Epidermal growth factor induces a trophectoderm lineage transcriptome resembling that of human embryos during reconstruction of blastoids from extended pluripotent stem cells

Abstract Objectives This study aims to optimize the human extended pluripotent stem cell (EPSC) to trophectoderm (TE)‐like cell induction with addition of EGF and improve the quality of the reconstructing blastoids. Materials and Methods TE‐like cells were differentiated from human EPSCs. RNA‐seq data analysis was performed to compare with TE‐like cells from multiple human pluripotent stem cells (hPSCs) and embryos. A small‐scale compound selection was performed for optimizing the TE‐like cell induction and the efficiency was characterized using TE‐lineage markers expression by immunofluorescence stanning. Blastoids were generated by using the optimized TE‐like cells and the undifferentiated human EPSCs through three‐dimensional culture system. Single‐cell RNA sequencing was performed to investigate the lineage segregation of the optimized blastoids to human blastocysts. Results TE‐like cells derived from human EPSCs exhibited similar transcriptome with TE cells from embryos. Additionally, TE‐like cells from multiple naive hPSCs exhibited heterogeneous gene expression patterns and signalling pathways because of the incomplete silencing of naive‐specific genes and loss of imprinting. Furthermore, with the addition of EGF, TE‐like cells derived from human EPSCs enhanced the TE lineage‐related signalling pathways and exhibited more similar transcriptome to human embryos. Through resembling with undifferentiated human EPSCs, we elevated the quality and efficiency of reconstructing blastoids and separated more lineage cells with precise temporal and spatial expression, especially the PE lineage. Conclusion Addition of EGF enhanced TE lineage differentiation and human blastoids reconstruction. The optimized blastoids could be used as a blastocyst model for simulating early embryonic development.

elevated the quality and efficiency of reconstructing blastoids and separated more lineage cells with precise temporal and spatial expression, especially the PE lineage.
Conclusion: Addition of EGF enhanced TE lineage differentiation and human blastoids reconstruction. The optimized blastoids could be used as a blastocyst model for simulating early embryonic development.

| INTRODUCTION
Mammalian embryogenesis starts with a totipotent zygote that can develop into a blastocyst containing epiblast (EPI), primitive endoderm (PE) and trophectoderm (TE). 1 The establishment of blastocyst models would address the problem of studying natural embryonic development in vivo and the limited available embryos in vitro. Mouse blastocyst-like structures termed blastoids were formed using mouse pluripotent stem cells (mPSCs). 2,3 Subsequently, multiple approaches were designed to generate human blastoids using human pluripotent stem cells (hPSCs). [4][5][6][7][8] These human blastoids separated cell populations specifically expressing some of the EPI, PE and TE lineage makers and exhibited similar functions to human blastocysts. However, cell populations incorrectly expressing lineage markers were observed during reconstruction, as described in previous works, 4,8 suggesting that these human blastoids still cannot fully represent human blastocysts at the transcriptomic and epigenetic levels. Therefore, a functional blastocyst model must be established to accurately summarize the cell organization and lineage composition of natural human blastocysts.
The establishment of TE plays an important role during the process of generating human blastoids. In natural human embryonic development, TE appears in the blastocyst stage, mediating the interaction at the foetal-maternal interface, and subsequently differentiates into cytotrophoblasts (CTs), syncytiotrophoblasts (STs) and extravillous trophoblasts (EVTs) as implantation occurs. [9][10][11] Impaired trophectoderm is thought to cause various miscarriages, preeclampsia 12 and intrauterine growth restriction. 13,14 Similar to reconstructing human blastoids, using hPSCs is a convenient way to study TE lineage development. The primed hPSCs were considered to be unable to induce functional TE-like cells [15][16][17][18] due to the primed pluripotency representing the postimplantation epiblasts. 1,19 Recently, efforts have been made to develop approaches for resetting naive hPSCs, [20][21][22][23][24][25] which result in similar features to those of human preimplantation embryos at the transcriptomic and epigenetic levels. [26][27][28] TE-like cells derived from naive hPSCs exhibited comparable gene expression patterns to TE cells and successfully differentiated into TE lineage derivatives (CTs, EVTs and STs). 9,[29][30][31] However, high demethylation and loss of imprinting generally occur during naive resetting, 32 which may disturb embryogenesis and placental development. It is worth studying the feasibility of TE-like cell induction or reconstruction of human blastoids using naive hPSCs. Human EPSCs are able to form endo-or extraembryonic tissues and exhibit higher chimeric efficiency than naive and primed hPSCs. [33][34][35] In our previous work, we differentiated human EPSCs into TE-like cells with morphological and transcriptional features similar to those of TE cells. Through assembly of undifferentiated human EPSCs, human blastocyst-like blastoids were formed with three separated cell lineages and exhibited a transcriptome similar to that of human blastocysts. 8

