key: cord-0819680-hqmn33xe
authors: Tietze, Ethan; Barbosa, Andre Rocha; Euclydes, Veronica; Cho, Hyeon Jin; Lee, Yong Kyu; Feltrin, Arthur; van de Leemput, Joyce; Di Carlo, Pasquale; Sawada, Tomoyo; Benjamin, Kynon J.; Brentani, Helena; Kleinman, Joel E.; Hyde, Thomas M.; Weinberger, Daniel R.; Ursini, Gianluca; McKay, Ronald; Paquola, Apua C.M.; Shin, Joo Heon; Erwin, Jennifer A.
title: Single-cell analysis of human trophoblast stem cell specification reveals activation of fetal cytotrophoblast expression programs including coronavirus associated host factors and human endogenous retroviruses
date: 2020-08-29
journal: bioRxiv
DOI: 10.1101/2020.08.29.273425
sha: 00ac157bf3766f9f41607b43587ac2dd5eab05b2
doc_id: 819680
cord_uid: hqmn33xe

The human placenta is increasingly a focus of research related to early child development and the impact of maternal hyperimmune states. The ability to model human trophectoderm disease states from human pluripotent stem cells, the nature of human pluripotent stem cell potency and the mechanisms regulating human trophectoderm specification remains poorly understood. Recent work suggests that only the naive state can give rise to trophectoderm and that primed iPSC generate mixed amnionic and mesoderm lineages. Here we identify conditions that efficiently drive the specification of primed iPSC to trophectoderm, named Trophoblast Stem Cell (TSC). iPS-derived-TSC share transcriptional, morphological and functional characteristics with human in vivo cytotrophoblasts including activation of human endogenous retroviruses, expression of COVID-19 associated host factors and generation of multinucleated syncytiotrophoblasts with a large fusion index. At high densities in 5% O2, iPS-derived-TSC form villi-like structures and express extravillous and syncytiotrophoblast proteins HCG-β and HLA-G. Using temporal single cell RNAseq, we define the molecular changes associated with specification under three separate conditions: 1) BMP4, 2) BMP4 and inhibition of WNT, 3) activation of EGF and WNT, inhibition of TGFbeta, HDAC and ROCK signaling (named TSC). With 9,821 high-quality single cell transcriptomes, we find that BMP4 gives rise to mesenchymal cells while TS conditions lacking exogenous BMP4 generate a stable proliferating cell type that is highly similar to six week placenta cytotrophoblasts. TFAP2A programs the specification of primed iPS cells to TSC without transitioning through a naive state. TSC specification independent of exogenous BMP4 will allow for robust and reproducible studies of the cytotrophoblast component of human placenta.

gene expression, such as a primate-specific ERV which functions as a placenta-specific 79 enhancer for corticotropin-releasing hormone (CRH), a hormone linked to the control of birth 80 timing in humans (Blond et al., 2000; Dunn-Fletcher et al., 2018) . Therefore, human cellular 81 systems are crucial to understand the many human and primate-specific properties of human 82 placenta. 83 84

Human induced pluripotent stem cells are powerful models to understand causal genetic 85 mechanisms of human disease. For organs such as the brain, rapid progress has led to iPS 86 models with increasing complexity and utility (Mansour et al., 2020) . For the human placenta, 87 which is increasingly implicated as a risk factor for developmental disorders associated with a 88 maternal hyperimmune state, the ability to faithfully model human trophectoderm disease states 89 from human pluripotent stem cells is a matter of debate. Several model systems have been 90 developed to differentiate primed and naive iPSC to trophectoderm. Soon after the derivation of 91 the first human stem cells, Bone Morphogenetic Protein 4 (BMP4) was shown to differentiate 92 human stem cells to syncytial cytotrophoblasts (Xu et al., 2002) . Since then, various BMP4-93 based methods have been described including the BAP protocol (BMP4, A83-01 an alk-5 94

inhibitor and PD173074 and FGF and VEGF inhibitor) ( Understanding if and how induced pluripotent stem cells can generate human trophectoderm is 116 central to understanding human pluripotent stem cell potency and the mechanisms regulating 117 human trophectoderm specification. In the present study, we first ask if BMP4 is required for 118 primed pluripotent stem cell specification to trophectoderm. In the absence of BMP4, we find 119 that activation of wingless/Integrated (Wnt) and epidermal growth factor (EGF) while inhibiting 120

