Changes in cellular composition shape the inductive properties of Hensen’s Node

Abstract

The establishment of the vertebrate body plan is orchestrated by the organizer, a specialized group of cells with inductive properties that guide axial specification during early development. In avian embryos, organizer cells reside within Hensen’s node, a transient structure located at the tip of the primitive streak. Despite its pivotal role during gastrulation, the cellular architecture of the Hensen’s node remains poorly understood. Here, we show that Hensen’s node is composed of two transcriptionally and functionally distinct organizer populations. In addition to anterior GSC-expressing cells associated with head induction, we identify a posterior population co-expressing organizer and mesodermal genes. These posterior cells exhibit trunk-inducing activity when transplanted into naïve tissue. Our findings reveal that the organizer is a dynamic and spatially compartmentalized structure, and that temporal changes in the relative abundance of anterior and posterior populations underlie shifts in its inductive capacity, ensuring coordinated patterning along the body axis.

The organizer is a signaling center in the embryo that orchestrates the formation of body axes. Here they show that Hensen’s Node contains two main cell types and that changes in their abundance alter its ability to induce head versus trunk structures.

Introduction

Two crucial steps in normal vertebrate development are the formation of the three germ layers and the specification of the body axes in the early embryo. Both processes are orchestrated by the organizer, a transient and dynamic structure in the gastrula formed by a group of cells with inductive properties^1–4. The organizer was first identified a century ago in classical transplantation experiments, which showed that this structure could induce a secondary axis when grafted into host embryos¹. Organizer cells regulate various signaling pathways to instruct nearby cells to form neural and mesodermal derivatives, which are organized along the antero-posterior axis of the embryo^3,5–7. The organizer is highly conserved among vertebrates, and it can be identified by the expression of a set of genes that include the homeobox transcription factor GOOSECOID (GSC)^8–11 and the BMP-antagonist CHORDIN (CHRD)¹². Notably, the organizer function is dynamic, and its inductive properties have been shown to change over developmental time^2,3,13,14. However, the molecular mechanisms driving these changes in inductive properties remain poorly understood.

In avian embryos, organizer cells reside within the Hensen’s node^15,16. Characterized in rabbit and guinea pig embryos by Victor Hensen in 1876, the node is a crucial signaling center that also differentiates into various embryonic tissues^17–19. Classic grafting experiments have demonstrated its powerful inductive capacity, showing that when transplanted into a host embryo, the Hensen’s node can induce the formation of a secondary body axis^18,20. The node forms at the anterior end of the primitive streak in the early gastrula²¹ and subsequently moves posteriorly during neurulation^22,23. Fate mapping studies have shown that node cells contribute to the central nervous system and the prechordal plate, as well as axial structures such as the notochord and somites^24–28. Thus, the Hensen’s node plays a critical role in forming anterior structures and in the sequential development of posterior tissues as it regresses. Despite its importance, the cellular composition of the node and the genetic programs that define its distinct cell populations are still largely unexplored.

In this work, we employed single-cell transcriptomics, high-resolution hybridization chain reaction fluorescence in situ hybridization (HCR-FISH), and experimental embryology to uncover the cellular composition and molecular identity of Hensen’s node. Our results reveal that the node is divided into two spatially distinct populations along the anterior-posterior axis, each defined by unique molecular signatures and inductive capacities. The anterior population, marked by GSC expression, is involved in the induction of cephalic structures^9,29,30. In contrast, the posterior population expresses mesodermal genes and is involved in trunk and caudal development. Both populations possess organizer activity, as shown by their ability to induce secondary axes when grafted into host embryos. Notably, the composition of the node is dynamic: during early gastrulation, the node is dominated by anterior organizer cells. In later stages, as the node regresses along the primitive streak, the emergence (and eventual predominance) of the posterior population of cells is observed. We propose that such cellular composition shifts underlie the node’s changing inductive properties, ensuring the coordinated formation of both anterior and posterior embryonic structures.

Results

The Hensen’s node is composed of two CHRD-positive cell populations

To characterize the cellular composition of the Hensen’s node, we microdissected the anterior portion of the primitive streak from gastrula stage chick embryos (Hamburger and Hamilton Stage 4; HH4, n = 38)³¹ and processed samples for single-cell RNA-seq (scRNA-seq) (Fig. 1a). This resulted in a dataset composed of 8134 high-quality cells. Unsupervised clustering and uniform manifold approximation and projection (UMAP)³² identified four cell clusters (Fig. 1b) within our samples. To determine the identity of these cell populations, we mapped the expression of specific markers of each cluster (Fig. 1c, Supplementary Fig. 1a and Supplementary Data 1) with HCR-FISH^33,34. We found that the upper cluster in the UMAP (in green), characterized by high expression of epithelial genes such as CDH11 (Fig. 1d–e), CLDN1 (Supplementary Fig. 1a), and SOX3 (Supplementary Data 1), contained the epiblast cells adjacent to the node. In contrast, the cluster occupying the lower part of the UMAP (in magenta) was composed of endodermal cells, identified by the expression of specific markers such as SOX17 (Fig. 1f, g). This was expected since the tissue samples we microdissected from gastrula-stage embryos also included the cells surrounding the node. Nevertheless, most of the cells in our dataset displayed robust CHRD expression and were thus part of the Hensen’s node (Fig. 1h, i).

Fig. 1. Single-cell transcriptomic analysis reveals molecular heterogeneity in the Hensen’s node.

a Schematic representation of the experimental design: 38 Hensen′s nodes were dissected from stage HH4 embryos and subjected to single-cell RNA-seq analysis. b UMAP visualization of all clusters from the dataset, colored by cell identity, depicting 8134 individual cells. c Dot plot showing the expression levels and percent expression of key marker genes for each cluster, indicating specific gene expression in each cell population. d UMAP displaying the expression of CDH11 in the epiblast cluster. e HCR-FISH showing CDH11 expression in the anterior epiblast around the Hensen′s node (n = 3/3 biological replicates with similar results). f UMAP displaying the expression of SOX17 in the endoderm cluster. g HCR-FISH showing SOX17 expression in the Hensen′s node and surrounding endoderm at HH4+ stage (n = 3/3 biological replicates with similar results). h UMAP displaying the expression of CHRD in the anterior and posterior node clusters. i HCR-FISH showing CHRD expression in the Hensen′s node at stage HH4 (arrow, n = 10/10 biological replicates with similar results). j UMAP displaying the expression of GSC, an organizer-specific gene, in the anterior node cluster. k HCR-FISH showing GSC expression in the Hensen’s node at stage HH4 (n = 10/10 biological replicates with similar results). l UMAP displaying the expression of LMO1 in the posterior node cells. m HCR-FISH showing LMO1 expression in the Hensen′s node at stage HH4 (n = 10/10 biological replicates with similar results). Scale bar = 200 μm.

