Scalable and Efficient Generation of Mouse Primordial Germ Cell-like Cells

Abstract

Primordial germ cells (PGCs) are the founder cells of the germline. The ability to generate PGC-like cells (PGCLCs) from pluripotent stem cells has advanced our knowledge of gametogenesis and holds promise for developing infertility treatments. However, generating an ample supply of PGCLCs for demanding applications such as high-throughput genetic screens has been a limitation. Here, we demonstrated that simultaneous overexpressing 4 transcriptional factors - Nanog and three PGC master regulators Prdm1, Prdm14 and Tfap2c - in suspended mouse epiblast like cells (EpiLCs) and formative embryonic stem cells (ESCs) results in efficient and cost-effective production of PGCLCs. The overexpression of Nanog enhances the PGC regulatory network and suppresses differentiation of somatic lineages, enabling a significant improvement in the efficiency of PGCLC production. Transcriptomic analysis reveals that differentiated PGCLCs exhibit similarities to in vivo PGCs and are more advanced compared to cytokine-induced PGCLCs. These differentiated PGCLCs could be sustained over prolonged periods of culture and could differentiate into spermatogonia-like cells in vitro. Importantly, the ability to produce PGCLCs at scale, without using costly cytokines, enables biochemical and functional genomic screens to dissect mechanisms of germ cell development and infertility.

Introduction

In mammals, gametogenesis is a complex and long process that is initiated by the specification of primordial germ cells (PGCs), the precursors of oocytes and sperm. Unraveling the mechanisms of PGC development is crucial for understanding key processes such as sexually dimorphic epigenetic inheritance and the bases of infertility, a relatively common health condition impacting ~15% of all couples ¹. In mammals, PGCs specified in the embryonic epiblast respond to extrinsic and intrinsic signaling pathways including bone morphogenetic protein (BMP) and WNT signaling. In mice, BMP4 activates WNT3, and subsequently a master mesodermal transcription factor (TF), brachyury (T). T activates critical early germline markers Prdm1 (also known as Blimp1), Prdm14, and Tfap2c (also known as AP2γ) ². Prdm1 is expressed in the precursors of PGCs, which then stimulates Prdm14 and Tfap2c transcription to activate the germline pathway. The core TF network containing Prdm1, Prdm14, and Tfap2c is critical for mouse PGC specification; it represses somatic differentiation while driving reacquisition of a pluripotency network and epigenetic reprogramming ². After migration to and colonization of the genital ridges, male germ cells differentiate into gonocytes. Thereafter, they become spermatogonial stem cells that subsequently seed recurrent rounds of mitotic expansion yielding spermatogonia that enter meiosis, and eventually, haploid spermatozoa.

Characterization of these gene regulatory pathways in PGCs formed the basis for creating mouse primordial germ cell-like cells (PGCLCs) from pluripotent stem cells in vitro ². Pluripotency has at least three distinct states that mirror early embryonic development: naïve or ground state ESCs resembling the preimplantation epiblast; formative ESCs, resembling early post-implantation gastrulating epiblast; and primed epiblast stem cells (EpiSCs), resembling late post-implantation gastrulating epiblast ^3-5. Naïve ESCs typically are cultured with inhibitors of fibroblast growth factor and mitogen-activated protein kinase (FGF/MEK) as well as glycogen synthase kinase 3β (GSK3β) in the presence of leukemia inhibitory factor (LIF) (common known as 2i+LIF medium). EpiSCs self-renew when exposed to Activin A and fibroblast growth factor 2 (FGF2). Naïve ESCs do not readily respond to germ cell inductive stimuli, while primed EpiSCs display limited germline competence. By modulating activin and WNT signaling, a formative state has been stably captured in vitro ^4-7. As an intermediate pluripotency state, formative ESCs possess dual competency for mouse chimera formation (when introduced into host blastocysts) and direct responsiveness to PGC specification.

The ability to model germ cell development in culture holds great potential for applications such as dissecting the genomic regulation of gametogenesis, and identifying genetic causes of infertility that would be important for assisted reproductive technologies (ART) involving gene corrections and optimization of in vitro gametogenesis (IVG). However, these applications remain challenging due to limited scalability of PGCLC generation. Conventional methods involve a transient conversion from naïve ESCs to germline-competent epiblast-like cells (EpiLCs), which presumably represent a formative state. Aggregated EpiLC clusters are then exposed to high concentrations of BMPs and supporting cytokines, leading to their differentiation into PGCLCs ⁸. Alternatively, PGCLC specification can be induced by overexpressing three TFs (Prdm1, Prdm14, and Tfac2c), and to a lesser extent, two TFs (Prdm1 + Tfap2c or Prdm14 + Tfap2c) or even a single TF (Prdm14), in EpiLC aggregates ⁹. Nanog, a core pluripotency factor expressed in PGCs, is not essential for emergence of PGCs or germline transmission, but its deletion results in a significant reduction in PGC numbers both in vitro and in vivo ¹⁰. Nanog is also expressed in naïve ESCs, and its expression undergoes a transient downregulation as cells transition into the EpiLC state. The overexpression of Nanog can promote efficient PGCLC differentiation by binding endogenous enhancers of Prdm1 and Prdm14, even without BMP4 and associated cytokines ¹¹. Furthermore, Nanog overexpression induces cytokine-free PGCLC specification by repressing Otx2, encoding a TF essential for anterior neuroectoderm specification ^12,13.

Promoting cell proliferation is another strategy for producing a large number of PGCs/ PGCLCs. However, long-term culture of mouse PGCs in vitro is prone to dedifferentiation to pluripotent embryonic germ cells (EGCs)^14,15. Saitou and colleagues found that Rolipram and Forskolin, which stimulate cAMP signaling, enables expansion of PGCLCs up to ~50-fold in the presence of SCF (Stem cell factor) on m220 feeders (stromal cells expressing a membrane-bound form of human SCF), and the expanded PGCLCs can contribute to spermatogenesis after transplantation to neonatal testes ^16-18. Nevertheless, PGCLCs cultured under these conditions cannot be expanded for more than 7 days. Although these methods have been instrumental in understanding and improving PGCLC formation, they are associated with either low efficiency or high costs (for reagents such as cytokines).