Transforming Growth Factor Beta (TGFb) histone deacetylase (HDAC) and Rho-associated 121 protein kinase (ROCK) efficiently specify multiple primed pluripotent stem cell lines to 122 trophectoderm. These factors were previously discovered to derive and grow trophoblast stem 123 cells from blastocysts and the early placenta (Okae et al., 2018) . We demonstrate that iPS-124 derived TS cells are capable of self-renewal for at least 30 passages, differentiate into syncytial 125 cytotrophoblasts and villus cytotrophoblasts and generate villi-like structures in low oxygen. We 126 used temporal single-cell sequencing analysis to elucidate the transcriptional landscape of this 127 specification compared to previously implemented protocols involving BMP4. We find that BMP4 128 conditions alone largely generate extraembryonic mesoderm cells. BMP4 combined with 129 inhibition of WNT shifts differentiation away from a mesodermal fate while upregulating 130 trophoblast associated lineage markers but cells are transcriptionally distinct from ex vivo 131

cytotrophoblasts. Trophoblast cells generated by TS conditions are transcriptionally highly 132 similar to ex vivo cytotrophoblasts and uniquely upregulate endogenous retroviral genes. 133

Specification of primed iPS cells to TS occurs without entering a naive state and TFAP2A and 134

ESRRG are central genes regulating the specification. Defining the molecular mechanisms 135 specifying the trophectoderm is key to understanding placental associated diseases and 136

improving iPS-derived models. 137 138

Results: 139 140

induced pluripotent stem cells without BMP4 142 143

In search of alternative methods to specify iPS cells to trophectoderm, we reasoned that primed 144 human pluripotent stem cells might be capable of specification to trophoblast stem cells directly 145 by activating wingless/Integrated (WNT) and epidermal growth factor (EGF) and inhibiting 146

Transforming Growth Factor Beta (TGFb) histone deacetylase (HDAC) and Rho-associated 147 protein kinase (ROCK Figure 1A and B). 48 hours post passage, cells were switched directly to TSC media. By 156 brightfield imaging over the subsequent 6 days, we observed that cells proliferated rapidly and 157 adopted a flatter appearance with some cells adopting a cobblestone appearance ( Figure 1A -158 D). After passaging, two morphologically distinct populations emerged. Circular colonies with an 159 epithelial-like appearance could be seen surrounded by phase bright fibroblastic cells ( Figure  160 1E,F). The inner epithelial-like cells continued to proliferate and appeared morphologically 161 similar to TS cells derived from primary human embryos ( Figure 1G ). Immunofluorescence 162 revealed that subpopulations of the colony cells expressed TP63 and/or KRT7 and minimally 163 expressed VIM while surrounding fibroblastic cells strongly expressed VIM ( Figure 1H ). Over 164 the course of several passages, the epithelial-like cells preferentially expand and reach purity 165 between passage 5-7. These TSC widely express both TP63 and KRT7 ( Figure 1I -K, 166

Supplemental Figure 1 ) without expressing the pluripotency genes SOX2 and NANOG 167 (Supplemental Figure 1 ). In summary, we find that activating wingless/Integrated (Wnt) and 168 epidermal growth factor (EGF) and inhibiting Transforming Growth Factor Beta (TGFb) histone 169 deacetylase (HDAC) and Rho-associated protein kinase (ROCK) directly in primed iPS lines 170 results in efficient specification of trophoblast stem cells. 171 172

Pluripotent stem cell lines vary in their propensity to differentiate to different cell types (Osafune 173 et al., 2008) . Therefore, we asked if TSC conditions without exogenous BMP4 could specify a 174 variety of primed human stem cell lines. We confirmed TS specification by VIM, KRT7 and TP63 175 immunofluorescence on iPS lines reprogrammed from dermal fibroblasts by sendai, named To confirm the ability to proliferate as TS-like cells, iPS-derived TS lines were passaged for up 179

to 32 passages and maintained KRT7 and TP63 expression ( Figure 1K ). 180 181

Hypoxia induces differentiation of iPS-derived TS cells into villi-like structures composed 182 of differentiated syncytial cytotrophoblasts and villus cytotrophoblasts 183 184

To investigate the maturation potential of iPS-derived-TSC, we evaluated the ability of iPS-185 derived TS cells to differentiate to mature trophectoderm cell types. Proliferative trophoblast 186 stem cells are bipotential stem cells with the capacity to differentiate to syncytial 187 cytotrophoblasts and extravillous cytotrophoblasts (Okae et al 2018) . Syncytial cytotrophoblast 188 is a terminally differentiated multinucleated epithelial layer that infiltrates the maternal 189 endometrium. The multinucleated cell forms from multiple cytotrophoblast cell fusions. When TS 190 cells were grown to high density, we observed the spontaneous formation of multinucleated 191 cells by brightfield microscopy. We confirmed that multinucleated syncytial cells are negative nuclei surrounded by the tight-junction protein ZO-1 positive cells (Figure 2A and B). 193 We observed large areas of syncytial formation fused nuclei (Figure 2A and S3). 194

Syncytiotrophoblasts produce and secrete high levels of human chorionic gonadotropin 195 (hCGB3), a hormone essential for cytotrophoblast differentiation, immuno-suppression, and 196 placental invasion (Cole, 2010) . We observe expression of hCGB3 in regions of the 197 differentiated cytotrophoblast cultures, indicating differentiation to syncytiotrophoblasts ( Figure  198 2C). Extravillus trophoblasts express human leukocyte antigen G (HLA-G), the main molecule 199 involved in maternal immune system evasion upon placental invasion of the endometrium. We 200 also readily detected regions of HLA-G positive cells, indicating iPS-derived TS cells are 201 capable of differentiation to extravillous cytotrophoblast ( Figure 2D ). Expression of hCGB3 and 202