CHRD-positive cells could be divided into two clusters within the UMAP, corresponding to anterior and posterior cell populations within the node (Fig. 1b, c). The anterior node cluster expressed high levels of transcription factor GSC (Fig. 1c, j, k), a canonical marker of the vertebrate head organizer²⁹. GSC plays a role in regulating gastrulation movements and represses the expression of WNT, BMP, and genes encoding ventral fates^9,29,35,36. This cluster also expressed other anterior genes such as CER1, ADMP, and OTX2 (Supplementary Fig. 1a, b). The second CHRD-positive population did not contain GSC transcripts and instead could be identified by the presence of LMO1 (Fig. 1l), which was expressed in the posterior portion of the node (Fig. 1m). This group of cells also displayed high expression levels of transcription factors known to regulate the formation and development of mesoderm, such as MSGN1 and MESP1, and mesenchymal markers, including ZEB2 and RND3 (Fig. 1c and Supplementary Fig. 1a). Gene ontology analysis of genes enriched in this posterior cell cluster revealed significant enrichment for terms associated with mesodermal programs, including mesoderm development and morphogenesis (Supplementary Fig. 1c). These results suggest that the HH4 Hensen’s node is composed of two CHRD-positive cell populations with distinct molecular profiles.

The cellular organization of the Hensen’s node

Next, we mapped the spatial distribution of the two CHRD-positive cell populations within Hensen’s node with multiplexed HCR-FISH, histological analysis, and image segmentation. Co-localization of CHRD (Fig. 2a) and genes enriched in the anterior and posterior node clusters allowed us to define how these cell populations are organized within the node. Triple staining for CHRD, GSC, and LMO1 confirmed that CHRD-positive node cells can be divided into anterior and posterior cell populations (Fig. 2a–d). GSC-positive cells were positioned anteriorly (Fig. 2b). In contrast, LMO1-positive cells were organized in two bilateral territories that surround the anterior portion of the primitive streak (Fig. 2c). Thus, GSC and LMO1 demarcate two spatial compartments within the node that are organized along the anterior-posterior axis (Fig. 2d–e). To confirm this, we examined additional markers of the posterior node cell cluster (Fig. 1c) with HCR-FISH. We found that CXCL14, POMC, and RND3 were also enriched in the posterior portion of the node (Fig. 2f–h). Multiplexed HCR confirmed co-expression of anterior (GSC and ADMP; Supplementary Fig. 2a–b) and posterior (LMO1, RND3, and CXCL14; Supplementary Fig. 2c–f) node markers in their respective populations. Other mesodermal genes, such as MSGN1, were expressed in the posterior node and also detected in the neighboring mesoderm and primitive streak (Supplementary Fig. 2g, h).

Fig. 2. Spatial segregation of organizer subpopulations in stage HH4 Hensen’s node.

a Whole-mount HCR-FISH showing CHRD expression throughout the Hensen’s node at stage HH4. b–d Anterior organizer cells marked by GSC are localized to the anterior region (b, cyan arrowhead), while posterior organizer markers such as LMO1 (c, magenta arrowhead) are enriched in lateral and posterior regions. d Merged image showing spatial segregation of GSC (cyan) and LMO1 (magenta). e Schematic summarizing the anterior-posterior compartmentalization of the node based on marker gene expression. The dotted line corresponds to the axial level of the transverse sections in (i, j, and k). f–h Additional posterior node markers (CXCL14, POMC, and RND3) are also enriched in lateral and posterior domains of the node (magenta arrowheads, n = 5/5 biological replicates with similar results). i, j Transverse section showing CHRD (i) and GSC (j) expression at stage HH4 (n = 3/3 biological replicates with similar results). k Segmentation analysis of transverse sections showing GSC-positive anterior organizer cells positioned ventrally and RND3-positive posterior organizer cells enriched in dorsal-lateral regions. l Schematic depicting the distribution of anterior (cyan) and posterior (magenta) node populations along the dorsal–ventral axis. D, dorsal; V, ventral. a–d (n = 3/3 biological replicates with similar results). Scale bar = 100 μm.

We next determined how the anterior and posterior node populations were organized along the dorsal-ventral axis of the embryo. To accomplish this, we performed multiplexed HCR-FISH for node genes on transverse cross-sections of gastrula-stage embryos. Analysis of the expressions of CHRD (pan-node marker), GSC (anterior node marker), and RND3 (posterior node marker) indicated that the anterior and posterior cell populations also occupied distinct dorsoventral positions within the stage HH4 node (Fig. 2i, j). To confirm this, we analyzed these samples with CellProfiler image segmentation³⁷ (Supplementary Fig. 3 and Supplementary Data 5), which allowed us to computationally isolate individual CHRD-positive cells (Fig. 2i and Supplementary Fig. 3) and compare the expression of GSC (Fig. 2j) and RND3 (Supplementary Fig. 3) within the node. This analysis revealed that GSC-positive cells are positioned in the ventral portion of the node, while most RND3-positive cells occupy a dorsolateral position within the structure adjacent to epiblast cells (Fig. 2k). While the two populations display distinct spatial biases, we also observe some intermingling of cell types at the boundaries of their territories (Fig. 2k), suggesting a gradual transition between anterior and posterior node domains. Similar to this, we also detect the co-expression of markers such as GSC and SOX17 in cells near the anterior node, consistent with transitional states as some organizer-derived cells begin to adopt mesendodermal fates (Supplementary Fig. 2i–n).

We confirmed this dorsoventral distribution of anterior/posterior node cells by examining embryos co-stained with GSC and TBXT, which in avian embryos is expressed in both node populations³⁸ (Supplementary Fig. 4a, b). Confocal microscopy revealed that GSC-positive cells could only be detected at deeper dorsoventral Z-sections within the node (Supplementary Fig. 4c–i), consistent with our image segmentation results (Fig. 2k). Together, these results allowed us to map the cellular organization of the Hensen’s node at the mid-gastrula (Fig. 2l). We found that the anterior and posterior node subpopulations are spatially segregated along the anterior-posterior and dorsal-ventral axes (Fig. 2e, l). At this developmental stage, anterior node cells are positioned at the center of the node. In contrast, posterior node cells occupy a more lateral position before and after ingression into the blastocoel (Fig. 2e, l). RNA velocity analysis^39,40 of stage HH4 dataset revealed early transcriptional divergence between anterior and posterior node populations, with vector fields indicating distinct developmental trajectories emerging within the node region (Supplementary Fig. 4j). This analysis also supports a gradual transition from a subset of anterior node cells toward the endodermal cluster, suggesting a developmental trajectory linking these organizer cells to the mesendodermal lineage, consistent with their known fate contribution^41,42. Together, the results show that by stage HH4, anterior and posterior node cells are already following separate transcriptional paths that may underlie their distinct behaviors and inductive properties.

Temporal segregation of node cell populations

As development proceeds, the Hensen’s node regresses toward the posterior end of the embryo^22,23. This event marks the transition from gastrula to neurula and is crucial for notochord formation and axial elongation^24,27,43. To define how the cellular composition of the node changes upon regression, we conducted scRNA-seq analysis of dissected stage HH6 nodes (Fig. 3a, n = 20). Hensen’s nodes from neurula stages were dissected and dissociated for scRNA-seq, which resulted in a dataset composed of 12,753 high-quality cells. This dataset contained the posterior node populations and neighboring cells from the notochord, neural tissue, endoderm, and presomitic mesoderm (Supplementary Fig. 5a, b, d and Supplementary Data 2). To identify the node cells, we isolated and re-clustered the CHRD-positive cells in our dataset (Supplementary Fig. 5c, e). Since differentiating notochord cells also express CHRD, we examined the expression of the notochord marker NOTO in the resulting UMAP (Supplementary Fig. 5f). This allowed us to distinguish notochord cells from the regressing Hensen’s node by identifying the CHRD-positive and NOTO-negative (Fig. 3b). Differential analysis of gene expression revealed that these cells retain the molecular signature of stage HH4 posterior node, including CXCL14, MSGN1, RND3, and POMC (Fig. 3b and Supplementary Fig. 5d, g–j). HCR-FISH revealed that these cells remain positioned at the posterior-lateral part of the node (Fig. 3c, d and Supplementary Fig. 5k, n). These results show that the regressing node cells retain the molecular program of the stage HH4 posterior node (Fig. 3).