Here, we report a simple and economical approach to induce the differentiation of PGCLCs from both formative ESCs and EpiLCs at scale by overexpressing four TFs (Prdm1, Prdm14, Tfap2c and Nanog). The resulting PGCLCs exhibit transcriptional and epigenetic profiles similar to those of late (E11.5) PGCs, and these could be further differentiated into spermatogonia-like cells. This system holds promise for facilitating high-throughput functional genomic studies of germ cell development, characterization of potential infertility-causing genetic variants, and improving IVG.

Results

Establishing a cytokine-free platform for efficient production of PGCLCs from EpiLCs.

To establish an efficient, cytokine-free system for generating PGCLCs, we began with a mouse ESC line containing enhanced GFP (eGFP) under the regulatory control of Stella (also called Dppa3/Pgc7), a PGC marker ¹⁹, and validated the ability of this line to form germline-transmitting chimeric mice after blastocyst microinjection (Fig. S1). The Stella-eGFP reporter enables real-time visualization of PGCLC development from ESCs in vitro. Next, based on the finding that Nanog promotes transition of ESCs to PGCLCs by binding the enhancer regions of Prdm1 and Prdm14 and inducing their expression ¹¹, we introduced a stably-integrated doxycycline (Dox)-inducible Nanog. To test the effectiveness of this transgene, we utilized a high efficiency enhancer activity detection system ²⁰ to derive ESC lines containing a targeted integrations of lacZ reporter under the control of Prdm1 or Prdm14 enhancers into the safe harbor H11 locus. Dox treatment of embryoid bodies (EBs) derived from these lines triggered β-galactosidase production, indicating the effectiveness of NANOG-stimulated transcription (Fig. S2A-D). Besides promoting expression of Prdm1 and Prdm14, NANOG represses Otx2, a TF that promotes differentiation of pluripotent cells towards a neuroectodermal fate ^11,13. Mutant cells lacking OTX2 enter the germline from an EpiLC state with high efficiency, even in the absence of the essential PGCLC-promoting cytokine signals ²¹. In summary, Nanog plays a crucial role in PGCLC differentiation by promoting the expression of PGC core regulatory TF-encoding genes Prdm1 and Prdm14, while simultaneously inhibiting the somatic fate inducer Otx2 (Fig. S2E).

Based on observations that overexpression of Prdm1, Prdm14 and Tfap2c can induce PGCLCs without cytokines ⁹, we hypothesized that simultaneous overexpression of four TFs (i.e. Prdm1, Prdm14, Tfap2c and Nanog, hereafter called “4TF”) in EpiLCs might robustly render PGCLCs in vitro. We isolated several Dox-inducible clonal cell lines harboring combinations of these TFs in Stella-eGFP ESCs that also contained rtTAM2 (reverse tetracycline-controlled transactivator mutant; Fig. 1A and B). The resulting ESCs, cultured in 2i+LIF medium, were differentiated to EpiLCs by two-day culture in the presence of Activin A and bFGF. To induce PGCLC differentiation, the EpiLCs aggregates were treated with Dox instead of cytokines. We first tested whether ectopic expression of one or more TFs drove EpiLC-to-PGCLC differentiation in U-bottom 96-well plates. Consistent with previous observations ¹¹, induction of Nanog alone led to ~1/3 of cells becoming eGFP⁺ (Fig. 1A). The efficiency rose to ~80% by simultaneous overexpression of all 4TFs. Cell lines harboring Nanog plus other TF combinations also showed substantial increases in the fraction of eGFP⁺ cells compared to Nanog alone (Figs. 1A, D and E).

Figure 1. Construction of 4TF-induced PGCLC differentiation platform.

A. Percentages of day 6 PGCLCs (Stella-eGFP⁺ cells) differentiated from Dox-inducible ESCs harboring distinct combinations of TFs. Each individual clonal line is represented by a dot. B. The process for constructing the 4TF-inducible ESCs. The bolded fonts highlight the 4TFs and rtTA. C. Schema illustrating the 4TF-induced PGCLC differentiation from ESCs harboring the Stella-eGFP reporter. D. Induction of PGCLCs (Stella-eGFP⁺ cells) in the presence or absence of Dox in EBs during the 6-day period. E. FACS patterns of eGFP⁺ cells (indicated by dashed lines) during differentiation. F. Relative fluorescence density of H3K9me2 and K3K27me3 in GFP⁻, GFP dim and GFP bright cells from Dox-induced day 6 EBs. G. Relative mRNA expression of Dnmt3a and Dnmt3b in ESCs, EpiLCs and day 6 EBs. H. Bisulphite sequence analysis of methylated cytosine in the DMRs of the imprinted genes in GFP⁺ and GFP⁻ cells in day 6 EBs. White and black circles represent unmethylated and methylated cytosines, respectively. Data in G are represented as the mean ± SEM. Data in F and G were analyzed using one-way ANOVA with Tukey’s post hoc test. Scale bar in D represents 100 μm.

In vivo, PGCs undergo substantial epigenetic reprogramming, characterized by the loss of H3K9me2 and the acquisition of H3K27me3, as well as global erasure of DNA methylation ². To test the fidelity of epigenetic changes in PGCLCs produced in our 4TF system, we examined histone modifications, CpG methylation status of imprinted genes, and capacity for dedifferentiation into pluripotent EGCs. Differentiated cells with higher eGFP fluorescence displayed lower H3K9me2 and higher H3K27me3 labeling (Fig. 1F). The mRNA levels of epigenetic modifiers Dnmt3a and Dnmt3b were increased in EpiLCs and subsequent downregulated in EBs (Fig. 1G), which aligns with previous findings ⁸. Bisulfite sequencing of PGCLCs (eGFP⁺ cells) revealed reduced methylation in differentially methylated regions for both paternally (H19) and maternally (Snrpn) imprinted loci, while eGFP⁻ cells (presumably somatic-like) exhibited a more methylated state (Fig. 1H). In addition, PGCLCs could de-differentiated to EGCs in vitro when transferred to 2i+LIF medium (Fig. S3). Collectively, our data indicate that simultaneously overexpression of 4TFs effectively induces the differentiation of EpiLCs into PGCLCs.