HLA-G was specific to iPS-derived TS cells and not detected in iPS cells (Supplemental Figure  203 1I). 204 205

To further induce differentiation of the iPS-derived-TSC, we tested if cells will respond to altering 206 oxygen concentrations. Hypoxia is a natural cue that stimulates syncytial cytotrophoblast 207 differentiation in vivo (Wakeland et al., 2017) , therefore we transitioned a high density TS cell 208 culture from atmospheric oxygen of 20% to 5%. In 5% oxygen, these cells demonstrate minimal 209 contact inhibition and form 3D villi-like structures that protrude from the monolayer surface of 210 the culture dish (Figure 2 E-H). When imaged by confocal microscopy, these villi-like structures  211  are approximately 50 microns in height. In summary, iPS-derived-TSC readily differentiate into  212  both HLA-G positive villus cytotrophoblasts and hCGB3 positive syncytial cytotrophoblasts.  213  214  Single cell RNA sequencing reveals unique transcriptional programs for iPS-derived TS  215  specification  216  217 To define the molecular events involved in the specification of iPS cells to trophectoderm, we 218 performed temporal single cell RNAseq analysis of human iPS cells specified to placental 219 trophoblast cells by three separate conditions: 1) addition of BMP4 (Xu et al., 2002) , 2) addition 220 of BMP4 and inhibition of Wnt with the small molecule inhibitor IWP2 (Horii et al., 2019) , 3) 221

Trophoblast Stem Cell condition (TSC) activation of EGF and Wnt, inhibition of TGFbeta, HDAC 222

and ROCK signaling to generate trophoblast stem cells ( Figure 3A ). Sequencing was generated 223

for iPS cells before differentiation (day 0), for BMP4 on days 4 and 6, for BMP4+IWP2 on days 224 2, 4 and 6, and on days 2, 4, 6 and 8 for TS condition, resulting in 10 single-cell transcriptomes 225

( Figure 3A ). In TS condition, the day 8 is a timepoint two days after the first passaging 226 annotated as p1. In the BMP4 and BMP4 + IWP2 conditions, no proliferative cells survived 227 passaging, therefore the p1 time point was not sequenced. Using the highly parallel droplet 228 based single cell sequencing method Drop-Seq (Macosko et al., 2015) , 9,821 high-quality cells 229

were obtained after removing cells with less than 1000 genes detected and more than 20% of 230 mitochondrial mapping. The average ratio of mapping reads for mitochondrial genes in all cells 231

were 3.9%, indicating good viability (Supplemental Figure 4A ). The total number of genes 232 detected ranged between 21804 for BMP4 + IWP2 day 4 to 27134 for TS day 4 (Supplemental 233 Figure 4B ) time point with between 563 to 1572 high quality cells per timepoint (Supplemental 234 Figure 4C ). 235 236

In order to explore the data in an unbiased manner, we performed dimensionality reduction for Figure 3D , Table S1 ). 260 261

To understand how transcriptional profiles change during the differentiation, we asked which 262 cells express canonical pluripotent, trophectoderm and mesoderm specific genes. We find that 263 the stem cell transcription factor SOX2 is highest expressed in the iPS stage and is absent from 264 the most differentiated BR, TS and SC clusters ( Figure 3E ). SOX2 expression is maintained at a 265 higher level in the initial days of the TS conditions compared to BMP WNTi and BMP treatment 266 and absent from the most differentiated TS cells. The Paired-like homeodomain transcription 267 factor 2 PITX2 is highly expressed in the later stages of BMP4 differentiation, suggesting that 268 BMP4 conditions generate placental stromal cells ( Figure 3F ). In single cell RNAseq of the 269 human placenta, PITX2 was identified as the top marker gene for a stromal fetal communicating initiated after blastocyst formation and has variable expression patterns in trophectoderm 283 (Niakan and Eggan, 2013) . Similar to TS cells derived from primary human embryos, CDX2 is 284 not expressed in the differentiated TS state, and there is sparse expression of CDX2 at d4 of all 285 differentiations ( Figure S5A ). 286 287

Next we asked which genes are preferentially expressed in each cluster by non-parametric 288

