Riding the crest to get a head: neural crest evolution in vertebrates

The neural crest gene regulatory network

Development of animal body plans is encoded in the regulatory genome, and modifications in gene regulatory programs lead to evolutionary change^37,38. Changes in morphological characters are driven by alterations in gene regulatory modules by gene innovation, gene duplication, and/or co-option of regulatory information from different tissues. Gene regulatory networks describe the regulatory interactions formed by transcription factors and cis-regulatory elements at each stage of development in a particular cell type³⁹. Integrated in these regulatory networks are cellular interactions mediated by signaling and receptor molecules and cues that dictate gene expression and function. Gene regulatory networks are broken into modules of regulatory information that include the genes and interactions that function at different time points of development. The given regulatory interactions in a cell at any given time represent its regulatory state⁴⁰. Conservation and changes in gene regulatory logic are at the core of our understanding of what drives evolution of the neural crest^2,9,26,41.

Formation of neural crest cells is controlled by a feed-forward gene regulatory network that controls their induction, migration, and differentiation (Figure 2d). As development proceeds, the neural crest progresses through successive regulatory states from specification to EMT to migration and ultimately to differentiation (Figure 2a)²⁴. Underlying each of these developmental milestones is a core gene regulatory network that describes the key interactions at each stage of development. The GRN is composed of developmental modules comprised of transcription factors and signaling molecules that interact in order to drive discrete steps of development (Figure 2d). By unraveling these networks, we can begin to understand the logic dictating how the neural crest arises, differentiates, and may have evolved to form unique derivatives in jawed vertebrates^2,26. The current view of the neural crest gene network has emerged over time and encompasses numerous organisms from basal vertebrates to mice and human pluripotent stem cells (hPSCs)^{2,23,24,42,43}.

Initiation of neural crest formation begins during gastrulation, as a series of signaling events including Wnts, FGFs, and BMPs refine the border between the forming neural and non-neural ectoderm^44–48. These signaling interactions promote regionalization along the mediolateral axis of the developing embryo and, at the presumptive neural plate border, activate downstream specification gene regulatory modules that include transcription factors such as Zic1, Msx1, Tfap2, and Pax3/7 (Figure 2d) ^{44,46,48–55}. Once the neural plate border has formed, the neural crest becomes specified as exemplified by expression of transcription factors including SoxE, FoxD3, Snai1/2, and Tfap2 ^52,54,56. Following specification, neural crest cells undergo an epithelial-to-mesenchymal transition to delaminate from the forming central nervous system. This delamination process is tightly controlled by regulatory interactions that coordinate a “Cadherin switch” to allow for the de-adhesion of precursors from the neural tube^57–62. Once free from the central nervous system, the neural crest cells activate a migratory gene network module to migrate extensively throughout the embryo^10,63–65. Upon reaching appropriate locations, they activate differentiation gene batteries that mediate differentiation into distinct derivatives based on their anteroposterior location (Figure 2d) ².

Along the anteroposterior body axis, the neural crest can be subdivided into four main subpopulations: cranial, vagal, trunk, and lumbosacral (Figure 2b)(Box 1). While neural crest cells at all axial levels form some common cell types like melanocytes and glia, there also are neural crest-derived structures that are unique to particular axial levels (Figure 2c). These unique derivatives are the consequence of deployment of axial-specific circuits that drive distinct fates related to their anteroposterior site of origin^13,66. For instance, in amniotes, only the cranial axial level is able to give rise to skeletogenic fates in vivo^15,16. Recently, a unique cranial specific circuit was found in chicken embryos that includes transcription factors Brn3c, Lhx5, Dmbx1, Tfap2b, Sox8, and Ets1²⁵. When components of this circuit (Sox8, Ets1, and Tfap2b) were ectopically expressed in the trunk neural crest, these were sufficient to impart skeletogenic potential by enabling trunk neural crest cells to form cartilage nodules after grafting to the head²⁵. However, since this circuit is insufficient to drive these fates when the trunk neural crest remained in the trunk, cranial-specific environmental cues and yet-to-be defined signals likely participate in the axial diversification of the neural crest. Still, understanding how subcircuits like this arose and evolved is important for elucidating how the neural crest was able to give rise to morphological novelties in different parts of the body.

