Complete human day 14 post-implantation embryo models from naive ES cells

Implantation triggers essential organizational changes in the human conceptus that enable gastrulation and subsequent development. These early post-implantation stages have a high incidence of pregnancy loss, and understanding them is important for fertility and developmental biology. Ethical and technical constraints limit direct study of implanted human embryos; existing ex vivo blastocyst-derived cultures fail to reproduce in utero events and structural organization. Integrated stem-cell-derived embryo models could bridge this gap, but must exhibit: (1) persistence of all key cell lineages (trophoblast-like, primitive endoderm-like, extra-embryonic mesoderm-like and pluripotent epiblast-like cells); (2) self-organization into fundamental embryonic compartments with correct morphology and orientation (embryonic disc-like, hypoblast-like, bilaminar disc-like, amnion-like, polarized yolk sac-like, chorionic cavity-like and surrounding trophoblast-like compartments); and (3) dynamic progression through morphologically defined milestones after aggregate formation. Building on prior work in mice showing naive ES cells can generate complete post-gastrulation SEMs, the study tests whether human naive pluripotent stem cells, without genetic modification, can self-assemble into fully integrated embryo models that advance through peri-implantation to peri-gastrulation hallmarks ex utero.

Prior work demonstrated mouse naive ES cells can be transiently directed toward extra-embryonic lineages via Gata4 and Cdx2 to form integrated SEMs advancing to early organogenesis. Human naive pluripotent culture conditions have improved (e.g., HENSM), and human naive cells exhibit accessible enhancers for GATA3/4/6, potentially enabling endogenous induction of extra-embryonic fates without transgenes. Previous human blastoid or attached embryo models lacked key defining lineages (e.g., trophoblast, visceral and parietal endoderm) and structural hallmarks (embryonic disc, bilaminar disc, yolk sac, chorionic cavity, surrounding trophoblast), and did not show dynamic progression. Human trophoblast stem cells derived in vitro are typically CDX2− and may represent later trophoblast stages than mouse TS cells, potentially limiting their integration into embryo-like aggregates. Collectively, these findings motivated a transgene-free, naive-ES-cell-only approach to generate human post-implantation SEMs.

Cell sources and priming: Human naive ES cells maintained in HENSM were used as the single starting source. For rapid derivation of primitive endoderm (PrE)-like and extra-embryonic mesoderm (ExEM)-like cells, a transgene-free induction protocol was established by modifying RACL conditions (omitting Activin A, termed RCL). Three-day RCL treatment yielded >50% PDGFRA+ cells from naive ES cells and generated both SOX17+ PrE-like and BST2+/FOXF1+ ExEM-like cells; N2B27 alone yielded fewer PDGFRA+ cells (~2.5-fold lower). For trophoblast (TE)-like induction, naive ES cells were treated with a BAP(J) regimen (BMP4 for 24 h, then A83-01/PD0325901 with JAK inhibitor for 48 h), producing TFAP2C+, CDX2+, TACSTD2+/ENPEP+ TE-like cells. Attempts using inducible GATA3 or CDX2 transgenes reduced viability or failed to form an enveloping trophoblast layer, so genetically unmodified cells were used henceforth. Aggregate formation: Three populations were aggregated: naive ES (HENSM), RCL-induced PrE/ExEM-like, and BAP(J)-induced TE-like cells. Aggregation used AggreWell and basal N2B27 with BSA to reduce stickiness. Optimal conditions: 120 cells per aggregate at a 1:1:3 ratio (naive:PrE/ExEM:TE). From day 3, aggregates were transferred to non-adherent 6-well plates on an orbital shaker. Culture medium (hEUCM2, adapted from mouse protocols) was used with gradually increasing FBS (20% day 4, 30% day 5, 50% day 6) to support structural organization and prevent TE-like attachment. Lineage omission tests: Systematic omission of each input validated lineage contributions: removing naive ES cells abolished epiblast formation; removing BAP(J)-induced TE-like cells eliminated the surrounding trophoblast compartment and prevented proper YS-like and bilaminar disc organization; omitting RCL-induced cells still allowed YS-like formation from naive-derived fractions but at reduced quality and efficiency. Phenotypic and functional assays: Immunofluorescence assessed lineage markers (e.g., OCT4/SOX2 for epiblast; SOX17/GATA6/GATA4/PDGFRA for hypoblast/YS; FOXF1/BST2/VIM for ExEM; CK7/SDC1/GATA3 for trophoblast; TFAP2A/ISL1 for amnion; T/BRA and CER1 for A-P axis; BLIMP1/PRDM1 for PGCs). Structural hallmarks (epiblast lumenogenesis, amniotic cavity, yolk sac polarity, chorionic cavity, connecting stalk, trophoblast syncytium and lacunae) were evaluated by confocal microscopy, 3D reconstructions, and morphological criteria. Secretion of hCG was assayed from culture supernatants. scRNA-seq (10x Chromium) profiled selected SEMs to define clusters and validate identities, with integration against an external human embryonic reference spanning zygote to CS7. Quantification of efficiencies across multiple batches and two ES lines (WIBR3 46XX; WIBR1 46XY) was performed per predefined morphological criteria.

