability in Nanos2^RFP-positive cells across samples. Importantly, at E15.5, the proportion of Nanos2^RFP-positive cells continued to increase in both strains. However, although 95% (±1.2) of 129 germ cells were Nanos2^RFP positive, a significantly smaller proportion of M19 germ cells was Nanos2^RFP positive (81±3.8%) (P≤0.01). Results of the E15.5 FACS analyses were confirmed by confocal microscopy of 129.OCT4::GFP, Nanos2^RFP and M19.OCT4::GFP, Nanos2^RFP testes. The significantly larger percentage of Nanos2^RFP-positive germ cells observed in 129 (97±0.5%) versus M19 (74±3.9%) testes (P≤0.01) was similar to that observed in the FACS analysis (Fig. 1B, Fig. S4). Therefore, we conclude that the size of the Nanos2^RFP-positive germ cell population is correlated with TGCT risk. Last, we aimed to determine whether nascent ECCs express NANOS2. Following transformation, ECCs strongly express NANOG and CDH1 (E-cadherin), either of which may be used as a marker to identify ECC foci starting at E15.5 (Heaney et al., 2012; Dawson et al., 2018). ECC foci were identified via immunolabeling for NANOG and examined for Nanos2^RFP expression. All ECC foci examined (n=8) in E15.5 M19 embryonic testes were found to be Nanos2^RFP negative (Fig. 1C).

To determine whether RFP detection corresponds to Nanos2 RNA expression, we used quantitative RT-PCR (QPCR) to compare Nanos2 levels in M19.OCT4::GFP, Nanos2^RFP-positive and Nanos2^RFP-negative germ cells. Compared with E13.5 germ cells, Nanos2 expression increased in E14.5 Nanos2^RFP-positive cells, which is consistent with previously reported studies on Nanos2 RNA expression patterns (Fig. 1D) (Dawson et al., 2018; Suzuki et al., 2006; Tsuda et al., 2006). Nanos2 expression remained unchanged from E14.5 to E15.5 in the Nanos2^RFP-positive population. A similar increase in RNA expression was observed in Nanos2^RFP-negative cells at E14.5 and E15.5; however, Nanos2 levels were half of those observed in Nanos2^RFP-positive germ cells at both timepoints (P≤0.01). Importantly, no significant difference in expression was detected between Nanos2^RFP-negative germ cells and GFP-negative somatic cells. We conclude that the presence of Nanos2 transcript but not detectable protein in Nanos2^RFP-negative germ and somatic cell populations is likely due to amplification of contaminating genomic DNA (Nanos2 is a single exon gene), as well as post-transcriptional regulation mediated by the Nanos2 3′UTR (Tsuda et al., 2006).

Having identified Nanos2^RFP-negative cells as having an expression pattern consistent with expectations of NANOS2-deficient germ cells, we asked whether these cells retain expression of Prdm1, which is normally expressed in PGCs and downregulated as sex-specific differentiation initiates (Ohinata et al., 2005; Vincent et al., 2005; Ancelin et al., 2006; Durcova-Hills et al., 2008). Importantly, downregulation of Prdm1 facilitates derivation of pluripotent embryonic germ cells (EGCs) from PGCs in vitro (Nagamatsu et al., 2015; Durcova-Hills et al., 2008, 2006). Moreover, it has been previously shown that integration into blastocysts (i.e. generation of chimeras) occurs only with EGCs, which have lost Prdm1 expression, and not in PGCs (Durcova-Hills et al., 2008; Leitch et al., 2014). Last, Prdm1 has been previously characterized as a ‘pluripotency gatekeeper’, preventing dedifferentiation via suppressing aspects of pluripotency such as Myc signaling (Nagamatsu et al., 2015). Therefore, we interpret Prdm1 as a key determinant of PGC identity and regulator of pluripotent capacity. Accordingly, we assessed Prdm1 expression via QPCR, comparing both Nanos2^RFP-positive and Nanos2^RFP-negative germ cells from M19.OCT4::GFP mice to germ cells from FVB.OCT4::GFP mice, a tumor-resistant strain (Heaney et al., 2012; Dawson et al., 2018). Interestingly, there was no significant difference in Prdm1 expression between Nanos2^RFP-positive, Nanos2^RFP-negative and FVB germ cells at any timepoints (Fig. 2A).

Concluding that Nanos2^RFP-negative cells are downregulating a key determinant of PGC identity at the appropriate timepoint, we then asked whether these cells initiate male sex-specific differentiation. DNMT3L is expressed at the onset of male sex specification, where it functions as a co-factor with DNMT3A and DNMT3B to facilitate de novo methylation of the genome (La Salle et al., 2004; Suetake et al., 2004). We previously reported that expression of Dnmt3l is significantly reduced and delayed in TGCT-susceptible germ cells (Dawson et al., 2018). Because NANOS2 promotes Dnmt3l expression and genome methylation in XY germ cells (Suzuki et al., 2016; Saba et al., 2014), we hypothesized that the Nanos2^RFP-negative germ cell population in TGCT-susceptible embryonic testes would be enriched for cells failing to express Dnmt3l. Therefore, we used QPCR to assess Dnmt3l expression in Nanos2^RFP-positive and Nanos2^RFP-negative germ cells that were FACS enriched from M19.OCT4::GFP, Nanos2^RFP embryonic testes (Fig. 2B). From E13.5 to E15.5, Dnmt3l expression increased in all germ cells, irrespective of Nanos2^RFP expression status. However, at both E14.5 and E15.5 expression of Dnmt3l was significantly lower in Nanos2^RFP-negative germ cells compared with Nanos2^RFP-positive germ cells (P≤0.0001). We confirmed our Dnmt3l QPCR analysis by immunostaining E15.5 M19.OCT4::GFP, Nanos2^RFP testes and counting Nanos2^RFP-positive and Nanos2^RFP-negative germ cells expressing detectable DNMT3L (Fig. 2C,D). Crucially, we observed a significant reduction in the proportion of germ cells with detectable DNMT3L protein in the Nanos2^RFP-negative population, while nearly all Nanos2^RFP-positive germ cells were DNMT3L positive (25.3±7.6 versus 91.7±2.5, respectively) (P≤0.0001). These results indicate that a population of germ cells not expressing Nanos2^RFP are disrupted in male sex-specific differentiation.

