Hirschsprung disease (HSCR) is a birth defect which affects gut motility. The absence of ganglion cells in the enteric nervous system (ENS) has been identified as the culprit of typical presentations among HSCR patients, including delayed meconium (a baby's first bowel movement), constipation, bowel obstruction and vomiting.1 Although most cases of HSCR are sporadic, ∼ 30% have other congenital malformations including Down syndrome.2 HSCR is a multifactorial disorder and genetic mutations have been one of the aetiologies. Most of the associated genes are implicated in the key processes of ENS development: (1) the migration of enteric neural crest cells (NCCs) and their colonization of gut mesenchyme by, (2) their proliferation, (3) neurogenesis and gliogenesis and (4) formation of ganglionated plexuses.3,4 Anatomical defects in the ENS are attributed to the defects of NCC colonization/proliferation/differentiation among HSCR patients.

ENS is an extensive neuroglial circuit in gut, which originates from NCCs in the neural tube and is the largest unit of the peripheral nervous system, consisting of approximately 400-600 million enteric neurons in humans. Two ganglionated plexuses constitute the ENS, namely the myenteric ganglia located between the circular and longitudinal muscle layers, as well as the submucosal ganglia in the submucosa. Because ENS is responsible for regulating motility patterns such as peristalsis, aganglionic bowel has no propulsive motility and fails to relax during luminal distension, delaying the passage of stool. The diseased aganglionic segment of the colon cannot relax and pass stool through the colon, creating an obstruction, leading to a distension of the normal ganglionated bowel rostral to the diseased segment (Fig. 1).

Here, we describe the overview of ENS development and the underlying molecular mediators. Apart from discussing the role of ENS neural activity in gut motility, we also highlight how perturbation in the activity of critical molecules, such as receptor tyrosine kinase RET, can predispose HSCR patients to structural malformations and deterioration in gut motility.

Following gastrulation during the third week of gestation (i.e., the establishment of ectoderm, mesoderm, and endoderm), ectoderm which overlays the notochord proliferates, fabricating the neural plate. Neurulation occurs consecutively and neural folds (i.e., the lateral bulges of neural plate) fuse to form the neural tube, the future spinal cord. Neighbouring neuroectoderm, on the other hand, dissociates from these folds to become NCCs, which displaces to the underlying mesoderm through one of the two pathways: (1) a dorsal route to become melanocytes through dermis, and (2) a ventral pathway to constitute neurons (including the enteric neurons) via the anterior half of each somite. NCC cell fate decision is restricted through a series of lineage-restriction events that involve coexpression and competition of genes driving alternative fate programs.5, 6, 7 A somite is derived from the paraxial mesoderm: each segment, known as a somitomere, appears in the cephalic region initially and this process continues cephalocaudally until 42 to 44 pairs appear by the end of the fifth week of gestation.8 The development of somites is closely correlated with the NCC colonization of the gut.

Among the segments of neural crest, NCCs from the vagal level, together with the NCCs from the sacral level of the neural tube, contribute to the ENS. Studies on chick embryos and mice further refine the axial level origins of the enteric NCCs. These NCC-derived precursors migrate from the vagal level (corresponding to somites 1-7) and the sacral level (caudal to somite 28) to the entire length of the gut; however, there are regional differences within the vagal NCCs.9 In quail, NCC from somites 1-2 colonize the oesophagus solely; those in the vicinity of somites 3-6 populate the stomach and midgut; and their counterparts from the levels 6-7 migrate to the hindgut.10,11 Taken together, the NCCs from different axial levels of the neural tube constitute the ENS: (1) vagal NCCs from somite levels 1-2 which are destined for the oesophagus, (2) migrating multipotent cells from levels 3-7 which contribute to bowel from the stomach to the hindgut, and (3) sacral NCCs which colocalise the gut caudal to the umbilicus (Fig. 2).

To form the ENS, these precursors migrate ventrally away from the notochord during the closure of the neural tube. The earliest waves of vagal NCC-derived cells (at somites 1-4) propagate to the pharyngeal arches while those adjacent to somites 4-7 penetrate ventrally beyond the dorsal aortae in mouse embryos.12 By embryonic day 2.5 (E2.5-E3) in chick,13 E9.5 in mice14 and week 4 in human embryos,15 vagal NCCs reach the foregut and populate the entire gastrointestinal (GI) tract through a long rostral-caudal journey. This linear migration takes 5 days in mice during which the NCC derivatives reach the cecum in two days and completely colonize the colon in the additional three days. Among humans, vagal NCCs migrate to the midgut in 1-2 weeks and complete their lateral migration by colonizing the hindgut at around week 7 of gestation.15,16

One prerequisite to the development of the ENS is how NCCs become committed to the enteric lineages. Evidence suggests that the migration of inappropriate lineages (for instance the melanocytic lineage) to gut can be circumvented by the accumulation of retinoic acid (RA) in the vagal somitic environment. RA acts on RA receptors in the migrating lineages and herein stimulates the expression of RET within these cells.17 Otherwise, Mice, deprived of retinaldehyde dehydrogenase 2 (RALD2), which is a main synthetic enzyme of RA, show defects in vagal NCC colonization and consequently ENS development.18 Collectively, the somitic environment mediates the upregulation of the RET gene in vagal NCCs and drives the NCC colonization of the gut.