Transcriptomic properties of 4TF-induced PGCLCs

To examine the fidelity of our 4TF-induced PGCLCs, RNA sequencing (RNA-seq) was performed (Fig. S4) and the transcriptomes were compared to those of in vivo germ cells (encompassing PGCs at E9.5, E10.5, and E11.5 and male/female germ cells at E12.5 and E13.5) and cytokine (Ck)-induced PGCLCs (those at day 2,4 and 6 after induction, referred to as Ck_D2/4/6PGCLC) ^16,22,23. Principal component analysis (PCA) showed that 4TF PGCLCs at days 2, 4 and 6 (4TF_D2/4/6PGCLC) clustered tightly as a group separate from naïve ESCs, EpiLCs and Ck_D2PGCLCs (Fig. 2A). Notably, 4TF-induced PGCLCs grouped closely with PGCs at E10.5 and E11.5, while Ck_D4/6PGCLCs grouped closely with E9.5 and E10.5 PGCs (Fig. 2A). These data suggest that 4TF-induced PGCLCs are more akin to late PGCs than Ck_D2/4/6PGCLCs. Pearson correlation coefficient analysis revealed that 4TF_D2/4/6PGCLCs are highly correlated with Ck_D4/6PGCLCs but not Ck_D2PGCLCs (Fig. 2B), consistent with the PCA. We next applied two-sided rank–rank hypergeometric overlap (RRHO) analysis to compare the gene expression signatures between samples in a threshold-free manner ²⁴. We observed robust overlap in both up- and down-regulated transcripts amongst differentiating cells in both Ck- and 4TF-induced groups relative to either ESCs or EpiLCs (Fig. 2C). In the Ck-induced group, ~60% of genes exhibited concordant differential expression (53% when compared to ESCs and 67% when compared to EpiLCs) between D2 and D6 PGCLCs. Interestingly, the concordance was higher for the D2 and D6 4TF-induced group (74% and 78%, respectively). In both induction systems, D4 and D6 PGCLC shared around 84% concordantly regulated genes (Fig. 2D and Fig. S5). These data demonstrated that 4TF-induced PGCLCs at early stages exhibited greater consistency in their transcriptomic changes compared those in Ck-induced PGCLCs.

Figure 2. 4TF-induced PGCLCs share similar transcriptomes with cytokine- and 3TF-induced PGCLCs.

A. PCA of indicated transcriptomes from 4TF- and cytokine (Ck)-induced PGCLCs, in vivo germ cells at indicated times of gestation, EpiLCs and ESCs. “D” = days of induction. PGCLCs were isolated by flow sorted for diagnostic markers. B. Heatmap of the Pearson correlation coefficients between PGCLCs. C. Comparison of differential expression profiles between D6PGCLC (x-axis) and D2/4PGCLC (y-axis) with RRHO2 maps. The extent of overlap transcriptome was represented by each heatmap colored based on −log10(p-value) from a hypergeometric test. Each map shows the extent of overlapped upregulated genes, versus ESCs or EpiLCs, in the bottom left corner, whereas shared downregulated genes are displayed in the top right corners. D. Percentage of concordant genes encompassing both up and down regulated genes compared with D6PGCLCs. E. GSVA score of Wnt signaling pathways in D2PGCLC between Ck-, 3TF- and 4TF-induced conditions (Ck_D2, 3TF_D2 and 4TF_D2). ESC, EpiLC and PGCLC in each group were used to calculate GSVA score and only the scores of D2PGCLC were displayed. F. Heatmap of selected genes associated with WNT signaling and targets.

In vivo, PGC specification follows WNT-dependent induction, which activates the master mesodermal TF, T. This triggers the core TF network for PGCs (i.e., Prdm1, Prdm14, and Tfap2c). To investigate the activity of WNT signaling in various PGCLC differentiation conditions, gene expression datasets from 3TF-induced PGCLCs at days 2 and 4 (referred to as 3TF_D2/4PGCLC) were retrieved ⁹ and compared to Ck- and 4TF-induced PGCLCs. Gene set variation analysis (GSVA) revealed strong enrichment of Wnt signaling pathways in Ck_D2PGCLCs (including downstream target genes Wnt3, Wnt5a, Wnt5b, Axin2, Mixl1, Cdx2, and T) as previously reported ⁹, but not in 3TF_D2PGCLCs or in our 4TF_D2PGCLCs (Fig. 2E and F).