We have previously shown increased mRNA expression of the pre-meiotic marker Stra8 in E13.5-E15.5 TGCT susceptible germ cells and that <1% of TGCT-susceptible germ cells inappropriately initiate prophase I of meiosis (Heaney et al., 2012). Stra8 expression is repressed by the retinoic acid catabolic activity of CYP26B1 until E15.5, at which time NANOS2 acts to repress Stra8 (Suzuki and Saga, 2008). Accordingly, we tested whether germ cells in M19.OCT4::GFP, Nanos2^RFP embryos express Stra8 in the absence of Nanos2^RFP. First, we examined FACS-enriched germ cells via QPCR and found a significant increase in expression of Stra8 transcript only in the E15.5 Nanos2^RFP-negative population (Fig. S5A) (P≤0.001). Interestingly, immunolabeling of E15.5 M19.OCT4::GFP, Nanos2^RFP testes revealed that detectable STRA8 protein expression was restricted to Nanos2^RFP-negative germ cells, although expression of this pre-meiotic factor was rare (Fig. S5B). Thus, although the E15.5 Nanos2^RFP-negative germ cell population is enriched for cells that have not initiated male germ cell sex-specific differentiation, few have initiated the meiotic program by E15.5.

Cyclin D1 (CCND1), which regulates and promotes G1/S phase cell cycle transition (Sherr and Roberts, 1999; Sherr, 1995), is normally expressed in post-migratory XX germ cells prior to entering the meiotic program and is aberrantly expressed in tumor-susceptible XY germ cells from E13.5-E15.5 (Heaney et al., 2012). In tumor-susceptible mouse strains, CCND1 is downregulated in most, but not all, germ cells by E15.5, is expressed in nascent ECCs, and contributes to abnormal germ cell proliferation and TGCT risk (Lanza et al., 2016). Thus, we wanted to determine whether maintenance of Ccnd1 expression through E15.5 is associated with absence of Nanos2^RFP expression. First, we analyzed Ccnd1 RNA expression in Nanos2^RFP-positive and Nanos2^RFP-negative germ cells (Fig. 3A). Expression of Ccnd1 was similar from E13.5 to E14.5, with no significant difference found between E14.5 Nanos2^RFP-positive and Nanos2^RFP-negative cells. At E15.5, Ccnd1 expression decreased in the Nanos2^RFP-positive population but not in Nanos2^RFP-negative cells, resulting in a fivefold difference in expression (P≤0.0001). Furthermore, we performed immunostaining for CCND1 in M19.OCT4::GFP, Nanos2^RFP embryonic testes at E15.5 and found that CCND1 protein expression was significantly enriched in Nanos2^RFP-negative germ cells, with 22.7% (±6.5) of Nanos2^RFP-negative germ cells expressing CCND1 compared with 5.7% (±2.0) of Nanos2^RFP-positive germ cells expressing CCND1 (P≤0.05) (Fig. 3B,C). In summary, our results show that the E15.5 Nanos2^RFP-negative population of germ cells is enriched for cells that: (1) have downregulated Prdm1, a crucial determinant of PGC identity and pluripotent capacity; (2) have not initiated the male sex-specific differentiation program (i.e. do not express DNMT3L); (3) have not expressed pre-meiotic factors (i.e. do not express STRA8); and (4) have retained abnormal expression of cell cycle markers (i.e. express CCND1).

After finding the Nanos2^RFP-negative germ cell population is enriched for cells aberrantly expressing CCND1, we tested whether this population is also enriched for cells that have remained mitotically active, which is a defining feature of germ cells susceptible to transformation. Using immunolabeling of M19.OCT4::GFP, Nanos2^RFP embryonic testes for KI67, we discovered that Nanos2^RFP-negative germ cells are significantly enriched for those actively proliferating, with 16.5% (±3.7) of Nanos2^RFP-negative germ cells expressing KI67 compared with 1.8% (±0.6) of Nanos2^RFP-positive germ cells expressing KI67 (P≤0.005) (Fig. 3D,E).

Next, we examined whether the Nanos2^RFP-negative population expresses Nanog, a marker of pluripotency that is maintained after E14.5 in the absence of Nanos2 (Dawson et al., 2018). We first used QPCR to analyze Nanog expression in Nanos2^RFP-positive and Nanos2^RFP-negative germ cells from M19.OCT4::GFP, Nanos2^RFP embryonic testes (Fig. 4A). Expression of Nanog decreased in Nanos2^RFP-positive germ cells at E14.5 and E15.5. However, Nanos2^RFP-negative germ cells failed to repress Nanog at the same rate as Nanos2^RFP-positive germ cells, resulting in a two- and threefold higher level of expression when compared with Nanos2^RFP-positive germ cells at E14.5 and E15.5, respectively (P≤0.0001). Results of the QPCR analysis were validated at the protein level by immunostaining E15.5 M19.OCT4::GFP, Nanos2^RFP embryonic testes for NANOG, whereby 20.5±8.2% of Nanos2^RFP-negative germ cells expressed NANOG compared with 2.0±1.0% of Nanos2^RFP-positive germ cells (P≤0.05) (Fig. 4B,C). Last, we analyzed the colocalization of NANOG and KI67. No significant difference was found between the proportion of NANOG-expressing cells positive or negative for KI67 (Fig. S6A,B). However, NANOG-positive germ cells trended towards a higher proportion of KI67 positivity. Taken together, these results indicate that a subpopulation of Nanos2^RFP-negative germ cells have neither suppressed pluripotent capacity nor entered G₀ mitotic arrest by E15.5.

