In vitro induction of patterned branchial arch-like aggregate from human pluripotent stem cells – Nature Communications

Spontaneous induction of NCC-like cells in hPSC aggregates

To induce NCCs from hPSC, we cultured hPSC aggregates in a defined medium (NCC_ind medium) (Fig. 1a). Day-5 aggregates were enriched in SOX10⁺ NCC-like cells. However, they contained a small number of PAX6⁺ neuroepithelial cell (NEC) and CDH1⁺ non-neural epithelium (NNE) cells (Fig. 1b). Knockout serum replacement (KSR), which is often used in neural tissue induction, reduced the induction of NCC-like cells (Supplementary Fig. 1)^13,14. Therefore, we added B-27 supplement instead of KSR to improve cell viability and we analyzed the gene expression during differentiation (Fig. 1c). The hPSC marker NANOG was downregulated on the first day. One of the earliest NCC markers, PAX7, was also upregulated. Subsequently, the early NCC marker FOXD3 and the late marker SOX10 were upregulated. We confirmed the expression of these genes using immunohistochemistry (IHC) (Fig. 1d). Under our experimental conditions, we detected cells co-expressing all these markers in the peripheral region of the aggregates from day 3. The expression of FOXD3 and SOX10 mostly overlapped at every stage tested, although the expression of FOXD3 seemed to be weaker on day 5 than on day 4. In model animals, such as chicks and zebrafish, FOXD3 and SOX10 are expressed in distinct but partially overlapping patterns, and FOXD3 seems to be downregulated as development proceeds^15,16. Weaker expression of FOXD3 at day 5 may reflect the downregulation observed in vivo. The expression of PAX7 was broader than that of the others, especially on day 4 (Fig. 1d, e). FOXD3^–/SOX10^–/PAX7⁺ cells might be earlier NCC precursor-like cells or other types of cells, such as dorsal NECs¹⁷.

a Scheme of the differentiation culture for the induction of NCC-like cells. b Immunostaining of the day-5 aggregate with antibodies for SOX10 (green), PAX6 (red), and CDH1 (blue). Scale bar, 100 µm. Three independent cultures were used for experiment and representative images are shown. c Real-time polymerase chain reaction analysis of time-course samples. Day-0 samples were undifferentiated hPSCs and used for normalization. Data are presented as mean ± standard error of mean (n = 3 independent experiments). Source data are provided as a Source Data file. d Immunostaining of the aggregates at days 3, 4, and 5 with antibodies for FOXD3 (green), PAX7 (red), and SOX10 (blue). Scale bar, 100 µm. Three independent cultures were used for experiment and representative images are shown. e Magnifications of areas surrounded by dotted lines in images of day-4 aggregates in (d). Scale bar, 20 µm.

Full size image

As day-5 aggregates contained non-NCC-type cells (Fig. 1b), we analyzed the content of aggregates derived from KhES-1 cells by using single-cell RNA-seq (scRNA-seq). A total of 1877 cells were divided into 14 clusters using unsupervised uniform manifold approximation and projection (UMAP) clustering (Fig. 2a, b). Clusters 0, 1, 3, 4, 5, 6, and 7 represented SOX10⁺ NCC-like cells, accounting for 71.9% of the total cells (Fig. 2c, d). The expression of PAX7 and FOXD3 was higher in cluster 5 than in the NCC-like clusters, indicating that cluster 5 represented nascent NCC-like cells (Fig. 2e, g). The presence of PAX7^high/FOXD3^high cells was confirmed using IHC (Fig. 1d). Cluster 8, adjacent to cluster 5, represented SP5⁺/PAX7 ⁺/FOXD3⁺ NPB-like cells (Fig. 2f, g)¹⁸. Cluster 8 was also characterized by extremely low SOX10 expression (Fig. 2b, d, g). Cluster 9 represented more immature cells, characterized by higher POU5F1 expression and the absence of SOX17 expression, a marker gene for endodermal cells (Endo) (Fig. 2h)¹⁹. Cluster 10 contained CDH1⁺/TFAP2A⁺ NNE-like cells, which also contained SIX1⁺ pre-placodal epithelium (PPE)-like cells (Fig. 2i)²⁰. Three types of neural lineage clusters were observed (Fig. 2j). Cluster 2 appeared to represent ISL1⁺/SIX1⁺ NCC-derived neuron-like cells (Fig. 2i, j)²¹ and cluster 11 contained PAX6⁺/EMX2⁺ NEC-like cells (Fig. 2j)²². Cluster 13 comprised EBF2⁺ placode-derived neuron-like cells (Fig. 2j)²³. Cluster 12 represented SOX17⁺ endo-like cells (Fig.2h). We also confirmed the presence of these cell types using IHC (Supplementary Fig. 2).

