An aggregation of human embryonic and trophoblast stem cells reveals the role of trophectoderm on epiblast differentiation

Abstract The interactions between extra‐embryonic tissues and embryonic tissues are crucial to ensure proper early embryo development. However, the understanding of the crosstalk between the embryonic tissues and extra‐embryonic tissues is lacking, mainly due to ethical restrictions, difficulties in obtaining natural human embryos, and lack of appropriate in vitro models. Here by aggregating human embryonic stem cells (hESCs) with human trophoblast stem cells (hTSCs), we revealed the hESCs robustly self‐organized into a unique asymmetric structure which the primitive streak (PS) like cells exclusively distributed at the distal end to the TS‐compartment, and morphologically flattened cells, presumed to be the extra‐embryonic mesoderm cells (EXMC) like cells, were induced at the proximal end to hTSCs. Our study revealed two potential roles of extra‐embryonic trophectoderm in regulating the proper PS formation during gastrulation and EXMCs induction from the human epiblast.


| INTRODUCTION
In early mammalian embryo development, three lineages contribute to the formation of the embryo, including the embryonic epiblast (EPI) which will generate the future organism, and two extra-embryonic tissues, primitive endoderm (PE, or hypoblast) and trophectoderm (TE), which will form the yolk sac and placenta. These extraembryonic tissues not only are necessary for nutrition but also play crucial roles in regulating embryo development before and during gastrulation. 1 The previous studies on embryonic and extra-embryonic tissue crosstalk were almost exclusively conducted in mice, in terms of interactions between the TE and EPI, and it has been reported that the mice polar TE, parts of TE that are located nearest to the EPI, affect the EPI differentiation and proliferation by regulating the NODAL signal. The mouse polar TE expresses and secretes the convertases to cleave the precursor protein of NODAL expressed by EPI, thus forming the full functional NODAL, and causing increased proliferation of EPI. The activated NODAL also diffuses into the adjacent polar TE cells to maintain convertase expression and the Bone Morphogenetic Protein 4 (BMP) transcription. BMP4 from the polar TE then diffuses into the epiblast to induce transcription of the cofactors of NODAL and WNT, 2,3 and further activate the gastrulation genes such as Brachyury (or TBXT, or T). 4 However, due to ethical restrictions and lacking research samples, the interactions between embryonic tissues and extra-embryonic tissues in human embryos are still largely unknown. 5 Recently, the in vitro embryo-like structures constructed using the mouse or human stem cells offer us an unique opportunity to study the mammalian embryo development in vitro. 6 In mice, researchers have successfully created embryo-like models for the peri-implantation and postimplantation stages by combining mouse ESCs with extra-embryonic stem cells such as TSCs and extra-embryonic endoderm cells (XENs). 7,8 These models, known as ETX-or ETS-embryos, can replicate several key spatial and temporal events that occur during embryo development. By incorporating extra-embryonic cells, these in vitro models offer several advantages over ESCs alone, especially in the context of modelling embryogenesis during gastrulation. These models can capture important developmental events such as primordial germ cell (PGC) specification, PS formation, symmetry breaking, axis formation, and more. [7][8][9][10] Those findings confirmed the significant role of extra-embryonic tissues during early development.
Here, we generate a 3D model by aggregating hESCs with hTSCs (ET-aggregate) to investigate the crosstalk between human trophectoderm and EPI. We found the lumenogenesis, symmetry breaking, asymmetric differentiation of PS-like cells, and EXMC-like cell induction in the ES-compartments under the interaction with hTSCs. [11][12][13] 2 | MATERIALS AND METHODS

| Ethical considerations
The study was approved by the Research Ethics Committee (Research licence 2019SZZX-008) of Sixth Affiliated Hospital of Sun Yat-Sen University.