Sacral NCCs, like the vagal counterparts, also congregate along the hindgut. However, they will not invade to fabricate a minority of post-umbilical neurons until the vagal NCCs have fully colonized the hindgut at E14.5 in mice.19 The reasons underlying the sacral ‘waiting period’ are not well understood, but the transient expression of repellent ligand may shed light on the phenomenon that vagal NCCs are more effective in the gut colonization. For instance, sacral crest-derived cells can express roundabout (ROBO) receptors, so they are repelled by the proximal gut where the repulsive molecule of ROBO receptors, SLIT2, is expressed. Meanwhile, without the ROBO receptors, the lateral migration of vagal NCCs along the proximal gut is not affected.10,20 Direct interaction of SOX10 with Cadherin-19 has been recently shown to direct the migration of sacral NCCs into the hindgut in mice.21 Collectively, it is not unexpected that NCC colonization of gut requires delicate migration pathways and complex regulations.

Various molecules influence NCC colonization, but none are as prominent as RET and endothelin receptor type B (EDNRB). RET, expressed in NCCs, is a co-receptor for four ligands, namely the glial cell line-derived neurotrophic factor (GDNF), neurturin, artemin and persephin. GDNF belongs to the TGFβ family and signals through the RET receptor. Nonetheless, RET is only stimulated when GDNF binds to GDNF family receptor α1 (GFRα1) in association with the glycosyl phosphatidylinositol (GPI) anchor.22,23 Prior to NCC colonization of the gut, GDNF is already accumulated along a rostral-to caudal gradient from the foregut to cecum. NCCs are attracted to the foregut mesenchyme initially where GDNF binds to GFRα1 and activates the RET receptors on NCCs, inducing the NCCs to enter the mesenchyme and migrate as far as to the cecum. This GDNF/RET-GFRα1, in addition, orchestrates the trans-mesenteric migration of NCCs from midgut to colon.24 Homeobox protein HOXB5, expressed in NCCs, can bind to the first intron of the human RET gene and activate RET transcription.25 Perturbation of HOXB5 signalling in mice led to down-regulation of RET and a HSCR phenotype in mice.26 Another key component is EDNRB and its ligand endothelins. Endothelin 3 (ET-3), after being cleaved by endothelin converting enzyme (ECE-1), becomes active along the gut mesenchyme and demonstrates the highest concentration in the cecum.4,27 Synergistically, this endothelin gradient can promote temporal and spatial rostral-to-caudal migration of NCCs along the gut mesenchyme in concert with the GDNF/RET pathway.

Mutations of RET and EDNRB genes are frequently associated with HSCR and affect not only the migration but also the proliferation and differentiation of NCCs (discussed in detail below). Induced pluripotent stem cell (iPSC) studies- the state-of-the-art technology of research –provided us insights into human NCC migration in vitro. Diseased iPSCs which carried RET mutations exhibited a retarded migration rate as compared to normal NCCs.28,29 Intriguingly, deletion of RET gene in normal iPSCs can also recapitulates this HSCR-related phenotype. The combination of gene-editing (CRISPR/Cas9) and iPSCs technologies has hence empowered scientists to unravel disease pathways in HSCR and even correct mutations in diseased cells.

Of note, the midgut elongates and enters the extraembryonic cavity, leading to the transient juxtaposition of presumptive ileum and proximal colon in week 6 of human embryonic development. This physiological umbilical herniation is coincident with the migration of vagal NCCs to midgut, and a large subset of NCCs within the dorsal mesentery (which vascularizes the small intestine and proximal colon) can colonize the colon without traversing the cecum, contributing to the colon ENS.24 Despite this transient shortcut, the lateral migration of NCCs to the distal gut warrants a great deal of time due to the concurrent growth of bowel, not to mention the delayed entry of NCC into the foregut among HSCR patients. Therefore, enteric NCCs from HSCR patients cannot harness this transient trans-mesenteric shortcut to colonize gut epithelium effectively, and consequently, the defect in forming ganglionated plexuses will have profound impacts on bowel activities.