a Uniform manifold approximation and projection of scRNA-seq data from the day-5 aggregates. Cluster number was determined according to the cluster size. Endo endoderm, NCC neural crest cell, NEC neuroepithelial cell, NNE/PPE non-neural ectoderm/pre-placodal epithelium, NPB neural plate border. b Dot plot of key marker genes to define each cluster. c Ratio of the annotated populations in the day-5 aggregates. d–f Feature plots of genes which characterize NCC-like cells and their progenitors g Violin plots of SOX10, PAX7, FOXD3, and SP5 (h–j) Feature plots of genes characterize immature cells, endo-like cells, NNE/PPE-like cells, NCC-derived neuron-like cells, NEC-like cells and placode-derived neuron-like cells.

Full size image

Although p75 is not specific to NCCs in human embryos²⁴, scRNA-seq revealed that p75 expression was mostly specific to NCC-like cells in day-5 aggregates (Supplementary Fig. 3a). IHC also showed that p75 expression was specific to SOX10⁺ cells (Supplementary Fig. 3b, c). Almost consistent with scRNA-seq data, 76.5% of the cells in day-5 aggregates derived from KhES-1 expressed p75 at a high level on average (Supplementary Fig. 3d). Although the induction efficiency of SOX10⁺ cells varied among the hPSC lines, the expression of p75 appeared to be specific to SOX10⁺ cells in all cell lines (Supplementary Fig. 3b–d). These results indicate that p75 could be a suitable surface antigen for isolating NCC-like cells in this experimental system. We also tested the multipotency of these NCC-like cells (Supplementary Fig. 4). These cells can differentiate into neuronal cells, glial cells, melanoblasts, and ectomesenchymal derivatives, including smooth muscle cells, osteoblasts, chondroblasts, and adipocytes.

In our method, NCC-like cells were spontaneously induced without additives, such as CHIR-99021, BMPs, and SB-431542 (SB) that are often used in other methods^4,5,6,7, indicating the importance of aggregate-intrinsic signals. We analyzed the role of aggregate-intrinsic WNT/BMP signals, considering their importance in vivo (Fig. 3a)⁸. WNT1/4 were expressed in aggregates and peaked on days 3 and 4. AXIN2, a well-known downstream target of the canonical WNT signal²⁵, also peaked around days 3 and 4. A non-canonical WNT, WNT7B, was upregulated on day 1. BMP2 was also upregulated around day 3, whereas BMP4 was upregulated between days 0 and 1, suggesting that BMP4 may play a role in the early phase of differentiation. To reveal the function of WNT/BMP signals in our method, we treated the aggregates with the WNT inhibitor IWP-2 or the BMP inhibitor dorsomorphin (Fig. 3b). We added these inhibitors from days 1, 2, or 3 to day 5 and analyzed the expression of marker genes for NCC-like cells (SOX10), NEC-like cells (PAX6), NNE-like cells (CDH1), and PPE-like cells (EYA1) (Fig. 3c). WNT inhibition starting on day 1 strongly suppressed SOX10 (Fig. 3c). This effect was weakened when IWP-2 was added on day 2. Treatment from day 3 showed almost no effect, suggesting that WNT signaling before day 3 was important for NCC-like cell induction. However, canonical WNT activity was upregulated from day 3, suggesting the importance of the early expression of non-canonical WNTs, such as WNT7B. BMP inhibition on day 1 also suppressed SOX10 expression, although this suppression was incomplete. This indicates that BMPs, such as BMP4, might act from day 0, and that their activities were sufficient to induce the differentiation of subsets of NCC-like cells. BMP inhibition on day 3 was also ineffective on SOX10 expression. For PAX6, CDH1, and EYA1, both WNT and BMP inhibition from day 3 had almost no effect, indicating that the fate of most precursors was determined by day 3 (Fig. 3c). WNT inhibition from days 1 and 2 slightly upregulated CDH1 and EYA1, indicating an increase in the number of NNE/PPE-like cells. Conversely, BMP inhibition on day 1 significantly reduced the expression of CDH1 and EYA1, indicating a reduction in the number of NNE/PPE-like cells. BMP inhibition from day 1 also resulted in the upregulation of PAX6, indicating an increase in the number of NEC-like cells, which was consistent with a reduction in NNE/PPE marker expression. Interestingly, WNT inhibition from days 1 and 2 also upregulated PAX6, although NNE/PPE markers were also upregulated under these conditions. We analyzed the aggregates treated with inhibitors from day 1 using IHC (Fig. 3d, e). Consistent with the results of real-time polymerase chain reaction (PCR), an increase in the number of PAX6⁺ cells and a decrease in those of SOX10⁺ and CDH1⁺ cells were observed in the dorsomorphin-treated aggregates. In contrast, SOX10⁺ cells disappeared completely from the IWP-2-treated aggregates (Fig. 3d). In IWP-2-treated aggregates, the counts of PAX6⁺ cells were increased, as well as in dorsomorphin-treated aggregates, but these PAX6⁺ cells expressed CDH1, suggesting that they were of the NNE/PPE-like lineage. PAX6 is also expressed in the anterior PPE, such as the lens/olfactory placodes²⁶. We confirmed that there were SIX1⁺/CDH1⁺ cells that did not express PAX6 in non-treated aggregates, while many SIX1⁺/CDH1⁺ cells co-expressed PAX6 in IWP-2-treated aggregates, suggesting that they became anterior PPE-like cells (Fig. 3e). Anteriorization by WNT inhibition is consistent with the role of WNT signals in the posteriorization of embryonic ectodermal tissue during early development²⁷. For dorsomorphin-treated aggregates, there were hardly any CDH1⁺ NNE/PPE-like cells. These results indicate that aggregate-intrinsic WNT and BMP signals regulate the positional identity of cells to induce NCC-like cells (Fig. 3f).

