Neural crest gene regulatory programs as drivers of development, diversification and disease

Abstract

The neural crest is an important stem cell population characterized by its multipotency, migratory behavior, and broad ability to differentiate into numerous derivatives throughout the vertebrate body, as diverse as cell types contributing to the cardiovascular system, craniofacial skeleton, peripheral nervous system, and pigmentation of the skin. The developmental trajectory of the neural crest is governed by a complex gene regulatory network (GRN) that mediates induction and specification at the neural plate border, emergence of neural crest cells from the neural tube, their migration through the periphery and cell fate determination en route to different final destinations. In this Review, we discuss the significant progress in investigating the neural crest GRN, which has increased our understanding of how neural crest cells impact vertebrate development and evolution, their role in adult tissue regeneration, and their contribution to diseases derived from abnormalities in neural crest cells.

Introduction

The evolutionary success of vertebrates is often attributed to the acquisition of the neural crest, a transient embryonic stem cell population. These cells originate within a broad stripe of ectoderm, known as the neural plate border, located between the prospective epidermis and neural plate, which will form the central nervous system. During neurulation, as the neural plate transforms into a cylindrical tube extending along the body axis, neural crest cells (NCCs) become specified within the dorsal aspect of the forming neural tube¹. They then undergo an epithelial-to-mesenchymal transition (EMT), enabling them to detach from the neural tube and migrate extensively throughout the body¹ (Fig. 1a). This remarkable migratory capacity, coupled with their broad developmental potential, enables NCCs to interact with diverse tissues and give rise to numerous cell types along the body axis. The multipotency of NCCs is driven by the transient reactivation or maintenance of a pluripotency gene regulatory network (GRN) modules, characteristic of early developmental stages^2–5. As a result, the neural crest is often referred to as the ‘fourth’ germ layer⁶ due to its critical role in diversifying the vertebrate body plan and producing a wide array of cell derivatives. Given its exceptional developmental versatility — surpassing the contributions typically attributed to any of the three primary germ layers — the neural crest behaves like a developmental hitchhiker, contributing cell types normally expected to arise from the mesoderm and, on occasion, even the endoderm^7,8. The entire neural crest developmental program from its induction to differentiation into distinct cell types is governed by a hierarchically organized GRN consisting of a unique combination of transcription factors and signaling molecules^9–13 (Box 1).

Figure 1 |. The neural crest developmental program.

a | The process of neural crest formation involves several key developmental milestones, including the establishment of the neural plate border (NPB), neural crest cell specification, delamination from the developing neural tube, migration to distant locations and differentiation into a variety of cell types. Each step is orchestrated by distinct subcircuits of the gene regulatory networks (GRN), which are shown in dashed rectangles. b |The axial identity of neural crest subpopulations, such as the cranial and cardiac neural crest, is governed by specific neural crest GRN subcircuits. c, d | Ectopic expression of key GRN subcircuit nodes in trunk neural crest cells (NCCs) can reprogram them into cranial-like (c) or cardiac-like (d) NCCs.

Box 1: Gene regulatory networks.

As the embryo develops from the one-cell stage to a complex body plan, its cells make critical fate decisions in an intricate progression orchestrated by gene regulatory networks (GRNs), which act as sophisticated control systems guiding cellular differentiation, tissue and organ formation. GRNs operate in a hierarchically organized manner, consisting of subcircuits — a small, specific subsets of regulatory interactions representing minimal functional units that perform discrete regulatory tasks. These subcircuits are formed through regulatory connections among specific nodes: individual units (such as a transcription factor, signaling effector, non-coding RNA, or epigenetic remodelers) that integrate signals from other units or external cues, and modulate the activity of other nodes in the network, either by activation or repression. Regulatory subunits are governed by clusters of cis-regulatory transcription factor binding sites. Rather than operating in a linear fashion, these subunits are interconnected^205–207 and organized into higher-level modules within the GRN that encompass multiple subcircuits, thereby enhancing GRN robustness and adaptability. The hierarchical position of regulatory changes within a GRN determines the nature of evolutionary alterations, which directly drive body plan modifications by altering the developmental regulatory program²⁰⁸. Indeed, GRN kernels, which represent evolutionarily conserved units within the network, are less affected by rapid evolution compared to other components, such as certain signaling interactions at the periphery of the network^209–211. In addition, the subdivision of the GRN into individual submodules — such as subcircuits, groups of transcription factors, or even signaling pathways — has profound implications for the emergence of morphological novelties. These submodules can be entirely or partially lost, or they may be redeployed to serve entirely different functions at different levels of organization, including organs, tissues or cell types, as happens with butterfly eye spots²¹², beetle horns²¹³, or treehopper helmets²¹⁴. GRNs thus provide a framework to understand how developmental processes shape evolutionary innovations, such as the neural crest and its derivatives, including the ‘New Head’ of vertebrates⁹⁵. The figure demonstrates key milestones in GRN and neural crest GRN research, including the pivotal studies^215,216 that served as a blueprint for this concept.

Over the past 15 years, advancements in the development of tools like CRISPR–Cas9 genome editing and single-cell biology (e.g. RNA-seq, assay for transposase-accessible chromatin with high-throughput sequencing (ATAC-seq), multiomics, machine learning, proteomics, metabolomics, embryonic stem cells (ESCs), induced pluripotent stem cells (iPSCs), high-resolution imaging, etc.) have revolutionized the study of GRNs. These innovations have enabled detailed analyses of regulatory networks, including that of the neural crest, and enabled predictions that can guide and facilitate experimental manipulation^14–18. These new tools have also deepened our understanding of what GRNs are and how they evolve, shedding light on their role in generating diverse cell types at various levels of the body axis, as well as their broader influence on evolution. In this Review, we provide an update of key innovations in the neural crest field, including how single-cell biology has provided a deeper understanding of cell lineage decisions and developmental trajectories during NCC migration and differentiation. We explore how new technologies have opened the door to studying neural crest development in a wide range of vertebrates that were previously inaccessible, providing insights into neural crest evolution. Finally, we discuss new insights that have emerged regarding the role of NCCs in disease progression.

Neural crest developmental program

Neural crest induction is initiated at the neural plate border (NPB) during gastrulation, driven by a complex network of signaling and transcriptional events. Key environmental signals shaping and refining the NPB include the WNT, fibroblast growth factor (FGF), NOTCH and bone morphogenetic protein (BMP) pathways. These signals originate from the non-neural ectoderm and underlying mesoderm underlying the neural plate¹⁹. Importantly, the neural plate border is established by intermediate levels of BMP²⁰, which are tightly controlled by BMP antagonists diffusing from surrounding tissues²¹. These signals in turn activate transcription of a core set of transcription factors known as ‘neural plate border specifiers’, including Pax3, Pax7, Zic1, Msx1, Msx2 or Gbx2, which further refine the NPB¹¹ (Fig. 1a). The expression patterns of NPB specifiers are highly conserved among vertebrates, implying that the mechanisms for NPB specification likely originated before the emergence of NCC-like cells. While non-vertebrate chordates express similar genes in the NPB, they lack bona fide NCCs and genes associated with neural crest GRN like SoxE genes²². Supporting this idea, several studies have demonstrated that orthologues of NPB specifiers also have a role in establishing lateral neural border identity in non-vertebrate chordates^23,24 and even in invertebrates, including protostomes^22,25. This raises the possibility that a ‘subcircuit’ responsible for NPB specification is deeply conserved across chordates and potentially shared with protostomes, aligning with an origin in stem bilaterians. Furthermore, this ectodermal domain can produce migratory cell derivatives, including pigment cells and peripheral neuronal subtypes characterized by conserved gene expression profiles²⁵. Although the conservation of this subcircuit remains speculative as long as individual linkages remain to be validated, it is intriguing to consider that the presence of nodes in the NBP subcircuit within non-vertebrate chordates could serve as a blueprint for the evolution of the neural crest developmental program within the NPB at the onset of vertebrate evolution.

The NPB region is not limited to generating NCCs; rather, it represents a more complex ectodermal domain. This region contains progenitors for various cell populations, including NCCs, neural ectoderm that will develop into the central nervous system, non-neural ectoderm, and cranial placodes²⁶. The question of whether the fate of NPB cells is predetermined has been the subject of extensive debate^19,26–28. Recent studies focusing on high resolution single-cell imaging and single cell RNA sequencing have revealed that the NPB is composed of a seemingly random mixture of intermingled cells, whose fates are determined only during the later stages of neurulation^17,29,30. Notably, cells within the NPB co-express markers of multiple lineages at both gastrula stage and later neurula stages²⁹. Surprisingly, a neural crest-specific signature only becomes distinguishable as the neural folds elevate and the neural tube closes¹⁷. In the chick embryo, it has been demonstrated that a simple modulation of a single node within the NPB subcircuit (Sox2 or Pax7) can induce developmental shifts in individual cell fates from neural to neural crest fate or vice versa²⁹.

