Establishment of a method to culture human embryos through implantation stages in vitro

To gain understanding of the developmental events undertaken by the implanting human embryo, we reasoned that we needed an in vitro culture system that would recapitulate these processes. To this end, we adapted the culture conditions we had previously established for mouse embryos at comparable stages7. Supernumerary pre-implantation human embryos (at either cleavage or blastocyst stage) donated for this project were thawed and placed in culture medium to recover, before the zona pellucida was removed (Methods). Embryos were scored by morphological criteria and those showing abnormalities (fragmented blastomeres or developmentally arrested) were discarded. We first plated human blastocysts on optical-grade dishes in 21% O 2 and in in vitro culture medium 1 (IVC1), which was replaced by in vitro culture medium 2 (IVC2) after 48 h (Fig. 1a, b)7. Time-lapse microscopy revealed that human blastocysts cultured in IVC1 underwent a series of contractions and expansions reflecting collapse and subsequent enlargement of the blastocyst cavity (Supplementary Video 1), as observed in IVF clinics10. However, by the end of day 7, the blastocyst cavity completely collapsed and embryos attached to the dish (Fig. 1b–d). At this time the attached embryos started to grow and this growth continued until day 12–13 (Fig. 1c, d and Supplementary Video 2). We previously found that either human cord serum or KnockOut Serum Replacement (KSR) in IVC2 permits peri-implantation mouse development11. Here we show human embryos develop in medium supplemented with KSR up to day 13 (Fig. 1b–d). The experiment was stopped at this point because the internationally recognized ethical limit for human embryo culture is up to day 14 or to the first signs of development of the primitive streak12.

Figure 1: Establishment of an in vitro system to study human implantation and early post-implantation morphogenesis. (a) Human embryos were thawed and cultured until the blastocyst stage (day 5–6 of development). The zona pellucida was removed and embryos were transferred to plates in IVC1 medium for imaging. (b–d) On the second day of culture, medium was changed for IVC2 with 30% KnockOut Serum Replacement (KSR). Shown are representative bright-field images of human blastocysts developing in vitro until day 12–13. All scale bars, 100 μm. These data involved the assessment of a total of 5 embryos collected across 3 experiments, out of which 3 showed correct development. Full size image

The trend in IVF clinics over recent years has been to culture pre-implantation human embryos in hypoxic conditions (5% O 2 ), as this favours embryo survival13 and inhibits differentiation in embryonic stem cells14,15. To investigate whether hypoxic conditions offer any advantage to human embryos cultured through implantation, we compared development of embryos cultured in 21% O 2 to ones cultured in 5% O 2 in terms of preservation of the pluripotent lineage. Initially, day 5–6 pre-implantation blastocysts had a group of inside pluripotent cells as indicated by the presence of the transcription factor OCT4 (Fig. 2a, b). However, after 4 days in culture, whereas 49% of embryos (N = 59) cultured in 21% O 2 maintained a cluster of epiblast cells showing expression of the pluripotency factor OCT4, none of the embryos cultured in hypoxic conditions (N = 20) had OCT4-expressing cells (Supplementary Table 1). In contrast, cells of embryos in hypoxic conditions showed signs of cell death (Fig. 2c). This suggests that preservation of the epiblast in vitro beyond day 7 might benefit from higher O 2 concentrations possibly because of the increase in embryo size, which may decrease the O 2 pressure in the core of the embryo. Hypoxia clearly has a complex influence on embryonic development and its effects need to be better understood in the intrauterine milieu to determine exactly how it affects both mouse and human embryogenesis16. As a result of these findings, here we cultured human embryos beyond day 7 in 21% O 2 , which we show permits further development of the pluripotent lineage in vitro.

Figure 2: Preservation of the pluripotent lineage in human embryos cultured through implantation stages in vitro. (a) Bright-field images of day 5–6 blastocysts. Asterisks indicate the inner cell mass. (b) Day 5–6 blastocysts were fixed and immunostained for lineage markers. Representative confocal Z sections of human blastocysts stained for OCT4 and GATA6, and OCT4 and CK7. (c) Human embryos cultured in either 21% or 5% O 2 were analysed at the indicated time points. Representative confocal Z sections of human embryos stained for OCT4, aPKC and F-actin. Note the absence of OCT4-expressing cells and the presence of fragmented nuclei in the embryos cultured in 5% O 2 . All scale bars, 50 μm. These data involved the assessment of a total of 59 embryos in 21% O 2 (out of which 29 embryos preserved the epiblast) and 20 embryos in 5% O 2 (out of which none preserved the epiblast) collected across 3 experiments. Full size image

