Notch coordinates self-organization of germ layers and axial polarity in sea anemone gastruloids Authors: Sanjay Narayanaswamy, Franziska Haas, Emmanuel Haillot, Alison G Cole, Elly M Tanaka, Ulrich Technau Published: 2026 ABSTRACT AbstractReaggregated early gastrula cells (“gastruloids”) of the sea anemone Nematostella vectensis are able to regenerate into whole polyps within a few days. However, the cellular and molecular mechanisms underlying restoration of body axis and germ layers remains largely unknown. Here, we show that mesodermal cells sort to the periphery of the gastruloid, where they form cell clusters with hitchhiking endodermal cells. One of these clusters immigrates at one pole, forming the future pharynx and inner layer. This morphogenetic behavior that enables the symmetry break of the organizer tissue, requires a Wnt-Notch signaling feedback loop. This highlights a hitherto unknown role of Notch signaling in self-organizing gastruloids. Conservation of Notch-mediated boundary formation between germ layers mirrors similar mechanisms in bilaterians. This demonstrates how adoption of ancestral regulatory networks enables a morphospace converging to similar body plans, thus contributing to evolutionary robustness.Subject terms: Differentiation, Embryogenesis Introduction The establishment of body axes governs morphological diversity across 600 million years of animal radiation. A key step preceding body axes formation is tissue symmetry breaking. In many model organisms, such as C. elegans, frogs and fruit flies, this event already occurs prior to or during fertilization, and therefore, there are limited tractable systems to study spontaneous symmetry breaking in a tissue1. Additionally, while conserved molecular feedback loops like those between Wnt/β-catenin signaling and T-box transcription factors underpin axial patterning2,3, their interplay with species-specific embryonic geometries and extraembryonic cues remains poorly understood. This knowledge gap stems from inherent limitations in studying native embryos, where maternal determinants and biomechanical constraints obscure cell-autonomous patterning principles1. Ex vivo and in vitro self-organizing systems, particularly gastruloids, have emerged as excellent systems to dissect these mechanisms. By recapitulating axis formation through self-organization rather than preprogrammed embryonic architecture, these models enable us to isolate intrinsic modes of self-organization. For instance, human and mouse gastruloids can autonomously pattern their body axes, mirroring the embryo, but without any inputs from the visceral endoderm or trophoblast. This highlights cell-autonomous symmetry breaking dependent on Wnt/Nodal signaling interactions rather than by external BMP gradients4. Additionally, unlike in embryos, gastruloids pattern in minimal biochemical and mechanical constraints, allowing us to uncover conserved developmental models across species by removing species-specific signaling and geometric constraints3. This explains why, despite having radically different embryonic geometries, in vitro systems often converge on similar axial elongation dynamics, suggesting an adaptation of conserved circuits rather than de novo innovations5. Finally, these systems allow for fine-tuning various experimental parameters such as cell proportions, helping us study aspects like dependence of symmetry breaking time on proportion of T+ cell populations6, something harder to test in vivo. Both in native embryos and gastruloids, symmetry breaking and axis elongation are driven by the presence of localized signaling centers or organizers. The concept of the organizer, first established by Spemann and Mangold in amphibians7, has since emerged as a unifying principle in developmental biology, with functionally similar regions identified across metazoans. This includes cnidarians, where earlier seminal transplantation experiments have led to the identification of the Hydra hypostome organizer8–11, and more recently to the blastopore lip organizer in Nematostella12. Comparative studies revealed that the molecular logic underlying embryonic organizer function in Nematostella is deeply conserved across Metazoa. The Wnt/β-catenin signaling pathway, central to organizer activity in Nematostella as well as other cnidarians12–18, also governs primary axis specification in bilaterians, where organizers such as the amphibian Spemann organizer and the mammalian node direct embryonic patterning through localized Wnt signaling19. In Nematostella, transplantation experiments and targeted activation of Wnt signaling demonstrate that co-expression of Wnt1 and Wnt3 can confer organizer properties to ectopic regions, paralleling the axis-inducing capacity of organizer tissues in bilaterians. This highlights the deep evolutionary conservation of Wnt/β-catenin signaling in axial patterning, predating the split between cnidarians and bilaterians12,20. Cnidarians also possess remarkable regenerative capacities, with many organisms in this phylum possessing the ability to reestablish their body plan from dissociated and reaggregated cells from embryonic stages through to adult polyps, thus exhibiting a form of extreme regeneration10,13,21–23. While this phenomenon has been observed in various cnidarian polyps, the lack of accessible embryonic stages in most cnidarians has limited comparisons to early embryonic self-organization in bilaterians. In this regard, the anthozoan Nematostella vectensis stands out as a tractable model: its embryonic cells, when dissociated and reaggregated, can reestablish its primary body axis and germ layer topology, ultimately giving rise to a viable polyp indistinguishable from those developed through unperturbed embryogenesis. This plasticity is orchestrated by the Wnt1 and Wnt3 expressing organizer cells at the blastopore lip, whose presence is both necessary and sufficient for axis reestablishment in aggregates, mirroring the role of these organizers in normal development12,24. In addition, cells of the inner layer of the gastrulae become mesenchymal post-dissociation, followed by ingression in a manner similar to endodermal cells during embryonic development in hydrozoans. While these cells of the inner layer remember their fate in the aggregates, aboral ectodermal cells exhibit plasticity and can be reprogrammed into inner layer cells24. Intriguingly, while Nematostella is morphologically diploblastic, recent work has shown that its germ layers exhibit molecular signatures homologous to the three germ layers of bilaterians, the blastopore lip and resulting pharynx and septal filaments have an endodermal identity (marked by FoxA), while the proximal mesentery corresponds to the mesoderm (marked by SnailA)25,26 (Fig. 1A, B). In this paradigm, the organizer cells (Wnt1/Wnt3+) reside within the endoderm tissue identity, reminiscent of the endodermal location of the Nieuwkoop center in amphibians27,28. Yet, how this population of initially randomly distributed organizer cells clusters and undergoes a symmetry break in order to polarize the aggregate is poorly understood. This raises the quintessential question about the mechanism that underlies the self-organization of the organizer. Schematic summarizing the current model of germ layer identities and their derivatives in Nematostella vectensis. A Signaling pathways cross-talk and marker expression underlying germ layer identities