As development progresses, the NPB undergoes significant morphological changes. It elevates to form the neural folds, which eventually fuse to create the neural tube (Fig. 1a). During this morphogenetic event, the interaction of NPB specifiers activates the expression of a group of transcription factors known as ‘neural crest specifiers’ genes, while simultaneously repressing alternative cell fates such as cranial placodes or neural ectoderm³¹. Since the GRN subunits function in an integrated manner, it is not surprising that some NPB specifiers, such as Pax3/7 and Tfap2a, also participate in the neural crest specification subcircuit, where they serve distinct roles¹¹. Among the well-established neural crest factors are the group of SoxE genes (including Sox8, Sox9 and Sox10), FoxD3, Ets1, Snai1/2, or Twist1. These factors collectively form a regulatory subcircuit that enables NCCs to maintain their multipotency while simultaneously preparing them for migration¹¹. Although urochordates have been shown to have neural crest-like cells that are bipotent and migratory³², they lack the most critical neural crest specifier genes such as the SoxE genes and interconnections within a bona fide neural crest specifier module.

Pre-migratory NCCs reside within the neural folds, where they will subsequently undergo EMT. This process entails substantial changes of their cellular properties including the loss of apicobasal polarity and dismantling of established connections with neighboring cells within the neural folds. Following EMT, NCCs delaminate either during or after elevation and fusion of the neural folds, depending on the species³³. The transformation from epithelial to mesenchymal cells involves profound changes that are regulated by factors which suppress epithelial characteristics while promoting mesenchymal traits, for example by modulation of cadherins during EMT^34–36. Notably, these processes are governed by neural crest specifiers, including FoxD3, Sox10, Twist1, or Snai1/2, which actively regulate cadherins like N-cadherin or Cadherin-6B^37,38 (Fig. 1a), molecules important for maintenance of tight junctions like Tetraspanin 18 (Tspan18)³⁸, or metalloproteases essential for degradation of the epithelial basal lamina³⁹. Together with additional markers like Zeb2 (ref.³⁴) or Lmo4 (ref.⁴⁰), these genes form the EMT subcircuit within the neural crest GRN, which is also influenced by the interplay of the same signaling pathways responsible for defining the NPB, such as BMP or WNT^41,42.

There are some regional differences in the transcriptional control of EMT in the head versus other axial levels; for example, Snai1/2, FoxD3, and Sox9 govern NCC EMT in the trunk region, whereas these transcription factors work in conjunction with Ets1, Sox5 and p53 in the cranial region^43–45 (Fig. 1). The reuse of the same transcription factors and signaling pathways highlights how a cell minimizes the need to evolve and maintain an extensive repertoire of unique regulators within a single GRN. Once NCCs undergo EMT and acquire the ability to migrate to distant locations across the embryo, they start processing signals from the surrounding tissues, both attractive and inhibitory, through interactions with other cells and the extracellular matrix.

Neural crest migration

The migration of NCCs, like any other form of cell migration, is a complex and highly coordinated process. It relies on the synchronization and integration of signals from neighboring tissues, extracellular matrix (ECM) proteins, and intercellular communication among the NCCs themselves⁴⁶ (Fig. 2a). The migration patterns of NCCs vary not only along different regions of the rostrocaudal [G] axis but also across vertebrate species. In some instances, NCCs migrate as a dispersed mass of individuals, whereas in other cases, they move collectively, forming chains, groups, or even single sheets. Individual migration is more commonly associated with amniotes, such as chick or mice, whereas collective migration is typical of anamniotes, like zebrafish or Xenopus⁴⁷. Although NCCs can follow different migratory routes in the head and trunk regions — likely influenced by distinct cues from the adjacent embryonic populations in these areas — the basic pattern of NCC migration remains very similar across vertebrates. A good example is the migration of cranial NCCs, which migrate through the head region in three broad and distinct migratory streams. However, it has been shown that minor alterations from this ‘conserved’ pattern, such as timing (heterochronic) or spatial (heterotopic) changes, may have a crucial role in shaping species-specific craniofacial characteristics³³.

Figure 2 |. Mechanistic and epigenetic control of neural crest development.

a | Modalities and mechanisms driving neural crest cell (NCC) migration. Migration of cranial NCCs is regulated by diverse mechanisms including co-attraction (mutual cell-cell attraction), contact inhibition of locomotion (cells becoming immobile upon contact and reversing direction), substrate stiffness sensing (cells responding to changes in the firmness of their environment), chemotactic ‘chase- and-run’ (cells producing signals that attract another cell type that in turn inhibits the producing cell), actomyosin contraction (changes in the cytoskeleton in response to external signals or contact with other cells), junctional remodeling (changes in cell adhesion and junctional proteins), and extracellular signals (influences of the environment). These coordinated processes ensure directional, collective, and confined migration during development. b | Signals targeted to the neural crest lead to histone modifications (e.g., acetylation, methylation) depending on the availability of substrates such as acetyl-CoA and S-adenosylmethionine (SAM). Various methyl transferases and demethylases (DNMTs, KDM4A, NSD3, PHF8, EHMT2, EZH2, TET2, TET3), histone deacetylases (HDACs), histone variants (H2A.Z.2) and chromatin remodellers (CHD7, PBAF, SRCAP, PHD12) are involved in regulating gene expression (e.g. of Sox genes, Twist, Snail2, Lhx and other developmental genes) to promote neural crest specification, migration and differentiation to various derivatives.

In the embryo, NCCs are positioned between the ectoderm, which forms the epidermis, and the underlying mesoderm. The extracellular matrix, composed of various molecules, lies between these tissues and acts as both a structural framework and a dynamic regulator of NCC migration. The ECM supports movement by providing migratory substrates, establishing inhibitory boundaries, and offering directional cues⁴⁸. As the ECM forms a three-dimensional environment around them, NCCs express matrix metalloproteinases, enzymes that degrade and remodel the ECM^39,49. This remodeling allows NCCs to navigate through the ECM while preserving their migratory capacity. To ensure NCCs do not invade non-target tissues and to guide their migratory streams, several repulsive signaling pathways are activated, including ephrins–Eph [G], semaphorins–neuropilins, and slit–robo signaling⁵⁰ (Fig. 2a).

At the heart of NCCs migration are cell polarity changes that facilitate the collective movement by enabling the outer cells of a migrating collective to generate protrusions and traction forces at the free edge, whereas inner cells exhibit reduced tension and cryptic protrusions^51–53. This polarity is largely established through contact inhibition of locomotion (CIL) (Fig. 2a), a phenomenon in which cells repolarize and change direction upon contact with their neighbors^54,55. In zebrafish or Xenopus, the developmental mechanism controlling CIL is based on a Rac–Rho gradient in which Rac1 is located at the free edge and Rho is found on the site of cell contacts^54,56,57. Thus, this gradient, influenced by WNT-mediated planar cell polarity [G] (PCP) signaling and N-Cadherin, promotes adhesion at contact sites while driving protrusion formation and migration at free edges. N-Cadherin, a key adhesion molecule, has a pivotal role in regulating this Rac–Rho gradient, recruiting the proto-oncogene tyrosine-protein kinase Src and focal adhesion kinase (FAK) to disassemble cell–matrix adhesions and generate the tension necessary for separation^54,55,58. As CIL tends to disperse cells, the balance for keeping the cells together during migration is maintained by co-attraction mechanisms. In this, the C3a ligand and its receptor C3aR have a key role, ensuring that dispersed cells are drawn back to the group, maintaining collective integrity⁵⁹. Thus, the interplay between CIL and co-attraction is finely tuned to support cohesive yet flexible migration.

The group of migratory NCCs can be divided into two main subpopulations, ‘leader’ and ‘follower’ cells, where leader cells guide the group forward, whereas follower cells adhere to directional cues. The transcriptomic analysis of chick NCCs have suggested that the leader and follower cells have distinct transcriptional programs⁶⁰. However, several studies have demonstrated that these roles are not fixed; rather, leader and follower cells frequently exchange positions, reflecting the dynamic and adaptable nature of NCCs migration^61,62.