Development of the embryonic and extra-embryonic lineages during human post-implantation morphogenesis in vitro

To gain temporal understanding of the cellular and developmental events through early post-implantation development, we fixed and immunostained human embryos at different time points of culture. We used KnockOut Serum Replacement to supplement the medium as it permits a more reproducible time course of development than human cord serum7. We have focused on development between days 7 and 11, because during these 5 days the embryo undergoes a major reorganization that is necessary for subsequent gastrulation9. Shortly after attachment (day 7–8 of development), we observed a clear separation between embryonic (OCT4-expressing, epiblast) and extra-embryonic (GATA6-expressing, hypoblast) cells, the respective progenitors for the fetus and yolk sac (Fig. 3a), whereas in the unattached blastocysts, this segregation of epiblast and hypoblast progenitors had not yet taken place (Fig. 2b). This contrasts with mouse embryos, where the epiblast and the hypoblast-equivalent (the primitive endoderm) segregate already before implantation17; but agrees with observations in human embryos18,19,20,21,22. We found that at day 7–8, the epiblast was formed by a cluster of about 20 OCT4-expressing cells surrounded by approximately 50 GATA6-expressing hypoblast cells (Fig. 3a and Supplementary Table 2). By day 8–9, GATA6-expressing hypoblast cells preferentially localized to one side of the epiblast, as occurs in human embryos developing in vivo9. This spatial organization was maintained throughout subsequent developmental stages (from day 9 to day 11; Fig. 3a and Supplementary Fig. 1a and Supplementary Videos 3 and 4). By day 11, the number of OCT4-expressing epiblast cells had significantly increased to an average of 328 cells, whereas the number of GATA6-expressing hypoblast cells increased up to 79 (Supplementary Table 2).

Figure 3: Analysis of embryonic and extra-embryonic lineages in human embryos cultured through implantation stages in vitro. Human embryos developing in vitro until day 11 were fixed and stained at the indicated time points. (a) Representative confocal Z sections of human embryos stained for OCT4 and GATA6. Right panels show the centre of all OCT4-expressing (white) and GATA6-expressing (green) cells. All positive cells were counted regardless of the fluorescence intensity value. (b) Representative confocal Z section of a day 9–10 embryo stained for OCT4, GATA6 and CK7. (c) Representative confocal Z section of a day 8–9 embryo stained for PAR6 and CK7. (d) Representative confocal Z section of human embryos stained for F-actin and DAPI. Arrowheads point to multinucleated cells. (e) 3D reconstruction of the cellular and nuclear shape of representative trophectoderm cells. Note that cells in close proximity to the epiblast have a single nucleus, whereas cells in the periphery of the embryo are multinucleated. All yellow rectangles indicate the regions in the embryos that are shown with higher magnification. All scale bars, 50 μm. These data involved the assessment of a total of 59 embryos collected across 6 experiments, out of which 29 showed preservation of the embryonic lineage. Full size image

To analyse development of trophectoderm, the progenitors of the placenta, we followed cytokeratin 7 (CK7), which is highly expressed in trophoblast derivatives23. We found that the trophectoderm cells surrounding the epiblast and hypoblast (Fig. 3b) presented a polarized epithelial phenotype, evident from the localization of the apical determinant PAR6 (Fig. 3c). From day 8 onwards, we noted the presence of CK7-expressing multinucleated cells in the outermost region of the embryo (Fig. 3d). To determine their exact position with respect to the epiblast–hypoblast cluster, we performed a three-dimensional (3D) reconstruction of cellular shapes, based on the computational segmentation of membranes and nuclei (Fig. 3e). This revealed two trophectoderm subpopulations: cells in proximity to the epiblast–hypoblast bilayer had a single nucleus, whereas those in the periphery of the embryo were multinucleated and exhibited characteristic lacunae (Fig. 3e and Supplementary Video 5). On the basis of the expression of CK7, their position and their cellular and nuclear shape, these two cell populations are likely to correspond to the cytotrophoblast and syncytiotrophoblast respectively, indicating that the development of post-implantation trophoblast derivatives can be recapitulated in the absence of any maternal tissue. Together these results demonstrate that the development of all three lineages progresses as human embryos develop in vitro.