To better characterize the transcriptome landscapes during PGCLC differentiation, we applied k-Means analysis of top 500 differentially-expressed genes and defined one cluster (C4) representing genes expressed in 4TF_D2/4/6PGCLCs (Fig. 3A). Gene Ontology (GO) analysis revealed that 11 of the top 20 pathways in this cluster were associated with reproductive processes (Fig. 3B). GSVA confirmed enrichment in germ cell development terms such as gamete generation, oogenesis, piRNA metabolic process, meiosis and reproduction, were enriched in PGCLCs compared to ESCs and EpiLCs (Fig. 3C). D2PGCLCs were predominantly enriched for Biological Processes including “S-adenosylmethionine metabolic process”, “leukocyte migration” and “inflammatory response” (Fig. S6). In contrast, the upregulated genes at days 4 and 6 were primarily enriched in processes related to gamete generation, meiotic cell cycle and reproduction (Fig. 3D and S7). We next compared the differentially expressed genes among Ck-, 3TF- and 4TF-induced PGCLCs. Across all three groups, the naïve pluripotency gene Klf4 was highly expressed in ESCs, while genes related to the formative state (Otx2, Pou3f1, Fgf5, Dnmt3a and Dnmt3b) were highly expressed in EpiLCs (Fig. 3E). Early PGC-related genes (e.g., Prdm1, Prdm14, Nanog, Dnd1, Tfap2c, Dppa3 and Itgb3) were notably expressed in Ck_D4PGCLC, 3TF_D2PGCLC and 4TF_D2PGCLC, indicating that both 3TF and 4TF induction systems cause molecular differentiation more rapidly than cytokine induction. Some PGC genes, including Rhox9, Rhox6, Rhox5 and Gm9, were highly expressed in Ck_D6PGCLC, 3TF_D4PGCLC and 4TF_D4PGCLC. Interestingly, late PGC, meiosis and piRNA-related genes were consistently expressed in 4TF-induced PGCLCs but not in Ck-induced PGCLCs (Fig. 3E and S7), explaining the enrichment of meiotic and piRNA metabolic pathways in 4TF-induced PGCLCs. Since there is no transcriptome data for 3TF_D6PGCLC, it uncertain whether they exhibit a similar phenomenon as 4TF_D6PGCLC (Fig. 3E). To comprehend the expression of meiotic and piRNA genes in 4TF-induced PGCLCs, we conducted a network analysis on the differentially upregulated genes and identified a sub-network constructed by germ cell specific genes (Fig. S8). During germ cell development, Tfap2c was found to be activated by Prdm1 and Prdm14 ², direct targeting Dmrt1 in PGCLCs ^25,26. Human DMRT1 upregulates late PGC markers (such as DAZL and PIWIL2), while DAZL enhances DMRT1’s impact by repressing pluripotency and promoting other late PGC genes (such as DDX4 and SYCP3) ²⁷. The regulatory axis involving Prdm1, Prdm14, Tfap2c, Dmrt1, Dazl, Ddx4 and Sycp3 genes facilitated the transition from early to late PGCLCs (Fig. S8).

Figure 3. Gene expression profiling of 4TF-induced PGCLCs.

A. Heatmap of k-means clustering of variably expressed genes in ESCs, EpiLCs, and D2-D6 4TF-induced PGCLCs (n = 500, k = 4). Genes were grouped into four clusters (“C1-4”) on the basis of expression similarity. B. Biological process GO term analysis of Cluster 4 (C4). Terms of germ cell development (highlighted by red) was highly enriched for terms in C4. C. Heatmaps showing the average GSVA enrichment score of selected germ cell development pathways. D. Representative GSEA enrichment plots depicting significant enrichment of germ cell signatures in D6PGCLCs compared with ESCs. E. Heatmap of representative gene expression of pluripotency and germ cells among ESCs, EpiLCs and PGCLCs. F. Further development of 4TF_D6PGCLC by aggregated with testicular somatic cells isolated from Kit^w/wv mice. The cultured aggregates were subjected to IF and quantification. Data in lower panel of F are represented as the mean ± SEM, and the single dot represents the percentage of double positive cells in a cluster of DDX4⁺ cells. Scale bar in F represents 50 μm.

In the 3TF induction system, ~30% of PGCLCs (Blimp1-mVenus⁺/Stella-eCFP⁺ cells) were differentiated, and this efficiency declined after day 4 ⁹. Conversely, in our 4TF induction system, we consistently observed a high level of differentiation efficiency beyond day 4. This variation prompted us to investigate the distinctive role of Nanog. GSEA revealed that pathways predominantly related to neurodevelopment were suppressed in D2PGCLCs compared to EpiLCs (Figs. S9A and B), and Otx2 was involved in the majority of these pathways (Fig. S9C). Otx2 plays a crucial role in brain development and exhibits reciprocal antagonism with Nanog. Induction of transgenic Nanog in EpiLCs shifts the balance between these mutual antagonists, favoring Nanog ²⁸. When Otx2 levels decrease sufficiently, this enables NANOG to influence the regulatory elements controlling genes associated with PGC-specific TFs ²⁸. These observations, together with the evidence that Nanog enhances the expression of Prdm1 and Prdm14, illustrates its dual role in both strengthening the PGC program and preventing the development of non-germline lineages.

We next examined the potential of PGCLCs to differentiate into spermatogonia-like cells in vitro. D6PGCLCs were aggregated with dissociated neonatal testicular cells from germ cell deficient Kit^W/Wv mice and co-cultured for 2 days in U bottom 96-well plates. Then, the cell aggregates were transferred to a trans-well plate and cultured in medium containing GNDF and bFGF that are essential for promoting spermatogonial growth ²⁹. After a two-week induction period, we conducted IF staining to assess whether the aggregates express late PGC or spermatogonia proteins. We detected cell clusters that expressed NANOS3 and DDX4 within the aggregates (Figs. 3F and S10). Over 70% of DDX4-positive cells also express GCNA (Germ cell nuclear antigen). Interestingly, certain DDX4-positive cells also contained PLZF (formally ZBTB16; Zinc finger and BTB domain-containing protein 16), a marker indicative of spermatogonia (Fig. 3F)³⁰. DDX4-positive cells did not express the Sertoli cell marker, SOX9 (Fig. 3F). These data suggest that the 4TF-induced PGCLCs have the capacity of further development in vitro.

Scale-up differentiation and long-term culture of PGCLCs

To validate the Stella-eGFP transgene as an accurate marker of PGCLCs, we analyzed differentiating cultures with two additional PGC surface markers, ITGB3 (also known as CD61) and SSEA1 ⁸. Doubly positive cells were shown to differentiate into functional gametes ³¹. As expected, the cell populations characterized by ITGB3⁺/SSEA1⁺ and ITGB3⁺/Stella-eGFP⁺ were significantly enriched in the day 6 EBs compared to ESCs and EpiLCs (Fig. 4A). Next, we sought to determine the scalability this 4TF-induction system. We seeded 2×10³ cells/ well to U bottom 96-well plates, 8×10⁴ cells/well to untreated 12-well plates and 1×10⁶ cells to a single 100mm bacteriological Petri dish, then assessed differentiation in day 6 EBs. The percentages of both ITGB3⁺/SSEA1⁺ and ITGB3⁺/Stella-eGFP⁺ populations were highest in 12-well plate and lowest in 96-well plates (Fig. 4A and B). Suspecting that EB size influences differentiation efficiency, we measured the diameters of day 6 EBs. In the U bottom 96-well plates, EB sizes were homogeneous and by far the largest (~2.5× on average) amongst the culture formats (Fig. 4C). The results indicate an inverse relationship between efficiency of PGCLC differentiation and EB size.