The directional migration of NCCs is orchestrated not only by co-attraction and CIL but also by chemoattractants secreted by adjacent tissues (chemotaxis). NCCs migrate toward chemoattractants such as stromal cell-derived factor 1 (SDF1), vascular epithelial growth factor (VEGF), or FGF8, which stabilize front cell protrusions and enhance directional movement⁶³ (Fig. 2a). In a phenomenon termed ‘chase-and-run’, NCCs express C-X-C chemokine receptor type 4 (CXCR4) and are attracted to cells located within the cranial placodes by SDF1, which activates Rac1 in NCCs at the front of the collectively migrating group to repel the NCCs upon contact with the placode, creating a coordinated migratory pattern. This heterotypic interaction, mediated by WNT-mediated PCP and N-Cadherin signaling⁶⁴, highlights the importance of chemical and mechanical cues in guiding NCCs migration. Other repulsive signals (e.g., EPHB4–Ephrin-B2) inhibit NCCs from invading non-neural crest tissue, thus preventing mixing of different cranial neural crest streams⁴⁶. Other chemorepellants such as semaphorins, the proteoglycan versican (Fig. 2a) and the BMP antagonist zinc finger protein DAN (also known as NBL1) are important to promote directional migration and maintain NCCs at high density, encouraging cell–-cell interactions fostering collective cell migration⁵⁰.

Mechanical properties of the microenvironment are also crucial for NCCs migration. The head mesoderm beneath the cranial neural crest serves as a substrate for migration. Stiffening of the head mesoderm through convergent extension (Fig. 2a), followed by an increase in cell density, results in the formation of an integrin–vinculin–talin complex in NCCs⁶⁵. This mechanosensation or durotaxis triggers neural crest EMT and initiates collective migration⁶⁶. In Xenopus, the mechanosensitive ion channel Piezo1 has been identified as a key regulator that enables NCCs to detect and react to mechanical signals in their environment^65,67. Piezo1 influences cytoskeletal dynamics by regulating Rac1 activity, which is essential for maintaining proper organization of migration streams and for interacting with repulsive signals such as semaphorins⁶⁷. In avian embryos, the application of electric fields has shown that directed galvanotaxis hinges on the influx of extracellular Ca²⁺ (ref.^68,69). The use of Ca²⁺ channel blockers or the elimination of Ca²⁺ completely disrupts movement toward the cathode, underscoring the importance of calcium in the guidance of migration⁶⁸. Additionally, endogenous electric fields have been shown to direct the coordinated movement of cranial NCCs in Xenopus laevis, driven by voltage-sensitive phosphatase 1 (Vsp1)⁷⁰. These electric fields are generated by ectodermal convergent extension, which produces tension gradients that trigger stretch-sensitive ion channels, positioning galvanotaxis as an important regulator of collective cell migration⁷⁰.

In general, the migration of NCCs is a fluid-like process, characterized by frequent position exchanges among cells within the group, which allows them to respond effectively to mechanical, chemical and electrical cues in the local environment. The intricate balance of adhesion, co-attraction, ECM remodeling, chemotaxis, durotaxis and galvanotaxis underscores the adaptability and complexity of NCCs migration during embryonic development. Despite significant advances, several fundamental questions remain. Given that mechanical, chemical and electrical cues all can influence cell migration, how do these inputs converge to mediate the NCCs response and does this vary according to axial level of origin? To date, most studies have focused on cranial NCCs, raising the question of whether the proposed mechanisms are universal or vary between neural crest subpopulations. Going forward, it will be important to distinguish general mechanisms that govern NCCs migration across vertebrates from those that are species-specific. Furthermore, the extent of neural crest plasticity remains poorly understood. Can differentiated NCCs be reprogrammed to regain migratory or even earlier developmental states, and if so, under what conditions? If achievable, such reprogramming could have profound medical applications, offering an autologous source of progenitor-like cells for repairing tissues and organs derived from the neural crest. Finally, although NCCs migration is clearly not random, the full repertoire of guidance cues directing their pathfinding toward specific tissues remains incompletely defined.

Neural crest GRN subcircuits

Rather than forming a single homogeneous population of cells along the embryonic body axis, the neural crest is comprised of distinct subpopulations that emerge sequentially as the embryo elongates along the anteroposterior axis through NCC proliferation¹. Although all neural crest subpopulations share a common GRN as their basic framework, they also display regional variations based on their specific site of origin from the dorsal neural tube at different levels of the body axis. For example, the cranial neural crest subpopulation gives rise to a unique set of developmental fates compared to the trunk neural crest. Notably, cranial NCCs can differentiate into cell types involved in forming hard tissues, such as odontoblasts (producing dentin) or osteoblasts (producing bone) as originally shown by Julia Platt in the late 1800’s^71,72. This suggests that specific subcircuits within the neural crest GRN, responsible for these developmental outcomes, may arise from the activation of GRN kernels specific to their axial identity.

The advent of RNA sequencing has provided valuable insights into the molecular circuitry underlying the difference in developmental potential among neural crest subpopulations. In chick embryos, the cranial neural crest subcircuit has been extensively studied and has been shown to include several transcription factors such as Dmbx1, Brn3c, and Lhx5, which are expressed during cranial neural crest specification, as well as Tfap2b, Sox8, and Ets1, which are expressed in emigrating and migrating cranial NCCs^11,12 (Fig. 1b). Notably, these factors are absent from the early trunk neural crest GRN^14,73. Interestingly, the ectopic introduction of three transcription factors — Tfap2b, Sox8, and Ets1 — in the trunk neural tube was sufficient to reprogram trunk NCCs into a cranial-like fate (Fig. 1c). These reprogrammed cells exhibited an unexpected gene expression profile, including genes associated with chondrogenesis, such as Runx2 and Alx1, which are not typically expressed in trunk NCCs. In addition, reprogrammed trunk NCCs transplanted into the head turned on the cranial-specific enhancer Sox10E2 (ref.⁷⁴), contributed to cartilage condensations, and thus co-opted cranial neural crest developmental identity¹⁴. Recent findings⁷⁵ have further illuminated the signaling pathways involved in this axial reprogramming. Specifically, mothers against decapentaplegic homolog 2 (SMAD2)- and SMAD3-mediated TGF-β signaling has emerged as a crucial regulator of cranial neural crest developmental potential. In the chick embryo, TGF-β signaling, in cooperation with low levels of WNT signaling, activates cranial-specific cis-regulatory elements and directly modulates gene circuits that support formation of craniofacial skeletal elements⁷⁵. Moreover, activation of TGF-β signaling in reprogrammed trunk NCCs shifts these cells towards a cranial-like identity with unique ectomesenchymal developmental capacity.

A similar subcircuit of the neural crest GRN has also been identified in the cardiac neural crest⁷⁶. Notably, this subcircuit closely mirrors the cranial subcircuit, comprising Ets1 and Sox8, with the addition of Tgif1 (Fig. 1b). Remarkably, the ectopic introduction of these factors into the chick trunk neural tube confers cardiac crest developmental potential rather than cranial. When transplanted into the location of ablated cardiac crest cells, these reprogrammed chick trunk NCCs migrated to the outflow tract, contributed to the septum, and rescued the congenital birth defect persistent truncus arteriosus caused by the absence of the cardiac neural crest⁷⁶ (Fig. 1d). In contrast, control trunk neural crest grafts lacking these factors were unable to achieve the same outcome. Thus, addition of three transcription factors to the trunk neural crest program was sufficient to reprogram their axial level identity.

These results collectively demonstrate that the simple redeployment of a new node within the neural crest GRN subcircuit can significantly alter cell fates. Furthermore, these findings underscore the modularity of the GRN and suggest that individual neural crest subpopulations possess unique, axial-specific transcription programs capable of reshaping the developmental potential of other neural crest subpopulations. Comparative transcriptomic analyses of cranial and trunk neural crest across vertebrates indicate that the trunk neural crest may represent the ancestral form of neural crest⁷³. As such, it may serve as the most suitable model for understanding the foundational blueprint of the neural crest GRN, and how the individual neural crest subpopulations evolved along the body axis. Despite these insights, several key questions remain unresolved. What are the specific cis-regulatory elements, upstream regulators and crosstalk between them responsible for driving subcircuit diversification across axial levels? How do these regulatory inputs evolve and interact to generate the functional heterogeneity observed in neural crest derivatives? Although current approaches allow us to manipulate neural crest subpopulations and generate, for example, early-stage neural crest-derived tissues (for example, pre-cartilaginous condensations [G]^75,76), mechanistic understanding and technical capability is still lacking to robustly generate fully differentiated neural crest-derived tissues de novo. Addressing these questions will be essential for decoding the full developmental logic of the neural crest and unlocking its full regenerative potential.

