Marmoset fibroblast reprogramming into cjiPSCs

To further establish the marmoset monkey as a model for studying germ cell development in vitro, we first generated a panel of cjiPSC lines. We followed our previously described approach to derive iPSCs from human, macaque, and baboon fibroblasts under feeder- and transgene-free conditions (Stauske et al, 2020; Rodríguez-Polo et al, 2022) (Fig 2A). To this end, primary fibroblasts isolated from one foetal and two postnatal marmosets were transfected with episomal plasmids for the transient expression of human OCT3/4, SOX2, KLF4, L-MYC, LIN28, and a small hairpin RNA against p53 (Okita et al, 2011). After transfection with the episomal plasmids, primary cells were maintained for ∼40 d in E8 medium or Universal Primate Pluripotent Stem Cell (UPPS) medium (Table S1) (Rodríguez-Polo et al, 2022), resulting in putative cjiPSC colonies with characteristic morphology (Fig 2A and B).

Between d30 and d40, cjiPSC colonies were manually picked and expanded in UPPS medium. After ∼5 passages, the cjiPSC lines gave rise to flat and compact colonies with sharp borders and cells with an apparently high nucleus-to-cytoplasm ratio, which is reminiscent of primed hiPSCs (Stauske et al, 2020). Between passages 5 and 10, we selected six cjiPSC lines, two derived from neonatal fibroblasts (DPZ_cjiPSC#1 and DPZ_cjiPSC#6) and four from foetal fibroblasts (DPZ_cjiPSC#2-5) (Fig 2C). However, in contrast to human or other NHP iPSCs under identical conditions (Stauske et al, 2020), the putative cjiPSC lines were less stable at early passages resulting in a high percentage of differentiated cells. To address this, we empirically tested the addition of different combinations of small molecules and cytokines to the UPPS medium. We were able to stabilise the cjiPSC lines by alternating the cell culture conditions every two to three passages between UPPS medium alone and UPPS medium together with Activin A (ActA) and LIF (Fig 2A). Around passages 10–15, six selected cjiPSC lines (DPZ_cjiPSC#1-6) could be stably maintained in UPPS medium alone (Fig 2C). After expansion, we confirmed pluripotency-associated alkaline phosphatase activity in the six cjiPSC lines (Fig 2D).

PCR analysis showed that five of six lines (DPZ_cjiPSC#2-6) lost the episomes by passage 15 (Fig 3A). In addition, semi-quantitative transcript abundance analysis by RT–PCR confirmed that the cell lines expressed the endogenous pluripotency factors OCT4, KLF4, and c-MYC at high levels (Fig 3B). In addition, four further characterised cjiPSC lines (DPZ_cjiPSC#1-3, 5) showed the protein expression of pluripotency factors OCT4A, LIN28, NANOG, and SOX2, as well as the pluripotency-associated glycans TRA-1-81 and TRA-1-60 (Fig 3C). Finally, we tested the differentiation potential of these cjiPSC lines using an embryoid body (EB) formation assay. Cell aggregates were generated and exposed to differentiation medium (see the Materials and Methods section) for 8d in suspension followed by 17d culture after EB attachment to gelatine-coated coverslips. Immunofluorescence stainings confirmed that the outgrowths of cell aggregates developed into representative cell types of the three embryonic germ layers, as judged by the expression of smooth muscle actin (mesoderm), α-fetoprotein (endoderm), and β-III-tubulin (ectoderm) (Fig 3D). Taken together, we have established three stable, transgene- and feeder-free pluripotent cjiPSC lines (DPZ_cjiPSC#2, DPZ_cjiPSC#3, and DPZ_cjiPSC#5).

Differentiation of cjiPSCs towards cjPGCLCs

To generate cjPGCLCs from cjiPSCs, we tested whether cjiPSCs (DPZ_cjiPSC#2, DPZ_cjiPSC#3, and DPZ_cjiPSC#5) are competent to be directly induced into cjPGCLCs. CjiPSCs were adapted to a modified UPPS medium containing forskolin, LIF, and a low concentration of ActA, which improves the stability of cjiPSCs (Petkov et al, 2020). Dissociated cjiPSCs were reaggregated in ultra-low attachment wells to allow the formation of EBs in the presence of BMP4, EGF, SCF, and LIF. At d2 of differentiation, small clusters of AP2Ɣ-SOX17 and/or AP2Ɣ-BLIMP1 double-positive cells were detected (Fig S1A). However, by d6 of differentiation AP2Ɣ-SOX17 and/or AP2Ɣ-BLIMP1 expression appeared to be mutually exclusive, suggesting that the cells adopted different cell fates during the time course of differentiation (Fig S1B).