| Establishment and culture conditions of human EPSCs
Human EPSCs were cultured in serum-free N2B27-LCDM medium under 5% CO 2 at 37 C and saturated humidity. The method and materials for the establishment and culture conditions of the human EPSCs were described previously. 8,33,36

| Gradient induction experiments with different concentrations of BMP4
The cells were cultured at 37 C, 5% CO 2 and saturated humidity.
Human EPSCs were differentiated with bone morphogenetic protein 4 (BMP4, Catalogue #314-BP-010, R&D Systems, CA) for several days. On the starting day, which we called Day 0, the cultured human EPSCs were digested into single cells by TrypLE Express (Catalogue #12604021, Gibco, CA), centrifuged and placed on 0.5% gelatin twice for 20 min each time to remove the feeder. Human EPSCs were seeded in plates pretreated with 1% Matrigel (Catalogue #354277, Corning, NY), which was diluted in DMEM/F12, and the walls were treated for at least 60 min in advance in BMP4 differentiation medium, which was described in our previous work. 8

| Different optimization schemes of BMP4 differentiation
The cells were cultured at 37 C, 5% CO 2 and saturated humidity.
The method of human EPSCs differentiated with BMP4 is as above.
In this experiment, the medium of Group BMP4 was composed of 25 ng/mL BMP4 differentiation medium. The medium of the BMP4 + EGF group was composed of 25 ng/mL BMP4 differentiation medium and 50 ng/mL human EGF (Catalogue #EGF L7, Peprotech, NJ). The medium of the BMP4 + VPA group was composed of 25 ng/mL BMP4 differentiation medium and 0.8 mM valproic acid (VPA, Catalogue #227-01071, Wako, Japan). The medium of the BMP4 + EGF + VPA group was composed of 25 ng/mL BMP4 differentiation medium, 50 ng/mL human EGF and 0.8 mM VPA. Components of the medium of Group HTSC were described previously. 11 The medium of the HTSC + BMP4 group was composed of HTSC medium with 25 ng/mL BMP4. The medium of Group HTSC À VPA was composed of HTSC medium without VPA. The medium of the HTSC -VPA + BMP4 group was composed of HTSC + BMP4 medium without VPA.

| RNA-seq library preparation and data analysis
Total RNA was isolated using TRIzol. Sequencing was performed on an Illumina X Ten sequencer with a 150 bp paired-end sequencing reaction.
The scRNA sequencing datasets of human embryos were downloaded from ArrayExpress E-MTAB-3929 38 and GSE109555. 39 29,31 ) and hTSCs (green 9,29,31 ). Hierarchical clustering was based on the genes with a fold change greater than 1.5 between all TE-like cells derived from naive and primed hPSCs TPM. (C) PCA of RNA-seq data from TE-like cells derived from human EPSCs, naive hPSCs, primed hPSCs and hTSCs derived from embryos and their derivatives (black 29 ). Each single symbol represents one sample in each cell line. (D) PCA of RNA-seq data compared with single-cell RNA-seq data from human early embryos in vivo. 38 Each single symbol represents one sample in each cell line or one cell in human early embryos at different embryonic days. (E) PCA of RNA-seq data compared with single-cell RNA-seq data from human early embryos in vitro. 39 Each single symbol represents one sample in each cell line or one cell in human early embryos at different embryonic days.
GSE167089, 9 GSE150616 30 and GSE138762. 31 All the bulk RNA-seq data analyses were performed with HISAT2 and Cufflinks using the UCSC human genome annotation (version hg19) with default settings for reads mapping and statistical analysis. The batch effects among the multiple datasets were removed using limma in R. Reads with unique genome location and genes with no less than 1 TPM in at least one sample were used for following analysis. Principal component analysis (PCA) and heatmap analysis were performed with the functions prcomp and pheatmap in R. TPM was normalized by log2 transformation, and the parameter scale was used in the pheatmap and pcromp functions.
The plot3D function was used to show the PCA results.
2.10 | Single-cell RNA-sequencing library preparation and data analysis Day 6 BMP4 + EGF blastoids were collected and prepared to establish an scRNA sequencing library, as described in our previous work. 8