a Real-time polymerase chain reaction (PCR) analysis of WNT/BMP and downstream genes in time-course samples. Day-0 samples were undifferentiated hPSCs. Data are presented as mean ± standard error of mean (SEM) (n = 3 independent experiments). Source data are provided as a Source Data file. b Scheme of the WNT/BMP inhibition experiment. IWP-2 or dorsomorphin was added on days 1, 2, and 3. Aggregates were analyzed by real-time PCR and immunohistochemistry on day 5. c Real-time PCR analysis of inhibitor-treated aggregates on day 5. Non-treated aggregates were used as control. Data are presented as mean ± SEM (n = 3 independent experiments). Two-tailed Student’s t-test was used. Source data are provided as a Source Data file. d Immunostaining of inhibitor-treated aggregates on day 5 with antibodies for SOX10 (green), PAX6 (red), and CDH1 (blue). Scale bars, 100 µm and 25 µm (magnifications). Three independent cultures were used for experiment and representative images are shown. e Immunostaining of inhibitor-treated aggregates on day 5 with antibodies for PAX6 (green), SIX1 (red), and CDH1 (blue). Scale bars, 100 µm and 25 µm (magnifications). Three independent cultures were used for experiment and representative images are shown. f Scheme representing a hypothesis about the effect of WNT/BMP inhibition. The left illustration indicates the spatial organization of each cellular domain in the early embryo. WNT and BMP forms gradients along anterior-posterior and medial-lateral axes, regulating positional information in the embryo. The right illustration indicates the hypothetical scheme of the effects of WNT/BMP inhibition in vitro. hPSC-derived aggregates generate NCC-like cells preferentially in NCC_ind medium in intrinsic WNT and BMP-dependent manner as default state. WNT inhibition may change the positional information of the aggregate into more anterior and medial identity and result in the induction of the anterior PPE-like cells. BMP inhibition may induce more medial positional identity and result in the induction of neural cells. NEC neuroepithelial cell, NNE non-neural ectoderm, and PPE pre-placodal epithelium.

Full size image

SB, which antagonizes activin/TGFß signal, was also used in some NCC induction methods^4,6. We tested the effect of SB on our NCC-like cell induction method by adding SB from day 0 onward. In SB-treated aggregates, the number of NCC-like cells expressing FOXD3 and SOX10 increased, while SOX17⁺ endo-like cells disappeared on day 4 (Supplementary Fig. 5). This result is consistent with the finding that SB inhibits the endodermal induction of hPSCs while facilitating neuroectodermal differentiation²⁸.