Next, we tested a two-step protocol that has been successfully applied to differentiate macaque and human PGCLCs from PSCs (Kobayashi et al, 2017). In this approach, PSCs are differentiated into pre-ME using ActA and CHIR, a small molecule inhibitor of GSK3, before being induced into PGCLCs by the addition of BMP4, EGF, SCF, and LIF. To this end, cjiPSCs (DPZ_cjiPSC#2 unless otherwise stated) or cynomolgus monkey ESCs (cyESCs), used as a positive control, were adapted to E8 medium and cultured on mouse embryonic fibroblasts. At d4 of differentiation, cynomolgus monkey EBs gave rise to large clusters of SOX17-AP2Ɣ– and NANOG-BLIMP1–positive cells (Fig S1C), consistent with a previous study (Kobayashi et al, 2017). In contrast, although a considerable number of cells within marmoset EBs expressed SOX17, only very few cells co-expressed AP2Ɣ (Fig S1D). Furthermore, NANOG and BLIMP1 expression was barely detectable in marmoset EBs, indicating that SOX17-positive cells may have acquired an endodermal rather than the PGC fate. These results suggest that cjiPSCs require different maintenance or differentiation conditions for the PGCLC fate than other primate pluripotent cells.

Next, we adapted cjiPSCs (DPZ_cjiPSC#2, DPZ_cjiPSC#3, and DPZ_cjiPSC#5) to TeSR-E8 (TESR) medium without feeders, which is regularly used to maintain human iPSCs (hiPSCs). cjiPSCs cultured in TESR medium retained undifferentiated morphology and expressed the pluripotency factor OCT4A (Fig S2A). However, only the differentiation of hiPSCs but not cjiPSCs into pre-ME, followed by induction into PGCLCs, resulted in large clusters of SOX17-AP2Ɣ– and NANOG-BLIMP1–positive cells (Fig S2B and C).

The culture of cjiPSCs in TESR medium did not result in successful cjPGCLC differentiation. Thus, we considered the origin of primate PGCs, which may be derived from or have the same precursors as extraembryonic amniotic cells (Sasaki et al, 2016; Chen et al, 2019; Castillo-Venzor et al, 2023). Because hESCs or hiPSCs in human expanded potential stem cell medium (hEPSCM, Table S1) have the developmental potential to develop into embryonic germ layers, hPGCLCs, and extraembryonic lineages (Gao et al, 2019), we sought to adapt cjiPSCs to hEPSCM. Although the transfer of cjiPSCs from TESR to hEPSCM resulted in cell death, cjiPSCs could be maintained in hEPSCM supplemented with TGFβ and FGF2, which promote pluripotency and self-renewal (James et al, 2005; Vallier et al, 2009; Chen et al, 2011; Gafni et al, 2013). cjiPSCs cultured in this medium, hereafter referred to as cjPSCM (marmoset PSC medium), homogenously express OCT4A, LIN28, and TRA-1-60 (Fig S3A), suggesting that pluripotency is maintained. To further characterise the pluripotent state of cjiPSCs cultured in TESR or cjPSCM, we performed RNA sequencing (RNA-seq). Differential gene expression analysis using DESeq2 (Padj < 0.05, log₂FoldChange > 2) revealed 857 up- and 527 down-regulated genes in cjiPSCs cultured in cjPSCM as compared to TESR medium (Fig 4A). The expression of core pluripotency factors including OCT4A, NANOG, and SOX2 was unchanged. This is in contrast to the expression profile of hPSCs cultured in hEPSCM, where these factors are up-regulated compared with primed conditions (Gao et al, 2019). Also, we could not detect changes in the expression levels of DNA methyltransferases DNMT3A, DNMT3B, and DNMT1. However, many transcription factors associated with differentiation, such as WNT3A, PAX6, FGF9, HEY1, and TBX-T, were significantly down-regulated in cjiPSCs cultured in cjPSCM. Accordingly, gene ontology (GO) classification with down-regulated genes showed enrichment of terms associated with cellular differentiation, including pattern specification process, regionalisation, embryonic organ development, and urogenital system development (Fig S3B). In contrast, transcription factors considered to be associated with naïve pluripotency including DPPA3, TFAP2C, KLF5, KLF4, GDF3, and ANPEP were significantly up-regulated (Fig 4A) (Nakamura et al, 2016; Boroviak & Nichols, 2017; Bergmann et al, 2022). These results indicate that cjiPSCs require a species-specific cocktail of factors for expanded pluripotency, as cjPSCM does not fully induce the transcriptional changes associated with this stem cell state. Consistently, the media composition to induce expanded pluripotency is different for mouse, porcine, and hPSCs (Yang et al, 2017; Gao et al, 2019).

