Endothelin signaling in development Free

More than 30 years ago, the first member of the endothelin peptide family, endothelin 1, was discovered based on its potent vasoconstrictive activity (Yanagisawa et al., 1988), followed by the discovery of two other endothelin peptides, endothelin 2 (EDN2) and endothelin 3 (EDN3) (Inoue et al., 1989), their receptors, endothelin receptor type A (EDNRA) (Arai et al., 1990) and type B (EDNRB) (Sakurai et al., 1990; Sakamoto et al., 1991) and their activating enzymes endothelin converting enzymes 1 and 2 (ECE1 and ECE2) (Xu et al., 1994; Emoto and Yanagisawa, 1995).

Endothelin signaling is a vertebrate-specific innovation that likely emerged in the common vertebrate ancestor, before the divergence of jawless (cyclostome) and jawed (gnathostome) vertebrate lineages (Hyndman and Evans, 2007; Martinez-Morales et al., 2007; Braasch et al., 2009; Kuraku et al., 2010), in parallel with neural crest cells (NCCs) (Box 1) (Square et al., 2016, 2020). Depending on the ligand, receptor and cell and tissue types involved, endothelin signaling can direct NCCs to migrate, proliferate, differentiate or maintain an undifferentiated state. Thus, endothelin signaling contributes to the functional diversity of NCCs, a crucial step in vertebrate evolution (Donoghue et al., 2008; Karim et al., 2021), and plays a conserved role in the development of tissues derived from NCCs, such as the craniofacial skeleton, cardiac outflow tract, pigmentation and the enteric nervous system (ENS) (Clouthier et al., 2010; Bondurand et al., 2018). Although the role of endothelin signaling in the development of these tissues and organs appears to work largely in a cell-autonomous manner in NCCs, it also appears that in some instances (including in melanocyte and ENS development), endothelin signaling is also required in non-NCCs to support NCC contributions to these tissues and organs (Wu et al., 1999; Barlow et al., 2003; Hou et al., 2004).

Here, we describe the components of the endothelin signaling pathway and their respective functions during embryogenesis, while highlighting model organisms and genetic tools used in these investigations (Table 1). We also summarize endothelin-associated developmental differences in humans and approaches used to understand the genetic, cellular and molecular basis of disease (Table 2).

The endothelin family of peptides is comprised of EDN1, EDN2 and EDN3 in tetrapods (Barton and Yanagisawa, 2019) (Table 1; Fig. 1). Teleosts have a fourth endothelin peptide, EDN4, which was likely present in the common vertebrate ancestor but was lost in tetrapods (Braasch et al., 2009). The endothelin genes, EDN1, EDN2 and EDN3, encode an ∼200 amino acid preproEDN, which is a biologically inert precursor that is subsequently converted to a biologically active peptide by proteases. preproEDN is first cleaved by furin-like proteases to generate a 37-41 amino acid inactive precursor, referred to as big-EDN (Denault et al., 1995; Kim et al., 2012). Membrane-bound metalloproteases ECE1 and ECE2 then convert big-EDN into a mature, 21 amino acid peptide (Xu et al., 1994; Emoto and Yanagisawa, 1995). Expression of endothelin components is regulated by a number of events, including circadian regulation (Kozakai et al., 2014; Richards et al., 2014), cell sheer stress (Malek et al., 1993; Harrison et al., 1998) and epigenetics (Welch et al., 2013). The transcriptional regulation of EDN1 has been most extensively studied, with at least ten signaling pathways capable of influencing its expression (Stow et al., 2011). However, the transcriptional regulation of endothelin ligands during development is poorly understood.

Endothelin peptides act on two receptor subtypes, EDNRA and EDNRB (Arai et al., 1990; Sakurai et al., 1990) (Fig. 1). Mammals have single genes for EDNRA and EDNRB, but some species have additional paralogs due to whole-genome duplication events. Zebrafish, for example, have two paralogs for ednra (ednraa and ednrab) and ednrb (ednrba and ednrbb), though at least ednraa and ednrab can act in a partially functionally redundant manner during craniofacial development (Nair et al., 2007). Avians also have two paralogs for Ednrb (Ednrb and Ednrb2), which display different expression patterns and functions during melanocyte development (Lecoin et al., 1998; Pla et al., 2005; Harris et al., 2008). Xenopus have an additional Endothelin receptor type C, though its function in development is unclear (Karne et al., 1993). Although progesterone can synergize with GATA2 to regulate Ednra expression (Zhang et al., 2013), as with endothelin ligands, transcriptional regulation of endothelin receptors during development is poorly understood.