Wilcoxon rank sum test (adj.p-value < 0.05; logFC > 0.25)( Figure 3I , 320 al., 2014). We identified cell type specific genes that were preferentially expressed in each cell 321 type/cluster of the 19 clusters from iPS-derived cells by the specificity index probability (pSI) 322 statistic at thresholds of p< 0.05. Next, we tested whether cell-type specific genes previously 323

identified by single cell studies of human placenta are over-represented in the cell-type specific 324 genes from our iPS specification by hypergeometric test and applied the Bonferroni correction 325 for multiple comparisons, considering all the tested gene lists [α = 0.05/ (19× (38+14)) = 5.1×10 -326 5 ]. Vento-Tormo identified a set of 38 cell types from maternal (decidua and blood) and fetal 327 tissues (placenta) of first trimester placenta and reported 30 cell-type specific genes per cell 328 type (Vento-Tormo et al., 2018). We found that the transcriptional profile from all iPS clusters 329 and clusters from d2-d4 timepoints did not have significant enrichment for any of the placental 330 cell type clusters ( Figure S6A ). In contrast, the most mature iPS-derived TS cluster was highly 331 enriched for the syncytial cytotrophoblast (SCT) and villous cytotrophoblast (VCT) cluster 332 (p=log10 -32 and log10 -17 ) ( Figure 4A ). The expression profile of the iPS-derived TS cluster was 333 highly specific for SCT and VCT with no significant enrichment for any other cell-type in the 334 maternal fetal interface. The D6 BMP4 cluster SC was most enriched in the two fibroblast 335 clusters (F1 and F2) (p=log10 -26 and p=log10 -13 ). Vento-Tormo described these placental 336 fibroblasts as mesenchymal stromal cells of fetal origin that derive from the primitive endoderm 337 expressing GATA4, GATA6, PDGFRa, and SOX17 ( Figure S6 and S7). The BMP4 + IWP2 338 conditions (BR) were not strongly enriched for any specific placental cell type and instead had 339

weaker enrichment for several clusters including syncytial cytotrophoblast (SCT), and maternal 340 decidua derived F1 endodermal cell type and dS1. In sum, iPS-derived TS cells are highly 341 enriched for expression of genes specific to syncytial cytotrophoblast and villous cytotrophoblast 342 of ex vivo human placenta. 343 344

Next we compared the expression profiles to early placental dataset from the 8 week and 24 345

week placenta (Liu et al., 2018) ( Figure 4B ). Again, we found strong enrichment for the iPS-346 derived TS cells in the eight week fusion competent CTB1 (cytotrophoblast, log10 -9 ). BMP4 alone 347

and Figure S7B ), we confirmed that TS cells expressed the highest level of cytotrophoblast 360

and EVT specific genes. In summary, iPS cells specified via TS conditions have a 361 transcriptional profile highly homologous to cytotrophoblasts taken directly from human placenta 362 while BMP4 differentiation generates placental mesenchymal cells. 363 364

Primed iPS specify to trophoblast by activation of TFAP2A without activation of naive 365 stem cell programs 366 367

It has been reported that primed stem cells are restricted in their potency and unable to 368 differentiate to TSC while naive stem cells readily differentiate to TSC because they are poised 369 to specify to trophectoderm (Dong et al., 2020) . Therefore, we asked if the primed iPS or TSCM 370 treated cells adopt a naive stem cell program during differentiation. For 12 previously identified 371 naive marker genes (Messmer et al., 2019), naive transcripts were nearly absent. Five genes 372 HORMAD1, ALPPL2, KHD3L, TRIM60, and HYAL4 had no observed expression in any cell 373 ( Figure 5A , To identify the most significant gene expression changes associated with the initial stages of TS 392 specification, we studied a subset of the sample using only cells from the iPS and TS day 2 393 samples and applied a single-cell trajectory inference and pseudotime estimation (STREAM) to 394 the single cell transcriptome data (Chen et al., 2019) . Briefly, single cells were ordered along 395 probabilistic trajectories and a numeric value referred to as pseudotime was assigned to each 396

cell to indicate how far it progresses along the dynamic differentiation. STREAM identifies an 397 initial branch composed of both iPS and TS day2 cells (S3-S0). This level of analysis also 398 identified branches containing IPS cells (S0-S2, S0-S4), reflecting the pluripotency continuum, 399

as well as a TS day 2 (S0-S1) branch reflecting the exit from pluripotency ( Figure 5C and Table  400 S5). On branch S3-S0, node S3 was predominantly composed of iPS cells while TS day 2 cells 401 are more abundant towards S0 ( Figure 5C , S8A,B). We next identified the transcriptional states 402 that contribute to the specification by identifying highly expressed genes within a branch that are were expressed not only along branch S3-S0 but also along branch S0-S1 ( Figure S8C ). The 411

branch S0-S1 captures the transition out of the pluripotent state and into the TS state with 412 upregulation of cytokeratins (KRT18, KRT19, KRT8) and calcium binding proteins (S100A11, 413 S100A10) known to be expressed in trophectoderm To understand the transcriptional changes associated with the full trajectory of iPS to 425 trophectoderm specification, we repeated the STREAM analysis with the complete iPS to d8 426 dataset. STREAM identifies 4 main branching points among all single cells ( Figure 6A ). The 427 early differentiation days for all conditions were mixed in pseudotime demonstrating a similar 428 transcriptional state at the initial phases of IPS commitment when compared in the context of d6 429 and d8 cells. Similar to the UMAP projection ( Figure 3B ,C), STREAM inferred mixing of two iPS 430 stages with cells from the d2 treatments. STREAM inferred three main branching points which 431 separated based on the most differentiated cells from each condition. To reveal the 432 transcriptional states that contribute to the branching, we carried out differential expression 433 analysis to identify highly expressed genes within a branch that are differentially expressed 434 compared to all other branches, termed leaf gene detection analysis in STREAM (see methods) 435