Metabolic and microenvironmental regulation of the neural crest

During embryonic development, stem cells communicate with their local microenvironment through mechanical cues that in turn influence cell fate and have a vital role in tissue patterning and organ formation. Thus, in addition to the inherent GRN, neural crest development is strongly influenced by various external influences, such as metabolic stress, mechanical forces and chemotactic factors. These macro- and micro-environmental factors shape the NCCs’ competency, fate choices, migration, and kinetics of differentiation in a spatiotemporally prescribed manner. For example, hydrostatic pressure of the blastocoel, an embryonic cavity in close contact with the prospective NCCs, can influence neural crest specification: An increase in hydrostatic pressure inhibits YAP signaling and disrupts Wnt activation in the responding tissue, which is necessary to initiate neural crest formation⁷⁷.

Varying oxygen levels also influence neural crest behavior, as evidenced by changes in hypoxia-inducible factors (HIF-1α and HIF-2 α) during neural crest EMT and migration^78,79. Studies in Xenopus and zebrafish embryos showed that knocking down HIF-1α using morpholinos inhibits NCC migration and chemotaxis, without affecting their initial induction⁷⁸. HIF-1α was found to directly activate Twist, which inhibits E-cadherin, allowing NCCs to detach from the neural tube. It also regulates the expression of CXCR4, the chemokine receptor essential for guiding⁷⁸. Chick embryos naturally experience hypoxia in ovo, activating HIF, which stimulates EMT and neural crest formation, whereas increased oxygen levels hinder EMT and decrease the population of migrating NCCs⁷⁸. Hypoxia affects the expression of EMT-related genes such as Snai2 and Sox10 (ref.⁸⁰). These findings collectively highlight the significance of temporary hypoxia and the activity of HIF-1α as a key regulator that coordinates EMT and chemotactic signals to facilitate effective migration of NCCs.

NCCs diversification involves the deployment of GRNs that determine cell identity and function. These GRNs are dynamically modulated by epigenetic modifications such as histone variants and histone modifications (e.g. acetylation, lactylation, crotonylation, methylation) and DNA methylation and demethylation (Fig. 2b). These histone and DNA modifications regulate chromatin structure throughout neural crest development, controlling both gene activation and repression that mediate cell fate decisions. In mouse cranial NCCs, cells in the early stages of migration exhibit bivalent chromatin regions characterized by the presence of histone 3 lysine 27 trimethylation (H3K27me3) and H3K4me2 at loci related to positional identity, establishing a transcriptional state that is ready to respond to subsequent stimuli⁸¹. During neural crest specification (Fig. 2b), epigenetic marks such as H3K4me1 and H3K27ac indicate poised and active enhancers, particularly at regions close to key neural crest-specifying genes like SOX10, FOXD3, and SNAI2, which are frequently associated with the transcriptional coactivator p300 (ref.⁸²). In addition, an accumulation of H3K4me3 at active promoters indicates ongoing transcription. Conversely, repressive modifications such as H3K27me3, H3K9me3, and H2AK119ub, which are introduced by Polycomb and heterochromatin remodelling complexes, function to suppress non-neural crest fate, thus promoting neural crest identity⁸³. As NCCs undergo EMT and initiate migration (Fig. 2b), chromatin remodeling is marked by a decrease in repressive marks and an increase in H3K27ac^15,81,82, which likely enhances chromatin accessibility and activates gene programs associated with motility. Additionally, histone variants like H3.3 and H2A.Z.2 have a vital role by influencing chromatin accessibility and neural crest lineage specification. H3.3 is necessary for the formation of cartilage and pigment cells derived from cranial neural crest⁸⁴, whereas H2A.Z.2 facilitates melanocyte identity through the activation of Mitf⁸⁵. Heterozygous truncations Snf2-related CBP activator protein (SRCAP) reduce H2A.Z.2 incorporation at AT-rich enhancers, thus impairing expression of genes critical for neural crest migration and craniofacial development. This defect underlies the tissue-specific anomalies seen in Floating-Harbor Syndrome (FHS), a rare genetic disorder characterized by distinctive facial abnormalities and frequently accompanied by hearing or gastrointestinal dysfunctions⁸⁶. Collectively, these studies highlight that histone variants such as H3.3 and H2A.Z.2, along with their specific binding partners (e.g., (PWWP domain-containing protein 2A) PWWP2A)⁸⁷, function as dynamic epigenetic regulators that influence chromatin accessibility and fate determination in the neural crest lineage.

Although histone modifications and variants are critical for accurate spatial and temporal activation of GRNs, substrate availability dictates how and when these modifications occur. Histone acetyltransferases (HATs) transfer an acetyl group from acetyl-CoA to the histones, allowing the chromatin to unwind, which gives transcription factors access to promoter sites that activate the GRN⁸⁸. The rate-limiting step for HATs is the availability of acetyl-CoA, which is tightly regulated and depends on the cell’s metabolic state and nutrient availability^89,90. Delaminating NCCs display a high NAD+ to NADH ratio, suggesting that aerobic glycolysis may be essential for maintaining a balance between histone acetylation and deacetylation in this stem cell population⁹¹. Likewise, migrating NCCs undergo increased aerobic glycolysis before delamination^91,92. This process produces lactate as a byproduct and leads to neural crest-specific histone lactylation, which is critical for promoting accessibility of neural crest enhancers (Box 2), expression of neural crest genes, and cellular behaviors like EMT and migration via the YAP–transcriptional enhancer factor TEF-1 (TEAD1) pathway [G]⁹³.

Box 2: The role of enhancers in neural crest biology.

Enhancers have a vital role in GRNs, acting as key components that integrate external signaling inputs with transcription factors. This integration ensures the precise timing and spatial regulation of critical developmental processes, such as cell fate determination. Enhancers are short DNA sequences containing multiple transcription factor binding sites, which can be located at various distances from their target promoters, sometimes hundreds of kilobases away, both up- and downstream^217,218. These elements work in conjunction with transcription factors and cofactors to regulate gene expression programs specific to tissues or developmental stages. A single gene is often regulated by multiple enhancers, forming clusters, some of which can form super-enhancers that drive expression of essential genes²¹⁹. An example is the Sox10 super-enhancer, which ensures robust expression of this crucial neural crest specifier gene¹⁵. Moreover, enhancer redundancy makes them an ideal target for evolution, allowing changes without jeopardizing function of the genes they regulate²²⁰.

Active chromatin enhancer landscapes play a key role in neural crest development. In particular, enhancers defined by the presence of the transcription coactivator p300 flanked by nucleosomes modified by histone 3 lysine 27 acetylation (H3K27ac) and lysine methylation (H3K4me1) have been shown to be critical for human NCCs development⁸². Synergistic function of the transcription factors TFAP2A and NR2F1/F2 in the neural crest is critical for permissive chromatin states, characterized by high levels of p300 and H3K27ac, marking active enhancers⁸². Such cooperative binding of transcription factors is thought to have a major role in overcoming the nucleosomal barrier threshold in transcription factor recruitment, subsequently promoting nucleosome eviction. In certain cases, the recruitment of transcription factors to enhancer regions occurs in a stepwise fashion, whereby ‘pioneer’ factors such as Foxd3 can initially access the nucleosomal DNA, either independently or through associations with chromatin remodelers, leading to enhancer priming and chromatin opening in a neural crest-specific manner¹³⁵.

The emergence of morphological innovations, such as species-specific characteristics, does not necessarily depend on the acquisition of new genes. Instead, the acquisition of novel regulatory elements is often sufficient to drive the origin of such traits. For example, it has been shown that although craniofacial diversity between humans and chimpanzees is influenced by similar fundamental factors, these factors are regulated differently at the enhancer level, leading to species-specific features²²¹. Enhancer evolution, therefore, has likely been a significant driver of morphological changes and species diversification. In addition to shaping species-specific traits, enhancers may encode ancestral cis-regulatory blueprints that govern gene expression patterns during neural crest development across vertebrates^222,223. Thus, clues regarding the evolution of the neural crest can be gleaned through the lens of enhancer evolution²²⁴. Recent technological advances in regulatory genomics, such as ultra-throughput, ultra-sensitive single-nucleus assay for transposase-accessible chromatin using sequencing (UUATAC-seq) and deep-learning models like Nvwa cis-regulatory element (NvwaCE)²²⁵, help to decode cis-regulatory grammar and accurately predict candidate cis-regulatory element landscapes from genomic sequences from both model and non-traditional organisms. Thus, these data hold the promise of identifying conserved cis-regulatory elements in the neural crest GRN, and revealing the effects of mutations within genomic regions. However, these algorithms currently lack the ability to infer the specific functions of individual enhancers or repressors, or to predict the impact of mutations in distal regulatory elements on gene expression. This limitation underscores the need for complementary functional assays to fully elucidate enhancer-mediated regulatory mechanisms.