Figure 4. Scale-up and long-term differentiation of PGCLCs.

A. Representative FACS pattern of Itgb3⁺/Ssea1⁺ and Itgb3⁺/Stella-eGFP⁺ cells in ESCs, EpiLCs and day 6 EBs from U bottom 96-well plates, 12-well non-treated plates and 100mm petri dishes. B. Quantification of PGCLC populations in day 6 EBs at different plates/dish. C. Boxplots showing the distribution of the diameter of day 6 EBs in different plates/dish. D. FACS pattern of Itgb3⁺/Ssea1⁺ and Itgb3⁺/Stella-eGFP⁺ cells in day 20 and 43 EBs cultured in 12-well non-treated plates. Data in B and D are represented as the mean ± SEM. Data in B and C were analyzed using one-way ANOVA with Tukey’s post hoc test. Scale bar in C represents 200 μm.

Next, we investigated the potential for maintaining PGCLCs over long periods. Day 6 EBs from a 12-well plate were dispersed and seeded into one well of an untreated 12-well plate in the presence of Dox. Cells were passaged every 4^th day, and after culturing EBs for 20 and 43 days, >70% of cells were ITGB3⁺/SSEA1⁺ and ITGB3⁺/Stella-eGFP⁺ (Fig. 4D).

Direct differentiation of PGCLCs from formative ESCs

We next asked whether 4TFs can efficiently induce PGCLCs from formative ESCs, a pluripotent stem cell that directly responds to PGC specification. The 4TF Stella-eGFP ESCs were transformed to a formative state through exposure to bFGF, Activin A and GSK3b inhibitor (CHIR99021), as described previously ⁵. Naïve ESCs typically grow as tightly packed colonies with a dome shape. Upon EpiLC induction, the cells quickly underwent a morphological conversion that includes flattening, diminished cell-cell interactions, and the formation of cellular protrusions. However, formative ESCs displayed distinctive morphologies (Fig. 5A). Immunolabeling of ZO1 (zonula occludens-1, also known as tight junction protein 1) revealed that formative ESCs and EpiLCs were more similar to one another than naïve ESCs (Fig. 5B). The regulation of Oct4 expression in naïve and primed pluripotent cells is differentially controlled by distal (DE) and proximal (PE) enhancers, respectively ³². The DE enhancer also exhibits activity in the formative state, albeit at a lower level compared to that in the naïve state ⁵. To monitor the exit from the naïve state during naïve-to-formative conversion, we compared the GFP signal intensity of Oct4-DE-GFP (“OG2”) iPSCs during this transition. Consistent with previous observations ⁵, the majority of formative iPSCs and EpiLCs were GFP⁺ but the average signal intensity was lower than that in naïve iPSCs (Fig. 5C). Collectively, these results confirm the establishment of formative ESCs/iPSCs from naïve cells.

Figure 5. Direct differentiation of PGCLCs from formative ESCs.

A. Representative morphologies of naïve ESCs, formative ESCs and EpiLCs. B. Immunofluorescence of tight junction protein ZO1 on naïve ESCs, formative ESCs and EpiLCs. C. GFP expression in naïve and formative OG2 iPSCs, and EpiLC differentiated from naïve OG2 iPSCs. D. Representative FACS pattern of Itgb3⁺/Ssea1⁺ and Itgb3⁺/Stella-eGFP⁺ cells in formative ESCs, EBs with Dox treatment. Scale bars in A and C represent 100 μm and in B represents 50 μm.

We next tested if formative 4TF ESCs respond directly to PGC specification. Following the removal of the feeder layer, Dox was added to promote PGCLC differentiation. FACS analysis of EB cells revealed distinct ITGB3⁺/SSEA1⁺ and ITGB3⁺/Stella-eGFP⁺ populations; such cells were not observed in undifferentiated formative ESCs (Fig. 5D). This differentiation system proved to be scalable (Fig. 5D), although the efficiencies were lower compared to those observed in EpiLC-to-PGCLC differentiation system described earlier (Fig. 4A and B). Collectively, these findings demonstrated that overexpressing 4TFs can directly and effectively induce differentiation of formative ESCs into PGCLCs.

To elucidate the differences in differentiation efficiencies between formative ESC and EpiLC, we compared the transcriptomes of our EpiLCs to a published formative ESC dataset ⁵. The RRHO analysis revealed significant overlap in co-regulated genes between formative ESC and EpiLC when compared to naïve ESC, indicating a shared cell state (Fig. S11A and B). Interestingly, genes associated with the formative state (Otx2, Pou3f1, Fgf5, Dnmt3a and Dnmt3b) exhibited higher expression in EpiLCs than those in formative ESCs (Fig. S11C). Thus, formative ESCs and EpiLCs represent distinct sub-stages, leading to variations in differentiation efficiencies.

Discussion

PGC development entails important regulatory and epigenetic events that lay the groundwork for subsequent sex-specific gametogenesis programs. Although in vitro PGCLC induction systems have facilitated molecular studies of the specification and developmental events of the germ cell lineage ^{9,11,21,33-37}, current techniques have shortcomings in terms of efficiency, scalability and long-term culture. Our 4TF-induction system addresses these limitations. We demonstrated that simultaneous overexpression of Nanog, Prdm1, Prdm14 and Tfap2c in suspended mouse EpiLCs and formative ESCs can render PGCLCs in an efficiently and cost-effectively manner. The induced PGCLCs exhibit late PGC hallmarks, and they can be further developed into spermatogonium-like cells in vitro. Overexpression of Nanog had the important effect of enhancing the PGC regulatory network while suppressing the formation of non-germline lineages, significantly improving PGCLC production.