Recapitulation of in vivo differentiation process during NCC-like cell induction

We appended the scRNA-seq data of day-2, day-3, and day-4 aggregates to the data of day-5 aggregates for further analysis of the developmental process of the aggregate (total 8466 cells, 23 clusters) (Fig. 4a). The cells in the cluster 15 were POU5F1⁺/NANOG⁺/EOMES^–, indicating that they exhibited a PSC-like state (Fig. 4b)²⁹. The early NPB marker TFAP2A, which can induce the expression of NCC marker genes when overexpressed in hPSCs³⁰, was upregulated in cluster 7 adjacent to cluster 15, and its expression continued throughout differentiation into NCC-like and NNE/PPE-like cells (Fig. 4c). The NPB markers, SP5 and IRX1, were upregulated in groups of clusters adjacent to cluster 7 (clusters 3, 6, 8, 9, 13, 14, 16, and 19) (Fig. 4d)^18,31. Another group of clusters (clusters 1, 2, 4, 5, 11, 17, and 20) contained SOX10⁺ NCC-like cells (Supplementary Fig. 6). In this group, some marker genes showed differential expression patterns, suggesting variations in cell state (Supplementary Fig. 6). Cluster 0 was the transition point between the NPB-like and NCC-like clusters expressing FOXD3 and SOX9 at high levels (Fig. 4e and Supplementary Fig. 6). Cluster 10 contained SIX1⁺ NCC-derived neuron-like cells (Fig. 4f). Clusters 12/18 contained KRT18 ^high NNE/PPE-like cells, and cluster 22 contained PAX6⁺ NEC-like cells (Fig. 4f). Cluster 21 contained SOX17 ⁺ endo-like cells (Fig. 4g). RNA velocity analysis suggested that most of the hPSCs differentiated into NPB-like states. However, only a small number of cells differentiated into SOX17⁺ endo-like cells (Fig. 4h). NPB-like cells seemed to differentiate into NCC-like-, NEC-like-, and NNE/PPE-like cells, suggesting that NPB-like cells are multipotent, as observed in model animals (Fig. 4h)^32,33,34.

a UMAP (Uniform manifold approximation and projection) of merged scRNA-seq data (day 2 to day 5). Individual cells are colored by cluster number or sampling date. Endo endoderm, NCC neural crest cell, NEC neuroepithelial cell, NNE/PPE non-neural ectoderm/pre-placodal epithelium, NPB neural plate border, PSC pluripotent stem cell. b–g Feature plots of genes associated with PSC-like cells, endoderm-like cells, NPB-like cells, NNE/PPE-like cells, NEC-like cells. h RNA velocities visualized as streamlines on UMAP. i UMAP of scRNA-seq data colored by inferred pseudotime without clusters 10, 18, 21, and 22. j Expression dynamics of PSC, NPB, and NCC-related genes along pseudotime.

Full size image

NPB is located between SOX2⁺ NE and TFAP2A⁺ NNE, and both genes are expressed in model animals³². In our scRNA-seq dataset, we observed the co-expression of SOX2 and TFAP2A NPB-like clusters at all stages analyzed (Supplementary Fig. 7a). We also confirmed co-expression in the aggregates using IHC (Supplementary Fig. 7b). Co-expression of SOX2 and TFAP2A clearly appeared on day 2. However, TFAP2A was already expressed in some SOX2⁺ cells weakly at day 1. From day 4, SOX2 expression decreased in most TFAP2A⁺ cells, including NPB-like, NCC-like, and NNE/PPE-like cells. This downregulation of SOX2 has been observed in the NCCs of chick embryos³². Some SOX2⁺/TFAP2A⁺ cells also co-expressed PAX7 from day 3, similar to the NPB cells in chick embryos (Supplementary Fig. 7b).

We analyzed the NPB clusters in more detail to identify the variety in the cell state. NPB clusters (clusters 3, 6, 7, 8, 9, 12, 13, 14, 16, and 19) were extracted according to the expression of SP5 and TFAP2A and re-clustered to identify differentially expressed genes (Supplementary Fig. 8a). SP5 and IRX1 were broadly expressed. The cranial NCC (cNCC) markers, LHX5, DMBX1, and OTX2, were also expressed in a large number of cells, although the expression patterns of DMBX1 and OTX2 were sparse and restricted, compared with LHX5, respectively. Lhx5 and Dmbx1 are regulators of the cNCC-specific gene regulatory network in chick embryo³⁵. Otx2 is crucial for head development³⁶. These results suggest that most NPB-like cells possess a cNCC-like identity.

The NPB-like cell clusters could be roughly classified into two groups (Supplementary Fig. 8c, d). One group expressed genes related to the anterior patterning of the embryonic ectoderm, such as CYP26A1 and HESX1^37,38. This group mainly comprised cells of early differentiation days (day 2: 85.0%, day 3: 14.6%, day 3: 0.4%, and day 4: 0%). In the other group, NPB-specified PAX3 and MSX1 were broadly expressed instead of CYP26A1 and HESX1. This group comprised cells from late differentiation days (day 2: 13.8%, day 3: 26.9%, day 3: 29.6%, and day 4: 29.7%). This result suggests that early anterior patterning of the ectoderm occurred in the aggregates, followed by further differentiation of NPB-like cells.