in Nematostella embryos. B Germ layer-derived structures in the mature polyp as well as their contribution to mesenteries (modified from Haillot et al. 2025 and Steinmetz et al. 2017). Here, we reveal the morphogenetic behavior and molecular identity of these germ layers in standardized gastruloids. We show that Notch signaling plays a crucial role in the polarization and symmetry break of the aggregate, which leads to the formation of the oral-aboral body axis as well as the topological segregation of the three germ layer identities in one step. Results Establishing a standardized and scalable system of cnidarian gastruloid generation Previous methods used to make cell aggregates in Nematostella employed dissociation of gastrulae, followed by centrifugation and cutting the pellet by eye into pieces of roughly uniform size24. There are some caveats with this method, including the high degree of compaction leading to potentially a large fraction of highly stressed/damaged cells, non-uniform sizes of aggregates leading to both a variable number of axes as well as spatiotemporal expression patterns of genes. This variability makes it difficult to identify crucial molecular and cellular mechanisms underlying this process of self-organization. To overcome this, we established a high-throughput workflow (see “Methods”) (Fig. 2A), which generates standardized Nematostella gastruloids minimizing cell stress and giving rise to a specific axis number with reproducibility. We found that a seed count of 10,000 cells robustly gives rise to gastruloids that develop into polyps with a single axis (Fig. 2B). Of note, we estimate that about 70% of seeded cells get incorporated into generating the gastruloid, based on the number of cells comprising the midgastrulae stage of the embryo24. The remaining cells are expelled through a hitherto unknown mechanism. To examine the effect of seed count on the number of axes, aggregates of different seed counts were made using a FoxA::mOrange2 reporter line (Table S4), with the expression of fluorophore marking the future pharynx of the polyp. The number of fluorescent poles was then quantified across biological replicates as a proxy for the number of discrete axes formed. The results confirm a highly reproducible and scalable system of gastruloid generation (Fig. 2C, Fig. S1). Interestingly, based on previous estimates of the number of blastopore lip cells present per embryo (Median cells per embryo: 6934; Median blastopore lip cells per embryo: 107)24, we calculate a requirement of ~155 blastopore lip cells (Wnt1/Wnt3+) per every 10 K seed count gastruloid to maintain the high axis inducing reproducibility of this system to give rise to a single axis. To get further insights about axis scaling, we also examined tissue patterning by RNA in situ hybridization staining for endodermal and mesodermal markers FoxA and SnailA, respectively, expressed along the Oral-Aboral (OA) axis. We found that the gastruloids formed repeating modules of FoxA and SnailA+ domains proportional to the seed count, a pattern expected under a Turing-type reaction-diffusion model, where addition of more cells resolves into multiple repeating patterning modules dependent on cell number (Fig. 2D)1. A standardized and scalable system for cnidarian gastruloid generation. A Schematic describing gastruloid generation from dissociated single cells of 24 hpf embryos. B Scaling of axes with increasing seed count in developing gastruloids (Scale bar: 200 μm). C Quantification of no. of axes across multiple biological replicates. 10,000 cells robustly give rise to gastruloids with a single axis, while 30,000 cells generate 2 body axes. Jitter added to discrete values for complete sampling to be visualized. D Double fluorescent in situ hybridization (dFISH) staining for endoderm (FoxA+) and mesoderm (SnailA+) markers showing corresponding molecular scaling of axes. Phenotype counts for 10,000 cell gastruloids: 23/26. Phenotype counts for 30,000 cell gastruloids: 15/23. Asterisk denotes oral pole. dpa; days post-aggregation (Scale bar:100 μm). Axial polarity and reestablishment of germ layer topology occurs concurrently Having established a scalable and reproducible system to generate a single-axis polyp, we wished to monitor the morphogenetic process of germ layer formation. To follow the epithelization of the ectodermal cells, the aboral ectoderm of early gastrulae were dissected and dissociated from sfGFP-β-catenin knock-in embryos29 followed by aggregation and imaging. As the ectodermal cells re-establish cell junctions, β-catenin is recruited to the apical-lateral membrane of all cells. Ectodermal epithelialization started as islands of a few cells, which then expanded to nearly complete epithelialization by 1 dpa (Fig. 3A–F’). Additionally, the ectodermal cells formed β-catenin+ membrane junctions with neighboring ectodermal cells (Supplementary Movie 1) through a previously described mode of short-range sorting in ectodermal cells of Hydra30. Recent work also demonstrated the role of the lateral polarity gene product Lethal giant larvae (lgl) and NCAM2 in establishing proper epithelialization and apico-basal polarization of Nematostella ectodermal cells31,32. Distinct morphogenetic germ layer behaviors in Nematostella gastruloids. A–F’ Rapid epithelialization of aboral ectodermal cell aggregates within a day, visualized using superfolderGFP-β-catenin transgenic line. The red square represents the same patch of cells for a close-up across all timepoints (B’-F’). G–L Ingression of peripheral mesodermal clusters visualized by SnailA::mCherry (red). SiR-actin (blue) was used to counterstain all cells. White arrow represents the points of ingression of the mesodermal cluster. M–T dFISH of foxA (green) and snailA (red) marking endodermal and mesodermal expression domains, respectively, in undissociated mid-gastrulae (N, O) and 4 h, 8 h, 12 h, 24 h and 48 h post aggregation (hpa) (P–T). LV and OV stand for Lateral and Oral view, respectively. Asterisk denotes oral pole. DAPI (gray) was used as a counterstain for all cells. The twister symbol represents dissociation and reaggregation of cells (Scale bar:100 μm). The morphogenetic movement of the mesoderm was visualized by dissociation and reaggregation of SnailA::mCherry transgenic line embryos24 (Supplementary Movie 2). SnailA is a marker of the whole inner cell layer, which has mesoderm-like profiles25,26, including the expression of Cadherin1 upon downregulation of the ectodermal Cadherin3 during the Cadherin switch33. Surprisingly, instead of forming the inner cell layer from the start, the initially randomly distributed mesodermal cells (mCherry+) first formed cell clusters with other mesodermal cells at the periphery of the aggregate, before ingressing collectively as a cluster (Fig. 3K) to form the inner layer (Fig. 3G–L). This is the behavior of all the mesodermal cells in the aggregate, including those located initially in the gastruloid center, previously thought to remain there through gastruloid development. Thus, the initially randomly distributed individual mesodermal cells either actively migrate or are passively pushed out to the periphery due to differences in cell adhesion forces in comparison to the ectodermal cells (Supplementary Movie 3). Notably, at the surface, they tend to attach to other mesodermal cells, forming rather compact clusters of about 20 cells. On average, each