Sequential TFs govern PGC differentiation, and specification of mouse PGCs requires coordinated actions of Prdm1, Prdm14, and Tfap2c. Here, we found that forced expression of Nanog alone, or in addition to combinations of these core TFs can drive PGCLC specification. The 4TF-induced PGCLC exhibits similar epigenetic reprogramming with those of Ck- and 3TF-induced PGCLCs. Interestingly, the 4TF-induction system exhibits superior performance compared to the 3TF (i.e., Prdm1, Prdm14 and Tfap2c) induction system, underscoring the pivotal role of Nanog. Given that NANOG can bind and activate enhancers of Prdm1 and Prdm14 in EpiLCs in vitro ¹¹, which we confirmed here, we conclude that Nanog overexpression strengthens the PGC core regulatory network in the 4TF system. NANOG also plays a role in suppressing somatic lineage induction development. OTX2, for example, shows reciprocal antagonism with NANOG, restricting mouse germline entry ²¹. In 4TF-induced PGCLCs, we observed downregulation of Otx2 and simultaneous upregulation of Nanog, mirroring the pattern seen in Ck- and 3TF-induced PGCLCs. OTX2 functions as a crucial TF for anterior neuroectoderm specification. Comparing 4TF_D2PGCLCs to EpiLCs, we observed significant downregulation of neurodevelopment pathways associated with Otx2. Therefore, the dual role of NANOG contributes to high efficiency PGCLC generation.

The transcriptional profiles of Ck- and 4TF-induced PGCLCs proved to be different and resembled PGCs at distinct stages. This disparity partially arises from the essential role of WNT signaling and mesodermal program in the Ck-induction system, whereas the 4TF-induction system circumvents these requirements. Additionally, the gene regulatory axis of early-to-late PGC development established in 4TF-induced PGCLCs propels their differentiation beyond what Ck achieved within the same timeframe. Consequently, in the 4TF but not Ck-induced PGCLCs, the late PGC regulatory network was triggered as indicated by expression of genes involved in piRNA processes (such as Piwil2, Mael, Tdrd5, Mov10l1, Tdrd7 and Tex19.2) and meiosis (such as Stra8, Sycp1/2/3, Meiob and Smc1b).

Gonadal and testicular somatic cells are essential to support the further differentiation of male PGCLCs in vitro ^17,38,39. Seminiferous tubule-like structures can be constructed by aggregating Ck-induced PGCLCs with gonadal somatic cells under a gas-liquid interphase culture condition ^17,39. Given that 4TF-induced PGCLCs exhibit more advanced development compared to Ck-induced PGCLCs, we opted for the readily accessible neonatal testicular somatic cells to facilitate PGCLCs differentiation, along with key cytokines (i.e., GDNF and bFGF) crucial for spermatogonial proliferation. While no seminiferous tubule-like structure was observed, clusters of DDX4⁺ cells were formed and surrounded by somatic cells. Similar to previous observations ^17,39, a subset of DDX4⁺ cells underwent further differentiation into cells with spermatogonium-like characteristics. However, further work is necessary to explore whether these cells can progress to later stages of spermatogenesis.

Efficient, scalable and simple PGCLC differentiation systems, such as that we report here, will be useful for optimizing IVG and facilitating high-throughput screening platforms. One interesting parameter we observed was a correlation between the EB size and differentiation efficiency. EB size likely impacts differentiation via factors such as the diffusion of soluble molecules and the degree of interactions in both extracellular matrix-cell and cell-cell adhesive contexts ⁴⁰. In addition to scalability and efficiency, a useful aspect of the 4TF system I that the derived PGCLCs persisted for prolonged periods in culture, in contrast to a 3TF-induction system where efficiency declined after day 4 ⁹. However, further work is necessary to characterize the properties of these long-term cultured or differentiated cells.

In mice, competence for germline specification distinguishes the formative state of pluripotency from the naïve and primed states ⁴¹. Given that both formative ESCs and EpiLCs respond directly to PGC specification, it was unclear whether exposure of formative ESCs to 4TF overexpression was compatible with efficient differentiation into PGCLCs. Recent advancements have identified four distinct sub-stages of formative pluripotent stem cells, exhibiting functional characteristics and WNT/ β-catenin signaling modulation ^4-7. In our study, we cultivated formative ESCs using one of these protocols ⁷, and revealed that 4TFs can drive both formative ESCs and EpiLCs into PGCLCs. Importantly, enables simplification of the PGCLC differentiation process, as formative ESCs can be stably maintained and would not require preliminary induction into the transient EpiLC-like state from naïve ESCs.

We envision that the high efficiency 4TF system we described here will be useful for several applications. One such application is for conducting high-throughput screening of genetic factors that influence gametogenesis. For example, massively parallel CRISPR-based whole genome screens require large number of cells to accommodate libraries containing tens of thousands of guide RNAs. Scalability is also important for screening small molecule libraries for chemicals that may further improve differentiation, or libraries of drugs or compounds in the environment that may impair or negatively impact germ cell development. Considering that genetic causes underlie ~50% of human infertility ¹, parallelized screening of genetic variants, particularly rare or minor alleles that are classified as variants of uncertain significance (VUS), for impacts on IVG. Functionally interpreting the VUS within genes that essential for gametogenesis is important for genetic counseling, clinical applications, and potentially genome editing to enable people with identified genetic causes of infertility to produce viable gametes in vitro or in vivo ^1,42,43.

Materials and Methods

Mice and cell lines

All animal usage was approved by Cornell University’s Institutional Animal Care and Use Committee, under protocol 2004-0038 to J.C.S. Four- to eight-day-old Kit^W/Wv male mice generated by crossing C57BL/6J-Kit^Wv/J and WB/ReJ Kit^W/J were used in this study. C57BL/6J-Kit^Wv/J (#000049), WB/ReJ Kit^W/J (#000692) and B6(Cg)-Tyr^c-2J/J (#000058) mice were purchased from The Jackson Laboratory. The Stella-eGFP transgenic ESC line (Stella-eGFP ESCs) with Stella flanking sequences coupled to eGFP as the reporter was kindly provided by Prof. M. Azim Surani ¹⁹.