Human epiblast polarization and pro-amniotic cavity formation

Next, we focused on the development of the epiblast because it undergoes its first major reorganization at implantation. One of the hallmarks of epiblast transformation is the acquisition of a polarized phenotype and the formation of a lumen, the prospective pro-amniotic cavity9,24, essential for the subsequent development of the body plan. However, the cellular mechanisms underlying lumen formation in humans remain unknown. Whereas some reports in higher primates point towards cell death within the epiblast as the main driver of lumen formation25, alternative reports suggest folding of the epiblast26, or a change of polarity in epiblast cells4. Importantly, these hypotheses are based on the observations of rhesus monkey embryos by electron microscopy as, so far, it has not been possible to definitively identify cell types and their features in human embryos at implantation. We found that at day 7–8 of development, the epiblast was a cluster of cells showing no signs of polarization (Fig. 4a and Supplementary Figs 1b and 2). However, by day 8–9, a group of OCT4-expressing epiblast cells became radially organized around a small central lumen. These cells were apico-basally polarized in 31% of the embryos (Supplementary Table 1), as determined by the polarized apical localization of actin and aPKC (the principal kinase of the apical Par polarity complex; Fig. 4a and Supplementary Fig. 1b). The formation of a small lumen at the exact site of incipient apical polarization indicates the onset of pro-amniotic cavity formation within the epiblast. Importantly, we did not observe any apoptotic cell, cellular debris or any other indication of cell death in the incipient pro-amniotic cavity (Supplementary Fig. 3). These results indicate that the epiblast becomes polarized by day 9 of in vitro development and suggest that apoptosis is not the mechanism driving lumenogenesis in human embryos.

Figure 4: Remodelling of the epiblast in human embryos cultured through implantation stages. Human embryos developing in vitro until day 11 were fixed and stained at the indicated time points. (a) Representative confocal Z sections of human embryos stained for F-actin, aPKC and OCT4. Arrowheads indicate the incipient pro-amniotic cavity. (b) 3D reconstruction of the pro-amniotic cavity (shown in red). The nuclei of OCT4-expressing epiblast cells are shown in grey. (c) 3D reconstruction of the cellular shape of representative OCT4-expressing epiblast cells. (d) 3D reconstruction of the prospective yolk sac (shown in blue). The nuclei of OCT4-expressing epiblast cells are shown in grey. (e) Day 10–11 embryo stained for F-actin, OCT4, GATA6 and aPKC. Note the presence of GATA6-expressing cells on both sides of the cavity (arrowheads). The prospective yolk sac is indicated with a dashed line. (f) Model of human embryo implantation morphogenesis based on our results and the Carnegie series. The main remodelling events that take place during this transition are: segregation of epiblast and hypoblast progenitors (day 7); polarization and pro-amniotic cavity formation in the epiblast (day 8–10); differentiation of the trophectoderm (TE) into cytotrophoblast and syncytiotrophoblast cells (day 8–10); formation of the prospective amniotic epithelium, the prospective yolk sac and the bilaminar disc (day 10–11). All rectangles indicate the regions in the embryos that are shown with higher magnification. All scale bars, 50 μm. These data involved the assessment of a total of 59 embryos collected across 6 experiments, out of which 9 showed pro-amniotic cavity formation. Full size image

Human bilaminar disc formation

The next expected remodelling event is the formation of the amniotic epithelium and the epiblast disc. In contrast to the mouse, in which the amnion is formed at gastrulation, the amnion is expected to form in humans shortly after implantation with those epiblast cells adjacent to the hypoblast acquiring a columnar shape, and those in contact with cytotrophoblast differentiating into squamous and flat amniotic epithelium. At this stage, both tissues line the pro-amniotic cavity24. We observed that as development of embryos progressed in vitro, the epiblast and the pro-amniotic cavity expanded in size (Fig. 4a, b and Supplementary Fig. 4 and Supplementary Table 2 and Supplementary Video 6), indicative of embryo growth. At day 10–11 of development, we could detect two morphologically distinct groups of OCT4-expressing cells. 3D reconstructions of cellular shapes revealed that OCT4-expressing cells located near the hypoblast had a columnar morphology whereas opposite-facing OCT4-expressing cells had a distinct squamous and flat shape (Fig. 4c and Supplementary Video 7). Moreover, these flat OCT4-expressing cells were in close proximity to cells with a higher nuclear volume (on average 2.5 times bigger than epiblast cells), a feature characteristic of trophoblast cells. These observations suggest that the two types of OCT4-expressing cells may represent the epiblast disc and the prospective amniotic epithelium.