gastruloid has about 3 clusters attached prior to ingression at 12 hpa (hours post aggregation) (Fig. S2). Notably, usually one of the clusters suddenly ingresses between 10–16 h post-aggregation, possibly only in regions of incomplete epithelization (Fig. S3). In order to describe the morphogenetic behavior of the endodermal cells, marked by the conserved endoderm transcription factor gene foxA, we used fine temporal double Fluorescent in situ hybridization (dFISH) (Fig. 3P–T). The results showed that an initially randomly distributed foxA+ endoderm (Fig. 3P) attached and intercalated with the snailA+ mesodermal cells already from 8 hpa. This, together with the fact that mesodermal cells form their clusters by migration and differential adhesion, is suggestive that either the endodermal cell population is also initially migratory, or is passively picked up by the mesodermal cells post-dissociation. As the mesodermal cluster ingressed into the gastruloid, the multiple endodermal foci visible at around 12 hpa (Fig. 3R) coalesced into a single primary pole. At 24 hpa, multiple smaller poles were still visible (Fig. 3S), suggesting some underlying possible mechanism of cluster competition to choose a final axis-determining cluster, such as those determined in other systems34. While foxA+ cells exhibited intercalation, brachyury+ cells, which are also primarily endodermal, clustered similarly near the mesoderm like foxA+ cells but with a clear boundary and no signs of intercalation (Fig. S4). While endodermal cells have attained a polarized distribution at one pole of the gastruloid at 24 hpf, they still remain a cell clump on the outside (Fig. 3S). Endoderm epithelialization and invagination are completed by 48 hpa (Fig. 3T). This also includes reestablishment of the aboral pole (six3/6+) (Fig. S5)24, thereby reestablishing the initial body plan. This body plan is roughly comparable with the postgastrula planula larva. Thus, there is only about one day of delay in body axis establishment between gastruloid and normal embryo. Notch signaling facilitates endoderm polarity reestablishment in gastruloids Notch signaling has been recently shown to induce the endoderm at the border of Delta + (ligand) mesoderm and Notch + (endoderm)26, making it a promising candidate regulating the endoderm-mesoderm boundary in Nematostella. Given that the endodermal cells intercalate within mesodermal clusters (Fig. 3Q, R), we first checked if disrupting Notch signaling modulated this boundary in the gastruloids. Notch signaling inhibition with LY-411575 caused disruption of the endoderm-mesoderm boundary (Fig. S6), followed by subsequent randomized distribution of foxA+ endodermal cells in gastruloids observed at 2 dpa (Fig. 4A). The resulting gastruloids also lost their ability to make oral structures such as a pharynx and tentacles, in a LY-411575 concentration-dependent manner (Fig. S7). Interestingly, a similar random distribution of foxA+ cells was also observed in BMP morphant gastruloids, implying potential cross-talk of multiple pathways in regulating polarized endoderm distribution24. Since ectopic activation of Notch signaling could induce ectopic endoderm in early stage embryos26, we asked whether the observed RNA in situ hybridization-stained endodermal population (Fig. 3P–T) was the initial pre-dissociation endodermal cells, or if they arose newly through reinduction during self-organization. Previous work showed that the presence of the endodermal cells, which also express the organizer genes wnt1 and wnt3 in the gastrula prior to dissociation, is essential to confer self-organizing potential to the gastruloids24. However, this experiment was carried out by dissecting the blastopore lip, which could have also removed other essential cell types in the ectoderm and mesoderm. Since Notch signaling inhibitor LY-411575 specifically ablates the endodermal cell identity without affecting ectodermal or mesodermal identity26, we sought to find out if pre-existing endoderm is essential for axis reestablishment in the gastruloids. Confirming the results of the previous study, complete ablation of the organizer cells expressing wnt1/wnt3 in the embryos (Fig. S8A) led to a loss of axis formation during gastruloid development (Fig. S8B). While this experiment does not rule out induction of new cells, it suggests that the initial endodermal cells are essential to reestablish the axis. Notch signaling maintains polarized endoderm distribution. A Gastruloids treated for 2 dpa with either DMSO or Notch signaling inhibitor LY-411575 and stained for FoxA depict dose-dependent loss of polarized endoderm distribution. Asterisk denotes oral pole. dpa;days post-aggregation. (Scale bar: 50 μm). B Gastruloids made from SuH Morpholino injected embryos phenocopy LY-411575 treatment compared to control Morpholino. Dotted lines denote polarized endoderm distribution in control. (Scale bar: 100 μm). C Primary axis induction (10 dpa) in gastruloids made from embryos co-injected with head organizers Wnt1/Wnt3 (green dots represent ectopic co-expression of constructs) and treated with either DMSO or LY-411575 shows Wnt signaling-mediated gastruloid axis formation is Notch-dependent. To understand which components of the Notch signaling cascade regulate this border, we first characterized an expanded panel of annotated canonical Notch signaling components in embryos (Fig. S9A). The primary axis is reestablished in the gastruloids by 24 hpa (Fig. 3P–S). This coincides developmentally with the transition from 24 hpf to 48 hpf (hours post fertilization) embryos, and therefore, we decided to validate the components in these embryonic stages. Inhibiting Notch signaling by LY-411575 treatment led to the downregulation of both the Notch receptor and its ligands delta and DLL. This suggests that Notch signaling positively regulates itself (Fig. S9B). The transition between 24 and 48 hpf also exhibits a split of the expression domains of delta and DLL within the mesoderm. Delta has an expression maximum at the boundary with the endoderm, whereas DLL, a paralog of delta, is expressed interestingly in the aboral-most mesoderm, at the boundary with the apical tuft (Fig. S9A). As the Notch receptor is also expressed in the aboral ectoderm, this receptor and ligand boundary might possibly play other roles, such as in the development of the apical tuft. Finally, the expression domain of annotated Notch signaling coactivator gene Suppressor of Hairless (SuH), also matched that of notch receptor. SuH was reported to be a potential Notch coactivator in Nematostella cnidogenesis35. Knockdown of SuH was also reported to phenocopy the Notch signaling inhibition phenotype in early embryos. Therefore, we also wanted to check if SuH was also involved in maintaining the polarity of the endoderm in gastruloids. Our results show that gastruloids made from SuH MO-injected embryos also phenocopy LY-411575 treated gastruloids (Fig. 4B), thus suggesting involvement of SuH and the canonical Notch signaling cascade in endoderm polarity maintenance. Finally, given the role of Notch signaling in maintaining polarized endoderm distribution, we also wanted to see if there was any interaction between Notch and Wnt signaling in the gastruloids. Gastruloids made from only aboral ectodermal cells are incapable of generating a body axis and germ layers, while the addition of aboral cells co-injected with wnt1/wnt3 can rescue