In response to NODAL signaling and prior to the onset of NANOS2 expression, post-migratory XY germ cells transiently express Otx2, which is required for the transition from naïve to primed pluripotency and for stabilization of the primed state (Wu et al., 2015; Weinberger et al., 2016; Acampora et al., 2013). We have previously reported nascent ECCs transition from a state of expressing both naïve and primed pluripotent markers to expressing only primed pluripotency markers by E18.5 (Dawson et al., 2018). We hypothesized that Otx2 expression is repressed, directly or indirectly, by NANOS2 to restrict the developmental potential of primed pluripotent XY germ cells at the onset of male sex-specific differentiation. Therefore, we assessed the expression of Otx2 in Nanos2^RFP-positive and Nanos2^RFP-negative germ cells from M19.OCT4::GFP, Nanos2^RFP embryonic testes (Fig. 4D). At both E14.5 and E15.5, Otx2 expression was reduced in the Nanos2^RFP-positive germ cell population when compared with E13.5 germ cells. However, E14.5 Nanos2^RFP-negative germ cells retained Otx2 expression, resulting in a 1.5-fold higher level of expression when compared with Nanos2^RFP-positive cells (P≤0.01). This trend continues at E15.5, with Nanos2^RFP-negative germ cells expressing Otx2 at threefold higher levels than Nanos2^RFP-positive germ cells (P≤0.0001). Therefore, the Nanos2^RFP-negative germ cell population is enriched for cells retaining characteristics of primed pluripotency – a prominent feature of nascent ECCs.

Our targeted RNA and protein expression analyses indicate that the Nanos2^RFP-deficient germ cell subpopulation in TGCT-susceptible mice are enriched for the primary features of germ cells susceptible to transformation into ECCs at E15.5. However, the question remained whether germ cells at risk of tumorigenic transformation had expressed NANOS2 earlier in development, subsequently lost NANOS2 expression and exited XY germ cell sex specification, or whether they never expressed NANOS2 and remained in a pluripotent-like state before directly transforming into ECCs. Therefore, we performed lineage tracing to determine whether ECCs arise from one or both of these two potential precursors.

We generated a Cre recombinase knock-in/knock-out allele at the Nanos2 locus (Nanos2^Cre) in 129 mice and backcrossed the allele to M19 (Fig. S1B). To identify germ cells in which Nanos2^Cre was active and to trace their progeny, 129 mice harboring the Rosa26 Brainbow 2.1 knock-in allele (Rosa26^Confetti) (Snippert et al., 2010) were generated and the reporter allele was backcrossed to M19. Following Nanos2^Cre-mediated recombination, one of four fluorescent reporters is constitutively expressed from the Rosa26^Confetti locus (Fig. 5A). Combining these alleles, we tested whether germ cell precursors of ECCs expressed NANOS2 at any point prior to transformation. M19.Nanos2^Cre/+, Rosa26^Confetti/+ embryonic gonads were harvested at E16.5 or postnatal day (P) 2, and ECC foci identified by strong E-cadherin (CDH1) immunostaining. Importantly, recombination efficiency of Nanos2^Cre was found to be 96.5% (±0.5) in P2 germ cells identified by weak CDH1 staining (data not shown). Assessment of 12 ECC foci from six mice revealed no ECCs expressing a Rosa26^confetti fluorophore, demonstrating that they arose from germ cells that did not express NANOS2 at any point during development (Fig. 5B).

Identifying that germ cells susceptible to transforming into ECCs reside in the Nanos2^RFP-negative population and that ECC precursors fail to express Nanos2^RFP, we assessed the transcriptomic profile of these cells to further elucidate mechanisms of transformation. To this aim, FACS-enriched Nanos2^RFP-positive and Nanos2^RFP-negative germ cells from E15.5 M19.OCT4::GFP, Nanos2^RFP embryos were analyzed by scRNA-seq (Fig. S7A,B). Single-cell transcriptomics were then performed with cells clustered into 13 groups based on principal component analysis. Next, the 13 groups were analyzed by hierarchical clustering based on expression of a priori genes associated with core pluripotency, naïve pluripotency, transition from naïve to primed pluripotency, primed pluripotency, male germ cell identity, germ cell identity, pre-meiotic germ cells, Sertoli cells, Leydig cells, endothelial cells, and the NODAL signaling pathway (Fig. 6A,B, Fig. S8). Somatic cells were classified by expression of Amh, Sox9 and Dhh (Sertoli cells, group 11); by expression of Col1a1 and Cyp11a1 (Leydig cells, group 7); and by expression of Icam2 and Cldn5 (putative endothelial cells, group 12) (Rehman et al., 2017; Li et al., 2014; Clark et al., 2000; Inoue et al., 2016; Xiao et al., 2012; Morrow et al., 2009). Pre-meiotic germ cells were classified by expression of Stra8 and Sycp3 (group 8). Interestingly, this pre-meiotic group consisted of both RFP-positive and RFP-negative germ cells. Differentiating germ cells were identifiable as groups 0, 1, 2, 3, 6 and 10, and primarily consisted of RFP-positive germ cells (Anderson et al., 2008; Botelho et al., 2001).