| Generation of ETAs
AggreWell 400 plate (STEMCELL Technologies, 34415) was prepared according to the manufacturer's protocol. Briefly, wells were rinsed with rinsing solution (Stem Cell Technologies, 07010), centrifuged for 5 min at 2000g and incubated with rinsing solution at room temperature for 20 min. After incubation, the wells were washed with 2 mL of Â1 Dulbecco's Phosphate Buffered Saline (DPBS), and 500 μL E8 medium with CEPT cocktail (Chroman 1, Emricasan, Polyamine, and trans-ISRIB) 19 was added to each well. The plate was spun for 5 min at 2000g, then placed at 37 C (5% CO 2 ) until use.
The hESC colonies were dissociated to single cells by incubation with Accutase (Invitrogen, A1110501) at 37 C for 8 min, and hESCs were pelleted by centrifugation for 3 min at 1200 rpm. Single cells were resuspended in E8 medium with CEPT cocktail and 14,400 hESCs per well were added in AggreWell plate. There was around 1.5 mL medium per well and 12 cells per microwell. The following day, hTSC colonies were dissociated to single cells by incubation with Try-pLE at 37 C for 8 min. hTSCs were pelleted by centrifugation for 4.5 min at 1100 rpm. After removing 500 μL E8 medium each well, hTSC single cells were resuspended in hTSC basal medium with 0.8 mM VPA, and CEPT cocktail and 80,000 hTSC single cells per well were added in AggreWell plate. All hESC and hTSC single cells were counted by a Luna Automated Fluorescence Cell Counter.
The AggreWell plate was then centrifuged for 3 min at 100g, and placed at 37 C under 5% CO 2 condition. The time when hTSCs were added into the plates was designated as 0 h, and aggregates formed after around 12 h of culture. For the first 48 h, hESC and hTSC aggregates were cultured at 37 C (5% CO 2 ) in mixed medium. After 48 h, the aggregates were transferred to 40% Matrigel (Corning, 354230) and cultured in medium 1 (APEL2 medium added both 3 μM CHIR99021 and 50 ng/mL EGF). After 12 h, replaced the medium 2 with CHIR99021 removed. The 72 h ETAs were changed to medium 3 (APEL2 added 3 μM CHIR99021, 5 ng/mL FGF2, and 50 ng/mL EGF) until the samples were collected at 96 h.

| Generation of ES-aggregates/TS-aggregates
Preparation of AggreWell 400 plate was same as above. hESCs or hTSCs will not be added into wells, but the same culture medium will be added. Other steps are the same as generation of ETA.

| Immunostaining
ETAs and ES-aggregates were fixed in 4% PFA for 30   F I G U R E 1 Legend on next page.

| Quality assessment and preprocessing of single-cell sequence data
For sequencing data generated by the 10Â Genomics platform, raw sequencing reads were aligned to GRCh38, and the count matrix was generated by using Cell Ranger (v6.1.1). The quality control for cells was carried out and cells were included if they met the following criteria: (1) the number of detected genes was between 1500 and 7500; (2) the percentage of mitochondrial gene sequences was <10%.

| Downstream analysis of single-cell RNA sequencing data
The following data analysis of single-cell RNA sequencing was mainly performed by using Seurat (v4.0.6). 20 Firstly, raw counts were normalized by using NormalizeData, and 2000 high variable genes (HVGs) were identified by using the FindVariableFeatures with default parameters. Then, dimensionality reduction was performed by using the combination of RunPCA and RunTSNE with dims = 1:10, followed the cluster identification by using FindNeighbors with dims = 1:10 and

| Identification of differentially expressed genes and GO enrichment analysis
The differentially expressed genes (DEGs) were identified by using the FindAllMarkers with the parameters logfc.threshold = 0.25. And GO enrichment analysis were performed by using the R package clus-terProfiler (v4.2.1). 22

| Inferring cell-cell communication with single-cell RNA sequencing data
The normalized expression data from scRNA-seq were used to infer the cell-cell communication by using CellChat (v1.6.1) 23 with default parameters and netAnalysis_signalingRole_heatmap was used to identify signals contributing most to outgoing or incoming signalling.

| Statistical analysis
Values were shown as mean ± SEM. Statistical parameters including statistical analysis and statistical significance reported in the figure legends and supplementary figure legends in Data S1 were obtained using t-test and ANOVA through GraphPad Prism8. Significance was defined as *P < 0.05; **P < 0.01; ***P < 0.001, ****P < 0.0001.

| Cavity formation in the ES-compartments of ETAs
The apical-basal polarization and lumenogenesis of epiblast were key morphogenetic events in early human embryo development.
At 24 h, immunostaining of F-actin showed that the EScompartment of ETAs was robustly self-organized into a lumenlike cyst structures with a tiny cavity, which gradually expanded during the following days (Figure 2A-C). Meanwhile, as a control, we cultured hESCs alone to generate ES-aggregates under the same condition as ETAs ( Figure 1A Besides, the basement membrane is important for lumenogenesis in ETAs. To determine the most appropriate Matrigel concentration for lumenogenesis, we detected the efficiency of lumenogenesis of the ETAs cultured in 10%, 40%, and 100% Matrigel ( Figure 2D). We found that the ETAs cultured in 40% Matrigel showed the best cavity formation efficiency of 92.7% ( Figure 2E) and the asymmetric structure of ES-compartment of 28.3% ( Figure 2F).

