Research Article

Dissecting primate early post-implantation development using long-term in vitro embryo culture

See allHide authors and affiliations

Science  15 Nov 2019:
Vol. 366, Issue 6467, eaaw5754
DOI: 10.1126/science.aaw5754

In vitro development of monkey embryos

Owing to technical and ethical limitations, the molecular and cellular mechanisms underlying primate gastrulation are far from clear (see the Perspective by Tam). Two independent studies used an in vitro culture system to study cynomolgus monkey embryo postimplantation development up to and beyond gastrulation (day 9 to day 20). Niu et al. observed in vivo morphogenetic events and used single-cell RNA sequencing and single-cell chromatin accessibility to study the distinct cell lineages in developing embryos. Ma et al. also observed that key events of in vivo early development were recapitulated in their system, and single-cell RNA-sequencing analysis revealed molecular signatures of postimplantation cell types. These systems will help elucidate the dynamics and regulation of gastrulation in primates, including possible relevance to human development.

Science, this issue p. eaaw5754, p. eaax7890; see also p. 798

Structured Abstract


The period from peri-implantation to gastrulation is critical for mammalian embryogenesis. During this time, connections between embryonic and maternal tissues are set up, and the primary germ layers and body plan are established. There is a substantial gap in our knowledge of early human postimplantation development because of technological limitations and ethical considerations. To extend the study of human embryogenesis to the postimplantation period, an in vitro culture system has been established that extends human blastocyst development to the pregastrulation stage (up to12 days) after fertilization, and the molecular and cellular events are revealed.


With the general prohibition of growing human embryos beyond 14 days, closely related surrogate species can be examined. In addition, improvements are needed for primate embryo culture to support extended growth periods. The establishment of an in vitro culture system that enables the development of primate embryos beyond the implantation period provides an accessible way to study molecular and cellular mechanisms that underlie postimplantation development, including gastrulation.


In this study, we have modified a human embryo in vitro culture protocol that enables the development of cynomolgus monkey (long-tailed macaque) embryos to develop up to 20 days after fertilization. The cultured cynomolgus embryos recapitulated key primate in vivo morphogenetic events, including amniotic and yolk sac cavitation, embryonic and extraembryonic lineage specification, specification of primordial germ-like cells (PGCLCs), and primitive streak cells. We demonstrated that the amniotic lumenogenesis is accompanied by the polarization of the epiblast (EPI); however, the polarization of PE was not observed during yolk sac cavitation. We used single-cell RNA-sequencing to delineate the developmental trajectories of EPI, trophoblast, and primitive endoderm (PE) lineages. We observed that accompanying the transition from the naïve to primer state, the metabolic mode of oxidative phosphorylation is no longer used in the EPI cells. Furthermore, the trophoblast differentiates in a stepwise manner, and expression of the trophoectoderm marker CDX2 decreases rapidly after day 11 in the trophoblast but maintains in the amniotic epithelium cells. In addition, we identified two types of PE lineage. Coordinated interactions were observed among EPI, trophoblast, PE, and extraembryonic mesenchyme cells during the postimplantation period. Furthermore, we showed that PGCLCs specified in vitro are similar to early-stage PGCs in vivo. Using the single-cell assay for transposase-accessible chromatin followed by sequencing (scATAC-seq), we also identified EPI, trophoblast, PE, and EXMC lineages. Last, scATAC-seq revealed that distal regions of chromatin in the EPI lineage exhibited higher accessibility than in other cell types.


In this study, we show that monkey embryos show robust development beyond 14 days after fertilization, surviving until day 20 without support from maternal tissue. We also provide insights into the transcriptional programs and chromatin dynamics that underlie monkey post-implantation development. Our system provides a platform to analyze molecular and cellular dynamics during primate early development. Last, our data may help guide the development of improved differentiation protocols for primate pluripotent stem cells.

Monkey embryos cultured in vitro recapitulate primate postimplantation embryogenesis in vivo.

Scheme of monkey postimplantation embryogenesis cultured in vitro. ICM, inner cell mass; EPI, epiblast; PE, primitive endoderm; TE, trophectoderm; AMEC, amniotic epithelium cell; PGC, primordial germ cell; VE, visceral endoderm; YE, yolk-sac endoderm; EXMC, extra-embryonic mesenchyme cell; and GAS, gastrulating cell.


The transition from peri-implantation to gastrulation in mammals entails the specification and organization of the lineage progenitors into a body plan. Technical and ethical challenges have limited understanding of the cellular and molecular mechanisms that underlie this transition. We established a culture system that enabled the development of cynomolgus monkey embryos in vitro for up to 20 days. Cultured embryos underwent key primate developmental stages, including lineage segregation, bilaminar disc formation, amniotic and yolk sac cavitation, and primordial germ cell–like cell (PGCLC) differentiation. Single-cell RNA-sequencing analysis revealed development trajectories of primitive endoderm, trophectoderm, epiblast lineages, and PGCLCs. Analysis of single-cell chromatin accessibility identified transcription factors specifying each cell type. Our results reveal critical developmental events and complex molecular mechanisms underlying nonhuman primate embryogenesis in the early postimplantation period, with possible relevance to human development.

From peri-implantation, when the blastocyst attaches to the uterine endometrium, to gastrulation when the primary germ layers and body plan are established, mammalian embryos undergo numerous cellular and molecular transitions (1). Mouse studies have revealed the molecular mechanisms that control gastrulation (27). However, there are considerable anatomical, physiological, and developmental differences between mice and humans, limiting the insights derived from mice that are applicable to human embryogenesis (810). Although our knowledge of early human development is improving, owing to the derivation of embryonic stem cells (ESCs) from human blastocysts, our understanding of the molecular and cellular events during early human postimplantation development remains limited.

Recently, an in vitro culture system that allows human embryos to develop beyond the blastocyst stage has facilitated study of the postimplantation period. Epiblast (EPI) expansion, lineage segregation, and amnion and yolk sac formation have been observed in vitro (11, 12), but embryos become disorganized by day 12. Beyond that point, improved culture conditions are needed. However, there are ethical considerations in generating human embryos, such as the 14-day rule. It is expected that examination of closely related species would reveal mechanisms conserved among different primates, including human. Among nonhuman primates, the monkey is our closest kin and represents an excellent surrogate species. In vivo derived cynomolgus monkey embryos have been used to analyze several primate postimplantation events, including amnion formation and germ cell specification. For example, single-cell RNA-sequencing (scRNA-seq) analysis has helped delineate the pluripotency program in EPI cells (13, 14). However, critical questions remain, including whether primate embryos can be cultured beyond the gastrulation stage in vitro and what the molecular determinants are of embryonic and extraembryonic lineage specification.