We also searched for genes with distinct expression patterns within NPB-like cell clusters, focusing on transcription factors (Supplementary Fig. 9). For example, FOXD3, ZFHX4, TFAP2B, DACH1, and EBF1 are expressed in cells located on the left side of the UMAP. These genes are also expressed in NCCs in vivo^39,40,41,42. PAX6, EN1, PAX2, PAX7, OTX2, GBX2, and EMX2, which are essential factors for the patterning of NECs around the diencephalon and mid/hindbrain, also exhibited distinct patterns^22,43, suggesting those factors may define the axial identity of NPB-like cells in the aggregates. DLX5, GRHL2, GRHL3, MAFB, and HAND1 also showed characteristic expression patterns in NPB-like cells, and their expression extended to NNE/PPE-like cell clusters. These genes are involved in the NNE/PPE lineage development in vivo^44,45,46,47. The tumor suppressor ELF3 is also expressed in similar patterns, although its role in NPB-like cells is unknown⁴⁸.

Furthermore, a trajectory analysis of NPB-like cells was performed to identify gene expression patterns specific to the trajectories bound for each ectodermal lineage (Supplementary Fig. 10a). Some genes exhibited differential expression patterns among the different lineages (Supplementary Fig. 10b). These genes included those known to be important for the development of each lineage, as earlier mentioned. There were also several lineage-specific genes whose roles in the differentiation of NPB-like cells were unknown, such as EBF1 and SEZ6 for the NCC-like lineage and ELF3, KLF6 for the NNE/PPE-like lineage. For example, EBF1 is known for its expression in migratory cNCC, but its function in NPB is unknown⁴¹, similar to that of ELF3.

To infer the gene expression dynamics during the differentiation of NCC-like cells, pseudotime analysis was performed on clusters, excluding NNE/PPE-like, NEC-like, and endo-like cell clusters (Fig. 4i). Markers for PSC, NPB, and NCC were upregulated along with pseudotime, indicating the recapitulation of the in vivo NCC induction processes to some extent (Fig. 4j). During in vivo development, EMT is associated with the induction of NCCs as they leave their origin. EMT is characterized by several features, such as the expression of EMT-inducing transcription factors, changes in the expression of cadherin genes, and degradation of the basement membrane⁴⁹. Gene ontology analysis of cluster markers revealed that in marker genes for cluster 0 of the merged scRNA-seq data, which was a junction between NPB and NCC-like cells, were enriched with genes related to EMT (i.e., GO:0001837 “epithelial to mesenchymal transition”) (Supplementary Fig. 11). Pseudotime analysis revealed temporal changes in the expression of several genes related to EMT, including transcription factors and cadherins (Fig. 5a). Consistent with their regulatory relationships, the expression of PAX3/7 and MSX1 was higher than that of FOXD3, ETS1, and SNAI2 ⁵⁰. Furthermore, the downregulation of CDH2 coincided with the expression of SNAI2 and LMO4⁵⁰. Changes in cadherin expression in the aggregates were analyzed by IHC. Epithelial cadherin CDH1 was expressed in the aggregates almost entirely on day 2 and then gradually restricted as differentiation proceeded (Fig. 5b). Most CDH1^– regions were occupied with FOXD3⁺ cells. SOX10⁺ cells in day-4 aggregates expressed the EMT-related cadherin CDH6 with a complementary pattern to CDH1, indicating the switching of cadherins during NCC-like cell differentiation (Fig. 5c). These results suggest that the in vitro induction of NCC-like cells may recapitulate the in vivo developmental processes to some extent (Fig. 5d).

a Expression dynamics of EMT-related genes along pseudotime. b Immunostaining of aggregates with antibodies for FOXD3 (green) and CDH1 (red), and counterstaining with DAPI (blue). FOXD3 and CDH1 were expressed in complementary manner. Scale bar, 100 µm. Three independent cultures were used for experiment and representative images are shown. c Immunostaining of aggregates with antibodies for CDH1 (green), CDH6 (red), and SOX10 (cyan). CDH1 and CDH6 were expressed in complementary manner. SOX10⁺ cells expressed CDH6 instead of CDH1. Scale bar, 100 µm. Three independent cultures were used for experiment and representative images are shown. d Schematic representation of the presumptive process of NCC (neural crest cell)-like cell induction from hPSCs (human pluripotent stem cells). hPSCs bifurcate into NPB (neural plate border)-like cells or Endo (endoderm)-like cells at first. Then NPB-like cells differentiate into NNE/PPE (non-neural ectoderm/pre-placodal epithelium)-like cells, NEC (neuroepithelial cell)-like cells, and NCC-like cells. NCC-like cell induction depends on aggregate-intrinsic WNT/BMP (bone morphogenetic protein) signals and is associated with EMT-like gene expression changes.