| Trophoblast stem cells induced asymmetric differentiation of embryonic stem cells in ETAs
WNT signalling was known to be important for symmetry breaking of human epiblast and was activated in epiblast during gastrulation. [24][25][26] To simulate the gastrulation, the ETAs were treated with 3 μM CHIR99021 (a WNT agonist) combined with 5 ng/mL (C0-C12) in the three structures at three different time points ( Figure 4B). The ES-derivates and TS-derivates were mainly separated into two different populations ( Figure 4C). The pluripotency markers SOX2 and POU5F1 were highly expressed in the ES-derivates, including clusters 0, 2, 3, 5, 8, 9, 11, and 12. The trophectoderm markers GATA3 and TFAP2A were highly F I G U R E 3 Legend on next page. expressed in the TS-derivates, including clusters 1, 4, 6, and 7 ( Figure 4D). Interestingly, compared with the ES-derivates in ES-aggregates, the clusters of the ES-derivates in ETAs were dramatically shifted ( Figure 4C), while the clusters of TS-derivates from both ETAs and TS-aggregates were generally overlapped ( Figure 4C), the dramatic changes in gene expression pattern were also not detected during the differentiation of hTSCs regardless of whether hESCs co-exist, which suggested an asymmetric interaction between hESCs and hTSCs, and the hTSCs differentiation was not obviously affected by co-culture with hESCs. Therefore, we will mainly focus on the ES-derivates herein.
Firstly, we identified two unique clusters C8 and C9 were specific and EOMES ( Figure 4I). Moreover, we observed a large population of cells in C11 also expressed these PS-and NM-specific markers ( Figure 4I), however, the C11 cells did not exhibit a high correlation with PS, NM, or other PS cell derivates ( Figure 4H). It is worth noting that the C11 was specific for ES-aggregates and this type of cells with confusing fate was completely eliminated from the ES-compartment of ETAs by co-differentiation with hTSCs ( Figure 4E). Collectively, we hypothesized a possible role for the extraembryonic trophectoderm in regulating the proper PS formation of the epiblast cells.
Next, we focused on the C8 which is a specific cluster for 72-96 h ETAs ( Figure 4E). The cells in the C8 demonstrated a much stronger interaction between the ES-compartment and the TS-compartment ( Figure 4G), which were negative for TBXT, and began to appear at 72 h and enriched at 96 h of ETAs, suggesting the C8 represents these undefined ETA-specific flattened cells ( Figure 3A) described above. To identify the C8, we compared C8 with the scRNA-seq dataset of CS 8-11 monkey embryo. 32 We found that C8 showed highest similarity to definitive endoderm (DE) and extra-embryonic mesoderm cells (EXMCs) ( Figure 4L). Considering the cells in C8 were derived from the pluripotent epiblast, and negative for the key endoderm markers GATA6, SOX17, and AFP ( Figure 4I), thus we speculated the C8 cluster as EXMC-like cells. To further characterize the C8, we compared the C8 with the scRNA-seq data from naive ES and ES-differentiated EXMCs. We found C8 was relatively similar to the early EXMCs, suggesting that hTSCs may facilitate the differentiation of epiblast into EXMCs ( Figure 4K). On the other hand, the origin of EXMCs in primates remains controversial and not well-established. It was first thought that the EXMCs in primate embryos were derived from the extraembryonic endoderm, while some recent works have indicated that the epiblast, not the endoderm, is responsible for EXMCs generation in both cynomolgus monkey embryos 29 and in vitro cultured hESCs, 33 or the EXMCs in primates have hypoblast and epiblast dual-origins. 32 Here we observed a type of cell with a flattened morphology that lined aside the contact surface between the hESCs and hTSCs in 72-96 h ETAs. These cells were differentiated from the hESCs and transcriptionally similar to the monkey EXMCs 32 and in vitro differentiated early human EXMCs. 33 Our results provided new evidence for the epiblast-origin of human EXMCs and raised a hypothesis that the EXMCs formation from human epiblast was induced by trophectoderm. However, further work is needed to validate the molecular regulatory mechanism of trophectoderm on EXMCs induction from the human epiblast.
In conclusion, the structure constructed in this study by aggregating hESCs and hTSCs could well simulate the asymmetric morphology and cell differentiation of human epiblast, and we provided a novel insight into the potential role of trophectoderm on the differentiation of epiblast to PS and EXMCs.