At the vagal axial level, neural crest cells contribute to the heart and gut, forming the outflow tract septum and enteric nervous system (Figure 2c). When the anterior vagal (called “cardiac”) neural crest is ablated in chick embryos, the outflow tract septum which connects the heart to the lungs fails to form properly, resulting in mixing of oxygenated and non-oxygenated blood, a defect highly reminiscent of a common human congenital heart defect⁶⁷. Only cardiac neural crest cells have the ability to form the outflow tract septum whereas trunk and cranial neural crest cells cannot do so. Recently, a neural crest gene subcircuit, comprised of transcription factors Tgif1, Sox8, and Ets1, was shown to be specific to this axial level and, when introduced ectopically, was able to confer this developmental potential to trunk neural crest cells grafted to the cardiac crest region⁶⁸. Thus, similar to the cranial crest-specific subcircuit that can confer cartilage forming ability, the cardiac crest-specific subcircuit appears to confer the ability to form a different derivative, mesenchymal cells of the outflow tract septum, onto a neural crest subpopulation in a region-specific manner. Future studies focusing on axial specific subcircuits in other neural crest subpopulations hold the promise of clarifying how these subcircuits act to drive cell type diversification along the anteroposterior body axis.

Other likely important players in the formation of distinct subpopulations along the anteroposterior axis are the Hox genes (Box 2). Differential Hox gene expression and their interactions with other neural crest gene network genes may be sufficient to account for subtle gene network differences observed along the anteroposterior axis and may act to modulate neural crest axial level differences in cell fates⁶⁹.

At its core, the neural crest gene regulatory network is remarkably similar in overall architecture and composition across all vertebrates, though species specific differences enable flexibility in morphological traits such as craniofacial features. Still, the overall network is vastly similar and adaptable such that modular components, such as axial specific subcircuits and differentiation gene batteries can be “plugged in” to the network to add to its evolvability and adaptability.

Origins of the neural crest GRN

Across vertebrates, groups of transcription factors, including Pax3/7, Msx1, Zic1, Tfap2, Snai1/2, FoxD3, and SoxE, and their regulatory interactions, are conserved in terms of expression in the neural crest and placement in a pan-vertebrate gene regulatory network (Figure 2d) ^2,26,70. These factors serve as a kernel that functions to establish the neural plate border and promote neural crest multipotency and migration. However, important gene regulatory differences between jawed (gnathostome) and jawless (cyclostome) vertebrates provide clues as to how novel cell types may have evolved under the umbrella of the neural crest. Both lamprey and hagfish are cyclostomes and form a monophyletic sister group to the jawed vertebrates^71–73. Far more is known about the gene regulatory network of lamprey than hagfish neural crest since it is very difficult to obtain live embryos from the latter^74–77.

Interrogation of neural crest gene network conservation and changes in the sea lamprey can provide insight into the formation of morphological novelties. For instance, work in the sea lamprey has shown that transcription factors Ets1 and Twist, two major players in the pre-migratory/migratory regulatory module of the neural crest, are absent from the early neural crest GRN⁷⁰. Ets1, which is essential for neural crest specification in jawed vertebrates, is instead expressed in late neural crest derivatives within the branchial arches as well as dorsal root ganglia, a trunk-derived neural crest cell type^8,70. This represents a significant change in the gnathostome gene network from the early ancestral vertebrate gene regulatory network where Ets1 was co-opted from later in development, or a more distal part of the network hierarchy, to drive specification of gnathostome neural crest specification potentially leading to novel cell fates. Another transcription factor, Twist, which is essential in frogs but dispensable in mice for neural crest development, is absent in lamprey migratory neural crest but present in later derivatives, providing another example of a distal node of the network that was co-opted to more proximal parts of the neural crest gene network in gnathostomes⁷⁰.