To address these questions, we used the human embryo culture protocol (11, 12) to culture cynomolgus embryos beyond gastrulation (~14 days) and up to 20 days in vitro. The cultured cynomolgus embryos underwent peri- and postimplantation events typical of higher primate species (9, 10, 15), including embryonic and extraembryonic lineage segregation, bilaminar disc generation, amniotic and yolk sac cavitation, and primordial germ cell–like cell (PGCLC) specification. scRNA-seq and single-cell assay for transposase-accessible chromatin–using sequencing (scATAC-seq) were used to analyze transcriptional and accessible chromatin profiles of cynomolgus postimplantation embryos at single-cell resolution, delineating the developmental trajectories of EPI, primitive endoderm (PE), trophectoderm (TE), and PGCs and characterizing transcription factor (TF) regulatory networks and signaling pathway interactions. Our results reveal the molecular details of early cynomolgus development and will help gain mechanistic insights into primate embryogenesis.


In vitro culture of cynomolgus embryos

To determine whether primate embryos could be cultured in vitro beyond gastrulation (~14 days), we used the human embryo culture protocol (11, 12) with several modifications: Advanced Dulbecco’s Modified Eagle Medium/Ham’s F-12 (Advanced DMEM/F12) was replaced with DMEM/F12, which was more suitable for monkey embryos culture, and 8 μM Rho-associated, coiled-coil containing protein kinase (ROCK) inhibitor, Y-27632, was included in culture media to prevent apoptosis (Materials and methods). Fertilized embryos were cultured to the blastocyst stage, the zona pellucida was removed, and denuded blastocysts were plated; 99.4% (n = 104 embryos) were attached to the cell-culture dishes by way of the polar TE, similar to the human embryos (11, 12). Mouse blastocysts implant by way of the mural TE in vivo and in vitro (16, 17), suggesting species-specific differences in the establishment of the embryo-endometrial epithelial interface. Most (92.31%, n = 104 embryos) cynomolgus embryos attached at day 10, and after attachment, the monkey blastocyst cavity expanded continuously until day 20 (21.74%, n = 46 embryos) (Fig. 1A). Inner cell mass (ICM) cells also continued to proliferate, forming a disc-like structure (Fig. 1B), with diameter increased from 113.75 by 86.28 μm (long axis by short axis, respectively) at day 9 to 216.62 by 178.33 μm at day 20 (Fig. 1B). We observed trophoblast villi-like structures sprouting from the embryos at day 15 and became evident at day 17 (Fig. 1C). In comparison, the human embryo blastocyst cavity collapsed soon after embryo attachment at day 8, and the disc-like and placenta villi–like structures were not clearly observed (11, 12).

Fig. 1 An in vitro culture system for monkey embryos.

(A) Representative bright-field images of monkey embryos developing in vitro until day 20 (n = 9 embryos for days 13, 15, and 17 and n = 7 embryos for other stages). Scale bars, 100 μm. Bottom right shows higher magnification of area indicated in dotted box in day 15. (B) Measurements of length of embryonic disc (micrometer by micrometer) (long axis by short axis) (n = 9 embryos of days 13, 15, and 17 and n = 7 embryos of other stages). (C) Representative bright-field images of villous-like structure at days 15 and 17 in cultured embryos (n = 9 embryos of each stage). Scale bars, 200 μm. Arrows indicate villous-like structure. EPI, epiblast; AM, amnion; VE, visceral endoderm; SYS, secondary yolk sac.

Lineage specification in cynomolgus embryos

Expression of lineage-specific markers—EPI (OCT4, NANOG), hypoblast (GATA6), and TE (CDX2)—was analyzed by means of immunofluorescence (IF) in in vitro cultured cynomolgus embryos. At day 8, OCT4 was principally expressed in the ICM and weakly expressed in the TE. GATA6 was expressed in all CDX2+ cells. GATA6+/CDX2+/OCT4+ cells were found in the ICM on day 8, suggesting that lineage specification was not complete in these cells (Fig. 2A). GATA6 expression is not detectable in human and mouse TE cells (13), suggesting species-specific differences in TE specification between humans and monkeys. By day 9, a few GATA6+/OCT4+ cells were detected in the ICM, and CDX2+/OCT4+ cells were detected in the TE, suggesting that lineage transitions are still occurring in these cells (Fig. 2B).

Fig. 2 Embryonic and extraembryonic lineage specification in monkey embryos cultured in vitro.

(A) Representative IF images of embryos at day 8 (n = 7 embryos). OCT4 (gray), GATA6 (red), CDX2 (green), and DAPI (blue). Scale bar, 100 μm. (Bottom) Magnification of ICM (yellow dotted lines). White dotted lines, ICM. Scale bar, 50 μm. Arrowheads, GATA6+/OCT4+ cells in ICM; arrows, CDX2+/OCT4+ cells in TE. (B) Representative IF images of embryos at day 9 (n = 4 embryos). OCT4 (gray), GATA6 (red), CDX2 (green), and DAPI (blue). Scale bar, 100 μm. (Bottom) Magnification of ICM (yellow dotted lines). White dotted lines, ICM. Scale bar, 50 μm. Arrowheads, GATA6+/OCT4+ cells in ICM; arrows, CDX2+/OCT4+ cells in TE. (C) Representative IF images of embryos at day 10 (n = 7 embryos). OCT4 (gray), GATA6 (red), CDX2 (green), and DAPI (blue). Scale bar, 250 μm. (Bottom) Magnification of inserts (yellow dotted lines) in top panels. White dotted lines, ICM. Scale bar, 100 μm. (D) Representative IF images of embryos at day 11 (n = 4 embryos) staining for OCT4/NANOG/GATA6. Scale bar, 50 μm. (Bottom) Zooms of inserts (yellow dotted lines) in top panels. Scale bar, 25 μm. Arrows, EPI; arrowheads: VE/YE. (E) Representative IF images of embryos at day 12 (n = 7 embryos) staining for OCT4/GATA6/CDX2. Scale bar, 100 μm. (Bottom) Zooms of inserts (yellow dotted lines) in top panels. Arrows, amniotic epithelium cells. White dotted lines, EPI. Scale bar, 50 μm. (F) Representative IF images of embryos at day 13 (n = 6 embryos) staining for OCT4/NANOG/GATA6. Scale bar, 100 μm. (Bottom) Enlargements of inserts (yellow dotted lines) in top panels. Scale bar, 50 μm. Arrows, amniotic epithelium cells. White dotted lines, EPI. (G) Representative IF images of day 14 embryos (n = 7 embryos) staining for OCT4/NANOG/GATA6. Scale bar, 100 μm. (Bottom) Enlargements of inserts (yellow dotted lines) in top panels. Scale bar, 50 μm. Arrows, the amniotic epithelium cells. White dotted lines, EPI. ICM, inner cell mass; TE, trophectoderm; EPI, epiblast; VE, visceral endoderm; YE, yolk-sac endoderm.