Vectors

To construct pPBhCMV1-Prdm1-pA and pPBhCMV1-Prdm14-pA, mouse Prdm1 and Prdm14 coding sequences were cloned into pPBhCMV1-Nanog-pA (a gift from M. Azim Surani) by PCR flanked with XhoI and NotI restriction sites. To construct targeting vectors containing enhancer-reporter transgenes, PCR-amplified sequences of mouse Prdm1 and Prdm14 enhancers ¹¹ bearing terminal NotI restriction sites were cloned into PCR4-Shh::lacZ-H11 (Addgene, # 139098). Plasmid clones were validated by Sanger sequencing. pPyCAG-Pbase and pPB-CAG-rtTA-IN (Addgene, #60612) were kindly provided by M. Azim Surani. pSpCas9(BB)-2A-PuroR (px459) V2.0 (Addgene, #62988), psPAX2 (Addgene, #12260), pMD2.G (Addgene, #12259) and pLV-tetO-Tfap2c (Addgene, #70269) were obtained from Addgene. Primer sequences are listed in Supplementary Table 1.

gRNA cloning

For CRISPR-assisted insertion of PCR4-Shh::lacZ-H11 vector into the H11 locus, we used a gRNA that was previously designed for this purpose ²⁰. Synthesized oligonucleotides were annealed and cloned into px459 vector. The sequences of all gRNAs are listed in Supplementary Table 1.

Lentiviral constructs

To isolate pLV-tetO-Tfap2c lentivirus, HEK293T cells (cultured in DMEM with 10% FBS, 1mM NEAA and penicillin–streptomycin) were transfected with 500μl Opti-MEM containing 30 μl TransIT^®-LT1 Transfection Reagent (Mirus, MIR2305), 10μg pLV-tetO-Tfap2c, 7.5μg psPAX2 and 2.5μg pMD2.G. Viral supernatants were collected at 48 and 72 hours, concentrated using Amicon Ultra-15 centrifugal filter units with an Ultracel-30 membrane (Millipore, Billerica, MA, USA), filtered through a 0.45 μM filter, and stored at −80°C until used.

ESC and iPSC culture

Mouse ESCs and iPSCs maintained on γ-irradiated feeder MEFs (mouse embryonic fibroblast cells) in DMEM containing 15% FBS, penicillin-streptomycin, NEAA, β-mercaptoethanol, 1,000U/ml LIF (Peprotech, 300-05). Naïve ESCs were cultured in 2i+ LIF medium containing 1×N2B27, 1 μM PD0325901 (Reprocell, 04-0006-02), 3 μM CHIR99021 (Reprocell, 04-0004-02) and 1,000 U/ml LIF on wells coated with 0.01% poly-L-ornithine (Sigma-Aldrich, P3655) and 10 ng/ml laminin (Corning, 354232), as described previously ³¹.

For the formative state conversion, naïve ESCs and iPSCs were dissociated into single cells and plated on MEFs plate in 2i +LIF medium. Follow the protocol describe previously ⁵, the next day, medium was changed to formative medium (N2B27 medium containing 10 ng/ml bFGF, 10 ng/ml Activin A and 3 μM CHIR99021) for further culture. It usually takes about 2-4 passages for fully conversion. Formative ESCs and iPSCs were maintained in formative medium.

Derivation of genetically modified ESC clones

A piggyBac transposon expression system was used to stably integrate single copies of various vectors that were doxycycline (Tet) inducible. To establish the Tet-on expression system, pPBhCMV1-Prdm1-pA, pPBhCMV1-Prdm14-pA and pPBhCMV1-Nanog-pA vectors were transfected using TransIT^®-LT1 Transfection Reagent (Mirus Bio^™, MIR2304) into ESCs together with pPBCAG-rtTA and pPyCAG-PBase vectors. After 1 week of Geneticin^™ (80 μg/ml; Life Technologies, 10131035) selection (all vectors except pPyCAG-PBase contain neomycin resistance genes), clones containing desired inserts were identified. Cells infected with pLV-tetO-Tfap2c lentivirus were selected by Zeocin^® (InvivoGen, ant-zn-05) to identify transformants.

To construct ESC lines containing enhancer-reporters, the targeting vector PCR4-Shh::lacZ-H11 and H11 gRNA-expression plasmid px459 were transfected into cells bearing Tet-inducible Nanog along. After puromycin selection for 48 hours, single colonies were expanded and correctly targeted clones were identified using primers listed in Supplementary Table 1.

Microinjection for chimera production

The ESCs were microinjected to B6(Cg)-Tyr^c-2J/J blastocysts by standard methods in Cornell’s Transgenic Core Facility ⁴⁴.

PGCLC differentiation

For PGCLC differentiation, EpiLCs were induced as described previously ³¹. Briefly, 1×10⁵ ESCs were placed per well of a 12-well plate coated with 16.7 μg/mL bovine plasma fibronectin (Sigma-Aldrich, F1141) in N2B27 medium containing 20 ng/mL Activin A (Peprotech, 120-14E), 12 ng/mL bFGF (Gibco, 13256-029), and 1% knockout serum replacement (KSR, ThermoFisher, 10828010). The medium was changed the following day. A total of 48 hours after plating, 2×10³ EpiLCs per well were seeded in a U-bottom 96-well plate (ThermoFisher Scientific, 268200) in GK15 medium (GMEM with 15% KSR, 0.1mM NEAA, 1mM sodium pyruvate, 0.1mM β-mercaptoethanol, penicillin-streptomycin, and 2mM L-glutamine) with Dox (Sigma, D9891). To induce PGCLCs from formative ESCs, after removing the feeder cells, 2×10³ cells per well were seeded in the U-bottom 96-well plate. Alternatively, 8×10⁴ cells per well were seeded in an untreated 12-well plate, or 1×10⁶ cells were seeded into a 100mm Petri dish in GK15 medium. Transgenes were induced by addition of 2 μg/ml Dox at day 0 of PGCLC induction. Cells were cultured at 37 °C in a 5% CO₂-95% air atmosphere.