A comparative analysis of the neural crest GRN that governs the ability of cranial neural crest cells to form the facial skeleton between amniotes and other vertebrates has shown that neural crest gene network components have been progressively added to the neural crest during the course of vertebrate evolution⁸. Several of the genes that are cranial crest specific in amniotes, instead of being added de novo to the cranial neural crest, appear to have been co-opted from more distal parts of the gene network to proximal modules at all axial levels then progressively restricted to the cranial axial level during the course of gnathostome vertebrate evolution⁸. Thus, basal vertebrates appear to have had a “basic” neural crest gene network that was relatively uniform along the body axis. With progressive evolution, genes were added and subcircuits built at different axial levels, resulting in subpopulations of neural crest cells with different migratory pathways and the ability to form distinct derivatives. The results of these recent comparative studies of gene regulatory control of neural crest development suggest that the New Head, as opposed to arising at the base of vertebrates in toto, arose progressively during the course of vertebrate evolution.

Two recent stories shed light on vertebrate-specific gene innovations and gene duplication events that enabled expansions and diversification of the neural crest. In one recent study, Scerbo and Monsoro-Burq (2020) show that the loss of vertebrate-specific Ventx2, an ortholog of mouse Nanog, leads to a loss of the neural crest multipotent state and skeletogenic potential⁷⁸. This genetic perturbation results in a neural crest that can only give rise to sensory neurons and pigmentation, similar to neural plate border derivatives found in invertebrate chordates. Further, Scerbo and Monsoro-Burq show that mouse Nanog is able to rescue the craniofacial phenotype of the Ventx2 depletion demonstrating a functional equivalence of Ventx2 and Nanog. Another recent study reports on the significance of genome duplication events that led to the expansion and diversification of neural crest subpopulations during vertebrate evolution. Endothelin ligands and receptors are unique to vertebrates and two rounds of genome-wide duplication events that occurred in basal vertebrates, the Edn signaling pathways components diverged and became specialized in order to expand the neural crest repertoire^79,80. These examples provide further evidence that throughout chordate evolution, the neural crest gene regulatory network was progressively elaborated to give rise to vertebrate novelties.

The advent of new systems-level techniques that are amenable to many research organisms has shed light on regulatory network changes and additions that drove the evolution of novel morphological characters of the neural crest. Initially, neural crest gene network interactions were studied by taking a candidate approach to identify genes expressed in the neural crest and then testing the effects of gene knock-down on other known neural crest markers. Recently, next-generation sequencing techniques including bulk and single cell RNA-seq, ChIP-seq, CUT&RUN, and assays for transposase-accessible chromatin using sequencing (ATAC-seq) have been applied to the study of neural crest development in several species ranging from jawed to jawless vertebrates^9,10,81,82. By comparing the global gene regulatory networks between these diverse vertebrates, it is possible to glean changes in the neural crest that have occurred as a function of evolutionary time.

Lending to our understanding of how novel programs evolve, recent single cell analyses throughout neural crest development in the mouse suggests a three-step fate selection mechanism where multipotent neural crest cells co-activate opposing regulatory programs for different fates, followed by repulsion of one program and commitment to a distinct fate⁸². From an evolutionary perspective, it is interesting to speculate that mechanisms such as these are at play to give rise to novel derivatives by co-option of fates from other cell lineages.

The gene regulatory network is not only useful for assessing regulatory changes that have occurred in the vertebrate lineage, but also for uncovering the homology of similar cell types in invertebrate chordates that provide clues regarding the origins of the neural crest (Figure 2d). Recent work in invertebrate chordates based on comparative gene regulatory network analyses suggest a more primitive origin of the neural crest than previously assumed.

Comparative approaches at gene network dissection help to uncover the foundations for evolutionary change via changes in linkages or subcircuits within the gene network, which in turn can inform upon new morphological novelties. Using information gathered from comparative regulatory analyses of GRNs, one can infer whether different morphological characters may have convergently evolved by parallel deployment of differentiation gene batteries. The more we learn about the neural crest gene network, the more we understand how mechanistic changes in the regulatory program have influenced the evolving vertebrate body plan and New Head.