After attachment (approximately day 10), few OCT4+/GATA6+ and SOX17+/OCT4+ cells were detected in the ICM. CDX2+/GATA6+ cells were detected mainly in the TE, implying complete lineage segregation of EPI and TE at this stage (Fig. 2C and fig. S1A). On day 11, GATA6+ cells formed a layer ventral to the EPI cells, indicating ongoing differentiation of the visceral and yolk sac endoderm (VE/YE) (Fig. 2D).

TE marker CDX2 was coexpressed with OCT4 and GATA6 in peri-implantation cynomolgus embryos (days 8 and 9) (Fig. 2, A and B). These markers are not coexpressed in mouse and human embryos (13, 18); CDX2 is expressed in mouse trophoblast stem cells (TSCs) but not in human TSCs (19). To study TE development in cynomolgus embryos, we analyzed CDX2, OCT4, and GATA6 expression. CDX2 was not detected in trophoblast derivatives from day 12 onward, whereas it was expressed in amniotic epithelium cells, which showed low or undetectable GATA6 levels (Fig. 2E and fig. S1B). From day 12 onward, two groups of OCT4+ cells were present: columnar cells in proximity to the VE/YE and squamous cells close to trophoblast, a sign of the amniotic epithelium differentiation (Fig. 2, F and G). NANOG was expressed at high levels in the EPI and was only weakly expressed or undetectable in the amniotic epithelium, whereas OCT4 was expressed both in the EPI and amniotic epithelium (Fig. 2, F and G, and fig. S1C). By contrast, mouse NANOG expression decreases after embryo implantation (20), representing species-specific difference.

Investigation of the expression of TE markers CDX2, GATA3, and CK7 (2123) revealed that CDX2 was not expressed in the CK7+ cells at day 12. CK7/GATA3 expression was maintained in subpopulations of trophoblast cells until day 17 (fig. S1D). These data imply that the putative placenta progenitors persisted at least until day 17, indicating that CDX2 alone is not a suitable marker for cynomolgus TS cells. Thus, cynomolgus embryos cultured in vitro could initiate embryonic and extraembryonic lineage specification in the absence of the maternal environment and maternal tissues.

Apical adherent junctions and the amnion

We asked whether in vitro cultured cynomolgus embryos could form amnion and yolk sac cavities. NANOG, FOXA1, and COL6A1 expression were used to delineate the EPI, VE/YE, and extraembryonic mesenchyme cells (EXMCs) (13, 15, 24). At day 11, we found a small lumen surrounded by NANOG+ cells. VE/YE cells (FOXA1+) lined beneath the EPI cells (Fig. 3A). At day 13, primary yolk sac composed of FOXA1+ cells was visible (Fig. 3B). The EPI continued to proliferate, forming a convex shape relative to the amnion cavity from day 17 onward, and the secondary yolk sac–like cavity was clearly observed beneath the NANOG-expressing EPI (Fig. 3C).

Fig. 3 Amnion and yolk sac cavity formation.

Embryos at indicated stage were stained with FOXA1 (PE marker), NANOG (EPI marker), and COL6A1 (EXMC marker) to monitor amnion and yolk sac cavity formation. (A) Representative IF images of embryos at day 11 (n = 4 embryos) staining for FOXA1/NANOG/COL6A1. Scale bar, 250 μm. (Bottom) Enlargements of inserts (yellow dotted lines) in top panels. Scale bar, 50 μm. Yellow dotted lines, EPI; white dotted lines, yolk sac; arrowheads, EXMC. (B) Representative IF images of embryos at day 13 (n = 6 embryos) staining for FOXA1/NANOG/COL6A1. Scale bar, 250 μm. (Bottom) Enlargements of inserts (yellow dotted lines) in top panels. Scale bar, 50 μm. Arrow, primary yolk sac cavity; white dotted lines, EPI; arrowheads, EXMC. (C) Representative IF images of embryos at day 17 (n = 6 embryos) staining for FOXA1/NANOG/COL6A1. Scale bar, 250 μm. (Bottom) Enlargements of inserts (yellow dotted lines) in top panels. Scale bar, 50 μm. Arrow, secondary yolk sac cavity; white dotted lines, EPI; arrowheads, EXMC. (D) Representative IF images of embryos at day 11 (n = 4 embryos) staining for OCT4/aPKC/E-CAD. (Bottom) Enlargements of inserts (yellow dotted lines) in top panels. Scale bar, 50 μm. Arrows, apical localization of aPKC; arrowheads, apical localization of E-CAD; white dotted lines, EPI. (E) Representative IF images of embryos at day 12 (n = 7 embryos) staining for OCT4/aPKC/E-CAD. (Bottom) Enlargements of inserts (yellow dotted lines) in top panels. Scale bar, 50 μm. Arrows, apical localization of aPKC; arrowheads, E-CAD apical localization; white dotted lines, EPI. (F) Representative IF images of embryos at day 11 (n = 4 embryos) and day 17 (n = 6 embryos) staining for FOXA1/aPKC/E-CAD. Scale bar, 50 μm. Arrows, FOXA1+ VE/YE cells; white dotted lines, EPI. PE, primitive endoderm; EPI, epiblast; EXMC, extra-embryonic mesenchyme cell.

In mice and humans, polarized EPI cells initiate amnion cavity formation (12, 16). We postulated that EPI polarization would trigger amnion cavity formation in cynomolgus embryos. To test this idea, we analyzed E-cadherin (E-CAD) and atypical protein kinase C (aPKC) expression from day 11 to day 17 to trace EPI cell (OCT4+) polarization. At day 11, E-CAD and aPKC were apically localized in EPI cells, which formed a rosette-like structure lining a lumen at the center of the EPI (Fig. 3D). After expansion of the amnion cavity, the polarized localization of E-CAD and aPKC persisted in EPI cells from day 12 (Fig. 3E) to day 17 (fig. S1E), suggesting maintenance of polarization of EPI during development.

To investigate whether a similar mechanism underlines yolk sac formation in cynomolgus embryos, we examined E-CAD and aPKC expression in VE/YE cells (FOXA1+). E-CAD was not detected, and aPKC was not apically localized in the yolk sac, either at day 11 or day 17 (Fig. 3F). Together, these results suggest that the apical localization of the adherent junctions plays no roles in amnion and yolk sac cavity formation in cultured cynomolgus embryos.