Upon NCC colonization, enteric NCC-derived progenitors will either proliferate to sustain migration along the entire length of the developing gut or differentiate into glia and neurons. Several trophic factors and morphogens have been implicated, including (1) GDNF/RET, (2) SRY-box 10 (SOX10) and (3) extracellular matrix (ECM) remodelling. Once the GDNF/RET-GFRα1 pathway is stimulated, RET becomes phosphorylated and its docking site (for example tyrosine 1062 of RET 9 isoform) binds to adaptor proteins so as to upregulate multiple intracellular signalling pathways including phosphatidylinositol 3-kinase (PI3K).27,30,31 PI3K exerts its effect on proliferation by (1) activating downstream AKT signalling to promote local activity of Rho GTPase RAC1 and CDC42, and (2) removing the inhibitory effects of glycogen synthase kinase-β (GSK3β) on neural growth.27 The PI3K/AKT signalling ensures cell survival by antagonizing pro-apoptotic FOXO transcription factors and p53 pathways as well.32 Transcriptional factors which influence the expression of RET (including SOX10) also play key roles in maintaining the potency and survival of NCCs. All the migrating NCCs express SOX10 to remain in a proliferative and undifferentiated state. Haploinsufficiency for SOX10, meanwhile, renders the colonization of gut NCCs ineffective and migration of NCCs from the neural tube is hindered by homozygous SOX10 mutation in mice, leading to colonic aganglionosis and HSCR.33,34

NCCs proliferate at the migrating front so that the migration to uncolonized areas advances. High proliferation speed and cell density are required for the directed translocation of wave front rostral-caudally, as the rate of proliferation in the rapidly growing intestine is faster than that in the stomach.35,36 This proliferation-dependent model could also be linked to the GDNF signalling pathway in that deficiency of GDNF diminished the proliferation and the density of NCCs, resulting in hypoganglionosis.23 Another paradigm, coined as contact inhibition of locomotion, argues that the directional migration is repelled away from the point of cell-cell contact and the repulsion could be mediated by noncanonical Wnt signalling – a planar cell polarity pathway which is abolished in Bardet-Beidel syndrome, an HSCR-associated ciliary gene dysfunction.37,38 However, these models cannot explain all the features in the time-lapse imaging studies. For instance, NCCs migrated in contact with their counterparts near the wave front and these ‘chains’ climbed over each other, demonstrating unpredicted trajectories in their drive to migrate.27,39

Despite these limitations, the models of migration can shed light on the non-cell-autonomous effects of ENS gene mutations to a certain extent. In a study of chimeric mice which contained both the wildtype and EDNRB−/− cells, mixing wildtype NCC cells with >50% EDNRB−/− cells can impair migration of NCCs, and these mice lacked neurons in the distal colon. On the contrary, normal neural development was observed along the entire length of colon in chimeras consisting of >50% wildtype cells.15,40 This compensatory mechanism is dependent on cell-cell interaction, given that the accelerated proliferation rate of normal NCCs can act as a remedy for the defective proliferation among NCC subpopulations whose genes involved in NCC proliferation (such as RET or FOXD3) were abolished.41 The other side of the coin is that aganglionosis and HSCR will be the consequences. When the population of incompetent cells at the wavefront suffices, they can exert the non-cell autonomous effects on the neighbouring normal NCCs to demolish overall proliferative capacity and population size.

Apart from the community effect, the interactions between NCCs and extracellular matrix (ECM) determine cell migration. Over-abundance of ECM components, including collagen type IV, laminin, tenascin, and fibronectin, was correlated with aganglionosis among HSCR patients4,42 and similar findings can be observed in EDNRBflex3/flex3 mice which demonstrated abnormal NCC colonization of gut.43 The loss of ECM-interacting molecules (including β1-integrin) also hampers NCC migration on fibronectin and tenascin-C (which are enriched in hindgut); meanwhile, adhesion molecules N-cadherin and NCAM control the efficiency of NCC migration and otherwise potentiate aganglionosis.44 Recent high coverage whole genome sequencing (WGS) studies further unearthed the ERBB-NRG cascade as an HSCR core pathway and discovered ERBB2-interacting candidates, β4-integrin, among 443 short-segment HSCR (S-HSCR).29 β4-integrin is known to activate focal adhesion kinase (FAK), a prerequisite for mediating RET signalling. Altogether, cell-cell interaction, environmental signals from the gut mesenchyme and the ECM niche govern the behaviours of NCCs.