EDNRA and EDNRB have different affinities for endothelin peptides. EDNRB is nonselective and binds to all EDN peptides with equal affinity, whereas EDNRA is selective and has the highest binding affinity for EDN1, moderate affinity for EDN2 and little to no affinity for EDN3 (Arai et al., 1990; Sakurai et al., 1990; Sakamoto et al., 1991). As EDN2 does not have an apparent role in embryonic development (Chang et al., 2013), this ligand will not be further discussed. The ligand selectivity for EDN1 is crucial for EDNRA function during development, as EDNRA nucleotide variants that enhance EDN3 binding affinity result in aberrant EDNRA signaling and developmental differences in mice and humans (discussed below) (Gordon et al., 2015; Sabrautzki et al., 2016; Kurihara et al., 2023). Endothelin peptide and receptor pairs involved in embryonic development can be classified into two general categories; EDN1-EDNRA, which drives craniofacial and outflow tract development, and EDN3-EDNRB, which drives melanocyte and ENS development. One of the key determinants that link these peptide-receptor pairs to specific developmental events is the coordinated expression of receptors and peptides in cells that are in close proximity (Nataf et al., 1996, 1998b; Brand et al., 1998; Clouthier et al., 1998; Yanagisawa et al., 1998a; Parichy et al., 2000; Barlow et al., 2003; Nair et al., 2007; Krauss et al., 2014; Square et al., 2016).

Endothelin receptors are members of the Class A G protein-coupled receptor (GPCR) family (Isberg et al., 2016). GPCR signaling pathways are mediated by heterotrimeric G proteins, which are composed of a Gα subunit and an obligate Gβ/Gγ heterodimer (Gilman, 1987). There are four families of Gα subunits, Gq/11, Gi/o, Gs and G12/13, that each activate or inhibit specific signaling effectors to regulate the production of secondary messengers, activity of kinase pathways, or gene expression (Fig. 1) (Neves et al., 2002). One of the direct signaling effectors for Gq/11 is phospholipase Cβ (PLCB) isoforms, which generates the second messengers inositol triphosphate (IP3) and diacylglycerol (DAG), which in turn stimulate intracellular calcium release and activation of calcium-dependent signaling pathways (Kadamur and Ross, 2013). Gs activates adenylyl cyclase (AC) isoforms, which leads to the production of cyclic AMP (cAMP) and activation of protein kinase A (PKA), whereas Gi/o attenuates this pathway by inhibiting AC. G12/13 activates RhoGTPases, which promote cell migration through regulation of actin-cytoskeleton dynamics. Both EDNRA and EDNRB can couple to all four Gα family members, though coupling preferences vary based on cell and tissue type (Aramori and Nakanishi, 1992; Takigawa et al., 1995; Kitamura et al., 1999; Kawanabe et al., 2002). In cranial and cardiac development, Gq/11 is the primary signaling mediator for EDNRA, though the contributions of other Gα families have not been ruled out (Sato et al., 2008). In melanocyte and ENS development, PKA is implicated downstream of EDNRB, but the roles of Gs or Gi/o in regulating PKA activity are unknown (Barlow et al., 2003; Hou et al., 2004; Goto et al., 2013; Bondurand et al., 2018).

Most of the bone, cartilage, connective tissue and portions of the nerves of the craniofacial complex are derived from cranial NCCs (Schilling and Le Pabic, 2014) (Fig. 2). NCCs migrate from the posterior midbrain and hindbrain rhombomeres (r) 1 and 2 into the mesenchyme of the first and second pharyngeal arches, transient bilateral structures on the ventral surface of the embryo (Serbedzija et al., 1992; Chai et al., 2000; Frisdal and Trainor, 2014). NCCs in the maxillary prominence of arch one are specified to adopt a maxillary identity, whereas cells in the mandibular portion of arch one are specified to adopt a mandibular identity (Minoux and Rijli, 2010) (Fig. 3A). The mandibular pharyngeal arch is further segmented along the dorsal-ventral axis into proximal, intermediate and distal patterning domains (Fig. 3A,B) that each give rise to specific lower jaw/middle ear structures (Fig. 3C).

Targeted deletion of Ece1, Edn1 or Ednra in mice results in neonatal lethality and numerous craniofacial anomalies, including retrognathia, homeotic transformation of the lower jaw into to upper jaw-like structures, mispositioning and malformation of the outer ears (pinnae), dysmorphic middle ear structures and malformation and mispositioning of the hyoid bone in the throat that results in mechanical asphyxia (Kurihara et al., 1994; Clouthier et al., 1998; Yanagisawa et al., 1998b; Ozeki et al., 2004) (Fig. 3D). In zebrafish, two large forward genetics screens yielded a collection of mutant alleles with ventral cartilage and jaw joint defects (Nusslein-Volhard, 2012), with several of these alleles mapping to genes that encode components of the EDN1-EDNRA pathway, including edn1, furina, plcb3 and mef2ca (Miller et al., 2000, 2007; Walker et al., 2006, 2007). Zebrafish embryos injected with morpholinos targeting edn1, ednraa or ednrab result in similar changes in ventral cartilages and loss of jaw joints (Kimmel et al., 2003; Nair et al., 2007). In humans, gene variants in signaling components within the EDN1-EDNRA pathway are associated with Auriculocondylar syndrome (Rieder et al., 2012; Clouthier et al., 2013; Gordon et al., 2013a; Kanai et al., 2022) and Mandibulofacial Dysostosis with Alopecia (Gordon et al., 2015; Kurihara et al., 2023), both of which present with facial differences, and Oro-Oto-Cardiac syndrome, which presents with both craniofacial and cardiovascular differences (Pritchard et al., 2020) (Table 2).