( Figure 6B , Table S6 ). For the zero IPS branch point, pluripotent-specific genes including 436

DNMT3A and POU5F1/OCT4 were highly differentially expressed. For the branch of cells 437 exposed to BMP4 (SC), mesoderm markers including PITX2 were identified as differentially 438

expressed. For the BMP+IWP2 branch, placental associated genes were identified including 439

RAB31, a small GTPases highly expressed in placenta that regulates intracellular membrane 440

trafficking of ligand bound EGFR (Chua and Tang, 2014) . For the TS branch, significantly 441 differentially expressed genes were associated with trophectoderm commitment and 442

implantation and preeclampsia including GATA3. We performed a Gene Ontology enrichment 443 analysis for biological processes using the top 300 significantly differentially expressed leaf 444 genes for the TS branch. Significantly overrepresented GO categories included placental 445 development, female pregnancy, viral processes (Figure 6 C,D). 446 447

Human endogenous retrovirus derived genes participate in a regulatory subnetwork 448 specifically within iPS-derived trophoblast stem cells 449 450

We found the set of genes involved in viral-like processes are upregulated in the later TS 451 differentiation timepoints ( Figure 6C ). This echoes an increasing body of evidence 452 demonstrating that endogenous retroviruses play a fundamental role in the evolutionary 453 diversification of the mammalian placenta (Johnson, 2019) . Therefore, we asked what is the 454 expression of human ERV-derived genes during trophectoderm specification? We found specific 455 upregulation of five ERV-derived genes in the most differentiated TS cell type ( Figure 7A ). 456 457

To understand the place of ERV-derived genes in the regulatory TS regulatory network, we 458

analyzed the participation of ERV-derived genes in a gene regulatory network. We built a 459 transcription factor and target gene network model using the Passing Attributes between 460

Networks for Data Assimilation (PANDA) algorithm to all clusters among the conditions. PANDA 461

integrates information from TF-sequence-motif data, gene expression and protein-protein 462 interaction (PPI) in a message-passing approach (Glass et al., 2013) . Briefly, PANDA's 463 algorithm makes the assumption that genes targeted by the same TF are more likely to be co-464 expressed than genes that are regulated by a different TF. Using a TF-sequence motif as initial 465 network model, the algorithm refines and updates the model by computing the congruence 466 between two measures: 1) Concordance in co-expression of genes presenting motif for a same 467 TF, and 2) Concordance in the regulatory profile of a set of TFs that interact with each other via 468

Protein-protein interaction data. Therefore, the likelihood of interaction between a TF and a 469 target gene is given by its edge weight, which states for the coordination of targets' expression 470

by a TF complex. PANDA outputs Z-score normalized edge weights, therefore positive and 471

negative weights can be seen as likely and unlikely interactions respectively. One network was 472 created for each previously identified cell cluster based on the expression profile of the top 100 473 cells in each cluster (based on PC1) with genes expressed in at least 10% cells. 474 We found that ERV interactions (Transcription Factors + ERV-derived genes) are prominent in 475 the most differentiated TS cluster and in no other cell clusters. We found regulatory interactions 476 for five ERV-derived genes (ERVH48-1, ERV3-1, ERVMER34-1, ERVW-1 and ERVFRD-1) are 477 highly unique to the TS cluster network, while only 0 or 1 ERV-derived gene was found in the 478 other cell clusters ( Figure 7B, S9A) . Clusters SC and SC 3 (BMP4 cells) exhibited interactions 479

for gene ERVH48-1, and clusters BR 1, BR 3 and BR 4 (BMP4+IWP2 cells) presented the gene 480

ERV3-1. Interestingly, genes encompassed in ERV regulatory interactions of iPS derived TS 481 cells were enriched in biological pathways related to hormone metabolism, cell differentiation 482 and the immune system (adj. p value < 0.01) ( Figure 7C , Table S7 ). In addition, we observed 483 ERV regulatory interactions are also enriched for categories such as placenta development, 484

trophectodermal cell differentiation and syncytium formation ( Figure 8A , Table S8)  519 and putative alternative entry receptors BSG, HSPA5 and DPP4 are more expressed. Other 520 potential entry proteases CTSB, CTSL and FURIN are all expressed at a higher level ( Figure  521 8B, Table S8 ). The placenta also protects the fetus from infection by the active transfer of 522 maternal antibodies to the fetus. Syncytiotrophoblasts transfer protective humoral maternal IgG 523

to the fetus via the neonatal Fc receptor encoded by the FCGRT gene (Leach et al., 1996) representation of the pseudotime with the clusters Figure S9 and C. Upset plot of shared genes within significant Gene Ontology enrichment terms generated 691 from the top 300 leaf genes for branch 1-3 (specification to TS). Gliogenesis p=7.92e-06, 692