Fig. 3. Temporal segregation of Hensen’s node cell populations.

a Schematic representation of the experimental design for single-cell RNA-seq, performed on 20 Hensen’s nodes at stage HH6. b UMAP plot showing the expression of CXCL14 and NOTO within CHRD-positive node cells. c, d HCR-FISH showing CXCL14 expression in the posterior node (node is outlined by a dotted line). White arrows highlight the localization of CXCL14/CHRD-positive cells in the posterior region of the node (n = 5/5 biological replicates with similar results). e–g UMAP plots displaying GSC and LMO1 expression in CHRD-positive cells from nodes at stages HH3 (e), HH4 (f), and HH6 (g). GSC-positive cells mark the anterior node present on HH3-HH4 stages, while LMO1-positive posterior cells emerge in the node by stage HH4 and remain at stage HH6. h, i HCR-FISH for CHRD, GSC, and RND3 at stage HH5 (n = 3/3 biological replicates with similar results). CHRD marks the entire Hensen’s node, GSC labels the anterior cells (cyan arrows), and RND3 is predominantly expressed in the lateral regions of the regressing node (magenta arrows). j–l Schematic depicting the spatial segregation of the anterior node (AN) and posterior (PN) node populations along the anterior-posterior axis at stages HH3 (j), HH4 (k), and HH6 (l). Scale bar = 100 μm.

Next, we utilized scRNA-seq to investigate the temporal segregation of anterior and posterior node populations. To complement our stages HH4 and HH6 datasets, we analyzed Hensen’s nodes dissected from early gastrula-stage embryos (HH3, Supplementary Fig. 6a, n = 20) when the primitive streak is still elongating and the node is just starting to form^44,45. Analysis of the molecular signatures of cell clusters in this dataset allowed us to identify the distinct cell types present at this developmental stage (Supplementary Fig. 6b, d and Supplementary Data 3). To examine how the composition of the node changes during these developmental stages, we projected the molecular signatures of the anterior (GSC and ADMP) and posterior (LMO1, CXCL14, MSGN1, POMC, and MESP1) organizer cells in the UMAPs of CHRD-positive cells of the three developmental stages (Fig. 3e–g and Supplementary Figs. 5g–j, 6e–m). At stage HH3, we could only detect the anterior signature within the CHRD-positive cells (Fig. 3e and Supplementary Fig. 6f–j), the earliest identifiable node population in our dataset. Posterior node cells were first detected at stage HH4 when both anterior and posterior cell populations co-exist within the structure (Fig. 3f and Supplementary Fig. 6k–m). Finally, at stage HH6, the anterior node signature is no longer detectable, indicating that only posterior cells remain in the regressing node (Fig. 3g and Supplementary Figs. 5d, m, n). Indeed, HCR-FISH analysis at stage HH5 shows that GSC-positive cells begin to leave the node and migrate toward the cranial region of the embryo (Fig. 3h, i). Previous fate mapping analysis shows that these cells will contribute to derivatives such as the prechordal plate and ventral foregut endoderm^9,28,46. These findings indicate that the node is initially established as an anterior structure (Fig. 3j) and that its cellular composition changes to include posterior cells once the primitive streak is fully extended (Fig. 3k). Subsequently, posterior cells dominate the node as it regresses along the primitive streak (Fig. 3l). These results illustrate how the cellular composition of the node changes during development and suggest that its inductive properties may also change accordingly.

To reconstruct the developmental trajectory of organizer cells, we computationally isolated CHRD-positive cells from our scRNA-seq datasets spanning stages HH3, HH4, and HH6 (Fig. 4a). Dimensionality reduction and re-clustering of this subset revealed six transcriptionally distinct populations corresponding to epiblast (EPI), anterior node (AN), intermediate node (IN), posterior node (PN), endoderm (EN), and notochord (NC) (Fig. 4b). To infer potential lineage relationships among these populations, we performed both RNA velocity and Monocle3 pseudotime analysis. RNA velocity (Fig. 4c) suggested directional transitions from epiblast into the anterior node, followed by divergence toward the posterior node, notochord, and endodermal fates. This general structure was supported by Monocle3 trajectory inference, which positioned the epiblast upstream of branching lineages corresponding to the same three terminal fates (Supplementary Fig. 7). Consistent with our HCR-FISH analyses, projection of developmental stage data onto the UMAP revealed that the anterior node cluster was primarily composed of stage HH3 and HH4 cells, while the posterior and intermediate node populations were largely made up of stage HH6 cells (Fig. 4d, e), indicating a shift in cell identity over time. The developmental trajectories inferred from this analysis support a model in which early anterior node cells give rise to endodermal, posterior node cells, and the notochord (Fig. 4f). This hierarchical organization positions epiblast cells upstream in the lineage and places posterior node derivatives at the terminus. These findings suggest that Hensen’s node is a temporally assembled and transcriptionally dynamic structure, where distinct organizer populations emerge in succession to coordinate sequential phases of axial induction.

Fig. 4. Developmental trajectories of Hensen’s node cells across different stages.

a Schematic representation of the Hensen’s node collected for scRNA-seq analysis at stages HH3, HH4, and HH6. b UMAP plot displaying 10,422 aggregated CHRD-positive cells from all three developmental stages (HH3, HH4, and HH6). c RNA velocity analysis showing the predicted lineage trajectories of CHRD-positive cells across developmental time. d UMAP plot showing the cell identities within the CHRD-positive population. e Stacked bar plot representing the proportion of each stage within the identified populations. f Proposed model of cell fate trajectory: Epiblast (EPI) cells give rise to the anterior node (AN) cells and contribute to notochord (NC). Anterior node cells further differentiate into endoderm (EN), notochord (NC), and posterior node (PN), which in turn give rise to paraxial mesoderm (PM).

Changes in neural-inducing properties of Hensen’s node

Previous studies have demonstrated that grafting the Hensen’s node into gastrula-stage embryos can induce the formation of a secondary body axis^18,20. GSC plays a critical role in this process⁴⁷, as it directly activates the transcription of secreted factors involved in embryonic induction, such as CHRD and NOG³⁰. However, our findings indicate that GSC is absent from the CHRD-positive posterior cells in the node (Fig. 2). We thus investigated whether posterior node cells exhibit organizer activity by performing transplantation experiments using transgenic chick embryos. As positive control, we first grafted entire stage HH4 Hensen’s nodes, which contain both anterior and posterior node regions, from GFP-positive donors into stage-matched wild-type hosts (Fig. 5a). These grafts induced the formation of a secondary axis, including anterior neural tissues marked by OTX2 expression (Fig. 5b, cand Supplementary Fig. 8a, b). Next, we grafted stage HH6 Hensen’s nodes, which lack the anterior domain, into stage HH4 hosts. These posterior node grafts induced the formation of a secondary trunk, including a notochord and somites, but failed to generate anterior neural structures (Fig. 5d, e), as confirmed by the absence of OTX2 expression (Supplementary Fig. 8c, d). These findings, along with previous studies using chicken-quail chimeras^26,48, demonstrate that posterior node cells retain organizer activity and can induce neural identity, but do not promote anterior neural fate.