this phenotype and will reestablish the OA axis24. To test if there was any interaction between the Notch and Wnt signaling pathways, we repeated this experiment with the gastruloids grown in Notch signaling inhibitor LY-411575 instead. We confirmed that the addition of wnt1/wnt3 expressing cells to the aggregates from aboral ectoderm rescued the axis forming capacity of the aggregates (Fig. 3C). However, when the Notch signaling inhibitor LY-411575 was added, the ability to generate an axis was lost completely, despite the presence of wnt1/wnt3 expressing cells (Fig. 4C). This suggests that Notch signaling acts downstream of Wnt Signaling in axial identity establishment. Since notch and delta expression in the early embryo were recently shown to depend on Wnt signaling26, our results demonstrate the presence of a feedback loop between Notch and Wnt signaling, which is essential for organizer symmetry breaking in the gastruloids. Notch signaling inhibition converts endoderm to transient endomesoderm To unravel how Notch signaling modulated the identity of the endoderm post-gastrulation, we first checked if it continues to play the role of maintaining endodermal gene expression as previously reported in embryos26. For this, we used embryos treated with LY-411575 from 24 hpf to 48 hpf (post-gastrulation time window) and stained for key endodermal markers foxA and brachyury as well as β-catenin target axin12,29. Notch signaling inhibition led to the downregulation of all these genes (Fig. S10), showing that it still continues to play a role in maintaining endodermal identity. To identify the broader contribution of Notch signaling in modulating the identity of the endoderm and the other germ layers, we generated scRNAseq datasets of the above-mentioned LY-411575 treated (and DMSO control) embryos, where cluster identities were annotated by label transfer from a reference atlas36(detailed in “Methods”). As validation of sufficient drug treatment strength, we found many of the genes previously reported to be downregulated by Notch signaling inhibition to also be reduced in expression in our datasets (Fig. S11). Upon cluster annotation, we could observe an expansion of the soxC+ secretory progenitor cells as well as a truncation of cnidocyte maturation, with a lack of soxA+ mature cnidocytes in the LY-411575-treated embryos (Fig. 5A). This is in alignment with earlier works that have demonstrated the dual role of Notch signaling in inhibiting neurogenesis by limiting the pool of progenitors, as well as promoting cnidogenesis35,37,38. We carried out Weighted Gene Co-expression Network Analysis (WGCNA) analysis to identify gene modules associated with specific cell type responses to the drug treatment (Fig. S12) (Refer to GEO repository-GSE307733_hdWGCNA_modules.RData.gz). One such interesting module was specific to the endomesoderm, here termed the EndoMeso module. This module included genes whose expression were primarily mesoderm-specific in control embryos, however, expanded into the endoderm in the LY-411575 treated embryos (Fig. 5C). This indicated that Notch signaling inhibition did not only lead to a downregulation of endodermal identity as previously seen (Fig. S10), but also to an adoption of mesodermal features, hence defining a transient endomesodermal state (Fig. 5A, B). Among the expanding genes of the EndoMeso module, an interesting candidate we investigated further was Cadherin1 (Fig. 5C). During gastrulation, in Nematostella, the early ubiquitous blastodermal Cadherin3 (Cdh3) becomes downregulated and replaced by Cadherin1 (Cdh1) in the invaginating mesoderm. As a result of this Cadherin switch, Cdh1 and Cdh3 form a sharp meso-endodermal boundary33 (Fig. 5D). Cdh3 was previously shown to play an important role in epithelialization of the ectoderm in Nematostella aggregates33. To test whether Cdh1 plays a role in adhesion of the mesoderm, we first knocked down cdh1 expression by injecting cdh1 morpholinos (MO) in embryos of a mesoderm-specific reporter line (β-Laminin::eGFP-CAAX), which we generated in order to monitor the organization of the mesoderm (Fig. S13A, B). Upon cdh1 knockdown (KD), the mesoderm loses its epithelial architecture in the embryos (Fig. S13C) in line with Cdh1 expected role in mesoderm adhesion. However, since Cdh1 is also expressed in the aboral ectoderm at the planula stage33, cdh1 knockdown also appeared to disrupt the ectoderm, which could also indirectly affect mesoderm organization. Therefore, to show mesodermal specificity, we knocked down β-catenin, which converts the identity of the whole embryo into embryonic mesoderm26, including an ectopic expression of mesodermal markers snailA and cdh1 throughout the embryo (Fig. S13D). These mesodermalized β-catenin KD embryos were injected either with control MO or with cdh1 MO, and then proceeded to dissociate and reaggregate these embryos. We observed that gastruloids made from β-catenin/control KD embryos aggregated properly; however, β-catenin/cdh1 KD aggregates fell apart, confirming the function of Cdh1 in adhesion of mesodermal epithelial cells (Fig. S13D). To verify the loss of Cadherin switch observed in the LY-411575 treated embryos in the scRNAseq dataset, we used dFISH which revealed a disruption of the mutually exclusive endodermal and mesodermal domains leading to the expected expansion of the mesodermal cdh1 expression into the FoxA+ endoderm (Fig. 5D). Additionally, we also observed a strong global downregulation of cadherin3 expression (Fig. S14), potentially through Notch signaling mediated downregulation of Wnt signaling (Fig. S10)26. Notch signaling inhibition converts endoderm to endomesoderm. A UMAP dimensional reduction of reference annotated scRNAseq cell clusters comparing DMSO and LY-411575 drug-treated embryos, treated in a window of 24 to 48 hpf. B Feature plots highlighting EndoMeso module eigengene expression (row1), as well as key EndoMeso module genes, cdh1 (row2) and snailA (row3), across both DMSO and LY-411575-treated embryos, highlighting expansion of mesodermal genes into endoderm upon Notch signaling inhibition. C Triple antibody staining in 2 dpf embryos for FoxA, Cdh1 and Cdh3 shows the endoderm expressing Cdh3 while the mesoderm expresses Cdh1 after the Cadherin switch (Scale bar: 100 μm). D Notch signaling inhibition causes loss of Cadherin switch with endoderm co-expressing foxA and cdh1 (Scale bar: 50 μm). Close-ups depict the boundary between endoderm and mesoderm. Note the sharp, mutually exclusive boundary between foxA and cdh1 in the DMSO control, compared to the overlapping expression in LY-411575 treated embryos (Scale bar: 20 μm). scRNAseq timecourse atlas suggests endoderm to transient endomesoderm conversion in gastruloids To verify if the modulation of endodermal identity we observe in embryos also occurs in our gastruloids, we monitored changes of cell identities throughout the development of the gastruloid using single-cell RNAseq. We used a 24 hpf gastrula library as our T0 control timepoint (Fig. 6A). After clustering, we annotated the clusters by query mapping to the latest Nematostella vectensis scRNAseq atlas36. We started by checking the expression of previously identified EndoMeso module genes, cdh1 and snailA (Fig. 5C). Indeed, both cdh1 and snailA are enriched in the endodermal cluster during early stages of aggregation, confirming an early transient endomesodermal cell identity. Subsequently, cdh1 