Full size image

Induction of branchial arch mesenchyme-like aggregates with specific regional identity

We tested the potential of these aggregates as in vitro models of early craniofacial development. As earlier mentioned, the expression of DMBX1, LHX5, OTX2 (Supplementary Fig. 8b), and ETS1 (Fig. 5a) suggested that the induced NCC-like cells possessed a cranial identity (Fig. 5a, Supplementary Fig. 8b)^35,36. cNCCs migrating into future craniofacial regions are HOX-negative in vivo⁵¹. We were unable to detect HOX-positive NCC-like cells in the aggregates, indicating that the cells in the aggregates possessed the ability to differentiate into the craniofacial mesenchyme (Supplementary Fig. 12).

We performed a long-term culture of the aggregates to induce craniofacial mesenchyme-like cells. The addition of basic fibroblast growth factor (bFGF) from day 4 enabled long-term culture, preventing the collapse of the aggregates during the culture period (Fig. 6a). bFGF-treated aggregates expressed the craniofacial mesenchyme markers PRRX1 and TWIST1 at higher levels than untreated aggregates (Fig. 6b)⁵². IHC of day-28 aggregates revealed that many cells expressed these genes, indicating the induction of craniofacial mesenchymal-like cells (Fig. 6d, e). The branchial arch markers, DLX2 and BARX1, were also expressed in day-28 aggregates (Fig. 6b, e, f)⁵³. These aggregates also expressed LHX6, a specific marker expressed on the oral side of the first branchial arch mesenchyme in a spatially restricted manner, as observed in vivo (Fig. 6b, e)⁵⁴. The first branchial arch is composed of the MX and MN, which give rise to the upper and lower jaws, respectively. bFGF-treated aggregates expressed POU3F3, which was strongly expressed in early MX⁹, but did not express the MN marker DLX5 (Fig. 6b and Supplementary Fig. 13), indicating that bFGF-treated aggregates acquired an MX-like identity (Fig. 6c).

a Scheme of the long-term culture and bright-field images of the aggregates. Aggregates were cultured with or without bFGF (basic fibroblast growth factor) from day 4. Scale bar, 400 µm. NCC: neural crest cell. Three independent cultures were used for experiment and representative images are shown. b Real-time PCR analysis of craniofacial mesenchyme-related genes. Data are presented as mean ± SEM (n = 3 independent experiments). Two-tailed Student’s t-test was used. Source data are provided as a Source Data file. c Schema of gene expression pattern in the first branchial arch. MN mandibular arch, MX maxillary arch. d Immunostaining of bFGF-treated aggregates on day 28 with antibodies for TWIST1 (green) and counterstaining with DAPI (blue). Scale bars, 200 µm. Three independent cultures were used for experiment and representative images are shown. e Immunostaining of bFGF-treated aggregates on day 28 with antibodies for DLX2 (green), LHX6 (red), and PRRX1 (blue). Scale bars, 100 µm. Three independent cultures were used for experiment and representative images are shown. f Immunostaining of bFGF-treated aggregates on day 28 with antibodies for BARX1 (green) and counterstaining with DAPI (blue). Scale bars, 100 µm. Three independent cultures were used for experiment and representative images are shown.

Full size image

The choice of fate between MX and MN is regulated by the peptide hormone EDN1^10,55. EDN1 is secreted from the epithelium surrounding the MN and is essential for the induction of MN specifiers, such as DLX5/6 (Fig. 7a). Loss of the EDN1 signal results in the transformation of the lower jaw into an upper jaw-like structure¹⁰, whereas overactivation transforms the MX into an MN-like identity⁵⁵. However, the NCC-like cells in the aggregates expressed the EDN1 receptor EDNRA (Fig. 7b), and the aggregates comprised only a small number of NNE-like cells, a possible source of EDN1. We hypothesized that low levels of EDN1 in the aggregates would result in the acquisition of an MX-like identity during long-term culture. Therefore, we cultured the aggregates with or without EDN1 and BMP4 from day 4 in the presence of bFGF (Fig. 7c). BMP4 is also secreted from the craniofacial epithelium and is important for the patterning of MN¹¹. When EDN1 and BMP4 were added, POU3F3 was downregulated, whereas DLX5 was upregulated (Fig. 7d). DLX5 expression was higher in the aggregates treated with both EDN1 and BMP4 than in those treated with either EDN1 or BMP4 alone. However, the cooperative effect of EDN1 and BMP4 on DLX5 expression has not been reported previously.