Undifferentiated germ cells (groups 4, 5) were noticeably distinct, separating into a unique clade and consisting primarily of RFP-negative germ cells. Compared with differentiated germ cells, undifferentiated germ cells maintain expression of core and naïve pluripotency factors, and do not express male germ cell markers. Interestingly, group 9, which was enriched for RFP-negative germ cells, clustered into a separate clade from all other groups. Group 9 is notable by expression of genes involved in core pluripotency and naïve pluripotency, while having the highest relative expression of primed pluripotency markers. Importantly, expression of Ccnd1, Nodal and Tdgf1, markers associated with ECCs and tumor susceptibility (Lanza et al., 2016; Dawson et al., 2018), was enriched in group 9 compared with all other germ cell groups. Last, group 9 lacked expression of markers of germ cell identity, male sex-specific differentiation, and pre-meiotic germ cells. Therefore, we concluded that the expression profile of group 9 is consistent with ECCs.

Next, we investigated the transcriptomic differences between undifferentiated and differentiated germ cells, and between the putative ECC-containing group and undifferentiated germ cells. To this aim, differential gene analysis was performed between: (1) the average of groups 2 and 3 (highly differentiated germ cells) against the average of groups 4 and 5 (undifferentiated germ cells); and (2) group 9 (ECCs) against groups 4 and 5 (Table S1). Gene Set Enrichment Analysis (Subramanian et al., 2005) of genes expressed lower in undifferentiated germ cells enriched for pathways associated with male germ cell sex specification and germ cell development (Fig. 7A, Table S2A,B). Genes expressed at higher levels in undifferentiated germ cells were associated with Myc signaling, including the G2M checkpoint genes, suggesting enrichment for proliferative capacity.

After establishing that undifferentiating germ cells in groups 4 and 5 are expressing genes implicated in proliferation and are deficient in genes implicated in male sex specification, we next performed a three-way comparison to examine group 9, which contains putative ECCs, against these two groups (Fig. 7B, Table S3A,B). Although undifferentiated germ cells are deficient in genes associated with male sex specification, compared with differentiated germ cells, this ECC group is further deficient in male sex specification. Furthermore, the more highly expressed genes in group 9 are mostly those from pathways involving MYC targets, glycolysis, the G2M checkpoint and E2F signaling, indicating that the ECC group is further dysregulated in male sex specification and is enriched for a pluripotent and proliferative capacity.

Next, we aimed to further clarify the highlighted genes and pathways to elucidate potential mechanisms of tumorigenesis. First, we noted that expression of genes associated with glycolysis is higher in the ECC-containing group 9 compared with the undifferentiated germ cells in groups 4 and 5 (Table S1). This is notable because: (1) primed pluripotent stem cells, and cells transitioning from naïve to primed pluripotency, use glycolysis over oxidative phosphorylation to meet their ATP demands; and (2) cancer cells ‘prefer’ glycolysis (the Warburg effect) (Warburg et al., 1927; Mathieu and Ruohola-Baker, 2017; Zhang et al., 2012; Chandrasekaran et al., 2017).

Knowing that ECCs display markers of primed pluripotency and that primed pluripotent cells rely on glycolysis, we further investigated this connection. Notably, ECCs were found to express the RNA-binding protein L1TD1 at a significantly higher level compared to undifferentiated germ cells (Fig. 6, Fig. S8, Table S1). L1td1 has been found to be necessary for the self-renewal of human primed pluripotent cells and cancer cells, and to regulate the core pluripotency factors Nanog, Pou5f1 and Sox2, likely through an interaction with Lin28a and/or Lin28b (Närvä et al., 2012). Furthermore, Lin28a and/or Lin28b have been shown to facilitate reprogramming to a primed pluripotent state, and in doing so adjusting to dependence on glycolysis (Zhang et al., 2016; Li et al., 2012). Last, Lin28a and/or Lin28b are regulated by MYC, which is crucial in reprogramming into a primed pluripotent state (Chang et al., 2009; Folmes et al., 2013; Marson et al., 2008; Logan and Nusse, 2004). Therefore, expression of L1td1 corroborates that ECCs are upregulating Myc signaling in a primed pluripotent state.

We have previously published that ECCs express nuclear-localized pSMAD2, indicative of NODAL signaling, which is known to drive expression of the core pluripotency factors Nanog, Pou5f1 and Sox2, hallmarks of ECCs in the E15.5 testis (Dawson et al., 2018; Pauklin and Vallier, 2015). Furthermore, NODAL signaling has been shown to drive the primed pluripotent state in vitro (Bertero et al., 2015; Vallier et al., 2009, 2005). Some markers of NODAL signaling, such as Acvr1b and Acvr1c, were not expressed at a significantly higher level in the group 9 ECCs compared with undifferentiated germ cells; however, NODAL co-factor CRIPTO (Tdgf1) was expressed at higher levels in this ECC-containing group (Fig. 6, Fig. S8). Together, these data indicate that there may be multiple pathways converging, potentially synergistically, to facilitate the transformation of germ cell precursors into ECCs.

Initiation of testicular germ cell tumors is first evident in mice at E15.5, at which time foci of pluripotent and proliferating ECCs are identifiable. Lineage commitment of XY germ cells is driven by the synchronized downregulation of genes essential for maintaining PGC identity and restricting pluripotent capacity during migration and concomitant upregulation of genes crucial for inducing male sex-specific differentiation. This coordinated change in the transcriptional profile of XY germ cells ensures appropriate cell fate determination in response to exogenous signals in the developing testis. Based on our results presented here, we propose that TGCTs in mice arise when the expression of genes crucial for restricting the pluripotent capacity of PGCs are downregulated but genes crucial for sex-specific lineage commitment are not activated (Fig. 8). This creates a developmental window during which controls on germ cell fate determination are absent and pluripotent capacity is unrestricted. We posit that germ cell transformation to ECCs mirrors the in vitro progression of mouse embryonic stem cells and epiblast stem cells along the naïve to primed pluripotency continuum, as well as the in vivo transition from the pre-implantation to post-implantation epiblast.