The mandibular pharyngeal arch is classically divided into proximal, intermediate and distal domains, with distal and intermediate domain patterning dependent on autonomous EDNRA signaling in NCCs. In mice, Ednra is expressed in migrating and post-migratory cranial NCCs (Clouthier et al., 1998). Edn1 expression is observed in the ectoderm, mesoderm and endoderm of the mandibular pharyngeal arches (Clouthier et al., 1998; Yanagisawa et al., 1998a; Miller et al., 2000), though studies in mice and zebrafish illustrate that primarily ectodermal EDN1 drives arch patterning (Nair et al., 2007; Tavares et al., 2012) (Fig. 3A,B). Ece1 is expressed by the ectoderm and underlying arch mesenchyme (Yanagisawa et al., 1998a, 1998b; Tavares et al., 2017). Similar expression patterns for edn1 and ece1 are observed in zebrafish (Miller et al., 2000), with ece1a expression induced by Nkx2.5/Nkx2.7 (Ikle et al., 2017). The Dlx family members Dlx5 and Dlx6, essential components of the gene regulatory network (GRN) governing mandibular identity (Beverdam et al., 2002; Depew et al., 2002; Ozeki et al., 2004; Ruest et al., 2004; Talbot et al., 2010; Vieux-Rochas et al., 2010) (Fig. 3A,B) are induced upon EDNRA activation and are required for subsequent induction of the basic helix-loop-helix transcription factor Hand2 in the distal domain (anlage of incisors and most of the mandible) and Nkx3.2 and Dlx3 in the intermediate domain (malleus, incus and the proximal mandible) (Thomas et al., 1998; Clouthier et al., 2000; Charité et al., 2001; Miller et al., 2003; Fukuhara et al., 2004; Tucker et al., 2004; Talbot et al., 2010; Barron et al., 2011; Funato et al., 2016). Although EDNRA signaling activity may directly induce mandibular patterning genes, a study in zebrafish suggests a permissive mechanism where EDNRA downregulates expression of the Nr2f family of nuclear receptors, which normally repress the expression of mandibular patterning genes (Barske et al., 2018).

EDN1-EDNRA signaling is mediated in part by Gq/11 and its downstream effectors PLCB and myocyte enhancer factor 2 (Mef2c), a MADS Box transcription factor (Offermanns et al., 1998; Ivey et al., 2003; Dettlaff-Swiercz et al., 2005; Miller et al., 2007; Verzi et al., 2007; Walker et al., 2007; Kanai et al., 2022) (Fig. 1). Disruption of Gq/11, PLCB isoforms or Mef2c by targeted deletion or mutations causes downregulation of endothelin-dependent patterning genes and produces craniofacial phenotypes reminiscent of endothelin loss. In mice, endothelin induces expression of Mef2c by direct transcriptional regulation, which is mediated in part by Ca2⁺/calmodulin-dependent protein kinase II (CaMKII) (Hu et al., 2015). In zebrafish, the expression of mef2ca is independent of endothelin signaling, though it is required for transduction of Edn1-mediated signaling (Miller et al., 2007).

When EDN1-EDNRA signaling is attenuated in mouse and zebrafish, NCCs of the ventral and intermediate domains fail to establish a mandibular identity and instead adopt a maxillary identity, resulting in homeotic transformation of most lower jaw structures and soft tissue into upper jaw-like structures and tissues (Kimmel et al., 2003; Miller et al., 2003; Ozeki et al., 2004; Nair et al., 2007; Ruest and Clouthier, 2009; Zuniga et al., 2010; Tavares et al., 2017; Barske et al., 2018) (Fig. 3C,D). In zebrafish, the second arch-derived operculum bones are also duplicated at the expense of the branchiostegal rays (Kimmel et al., 2003). Similar changes are observed when Ednra signaling is disrupted in chick, frog and humans, indicating the conserved nature of this developmental mechanism (Kempf et al., 1998; Miller et al., 2000; Bonano et al., 2008; Pritchard et al., 2020; Kanai et al., 2022).

The window of necessity for EDN1-EDNRA signaling is between the arrival of NCCs in the arch and the termination of migration (Ruest and Clouthier, 2009). In wild-type mice, administration of an EDNRA-specific antagonist by gavage between embryonic day (E) 8.25 and E9.25 (either once or twice separated by 12 h) resulted in defects in jaw and middle ear structures, with treatment between E8.75 and E9.25 producing the strongest effect (Fukuhara et al., 2004; Ruest and Clouthier, 2009). Treatment after E9.5 had no effect on craniofacial development. Gene expression changes at E10.5 in the mandibular arch of antagonist-treated embryos and in cultured E10.5 mandibular arches treated in vitro with an antagonist (Fukuhara et al., 2004) were similar to those observed in Ednra^−/− embryos. Conditional ablation of the Ednra gene in mice using the Wnt1-Cre strain, which is active around E8.0, also recapitulated the jaw phenotype of Ednra^−/− embryos, while conditional ablation of the Ednra gene using the Hand2-Cre strain, which results in gene deletion after E9.25, did not disrupt jaw development. This further highlights the strict temporal requirement for EDNRA signaling in craniofacial development.