| Statistical analysis
Statistical analyses were performed with GraphPad Prism 8 software using We converted human induced pluripotent stem cells (hiPSCs) into human EPSCs by an established protocol. 8,33,36 The obtained human EPSCs exhibited a dome shape ( Figure S1A), high alkaline phosphatase activity ( Figure S1B), a normal karyotype ( Figure S1C) and high expression of pluripotent cell markers ( Figure S1D). Teratoma assays verified the ability to differentiate into three germ layers ( Figure S1E).
According to previous work, 8

| Human blastoid construction is enhanced using TE-like cells derived from human EPSCs induced by EGF
To identify the function of TE-like cells derived from human EPSCs under different induction conditions, we carried out the reconstruction of blastoids according to our previous work. In the threedimensional culture system early on the third day, cell aggregates of the BMP4 and BMP4 + EGF groups formed small cavities ( Figure 4A).
On Day 6, the BMP4 and BMP4 + EGF groups formed obvious blastocyst-like structures, similar in shape to human blastocysts, including a dense inner cell mass (ICM), a cavity and circular surrounding cells ( Figure 4A). Few blastoids were generated from the HTSC + BMP4 and VPA groups with poor shapes ( Figure 4A). Blastoids generated from the BMP4 and BMP4 + EGF groups accurately expressed ICM markers (OCT4, SOX2), TE markers (GATA3, TFAP2C, CDX2, CK8, CK7) and PE maker (GATA6) through immunofluorescence staining and microscoping with or without the Z axis ( Figures 4B and S3A). However, the cell aggregates derived from other groups incompletely expressed lineage-specific genes ( Figure S3B). In comparison to natural human blastocysts, these human blastoids generated from the BMP4 and BMP4 + EGF groups exhibited similar shapes, including similar average diameters ( Figure 4C) and ICM ratios ( Figure 4D). In addition to EGF, the efficiency of blastoid generation was 1.5 times higher than that in the BMP4 group ( Figures 4E, S4A and S4B), suggesting that the addition of EGF strengthens the establishment of blastoids by optimizing the induction of TE-like cells.
To further assess the role of EGF in TE-like cell induction by BMP4, we analysed the DEGs between the TE-like cells induced by BMP4 and BMP4 + EGF ( Figure S4C). KEGG and GO term analyses showed that the BMP4 + EGF-TE-like cells further enhanced TE lineage-related signalling pathways, such as the ECM-receptor interaction, TGFβ signalling pathway and Hippo signalling pathway, compared with the BMP4-TE-like cells ( Figure S4D). Moreover, adding EGF enhanced the WNT and pluripotency of stem cell signalling pathways ( Figure S4D), which is consistent with previous reports. 11 Both BMP4 and BMP4 + EGF blastoids collapsed under extended culturing ( Figure S4E). Compared to the previous results of generating blastoids using mouse EPSCs independently, 62 we did not achieve blastocystlike structures using human EPSCs alone or TE-like cells induced with BMP4 treatment ( Figure S4F). To identify whether blastoids simulated postimplantation morphogenesis, we used the in vitro culture condition (IVC) of human embryos to extended culturing of BMP4 or BMP4 + EGF blastoids. The morphological characteristics of blastoids on Days 8, 10 and 12 were consistent with previous work, 8 exhibiting 31.1% and 40% adherence rates ( Figure S4G) and correct expression of ICM and TE lineage markers ( Figure S4H). In summary, these results indicate that the BMP4 + EGF induction system functionally optimizes the differentiation of TE-like cells and improves the efficiency of generating human blastoids.