From specification to migration into the developing gonad, PGCs exist in a naïve pluripotent-like state (Saitou and Yamaji, 2010). However, genes crucial for establishing and maintaining PGC identity, such as Prdm1, restrict their pluripotent potential until the initiation of sex-specific differentiation (Surani et al., 2007; Durcova-Hills and Surani, 2008). Intriguingly, repression of Prdm1 (directly or indirectly by FGF2) is crucial for in vitro derivation of pluripotent embryonic germ cells (EGCs) from migratory-stage PGCs in vitro (Durcova-Hills et al., 2008). Incorporation of EGCs into chimeric mice likely requires the loss of Prdm1 during derivation from a PGC. Isolated PGCs are unable to integrate into the blastocyst while EGCs can (Durcova-Hills et al., 2006; Nagamatsu et al., 2015; Leitch et al., 2014). Furthermore, deletion of Prdm1 in PGCs permits the in vitro derivation of pluripotent ‘EGC-like cells’ in the absence of FGF2 and membrane-bound stem cell factor (mSCF); these EGC-like cells can form teratomas when injected into nude mice (Nagamatsu et al., 2015). Last, it has been shown that Prdm1 represses Myc, which plays a crucial role in establishing pluripotent capacity (Durcova-Hills and Surani, 2008; Nagamatsu et al., 2015). Here, we show that the endogenous timing of Prdm1 downregulation is maintained in TGCT-susceptible germ cells when compared with TGCT-resistant germ cells. Additionally, using a Nanos2^RFP reporter, we show that a subpopulation of TGCT-susceptible germ cells fails to express Nanos2^RFP and initiate sex-specific differentiation by E15.5. Importantly, reduced Prdm1 expression is observed in both Nanos2^RFP-positive and Nanos2^RFP-negative germ cells, indicating that downregulation of PGC identity is not necessarily dependent on the initiation of sex specification.

Using our Nanos2^RFP reporter, we found that Nanos2^RFP-deficiency significantly enriches for germ cells that have not initiated critical sex-specific differentiation events. Using a Cre-based lineage-tracing system, we demonstrate that the germ cell precursors that transform into ECCs do not initiate expression of Nanos2. Based on these observations, we propose that failure to express Nanos2 and disrupted sex-specific differentiation are functionally required for germ cell transformation. Supporting this conclusion, we previously demonstrated that Nanos2 deficiency increases TGCT incidence in 129 mice and that Nanos2 deficiency in TGCT resistant mice causes embryonic germ cells to retain proliferative and pluripotent capacities (Dawson et al., 2018). Presently, our study shows that germ cells that fail to initiate sex specification are an at-risk population for tumorigenic transformation, and that failure to initiate sex specification is a requisite precursor event to tumorigenesis.

Paracrine and autocrine signals of the developing testis, including the TGFβ superfamily ligands activin and NODAL, are essential for establishing male germ cell identity (Wu et al., 2013). Curiously, these same factors are crucial for establishing and maintaining primed pluripotency in vitro (Weinberger et al., 2016). It has been previously shown that XY germ cells transiently express OTX2, which is required for naïve to primed pluripotency transition and for stabilizing the primed state, prior to the onset of Nanos2 expression and sex-specific differentiation (Wu et al., 2015). Based on this observation, we previously hypothesized that XY germ cells transiently enter primed pluripotency in response to activins and NODAL signaling, and that activation of Nanos2 is essential for directing differentiation out of this primed state (Dawson et al., 2018). In TGCT-susceptible germ cells that do not express NANOS2, we proposed that a stable primed pluripotent state is established (Fig. 8). Here, our scRNA-seq data support a crucial role for NODAL in ECC formation and maintenance of primed pluripotency. The ECC-containing group had the highest relative expression of Nodal and the co-receptor Tdgf1 (Cripto), indicating amplification of this signaling pathway.

Although germ cells that do not express NANOS2 by E15.5 give rise to ECCs, it is unlikely that the entire NANOS2-negative germ cell population undergoes transformation. In fact, when analyzed by immunofluorescence, only ∼20% of E15.5 Nanos2^RFP-negative germ cells demonstrate any of the developmental anomalies associated with tumorigenesis (i.e. 22.7% of Nanos2^RFP-negative germ cells express CCND1, 16.5% express KI67 and 20.5% express NANOG). Moreover, only 8.9% (±4.6) of NANOG-expressing cells express KI67, potentially indicating a more tumor-susceptible subpopulation. It has been previously shown that Nanos2 deficiency in TGCT-resistant strains leads to initiation of meiosis, meiotic catastrophe and apoptosis (Suzuki and Saga, 2008; Bourc'his and Bestor, 2004). Importantly, we show that some Nanos2^RFP-negative M19 germ cells express the pre-meiotic marker Stra8 at E15.5 when NANOS2 becomes responsible for suppressing Stra8 and meiotic differentiation (Suzuki and Saga, 2008). We hypothesize that pre-meiotic XY germ cells are likely destined for cell death rather than transformation. We previously demonstrated that a small population of embryonic XY germ cells in TGCT-susceptible embryos stall in prophase I of meiosis (Heaney et al., 2012). Nicholls et al. recently showed that TGCTs in mice do not arise from meiotic cells (Nicholls et al., 2019). Thus, in the absence of NANOS2, we propose that localized concentrations of NODAL and retinoic acid may drive tumorigenic or meiotic fate determination.