Invertebrate neural crest-like cells

Central to the understanding of vertebrate evolution is uncovering when bona fide neural crest first appeared. Recent evidence from invertebrates suggests that the neural crest evolved in a step-wise fashion throughout the evolution of deuterostomes. Cell types that share neural crest features such as molecular signatures, location of origin, and derivatives may represent a lineage that is homologous to the neural crest, co-opted from other tissues and incorporated into an evolving neural plate border population⁸³. While one cannot rule out that these cell types arose by means of convergent evolution, evidence suggests that the cell types and regulatory programs implemented in the formation of these cell types in invertebrate chordates may represent an ancestral state. As a case in point, the neural crest-like cell types that have been identified to date lack multipotency and extensive, long-range migratory ability. Comparative gene regulatory studies have now enabled the investigation into neural crest-like cell type evolution in invertebrates by assessing the presence of regulatory programs in cell types that don’t necessarily have all distinguishing characteristics of the vertebrate neural crest cells like multipotency or long-range migratory ability yet share some common gene signatures. Future comparative gene regulatory studies aimed at uncovering why invertebrate chordate neural crest-like cells lack multipotency programs is important for understanding the origins of vertebrate neural crest.

In urochordates, cell types similar to pigment cells and neurons have been found with gene regulatory similarities to neural crest and cranial placode cells (Box3) that reflect a preliminary neural crest gene network (Figure 2d). This network consists of homologues to Id, Zic, Pax3/7, Mitf, Msx, Snai, Ets1, and FoxD that are expressed in cells that are located at edges of the lateral neural plate; however, the cells within these expression domains neither migrate extensively nor retain multipotency properties to give rise to a wide variety of derivatives (Figure 2d) ^4,35. Further, progenitors of the pigmented cells in the lineage of otolith and ocellus, derived from the a9.49 cell lineage in the ascidian Ciona intestinalis, normally remain in the central nervous system but have the ability to extensively migrate upon misexpression of Twist⁴. Finally, a cell type that originates from the b8.20 and b8.18 cell lineages, has been found to arise in the posterior lateral plate border, then migrate away from the central nervous system to give rise to neurons call bipolar tail neurons (BTNs) that are similar to vertebrate sensory neurons of dorsal root ganglia. BTNs express transcription factors Neurog and Islet, which are both required for vertebrate sensory neuron differentiation, suggesting that these cells represent neural crest-like cells or placode-like cells on both a morphological and molecular level³.

Beyond derivatives formed from neural crest-like cells, it was also recently shown that there are parallels between the compartmentalization of the lateral plate in Ciona and the neural plate in vertebrates. Both systems require similar network interactions in order to drive the formation of different sub-domains across the lateral organization of the neural plate, including Six1/2, Pax3/7, and Msxb⁸⁴. In urochordates, this organized lateral plate will give rise to sensory cells that are similar to both vertebrate cranial placode and neural crest derivatives. Furthermore, relatively minor gene network perturbations lead to a fate switch of one sensory cell type to another. These data suggest that cranial placodes and neural crest may have arisen from a common precursor population and only after reorganization of the lateral plate did the pre-placodal domain become distinct from the rest of the neural plate border (Box 3) ⁸⁴.

In the cephalochordate amphioxus, homologues of neural plate border specifiers Msx, Zic, and Pax3/7 are expressed in cells that are found in the lateral edges of the neural plate⁸⁵. Amphioxus also exhibits expression of Snail in an ependymal cell in the neural tube, but this cell type is not migratory⁸⁶. However, while many of the genes that are expressed in the neural crest in vertebrates are present in the cephalochordate genome, none of the cells that express these genes are multipotent, migrate, or differentiate into neural crest cell types^87,88. Still, it may be possible that the amphioxus possesses cells with some homology to neural crest cells, such as pigment cells of the ocellus, but this still requires further investigation.

The New Head hypothesis states that neural crest and cranial placodes arose in rudimentary form in urochordates, but new evidence suggests a more primitive origin for neural crest-like cells in sea urchins, a basal deuterostome. Recently, it was reported that a population of neurons in the sea urchin, Lytechinus variegatus, share similar features to BTN cells of urochordates, which share features with neural crest cells⁸⁹. These ciliary band neurons arise at the border of the neuroectoderm and non-neural ectoderm in the sea urchin larva, migrate from bilateral sites of origin, express ngn, and differentiate into afferent sensory neurons that are required for swimming behavior⁸⁹. One possible interpretation is that appearance of the neural crest lineage was not so sudden but rather, a neural crest-like condition was a continuous character that existed in multiple states and was remodeled in a step-wise fashion over the course of deuterostome evolution. A caveat, however, is that one cannot rule convergent evolution of some of these cell types.