As enhancers are important for normal development, their dysfunction can have significant consequences, contributing to a range of disorders collectively known as ‘enhanceropathies’. Recent research has shown that the craniofacial disorder Pierre Robin sequence results from the loss of extreme long-range enhancers of SOX9, which regulate the stage-specific expression of this gene in early migrating cranial NCCs²²⁶. This loss reduces SOX9 levels precisely when these cells are forming the mandibular and midfacial structures, leading to impaired chondrogenesis and craniofacial defects. Given that combinatorial binding of transcription factors to their cognate motifs largely determines the spatiotemporal activity of developmental enhancers, future research examining the possible causal role of other single nucleotide polymorphisms (SNPs)⁸² in ‘neural crest enhancers’ may provide important insights into the origin of developmental disorders.

Histone and DNA methylation are typically associated with silenced chromatin states. However, exceptions, such as methylation of H3K4, H3K36, and H3K79, are generally associated with active transcription⁸⁸. Histone and DNA methyltransferases transfer the methyl group from a donor molecule such as S-adenosylmethionine (SAM) to their targets and modulate their function, facilitating global chromatin changes. The folic acid synthesis pathway produces SAM, thus, influencing the epigenetic landscape of NCCs by impacting histone and DNA methylation. In addition, decreased folic acid levels disrupt methionine metabolism and cause cellular accumulation of the metabolite homocysteine, which promotes NCCs migration but prevents their differentiation⁹⁴.

Collectively, these studies highlight the importance of metabolic substrates throughout the stages of neural crest development — from specification and migration to differentiation — and provide insights into how these metabolic pathways are modified to carry out cell-specific roles. Taken together, these studies show that coordinated epigenetic modifications contribute to the characteristic plasticity of the neural crest, ensuring precise spatial and temporal regulation of GRNs during their induction, migration, and differentiation. Open questions remain regarding how the myriad of epigenetic changes are coordinated in space and time as well as the role of the changing macroenvironment and how this may affect epigenetic factors that drive the species-specific acquisition of new traits in response to environmental pressures.

Neural crest as a driver of evolutionary change

The emergence of NCCs together with cranial ectodermal placodes represents a pivotal milestone in vertebrate evolution that had a critical role in the development of the vertebrate ‘New Head’ [G] and significantly enhanced the complexity of the chordate body plan. Acquisition of these cell types enabled a transition from passive filter-feeding, characteristic of non-vertebrate chordates, to the active predation seen in most vertebrates⁹⁵, largely due to the advent of specialized sensory systems and craniofacial skeletal structures adapted to various feeding strategies^95,96. Thus, the remarkable versatility of NCCs is the foundation of the evolutionary success of vertebrates. Their ability to migrate throughout the developing body and differentiate into a vast array of specialized cell types underpins the formation of the structures facilitating evolutionary predation and expansion of the vertebrate brain.

The neural crest gives rise to two main categories of derivatives: ectomesenchymal cells, such as chondroblasts, osteoblasts, odontoblasts, pulp cells, adipocytes, pericytes, and cardiomyocytes, and non-ectomesenchymal cells, including neurons, glia, and pigment-producing cells like melanocytes or iridophores¹ (Fig. 3a). As with other submodules of the neural crest GRN, the differentiation subcircuits are dynamically shaped by various environmental signaling cues, such as WNT, BMP or TGF-β pathways¹¹. These external signals activate subcircuits that guide the differentiation of specific cell types. For instance, differentiation into the neuronal lineage begins with activation via BMP signaling. This activates transcription factors like Sox10, which subsequently promotes the expression of other key transcription factors, including Phox2b, Ascl1, and Tfap2a. These factors, in turn, activate downstream genes, such as Hand2, Phox2a, Gata 2, Gata3, Th, and Dbh^{11,12,97–99} (Fig. 3a). Similarly, other cell fates are regulated by specific subcircuits of the neural crest GRN, in which Sox10 (enteric neurons, melanocytes, or Schwann cells) versus Sox9 (chondrocytes) serve as important regulatory genes¹¹. Ultimately, the diversity of neural crest derivatives is a result of their exposure to distinct combinations of external signals and transcription programs during migration and the very early phase of differentiation. These environmental and internal cues act in concert to shape the terminal fates of NCCs, ensuring the proper development of a wide array of specialized cell types.

Figure 3 |. Developmental potential of the neural crest and neural crest-derived tumors.

a | Neural crest derivatives are categorized into two groups: ectomesenchymal and non-ectomesenchymal. Neural crest differentiation is orchestrated by distinct GRN subcircuits, leading to the formation of different cell types, such as chondroblasts, Schwann cells, melanocytes, sympathetic neurons, and enteric neurons among others. The neural crest also gives rise to Schwann cell precursors (SCPs), which serve as developmental precursors (solid arrow) to mature Schwann cells. Notably, SCPs have the potential to generate multiple neural crest-derived cell types (dashed arrows). Furthermore, these cell types can arise through dedifferentiation from mature Schwann cells following peripheral nerve injury, contributing various regeneration processes. b | Progenitors of the neural crest lineage exhibit remarkable diversity and are found in various locations throughout the human body. These progenitors can become cancerous by hijacking neural crest developmental programs and repurposing these circuits for uncontrolled proliferation, survival, invasion, and therapy resistance, as has been shown in meningiomas, melanoma, neurofibroma and neuroblastoma cases. The gene regulatory networks (GRNs) subcircuits underlying these cell changes are cell type-specific, and minor rewiring of the subcircuits can lead to cell state switching, as shown in the case of adrenergic (ADRN) and mesenchymal (MES) neuroblastoma. The diamond icon on the line indicates regulation via protein-protein interaction.

Beyond its role in generating novel cell types, the neural crest has a dynamic role in embryonic development by reciprocal interactions with surrounding tissues. For example, interactions with neighboring ectoderm, mesoderm or endoderm are essential for the formation of various structures, including craniofacial cartilage and bones, teeth, feathers, and more¹⁰⁰. Much of the evidence supporting the importance of neural crest interactions comes from studies using interspecies chimeras, particularly quail–chick or quail–duck chimeras^101,102. Although these interactions largely dictate the general craniofacial pattern, minor deviations from this developmental blueprint give rise to species-specific features, which are finely tuned by natural selection. Notable examples include studies on Darwin’s finches and cichlid fishes, which have undergone adaptive radiations, resulting in a diverse array of craniofacial morphotypes — including a stunning repertoire of beaks or jaws adapted to different feeding strategies^103,104. Research has shown that modulating WNT and BMP signaling pathways in the epithelium^105–110, regulating expression of the craniofacial-specific transcription factor Alx1 (refs.^111,112) along with cell cycle regulation¹¹³ and bone resorption¹¹⁴ in the neural crest, significantly influences the size and shape of beak and jaw primordia. Genome-wide association studies [G] have further elucidated the mechanisms underlying craniofacial diversity, identifying key genomic regions and regulatory elements associated with cranial NCCs^115,116. Beyond differentiating into various cell types and tissues, NCCs have an important role in shaping non-neural crest structures, including mesodermal cranial musculature. Specifically, cranial NCCs contribute to the formation of connective tissues surrounding muscle fibers^117,118. These cells are essential not only for proper muscle development and attachment but also for influencing species-specific patterns¹¹⁹. This body of evidence suggests that NCCs actively shape species-specific morphologies, rather than serving as passive components, and have a pivotal role in coordinating the developmental processes driving the evolution of diverse craniofacial structures unique to each species.