Human prospective yolk sac formation

The next expected event of human embryo development is the reorganization of hypoblast cells to give rise to a second cavity, the primary yolk sac9,27, which will provide the blood supply to the developing fetus. Remarkably, we found that human embryos developing in vitro established a second cavity at day 11 of development (Supplementary Video 8). This cavity was localized below the epiblast and was surrounded by GATA6-expressing cells, highlighting its hypoblast origin (Fig. 4d, e and Supplementary Video 9). Thus, both expression of characteristic lineage markers as well as morphological features suggest that formation of the prospective yolk sac can be also recapitulated in human embryos developing in vitro (Fig. 4f).

Human pluripotent stem cells (hPSCs) recapitulate the events of polarization and lumenogenesis

As the above results revealed that epiblast polarization and lumen formation are the first features of epiblast morphogenesis after implantation, we next investigated the molecular pathways that could be responsible for this critical remodelling. Given the ethical restrictions and the limited number of human embryos available for functional studies, we turned to human embryonic stem cells (hESCs) and human induced pluripotent stem cells (hiPSCs) as alternative models. First, to determine whether the extracellular matrix (ECM) participates in inducing epiblast polarization and lumen formation, we cultured individual hESCs in 3D Matrigel. We found that surrounding cells with ECM enabled lumen formation by 24 h after cell plating, and so it was sufficient to have two sister cells from a single cell division for this to take place (Fig. 5a), in agreement with other recent observations28. As the hESCs continued to divide, they self-organized around a central lumen. These hESC cysts were apico-basally polarized, as demonstrated by the polarized localization of aPKC (Fig. 5a), PAR6 (Fig. 5b), centrosomes (Fig. 5c) and the Golgi (Fig. 5d). They retained expression of pluripotency markers such as OCT4 (Fig. 5a), mimicking the preservation of OCT4 in the epiblast of in vitro-cultured human embryos (Fig. 3). To confirm these results, we next plated hiPSCs into 3D Matrigel as we did with hESCs. We found that hiPSCs also polarized and began to form lumens between two cells, while retaining OCT4 expression (Fig. 6a). Lumens formed on the apical side of polarized cells that exhibited apical microvilli and a polarized distribution of aPKC, PAR6 and the Golgi (Fig. 6a–c). This polarized organization was preserved in the presence of the ROCK inhibitor Y-27632 (Fig. 6d), commonly used to avoid cell death at the time of plating of hPSCs. We could also detect lumen-directed secretion of the anti-adhesive protein podocalyxin (PODXL), implicated in lumen formation through charge repulsion29,30 (Fig. 6e, f). The observation that hPSC cysts develop lumens between two cells as they become polarized suggests that lumen formation is not a consequence of cell death. To test this possibility directly, we treated hiPSCs with a caspase 3 inhibitor to prevent cell death8 and found that polarization and lumen formation were not affected (Fig. 6g, h). Taken together, our results indicate that lumen formation is a consequence of cellular polarization and not cell death and lead us to suggest that similar molecular pathways may drive lumenogenesis in human embryos soon after implantation.

Figure 5: Self-organization of hESCs in response to ECM signalling. hESCs were plated in a 3D matrix of Matrigel and analysed at the indicated time points. (a) hESCs stained for OCT4 as a marker of pluripotency, aPKC as a marker of apical polarization and F-actin to reveal the general organization of the cells. (b) hESCs stained for PAR-6 and F-actin. The size of the lumen was quantified. Data are shown as average ± s.e.m. Note the increase in lumen size with time (n = 10 hESC cysts per time point). Unpaired two-tailed Student’s t test with Welch’s correction, ∗∗P = 0.0049. (c) hESCs stained for the centrosome marker γ-tubulin and F-actin. Arrowheads indicate the polarized apical localization of the centrosome. The nucleus centrosome angle with respect to the nucleus–nucleus axis is shown. Each dot represents an individual centrosome. Data are shown as average ± s.e.m. (n = 19 centrosomes). (d) hESCs stained for the Golgi marker GM130 and F-actin. Arrowheads indicate the polarized apical localization of the Golgi. All scale bars, 10 μm. These data involved the assessment of a minimum of 10 hESC cysts per panel, all of which showed the indicated phenotype. Images are representative of a minimum of 2 independent experiments per panel. Full size image