The boundary between upper and lower jaw patterning domains is established by the antagonistic interactions between EDN1-EDNRA and JAGGED-NOTCH signaling pathways. NOTCH is a cell-surface receptor for JAGGED and other related transmembrane ligands (Mumm and Kopan, 2000). In zebrafish, Jag1b and Notch2 establish the dorsal patterning domain (for maxilla and upper face) while also antagonizing the Edn1-Ednra signaling pathway to correctly position the dorsal/intermediate domain of the first and second pharyngeal arches (Zuniga et al., 2010; Barske et al., 2016). In contrast, Edn1-Ednra signaling antagonizes Jag1b-Notch2 signaling by inhibiting expression of jag1b and transcriptional targets, such as hey1 (Zuniga et al., 2010; Barske et al., 2016). Absence of Edn1-Ednra signaling causes Jag1b-Notch2 signaling activity to expand ventrally, subsequently contributing to the observed homeotic transformations. JAGGED-NOTCH signaling may have species-specific roles in the dorsal arches; mouse embryos lacking Jag1 had midfacial hypoplasia, a persistent foramen in the frontal bone and defects in the hyoid bone; however, defects in the upper and lower jaw were not observed (Teng et al., 2017).

The regional confinement of both EDN1-EDNRA and JAGGED-NOTCH signaling is likely controlled by several mechanisms, one of which involves the transcription factor SIX1. In mice, SIX1 represses Edn1 expression in the dorsal portion of the first pharyngeal arch endoderm while also inducing Jag1 expression (Tavares et al., 2017) (Fig. 3A,B). Targeted inactivation of Six1 leads to an expansion of Edn1 expression along the dorsal endoderm, resulting in aberrant EDNRA signaling and downregulation of Jag1 expression in the proximal arch. Consequently, the zygomatic process of the maxilla expands into a pseudo-condylar process that inserts into the temporal mandibular joint (TMJ) adjacent to the true mandibular condyle.

Although the mandibular arch has historically been divided into three primary domains, recent advances in single-cell RNA-sequencing in the mouse have illustrated that the arch can be subdivided further into at least 15 clusters of cells with distinct gene expression patterns (Xu et al., 2019; Yuan et al., 2020). This indicates that induction of EDNRA signaling leads to establishment of multiple developmental domains within the mandibular arch that likely drive patterning and development of distinct elements (or portions of distinct elements).

EDN1-EDNRA also works in concert with other signaling pathways to pattern the mandibular arch. In zebrafish, signaling by bone morphogenic protein (Bmp) ligands and Bmp receptors (Bmpr) initiates expression of edn1 in the ventral ectoderm (Alexander et al., 2011), which then upregulates expression of a Bmp antagonist, Gremlin 2, in the intermediate and dorsal domains, restricting Bmp activity to the ventral domain (Zuniga et al., 2011). This suggests that Bmp is first necessary to induce Edn1, with Bmp and Edn1 then adopting independent roles in establishing the ventral and intermediate domains, respectively. Although this has been supported by computational models (Meinecke et al., 2018), it is unclear why endogenous Bmp cannot compensate for complete loss of Edn1-Ednra signaling in mouse and zebrafish. This may be due in part to Bmp4 expression being dependent on EDN1-EDNRA signaling in mice (Ruest et al., 2004). Furthermore, loss of BMP signaling in mice causes lower jaw hypoplasia and malformations but does not result in the homeotic transformations observed in Edn1 and Ednra mutants (Bonilla-Claudio et al., 2012). Additional work is needed to clarify the relationship between BMP-BMPR and EDN1-EDNRA signaling pathways.

WNT signaling also appears to work upstream of or in parallel with BMP-BMPR and EDN1-EDNRA during mandibular arch patterning. Disruption of Wnt signaling in zebrafish leads to loss of ventral arch gene expression and expansion of dorsal gene expression into the ventral arch, resulting in craniofacial phenotypes resembling edn1 mutants (Alexander et al., 2014). These changes occur due to a likely requirement for WNT signaling in establishing competence of cranial NCCs to respond to Bmp and Edn1. In mice, targeted inactivation of a WNT co-receptor, Lrp6, disrupts arch gene expression and later lip fusion but does not affect lower jaw development (Pinson et al., 2000; Song et al., 2009). In addition, targeted deletion of R-spondin 2 (Rspo2), a positive regulator of WNT/β-catenin signaling, results in lower jaw defects in mice, potentially by regulating Edn1 expression either directly or through FGF8 (Jin et al., 2011). In addition, targeted deletion of Fgf8 in the pharyngeal arch ectoderm in mice disrupts expression of multiple mandibular arch patterning genes, including Edn1 (Trumpp et al., 1999). However, the resulting phenotype in Fgf8 mutant mice (arch hypoplasia and subsequent loss of most mandibular arch-derived structures) does not resemble the Edn1 mutant phenotype (Kurihara et al., 1994; Ozeki et al., 2004). These discrepancies motivate additional work to clarify the relationship between WNT, BMP, EDN1 and FGF8 signaling pathways. Finally, Sonic hedgehog signaling is a crucial downstream mediator of EDN1 activity, working with HAND2 to pattern the lower jaw (Elliott et al., 2020) while also confining BMP activity (Xu et al., 2019).