Neural crest populations along the antero-posterior axis

The neural crest is not a uniform population along the body axis. Instead, it is divided into distinct subpopulations based on their axial origin: cranial, vagal including cardiac, trunk, and sacral NCCs. Although individual neural crest subpopulations are specialized to produce distinct derivatives, certain cell types, such as melanocytes and peripheral glia, are generated at all axial levels of the embryo¹, likely because their widespread presence is essential for functions throughout the entire body. In contrast, other neural crest derivatives are confined to specific regions, with each region giving rise to specialized structures identified originally through quail–chick chimera transplantation studies^120–122. For example, cranial NCCs form cartilage¹²¹, cardiac NCCs contribute to the heart ectomesenchymal structures¹²³, and vagal and sacral NCCs are uniquely responsible for developing the enteric nervous system^124,125. In contrast, trunk NCCs of amniotes lack the ability to form ectomesenchymal derivatives or contribute to the enteric nervous system, but instead differentiate into pigment cells, sensory and sympathetic ganglia¹.

Heterotopic transplantation experiments have demonstrated that when trunk NCCs are grafted into the cranial region, they failed to develop into typical cranial neural crest derivatives, such as cranial skeletal tissues or led to a reduced number of trigeminal ganglion neurons^126–128. Interestingly, experiments on chick trunk neural crest have shown that when these cells are co-cultured in bone-growth-optimized media, they can differentiate into skeletal cells in vitro¹²⁹, which does not occur in vivo. After being transplanted into the cranial region, these ‘enhanced’ trunk NCCs were capable of generating cranial derivatives in vivo¹²⁹. This indicates that trunk NCCs may represent a simplified neural crest subpopulation. Moreover, as discussed above, the developmental program of trunk NCCs can be readily reprogrammed to generate derivatives typically associated with other neural crest subpopulations^14,76. This suggests that the transcription program of trunk NCCs may be the most representative of the ancestral neural crest developmental program.

To better understand the evolution of the neural crest developmental program, a comparative analysis of key evolutionary lineages of vertebrates and non-vertebrate chordates is essential. A recent study⁷² sheds light on this evolutionary process by examining key nodes of the neural crest GRN across major vertebrate lineages, including both jawless and jawed vertebrates. The findings support the hypothesis that the individual axial subcircuits of the neural crest GRN evolved gradually through the co-option of neural crest nodes during vertebrate evolution. Indeed, the study demonstrated that in the cartilaginous little skate, transcription factors such as Ets1, SoxE, and Tfap2 are expressed beyond the cranial region. In contrast, the cranial ‘subcircuit’ of ray-finned fish, such as zebrafish, includes these three genes along with two additional nodes, Lhx5 and Dmbx1. Similarly, in tetrapods, another node, Brn3, is incorporated into a ‘cranial crest specific subcircuit’. Moreover, data from the jawless sea lamprey show that only two candidate nodes of the neural crest GRN, Tfap2 and SoxE, are expressed along its body axis, whereas the other genes are not expressed at all⁷³. These findings suggest that a certain level of axial identity within the cranial subcircuit exists in non-tetrapod lineages, although it is significantly more elaborated in tetrapods. Overall, the cranial subcircuit likely evolved through the gradual incorporation of new nodes into an ancestral ‘trunk-like’ neural crest GRN, resulting in a distinct axial-specific regulatory circuit. Although it might be tempting to link this to the developmental inability of trunk NCCs to form skeletogenic derivatives (bone and dentin), recent studies suggest otherwise: One study showed¹³⁰ that trunk NCCs, particularly in early branching vertebrate lineages, are capable of giving rise to skeletogenic tissues, giving rise to the protective dermal skeleton. Therefore, the cranial subcircuit nodes identified so far are not indispensable for the formation of hard tissues, and the skeletogenic ability of neural crest appears to be an ancestral characteristics^130,131 (Box 3).

Box 3: Origin of bones and dentin from trunk neural crest cells.

Neural crest subpopulations differ in their ectomesenchymal potential to generate hard tissues, such as bone or dentin. In amniotes such as the chick embryo, this ability is confined to the head and neck¹²¹ and absent from the trunk. However, fossil evidence of early vertebrates suggests their bodies were partially or entirely covered by dermal armor, composed of bone and/or dentin, characteristic of various stem vertebrates²²⁷. This suggests that the trunk neural crest may have once produced hard tissues²²⁸, with that ability later becoming restricted to the head during vertebrate evolution^229,230.

In living vertebrates, fish scales are a prominent example of a postcranial dermal skeleton. Initial evidence for trunk NCC contribution to hard tissue came from the observation that neural crest-derived tumor pigment cells could differentiate into scale-like structures in vitro²³¹. Subsequent in vivo analyses suggested that trunk NCCs can form postcranial dermal skeleton, such as fin rays^232,233. However, later studies suggested that the fin rays and scales of teleost fish (e.g. zebrafish and medaka) are mesoderm-derived^229,234,235. In contrast, dentinous tooth-like denticles covering the bodies of the little skate, a cartilaginous fish, originate solely from trunk NCC¹³¹. This raised questions as to whether the dermal armor of early vertebrates consisted of a neural crest-derived dentinous layer and a mesoderm-derived bony layer, or both. A caveat is that scales of zebrafish and medaka are highly derived, formed by a weakly mineralized plywood-like structure, elasmodin²³⁶. Recent insights into this question arose from studying sturgeon dermal scutes, comprised of only bone¹³⁰. Fate mapping of trunk NCC and transcriptomic analysis of sturgeon scutes revealed that the osteocytes forming the bone are exclusively derived from trunk NCC¹³⁰, suggesting that the postcranial dermal skeleton in early vertebrates likely originated from trunk NCC. Derived from ancient ganoid scales comprised of bone and dentin, scutes have lost their dentinous layer during evolution. Together with evidence from cartilaginous fish¹³¹, this suggests that the skeletogenic ability was likely present along the entire rostrocaudal axis in early vertebrates¹³⁰, providing a novel view of vertebrate skeletal evolution.

In addition to fish scales, several tetrapod lineages possess extensive postcranial dermal skeletons (e.g. osteoderms, turtle shells) though to be derived from late-migrating trunk cells in the region of the anterior turtle carapace, the gastralia of alligators, and the osteoderms of armadillos that express markers often associated with NCCs, including Human Natural Killer-1 (HNK1)^237,238. A recent murine scRNA-seq study⁹⁷ uncovered a cryptic bifurcation in the posterior trunk neural crest, whereby cells can differentiate into either neuroglial or mesenchymal fates. Thus, it is possible that strong mesenchymal signaling could bias trunk NCCs toward a skeletogenic fate. Together, these findings suggest that the developmental potential to form hard tissue may have been uniformly present in the neural crest along the body axis, later lost in the trunk of some vertebrate lineages but independently reactivated in others. This raises the intriguing possibility that bone formation is an ancestral property of the neural crest, later coopted by the mesoderm to generate dermal and endochondral bone. While this scenario remains speculative, future detailed developmental and functional analyses may shed light on this important question.

The division of the neural crest along the anteroposterior (rostrocaudal) axis into individual subpopulations is not determined solely by the interactions of a few transcription factors. Epigenomic analyses have uncovered the presence of numerous axis-specific enhancers. For example, Sox10 enhancers, such as Sox10E2 and Sox10E1, are specifically active in the cranial and vagal-trunk regions, respectively¹³². Similar region-specific patterns have been observed for other Sox10 and FoxD3 enhancers in chick embryos^15,133. A key question is whether these axial-specific enhancers are conserved across vertebrates¹³⁴ and if they have been progressively incorporated during vertebrate evolution. By integrating epigenomic and transcriptional profiling of NCCs at both the population and single-cell level with in vivo genome and epigenome engineering^{15,82,135,136}, it is possible to uncover the regulatory mechanisms underlying neural crest development and axial identity. Thus, this approach can help identify neural crest-specific enhancers, super-enhancers [G], transcription factors, and cis-regulatory elements. Combined with robust functional validation, this facilitates a detailed reconstruction of the neural crest GRN.

An additional well-defined subpopulation of the neural crest are the vagal NCCs, which contribute to the formation of the enteric nervous system (ENS) in vertebrates^124,125. Interestingly, the analysis of the neural crest of sea lamprey has shown that jawless vertebrates may lack vagal and possess only cranial and trunk neural crest subpopulations¹³⁷. Analysis of the lamprey neural crest transcriptional network has provided valuable insights into the developmental program, shedding light on how it is conserved — or diverges — among vertebrates. Furthermore, unlike findings in zebrafish, lamprey data revealed a lack of key nodes in the vagal subcircuit that are essential for ENS development¹³⁴.