• Transgene-free, human naive ES-cell-derived aggregates self-assembled into fully integrated SEMs containing epiblast-like, hypoblast/YS-like, ExEM-like, and enveloping trophoblast-like compartments, recapitulating key hallmarks up to ~13–14 d.p.f. (Carnegie stage 6a). • Efficient lineage induction from naive ES cells: RCL conditions produced >50% PDGFRA+ PrE/ExEM-like cells; BAP(J) generated TE-like cells with high TFAP2C/CDX2/TACSTD2 expression that uniformly surrounded aggregates; transgenes (iGATA3/iCDX2) were unnecessary and in some cases detrimental to viability/integration. • Correct spatial organization and dynamics: formation of an embryonic disc-like and bilaminar disc-like structure; epiblast polarization and lumenogenesis by day 6; dorsal squamous amnion-like and ventral pseudostratified epiblast-like layers; polarized YS-like compartment with VE-like (columnar) and PE-like (squamous) cells; ExEM-like mesenchymal meshwork forming a chorionic cavity-like space and a connecting stalk-like structure. • Axis formation and germ cell specification: anterior–posterior symmetry breaking with T/BRA expression opposite CER1+ AVE-like cells from day 6 (~1.02% efficiency for A–P axis detection); detection of PGC-like cells (OCT4+SOX17+BLIMP1/PRDM1+) in day 8 SEMs. • Trophoblast maturation: outer syncytiotrophoblast-like layer expressing SDC1, CK7, GATA3, with lacunae-like structures in ~90% of trophoblast-surrounded aggregates; HCGB protein localized in trophoblast and detectable soluble HCG in medium; multinucleated syncytium with microvilli-like structures. • scRNA-seq identified 13 clusters comprising epiblast-like (including posterior and committed sub-states), YS-like (including SYS-like), ExEM-like, amnion-like, syncytiotrophoblast-like, and a cytotrophoblast-like subcluster (within the amnion cluster by marker analysis). Projection onto a human embryonic reference confirmed post-implantation identities; some cells mapped near primitive streak. • Efficiencies: properly organized SEMs at day 6 occurred in 1.64% (WIBR3) and 1.09% (WIBR1) of starting aggregates; aggregates surrounded by trophoblast-like cells at day 6 were 48–74%; advanced day 8 SEMs with expanded lumina occurred at ~0.42% of starting aggregates; a reporter ES line yielded structured SEMs at ~0.08%.

The study demonstrates that human naive ES cells alone, without genetic modification or additional extra-embryonic cell lines, can generate fully integrated embryo models that recapitulate essential lineage composition, spatial compartmentalization, and morphological dynamics of early post-implantation human development. This addresses a critical gap caused by ethical and logistical constraints on in utero studies and surpasses prior in vitro models that lacked defining structures or dynamism. Human naive ES cells appear particularly poised to generate extra-embryonic fates via endogenous GATA3/4/6 activation under defined cues, enabling TE- and PrE/ExEM-like induction without transgenes, unlike mouse protocols. Functional lineage omission experiments show that a surrounding trophoblast-like layer is necessary for proper YS-like and bilaminar disc organization, highlighting cross-compartment interactions. scRNA-seq, while insufficient alone to define embryo models, corroborated microscopy-based evidence of appropriate cell identities and maturation trajectories (e.g., progression from YS-like to SYS-like, epiblast commitment and posteriorization, emergence of PGC-like cells). Collectively, these findings establish a platform to interrogate early human peri-implantation to peri-gastrulation events, including axis formation, amniogenesis, yolk sac remodeling, chorionic cavity and stalk formation, and trophoblast maturation and signaling.

This work establishes a transgene-free protocol to generate complete human post-implantation stem-cell-derived embryo models exclusively from naive ES cells. The models recapitulate the organization and dynamic progression of epiblast, hypoblast/YS, ExEM, and trophoblast compartments up to features characteristic of Carnegie stage 6a, including amniogenesis, A–P axis formation, PGC-like cells, chorionic cavity and connecting stalk, and syncytiotrophoblast lacunae with hCG production. The approach enables experimental access to previously inaccessible developmental windows and provides a basis for studying early pregnancy loss, congenital defects, and improving directed differentiation protocols. Future directions include: increasing efficiency and synchrony; testing whether spontaneous PrE/ExEM emergence from naive ES cells in basal conditions can replace RCL pretreatment; embedding SEMs in defined extracellular matrices to enhance trophoblast maturation; and pushing development beyond gastrulation to early organogenesis, potentially yielding authentic somatic cell types for regenerative applications.

Current SEM formation efficiency is low with notable asynchrony (up to ~2 days variance) and batch-to-batch variability, limiting throughput. Proper organization was not achieved from primed ES cells. Trophoblast survival during dissociation reduced scRNA-seq capture, hindering resolution of distinct cytotrophoblast clusters. Structural resemblance, while highly similar, is not identical to in utero embryos, and long-term development beyond the reported stages was not demonstrated here. Complete SEMs could not be formed from naive ES cells alone without TE-like induction, and blastoids did not progress under SEM culture conditions.