EDN1-EDNRA signaling is required for several aspects of cardiac development, including remodeling of the cardiac outflow tract, ventricle septation and valve formation, as well as the cardiac conduction system.

Each pharyngeal arch initially has an associated artery, with the paired arch arteries 1 and 2 regressing into capillary beds of the face and neck (Kirby and Waldo, 1995; Waldo et al., 1996). Paired right and left pharyngeal arch arteries (PAAs) 3, 4 and 6 give rise to the cardiac outflow tract (Kirby et al., 1997), comprised of the ascending aorta, aortic arch, common carotid arteries and portions of the subclavian arteries (Fig. 4A,B). This remodeling is driven by ventrally-migrating cardiac NCCs that surround the endothelial cells demarcating each nascent arch artery lumen (Fig. 5) and eventually form the smooth muscle portion of the vessels (tunica media) (Kuratani and Kirby, 1991; Hutson and Kirby, 2007; Waldo et al., 1996). The endothelium of PAAs 3, 4 and 6 express Edn1 and Ece1, providing paracrine signaling to the surrounding Ednra-expressing cardiac NCCs (Kurihara et al., 1995; Clouthier et al., 1998; Yanagisawa et al., 1998a,b; Jiang et al., 2000). Ece1 is also expressed in the mesenchyme surrounding the arch arteries (Yanagisawa et al., 1998a,b). In Ednra, Edn1 or Ece1 mouse mutants, cardiac NCCs surround the PAAs but cannot facilitate remodeling, resulting in a variety of arch artery defects (Kurihara et al., 1995; Clouthier et al., 1998; Yanagisawa et al., 1998a,b, 2000) (Fig. 4C-G). The mechanisms by which cardiac NCCs coordinate vessel remodeling is not fully understood.

A subset of cardiac NCCs continue past the arch arteries, migrating into the cardiac outflow cushions, where they form two prongs of condensed cells in the distal outflow cushions (Fig. 5) (Kirby et al., 1983, 1985; Kirby and Waldo, 1995). Along with a shelf of tissue in the aortic sac, they form the aorticopulmonary septation complex, which divides the singular outflow into the pulmonary and aortic outflows and aligns the outflows with the respective ventricles. NCCs also contribute to both semilunar and atrioventricular valves and form the muscular portion of the ventricular septum separating the right and left ventricles (Waldo et al., 1998). EDNRA signaling, acting through CDC42, is required for cardiac NCC migration through the outflow tract (Fritz et al., 2019), with its absence resulting in persistent truncus arteriosus (failure to separate the aortic and pulmonary outflows) (Clouthier et al., 1998; Yanagisawa et al., 1998a,b). Compared with Ednra, Edn1 and Ece1 mutant embryos, defects in aorticopulmonary septation are more severe in embryos lacking both ECE1 and ECE2 [Ece1/Ece2 double knockout (dko) embryos] (Yanagisawa et al., 2000). Conotruncal alignment defects are also observed in the absence of EDN1-EDNRA signaling, including overriding aorta and perimembranous interventricular septal defects (Clouthier et al., 1998). In addition, a majority of both Ece1 knockout embryos and Ece1/Ece2^dko embryos die between E13.5 and E18.5. Both groups show poor endocardial cushion development along with peripheral dilation and general edema, suggesting that lethality arises from defects in atrioventricular valve development (Yanagisawa et al., 2000). The observed edema could also arise from poor myocardial development. Ednra^−/− embryos have a thin myocardium (Clouthier et al., 1998; Asai et al., 2010), potentially due to an early requirement for EDNRA signaling in a region of the cardiac crescent from which ventricular myocardial cells arise (Abu-Issa and Kirby, 2007; Asai et al., 2010). Overall, these findings illustrate a broad requirement of EDN1-EDNRA signaling in cardiac NCC differentiation.

In the chick, endothelin signaling is required for cardiac conduction system development (Gourdie et al., 1998; Takebayashi-Suzuki et al., 2000; Hall et al., 2004). Embryonic myocytes ubiquitously express Ednra, whereas Edn1 and Ece1 are expressed by endocardial and arterial endothelial cells (Takebayashi-Suzuki et al., 2000). Such an arrangement argues for a role of Ednra signaling in myocardial development. Indeed, ectopic expression of Edn1 and Ece1 in myocytes results in cell-autonomous expression of Cx42, sMyHC and AMyHC, which are markers for Purkinje fibers, the impulse conducting cells of the heart (Takebayashi-Suzuki et al., 2000). This suggests that Ednra-expressing myocytes that are in direct association with Edn1/Ece1⁺ endothelial cells of the heart drives formation of the Purkinje lineage (Mikawa and Fischman, 1992; Takebayashi-Suzuki et al., 2000). The transcriptional regulation of Ece1 in this setting is not well understood, but it appears that biomechanical forces exerted during cardiac development induce Ece1 expression in a spatially restricted pattern necessary for proper conduction system architecture (Hall et al., 2004).