a Schema of the expression of EDN1 (endothelin 1) and DLX5/6 in first branchial arch. MX: maxillary arch. b Feature plot of EDNRA in scRNA-seq data. c Scheme of long-term culture of aggregates. Aggregates were cultured with or without EDN1 and BMP4 (bone morphogenetic protein 4) in the presence of bFGF (basic fibroblast growth factor) from day 4 and analyzed by real-time PCR and immunohistochemistry at day 14. NCC: neural crest cell. d Real-time PCR analysis of POU3F3, and DLX5, which are expressed in MX and MN, respectively. Data are presented as mean ± SEM (n = 4 independent experiments). Two-way ANOVA with Tukey–Kramer test was used. Source data are provided as a Source Data file. e Schema of the spatial organization of domains in the first branchial arch. f Real-time PCR analysis of the region-specific genes. Data are presented as mean ± SEM (n = 4 independent experiments). Two-way ANOVA with Tukey–Kramer test was used. Source data are provided as a Source Data file. g Hypothetical representation of the effect of EDN1 and BMP4 on regional identity of the aggregates. The aggregates switched their identity from the MX-like to the MN-like, when treated with EDN1 and BMP4 during the long-term culture. h Immunostaining of non-treated aggregates and EDN1/BMP4-treated aggregates on day 14 with antibodies for DLX5 (green), GSC (red), and DLX2 (blue). Scale bar, 100 µm. i Immunostaining of non-treated aggregates and EDN1/BMP4-treated aggregates on day 14 with antibodies for POU3F3 (green), HAND2 (red), and DLX2 (blue). Scale bar, 100 µm. Three independent cultures were used for experiment and representative images are shown.

Full size image

The MN is divided into several subdomains (Fig. 7e). We analyzed the expression of domain-specific genes to characterize the MN-like cells induced by EDN1/BMP4 (Fig. 7f). EDN1/BMP4-treated aggregates expressed the distal region markers HAND2 and MSX2. However, MSX2 expression was upregulated when BMP4 was added alone. EDN1/BMP4-treated aggregates also expressed the intermediate region markers DLX3 and NKX3.2⁵⁶. Regarding the oral-aboral axis, the aboral marker GSC was upregulated in the EDN1/BMP4-treated aggregates, whereas the oral side marker LHX6 was downregulated. Furthermore, in the EDN1/BMP4-treated aggregates, WNT5A, which regulates the proper morphogenesis of MN, was also upregulated⁵⁷. We confirmed the expression of these markers in EDN1/BMP4-treated aggregates using IHC (Fig. 7h, i). These results suggested that the aggregates were similar to the in vivo branchial arch mesenchyme in their responsiveness to signaling factors, and that EDN1 and BMP4 cooperatively induced the identity of the aboral side of the distal, intermediate regions of the MN (Fig. 7g).

Formation of the patterned branchial arch-like aggregates through temporary treatment

The MX and MN can be recognized as the dorsal and ventral sides of the same tissue, as the most dorsal side is POU3F3⁺ MX and the ventral side is HAND2⁺ region of the MN (Fig. 8a). At day14, in bFGF-treated aggregates that differentiated into an MX-like state, most NCC cells still expressed SOX10 (Supplementary Fig. 14). In contrast, in EDN1/BMP4-treated aggregates that differentiated into an MN-like state, most mesenchymal cells expressed DLX5 instead of SOX10, indicating fate commitment into the MN-like lineage. However, in these aggregates, a subset of cells still expressed SOX10, suggesting that they were not committed to the MN-like lineage and could differentiate into the MX-like state if EDN1 and BMP4 were removed from the medium. We tested whether single aggregates contained both POU3F3⁺ (dorsal MX-like) and HAND2⁺ (ventral MN-like) cells by removing EDN1 and BMP4 from the medium on day14 (Fig. 8b). The temporarily treated aggregates (condition 2 in Fig. 8b) expressed both POU3F3 and HAND2 at higher levels than the EDN1/BMP4-treated (condition 3 in Fig. 8b) and non-treated aggregates (condition 1 in Fig. 8b) on day 21(Fig. 8c). Furthermore, POU3F3⁺ and HAND2⁺ cells formed spatially separated domains in temporarily treated aggregates, indicating mutually exclusive fate choice and self-patterning properties of cells (Fig. 8d).