Fig. 5. Induction of posterior neural identity by posterior organizer cells.

a Schematic of the grafting experiment: the Hensen′s node from a GFP + donor embryo at stage HH4 or HH6 was transplanted into the marginal zone of a wild-type stage HH4 host. b, c Whole-mount immunostaining for OTX2 (magenta) shows that stage HH4 node grafts induce a secondary axis (arrowhead) that contains anterior neural tissue (pink arrow, n = 5/5 biological replicates with similar results). d, e In contrast, HH6 node grafts induce posterior axial structures (arrowhead), including CHRD-expressing notochord (magenta arrow, n = 5/5 biological replicates with similar results). f Schematic of enhancer reporter experiment: stage HH4 or HH6 donor nodes were grafted into transgenic stage HH4 host embryos, electroporated with SOX2N1::GFP (N1) and SOX2N2::mCherry (N2) reporter constructs. g, h Representative images showing differential enhancer activation following grafting of stage HH4 (g, n = 12/16 biological replicates with similar results) or HH6 (h, n = 11/19 biological replicates with similar results) node tissue. i Box plots showing the distribution of N1-GFP+ and N2-mCherry+ cells per embryo after transplantation of nodes at stages HH4 or HH6. Each dot represents a single embryo (biological replicate); n = 12 embryos (HH4), n = 11 embryos (HH6). Boxes indicate the median (center line), 25th and 75th percentiles (box), and whiskers denote values within 1.5 × interquartile range. A two-way binomial generalized linear model (GLM) was used to assess the effect of stage on marker expression. For GFP+ cells, the stage effect estimate (HH6 vs HH4) was − 1.203 (z = − 9.78, p = 1.5 × 10^-22, 95% CI: − 1.444 to − 0.962). For mCherry+ cells, the estimate was + 1.203 (z = 9.78, p = 1.5 ×  10^-22, 95% CI: 0.962 to 1.444). Source data is provided as a Source Data file.

To test whether distinct node populations not only orchestrate neural induction but also contribute to the spatial patterning of neural tissue, we used enhancer-reporter constructs derived from the SOX2 locus. The SOX2N2 and SOX2N1 enhancers are active in anterior and posterior regions of the neural plate, respectively, and can serve as readouts for spatial identity during neural induction⁴⁹. Specifically, SOX2N2 enhancer is active in the anterior neural plate, encompassing prospective forebrain and midbrain regions, while SOX2N1 enhancer is active in the posterior neural plate, marking areas fated to become the spinal cord. We generated stage HH4 transgenic hosts carrying SOX2N1::GFP (N1-GFP) and SOX2N2::mCherry (N2-mChe) constructs and grafted wild-type HH4 or HH6 nodes into these embryos (Fig. 5f). Notably, because only host epiblast cells carry the enhancer reporters, these experiments allow direct visualization of neural induction in the host tissue. Grafts of stage HH4 nodes led to robust activation of the SOX2N2 in the secondary axis (Fig. 5gand Supplementary Fig. 8e–h), consistent with the induction of anterior neural identities. Imaging of the secondary axis showed spatial separation between SOX2N2 expression in anterior regions and some SOX2N1 expression in more caudal regions of the induced axis (Supplementary Fig. 8e, f). In contrast, stage HH6 node grafts predominantly activated the SOX2N1 enhancer (Fig. 5h and Supplementary Fig. 8g, h), suggesting that posterior node cells preferentially induce neural fates with posterior character. Indeed, quantification of enhancer activation across multiple embryos revealed that stage HH4 grafts (n = 12, Supplementary Table 1, Supplementary Fig. 8f) consistently induced both enhancers, whereas stage HH6 grafts (n = 11, Supplementary Table 1 and Supplementary Fig. 8h) led almost exclusively to SOX2N1 activation (Fig. 5i). These results indicate that anterior and posterior node populations differ in their ability to induce and pattern neural tissue, with posterior node cells biased toward the caudal neural fates.

Induction of caudal mesoderm by posterior node cells

Next, we determined the cohort of cell types induced by the posterior node by performing scRNA-seq on secondary axes. In this experiment, we inverted our transplantation scheme and grafted wild-type stage HH6 nodes into GFP-positive stage HH4 host embryos (Fig. 6a). This approach enabled us to confidently identify induced cells even with lower-depth single-cell transcriptomes, as the presence of GFP transcripts ensured they originated from the host. After a 16 h incubation, we dissected secondary axes from five embryos, dissociated the tissue, and profiled a total of 7750 cells. These were grouped into 16 clusters (Supplementary Fig. 9a), including mesodermal, neural, ectodermal, and extra-embryonic populations. To examine the diversity of host contributions, we isolated and re-clustered 2372 GFP-positive cells (Supplementary Fig. 9b, c). This dataset revealed 14 distinct clusters corresponding to the cell types induced by the posterior node population. These included posterior mesoderm cells expressing notochord and presomitic mesoderm genes, as well as axial-specific markers like HOXB5 and CDX4, indicating their posterior identity (Fig. 6b, Supplementary Fig. 9dand Supplementary Data 4). In addition, neural plate progenitors, non-neural ectoderm, and neural cells were identified among the induced populations, marked by the expression of genes such as SFRP2, TFAP2A, SOX2, and PAX6 (Supplementary Data 4). Non-induced cell types from the surrounding graft region, such as blood cells, endodermal cells, and extra-embryonic tissue, were also present in the dataset (Supplementary Fig. 9e–g). To examine the extent of host cell induction, we focused on the subset of 3453 cells with ectodermal or mesodermal identity from the initial clustering analysis (Supplementary Fig. 9a). This analysis showed that 1188 cells (34%) from this group were GFP-positive, indicating that a substantial portion of the secondary axis was derived from induced host cells.

Fig. 6. Posterior node cells can induce caudal mesoderm.

a Schematic depicting the transplantation assay, where wild-type (WT) stage HH6 nodes were transplanted into GFP-positive HH4 hosts, followed by dissociation for single-cell RNA-seq (n = 5). b UMAP plot displaying GFP-positive host cells across different clusters, showing the formation of ectodermal and mesodermal cells. Gray dots represent unconverted cells. c, d Schematic of node transplantation experiments into transgenic embryos. WT stage HH4 embryos were electroporated with CHRD-E (notochord enhancer) or TCF15-E (somite enhancer), followed by node grafting from GFP-positive HH6 embryos. e After ~ 16 h, CHRD-E activity is detected in the notochord, confirming the induction of trunk-specific structures in the host (n = 10/17 biological replicates with similar results). f High magnification image of the marked region within the transplanted cells shown in panel (e), showing activation of CHRD-E within the induced secondary axis. g TCF15-E activity is observed in the somite region (arrows), confirming the induction of paraxial mesoderm in host embryos (n = 6/10 biological replicates with similar results). h HCR-FISH showing CHRD (notochord – magenta, arrowhead) and LMO2 (blood islands - blue) expression in embryos after 16 h post-transplantation of stage HH6 GFP-positive node (n = 5/5 biological replicates with similar results). i, j High magnification image of the marked region shown in panel (h). GFP-positive cells from the donor form posterior structures (somites, notochord) and are shown in green (i). HCR-FISH for CHRD (magenta) showing that some notochord cells are GFP-negative (j). Scale bar = 100 μm. NC = Notochord, S = Somite.