EDN3-EDNRB signaling is required for the development of melanocytes, which provide pigmentation to skin and hair and contribute to intermediate cells in the stria vascularis of the mammalian cochlea (Mort et al., 2015; Ritter and Martin, 2019) and the network of neurons and glia of the ENS that control the complex functions of the gastrointestinal tract (Schneider et al., 2019).

The roles of EDN3-EDNRB signaling in pigmentation and ENS development are widely conserved in vertebrates, though some species-specific differences exist. Spontaneous mutations in EDNRB that cause similar defects to pigmentation and/or ENS development have been identified in rat, horse, zebrafish, pig, goat and chicken (Bondurand et al., 2018). In humans, variants in EDN3, EDNRB or ECE1 are associated with isolated and syndromic Hirschsprung disease (HSCR) (Karim et al., 2021) (Box 2). Below, we discuss the roles of EDN3-EDNRB signaling in melanocyte and ENS development, highlighting species-specific differences.

Melanocytes are derived from NCCs from all axial levels (Freyer et al., 2011; Mort et al., 2015) (Fig. 2), with the role of EDN3-EDNRB in melanocyte development best characterized in trunk NCC derivatives. After delamination from the neuroectoderm, trunk NCCs transiently accumulate in a region between the dorsal neural tube and somites called the migration staging area (MSA) before bifurcating into two migratory paths that each give rise to two distinct NCC populations (Erickson et al., 1992) (Fig. 6A). The first population follows a ventral migratory pathway and gives rise to the neurons and glia of the peripheral nervous system, whereas the second population takes a dorsolateral path and gives rise to melanocytes (Serbedzija et al., 1990; Erickson and Goins, 1995; Reedy et al., 1998; Dupin and Le Douarin, 2003; Thomas and Erickson, 2008). The melanocyte progenitors, or melanoblasts, eventually invade and colonize the dorsal surface of the embryo. Throughout this process, Ednrb is expressed in trunk NCCs and melanoblasts, with Edn3 expressed in the dorsal ectoderm (Nataf et al., 1996, 1998b; Lee et al., 2003; Hou et al., 2004). In zebrafish, melanocytes are derived from both ventral and dorsal-migrating trunk NCCs (Mort et al., 2015). In Xenopus, melanocytes are derived primarily from the ventral-migrating populations of NCCs (Theveneau and Mayor, 2012).

Avian and rodent models have provided a great deal of insight into the role of EDN3-EDNRB signaling in melanoblast migration (Dupin and Le Douarin, 2003). In mice containing a lacZ cassette in the Ednrb locus (Ednrb^lacZ) (Lee et al., 2003), trunk NCCs first appear at the MSA around E10 and are specified into melanoblasts, indicated by the expression of dopachrome tautomerase (Dct) (Mackenzie et al., 1997). Subsequently, they migrate into and colonize the dorsal surface between E10 and E12.5. In Ednrb^lacZ/lacZ animals, lacZ⁺ and Dct⁺ melanoblasts appear in the MSA at E10 but fail to migrate into the dorsolateral path. These results indicate that EDN3/EDNRB signaling is required for dorsolateral migration of melanoblasts but not for fate specification of NCCs (Pavan and Tilghman, 1994; Lee et al., 2003). This is consistent with experiments in mice that show EDN3-EDNRB signaling is required between E10 and E12.5 for normal coat pigmentation (Shin et al., 1999).

EDN3-EDNRB signaling likely plays a cell-autonomous role in melanoblast migration (Druckenbrod et al., 2008), though the underlying mechanisms are not fully understood (Fig. 6B). Studies using avian models suggest that EDN3-EDNRB signaling serves a pathfinding role in the migratory process (Pla et al., 2005; Harris et al., 2008). Aves have two types of EDNRB receptors: Ednrb and its paralog Ednrb2 (Lecoin et al., 1998). Although trunk NCCs initially express Ednrb, the subpopulation that adopts a melanocyte fate downregulates Ednrb and upregulates Ednrb2, driving melanoblast migration through the dorsolateral pathway. While the extracellular environment of the dorsolateral pathway contains repulsive signaling molecules (Harris et al., 2008), melanoblasts appear to overcome these repulsive cues through the cooperative effects of Ednrb2 and EphB2, a gene that encodes for Ephrin receptor B2 (Harris et al., 2008). In contrast, trunk NCCs expressing Ednrb are unable to enter the dorsolateral pathway (Pla et al., 2005). Future work is needed to better understand the functional differences between Ednrb and Ednrb2 and whether similar signaling mechanisms underlie melanoblast migration in mammals.