a Schema of the axial organization of the 1st BA. The proximal (dorsal) side is POU3F3⁺ MX (maxillary arch), and the distal (ventral) side is HAND2⁺ distal MN (mandibular arch). b Schematic of experiment. The duration of treatment with EDN1 (endothelin 1) and BMP4 (bone morphogenetic protein 4) differed among conditions 1, 2, and 3. The aggregates were analyzed using real-time PCR and immunohistochemistry on day 21. bFGF: basic fibroblast growth factor. c Immunostaining of the day-21 aggregates cultured in each condition with antibodies for POU3F3 (green), HAND2 (red), and DLX2 (blue). Scale bars, 100 µm (the left one) and 200 µm (the others). Three independent cultures were used for experiment and representative images are shown. Cond. 1–3 condition 1–3. d Real-time PCR analysis of the region-specific genes. Data are presented as mean ± SEM (n = 3 independent experiments). One-way ANOVA with Tukey–Kramer test was used. Source data are provided as a Source Data file. e Scheme for differentiation of first branchial arch-like aggregates. When cultured with bFGF alone, the aggregates differentiated into MX-like states, whereas they differentiated into MN-like states in the presence of EDN1 and BMP4. Temporary treatment with EDN1 and BMP4 induced the formation of single aggregates containing both POU3F3⁺ Mx-like and HAND2⁺ Mn-like regions.

Full size image

We analyzed previously reported scRNA-seq data from the mouse branchial arch on embryonic day 9.5 (E9.5) as a reference⁵⁸. Hox-negative cells were extracted from NCC clusters and re-clustered (Supplementary Fig. 15a). There were clusters positive for the MN markers Dlx5 and Hand2 (clusters 1, 2, and 7) and clusters positive for the MX arch marker Pou3f3 (clusters 0, 3, 5, and 6). Several presumptive markers for these clusters were identified (Supplementary Fig. 15b). On day 21, EDN1/BMP4-untreated aggregates expressed the presumptive MX arch markers EBF1 and POU3F1 at higher levels, and EDN1/BMP4-treated aggregates expressed the presumptive MN marker DLK1 at higher levels (Supplementary Fig. 15c). Temporarily treated aggregates expressed high levels of both markers. The scRNA-seq of E9.5 mouse embryo showed that Pou3f3 and Hand2⁺ were expressed in a mutually exclusive manner (Supplementary Fig. 15d). IHC also showed the separation of Pou3f3⁺ and Hand2⁺ cells in the first branchial arch at E9.5, as in case of temporarily treated aggregates (Supplementary Fig. 15e).

As the branchial arch mesenchyme differentiates into craniofacial bones and cartilage in vivo¹, we analyzed the osteogenic/chondrogenic potential of the aggregates by extending the culture period (Fig. 9a). On day 35, temporarily treated aggregates expressed RUNX2, a marker for osteogenic/chondrogenic lineage cells (Fig. 9b)⁵⁹. Some RUNX2⁺ cells express SP7, an osteogenic cell marker, indicating their differentiation into this lineage⁵⁹. Furthermore, in the aggregates, there were cell condensates positive for SOX9 and NKX3.2, which were expressed in the Meckel’s cartilage of mouse embryos (Fig. 9c)^60,61. These results suggest spontaneous differentiation of NCC-like cells into osteogenic/chondrogenic lineages in culture.

a Schema of the long-term culture. After the wash out of EDN1 (endothelin 1) and BMP4 (bone morphogenetic protein 4), the aggregates were cultured in NCC_ind medium supplemented with only bFGF (basic fibroblast growth factor). NCC neural crest cell. b Immunostaining of the day-35 aggregates with antibodies for SP7 (green), RUNX2 (red), and counterstaining with DAPI (blue). Scale bar, 200 µm. Three independent cultures were used for experiment and representative images are shown. c Immunostaining of the day-35 aggregates with antibodies for NKX3.2 (green), SOX9 (red), and counterstaining with DAPI (blue). Scale bar, 200 µm. Three independent cultures were used for experiment and representative images are shown.

Full size image

• SEO Powered Content & PR Distribution. Get Amplified Today.
• PlatoData.Network Vertical Generative Ai. Empower Yourself. Access Here.
• PlatoAiStream. Web3 Intelligence. Knowledge Amplified. Access Here.
• PlatoESG. Carbon, CleanTech, Energy, Environment, Solar, Waste Management. Access Here.
• PlatoHealth. Biotech and Clinical Trials Intelligence. Access Here.
• Source: https://www.nature.com/articles/s41467-024-45285-0