In humans, TGCTs arise in utero from abnormal male germ cell development (Moch et al., 2016). Human TGCTs are classified into three different subgroups: (I) teratomas, yolk sac tumors and embryonical carcinomas; (II) seminomas and non-seminomas (teratomas, yolk sac tumors and choriocarcinomas); and (III) spermatocytic seminomas of the elderly (Oosterhuis and Looijenga, 2005). Both type I and type II tumors originate during embryogenesis; however, type I tumors present during infancy and type II present during adolescence to early adulthood (Oosterhuis and Looijenga, 2005). Furthermore, type II tumors originate from germ cell neoplasia in situ (GCNIS) precursors, while type I tumors appear to develop from germ cells without a GCNIS intermediate (Pierce et al., 2018). Recently, Nettersheim et al. proposed that type I and type II TGCTs in humans both originate from post-migratory PGCs: type I precursors such as ECCs arise from germ cells that downregulate PGC-associated genes, maintain pluripotency and fail to initiate the sex-specification transcriptional program, whereas type II GCNIS precursors arise from germ cells that maintain PGC-associated genes, pluripotent capacity and do not express sex-specification genes (Skakkebæk et al., 1987; Nettersheim et al., 2016). For type II tumors, retention of PGC-associated genes likely stabilizes the pluripotent-like state of GCNIS until puberty, much as they do during PGC migration. Our data provide in vivo experimental evidence to support this model and indicate that the TGCTs of 129 mice accurately models type I tumor development in humans. Crucially, the proposed similarities in the developmental origins of type I and II tumors explain the significant overlap in heritable risk factors.

Genome-wide association studies (GWAS), targeted re-analyses and epidemiological studies have found that approximately one-third to one-half of TGCT risk is attributable to genetic factors, with many overlapping risk loci associated with type I and type II tumors (Litchfield et al., 2015b). GWAS studies have implicated many genes associated with embryonic germ cell development in human TGCT risk, indicating that dysfunction of XY germ cell development is crucial to the initiation of these tumors in utero. Importantly, NANOS2 has not been associated with TGCT risk in humans. Presently, it is unclear what role NANOS2 plays in human TGCT initiation. Potentially, NANOS2 may remain to be identified via GWAS; hundreds if not thousands of additional variants yet to be identified (Litchfield et al., 2015a,b, 2016, 2018). Alternatively, within the Caucasian population included in GWAS, there may not be variation at the NANOS2 locus that directly influences NANOS2 expression. Last, due to its importance in reproduction, variation at this locus may have been selected against.

Irrespective of whether genetic variation at the NANOS2 locus contributes to genetic susceptibility, we propose that many of the germ cell developmental genes that associate with TGCT risk in humans, including DAZL, likely influence tumor susceptibility through NANOS2 expression and disruption of sex-specific differentiation. Dazl-deficient embryonic XX and XY germ cells do not express factors crucial for sex-specific differentiation (e.g. NANOS2 in XY germ cells; REC8 in XX germ cells) and Dazl deficiency on the 129 inbred background increases TGCT incidence (Nicholls et al., 2019; Koubova et al., 2014; Gill et al., 2011). Therefore, we propose that tumor initiation occurs due to a failure in sex-specific differentiation in the embryonic testis, regardless of the genetic or environmental insult (i.e. loss of Dazl, Nanos2 and Dnd1, etc.). Crucially, some TGCT risk genes (e.g. KITLG and PRDM14) have definitive roles in the development of germ cells prior to populating the gonad. It is possible that perturbations of embryonic germ cell development at various stages can disrupt post-migratory germ cell differentiation and subsequently cause transformation in the gonad (Fig. 8).

Last, our scRNA-seq data implicate Myc in the initiation of tumorigenesis and maintenance of ECCs. Crucially, Myc expression has been found to be inhibited by Prdm1, suggesting that the normally timed reduction in Prdm1 expression observed in tumor-susceptible germ cells may permit Myc activation (Magnúsdóttir et al., 2013; Saitou and Yamaji, 2012). Importantly, Myc expression has previously been identified in human TGCTs and ECCs (Shuin et al., 1994; Ben-Porath et al., 2008; Schaub et al., 2018; Skotheim and Lothe, 2003; Skotheim et al., 2002; Juric et al., 2005). Furthermore, our scRNA-seq data indicate that ECCs have increased expression of L1td1 and Lin28. Myc signaling has been shown to be intertwined with Lin28 to promote proliferation and pluripotent capacity in tumor stem cells, including human germ cell malignancies (Li et al., 2012; Närvä et al., 2012; Zhang et al., 2016; Chang et al., 2009; West et al., 2009). L1td1 has been identified as a co-factor for Lin28 and is crucial to human stem cell self-renewal and proliferation of tumorigenic cells (Närvä et al., 2012). Furthermore, our scRNA-seq data implicate Lef1 in tumorigenesis. Myc has been identified as a downstream target of LEF1 when bound to pSMAD2 (Labbe et al., 2000); we have previously found that ECCs strongly express pSMAD2, suggesting a potential promotion of Myc expression (Dawson et al., 2018). Future studies aimed at further elucidating the axis of L1td1, Lin28, Lef-1 and Myc signaling, including defining which pathway(s) are inducing Myc, may lead to the identification of therapies with fewer long-term side effects.

All protocols were approved by the BCM Institutional Animal Care and Use Committee. 129S1/SvImJ (129, JR#002448) and FVB/NJ (FVB, JR#001800) were obtained from the Jackson Laboratory. Chromosome substitution strain 129.Chr19^MOLF/Ei (M19) mice were obtained from our research colony (Matin et al., 1999). 129.Gt(ROSA)26Sor^{tm1(CAG-Brainbow2.1)Cle}/J (Confetti) ES cells were a gift from Dr Hans Clevers (Hubrecht Institute for Developmental Biology and Stem Cell Research, Utrecht, The Netherlands) and chimeric founder animals were generated from these cells. Founders were subsequently backcrossed and maintained on the 129S1/SvImJ background. A germ-cell specific Oct4ΔPE::GFP transgene (Youngren et al., 2005) was previously established on the 129 inbred background and was backcrossed on the M19 background (Heaney et al., 2009). A separate Oct4ΔPE::GFP transgene was previously established on the FVB inbred background (Anderson et al., 1999). Both transgenes were produced with the GOF-1/ΔPE/EGFP construct (Yoshimizu et al., 1999).