To further characterise the pluripotent state of cjiPSCs in cjPSCM, we compared the transcriptomes of cjiPSCs with available published datasets of hPSCs in naïve, formative, expanded potential and primed conditions (Gao et al, 2019; Kinoshita et al, 2021). Principal component analysis shows that hPSCs are separated across the pluripotency spectrum (Fig S3C). In particular, naïve hPSCs are distant from all other samples, whereas primed hPSCs are separated from formative and expanded pluripotent hPSCs along PC2. Interestingly, we observed a comparable shift of cjiPSCs cultured in cjPSCM from cjiPSCs in TESR medium. Similar to formative and expanded pluripotent hPSCs, cjPSCs in cjPSCM express a subset of genes associated with naïve and formative pluripotency including ANPEP, PECAM1, POU5F1, NANOG, KLF5, DPPA3, NLRP7, DPPA3, KLF5, OTX2, and TFAP2C, as well as primed pluripotency including FAT3, THY1, and SOX11 (Fig 4B). However, there are also notable differences including the up-regulation of SOX17 and FOXC1 in cjiPSCs. Taken together, these results indicate that cjPSCM induces gene expression changes in cjiPSCs that may lead to a partial reversion from primed to a formative-like state of pluripotency.

Next, we asked whether cjiPSCs cultured in cjPSCM gain competence to differentiate into pre-ME and cjPGCLCs. Indeed, immunofluorescence analysis of d4 EBs shows the expression and colocalisation of SOX17, AP2γ, NANOG, and BLIMP1 in a small number (6.9%, SEM: 1.51%) of cells (Fig 5A). In contrast, the definitive endoderm marker FOXA2 was only detected in a few SOX17 single-positive cells (Fig S3D). In addition, we empirically tested different conditions and found that the removal of ActA from the pre-ME differentiation medium apparently increased the number (9.9%, SEM: 2.65%) of cjPGCLCs (Fig 5B). Importantly, the differentiation of two additional cjiPSC lines (DPZ_cjiPSC#3 and DPZ_cjiPSC#5) in cjPSCM led to SOX17-BLIMP1 double-positive cells in d4 EBs (Fig S4A and B). Taken together, these results suggest that cjiPSCs cultured in feeder-free conditions with cjPSCM can be differentiated into pre-ME and induced into cjPGCLCs.

These results prompted us to test whether cjiPSCs in primed conditions with TESR medium, give rise to cjPGCLC-competent pre-ME if ActA is omitted during differentiation. Although immunofluorescence analysis of d4 EBs shows the expression of SOX17, AP2Ɣ, and BLIMP1 in a low number of cells, the expression of these PGC markers was largely not colocalised (Fig S5A), which is reminiscent of EBs induced from pre-ME with ActA (Fig S2C). To characterise this further, we performed RNA-seq with pre-ME induced with or without ActA from cjiPSCs in TESR medium. Differential gene expression analysis using DESeq2 (Padj < 0.05, log₂FoldChange > 2) showed significant up-regulation of 176 genes, many of which are associated with mesoderm development, including EOMES, SNAI1, GSC, HAND3, and WNT3 (Fig S5B and C). Consistently, GO classification with up-regulated genes shows enrichment for terms such as muscle tissue development, pattern specification process, gastrulation, and mesoderm development (Fig S5D). Considering that WNT signalling is required for PGC specification (Chen et al, 2017, 2019), it may be that the up-regulation of various WNT-associated genes upon the addition of ActA primes pre-ME from cjiPSCs in TESR medium towards the mesoderm fate. Taken together, these results demonstrate that cjiPSCs in a pluripotent state associated with cjPSCM in contrast to TESR medium give rise to cjPGCLC-competent pre-ME.

cjPGCLCs express genes associated with the PGC fate

We asked whether cjPGCLCs up-regulate the germ cell–specific transcriptional programme, which entails the up-regulation of genes associated with pluripotency and the germ cell fate, whereas genes associated with somatic fates are down-regulated (Kurimoto et al, 2008; Irie et al, 2015). To this end, we differentiated cjiPSCs in cjPSCM with or without ActA into pre-ME, which were then induced into cjPGCLCs. CjPGCLCs were isolated by FACS from d4 EBs after staining for the cell surface markers INTEGRINα6 (INTα6) and CXCR4 (Fig 6A–C), which are routinely used to sort primate PGCLCs (Sasaki et al, 2015; Kojima et al, 2017; Mitsunaga et al, 2017; Sakai et al, 2020; Seita et al, 2023).