Since notochord and paraxial mesoderm cells were relatively rare in our scRNA-seq dataset, we also investigated the capacity of the regressing node to induce posterior mesoderm with additional experiments. We grafted a GFP-positive regressing Hensen’s node (HH6) into stage HH4 embryos electroporated with reporter constructs containing CHRD or TCF15 enhancers (Fig. 6c, d). We identified the CHRD enhancer using our stage HH4 Hensen’s node ATAC-seq dataset and demonstrated its activity in the notochord (Supplementary Fig. 9h, i). TCF15 enhancer was previously described, and it is active in presomitic mesoderm and somites⁵⁰. Since only host cells have been transfected with the enhancer construct, the reporter gene activity indicates mesoderm induction. Consistent with the induction of notochord cells by the graft, we observed activation of the CHRD enhancer in the midline of the induced secondary axis (Fig. 6e, f, n = 10/17, Supplementary Table 1). Adjacent to the notochord, we also observed activation of TCF15 enhancer in the somite epithelium (Fig. 6g, Supplementary Fig. 9j–ln = 6/10 and Supplementary Table 1). These cells also expressed endogenous LEF1 (Supplementary Fig. 9j–l), which is expressed at high levels by paraxial mesoderm^51,52. As expected, we also observe the contribution of the GFP donor tissue to the somites (Fig. 6f and Supplementary Fig. 9j, l), consistent with previous reports that the medial part of these structures is derived from the node²⁴. HCR-FISH for CHRD in the secondary axis also revealed that the notochord presented GFP-negative cells and was thus induced from host tissue (Fig. 6h, i).

Our transplantation experiments indicate that the posterior node displays organizer activity and that these cells can drive cellular diversification in stage HH4 epiblast cells. We next examined the regressing node’s ability to induce naïve cells, such as the early blastoderm and embryonic stem cells, into mesoderm. In the first experiment, we isolated the anterior blastoderm from HH3+ host embryos by dissecting away the primitive streak (Supplementary Fig. 10a). We then grafted stage HH6 GFP-positive nodes in the isolated blastoderm, which was free of any mesoderm-derived cells (Supplementary Fig. 10a). The transplanted node successfully induced the expression of mesodermal and ectodermal markers, specifically BRACHYURY (BRA) and SOX2, respectively (Supplementary Fig. 10b–f'', n = 5/7), demonstrating its capacity to direct naïve cells toward these lineages. In the second experiment, we co-cultured stage HH6 Hensen’s nodes with human embryonic stem cells (hESCs) engineered with reporters for the three germ layers: SOX2-mCitrine (ectoderm), BRA-mCerulean (mesoderm), and SOX17-tdTomato (endoderm) (RUES2-GLR)⁵³. This system allowed us to visually track the induction of germ layer markers in response to the transplanted node (Supplementary Fig. 10g, n = 3/3). Pluripotent cells express SOX2, and a few cells have begun to express SOX17 (Supplementary Fig. 10h), while BRA expression is absent (Supplementary Fig. 10i, j). Upon transplantation, the hESCs adjacent to the explants expressed the mesodermal marker BRA (Supplementary Fig. 10k–o), further confirming the node’s inductive capacity. Together, these results demonstrate that the posterior cell population in Hensen’s node possesses the capacity to induce the formation of posterior structures and cell types, including both axial and paraxial mesoderm, thereby functioning as an organizer in the avian embryo.

Discussion

Here, we provide a detailed characterization of the molecular signatures and inductive properties of the cell populations within Hensen’s node. Our findings reveal that the node is composed of two molecularly distinct populations, anterior and posterior node cells, which function as the head and trunk organizers, respectively. These populations exhibit distinct transcriptional identities, underscoring the molecular and functional heterogeneity of the organizer. Fate map experiments conducted by Storey and colleagues^24,26,54,55 indicate that distinct portions of the node also have distinct developmental fates⁵⁶. Their work identified the medial-anterior region of the node as giving rise to notochord cells, while the posterior-lateral regions contribute to somite formation^24,26. Our study builds upon these findings by demonstrating that these anatomical territories correspond to molecularly distinct cell populations with unique transcriptional programs and inductive capacities.

The division of the organizer into transcriptionally distinct anterior and posterior populations is not unique to the chick and appears to be a deeply conserved feature of vertebrate development. In zebrafish, Saude and colleagues (2000)¹⁴ demonstrated that anterior and posterior organizer activities are spatially segregated within the embryonic shield, the zebrafish equivalent of the node in amniotes and the blastopore lip in amphibians. This compartmentalization is reflected in distinct transcription factor expression: gsc marks cells that induce anterior structures, while floating head/noto marks cells associated with posterior fates. Gritsman et al. ⁵⁷ showed that this anteroposterior polarity is established before gastrulation and is regulated by differential Nodal signaling, with gsc-positive prechordal plate progenitors positioned closer to the margin than noto-expressing notochordal cells. Comparable spatial organization has also been observed in Xenopus, where fate-mapping and functional experiments have long distinguished between anterior and trunk organizer domains within the organizer. The early Xenopus organizer expresses gsc, chordin, and cerberus and is required for head induction^12,29,47,58, whereas posterior regions express xnot and brachyury, contributing to notochord and mesoderm formation^59,60.

In the mouse, recent single-cell atlases⁶¹ have also examined anterior and posterior compartments within the node and axial mesendoderm. These studies revealed early segregation of Gsc and Foxa2 expression domains, alongside Nodal-dependent specification of the anterior mesendoderm, which contributes to head structures, and posterior domains that give rise to the notochord⁶¹. The anterior primitive streak and node harbor organizer-like populations that express Gsc and Otx2, whereas posterior node regions express T (Brachyury), Mixl1, and EMT-associated factors, resembling the molecular signature we report in chick. Our identification of anterior GSC-positive and posterior LMO1/RND3-positive populations in Hensen’s node reinforces the view that the subdivision of organizer function is a conserved feature of vertebrate development. Nevertheless, our findings also highlight possible species-specific aspects of organizer regulation. For example, in the chick, BRACHYURY is expressed in both head and trunk organizer cells, while NOTO expression appears restricted to the forming notochord and is absent from posterior node cells. Alternatively, these patterns may reflect differences in developmental timing or sampling. For example, scRNA-seq analysis in zebrafish⁶² identified an intermediate population co-expressing gsc, brachyury (ta), and noto, suggesting similar transcriptional states may exist in other species but remain undetected. High-resolution comparative studies will be essential to distinguish biological variation from technical limitations.

These findings provide a mechanistic foundation for a long-standing model of organizer function. Our results align with the hypothesis first proposed by Otto Mangold⁶³, who suggested that the organizer has distinct head and trunk inductive capacities. Mangold’s experiments demonstrated that early-stage organizers induce head structures when transplanted, while later-stage organizers promote trunk formation⁶³. Our findings provide a cellular and molecular basis for Mangold’s hypothesis by identifying two distinct populations in the node - one responsible for head induction and the other for trunk development. As the node regresses during gastrulation, its cellular composition shifts, transitioning from predominantly anterior organizer cells to posterior cells. This shift in cell populations likely underlies the changing inductive properties of the node, explaining why early-stage nodes induce head structures while later-stage nodes promote trunk formation. Our data suggests that distinct gene regulatory circuits govern these two populations. GSC is a key regulator in the anterior cells, orchestrating a program geared toward head formation^29,30,64,65. In contrast, the posterior cells appear to rely on a genetic circuit involving LMO1, mesodermal transcription factors, and EMT-related genes. While our transcriptomic analysis suggests that these populations may also have differing secretory profiles, which could contribute to their distinct inductive roles, this aspect requires further investigation.