WNT signaling during primitive streak induction

Ethical considerations regarding human embryos means primitive streak (PS) induction remains poorly understood. Cynomolgus embryos develop for up to 20 days in culture, suggesting that cynomolgus PS induction may be recapitulated in vitro. To test this idea, we analyzed BRACHYURY (also known as T) expression, a PS marker, from day 11 (24, 25). On day 11, T-expressing cells localized mainly to the dorsal amnion cavity, and OCT4 expression in these cells was weak. Around day 13, some of these cells were found underneath the EPI, and OCT4 expression was further reduced (Fig. 4A). Because T+-expressing EPI cells migrated into the space between the EPI and VE during gastrulation and the orientation of nuclei of migrating cells changed (26, 27), we observed that T+ cell nuclei outside EPI were progressively oriented parallel rather than perpendicular to the EPI layer [day 15 (Fig. 4B) and day 18 (fig. S2A)]. Approximately 63.83% (n = 30 nuclei) of nuclei of T+ cells (n = 47 cells), outside the EPI, were parallel to the EPI layer, whereas 79.30% (n = 157 cells) of T+ cells within the EPI (n = 198 cells) were perpendicular to EPI layer in total identified T+ cells (n = 245 cells) from day 11 to day 20 (Fig. 4C).

Fig. 4 Development of PS in cultured embryos.

(A) Representative IF images of embryo cryosections at day 11 (n = 4 embryos) and day 13 (n = 6 embryos) for OCT4 (red) and T (green) merged with DAPI (blue). Scale bar, 100 μm. (Bottom) Enlargements of inserts (yellow dotted lines) in top panels. Scale bar, 50 μm. Arrows, ThighOCT4low cells outside EPI. (B) Representative IF images of embryos at day 15 (n = 6 embryos) for E-CAD (gray), NANOG (red), and T (green) with merges with DAPI (blue). Scale bar, 100 μm. (Bottom) Enlargements of inserts (yellow dotted lines) in top panels. Scale bar, 50 μm. T+ cells outside EPI (arrowhead) is T+NANOGlow, and T+ cells within EPI (arrows) are TlowNANOG+. T+ cells outside EPI (arrowhead) show nucleus relatively parallel to the EPI layer (layer of NANOG positive cells). (C) Quantification of cells (n = 245 cells) with nuclei oriented parallel or perpendicular to the EPI layer in T+ cells outside (ThighOCT4low/ T+NANOGlow or T+NANOG-) or within (TlowOCT4high/ TlowNANOGhigh) EPI. (Left) Percentage of T+ cells outside (red bar, n = 47 cells, 19.18%) and within (blue bar, n = 198 cells, 80.82%) EPI. (Middle) Percentage of cells with nuclei oriented parallel (light blue bar, n = 30 cells, 63.83%) or perpendicular (orange bar, n = 17 cells, 36.17%) to the EPI layer in T+ cells outside EPI. (Right) Percentage of cells with nuclei oriented parallel (green bar, n = 41 cells, 20.70%) or perpendicular (yellow bar, n = 157 cells, 79.30%) to the EPI layer in T+ cells within EPI. (D) Violin plot of SNAI2, MIXL1, and OTX2 expression in EPI (EPI-A, EPI-B, and EPI-C) and gastrulating cells (Gast) identified in Fig. 6G. (E) Representative IF images of embryo cryosections at day 14 (n = 7 embryos) and day 17 (n = 6 embryos) for OCT4 (green) and OTX2 (gray) merged with DAPI (blue). Scale bar, 100 μm. (Bottom) Enlargements of inserts (yellow dotted lines) in top panels. Scale bar, 50 μm. White dotted lines, EPI; arrowheads, OTX2+ gastrulating cell. (F) Violin plot of WNT3, WNT3A, WNT5A, and WNT5B expression in EPI (EPI-A, EPI-B, and EPI-C) and gastrulating cells (Gast) identified in Fig. 6G. (G) Violin plot of WNT inhibitors expression in EPI (EPI-A, EPI-B, and EPI-C) and gastrulating cells (Gast) identified in Fig. 6G. CER1 is highly expressed in Gast (P < 0.05). (H) Representative IF images of embryo cryosections at day 13 (n = 6 embryos) and day 17 (n = 6 embryos) for OCT4 (green), WNT3 (red), and WNT5A (gray) merged with DAPI (blue). Scale bar, 100 μm. (Bottom) Enlargements of inserts (yellow dotted lines) in top panels. Scale bar, 50 μm. White dotted lines, EPI; arrowheads, gastrulating cell. EPI, epiblast; Gast, gastrulating cell.

PS induction is characterized by the epithelial-mesenchymal transition (EMT) (1, 26); we therefore analyzed expression of EMT maker genes, such as SNAI2 and MIXL1, in identified gastrulating cells, which were identified by using markers characterized in a previous report (13). SNAI2 and MIXL1 were significantly elevated in gastrulating cells (Fig. 4D). The PS marker, OTX2 (2830), was also expressed in gastrulating cells (located under the EPI layer) (Fig. 4, D and E), indicating that PS formation occurs in cultured cynomolgus embryos.

We asked whether the WNT signaling pathway is involved in PS induction because it is in mouse embryos and in human ESCs (29, 31, 32). High WNT3, WNT5A, WNT5B, and WNT8A expression levels (Fig. 4F and fig. S2B) were found in gastrulating cells by use of scRNA-seq, similar to human ESCs (29). WNT pathway repressors, such as the secreted proteins CER1, are required for mouse gastrulation (33) and induction of PS in human ESCs (29). scRNA-seq revealed that CER1 was highly expressed in gastrulating cells (Fig. 4G). IF staining of cryosections from embryos cultured in vitro at day 13 and day 17 revealed WNT3 and WNT5A expression in EPI and gastrulating cell, which was confirmed with scRNA-seq profiles (Fig. 4H). scATAC-seq analysis revealed open chromatin regions over gene regulatory regions of WNT signaling family members in gastrulating cells (fig. S5E). These results imply a putative role for WNT signaling in PS formation in cultured cynomolgus embryos.

Cynomolgus PGCLC specification in vitro

Primate primordial germ cell (PGC) specification occurs during the second or third week after fertilization (34), within the time frame of our embryo culture. SOX17+/CDX2+ cells were detected in amniotic epithelium at day 13 (Fig. 5B). Because SOX17 is a critical specifier of primate PGC (14, 35), we postulated that PGCLC could emerge in amnion. Expression of two primate-specific PGC markers (SOX17/TFAP2C) was detected from day 13 to day 17. SOX17+/TFAP2C+ cells were detected within the amnion, and SOX17 was expressed in cells within the amnion epithelium and beneath the EPI (gastrulating cells) at day 13 (Fig. 5A). From days 14 to 16, SOX17+/TFAP2C+ cells were detected among amniotic cells and at the junction region between amnion and EPI (Fig. 5A). At day 17, SOX17+/BLIMP1+ PGCLCs were found underneath the EPI (fig. S3). Expression of EOMES, another transcription factor required for PGCLC specification in humans PSCs (36, 37), was not detected in SOX17+/TFAP2C+ cells at day 12 or day 17, whereas EOMES+/SOX17+ cells were detected in gastrulating cells, which may represent precursors of the definitive endoderm (Fig. 5C) and is consistent with the amnion being a PGC source in vivo (14).