The Nanos2^RFP [Nanos2^{em1(tagRFP)Jahe}] and Nanos2^Cre [Nanos2^em2(cre)Jahe] alleles were generated on the 129S1/SvImJ inbred background using CRISPR/Cas9 genome editing in zygotes (Fig. S1A,B). For Nanos2^RFP, guide RNA 5′-GCGATAATTGAGACCCTGTG(agg) (chr7:18988009-18988031, GRCm38/mm10) was injected into pronuclear stage embryos with Cas9 mRNA and a long single-stranded DNA oligo (Integrated DNA Technologies) containing a P2A-tagRFP and 100-200 bp homology arms as previously described (Lanza et al., 2018). For Nanos2^Cre, guide RNA pair [5′-CTGAGTTTCGCCCACTGCGT(cgg) (chr7:18987970-18987992; GRCm38/mm10) and 5′-GCGATAATTGAGACCCTGTG(agg) (chr7:18988009-18988031, GRCm38/mm10)] was co-injected into pronuclear stage embryos with Cas9 mRNA and a long single-stranded DNA oligo (Integrated DNA Technologies) containing Cre recombinase and 100-200 bp homology arms. Founder mice and N1 offspring were screened by PCR and Sanger sequencing for the allele of interest and alleles were maintained with backcrosses to 129S1/SvImJ inbred mice.

All alleles were transferred to the 129.Chr19MOLF/Ei (M19) background, as previously described (Heaney et al., 2009). PCR genotyping for Nanos2^RFP was performed using primers 5′-AGGGTGAGGTAGCTGAGGAG and 5′-GCCCTCACAATCCAAAAGAA. PCR genotyping for Nanos2^Cre was performed using primers 5′-TGATGGACATGTTCAGGGATC and 5′-CAGCCACCAGCTTGCATGA. PCR genotyping for Confetti was performed using the primers 5′-CTCCTGGCTTCTGAGGACC (wild-type forward), 5′-CCAGATGACTACCTATCCTC (wild-type reverse) and 5′-GACCACTACCAGCAGAACAC (knock-in forward). Genotyping primers are listed in Table S4A.

For immunofluorescence and fluorescence-activated cell sorting experiments using Nanos2^RFP, animals homozygous for Nanos2^RFP and Oct4ΔPE::GFP were intercrossed and E0.5 was assumed to be noon of the day a vaginal plug was observed. Pregnant females were euthanized and gonads were removed from embryos in ice-cold 1×PBS. Gonad morphology was used to determine the sex of E13.5 and older embryos. For Confetti and Nanos2^Cre lineage-tracing experiments, M19 animals homozygous for Confetti were bred to animals heterozygous for Nanos2^Cre to produce M19.Nanos2^Cre/+, Rosa26^Confetti/+ and M19.Nanos2^+/+, Rosa26^Confetti/+ mice. Testes were collected at either E18.5 or postnatal day 2 and tissue harvested for Cre genotyping as described above.

Fluorescence-activated cell sorting for germ cells expressing Nanos2^RFP and Oct4ΔPE::GFP from embryonic gonads, RNA preparation, reverse transcription, and quantitative real-time PCR analysis and normalization of gene expression were performed as previously described (see Table S4B for primers) (Lanza et al., 2016; Heaney et al., 2009). Female littermates were used as negative controls and gates set to <1.0% of female germ cells as Nanos2^RFP positive. These gates were then used to determine Nanos2^RFP-positive germ cells in males. One-way ANOVA with the Bonferoni post-tests for multiple comparisons were used to determine significant differences in RNA expression between E15.5 Nanos2^RFP-positive germ cells, E15.5 Nanos2^RFP-negative germ cells and somatic cells. Unpaired t-tests were used to determine significant differences in RNA expression between E14.5 Nanos2^RFP-positive germ cells and E14.5 Nanos2^RFP-negative germ cells, as well as E15.5 Nanos2^RFP-positive and E15.5 Nanos2^RFP-negative germ cells.

Gonads were collected from embryos or post-natal mice and processed for sectioning and immunofluorescence as previously described (Heaney et al., 2009). Sections were incubated with primary antibodies in donkey blocking solution overnight at 4°C. For secondary detection by confocal microscopy, sections were incubated with a secondary antibody in donkey blocking solution for 2 h at room temperature. Nuclei were counterstained with DAPI in hardset mounting medium (Vector Laboratories H1500). Sections for Confetti×Nanos2^Cre lineage tracing were mounted in hardset mounting medium without DAPI (Vector Laboratories H1400). CFP signal from the Confetti allele was amplified for imaging using an anti-CFP primary antibody. Fluorescence of Confetti fluorophores, RFP, OCT4::GFP transgene and secondary antibodies were imaged with a Nikon A1-Rs inverted Laser Scanning Microscope or Zeiss LSM 780 inverted confocal microscope. GFP and YFP signals from the Confetti allele were combined into one GFP channel due to imaging channel constraints. No-primary controls of all secondary antibodies were performed with no false-positive signals found (data not shown). All antibodies are listed in Table S5A,B.

Oct4ΔPE::GFP-positive E15.5 M19 germ cells positive or negative for Nanos2^RFP, and positive or negative for NANOG, CCND1, KI67 or DNMT3L were counted as previously described (Heaney et al., 2009). Cells in ECC foci were not included in counts. Unpaired t-tests were used to detect significant differences in the percentage of NANOG, CCND1, KI67 or DNMT3L between Nanos2^RFP-positive and Nanos2^RFP-negative germ cells. To determine the efficacy of Nanos2^Cre-mediated recombination, germ cells from P2 M19-Nanos2^Cre/+;Rosa26^Confetti/+ were identified by weak CDH1 staining and counted for expression of a Confetti reporter (data not shown).