is reduced by 24 hpa and 48 hpa in this cell cluster, adopting a more typical endodermal cell identity (Fig. 6B). This coincides with the reestablishment of cell contacts between the endoderm and mesoderm, followed by proper resegregation of endoderm and mesoderm. The presence of the transient endomesodermal cell identity marked by foxA/cdh1 expression was also verified by Hybridization Chain Reaction (HCR) with an initial major overlap between foxA+ and cdh1+ domains at 4 hpa and a receding overlap with continued development at 24 hpa (Fig. 6C). In the transcriptome of the transient endo-mesodermal cell identity, we found many genes characteristic of cells undergoing epithelial mesenchymal transition (EMT), which is consistent with the previous finding that the mesodermal cells undergo a partial EMT phenotype during gastrulation39. scRNA-seq gastruloid atlas confirms endoderm conversion to endomesoderm. A UMAP dimensional reduction showing different tissue cell clusters from the gastruloid time course starting from an undissociated 24 hpf gastrula (T0) to dissociated and reaggregated gastruloids from 4 h to 48 hpa. B Feature plot showing expression dynamics of example EndoMeso module genes cdh1 and snailA across gastruloid development time course. Note the transient expansion of mesodermal markers into the ecto-endodermal territory. C HCR expression profile of endodermal marker foxA and mesodermal marker cdh1 at 4 hpa and 24 hpa in gastruloid development. D EndoMeso module and mesodermal module projection scores calculated for all annotated clusters across timepoints of gastruloid development. We also calculated the module score of the entire EndoMeso module gene set for our gastruloid development dataset (Fig. 6D). This score measures the average expression of the genes in the EndoMeso module, adjusted by subtracting the background expression of control gene sets that have similar expression levels. As expected, the score is highest for the mesodermal cluster as the expression of the genes is generally mesoderm specific; however, we also observed an elevation in the endoderm cluster, especially in the gastruloids compared to the pre-dissociation 24 hpf embryo (Fig. 6D). This was also evident in the module score ranking, with the gastruloid endoderm having the second highest rank after the mesoderm (Table S3). To show similarity of the EndoMeso module gene set to the mesoderm, we also identified differentially expressed genes specific to the mesodermal cluster from the wildtype scRNAseq atlas36 and calculated the module score again for these genes in the gastruloid dataset. In alignment with the EndoMeso module, scores for the mesoderm module also behaved in a similar fashion, with the highest scores in the mesoderm and an elevation in the endodermal clusters of the gastruloid libraries (Fig. 6D). Taken together, this confirms the transient conversion of the endoderm to an endomesodermal cell identity after dissociation and subsequent reaggregation. Discussion Our findings reveal a unique mechanism of self-organization where both germ layer topology and axial polarity are reestablished through coordinated Notch and Wnt signaling. In normal embryos, Notch signaling plays a crucial role in segregating the endoderm and mesoderm identity, by inducing and maintaining the endodermal identity26, which is marked by the sharp boundary between ecto/endodermal Cadherin3 and mesodermal Cadherin133. Upon dissociation, the interaction of the mesodermal Delta ligand with the ecto/endodermal Notch receptor and hence Notch signaling in the endodermal tissue becomes disrupted (Fig. 7). This abolishes transiently the maintenance of the endodermal identity signaling inhibition, hence leading to an endo-mesodermal profile. In embryos, this is mimicked by treatment with the Notch signaling inhibitor LY-411575, which leads to an upregulation of (mesodermal) Cdh1 expression in the FoxA+ endodermal cells. This indicates that normally Notch (and Wnt) signaling in the endodermal cells directly or indirectly suppresses cadherin1. The transient upregulation of Cdh1 in the individual FoxA+ endodermal cells presumably allows them to adhere to the Cdh1/SnailA+ mesodermal cells during the process of axial polarization and restoration of the correct germ layer identity topology (Fig. 6). This also explains the observed clustering of these endo-mesodermal cells with mesodermal cells (Fig. 3P–T) prior to reestablishing endodermal identity through re-initiation of Notch signaling. Besides upregulation of snailA and cdh1, the transcriptomic profile of the endodermal cells post-dissociation also displays a number of genes characteristic of mesenchymal or migratory cells. This is in line with the partial EMT state of mesodermal cells during gastrulation (Kraus et al., 2003). This presumably facilitates the migratory behavior of endodermal cells prior to clustering with the mesodermal cells. Alternatively, they are passively “collected” by the mesodermal cells due to their transient similarities in cell identity. In summary, while Notch signaling is required for axial polarization, its transient disruption due to the dissociation is necessary to allow the attachment of endodermal and mesodermal cells. If Notch signaling is continuously inhibited in aggregates by treatment with LY-411575, reclustering of endodermal and mesodermal cells and hence axial polarization is abolished. When the endodermal cells again form a clear boundary with the mesodermal cluster, Notch signaling is re-established, and cdh1 and other mesodermal genes are downregulated in the endodermal cells. Summary of self-organization of Nematostella gastruloids. See text for explanation. While EMT is a typical feature of mesoderm, endoderm undergoing EMT can also be found in some mammals, such as mouse embryos, where the definitive endoderm undergoes EMT to leave the primitive streak along the mesodermal wings before undergoing MET to form the gut endoderm40. This suggests a relatively high level of evolutionary plasticity in the regulation of epithelial/mesenchymal cell state in animal embryogenesis. While upstream signaling pathways driving mouse epiblast migration are well studied, it is still unclear what drives the EMT-MET cycle of the definitive endoderm41. Based on our results, cell-cell signaling pathways like Notch signaling could be a plausible candidate. Interestingly, in zebrafish embryos, endoderm internalization is mediated by activation of the N-Cadherin (Cdh2), triggering them to become migratory42. Despite not being orthologs, both Nematostella and bilaterian cadherins exhibit functional convergence in mediating cell adhesion and actin binding dynamics33,43. In addition, their cadherin switches from Cdh3 to mesoderm-specific Cdh1 during gastrulation33, parallels the E- to N-Cadherin switch in vertebrates44. It remains to be seen if the cdh1 expression in the endoderm of the developing Nematostella gastruloids can trigger endoderm EMT. Our results also have important implications for the theory of self-organization. Turing-type reaction-diffusion models, like the Gierer-Meinhardt model, have provided a theoretical framework for pattern formation and self-organization from almost homogenous conditions. In these classical Turing models, the short-range autocatalytic activator would cross-activate a long-range inhibitor, leading to local activation and lateral inhibition45,46. Such a system