| Single-cell transcriptome analysis of three lineages of blastoids
In our previous work, 8 Figure S5B). Next, we wondered whether the increased PE cell population of the BMP4 + EGF blastoids truly resembled the PE lineage.
The Pearson correlation coefficient showed that this PE cell population was close to PE cells from E7 38 and Day 6 63 human embryos ( Figure 5E) with similar PE lineage gene expression levels ( Figure 5F).
In comparisons of imprinting gene expression patterns in human three embryonic lineages, the expression levels of imprinting genes in the pre-implantation epiblasts (EPI) were higher than those in TE and PE, no matter at E5, E6 or E7 ( Figure 5G). We also found the similar expression patterns in the BMP4 + EGF blastoids but not in the BMP4 blastoids ( Figure 5G ) and pluripotent markers (OCT4, SOX2) by immunofluorescence staining. All images were taken at Day 6 during induction. Scale bars indicate 100 μm. (C) Diameter was quantified between human blastoids generated from the BMP4 and BMP4 + EGF groups and blastocysts. Data are the mean ± SD (n = 10 blastoids). *P < 0.05, **P < 0.01. (D) The ICM cell ratio was quantified between human blastoids generated from the BMP4 and BMP4 + EGF groups and blastocysts, data are mean ± SD (n = 10 blastoids). *P < 0.05, **P < 0.01. (E) Derivation efficiency was quantified between human blastoids generated from the BMP4 and BMP4 + EGF groups. Data are the mean ± SD (n = 9 times). *P < 0.05, **P < 0.01  In early pregnancy, EGF and EGFR are expressed in the trophoblast cells of the placenta and stimulate cell proliferation and hormone production. 66 The EGF/EGFR and VEGF/VEGF receptor (VEGFR) loops may play a major role among the autocrine and paracrine loops related to TE proliferation. 67 A previous work reported that EGF downregulated KISS1 expression by activating the PI3K/AKT signalling pathway to stimulate human TE cell invasion. 46 EGF is necessary for epithelial stem cell proliferation and cooperates with forskolin, an EPAC/RAP1 agonist, to enhance the formation of sac-like structures. 11 Additionally, EGFR is commonly used as a surface marker for labeling CTs and hTSCs. 31 Previously, some important regulatory genes have been found in mice, such as Cdx2 and Rif1. Knocking out Rif1 plays critical roles in the regulation of trophoblast cells in mice. 68 Notch and Hippo signalling pathways cross talk to activate Cdx2 expression in mouse preimplantation embryos and promote the specialization of trophectoderm. 69 Overexpressing Cdx2 converted haploid ESCs to TSCs, 70 suggesting Cdx2 influenced the pluripotency. There is a crossregulation of the Nanog and Cdx2 promoters during trophoblast differentiation and Cdx2 can represses Oct4 expression. 71 These indicated that Cdx2 and Rif1 are key regulators for specialization of TE lineage cells. To verify the function of these regulators in human TE lineage cells, it is necessary to establish a cell line resembling the TE in vivo.
However, the current method of differentiating human TE lineage cells still needs to be optimized. In this study, we found the expression of Cdx2 in the induced TE-like cells but not in the blastoids. In the previous work, Cdx2 is mentioned as a medium TE maker which is temporarily up-regulated during the preimplantation embryos, 72 We suspect Cdx2 initiates TE differentiation in the early stage, but fails to sustain in subsequent development.
In summary, by assessing TE-like cells derived from multiple hPSCs, we demonstrated that TE-like cells derived from human EPSCs