All these findings raise the intriguing possibility that the vertebrate ancestor possessed a single, continuous neural crest population extending uniformly along the rostrocaudal axis. During vertebrate evolution, this uniform ancestral neural crest, along with its GRN, may have served as an evolutionary substrate for the formation of distinct neural crest subpopulations with specialized developmental capabilities, driven by the progressive refinement of GRN subcircuits. Whereas recent studies have focused mostly on anterior axial levels — such as cranial and vagal NCCs — open questions remain about the trunk neural crest and its relationship to the ancestral neural crest, from which it may be derived. Another important question is how the elaboration of neural crest populations along the body axis relates to the transition from water to land. These questions highlight the importance of further research on non-model organisms positioned at critical nodes along the tree of life. For example, future analyses should examine other species important for our understanding of water to land transition like non-teleost fishes (bichirs, or sturgeons), lungfishes and early branching lineages of amphibians.

Function of the neural crest in adult regeneration

In addition to forming most of the peripheral nervous system and craniofacial skeleton, NCCs make critical contributions to the heart. Lineage-tracing experiments have shown that NCCs differentiate into cells making up the smooth muscle of the arterial vasculature^138,139, the septum and valves that separate the atrial and ventricular compartments of the heart^140,141, and neuronal and glial derivatives that mediate its parasympathetic innervation^142,143. Ablation of cardiac NCCs in the chick embryo results in congenital heart defects, including persistent truncus arteriosus, and leads to reduced heart rates, failed heart looping, defective myocardial maturation, and reduced ejection fraction¹⁴⁰. Surprisingly, it has also been shown that NCCs in chick and mouse contribute to cardiomyocytes in the ventricle, a feature once thought to be limited to non-amniotic vertebrates like zebrafish^123,144.

The contribution of the cardiac neural crest to cardiomyocytes across diverse species raised the intriguing possibility that NCCs may be involved in adult heart regeneration. Adult zebrafish hearts retain extensive regenerative capacity throughout life¹⁴⁵. Interestingly, many genes of the embryonic neural crest gene regulatory program appear to be reactivated during the process of zebrafish heart regeneration^123,146. Moreover, depletion of Sox10-expressing NCCs results in a failure to regenerate¹⁴⁷. As the hearts of postnatal mice display regenerative ability for one week after birth¹⁴⁸, it will be interesting to explore the possibility that changes in murine NCCs shortly after birth may contribute to the loss of regenerative ability of the mammalian heart. One intriguing possibility is that nerve-associated Schwann cell precursors (SCPs)^149,150, which are neural crest-derived, may be the source of cells that contribute to heart regeneration.

SCPs (Box 4) may also have a potential role in the regeneration of other adult tissues. Indeed, multiple recent studies have shown that SCPs actively contribute to regeneration not only of peripheral nerves^151,152, but also other tissues and organs, including the skin¹⁵³ and hard tissues such as bones^154,155 and teeth¹⁵⁶. Schwann cells and SCPs may represent a unique pool of late stem cells present along the peripheral nerves throughout the body. Following injury, Schwann cells can reactivate the EMT GRN subcircuits, enabling them to detach and revert to a neural crest-like state, effectively becoming SCPs. This extraordinary developmental plasticity enables SCPs to regain multipotency and facilitate tissue regeneration by remodeling the wound microenvironment. Interestingly, recent studies^157,158 suggest that Schwann cells, and consequently SCPs, are attracted/migrate to tumors such as neuroblastomas, where they exhibit repair-like responses similar to those observed during tissue regeneration. These findings raise the intriguing possibility that SCPs retain embryonic-like multipotency, enabling them to contribute not only to tissue regeneration but also to the repair of tumor-affected tissues. However, open questions remain regarding the GRN modules that regulate SCPs, and further investigation is required to better define this cell type and its role in development and regeneration. Such insights would be critical for improving tissue regeneration strategies and developing more effective cancer therapies.

Box 4: Schwann cell precursors as late stem cells.

Schwann cell precursors (SCPs) are a fascinating type of glial progenitor that arises from the neural crest²³⁹. Their primary function is to differentiate into terminal Schwann cells, both non-myelinating and myelinating cells that wrap around peripheral nerve axons to enable saltatory conduction. However, some SCPs retain a stem cell-like state throughout development and, in some cases, into adulthood²⁴⁰. Interestingly, mature Schwann cells exhibit an exceptional ability to detach from peripheral nerves and dedifferentiate back into SCPs (Fig. 3a) under specific conditions. Unlike other neural crest derivatives, SCPs are unique in retaining their multipotent state, enabling them to differentiate into a broad range of cell types — nearly the same spectrum as their parent embryonic neural crest population²³⁹. These include odontoblasts¹⁵⁶, chondrocytes²⁴¹, carotid body cells²⁴², neuroendocrine cells²⁴³, sympathetic neurons¹⁵⁷ or enteric neurons^137,244. This distinctive trait classifies SCPs as ‘late stem cells’, remaining present along nerves throughout the body.

Unlike other neural crest-derived cell types, which gradually differentiate during embryonic development, SCPs retain the expression of several core neural crest specifier genes, such as Sox10, FoxD3, Tfap2a/b, among others²⁴⁵. Moreover, SCPs are morphologically distinct from migratory NCCs and are associated with nerves²⁴⁶. As a result, deciphering the transcriptomic identity of SCPs has posed a longstanding challenge in the neural crest field. However, a recent single-cell ‘transcriptomic atlas’ of the entire murine neural crest, spanning from the induction to postnatal developmental stages²⁴⁵, revealed that late migratory NCCs and SCPs share similar transcriptional profiles and transit through a multipotent hub state, characterized by the expression of markers such Sox8, EndrB, Serpine2, and Itga4. This hub state consists of intermingled cells predisposed to distinct NCCs fates, including pigment cells or enteric neurons²⁴⁵. Interestingly, proto-neural crest cells are thought to have originally followed sensory and motor nerves as they traversed the bodies of early vertebrates, and thus they might resemble SCPs²⁴⁷. It is intriguing to speculate that both these cells and SCPs were likely regulated by a similar, possibly ancestral, genetic program. Acting as late-stage stem cells, SCPs appear to be important for contributing to many tissues during postnatal development.

Hijacking of the neural crest developmental program in cancer

During development, the neural crest gene regulatory program balances multipotency with specification, migration, and differentiation. In neural crest-derived tumors, these fundamental aspects of neural crest biology are exploited for malignant transformation, tumor progression, and resistance to treatment. This is achieved by co-opting molecular programs underlying EMT and migratory ability. Ectopically activating the neural crest GRN subcircuits enables neural crest-derived tumors to ‘hijack’ normal embryonic processes for invasion and metastasis. This has led to the hypothesis of a neural crest-like cell in certain types of cancers and can reflect a ‘cancer stem cell’ (CSC) phenotype.

Neural crest-derived tumors contain transcriptionally heterogeneous cell states that differ in their proliferative and metastatic ability. A notable example is neuroblastoma (Fig. 3b), a pediatric tumor that arises from trunk NCCs that form sympathetic ganglia and adrenomedullary cells¹⁵⁹. MYCN gene amplification is the most significant driver of sporadic neuroblastoma, whereas mutations in ALK and PHOX2B are linked to familial neuroblastoma^160,161. Based on super-enhancer and transcriptional signatures, neuroblastoma tumors are broadly categorized into adrenergic (ADRN) and mesenchymal (MES) subtypes, with the former characteristic of committed sympathoadrenal lineage and the latter similar to undifferentiated precursors with elevated therapy resistance^162,163. CRISPR–Cas9 screens have revealed HAND2, ISL1, PHOX2B, GATA3, TBX2, and ASCL1 as neural crest core regulatory genes critical for the growth and survival of MYCN amplified neuroblastoma¹⁶⁴. Patient-derived neuroblastoma cells behave similarly to CSCs, showing elevated tumor-initiating potential and chemoresistance, and have signatures typically reminiscent of bipotent progenitors, SCPs, neuroblasts, and chromaffin cells [G] that help determine overall risk status¹⁶⁵.

Neural crest-derived tumor melanomas (Fig. 3b) originate from neural crest-derived melanocyte stem cells (McSCs) and differentiated melanocytes¹⁶⁶. The oncogene, BRAF^V600E, is expressed in 80% of benign human naevi, but only a fraction progress to melanoma¹⁶⁷. The transcription factor SOX10 is re-expressed in human and mouse nevus melanocytes and zebrafish melanoma-initiating cells^166,168. The re-expression of SOX10 and downstream McSCs and melanoblast genes is mediated by epigenetic factors that modulate enhancer accessibility¹⁶⁹. In addition, MITF, a transcription factor downstream of SOX10, functions as a molecular rheostat in melanoma progression, where MITF^high cells are proliferative, and MITF^low cells are invasive and resistant to therapies¹⁷⁰. Such a high degree of phenotypic plasticity can be attributed to melanoma cancer stem cells marked by high expression of NGFR (CD271), JARID1B, ALDH1, and ABCB5, which exhibit slow-cycling behavior, evasion of the immune response, and resistance to BRAF inhibitors^171–174. Single-cell analysis of human and murine melanomas has revealed seven different melanoma states, two of which showed expression of neural crest-stem cell-like signatures¹⁷⁵. Additionally, four distinct drug-tolerant transcriptional states are found in patient biopsies and PDX models with distinct MAPK dependencies, translating into different therapy responses¹⁷⁶.