The shifts in cellular composition we observed within the Hensen’s node also explain the temporal regulation of its inductive abilities. These changes are driven by extensive morphogenetic events during node development. First, there is the migration of anterior node cells that leave the structure to contribute to head formation. This is followed by the regression of the node, which moves posteriorly along the primitive streak during notochord formation. These morphogenetic movements lead to changes in the node’s composition, which we have observed across different developmental stages, as anterior organizer cells are gradually replaced by posterior, trunk-inducing cells. Trajectory analyses using RNA velocity and Monocle3 pseudotime inference suggest dynamic cell state transitions within the node, with epiblast cells giving rise to anterior node cells, which subsequently contribute to posterior node, notochord, and endodermal derivatives. Experimental lineage tracing will be required to validate these trajectories. These observations raise important questions about how trunk organizer cells are specified and how their emergence influences the behavior of head organizer cells. For example, the appearance of trunk organizer cells may initiate signaling events that prompt anterior cells to exit the node. Taken together, our findings support a model in which the node undergoes continuous remodeling through cell-cell interactions and cell-stage transitions, allowing it to coordinate successive inductive events along the body axis.

Beyond providing insights into embryonic development, our findings also have important implications for regenerative medicine and evolutionary biology. Our single-cell analysis of secondary axes induced by node grafts, complemented by enhancer-reporter assays in host tissue, demonstrates that the trunk organizer population is capable of inducing a broad range of caudal cell fates, including both mesodermal and ectodermal derivatives (Fig. 6). Understanding how specific cell populations within the organizer control tissue induction and axis formation could inform strategies for engineering tissues and organs from pluripotent stem cells⁶⁶. For instance, harnessing the inductive properties of these cells may provide a foundation for developing therapies to repair or replace damaged tissues, particularly in cases of congenital defects linked to early patterning errors⁶⁷. Moreover, our work contributes to the broader understanding of how the organizer has evolved across species, offering clues into how vertebrate body plans have diversified through evolutionary time. The identification of specialized organizer populations within the Hensen’s node raises intriguing questions about how these mechanisms may vary across different organisms and what evolutionary pressures shaped the emergence of such distinct yet coordinated inductive functions.

Methods

Ethics statement

This work was conducted with ethical regulations and guidelines approved by BCH ESCRO (BCHSC#2022.10.07).

Embryo collection and tissue dissociation

Fertilized White Leghorn chicken eggs were obtained from the Department of Animal Science, University of Connecticut. Transgenic GFP chicken eggs were obtained from Clemson University through Dr. Susan Chapman. Eggs were incubated at 37-38 °C and 50% humidity until reaching the desired developmental stage. Embryos were collected using filter paper as a carrier⁶⁸ and staged according to Hamburger and Hamilton³¹. The Hensen’s node of 20 embryos at HH3 (12-13 h), 38 embryos at HH4 (18–20 h), and 20 embryos at HH6 (25-26 h) were fine dissected using spring scissors and dissociated using Accumax (Sigma #A7089) for 15–20 min at room temperature to generate a single-cell suspension.

Single-cell RNA-Seq library preparation and sequencing

For single-cell RNA sequencing (scRNA-seq), the dissociated cells were washed with 1X PBS containing 0.04% of BSA and filtered using a Flowmi cell strainer (Millipore-Sigma), porosity of 40 µM to remove cell debris and clumps. The gene expression library preparation and sequencing were conducted by Cornell University BRC Genomics Core Facility, following the manufacturer’s instructions (10x Genomics). The libraries were prepared using the NextSeq 500/550 75 bp kit and sequenced using the NextSeq 500 instrument. The sequencing run was configured with the following read lengths: an initial read of 28 bases, an 8-base pair index read for sample multiplexing, and a final read of 56 bases.

Single-cell RNA-Seq data processing and analysis

FASTq files of each sample generated from the 10x Genomics Chromium platform were aligned to the Gallus gallus GRCg6a.101 reference genome using 10x Genomics Cell Ranger software v5.0.1 with default parameters. Stages HH3, HH4, and HH6 Hensen’s node-filtered expression matrices were analyzed using the R package Seurat (version 4.1.3). We filtered out the low-quality cells by removing single cells with less than 2000 (HH3), 1000 (HH4), and 500 (HH6) genes, and with a percentage of reads that map to the mitochondrial genome above 5%. After the filtering step, 4849 (HH3), 8694 (HH4), and 15,095 (HH6) cells remained, and the data were normalized using the scale factor = 10,000. Next, we identified the highly variable expressed genes that were used in the downstream analysis by using the function FindVariableFeatures. The Seurat pipeline was used to perform linear dimensional reduction and unsupervised clustering. To generate the UMAP of the 4849 (HH3), 8694 (HH4), and 15,095 (HH6) cells, we used the first 20 principal components. In the clustering step, the number of dimensions used to find the neighbors was 20, and the resolution was 0.3 (HH3 and HH4) and 0.5 (HH6). Next, to identify and remove the potential doublets, we used the DoubletFinder tool. A total of 250 (HH3), 466 (HH4), and 868 (HH6) doublets were identified and removed from the datasets. A total of 4599 (HH3), 8228 (HH4), and 14,227 (HH6) cells passed the criteria described above. Cells with high content of mitochondria biogenesis and cell cycle genes were excluded from the downstream analysis. This adjustment resulted in a final cell count of 4512 for HH3, 8134 for HH4, and 12,753 for HH6. Differential expression analysis for each cluster was then conducted using the FindAllMarkers function. Genes on the female chromosome W were excluded from the HH3 and HH6 datasets to prevent sex-biased analysis and ensure the integrity of cell populations. Cell identity for each cluster was assigned based on the expression profile of known marker genes. CHRD-positive cells were identified from the scRNA-seq dataset by selecting cells with a scaled expression level of CHRD greater than 0.5. The subset of CHRD-positive cells was re-clustered using the FindNeighbors function on the first 20 principal components with a resolution of 0.5. The total CHRD-positive cells were 1759 for HH3, 2215 for HH4, and 6538 for HH6.