Fig. 5 PGCLCs originate from amnion.

(A) Representative IF images of embryos at days 13 (n = 6 embryos), 14 (n = 7 embryos), 16 (n = 7 embryos), and 17 (n = 6 embryos) stained for TFAP2C/SOX17/T. Scale bar, 100 μm. (Bottom) Enlargements of inserts (yellow dotted lines) in top panels. Scale bars, 50 μm. Arrows, PGCLCs within amnion; arrowheads, PGCLCs beneath the EPI; asterisk, SOX17/ TFAP2C cells. (B) Representative IF images of embryos at day 13 (n = 5 embryos) stained for CDX2/SOX17. (Bottom) Enlargements of inserts (yellow dotted lines) in top panels. Arrowhead, CDX2+/SOX17+ cell. Scale bars, 50 μm. (C) Representative IF images of embryos at days 12 (n = 7 embryos) and 17 (n = 6 embryos) stained for TFAP2C/EOMES/SOX17. Arrows, SOX17+EOMES+ cells; arrowheads, TFAP2C+SOX17+ cells. Scale bar, 100 μm. (Bottom) Enlargements of (yellow dotted lines) in the top panels. Scale bar, 50 μm. (D) Violin plot of PGC marker genes expressions in identified PGCLCs. (E) Violin plot of EMT marker genes [VIMENTIN (VIM), CDH1, SNAI1, SNAI2, and SNAI3] expressions in identified PGCLCs. PGCLC, primordial germ cell-like cell; EMT, epithelial-mesenchymal transition.

T can induce PGCLCs from germ-cell competent human PSCs (35, 38) or were induced when the PGC specified in vivo (14). T+/SOX17+/ TFAP2C+ cells were found in amnion from day 14 (Fig. 5A). From day 16, T was expressed in SOX17+/ TFAP2C+ cells and in the PS (SOX17/ TFAP2C cells) (Fig. 5A), which is consistent with data from in vivo derived embryos (14).

We examined the expression of PGC markers using scRNA-seq in putative PGCLCs (expressing SOX17, TFAP2C, and PRDM1 but not SOX2) (cells are identified in fig. S8A). The identified PGCLCs expressed PGC markers—including OCT4, NANOS3, SOX17, TFAP2C, and PRDM1 (Fig. 5D)—and also genes associated with the EMT, including VIMENTIN and CDH1 (Fig. 5E), which characterize early-stage PGCs in vivo (14), demonstrating the emergence of PGCLCs in cultured embryos, with a similar spatial location and molecular signatures as their in vivo counterparts.

Transcriptional regulation of lineage specification

To study lineage specification in the cultured cynomolgus embryos, scRNA-seq and scATAC-seq were performed on profile embryo transcriptomes at different developmental stages; 1014 and 1198 single cells from pre- and postimplantation embryos at seven time points (days 9, 11, 13, 15, 17, 19, and 20) were analyzed, respectively. After filtering, 600 and 978 cells were used for the final scRNA-seq and scATAC-seq analyses, respectively (table S1). scRNA-seq captured, on average, 4568 genes per cell (fig. S4A). Using t-distributed stochastic neighbor embedding (t-SNE) analysis, we identified four clusters of cells in vitro, the same cell types as identified in vivo (Fig. 6A) (13): EPI cells expressed high levels of pluripotency markers (OCT4 and NANOG); TE cells expressed KRT7 and GATA2; PE cells expressed GATA4, GATA6, and DPPA3; and EXMC cells also expressed GATA4 and GATA6 but not DPPA3 (Fig. 6B). t-SNE analysis showed that EPI, PE, TE, and EXMC of the in vitro cultured embryos were clustered with their in vivo counterparts (fig. S4, B and C) (13). EPI, PE, TE, and EXMC of the in vitro cultured embryos exhibited a close correlation with their in vivo counterparts, as did expression of key genes (fig. S4D). In vivo embryos at days 16, 18, and 20 were collected for low-input RNA-sequencing (Geo-seq) (table S1). t-SNE analysis revealed that EPI, PE, TE, and EXMC of in vitro cultured embryos clustered with those of their in vivo counterparts (Geo-seq) (fig. S4E). scATAC-seq analysis revealed six cell clusters, including EPI, gastrulating cells (named Gast a and Gast b), PE, EXMC, and TE (fig. S5, A and B), similar to the scRNA-seq analysis. Transcription factor motifs, such as OTX2 in gastrulating cells, were identified (fig. S5C). Furthermore, a clustered heatmap of differential chromatin accessibility in each cluster, at the proximal and distal sites of marker genes, revealed that a majority of distal sites–accessible genes were enriched in EPI and gastrulating cells, and proximal sites–accessible genes were enriched in other cell types (fig. S5D).

Fig. 6 Transcription landscape of monkey peri- and postimplantation embryos.

(A) t-distributed stochastic neighbor embedding (t-SNE) plot of cells at representative stages (days 9, 11, 13, 15, 17, 19, and 20). Cells were identified as epiblast (EPI), primitive endoderm (PE), trophectoderm (TE), and extraembryonic mesenchyme (EXMC) cells. (B) Expression of lineage-specific marker genes exhibited on t-SNE plots. A gradient of gray, yellow, and red indicates low to high expression. (C) t-SNE plot of TE cells at seven time points. (D) (Left) Pseudotime construction of single TE cells colored according to embryonic stage. (Right) Expression patterns of CDX2, TCEAL4, GCM1, GATA3, SOX2, and GATA4 exhibited on pseudotime construction. Dot-size gradient indicates low to high expression. (E) (Left) t-SNE plot of PE cells at seven time points, revealing two cell types, with (right) expressions of cell-type specific genes. (F) Gene Ontology (GO) term analysis: Cluster 1, enriched for lipid metabolism and transport; cluster 2, enriched for transcription and protein synthesis. (G) t-SNE plot of EPI cells at seven time points. Cells are designated as EPI-A, EPI-B, EPI-C, and gastrulating cell (Gast). (Inset) Single cells colored according to embryonic stage. (H) GO term analysis of classified EPI cells. (I) Violin plot of FGF/WNT/NOTCH signaling components expression in the various lineages.

We explored TE specification by examining transcriptional changes, revealing a continuous shift in the TE transcriptome from days 9 to 20, suggesting that TE cells differentiate in a stepwise fashion (Fig. 6C). On the basis of pseudotime analysis, TE cells were subdivided into two types accompanying the reduced expression of CDX2 from day 11 (Fig. 6D): Cells expressing high levels of TCEAL4 were likely cytotrophoblasts (CTs), and cells expressing high levels of GCM1 were likely syncytiotrophoblasts (STs) (11, 39). At day 11, yolk sac TE (ysTE) cells (11) were also found in the cultured cynomolgus embryos. These cells are positive for TE markers (such as GATA3) and express low levels of the EPI markers (such as SOX2) and PE markers (such as GATA4) (Fig. 6D).