Cell culture and explant assays have suggested that EDN3-EDNRB plays additional roles throughout different stages of melanocyte maturation (Saldana-Caboverde and Kos, 2010). In uncommitted trunk NCCs, EDN3-EDNRB inhibits differentiation and promotes proliferation (Lahav et al., 1996, 1998). After trunk NCCs have adopted a melanoblast identity, EDN3-EDNRB works cooperatively with KIT ligand and KIT tyrosine kinase to promote survival, proliferation and terminal differentiation to melanocytes (Lahav et al., 1998; Reid et al., 1996; Opdecamp et al., 1998; Ono et al., 1998; Hou et al., 2004; Aoki et al., 2005). Terminal differentiation is likely mediated by protein kinase C (PKC), a signaling effector downstream of EDNRB (Ono et al., 1998; Hou et al., 2004). The mechanism of signaling crosstalk between EDN3-EDNRB and KITL-KIT, and the G protein pathways that mediate PKC signaling in terminal differentiation, are not fully understood. Furthermore, it is unknown whether the effects of EDN3-EDNRB on cultured cells and explants also apply to trunk NCCs and melanoblasts in vivo. (Lahav et al., 1996, 1998; Reid et al., 1996; Opdecamp et al., 1998; Hou et al., 2004).

The ENS is comprised of neurons and glial cells that form a network of ganglions that regulate gastrointestinal functions such as peristalsis, secretion and absorption (Kang et al., 2021). These neurons and glial cells are derived from vagal and sacral NCCs, which we refer to collectively as enteric neural crest cells (ENCCs). Vagal NCCs contribute to the entire ENS, whereas contributions by the sacral NCCs are limited to a small proportion in the distal colon (Fig. 2, Fig. 6C) (Hutchins et al., 2018). In zebrafish, the ENS is likely derived exclusively from vagal NCCs (Kuil et al., 2020).

EDN3-EDNRB signaling is required for ENCCs to migrate into and colonize the hindgut, with loss of this signaling pathway resulting in an aganglionic distal colon and additional defects in ganglionated areas (Zaitoun et al., 2013; Cheng et al., 2016; Bondurand et al., 2018; Bhave et al., 2021). Similar defects are observed in humans with mutations in EDN3 or EDNRB (Box 2). Ednrb is expressed in both ENCCs and mesenchymal cells of the gut (Nataf et al., 1996, 1998a; Leibl et al., 1999; Wu et al., 1999; Barlow et al., 2003). Edn3 is expressed in the gut mesenchyme; in mice, expression occurs in a dynamic pattern over time, first being expressed in the midgut and hindgut (∼E10.5) before shifting to the cecum and proximal colon (∼E14.5) (Barlow et al., 2003) (Fig. 6C).

Distal colon aganglionosis in EDN3-EDNRB mutant animals is caused by the failure of ENCCs to migrate into the distal gut. ENCCs colonize the fetal gut in highly stereotyped migratory behavior in all model organisms studied thus far (reviewed by Bondurand et al., 2018; Kang et al., 2021). In mice, ENCCs derived from vagal NCCs adjacent to somites 1-7 enter the foregut around E9-E9.5 and advance through the intestine as a wavefront in a rostral-caudal direction (Young et al., 2004; Druckenbrod and Epstein, 2005) (Fig. 6C). By E11.5, the wavefront arrives at the edge of the ileo-cecal junction and stalls for ∼12 h before colonizing the hindgut (Kapur et al., 1995; Lee et al., 2003; Nishiyama et al., 2012). In Edn3 and Ednrb mutant mice, ENCCs reach the ileo-cecal junction but fail to advance into the hindgut (Kapur et al., 1992; Barlow et al., 2003; Kruger et al., 2003; Druckenbrod and Epstein, 2005) (Fig. 6C). Sacral NCCs also fail to properly colonize the distal colon in Ednrb mutant mice (Burns et al., 2000; Lee et al., 2003; Erickson et al., 2012). This ENCC migration defect is attributed to multiple mechanisms.

During gut development, proliferation and differentiation of ENCCs must be properly coordinated in order to maintain a sufficient pool of progenitor cells to colonize the entire gut (Nagy and Goldstein, 2006). This balance between proliferation versus differentiation is maintained in part by EDN3-EDNRB signaling, which promotes proliferation and inhibits differentiation of ENCCs (Lahav et al., 1996; Reid et al., 1996; Hearn et al., 1998; Wu et al., 1999). Loss of this signaling pathway causes premature differentiation of ENCCs into neurons and glial cells, thus depleting the pool of progenitor cells required to colonize the distal colon (Hearn et al., 1998; Wu et al., 1999; Lee et al., 2003; Bondurand et al., 2006; Nagy and Goldstein, 2006; Watanabe et al., 2017).