Whole-mount HCR-FISH in chick embryos

Hybridization chain reaction RNA fluorescent in situ hybridization (HCR-FISH, Molecular Instruments HCR protocol v3.0) was performed on stage HH4 (18–20 h) and HH6 (25-26 h) whole mount chick embryos. Briefly, embryos were collected and fixed in 4% paraformaldehyde for 1-2 h at room temperature. After fixation, embryos were dissected to remove the vitelline membrane, washed with PBS containing 0.1% Tween (PBST) on ice, and then stored in methanol at − 20 °C until use. For the rehydration process, embryos were subjected to a series of graded Methanol/PBST washes on ice. During the detection step, embryos were pre-hybridized with probe hybridization buffer for 30 min at 37 °C, followed by incubation with 4 pmol of each probe set overnight. Excess probes were removed by washing the embryos with probe wash buffer at 37 °C. Embryos were then pre-amplified with amplification buffer and incubated with hairpin overnight at room temperature, protected from light. The excess hairpins were removed by washing the embryos with 5X SSCT at room temperature. Embryos were incubated with DAPI for 1 h, followed by sample mounting for microscopy. Imaging was conducted at Cornell University BRC imaging core facility using a Zeiss LSM880 confocal multiphoton inverted microscope with a 20x objective lens. Additional imaging was also performed at the Core for Imaging Technology & Education (CITE) at Harvard Medical School, using a Nikon AX R point scanning confocal microscope, also equipped with a 20x objective lens. In addition, embryos were imaged using the upright Zeiss Axio Zoom fluorescence microscope available in our lab. Image analysis and processing were performed using Fiji software. The target mRNA probe sets, amplifiers, and buffers were purchased from Molecular Instruments (https://www.molecularinstruments.com/). Probes purchased from IDT were designed using the custom software created by the Özpolat Lab⁶⁹. These probes were amplified with Molecular Instruments hairpins. All HCR-FISH experiments were performed in multiple embryos (n ≤ 10) per condition to ensure reproducibility.

HCR-FISH on tissue sections

For HCR in tissue sections, embryos were collected and fixed in 4% paraformaldehyde for 1 h at room temperature. Following fixation, embryos were dissected, washed with DEPC-PBS, and incubated with 5% sucrose/PBS for 3 h at room temperature. Embryos were then transferred to 15% sucrose/PBS overnight. The next day, embryos were incubated with a 1:1 mixture of OCT and 30% Sucrose/PBS, embedded in disposable plastic cryomolds, and stored at − 80 °C until cryosectioning. Tissue sections of 10 μM were obtained using the CryoStar NX50 (Thermo Fisher). HCR-FISH was performed following the HCR RNA-FISH protocol for fixed tissue sections provided by Molecular Instruments (https://www.molecularinstruments.com/hcr-rnafish-protocols). In brief, sections on slides were fixed in ice-cold 4% paraformaldehyde for 15 min at 4 °C. The slides were then washed with a series of ethanol at room temperature and then incubated in 1X PBS. For the detection stage, slides were pre-hybridized for 10 min inside a humidified chamber pre-warmed to 37 °C, followed by incubation with 0.4 pmol of each probe set overnight. Excess probes were removed by washing the embryos with probe wash buffer at 37 °C. Next, the sections were pre-amplified with amplification buffer in a humidified chamber for 30 min at room temperature and incubated with hairpin solution in the dark overnight, also at room temperature. Finally, excess hairpins were removed by washing the embryos with 5X SSCT at room temperature. The slides were incubated with DAPI for 1 h, followed by sample mounting for imaging. Tissue sections were imaged using a Zeiss Imager.Z2 fluorescent microscope with a 20x objective lens.

Image segmentation and analysis

Cell segmentation was performed using the open-source image software Cellprofiler (v.4.2.1 www.cellprofiler.org). A custom pipeline (Supplementary Data 5) was designed to segment the nuclei, define the CHRD-positive domain, and quantify marker expression at the single-cell level. The Hensen’s node cells can be identified by CHRD expression. Non-Hensen’s node cells were excluded, and the intensity of the expression for anterior and posterior node markers were analyzed to characterize the spatial distribution of these two populations within the Hensen’s node. This pipeline effectively detects the distinct compartments where each cell population is present.

Pseudotime trajectory analysis

RNA velocity analysis was performed using the Python package scVelo⁴⁰. Initially, the spliced and unspliced counts matrix of stage HH4 Hensen’s node was constructed using the Velocyto command line tool³⁹. Primary data processing of stage HH4 Hensen’s node scRNA-seq was performed using Seurat in R. To convert the Seurat object into a format compatible with scVelo, we exported all necessary metadata, including the metadata table, expression counts matrix, dimensionality reduction matrix, and gene names. Next, the data was loaded in Python, where the spliced/unspliced counts matrix and the Seurat metadata were integrated. RNA velocity was computed using the scVelo dynamical model. The above-described method was also employed for RNA velocity analysis of stages HH3, HH4, and HH6 Hensen’s node-integrated dataset. Single-cell RNA-seq Seurat objects from the three developmental stages were integrated, and the batch effect was corrected using the Seurat package. Trajectory analysis was also performed using Monocle 3 (v1.4.25) to reconstruct developmental progression across stages HH3, HH4, and HH6. The Seurat object containing integrated single-cell RNA-seq data was converted to a Monocle object. Dimensionality reduction and clustering results from Seurat were transferred to ensure consistency between analyses. Bootstrap-based cluster stability was performed to assess the robustness of clustering across developmental stages (HH3, HH4, and HH6). A total of 200 bootstrap iterations were conducted, each time randomly sampling 1750 cells from each developmental stage from the integrated Seurat object. Each subsample was clustered independently, and the cluster assignments were compared to the original clustering using the Adjusted Rand Index (ARI). The ARI distribution (median ≈ 0.222) indicates moderate but consistent cluster reproducibility.

Enhancer cloning

The CHRD enhancer was identified from ATAC-seq datasets generated by the Simoes-Costa lab in stage HH4 Hensen’s node tissue. A region of high accessibility in the CHRD locus was amplified from HH10 chicken genomic DNA and cloned into the pTK-mCherry reporter vector⁷⁰. The DNA fragment was amplified using primers with overlapping ends to facilitate cloning via Gibson assembly (Invitrogen #A46627). Primers used: Fw - tcttacgcgtgctagcccACCGGACATGCAGGCAAATA and Rv - atcgcagatctcgagcccCACGTAGCCCCTCACAGC. When tested in-vivo, this element drove specific reporter activity in the notochord, consistent with the endogenous CHRD expression pattern. A cis-regulatory element (CRE) for TCF15 previously described by Mok and colleagues⁵⁰ was amplified and cloned into the pTK-mCherry reporter vector as described above. This CRE drives reporter activity in presomitic mesoderm, somites, and lateral plate mesoderm. The following primers were used to clone the DNA fragment: FW - tcttacgcgtgctagcccTCTCCAGCTCTGTTCAAGGG and RV - atcgcagatctcgagcccGCAGTGGAAAATCCAAGTCCC. SOX2 enhancers N1 and N2, previously described by Uchikawa and colleagues⁴⁹, were also cloned into pTK-GFP⁴⁹ and pTK-mcherry⁷⁰ expression vectors.

Embryo preparation and enhancer electroporation

Wild-type host embryos at stages HH3 + /HH4 were collected using the filter paper-based easy-culture method⁶⁸. CHRD and TCF15 enhancer plasmids were transfected into the embryos by ex-ovo electroporation. Briefly, 1-2 μg/μL of the expression vector was injected into the space between the epiblast tissue and vitelline membrane to cover the entire embryo. Embryos were electroporated with platinum electrodes using the following parameters: 5 pulses of 5 V, 50 ms on, 100 ms off. Electroporated host embryos were then transferred to 35 mm Petri dishes containing Bacto-agar+albumen (1:1) substrate supplemented with 10% glucose and were kept at 37 °C until use. Bacto-agar solution adjusted to 0.8%.