The transcriptomic profiles delineate the developmental trajectory of the PE cells revealed two clusters of cells. Cluster 1 and cluster 2 expressed high levels of APOA2 and CXCR4, respectively, representing potential VE/YE differentiation (Fig. 6E and table S2). Gene Ontology (GO) term analysis revealed that cluster 1 was enriched for genes involved in lipid metabolism and transport, whereas cluster 2 was enriched for genes mediating transcription and protein synthesis (Fig. 6F). We also identified EXMCs, which are only found in higher primates (15). EXMCs exhibited similar marker gene expression to PE cells, suggesting that EXMCs might originate from the PE (fig. S6A). Furthermore, the GO terms analysis showed that pathways involved in extracellular matrix and EMT were enriched in EXMCs (fig. S6B and table S3). Transforming growth factor–β (TGFβ) signaling regulates the EMT. We found that TGFβ family members were highly expressed in EXMCs (fig. S6C), presumably suggesting that PE cells differentiate into EXMCs through the EMT, a process likely regulated by the TGFβ signaling pathway, which is consistent with their putative biological function in vivo (13, 24).

Amniotic epithelial cells form a thin membrane that contains the amniotic fluid, protecting the developing embryo or fetus (40, 41). There has been limited characterization of amniotic epithelial cells during postimplantation in primates. Monkey amniotic epithelial cells expressed the markers OCT4, CDX2, and TFAP2C but not NANOG (Figs. 2E and 5A). We therefore analyzed amniotic epithelia cell scRNA-seq profiles on the basis of the expression of these markers and determined the relationship between EPI, EXMC, PE (VE/YE), and amniotic epithelia cells. Amniotic epithelium cells exhibited a profile highly similar to that of EXMC (fig. S7A), suggesting a potential developmental relationship between these two types of cells. We identified differential-expressed genes between these four types of cells but did not detect significant gene expression differences between amniotic epithelium and EPI cells (fig. S7B), probably because of the small numbers (n = 12) of scRNA-seq profiles of amniotic epithelium cells and the similarity between amniotic epithelium and EPI cells. GO terms analysis showed that TGFβ signaling was enriched in amniotic epithelium cells (fig. S7C).

Examination of the pluripotency transition from the naïve to primed state in EPI cells during peri- to postimplantation stage revealed four clusters of EPI cells. Clusters 1 to 3 represented early- to late-stage EPI (EPI-A, EPI-B, and EPI-C, respectively), and cluster 4 represented gastrulating cells (Gast), because t-SNE analysis showed EPI-A and EPI-B were clustered with in vivo early-stage EPI cells, EPI-C was clustered with in vivo late-stage EPIs, and Gast was clustered with in vivo gastrulating cells (Fig. 6G and fig. S4C). We examined the expression of naïve and primed pluripotency-related genes and found that naïve genes (such as TBX3, KLF4, and KLF5) were highly expressed in early-stage EPI cells (EPI-A and EPI-B), whereas primed-specific genes (such as ID1, ZIC2, and FGFR1) gradually increased in later-stage EPI cells (EPI-C and Gast) (fig. S6D). GO term analysis revealed genes enriched in EPI-A were involved in oxidative phosphorylation and mitochondrial respiratory, whereas genes enriched in EPI-B and EPI-C were related to ribosome biogenesis and translation. Genes up-regulated in Gast were involved in embryonic morphogenesis, and mitochondrial oxidative metabolism was no longer enriched in these cells (Fig. 6H and table S4). Thus, naïve-state EPI cells favor the metabolic model of oxidative phosphorylation, as previously shown in PSCs (42, 43).

To explore signaling pathways potentially involved in lineage specification, we examined expression of the canonical signaling pathway. We observed that FGF (FGF2) and WNT ligands (WNT3A and WNT5B) were highly expressed in the EPI, whereas FGF (FGFR1 and FGFR2) and WNT receptors (FZD4 and FZD7) were enriched in PE and TE cells. IHH, a hedgehog pathway ligand, was highly expressed in PE cells (Fig. 6I). DLL3, a NOTCH pathway ligand, was highly expressed in EPI, and the receptor NOTCH2 was highly expressed in EXMC (Fig. 6I). Because PE (VE/YE), EXMC, and EPI may form a niche coordinating-lineage specification, we determined potential cellular communication interactions among PE (VE/YE), EXMC, EPI, and TE using public ligand-receptor databases (Materials and methods). EXMC and PE (VE/YE) frequently interacted with other cell types through ligand expression, and EPI cells mainly expressed receptors to receive the ligand from PE (VE/YE), EXMC, and TE (fig. S7D). For example, insulinlike growth factor 2 (IFG2) and granulin (GRN) were enriched in EXMC and PE (VE/YE), respectively. TGFBR3 was specifically expressed in EPI (fig. S7D). Overall, our study suggests the specific expression spectrum of ligands and receptors and corresponding interactions within PE (VE/YE), EXMC, EPI, and TE.

Transcriptional regulation of PGCLCs specification

To investigate potential transcriptional mechanisms underlying cynomolgus PGCLCs specification, PGCLC candidates (expressing SOX17, TFAP2C, and PRDM1/ BLIMP1 but not SOX2) (fig. S8A) (14) were compared with EPI cells because our data indicate that PGCLCs may originate from the amnion. Pseudotime analysis revealed two types of PGCLCs, one clustered with EPI cells (PGCLC-EPI) and one that did not (PGCLC) (fig. S8B). GO term analysis revealed that genes enriched in PGCLC-EPI cells were involved in ribosome biogenesis, whereas genes enriched in PGCLC cells were related to responses to fibroblast growth factor and cell-cell signaling by WNT (fig. S8C and table S5). These findings suggest heterogeneity in PGCLC population during development in vitro.

Genes differentially expressed between PGCLCs and the EPI included enrichment of the GO terms cell substrate adhesion and extracellular structure organization in PGCLCs (fig. S8D and table S6), implying that migration is a characteristic of PGCs, which is in agreement with expression of EMT-related genes in cynomolgus PGCs in vivo (14). We examined the correlation between PGCLCs and in vivo PGCs (early and late stage) (14). PGCLCs exhibited close correlation with in vivo early-stage PGCs (fig. S8E). Thus, PGCLCs specified in vitro exhibited a transcriptome similar to that of early in vivo PGCs.