Neurofibromatoses (Fig. 3b) are tumor suppressor syndromes characterized by multiple nervous system tumors and Schwann cell neoplasms, classified into neurofibromatosis type 1 (NF-1), neurofibromatosis type 2 (NF-2), and Schwannomatosis¹⁷⁷. NF-1 is caused by mutations of the NF1 gene encoding neurofibromin, a GTPase-activating protein that functions upstream of the RAS–MAPK, RAF–MEK–ERK, PI3K–AKT–mTOR, and cAMP–PKA pathways¹⁷⁸. Characterized by neurofibromas on spinal, peripheral, or cranial nerves, loss of neurofibromin in Schwann cells or neuronal progenitors leads to increased cell growth, tumorigenesis, and survival through hyperactivation of RAS^179,180. NF-2 is an autosomal dominant disorder resulting from germline or mosaic mutations in the NF2 tumor suppressor gene, leading to multiple benign tumors. NF-2 tumors are mainly composed of Schwannomas and meningiomas¹⁸¹. In humans, mice, and zebrafish, NF2 is highly expressed in NCCs, Schwann cells, meningeal cells, neurons, oligodendrocytes, optic neuroepithelial compartments, and lens fiber cells^182,183. The NF2 gene encodes Merlin, a tumor suppressor that functions as a membrane–cytoskeleton linker inhibiting cell proliferation through WNT–β-catenin, NOTCH, RAS, RAC and RHO, TGF-β, HIPPO, and receptor tyrosine kinase pathways^184–186. Schwannomatosis appears similar to NF-2 tumors but arises through germline mutation of SMARCB1 or LZTR1, and other genes yet to be discovered^187,188.

Another neural crest-derived cancer that has alterations in the NF2 gene is meningioma, a predominantly intracranial tumor of the meninges¹⁸⁹. Meningiomas (Fig. 3b) arise due to gene inactivation or mutations in the primordial meningeal cells, which arise from the neural crest or mesoderm, depending on the axial level^190,191. Lineage-tracing techniques have demonstrated that the forebrain meninges are neural crest-derived^192,193. In contrast, the mid-hindbrain meninges are mesoderm-derived, closely recapitulating the amniote skull vault origins^192,193. Accordingly, benign meningiomas can be traced back to their origin¹⁹⁴. Oncogenic mutations have been identified in AKT1, SMO, KLF4, TRAF7, and in genes encoding the epigenetic modifiers KDM5C, KDM6A, and SMARCB1 in tumors lacking NF2 mutations¹⁹⁵. Epigenetic profiling of global H3K27 acetylation of meningiomas robustly stratifies the tumors into three biological groups: adipogenesis- or cholesterol-related, mesodermal, and neural crest-derived. Interestingly, the neural crest-derived subtype has several features in common with pediatric radiation-induced meningiomas and Schwannomas, including poor prognosis, WHO grade II & III tumors, male predominance, and rapid recurrence¹⁹⁶. The molecular underpinnings of meningioma and the potential role of neural crest GRNs are an emerging theme that require further investigation^197,198.

NCC-derived tumors are notoriously difficult to treat, partly because of their developmental plasticity and ability to switch between cell states. These tumors frequently metastasize to distant locations derived from different germ layers. For example, the most common site for neuroblastoma metastasis are bone and bone marrow (mesoderm)^199–201, whereas melanoma frequently metastasizes to the lungs and liver (endoderm)^202,203. Interestingly, recent studies indicate that NCCs may either hold onto or regain a pluripotency-like profile during early development. This could explain their capacity to produce mesenchymal and neuroglial derivatives in vivo and endoderm signatures under in vitro conditions. In a zebrafish model of Ewing Sarcoma, overexpression of the onco-fusion protein EWSR1:FLI1 led to the reprogramming of trunk NCCs to a mesoderm-like state by hijacking developmental enhancers of neural crest genes responsible for mesoderm reprogramming²⁰⁴. This raises the question of whether neural crest-derived tumors also retain the ability to be pluripotent or multipotent-like cells. If so, when do they acquire these properties? Although in vivo cancer initiation events are challenging to isolate, the advent of low input, high throughput techniques hold the promise of uncovering these transient populations and, thus, new therapeutic strategies.

Conclusion

By tracing neural crest gene regulatory programs from induction to differentiation of diverse derivatives, the past decade has provided remarkable insights into the increasing complexity and evolution of the neural crest, a uniquely vertebrate cell type. Technological advances such as CRISPR, single-cell biology, and machine learning have enabled a more comprehensive examination of neural crest gene regulatory events. This has informed not only upon neural crest development and evolution, but also possible roles in adult tissue regeneration and tumor formation.

Despite significant progress, our understanding of the neural crest GRN remains incomplete. Although modules of the GRN have been identified, the subcircuits often remain hypothetical. Emerging high-throughput technologies provide unbiased approaches for exploring the neural crest GRN, moving beyond traditional candidate-driven approaches. Furthermore, comparative studies of non-model organisms at key phylogenetic positions using these technologies offer unique opportunities to reconstruct the ancestral GRN and uncover the mechanisms underlying neural crest regionalization and identification of potential species-specific circuits. Future research should target all neural crest subpopulations, integrate current technological advancements, and take advantage of comparative analyses. Although continued advances in sequencing technologies and machine learning will provide deep insights, validating and functionally testing these predictions in developing embryos remains essential.

Although several core genes have been identified for neural crest differentiation modules related to chondroblasts, melanocytes, and enteric neurons, there is a paucity of information regarding vertebrate evolutionary novelties originating from the neural crest, such as odontoblasts and osteoblasts. Since the earliest bones likely evolved within the dermal exoskeleton, deciphering these subcircuits may clarify whether the developmental program of endochondral bones represents an adaptation or an integration of the dermal exoskeletal program.

Beyond developmental and evolutionary insights, understanding the neural crest GRN also has significant medical relevance. Neural crest-derived tumor cells and repair cells appear to reactivate embryonic neural crest developmental programs. For example, a single mutation in a neural crest kernel could collapse parts or the entire GRN, potentially causing profound effects — beneficial or harmful — on development and regeneration. As the embryo is our greatest teacher, future GRN studies that couple “modern-omics” approaches with classical embryological validation and functional testing may hold the key for unlocking the secrets of the neural crest.

Glossary

New head hypothesis

An evolutionary developmental concept proposing that the emergence of neural crest cells and cranial placodes enabled the transition from filter-feeding chordates to vertebrates by enabling the formation of a complex head with centralized nervous system, sensory organs, jaws, and teeth.

Chromaffin cells

Neuroendocrine cells of the adrenal medulla that mediate physiological stress responses induced by various external stimuli.

Ephrins–Eph signaling

A cell-cell communication system mediated by interactions between Eph receptors, a family of receptor tyrosine kinases, and their membrane-bound ephrin ligands that is essential for normal development and implicated in disease processes such as cancer.

Genome-wide association studies (GWAS)

Genetic analytical method that assess genome-scale DNA variation across many samples to identify genomic regions associated with particular traits of disease susceptibility.

Planar cell polarity

Mechanism by which cells maintain a common directional orientation, enabling coordinated cell behavior during morphogenesis, and whose disruption results in severe developmental malformations.

Pre-cartilaginous condensations

Cluster of mesenchymal cells that aggregates at sites of prospective cartilage during early chondrogenesis.

Rostrocaudal

An anatomical directional running from the head (anterior) toward the trunk/tail (posterior) of the body axis.

Super-enhancers

Genomic regions composed of densely clustered enhancers that collectively drive high-level transcription of genes important for establishing and maintaining cellular identity or specialized function.

YAP–TEAD signaling pathway

A developmental signaling module often associated with the Hippo pathway in which YAP serves as a transcriptional co-activator with TEAD factors to regulate gene expression programs that govern cell proliferation, differentiation, and tissue growth during development.