Transplantation of Hensen’s Node into electroporated hosts

Hensen’s nodes from stage HH6 GFP donor embryos were finely dissected and transplanted into the marginal zone between the area opaca and area pellucida of wild-type stage HH4 host embryos. These host embryos had been previously electroporated with either CHRD (n = 10/17) or TCF15 (n = 6/10) enhancers. To allow further development, embryos were cultured at 37 °C with 50% humidity for 16–18 h. Subsequently, embryos were fixed, stained with DAPI, and processed for imaging.

Immunofluorescence on transplanted embryos

Grafted embryos were fixed in 4% paraformaldehyde for 30 min at room temperature, followed by three washes with PBST (1X PBS + 0.5% Triton X-100) for 10 min each. Next, embryos were blocked and permeabilized with PBST supplemented with 3% normal donkey serum (NDS) and 1% bovine serum albumin (BSA) at room temperature for 1 h. After the blocking step, wild-type host embryos were incubated with anti-OTX2 antibody (1:300, R&D Systems #AF1979), and host embryos electroporated with TCF15 were incubated with anti-LEF1 antibody (1:200, Abcam #ab137872) diluted in blocking buffer overnight at 4 °C. The following day, embryos were washed three times with PBST for 15 min each and then incubated with secondary donkey anti-goat antibody conjugated with Alexa Fluor 647 (Molecular Probes) for 2 h at room temperature. After the secondary antibody incubation, embryos were washed, stained with DAPI, and mounted with fluoromount for imaging. Images were acquired using a Nikon AX R point scanning confocal microscope, also equipped with a 20x objective lens. In addition, embryos were imaged using the upright Zeiss Axio Zoom fluorescence microscope available in our lab.

Single-cell RNA-seq of induced secondary axes

GFP host embryos at stage HH3 + /HH4 were collected and cultured as previously described (see Embryo preparation and enhancer electroporation). Wild-type HH6 Hensen’s nodes were then dissected and grafted into the marginal zone between the area opaca and area pellucida of the GFP host embryos. Embryos were cultured for 16–18 h to allow further development. Next, secondary axes (n = 5) induced by the graft were micro-dissected and dissociated into a single-cell suspension as described above. scRNA-seq library preparation and sequencing were performed according to the manufacturer’s instructions (10x Genomics). Libraries were prepared using the Illumina NextSeq 500/550 High Output kit v2.5 (75 Cycles) and sequenced using the NextSeq 500/550 instrument at the Boston Children’s Hospital Molecular Genetics Core facility. The final library was paired-end sequenced at (28 × 10 ×10 × 44).

Single-cell RNA-seq data processing of secondary axes

To identify the GFP host cells, we modified the Gallus gallus GRCg6a.99 reference genome to include the eGFP sequence. The FASTq files were then aligned to this modified reference genome. The filtered expression matrix was analyzed using the R package Seurat. Low-quality cells with fewer than 200 genes and a mitochondrial read percentage above 5% were removed, leaving a total of 8249 cells. The data was normalized using the scale factor = 10000. Highly variable-expressed genes for downstream analysis were identified using the FindVariableFeatures function. The Seurat pipeline was employed for linear dimensional reduction and unsupervised clustering. The first 20 principal components were used to generate the UMAP plot. For clustering, the number of dimensions used to find the neighbors was set to 20, and the resolution was 0.3. DoubletFinder was then used to filter out 499 doublets. From the 7750 remaining singlets, 2372 GFP-positive cells were isolated, and differential expression analysis for each cluster was performed using the FindAllMarkers function. Cell identities were assigned to each population based on the expression of known marker genes found in their gene expression profiles.

Quantification of SOX2 enhancer activity following node grafts

Wild-type host embryos at stages HH3 + /HH4 were collected and electroporated with SOX2N1::GFP (N1-GFP) and SOX2N2::mCherry (N2-mChe) enhancer constructs at a final concentration of 1 µg/µL, as previously described (see Embryo preparation and enhancer electroporation). Following electroporation, stage HH4 and HH6 Hensen’s nodes were dissected from wild-type embryos and grafted into the marginal zone between the area opaca and area pellucida of the transgenic host embryos. Embryos were cultured for 15–16 h, then collected, fixed, and mounted for imaging. Grafted tissues were imaged using a Nikon AX R point scanning confocal microscope, also equipped with a 20x objective lens. Quantification of SOX2N1::GFP- and SOX2N2::mCherry-positive cells was performed using Fiji’s particle analysis tool for each individual embryo. A fluorescence threshold was first applied to distinguish labeled cells from the background signal. The watershed function was then used to separate adjacent cells, followed by segmentation and counting using the analyze particles function, which selected objects within the 3–10 µm size range. A binomial generalized linear model (GLM) analysis was performed to evaluate changes in the proportion of SOX2N1::GFP and SOX2N2::mCherry-labeled cells following stages HH4 or HH6 node transplantation. For each embryo, the proportion of labeled cells was calculated by dividing the number of GFP + or mCherry + cells by the total number of fluorescent cells. The total number of fluorescent cells per embryo was used as a weighting factor to account for variation in sample size across embryos. Predicted proportions and 95% confidence intervals were calculated on the logit scale and converted to probabilities using the inverse logit function. Statistical significance was determined from model coefficients, with p-values < 0.05 considered significant.

Co-culture of human ESC reporter line with chick Hensen’s node

RUES2-GLR⁵³, a human embryonic stem cell (hESC) reporter line containing SOX2-mCitrine, BRACHYURY-mCerulean, and SOX17-tdTomato, was kindly donated by the Brivanlou lab at The Rockefeller University. RUES2-GLR cells were cultured in 35 mm ibidi dishes with mTser plus medium for 24 h before the addition of the Hensen’s node. Hensen’s nodes at stage HH5 + /6 were dissected and washed in PBS, then added to the hESC culture. Co-cultured cells were incubated for 14 h, fixed with 4% paraformaldehyde for 15 min at room temperature, and then washed with 1X PBS and stained with DAPI. Images were acquired using a widefield inverted Nikon Ti2 fluorescence microscope with monochrome and color cameras.

Transplantation of Hensen’s node into naïve anterior epiblast

Embryos at stage HH3+ were collected using filter paper and cultured in Petri dishes containing Bacto-agar+albumen (1:1) substrate supplemented with 10% glucose. The entire primitive streak and surrounding epiblast were then removed using a tungsten needle, leaving the naïve anterior epiblast as the host tissue. The Hensen’s node from stage HH6 GFP embryos was dissected and grafted into the marginal zone of the previously dissected host epiblast. The tissue was cultured for around 16 h at 37 °C and 50% humidity and then fixed in 4% paraformaldehyde for 15 min at room temperature. For immunostaining, the samples were incubated with anti-SOX2 (1:200, Abcam #ab97959) and anti-BRACHYURY (1:300, R&D Systems #AF2085) antibodies. Images were acquired using a Nikon point scanner confocal AX R microscope with tunable GaAsP PMTs and resonant scanner at 20x magnification.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Conceptualization: T.Y.K. and M.S.C.; Methodology: T.Y.K., M.R. and M.S.C.; Investigation: T.Y.K.; Visualization: T.Y.K. and M.S.C.; Funding acquisition: M.S.C. and T.Y.K.; Project administration: M.S.C.; Supervision: M.S.C.; Writing: T.Y.K. and M.S.C.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data generated in this study is available in the Gene Expression Omnibus (GEO) repository under the accession number GSE228354, including raw data as well as processed/normalized files. Source data are provided in this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-63154-2.