Gastrulation in primates involves extensive remodeling of the embryo’s transcriptional landscape to facilitate the formation of the body plan (1, 7). However, information about postimplantation development in primates remains limited because of ethical considerations, technical limitations, and high research costs. In vitro culture systems have been developed to study early postimplantation development in humans (11, 12). Here, we modified the human embryo culture system for cynomolgus embryos, enabling them to develop in vitro for up to 20 days and recapitulate the specification of embryonic and extraembryonic lineages, the cavitation of amnion and yolk sac, PS formation, and PGC specification. The structure of the embryos, especially amnion and yolk sac cavity, varied among cultured embryos, which was not observed from in vivo derived embryos. This discrepancy can potentially be attributed to different microenvironments, such as mechanical and biochemical cues, between in vivo and in vitro. The structure of cultured embryos eventually collapsed around 20 days, which highlights the need to further improve culture parameters to enable more faithful and extended monkey embryo development in vitro.

Single-cell -omics analyses of the cultured cynomolgus embryos revealed molecular details of how the EPI, PE, TE, and EXMC lineage are specified. We found that cynomolgus PE cells differentiated into putative visceral and yolk sac endoderm and also EXMCs, during which PE cells undergo the EMT, regulated by the TGFβ signaling pathway. In mice, PE differentiate into parietal and visceral endoderm. Then the latter further differentiates into distal and anterior visceral endoderm (DVE/AVE) (10). Whether monkey PE cells follow a similar developmental path warrants further studies. We saw evidence of the metabolic state transition from early- to late-stage EPI, accompanying the switch from naïve to primed pluripotency. This is consistent with the notion that interactions between the metabolome and histone modifications drive the metabolic switch from naïve to primed pluripotency in ESCs (44). Our results support the notion that PGCs originate from the amnion in vivo (14). The transcriptional profiles of a small group of the PGCLCs were similar to that of the EPI, suggesting that these PGCLCs may also have been derived from the EPI.

One of the advantages of in vitro monkey embryo culture is that it provides an accessible platform to gain in-depth insights into the dynamic molecular and cellular changes during primate early development on the basis of the improvement of culture conditions that could support more homogenous and extended embryo development. Combining the embryo culture system with live-cell imaging, lineage tracing, signaling pathway perturbation through activators and inhibitors, and single-cell -omics analyses will help determine key factors and pathways underlying the specification and development of different lineages and thereby guiding the development of improved differentiation protocols from primate PSCs, including humans.

We have cultured primate embryos beyond day 14 and provide a comprehensive analysis of the transcriptional features of postimplantation embryos using scRNA-seq and scATAC-seq. This culture system provides a platform with which to dissect the mechanisms underlying primate embryonic development.

Materials and methods

Embryo in vitro culture

The monkey blastocysts were collected at 6 to 7 days post fertilization. Zona pellucida of blastocyst was removed by exposure to hyaluronidase from bovine testes about 30s. Zona-freed blastocysts were seeded on 8-well μ-plates (80826, Ibidi) with pre-equilibrated in vitro culture medium 1 (IVC1). After embryos attached to the well, half of the IVC1 was exchanged with in vitro culture medium 2 (IVC2). Thereafter, half of the medium were replaced daily with fresh IVC2, and monkey embryos were harvested at the indicated stage. Embryo culture was performed at 37°C in 5% CO2. All chemicals were from Sigma Chemicals unless otherwise stated.

IVC1: DMEM/F12 (11320-033, Thermo Fisher Scientific) supplemented with 20% (vol/vol) heat-inactivated fetal bovine serum (FBS) (35-076-CV, Corning), 2mM l-glutamine (25030; Thermo Fisher Scientific), Penicillin (25 units/ml)/streptomycin (25 μg/ml) (15070-063; Thermo Fisher Scientific), 1× ITS-X (51500-056; Thermo Fisher Scientific), 8 nM β-estradiol, 200 ng/ml progesterone and 25 μM N-acetyl-l-cysteine.

IVC2: 20% (vol/vol) heat-inactivated FBS of IVC1 was replaced with 30% (vol/vol) KnockOut Serum Replacement (10828010, Thermo Fisher Scientific).

Single cell collection and embryo laser capture microdissection

After washing with phosphate buffered saline (PBS) (MA0008, meilunbio), the embryo was digested with 0.1% trypsin (25200-072, Gibco) at 37°C for 5 to 10 min. Then the trypsin was neutralized with 2% FBS (04-002-1A, Biological Industries) and FBS was washed off with cold PBS with 0.1 to 1% bovine serum albumin (BSA). Finally, individual cells were picked into a lysis medium on ice with a mouth pipette. All procedures were performed under stereomicroscope.

The transcriptome of in vivo embryos was obtained according to Geo-seq method (4). Briefly, one monkey embryo at day 16, two at day 18 and two at day 20 were embedded in optimal cutting temperature compound (OCT; 4583, Tissue-Tek OCT, skura). Embryos were cryo-sectioned along the anterior-posterior axis. Populations of approximately 20 cells from putative EPI, TE, EXMC or VE/YE were collected by laser microdissection and processed for low-input RNA sequencing based on Smart-seq2 method.

IF analysis

Monkey in vivo and in vitro postimplantation embryos were harvested and fixed in 4% paraformaldehyde in PBS for 30 min at room temperature. Then embryos were washed with PBS, and then day 9 and day 10 embryos were stained directly. Other stages of embryos were dehydrated in sucrose (57-50-1, meilunbio) solutions each for 6 hours with increasing concentration from 15% to 30% (15%, 20% 30%; (w/vol). Next, embryos were embedded in OCT, frozen and stored at -80°C. Then the samples were prepared as cryosections with 8 μm thickness on pre-treated glass slides (1A5105, CITOTEST), and air-dried for 1 hour. After permeabilized in PBST (PBS with 0.3% Triton X-100) (0694, Amresco) for 30 min at room temperature, samples were blocked with 3% (w/v) BSA and 10% (w/vol) FBS (04-001-1ACS, Biological Industries) in PBS for overnight at 4°C. Slides were then incubated with primary antibodies overnight at 4°C. After wash, fluorescence-conjugated secondary antibodies, and 4',6-diamidino-2-phenylindole (DAPI) were incubated with the slides in dark at 25°C for 2 hours. Images were taken using Leica TCS SP8 confocal microscope (Leica). For antibodies used, see table S7.


Plate-based single-cell cDNA synthesis, amplification and library construction were performed as previously described (45, 46). Briefly, single cell was placed into the lysis buffer by mouth pipette. The reverse transcription reactions and pre-amplification were performed using SuperScript II (18064-071, Invitrogen), and KAPA HiFi HotStart ReadyMix (KK2602; KAPA biosystems,), respectively. For library construction, the cDNA was fragmented by Tn5 transposase (BGE005, BGI) mix at 55°C for 15 min. After that, 0.25% SDS was added to each well to stop the fragmentation, followed by polymerase chain reaction (PCR) amplification using KAPA HIFI Hotstart polymerase mix (KK2602, KAPA biosystems) and primer sets with 25 cycles of denaturation. Then the libraries were purified with AMPure XP magnetic beads (A63881, Beckman Coulter), and were circularized by incubating with T4 DNA ligase (EL0011, Fermentas) at 37°C for 1 hour. After removal of linear DNA, the libraries were then purified by PEG32 (MP07123, CG) and sequenced on BGISEQ-500 (BGI).