Leukocyte activation involved in immune response p=1.72e-05, Reproductive system 693 development p=4.05e-05, Female pregnancy p=7.95e-05, Viral Process p=.0004, Blood Vessel 694

Morphogenesis p=.003, Placenta Development p=.008 695 D. Gene expression of CTSB along the pseudotime trajectory. CTSB is a significant leaf gene 696

for branch 1-3 that is annotated as associated with all GO processes listed in (6C). 697 698 Supplemental Tables  721  Table S1 : Number of cells from each condition/day that are assigned to the cluster 722 Table S2 : Differentially expressed genes in each cluster where a single cluster is compared to 723 all other clusters 724 Table S3 : Differentially expressed genes analyzed by pairwise comparisons of one cluster 725 compared to another cluster 726 Table S4 : Expression analysis of naive, primed and BMP signaling during differentiation 727 Table S5 : Stream pseudotime analysis of transition and differentially expressed leaf genes for 728 iPS to TSd2 729 Table S6: Stream pseudotime analysis of transition and differentially expressed leaf genes for  730  all cells  731  Table S7 : ERV gene regulatory network and associated GO enrichment. 732 PSC culture 738 6-well plates were coated with reduced growth factor Cultrex (1mg/12 ml DMEM/F12) at 37°C 739 for at least one hour. Human ES and iPS were maintained on coated plates in Stemflex media. 740

Media was changed every 48hrs in accordance with manufacturer recommendations. Cells 741

were passaged in small clusters using Versene solution and a split ratio of 1:10-1:12. For 742 cryopreservation, cells were suspended in PSC media and temporarily stored in a deep freezer 743 at −80°C before being transferred to liquid nitrogen for long term storage. 744 745

Differentiations and Cell Maintenance 746 6-well plates were coated with reduced growth factor Cultrex (1mg/12 ml DMEM/F12) at 37°C 747 for at least one hour. PSC were harvested in small clusters using Versene solution without 748

centrifugation, seeded at a passage ratio of 1:12, and cultured in 2 mL of Stemflex medium. 749

After 24 hours, cells were washed with DPBS and 2 mL of appropriate media was added to 750 each differentiating well. For TS differentiation, TS medium [DMEM/F12 with Glutamax 751 supplemented with 0.1 mM 2-mercaptoethanol, 0.2% FBS, 0.3% BSA, 1% ITS-X supplement, 752

1.5 μg/ml L-ascorbic acid, 50 ng/ml EGF, 2 μM CHIR99021, 0.5 μM A83-01, 1 μM SB431542, 753 0.8 mM VPA and 5 μM Y27632] was added to each well. For BMP4 based differentiations 10 754 ng/ml BMP4 and/or 2uM IWP2 was added to basal differentiation medium (DMEM/F12 with 755

Glutamax supplemented with ITS and L-ascorbic acid 2-phosphate magnesium) was added to 756 each well. Every 24 hours 2 mL of fresh medium was added to each well. Cells were collected 757 on day 2, 4, and 6 for RNA sequencing. For the TS condition, after 6 days of differentiation cells 758

were passaged using TrypLE express at a split ratio of 1:3 and plated on new Cultrex or col IV 759

(5 ug/mL) coated plates. After the first passage, cells were fed every 48 hours and split at a ratio 760 of 1:3- Preprocessing of Drop-seq data 828

Raw sequencing data was preprocessed using the pipeline "Drop-seq Alignment Cookbook" 829 v2.0.0 found at https://github.com/broadinstitute/Drop-seq/releases/ and described in (Macosko 830 et al., 2015) . Briefly, paired-end reads were filtered in order to remove read pairs with any base 831 with quality of less than 10 in both cellular and molecular barcodes. SMART adapters at 5' end 832 and polyA tails at 3' end with 6 or more bp were removed from the second pair, and then 833 aligned to the reference human (GRCh38) genome using HISAT2 v2. To profile the expression patterns of the 19 clusters detected in this work, we applied the Cell-894

Specific Expression Analysis (CSEA) as implemented in the pSI R package (Xu et al., 2014) . 895

This approach compares each profile with all the others and identifies genes expressed in one 896 group but not in the others, calculates a score (specificity index, SI) for each gene and attributes 897 a statistical level of significance (pSI). Here, we used the normalized expression matrix to 898 compute cluster-level median gene expression for each gene, which defined 19 cluster profiles. 899

Then, cluster profiles were used to compute SI and pSI for each gene, within each profile. 900

Cluster-specific gene lists were obtained applying the statistical threshold, pSI < 0.05. 901 902