is able to give rise to either a single (body) axis or to periodic patterns as observed in cases like digit patterning.Wnt/β-catenin signaling, with its slow and short diffusing ligands and quick intracellular β-catenin stabilization, is a strong candidate for a Gierer-Meinhardt activator capable of generating polarity from initially homogeneous conditions45,47. Recent evidence from mammalian gastruloids demonstrates that Wnt polarity can emerge from an initially homogenous population of cells through self-reinforcing β-catenin/Tcf feedback loops48. Pattern formation by a Turing-type reaction-diffusion system assumes that self-organization is based solely on the re-establishment of signaling gradients, without the requirement of cell sorting. In mammalian systems, the number and size of poles are regulated by multiple factors, such as the size and shape of gastruloids49, adhesion-based sorting4,48, as well as the production of Wnt antagonists such as Dkk1 (Dickkopf-1)50–53, which meet the key criteria of a Gierer-Meinhardt inhibitor. In the cnidarian Hydra, the head is maintained by an autoregulatory Wnt3/β-catenin circuit54, with the number of heads controlled by the inhibitor Sp5, a zinc-finger transcription factor, whose knockdown leads to the expected multiheaded polyp phenotype, as well as multiple head organizer poles in aggregates made from dissociated and reaggregated polyps10,11,55. In Nematostella gastruloids, it was previously shown that the presence of some wnt1/wnt3 expressing cells is required to establish a body axis24. These wnt1 and wnt3 expressing cells are found in the blastopore lip of the gastrula56, which also can act as axis organizer when transplanted to an aboral position12. They form a feedback loop with foxA and brachyury2, and together they demarcate the tissue with a transcriptomic profile of the bilaterian endoderm and hence are here termed endoderm25,26. Notably, self-organization of a whole sea anemone polyp from gastruloids does not only imply the formation of a body axis, but also the re-establishment of the topology of three germ layer identities. Therefore, we propose a dual, connected activator system with Wnt/β-catenin signaling driving broad axial patterning and Notch signaling as its coactivator in establishing and maintaining endodermal identity. We emphasize the hitherto unappreciated significance of Notch signaling for the process of self-organization in sea anemone gastruloids. Of note, recent work in regenerating mouse intestinal epithelium demonstrated that Notch/Delta-dependent lateral inhibition is coupled to a bistable cell fate switch57. Here, transient tissue-density-dependent heterogeneity in the mechanosensor YAP1 pre-patterns the expression of the Notch ligand DLL1 via the FOXA1, and DLL1–Notch feedback creates two stable fate states (high-DLL1 ‘sender’ cells and low-DLL1 ‘receiver’ cells). This bistability allows patterns to remain even after the initial YAP1 heterogeneity is lost, providing a ‘memory’ of earlier symmetry-breaking signals. Perturbing Notch signaling alters both the fraction and spatial arrangement of DLL1+ cells, demonstrating that lateral inhibition here not only refines boundaries but stabilizes the pattern formation once completed. In the sea anemone gastruloids reported here, due to the rapid epithelialization and small size, there may not be a requirement for classical fast diffusing Turing inhibitors. This system could achieve axis polarization simply through Wnt/Notch signaling-driven symmetry breaking, Notch-mediated boundary refinement and potential cadherin-dependent cell sorting. By analogy with the mouse intestinal regeneration model, such a Wnt/Notch circuit in gastruloids could operate via a similar bistable lateral inhibition module, preserving organizer versus non-organizer identity and stabilizing against later noise or fluctuations. This pattern scales with the number of cells (Fig. 2B, D; Fig. S1), as 30,000-cell gastruloids robustly form polyps with 2 axes. This shows that multiple organizer poles can emerge if they are at a sufficient distance from each other, indicating the range of inhibition. This scenario mirrors reports of multiple axes in large mouse gastruloids1, with larger tissue sizes warranting inhibitors for wavelength control as in Turing patterns. More experiments studying larger gastruloids would be needed to garner insights into the relevance of Turing models in axial pattern formation. Cnidarians in general exhibit extreme regenerative capabilities, with multiple species that can self-organize from being dissociated and reaggregated23. The various parameters and mechanisms underlying their self-organization have, however, only been studied in detail primarily in Hydra10,13,22 and Nematostella24 and more recently in Hydractinia58. In Hydra, germ layer topology and axial polarity are largely decoupled with head organizers capable of emerging de novo, driven by a combination of biochemical and mechanical cues10. In aggregates made from Hydractinia polyps, self-organization is a 2-step process with the first step being Sphingosine-dependent i-cell migration towards the epidermis to form clusters, followed by Wnt signaling within these clusters that reestablish the head-foot axis. Additionally, all somatic cells are replenished by the i-cells58,59. In Nematostella gastruloids, although not directly comparable in terms of ontogeny to Hydra or Hydractinia, exhibits an alternative mode of self-organization where axial polarity and germ layer organization are reestablished in parallel, mediated by Notch signaling. The divergence of mechanisms across cnidarians indicates that the principles of self-organization among cnidarians may not be universal, motivating a need to describe the variable potential of developmental robustness observed within cnidarians10. Since the starting point is early embryonic cells, we consider the Nematostella gastruloids as a useful cnidarian in vitro embryonic self-organization system that is comparable to vertebrate organoids and gastruloids. We propose that the morphospace that is available to the organisms is larger than anticipated, reflecting a developmental robustness, but which also may be a source for evolutionary plasticity of body plans on the basis of conserved regulatory networks. This comparison would help us learn more about the degree of developmental plasticity in early branching animals that may have allowed for the subsequent diversification of body plans. Methods Animal culture and spawning Nematostella vectensis polyps used for spawning were F2 offspring of the original CH2 and CH6 individuals, isolated by Cadet Hand60. This is also the strain that was sequenced as a draft genome61 and at the chromosome scale62. Nematostella polyps were grown in 16%ₒ Nematostella medium (hereafter NM) at 18 °C as previously described63. These animals were fed 5x a week with Artemia nauplii and spawning was induced by a 9 h light exposure window at 25 °C, followed by incubation for 2.5 h at 18 °C. Eggs were then collected and fertilized with sperm for 45 min. Fertilized eggs were dejellied in a solution of 3% L-cystein/NM solution64 for 30 min and then cleaned thoroughly to remove residual L-cystein using NM. Subsequently, embryos were then raised at 21 °C. Gastruloid generation Low adhesion 96-well plates were prepared by coating sterile non-coated 96-well plates with BIOFLOAT™FLEX coating solution (FaCellitate) using manufacturer instructions. Mid-gastrula stage embryos were then dissociated using Calcium Magnesium-free seawater (hereafter CMF) until the suspension consisted of only single cells with no noticeable clumps24. Tubes were filled with filtered NM and centrifuged at 400 x g for 4 min. Supernatant was discarded, and the cells were washed once with filtered PBS before being finally pelleted and resuspended in filtered PBS. The concentration and viability of the cells were then stained with ViaStain™AOPI staining solution (Nexcelom) and quantified using Cellometer X2 using parameters appropriate for gastrula stage cells. Resultant concentrations were used to calculate required volumes for different seed counts. 