Plate-based single-cell transposition reactions were performed with transposase mixture at 37°C for 15 min with agitation at 300 rpm, following with a modified FAST-ATAC method (47). The release buffer was added and the reaction was maintained at 50°C for 15 min. Then plasmid DNA (30 ng) was added as the carrier DNA to the mixture. Afterwards, the DNA was purified with Ampure XP beads. Then the DNA was pre-amplified with NEBNext High-Fidelity 2× PCR Master Mix (M0541, New England Biolabs) and transposase adapters. The pre-amplified DNA was purified with Ampure XP beads and used for libraries construction as previously described (48). Briefly, DNA was amplified for 14 cycles using the NEBNext High-Fidelity 2× PCR Master Mix and barcode primers. Then the libraries were size-selected with Ampure XP beads for fragments between 150 and 700 base pairs (bp) and sequenced on BGISEQ-500.

scRNA-seq and scATAC-seq data pre-processing

For RNA-seq data, adapters and low quality reads (N rate > 0.2) were removed by Cutadapt (v1.15) (49). Then raw reads were mapped to Macaca fascicularis (Macaca_fascicularis_5.0) genome by STAR (v2.5.3) program (50). We calculated the transcripts per million mapped reads (TPM) as expression level using RSEM v1.3.0 with default parameters (51). The single cells with mapped reads >1 million and more than 2000 genes with TPM value >1 were used to further analysis.

For ATAC-seq data, the inherent sequence of Tn5 for each cell was removed using cutadapt (v1.16) (49) and aligned to the cynomolgus reference genome (Macaca_fascicularis_5.0) using bowtie2 (v2.2.5) with the parameter –X 2000 (52). SAMtools (v1.9) (53) was used to filter reads for alignment quality of >Q30 and Picard tools (v2.6.0) (Broad Institute; was used to remove duplicate reads. Cells with usable fragments under 10000 and promoter regions (500 bp around transcriptional start site) with ratio of fragments under 10% were filtered out. The fragments were aggregated and reference peak were called using Sambamba (v0.6.6) (54) and MACS2 (v2.1.2) (55) respectively. Then the number of fragments per reference peak per cell was counted using chromVAR (v1.4.0) (56) to construct the matrix of fragment counts in peaks X.

Cell clustering and differentially expressed gene analysis

Cell clustering and differential gene expression analysis were performed by Seurat (v2.3.4), a package in R (57). Principal components analysis (PCA) was performed to select principal components for clusters finding based on a jack straw method. Then graph-based clustering approaches were applied to identify cell clusters and t-distributed stochastic neighbor embedding (t-SNE) was used for visualization of distance between cells in the reduced 2D space.

t-SNE analysis of scATAC-seq data

chromVAR (56) were used to perform t-SNE analysis and Cicero (v1.0.11) (58) were used to define clusters in t-SNE embedding.

Transcription factor motifs analysis and differentially accessible sites identification

TF motif deviations were obtained from “deviationScores” function in chromVAR (56). To identify differentially accessible sites, we adopt the method in previously report (59). We randomly sampled 20 cells from each of the clusters identified above to generate the reference panel. Then we implement “binomialff” test using Monocle 2 package (60, 61) to get differentially accessible sites at a 1% FDR threshold (Benjamini-Hochberg method). We calculated specificity scores for differentially accessible sites based on Jensen-Shannon divergence using method in previous report (59). We set 0.02 for specificity score threshold to determine cluster specific accessible sites. To investigate chromatin accessibility of WNT signaling family members, we calculated “gene activity scores” to evaluate the chromatin accessibility degree of genes using Cicero (58).

Functional enrichment analysis

GO analysis of differentially expressed genes (DEGs) was performed using ClueGO (v2.5.2) (62, 63) as previously described (64).

Single cell pseudotime analysis

For scRNA-seq data, the pseudo-temporal analysis was performed using the R package Monocle2 (60, 61, 65).

Similarity analysis between in vivo and in vitro embryos

scRNA-seq dataset of our cultured and previous published in vivo embryos were aligned by using canonical correlation analysis (CCA) of Seurat R package (57). We selected the union of the top 1000 genes of highest variance for our and published dataset and calculated 20 canonical correlates (CCs) with diagonal CCA. The first 10 CCs were used for t-SNE visualization. Then, the correlation analysis was employed to detect the correspondence of cell subtype for in vivo and in vitro cultured embryo cells by using expression matrix of 200 high variable genes that contributed to the first 20 CCs. Also, scRNA-seq data of cultured embryos and Geo-seq data of in vivo embryos were aligned using CCA. To compare PGCLC with published in vivo early and late PGCs, the correlation analysis was performed with expression matrix of in vivo early and late PGCs DEGs.

Analysis of cross-talk between different cell types

CellphoneDB (v2.0.1) (66) was used to build signaling cross-talks and their directionality between different cell types. The R package Circlize (67) was used for the interaction visualization.

Supplementary Materials

Materials and Methods

Figs. S1 to S8

Tables S1 to S7

References (4574)

References and Notes

Acknowledgments: We thank Y. Hou and S. Liu for discussion of this study and L. Xu and Y. Wang for their technical assistance. We are grateful to K. Chen, X. Lin, and X. Liu for helpful feedback on the manuscript. Funding: This work was supported by the National Key Research and Development Program (2016YFA0101401), the National Natural Science Foundation of China (81760271), Major Science and Technology Projects of Yunnan Province (2017ZF028), Key Projects of Basic Research Program in Yunnan Province (2017FA010), and Frontier Research Program of Guangzhou Regenerative Medicine and Health Guangdong Laboratory (2018GZR110105013). Author contributions: Y.N., J.C.I.B., T.T., and W.J. designed the study and supervised overall experiments. Y.N., C.S., Y.K., N.S., and H.W. performed samples collection work, including superovulation, micromanipulation, and animal care. N.S., C.Li, C.Liu, J.W., L.L., W.S., E.Z., L.Z., Z.L., S.F., G.C., X.L., and G.P. carried out experiments or contributed critical reagents and protocols. Y.L., Z.H., J.W., X.D., J.C.I.B., and T.T. analyzed the data and performed statistical analyses. Y.N., J.W., J.C.I.B., W.J., and T.T. wrote the manuscript. Competing interests: The authors declare no competing financial interests. Data and materials availability: All sequencing data were deposited at the National Center for Biotechnology Information Sequence Read Archive under accession no. SRP175059. The data were also deposited at the China National GeneBank (CNGB) Nucleotide Sequence Archive (CNSA; under accession no. CNP0000231.

Stay Connected to Science

Navigate This Article