Finally, we tested whether cell-specific genes from previous studies are over-represented in our 903 clusters. We used the hypergeometric test and applied the Bonferroni correction for multiple 904

comparisons, considering all the tested gene lists [α = 0.05/ (19× (38+14)) = 5.1×10 -5 ]. 905 906

PANDA regulatory networks 907

Gene regulatory networks were constructed using the pandaR v1.14.0 R package which 908

integrates multiple types of data to infer direct interactions (edges) between TFs and target 909

genes. PANDA initiates with a prior regulatory network that can be built by mapping TF binding 910 sites to the genome, and refines this initial network integrating gene expression data of target 911 genes and TF PPI data. The main idea behind the algorithm is that target genes from a given 912 TF are likely co-expressed, and TFs that interact with each other are more likely to regulate a 913 similar group of target genes. These two assumptions are used to infer the edge weight of each 914 TF-target interaction, and basically reflects the congruence between the regulatory profile of a 915 TF with target gene co-expression. Iteratively the algorithm refine the initial network structure 916 and infer a final consensus regulatory network. 917

A regulatory network was built for each cluster using as initial network a TF-motif binding map 918 described in (Sonawane et al., 2017) and downloaded in 919 (https://sites.google.com/a/channing.harvard.edu/kimberlyglass/tools/resources). Gene 920 expression data was library size normalized, scaled, and natural-log transformed. Genes 921 expressed in less than 10% of the cells of each cluster were filtered out. PPI network of TFs 922 was built using STRING database v11.0 (downloaded from https://string-db.org/) and score 923 interactions were divided by 1000 to initiate the PANDA. 924

Edges with negative weights estimated by PANDA were discarded in order to explore only 925

interactions with greater evidence. 

Assessment of the cellular localization of the 950 annexin A2/S100A10 complex in human placenta

Gene expression 953 and in situ protein profiling of candidate SARS-CoV-2 receptors in human airway epithelial cells 954 and lung tissue

The fetal origins of mental illness

Nephronectin plays critical roles in Sox2 expression 960 and proliferation in dental epithelial stem cells via EGF-like repeat domains

BRACHYURY and CDX2 mediate 963 BMP-induced differentiation of human and mouse pluripotent stem cells into embryonic and 964 extraembryonic lineages

ClueGO: a Cytoscape plug-in to decipher 967 functionally grouped gene ontology and pathway annotation networks

Genomewide screening for 970 fusogenic human endogenous retrovirus envelopes identifies syncytin 2, a gene conserved on 971 primate evolution

Defining the three cell lineages of the human 974 blastocyst by single-cell RNA-seq

Molecular characterization and placental expression of HERV-W, a new human 977 endogenous retrovirus family

An envelope glycoprotein of the human endogenous 980 retrovirus HERV-W is expressed in the human placenta and fuses cells expressing the type D 981 mammalian retrovirus receptor

Small molecule-mediated disruption of Wnt-dependent signaling in 984 tissue regeneration and cancer

Single-cell trajectories reconstruction, exploration 987 and mapping of omics data with STREAM

PRSS23 is essential for the Snail-dependent endothelial-to-mesenchymal transition during 990 valvulogenesis in zebrafish

Engagement of the small GTPase Rab31 protein and its 992 effector, early endosome antigen 1, is important for trafficking of the ligand-bound epidermal 993 growth factor receptor from the early to the late endosome

Biological functions of hCG and hCG-related molecules

Captured retroviral envelope 998 syncytin gene associated with the unique placental structure of higher ruminants

Derivation of trophoblast stem cells from naïve human 1002 pluripotent stem cells

Isolation of primitive endoderm, mesoderm, vascular endothelial and 1005 trophoblast progenitors from human pluripotent stem cells

Anthropoid primate-specific retroviral 1008 element THE1B controls expression of CRH in placenta and alters gestation length

Development of a 1011 versatile enrichment analysis tool reveals associations between the maternal brain and mental 1012 health disorders, including autism

Paracrine and epigenetic control of trophectoderm 1015 differentiation from human embryonic stem cells: the role of bone morphogenic protein 4 and 1016 histone deacetylases

An ex vivo model for imprinting: 1018 mutually exclusive binding of Cdx2 and Oct4 as a switch for imprinted and random X-1019 inactivation

Neuronal 1021 leucine-rich repeat protein-3 amplifies MAPK activation by epidermal growth factor through a 1022 carboxyl-terminal region containing endocytosis motifs

Establishment of porcine and 1025 human expanded potential stem cells

Passing messages 1027 between biological networks to refine predicted interactions

BMP4 regulates the hematopoietic stem cell niche

MEIS2C and MEIS2D promote tumor progression via Wnt/β-catenin and hippo/YAP signaling in 1032 hepatocellular carcinoma

The trophoblast is a component of the innate immune 1034 system during pregnancy

Normalization and variance stabilization of single-cell 1036 RNA-seq data using regularized negative binomial regression

PITX2 deficiency and associated human disease: 1039 insights from the zebrafish model