10,000 cells were seeded to generate standardized single-axis gastruloids. The cells were then seeded in wells containing filtered NM and spun gently and quickly at 250 g for 5 min. Plates were incubated at 21 °C and cell debris was cleaned 1 dpa, and gastruloid development was subsequently monitored. Live imaging Gastruloid epithelialization (Fig. 2A–F’) was imaged by dissecting out the aboral ectodermal half of β-catenin-sfGFP KI line midgastrula embryos29 and dissociated and reaggregated as described previously. These gastruloids were allowed to develop until 3 hpa when the cells had sufficiently adhered to each other, and then embedded in 2% low melting agarose and imaged on a spinning disk microscope. The site of ingression of mesodermal cells (SnailA + ) (Fig. S3) was imaged by crossing β-catenin-sfGFP KI females and SnailA::mCherry reporter males, and resulting embryos were used to generate gastruloids and were imaged similarly on the spinning disk confocal. Long-term tracking of ingressing peripheral mesodermal (SnailA+) cell clusters during gastruloid development (Fig. 2G–L) was carried out free-floating in NM within special teflon wells using the Light sheet microscopy (Viventis). Cloning and gene expression analysis (Colorimetric + Fluorescent) Transcripts of genes of interest were amplified from cDNA using primers listed in Table S1. PCR products were cloned into pGEM-T (Promega, #A3600). Both FITC or DIG-labeled RNA probes were then generated through in vitro transcription with SP6 or T7 polymerase. In situ hybridization was carried out as described previously12,26. After incubation with anti-Digoxigenin AP (1:4000) or anti-Fluorescein/Digoxigenin POD (1:100), embryos and gastruloids were washed 10 times 10 min with TBST (0.5% Tween). Background was removed by washing embryos 6 times with TBST (0.5% Tween) and kept overnight at 4 °C. Double fluorescent in situ hybridization (dFISH) was performed similarly to single in situ hybridization according to the previously reported protocol65. Fluorescent staining was revealed with the TSA Plus Fluorescein and TSA Plus Cy3 detection Kits (AKOYA Biosciences). Post-staining, embryos were infiltrated with Vectashield (VectorLabs) and imaged. The following modifications were made while staining gastruloids: a. Embryos were grown in 0.5% DMSO from fertilization until prior to dissociation to improve staining in subsequently made gastruloids. b. Given their much smaller volume due to compaction, gastruloids were regularly spun down between ISH steps to prevent loss of material. Hybridization chain reaction (HCR) staining was done as previously reported66 with some modifications. Probe pools for genes cdh1, foxA, brachyury and six3-6 were generated (Molecular Instruments HCR v3.0) using sequences (Table S2). Gastruloids were fixed at 4 hpa and 1 dpa respectively with 4% PFA diluted in 1 x PBS for 1 h at room temperature (RT), washed 3 times with PTw (PBS + 0.5% Tween-20) and stored in 100% Methanol at −20 °C until needed for staining. Gastruloids were washed with 50% Methanol/50% PTw 2 times, followed by 3 washes in PTw. The rehydrated gastruloids were then washed 2 times for 5 min in 5 x SSCT (5x Sodium saline citrate buffer, 0.1% Tween-20). They were then prehybridized in Probe hybridization buffer (Molecular Instruments) for 30 min at 37 °C. They were then hybridized in hybridization buffer consisting of the probe set (0.8 pmol probe in 100 ul hybridization buffer) overnight at 37 °C. Samples were then washed 4 times for 10 min and 2 times for 3 min with Probe wash buffer (Molecular Instruments) at 37 °C, followed by 5 times 5 min wash with 5 x SSCT at RT. Samples were subsequently incubated in Amplification buffer (Molecular Instruments) at RT for 30 min. Parallel, hairpins h1 and h2 (12 pmol) were heated separately at 95 °C for 90 s, snap cooled to RT in the dark, and added to 100 ul of Amplification buffer. Gastruloids were then incubated in this hairpin/amplification buffer mix overnight at RT in the dark. Post signal amplification, samples were washed with 5xSSCT: one time for 1 min, three times for 5 min, and two times for 30 min. Samples were then infiltrated with Vectashield (VectorLabs) and imaged using the Leica Stellaris 5 CLSM. Inhibitor treatments Inhibitor stocks were dissolved in DMSO. Notch signaling was inhibited using a specific gamma-secretase inhibitor (LY411575, MedChemExpress, #HY-50752). In the case of embryo treatments, a concentration range of 40 μM–70 μM was used, treating from 24 hpf to 48 hpf. For gastruloids, a much lower concentration range of 0.1 nM–20 nM was used due to their initial treatment as a clump of single cells, making them more sensitive to inhibition than compact embryos. The treatment window was from 0 hpa to the time of sampling. To generate scRNAseq libraries to capture Notch inhibition response of embryos, 50 μM LY411575 was used. Antibody staining Staining was carried out as described previously33. In brief, samples were fixed in 4% PFA in 1xPBS for 1 h at RT, following which they were incubated in ice-cold acetone at −20 °C for 7 min. For samples being stained for cdh1 alone, samples were fixed in Lavdovsky’s fixative as reported previously33. Then they were washed 5 times in PTx (PBS + 0.2% Triton-X) and blocked in blocking solution (20% Sheep serum, 1% Bovine serum albumin (BSA) in PTx) for 2 h at RT. Primary mouse anti-Cdh3 antibody (1:1000), rat anti-Cdh1 (1:500), rabbit anti-FoxA (1:500), rabbit anti-SnailA (1:500) and rabbit polyclonal anti-GFP (1:500; (abcam290, RRID:AB_303395)) were diluted in blocking solution, and samples were incubated in the relevant primary antibodies overnight at 4 °C. They were then washed in PTx at RT (10 × 10 min each) and blocked again in blocking solution for 2 h at RT. They were then incubated in relevant secondary antibody solutions of goat anti-mouse Alexa Fluor 633 (1:1000), goat anti-rat 568 (1:1000) and goat anti-rabbit 488 antibodies with DAPI (1:1000) diluted in blocking solution overnight at 4 °C. They were then washed in PTx (10 × 10 min) and infiltrated with Vectashield (VectorLabs) and imaged using the Leica Stellaris 5 CLSM. Plasmid and morpholinos microinjection For knockdown experiments, previously published antisense translation blocking Cadherin1 morpholino33, splice blocking SuH morpholino35 and translation blocking β-catenin morpholino67 (Gene Tools Inc., USA) were injected into embryos at 750 μM, 250 μM and 1 mM, respectively, for the experiments. For organizer gene overexpression experiments, plasmids encoding the expression of Wnt1 and Wnt312 driven by ubiquitous promoters were microinjected into zygotes at 20 ng μl − 1 each. All microinjections were carried out with 