Testes of four E15.5 M19.OCT4::GFP;Nanos2^RFP mice were dissociated and pooled for FACS sorting to enrich for E15.5 Nanos2^RFP-positive and Nanos2^RFP-negative germ cells, as described above, for single-cell sequencing. 17,589 and 3141 single-cell transcriptomes were analyzed from Nanos2^RFP-positive and Nanos2^RFP-negative populations, respectively. Cells were combined and pelleted at 300 g at 4°C, sorting buffer decanted and resuspended in 0.04% bovine serum albumin (Amresco, 0332-100G) in PBS (free of Ca²⁺, Mg²⁺ and EDTA; Hyclone SH30256.01) at ∼1000 cells/µl, in a volume no more than 46.6 µl. A single-cell 3′ Gene Expression Library was prepared according to chromium single cell 3′ reagent kit v3 (10× Genomics). In brief, a single-cell suspension was loaded on a Chromium controller (10× Genomics) to generate single-cell GEMS (gel beads-in-emulsions) where full-length cDNA was synthesized and barcoded. Subsequently, the GEMS are broken and cDNA from each single cell is pooled. Following clean-up using Dynabeads MyOne Silane Beads (Thermo Fisher, 370020), cDNA is amplified by PCR. The amplified product is fragmented to optimal size before end-repair, A-tailing and adaptor ligation. The final library was generated by amplification.

The Genomic and RNA Profiling Core first conducted Sample Quality checks using the NanoDrop spectrophotometer and Agilent Bioanalyzer 2100 (High Sensitivity DNA Chip, p/n 5067-4626). To quantitate the adapter ligated library and confirm successful P5 and P7 adapter incorporations, we used the Applied Biosystems ViiA7 Real-Time PCR System and a KAPA Illumina/Universal Library Quantification Kit (p/n KK4824). GARP then sequenced the libraries on the NovaSeq 6000 System using the S1 flowcell.

A QPCR assay was performed on the libraries to determine the concentration of adapter ligated fragments using the Applied Biosystems ViiA 7 Quantitative PCR instrument and a KAPA Library Quant Kit (p/n KK4824). All samples were pooled equimolarly and re-quantitated by qPCR, and also re-assessed on the Bioanalyzer.

Using the concentration from the ViiA7 qPCR machine above, 250 pM of equimolarly pooled library was loaded onto a NovaSeq S2 flowcell (Illumina p/n 20012862), using the Standard Workflow loading conditions specified by the manufacturer and amplified by exclusion amplification (ExAMP) for patterned flowcells using the Illumina NovaSeq 6000 sequencing instrument. PhiX Control v3 adapter-ligated library (Illumina p/n FC-110-3001) was spiked-in at 3% by weight to ensure balanced diversity and to monitor clustering and sequencing performance. A paired end run, using 28 cycles for Read 1, eight cycles for Index 1 (i7) Read and 91 cycles for Read 2 was set to sequence the flowcell on a NovaSeq 6000 Sequencing System and achieved an average of ∼380 million read pairs per sample. FastQ file generation and QC assessment was achieved using the 10X Cell Ranger software for 10X Chromium Platforms.

Single-cell data were aligned to the murine reference genome (mm10) using CellRanger from 10X Genomics and processed using Seurat v.3.1.1 (Butler et al., 2018; Stuart et al., 2019b). Nanos2^RFP-positive cells with >5% mitochondrial expression counts or expressing less than 1000 or more than 5000 unique genes were removed from further analysis. Nanos2^RFP-negative cells with >5% mitochondrial expression counts or expressing less than 2500 or more than 7500 unique genes were removed from further analysis. For each of the samples, Seurat employed a global-scaling normalization method to normalize the gene expression for each cell by multiplying by 10,000 and log transforming the resulting expression levels. Data were then scaled prior to dimensional reduction using the top 2000 most variable features. Cells from each sample were then clustered using the default Louvain algorithm with a resolution parameter equal to 0.5. To visualize clusters, t-distributed Stochastic Neighbor Embedding (t-SNE) analysis was performed on the resulting principal components estimated by the clustering analysis. Marker genes were used to identify the groups containing cell types of interest used for further analysis. Differential gene analysis was performed using the Seurat ‘FindAllMarkers’ function. Clusters of interest were compared and genes with an absolute natural log fold change≥0.25 and Bonferroni adjusted P value<0.05 were reported. Pathways analysis was performed using Gene Set Enrichment Analysis (Subramanian et al., 2005). Each gene set was analyzed using the ‘Compute Overlap’ feature and pathways with an FDR<0.05 were reported.

We thank the Genetically Engineered Rodent Models, Single Cell Genomics, Genomic and RNA Profiling, Cytometry and Cell Sorting, and Integrated Microscopy and Optical Imaging and Vital Microscopy Cores at Baylor College of Medicine for assistance with mouse production, single-cell sequencing, FACS and imaging. Resources accessed through cores were supported by a National Institutes of Health grant (P30CA125123 to the Dan L. Duncan Comprehensive Cancer Center). The Single Cell Genomics Core, and Genomic and RNA Profiling Cores were supported by the National Institutes of Health (S10OD018033, S10OD023469, S10OD025240 and EY002520 to Rui Chen). The Cytometry and Cell Sorting Core was supported by a Cancer Prevention and Research Institute of Texas grant (RP180672) and by a National Institutes of Health grant (RR024574). Optical Imaging and Vital Microscopy Cores were supported by National Institutes of Health grants (DK56338 and ES030285) and by Cancer Prevention and Research Institute of Texas grants (RP150578 and RP170719). We thank Dr Isao Suetake (Institute for Protein Research, Osaka University, Japan) for providing the DNMT3L antibody and Dr Hans Clevers (Hubrecht Institute for Developmental Biology and Stem Cell Research) for providing 129 ES cells harboring the Confetti allele.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/148/9/dev197111