RNA-seq analysis of pre-ME induced with ActA (pre-ME + ActA) resulted in significant up-regulation of differentiation-associated genes associated with GO terms such as pattern specification process and mesoderm development (Fig S6A). This includes EOMES and WNT3A (Fig S6B), which are involved in PGCLC specification in primates (Chen et al, 2017, 2019). Similarly, pre-ME induced without ActA (pre-ME − ActA) showed up-regulation of genes associated with pattern specification process and BMP signaling pathway (Fig S6C), which, however, did not include EOMES (Fig S6D). Instead, we noticed a pronounced transcriptional increase of TFAP2A (Fig S6D), which is required for the induction of the hPGC fate from progenitor cells (Chen et al, 2019; Castillo-Venzor et al, 2023). Direct comparison of pre-ME + ActA with pre-ME − ActA revealed 89 and 198 of up- and down-regulated genes (Padj < 0.05, log₂FoldChange > 2), respectively. Most notably, HOXA genes in pre-ME − ActA were significantly down-regulated (Fig S6E). These data show that the expression of somatic differentiation markers is up-regulated in pre-ME, which is more pronounced in pre-ME induced with ActA. Moreover, the up-regulation of TFAP2A in pre-ME − ActA might be an indicator of cjPGCLC competence.

Finally, we asked whether marmoset pre-ME retain the expression of subsets of genes associated with pluripotency, as it was shown for human pre-ME (Kobayashi et al, 2017; Tang et al, 2022; Alves-Lopes et al, 2023). Analysis of gene expression profiles in marmoset cjiPSCs and pre-ME compared with human ESCs and pre-ME using published RNA-seq datasets (Tang et al, 2022) confirmed that they share the expression of subsets of genes associated with naïve and primed pluripotency (Fig S7).

The INTα6-CXCR4 double-positive cjPGCLCs induced from pre-ME + ActA or pre-ME − ActA as compared to INTα6-CXCR4 double-negative somatic cells showed up-regulation of key PGCLC markers, including SOX17, TFAP2C, PRDM1, EOMES, NR5A2, NANOG, OCT4 (POU5F1), and DPPA3 (Fig 6D and E). Importantly, this was also the case for NANOS3, which is specifically expressed in the germline (Tsuda et al, 2003). Conversely, somatic genes including HOXD4, HOXB6, MIXL1, and HAND1 were down-regulated. Also, the pluripotency factor SOX2 was expressed but down-regulated, which is one of the key differences between rodent and primate PGC development (Irie et al, 2015; Sasaki et al, 2015; Sugawa et al, 2015).

Next, we compared the transcriptomes of cjPGCLCs with published datasets for mPGCLCs and human PGCLCs (hPGCLCs) (Sasaki et al, 2015; Tang et al, 2022). Pearson’s correlations show that the transcriptomes of d4 cjPGCLCs cluster with d4 hPGCLCs (0.59–0.63) (Fig 7A). In contrast, both hPGCLC and cjPGCLC transcriptomes correlate less (0.25–0.41) with mPGCLCs at d4 or d6 of differentiation.

To characterise this further, we analysed the gene expression profile of various PGC markers between PGCLC samples, including published datasets for embryonic human PGCs at week 7 of development (Tang et al, 2022). A large number of genes showed a similar expression profile in PGCLCs of all three species including the expression of POU5F1 (OCT4), NANOG, PRDM1, TFAP2C, and KIT (Fig 7B). Notably, PRDM14 could be only detected in hPGCLCs and mPGCLCs, indicating species-specific differences. Importantly, we also observed primate-specific up-regulation of PGC-associated genes including SOX17, KLF4, TFAP2A, GATA3, and TBX3 in hPGCLCs and cjPGCLCs as opposed to mPGCLCs. Taken together, these data suggest that cjiPSCs cultured in cjPSCM in feeder-free conditions can be induced into cjPGCLCs with a transcriptional profile reminiscent of hPGCLCs.