An Improved Two-Step 1041

Protocol for Trophoblast Differentiation of Human Pluripotent Stem Cells

Thymosin beta 4 silencing 1044 suppresses proliferation and invasion of non-small cell lung cancer cells by repressing Notch1 1045 activation

Origins and evolutionary consequences of ancient endogenous 1047 retroviruses

R-Spondin2 is a secreted activator of Wnt/beta-catenin signaling and is required for Xenopus 1050 myogenesis

HISAT: a fast spliced aligner with low 1052 memory requirements

GATA2/3-TFAP2A/C transcription factor network 1055 couples human pluripotent stem cell differentiation to trophectoderm with repression of 1056 pluripotency

Initial sequencing and analysis of the human genome

Isolation from human placenta of the IgG transporter, FcRn, and localization to the 1062 syncytiotrophoblast: implications for maternal-fetal antibody transport

Integrin α2 marks a niche of trophoblast progenitor cells in first trimester human placenta

Proteomic analysis on the alteration of protein expression in the placental villous 1069 tissue of early pregnancy loss

Single-cell RNA-seq reveals the diversity of trophoblast subtypes and 1072 patterns of differentiation in the human placenta

Establishment of human trophoblast stem cells from 1074 human induced pluripotent stem cell-derived cystic cells under micromesh culture

Thymosin beta4 induces angiogenesis through 1077 Notch signaling in endothelial cells

Highly Parallel Genome-wide Expression 1080 Profiling of Individual Cells Using Nanoliter Droplets

Cellular complexity in brain organoids: 1082 Current progress and unsolved issues

Transcriptional Heterogeneity in Naive and Primed Human Pluripotent 1085 Stem Cells at Single-Cell Resolution

Interferon-α 1087 inducible protein 6 impairs EGFR activation by CD81 and inhibits hepatitis C virus infection

Two distinct trophectoderm lineage stem cells from human pluripotent 1091 stem cells

Initial sequencing and comparative analysis of the mouse genome

Placental adaptive responses and fetal programming

Analysis of human embryos from zygote to blastocyst 1099 reveals distinct gene expression patterns relative to the mouse

Interaction between Oct3/4 and Cdx2 determines trophectoderm differentiation

Synthesis and characterization of insulin-like growth factor-binding protein (IGFBP)-7

Recombinant human mac25 protein specifically binds IGF-I and -II

Derivation of human trophoblast stem cells

Marked differences in differentiation propensity among 1112 human embryonic stem cell lines

Pregnancy-associated plasma protein-A2 (PAPP-A2), a novel insulin-like growth factor-1115 binding protein-5 proteinase

Mechanisms of trophoblast-virus interaction

Single-Cell RNA-Seq Reveals Lineage 1120 and X Chromosome Dynamics in Human Preimplantation Embryos

Does the human placenta express the canonical cell entry mediators 1123 for SARS

YAP is essential for tissue tension to ensure 1126 vertebrate 3D body shape

Specification of trophoblast 1128 from embryonic stem cells exposed to BMP4

Genetic control of early cell lineages in the mammalian embryo

Generation of four postmortem dura-derived 1133 iPS cell lines from four control individuals with genotypic and brain-region-specific transcriptomic 1134 data available through the BrainSEQ consortium

Tempting fate: BMP signals for cardiac 1136 morphogenesis

Cytoscape: a software environment for integrated 1139 models of biomolecular interaction networks

A single-cell RNA expression map of human 1141 coronavirus entry factors

Understanding Tissue-Specific Gene 1144 Regulation

Integrated 1146 analysis of single-cell embryo data yields a unified transcriptome signature for the human pre-1147 implantation epiblast

Comprehensive Integration of Single-Cell 1150 Data

A single-cell survey of the human first-trimester placenta 1153 and decidua

Embryonic stem cell lines derived from human blastocysts

Early cell fate decisions of human embryonic stem 1159 cells and mouse epiblast stem cells are controlled by the same signalling pathways

Single-cell reconstruction of the 1163 early maternal-fetal interface in humans

Transplacental transmission of SARS-CoV-2 infection

Hypoxia Directs Human Extravillous Trophoblast 1169 Differentiation in a Hypoxia-Inducible Factor-Dependent Manner

MUC15 inhibits dimerization of EGFR and PI3K-AKT signaling and is 1172 associated with aggressive hepatocellular carcinomas in patients

Dynamic stem cell states: 1175 naive to primed pluripotency in rodents and humans

BMP4 initiates human embryonic stem cell differentiation to trophoblast

Cell type-specific 1180 expression analysis to identify putative cellular mechanisms for neurogenetic disorders

Comparison of syncytiotrophoblast generated from human embryonic 1184 stem cells and from term placentas

Human cardiovascular progenitor 1187 cells develop from a KDR+ embryonic-stem-cell-derived population

Short-term BMP-4 treatment initiates mesoderm induction in human embryonic stem 1190 cells