ENCC migration is also coordinated by functional interactions between EDN3-EDNRB and rearranged during transfection (RET) receptor tyrosine kinase and its ligand Glial cell line-derived neurotrophic factor (GDNF). GDNF-RET is a crucial signaling pathway in ENS development that is required for ENCC survival, proliferation and chemoattraction (Young et al., 2001; Natarajan et al., 2002). Both GDNF and EDN3 are highly expressed in the cecum when the ENCC wavefront arrives at the ileo-cecal junction and EDN3 not only augments the proliferative effects of GDNF but also antagonizes the chemoattractive properties of GDNF (Hearn et al., 1998; Barlow et al., 2003; Kruger et al., 2003; Nagy and Goldstein, 2006). This signaling crosstalk likely permits continued ENCC migration towards the hindgut. Further, the dual effects of EDN3-EDNRB on GDNF-RET signaling are mediated by inhibition of PKA, though the exact mechanisms of signal crosstalk is not yet fully understood (Barlow et al., 2003; Goto et al., 2013; Gazquez et al., 2016; Hao et al., 2019). EDNRB and RET also regulate one another's expression levels through a shared gene regulatory network (Chatterjee et al., 2016; Chatterjee and Chakravarti, 2019). The cis-regulatory elements (CREs) that control EDNRB and RET expression in ENCCs share binding motifs for transcription factors SOX10 and GATA2 (Chatterjee and Chakravarti, 2019). Further, expression of both Sox10 and Gata2 is positively regulated by EDN3-EDNRB and GDNF-RET signaling in a feedforward mechanism, suggesting that disruptions to one signaling pathway can also lead to disruptions to both. Further work is needed to better understand how interactions between EDN3-EDNRB and GDNF-RET at the levels of cell signaling and gene regulation impact ENCC biology during ENS development.

The ENCC migration defect is also likely caused by changes to the composition of extracellular matrix (ECM) proteins in the gut mesenchyme (Nagy and Goldstein, 2006). In Edn3 and Ednrb mutant animals, the expression of numerous ECM components such as collagens, proteoglycans, laminins and laminin receptors are upregulated in the distal colon (Tennyson et al., 1990; Rothman et al., 1996; Fu et al., 2020). Many of these ECM components inhibit ENCC migration in vitro, leading to the hypothesis that failure of ENCC migration into the distal hindgut is attributed in part to upregulation of inhibitory ECM components (Druckenbrod and Epstein, 2009). Contrary to this model, a recent study has found that a pro-migratory ECM component, laminin β1, is upregulated in the distal hindgut mesenchyme of conventional (Ednrb^lacz/lacz) and NCC-specific (Wnt1-Cre;Ednrb^flex3/flex3) Ednrb knockout mice (Fu et al., 2020). Further, application of YIGSR, a pentapeptide mimic of the receptor binding region of laminin β1, on whole gut explants from E12.5 Wnt1-Cre;Ednrb^flex3/flex3 mice restored ENCC colonization of the distal colon and promoted directional cell migration of ENCCs in midgut slice cultures from E12.5 wild-type mice in a GDNF-dependent manner. This effect was further enhanced by β1-integrin, another pro-migratory ECM component implicated to interact with EDN3-EDNRB (Gazquez et al., 2016). ENCCs in Ednrb mutant mice also exhibited downregulation and changes to subcellular localization of the 37/67-laminin receptor (a laminin β1 receptor). Elucidating the functional interactions between EDN3-EDNRB signaling and expression of ECM components will be essential for advancing stem cell-based regenerative therapies for HSCR (Box 2).

In addition to distal colon aganglionosis in Ednrb^−/− mice, structural and functional deficits have also been reported in ganglionated regions of the proximal colon, small intestine and stomach (Ro et al., 2006; Roberts et al., 2008; Zaitoun et al., 2013; Cheng et al., 2016; Bhave et al., 2021). These regions display abnormal intestinal motor function and ENS structure, characterized by irregular proportions of nitrergic and cholinergic neurons and irregularities in the number and sizes of ganglia and neuronal fiber density (Cheng et al., 2016; Bhave et al., 2021). The causative mechanisms are unknown, though comparison of ENCC-derived neuronal subtypes in the small intestine between E14 wildtype and Ednrb^−/− embryos using single cell RNA-seq data revealed that a cell cluster expressing Neurod6 and Gad2 is absent in Ednrb^−/− embryos (Bhave et al., 2022). These genes represent markers of GABAergic neurons, but further work is needed to understand how the loss of these neurons lead to functional deficits in the ganglionated regions of Ednrb^−/− mice.

Research over at least the last 30 years has clearly established that endothelin signaling is essential for many aspects of embryonic development, with its disruption contributing to congenital disorders in humans. In the case of HSCR, this knowledge has led to promising developments in a stem cell-based strategy to restore distal colon innervation. In addition, investigations across a wide variety of vertebrates have illustrated the deep evolutionary conservation of endothelin function in the development of tissues and organs that are fundamental for species survival. Looking ahead, new findings from endothelin research will be driven by ongoing innovations in genome editing tools, -omics, imaging technologies and biosensors for cell signaling. This next phase of endothelin research will undoubtedly provide novel insights into mechanisms of embryonic development, with major implications for improving the diagnosis and treatment of individuals with endothelin-associated congenital disorders.

The authors would like to thank the Walker family for sharing their experiences, the numerous scientists whose work is part of this review and Dr Masashi Yanagisawa, who supported and encouraged D.E.C. during the initiation of this work.