0125 mg/ml of fluorescent Dextran-Alexa488 or Alexa568. Additionally, in experiments involving suH knockdown in gastruloids, the highest concentration of morpholino giving the exact same stage of gastrulation compared to the control was identified. This was done in order for the gastruloids generated from the control and knockdown gastrulae to have the same starting point (Fig. S15). Transgenic lines For the generation of the mesodermally expressed reporter line β-Laminin::eGFP-CAAX and endodermally expressed reporter line FoxA::mOrange, we cloned a 4.6-Kb fragment upstream of the translation start site of the gene Beta-Laminin (gene model: NV2.1719) and a 5.9-Kb fragment upstream of the translation start site of the gene FoxA (gene model: NV2.11441), respectively, into a PCRII-TOPO vector. The mesodermal line has a promoter driving the expression of eGFP conjugated to a membrane tag, and the endodermal line has a promoter driving the expression of mOrange. This was followed by an SV40 polyadenylation signal. These constructs were then injected into wild-type zygotes as described68 to generate mosaically transgenic F0 animals, which were then crossed to generate fully heterozygous F1 animals. These were then intercrossed to generate F2 animals that were then used for the experiments. Refer Table S4. for the sequence of Beta-Laminin and FoxA regulatory regions cloned. Generation and analysis of single-cell RNA sequencing (scRNAseq) libraries Single-cell suspensions Samples either after drug treatment or post-aggregation were collected in tubes. They were then washed with ice-cold Calcium Magnesium Free (CMF) seawater briefly, and the samples were allowed to settle. The CMF was then removed, and they were resuspended in ACME (ACetic-MEthanol) prepared as described previously69, and the tubes were left on the rocker for 20 min. The samples were then dissociated using a cut pipette tip every 5 min during the 20 min window, and their dissociation status was checked until sufficient. Tubes were then spun at 400 x g for 4 min at 4 °C, and the supernatant was carefully removed without disturbing the loose cell pellet. The pellet was then resuspended with 1xPBS + 0.5% BSA + RNaseOUT (40 U/ml), and DMSO was added to be 10% of the final volume. This cell mix was then stored at −80 °C until the day of library preparation. The tubes were then left to thaw at RT and then spun again at 400 x g for 4 min at 4 °C, after which the cell pellet was resuspended in 1xPBS + 0.5% BSA + RNaseOUT (40 U/ml), and the viability and concentration were then quantified using ViaStain™AOPI staining solution (Nexcelom) and the Cellometer X2. All processed samples had a viability of at least 85%. Library preparation - species mixing Gastruloid libraries were generated in the format of mixed species libraries. This is where single-cell suspensions of Nematostella vectensis were mixed with suspensions from a different species (here, Aurelia coerulia), maximizing sequence divergence so sequences from each of the species can be mapped to their reference genomes unambiguously. The suspensions were mixed using a 1:1 ratio, and care was taken to keep the cells on ice prior to loading. The mixed cell suspension was processed using the Chromium Single Cell 3’ Gene expression Solution (10X Genomics) according to the manufacturer’s instructions. Suspensions were loaded onto the Chromium controller to generate Gel Bead-in-Emulsions (GEMs). Following reverse transcription and amplification, libraries were constructed using the Single Cell 3’ v3.1 chemistry, using recommended PCR and clean-up steps. Sequencing and data analysis Libraries were sequenced on an Illumina HiSeq platform, targeting a sequencing saturation of at least 70% per library. Sequencing reads were mapped to the Nematostella vectensis genome62 using standard parameters of Cellranger v7.1.070. Filtered matrices were then used to proceed with analysis using Seurat v571,72. The libraries were filtered for cells with low read and feature counts and analyzed using our scripts (see data availability). Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Peer Review file Description of Additional Supplementary Files Supplementary Movie 1 Supplementary Movie 2 Supplementary Movie 3 Reporting Summary Source data Source Data Acknowledgments We thank the Nematostella facility staff members led by Angela Caballero. We would like to thank the Core Facility for Cell Imaging and Ultrastructure Research (CIUS) of the Faculty of Life Sciences. We would also like to thank the BioOptics facility of the IMP/IMBA/GMI for microscopy support, in particular, Alberto Moreno-Cencerrado, Gabriele Bradamante, and Pawel Pasierbek for technical assistance and data acquisition with lightsheet (LS1, Viventis) and widefield microscopy (CD7, Zeiss), for the design of the microscopy experiments, and for helping in the design of customized sample chambers produced by Martin Colombini of the Mechanical Engineering Center (IMP). We thank Lukas Hille (Uni Vienna) for his help with image analyses of the gastruloids, Juan Montenegro for his help with data management and accessibility, and Grigory Genikovich for help in generating the β-Laminin::eGFP-CAAX reporter line. For the purpose of Open Access, the authors have applied a CC BY public copyright license to any Author Accepted Manuscript (AAM) version arising from this submission. Author contributions S.N. and U.T. conceived the study, S.N. carried out most of the experiments, and S.N. and U.T. analysed the data. F.H., E.H., and A.G.C. contributed experimental data, S.N. and U.T. wrote the paper, E.M.T. edited the draft of the paper, and U.T. provided financial support. Peer review Peer review information Nature Communications thanks Thomas Holstein and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available. Funding This work was supported by grants of the Austrian Science Fund FWF to U.T. (P34404; P34870), Austrian Science Fund DocFund-Scorpion (DOC72), the Research Platform-SINCEREST of the University of Vienna (369100) and a PhD completion grant by the Konrad Lorenz institute (KLI), Klosterneuburg to S.N. Data availability Raw sequencing data from captured cells have been deposited into the GEO repository GSE307733. All processed R objects used for analyses and a list of identified module genes are also deposited in the aforementioned GEO repository (GSE307733). Source data are provided with this paper. Code availability Scripts used for analyses of single-cell data are available at https://github.com/technau/Gastruloids (TechnauLab and hitwhicks, 2026). Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Supplementary information The online version contains supplementary material available at 10.1038/s41467-026-74441-x. References Associated Data Supplementary Materials Supplementary Information Peer Review file Description of Additional Supplementary Files Supplementary Movie 1 Supplementary Movie 2 Supplementary Movie 3 Reporting Summary Source Data Data Availability Statement Raw sequencing data from captured cells have been deposited into the GEO repository GSE307733. All processed R objects used for analyses and a list of identified module genes are also deposited in the aforementioned GEO repository (GSE307733). Source data are provided with this paper. Scripts used for analyses of single-cell data are available at https://github.com/technau/Gastruloids (TechnauLab and hitwhicks, 2026).