Flow cytometry

To flow sort PGCLCs, single-cell suspensions were stained with 1 μg/ml propidium iodide (PI, Invitrogen; P3566) and filtered through a 70 μm cell strainer. eGFP⁺ cells were sorted using a BD FACSMelody 4-Way Sorter. Suspended cells were washed with PBS and fixed in 70% ethanol. For staining, cells were washed in MACS buffer (1×PBS, 0.5%BSA and 2mM EDTA) and blocked with 5% goat serum (Sigma; NS02L) in MACS buffer at 4°C for 1 hour. Cells were then stained with PE (Phycoerythrin) mouse/rat anti-CD61 (1:500, Invitrogen; 12-0611-82) and Alexa Fluor^® 647 mouse/human anti-SSEA1 antibodies (1:200, Biolegend; 125608) at 4°C for 1 hour. Cells were then washed in MACS buffer twice and filtered through a 35 μm cell strainer (Falcon; 352235). The stained cells were analyzed using a BD FACSymphony A3 cytometer. FACS data were analyzed using FlowJo^™ v10.4 or flowCore package in R.

Cell aggregate cultures for differentiation into spermatogonia

Sorted PGCLCs were mixed with dissociated neonatal testicular cells from Kit^W/Wv mice with 10:1 ratio and co-cultured for 2 days in a U bottom 96-well using GK15 medium. The aggregates were subsequently transferred to a trans-well plate in N2B27 medium containing 20 ng/ml bFGF, 40 ng/ml GDNF (PeproTech, 450-44), 1,000 U/ml LIF, 1×Insulin-Transferrin-Selenium (ThermoFisher Scientific, 41400045) and 15% KSR for additional 2 weeks. The aggregates were cultured at 37 °C in a 5% CO₂-95% air atmosphere.

Cell staining

ALP staining was performed using the Vector Red Alkaline Phosphatase Substrate Kit (Vector Laboratories, Burlingame, CA, USA) according to the manufacturer's directions. LacZ staining was performed the Senescence β-Galactosidase Staining Kit (Cell Signaling Technology, #9860) according to the manufacturer's directions.

For cell IF staining, the cells growing on cover slides were washed with PBS, fixed in 4% paraformaldehyde (PFA) for 10 min at RT, washed twice with PBS. For section IF staining, cell aggregates were fixed in 4% PFA and embedded in paraffin. The 6 μm sections were deparaffinized and performed antigen retrieval using sodium citrate buffer. The slides were incubated for 10 min at 37 °C in blocking buffer (PBS containing 10% normal goat serum). Then, they were incubated overnight in a humidified chamber at 4 °C with the following antibodies: Ddx4 (Abcam, ab13840, 1:500), Sox9 (Abcam, 1:100), Plzf (Thermo Fisher Scientific, 39987, 1:100), Nanos3 (Abcam, ab70001, 1:500), Zo1 (Thermo Fisher Scientific, 40-2200, 1:100), E-cadherin (1:100), Ssea1 (Millipore, MAB4301, 1:100) and Sox2 (Thermo Fisher Scientific, MA1-014, 1:100). After washing twice with PBS, the slides were incubated at 37 °C for 30 min with a 1:500 dilution of secondary antibody goat anti-rabbit IgG 594 (Molecular Probes, A11012) and/or goat anti-mouse IgG 488 (Molecular Probes, A11001), then incubated at 37 °C for 5 min with 500 ng/mL of DAPI and mounted using Vectashield antifade mounting medium (Vector, H-1000).

Bisulfite sequencing

Methylation patterns in the differentially methylated domains (DMDs) of the paternally imprinted gene H19 and maternally imprinted gene Snrpn were determined using bisulfite genomic sequencing as previously described ⁴⁵. Genomic DNA was isolated and bisulfite treatment was conducted with an EZ DNA methylation-lightning kits (ZYMO Research, D5030) according to the manufacturer’s protocol. Bisulfate-converted DNA was subjected to PCR amplification of the H19 and Snrpn DMDs. The primer sequences and PCR conditions for amplification are previously described ⁴⁵. PCR products were sub-cloned into the pGEM-T Easy vector (Promega, A1360) and sequenced. Sequences were determined and analyzed using BiQ Analyzer software ⁴⁶.

Reverse transcription and qRT-PCR

Total RNA extraction from cells was performed by E.Z.N.A.^® Total RNA Kit I (Omega BIO-TEK) according to the manufacturer’s instructions or extracted with TRIzol^® reagent (Thermo Fisher Scientific, 15596018). Reverse transcription was performed using a qScript^™ cDNA SuperMix kit (Quantabio, 95048). qRT-PCR was performed with a C1000 Touch^™ Thermal Cycler (Bio-Rad) amplification system using RT² SYBR^® Green qPCR Mastermixes (Qiagen) and primer sets specific for each gene (Table S1). The relative levels of transcripts were calculated using the 2-^ΔΔC_T method and gene expression levels were normalized to Gapdh.

RNA-seq and data analysis

Total RNAs were purified from naïve ESCs, EpiLCs, and PGCLCs (Stella-eGFP⁺) at days 2, 4 and 6 (three biological replicates for each) using TRIzol^® reagent. Purified RNAs were used for Illumina RNA-seq library preparation with NEBNext Ultra II Directional RNA Library Prep Kit (NEB, E7765), and a minimum of 20 million raw reads were obtained. The raw counts were then aligned to mouse mm10 genome with ENSEMBL gene annotations. Differential analysis of RNA-seq data was done using the R package DESeq2. RRHO2 analysis was performed also using an R package ²⁴. k-Means clustering analysis was performed using iDEP 1.12 ⁴⁷. GSEA was performed to explore enriched pathways and interpret RNA-seq data using predefined gene sets from the Molecular Signatures Database (v.7.1) in R packages ⁴⁸. GSVA was carried out using the GSVA package in R ⁴⁹. Gene sets enriched in the GSEA were organized into a graphical network produced using EnrichmentMap ⁵⁰ plugins in Cytoscape (v.3.10.1) ⁵¹, as described previously ⁵². The online STRING database (version 12.0) was used for generation of protein interaction ⁵³. The upregulated differentially expressed genes (log2FC>2, p<0.05 and count>250) from PGCLCs on days 2, 4 and 6 compared to ESC were integrated for the network analysis.