From the data in invertebrates thus far, we can elaborate on crucial points in the New Head hypothesis involving origin of the neural crest and cranial placodes (Figure 3). One could speculate that regulatory programs were progressively co-opted from neighboring tissues by means of germ layer rearrangement and compartmentalization of the neural plate. The vertebrate acquisition of a multipotent state and more complex gene regulatory network modules resulted in a neural crest that gives rise to more elaborate derivatives, including ectomesenchymal cell types, by co-option of new differentiation gene batteries. It is also important to note that homologies drawn from molecular similarities alone are not conclusive but will need to be supplemented with more information on morphological and behavioral similarities, as well as more expression and genome-wide similarities.

Neural crest cell types in vertebrates

Comparisons between two major groups of living vertebrates, the jawed (gnathostome) and the jawless (cyclostome) vertebrates, have shed light on the origin of the vertebrate neural crest and the means by which it has evolved. By comparing extant vertebrate organisms, it can be concluded that emergence of the vertebrate lineage was accompanied by the introduction of neural crest cells that acquired novel derivatives, multipotency, and extensive migratory ability.

By studying the neural crest in these two groups, shared, derived traits of the early neural crest can be identified. Neural crest cell types that are shared among vertebrates include neurons and glia of the peripheral nervous system, pigment cells, cellular pharyngeal cartilage, cardiac valves, and chromaffin cells. However, many of these cell types, including cranial derivatives such as jaws and odontoblasts that produce dentin, a vagal-derived enteric nervous system, and trunk derived sympathetic ganglia, are absent in cyclostomes^34,90. These most likely arose in stem gnathostomes by modifications in gene regulatory network architecture that gave rise to new morphological novelties. Emergence of these novel gnathostome cell types also coincides with the refinement of neural crest axial subpopulations and their unique developmental potentials (Box1).

Assembly of axial specific transcriptional circuits occurred progressively throughout vertebrate evolution to give rise to distinct axial derivatives. Skeletogenic potential is a derived feature, arising later in vertebrate evolution but initially emerging along the entire anteroposterior axis⁸. Addition of cranial-specific circuits resulted in progressive restriction of skeletogenic fate to the cranial population in amniotes. In support of this, invertebrate chordate neural crest-like cell types lack skeletogenic potential but possess the ability to form pigment cells and neurons, traits common to all neural crest axial levels^3,4. These data suggest that the New Head arose progressively by first acquiring skeletogenic potential at all axial levels then becoming restricted to the cranial levels by addition of neural crest gene network nodes.

While jaws are a clear gnathostome novelty, the origin of the vertebrate head skeleton did not depend on the evolution of a new skeletal tissue, but rather on the spread of this tissue throughout the head and modification of the anterior pharyngeal arches⁹¹. While ectomesenchymal derivatives and gnathostome novelties such as jaws have been restricted to the cranial neural crest, odontoblasts, or cells that produce dentin, may have originated along the length of the body axis before becoming restricted to cranial regions. While the trunk neural crest is often regarded as non-skeletogenic in gnathostomes, it has recently been shown to give rise to an ectomesenchymal cell type in the little skate, Leucoraja erinacea⁹². Using DiI cell lineage tracing, Gillis and colleagues revealed that the dermal denticles in the trunk region are derived from neural crest cells, thus revealing a trunk origin for odontoblasts. A small circuit of transcription factors that is sufficient to confer ectomesenchymal ability in the trunk of amniotes was recently shown to be expressed along the length of the little skate anteroposterior axis, lending support to the presence of ectomesenchymal potential in the skate trunk neural crest cells⁸.

The enteric nervous system in gnathostomes arises from vagal and sacral neural crest populations^17,28. Recent evidence from Green and colleagues (2017) shows that lamprey lacks a vagal subpopulation of neural crest and only possesses cranial and trunk neural crest populations; however, the sea lamprey has an enteric nervous system. To determine the evolutionary origin of the vertebrate enteric nervous system, they performed DiI lineage tracing and found that the enteric neurons of the lamprey are derived from late migrating trunk neural crest-derived Schwann cell precursors³⁴. Further gene regulatory analyses of neural crest subpopulations across diverse species will augment understanding of the evolution of the neural crest and the vertebrate body plan.