After proliferation and migration, an individual NCC must differentiate into a neuronal or a glial progenitor. Any premature shift from proliferation to differentiation can result in an inadequate colonization of gut and hence differentiation should be coordinated with the proliferation of enteric progenitors. These precursors express central regulators – paired-like homeobox 2B (PHOX2B), SOX10 and RET to differentiate into neurons (neurogenesis) or glia (gliogenesis). In the developing ENS, neuronal differentiation commences after vagal NCCs colonize the foregut. During migration, SOX10-expressing NCCs encompass bipotent progenitors, fate-restricted neurogenic progenitors and gliogenic precursors. SOX10 induces the expressions of PHOX2B and achaetescute homologue 1 (ASCL1).45 PHOX2B regulates RET expression46, and single-cell RNA sequencing of enteric NCCs revealed that coordinated expression of RET and PHOX2B were correlated with neurogenesis.47 NBPhox and ASH-1, encoded by PHOX2B and ASCL1 respectively, suppress the expression of SOX10 so that neurons will be generated.48 Therefore, haploinsufficiency of PHOX2B was reported in congenital central hypoventilation syndrome which can occur with HSCR in 16% cases.49 In addition, iPSCs carrying RET mutations were differentiated into neurons less effectively.28 Other transcription factors, such as HAND2, also promote neurogenic commitment of bipotent NCCs. Concurrently, glia-committed progenitors should express SOX10 and ERBB3 as identified in hierarchical clustering of NCC transcriptomes.47 RET activity should be downregulated during gliogenesis although it begs the question of how SOX10 and RET coordinate their roles in neural and glial fate acquisition.

Neurogenic progenitors expressing early neuronal markers (Tubb3 and Elavl4) differentiate to different subtypes of enteric neurons, including serotonergic, cholinergic, dopaminergic, peptidergic, nitrinergic and GABAergic neurons, at different developmental stages.27 After exiting the cell cycle, precursors of different subtypes express neurochemical markers under the influence of transcription factors (HAND2 and ASCL1). Deletion of HAND2 prevents the specification of vasoactive intestinal peptide (VIP)-expressing neurons, as well as nitrinergic and calretinin-expressing counterparts.50 ASCL1, on the other hand, is required for the development of serotonin (5-HT) neurons. Early neuronal activity is also attributed to Neurotrophin-3 (NT-3) and BMP4 signalling pathways.51,52 Collectively, neurogenesis and gliogenesis should be finely tuned, and alteration of factors including RET and SOX10 will have profound consequences on the functions of the gut.

Neurons and glia are then aggregated to form two interconnected plexuses of ganglia – the submucosal plexuses in the submucosa and the myenteric plexuses in the muscularis externa. The myenteric plexus arises before the submucosal plexus and a recent study on the spatial organization of ENS clones suggested that the glial clones were detected in mucosa only when sister cells colonized the submucosa, supporting the ‘outside-in’ development of the enteric plexuses.47 Further investigation of ENS clones delineated that bipotent neuroglia (NG) cells, glia-containing and neuron-containing clones were all present in the myenteric plexus while NG clones and glial counterparts contributed to the submucosal plexus and mucosa.47 Therefore, it is postulated that each column of progenitor cell overlays the myenteric plexus and generates descendants (NG and glial progenies) which compose the submucosal plexus. Netrin is involved in this radial migration of NCC. In the developing bowel, gut mesenchyme produces netrin so that migrating NCCs which express the netrin receptor, deleted in colon cancer (DCC), are attracted. Considering that DCC−/− mice had no submucosal plexus,53 the DCC-netrin pathway should presumably facilitate the radial migration of ENS clones to the submucosal plexus. Studies on the sonic hedgehog (SHH) pathway also suggested that the improper colonization of mucosa can be averted when SHH signals in gut epithelium repelled the NCCs.54, 55, 56 Overall, enteric neuronal connectivity is a complex process for which most signals are yet to be discovered. In the context of HSCR, formation of the ganglionated plexus will be in vain unsurprisingly when NCCs can hardly colonize the gut mesenchyme, let alone differentiate into neurons and glia.

The defects in NCC colonization of the gut have always resulted in an absence of ganglionated plexuses (aganglionosis) at the most distal region of the colon, that regulate the activity of the colon, although the length of affected gut varies among HSCR patients. 80% of children with HSCR suffer from aganglionosis which is limited to the rectosigmoid bowel (known as S-HSCR) and 20% of these patients are diagnosed with long-segment HSCR (L-HSCR) when aganglionosis extends proximal to the sigmoid colon. Among those L-HSCR patients, a minority unfortunately have total colonic aganglionosis (i.e., aganglionosis extending to/beyond the ileocecal region).1 Most of the symptoms can be attributed to the absence of propulsive motility along the aganglionic bowel and here we discuss how gut motility is attenuated in aganglionic regions.

The ENS is composed of intrinsic primary afferent neurons (IPANs), motor neurons and interneurons which connect with each other to form the peristatic reflex circuit. The myenteric plexus contains cholinergic neurons mainly, while the submucosal plexus encompasses IPANs, cholinergic and VIP-expressing neurons.27,57 Different neurotransmitters are secreted by different subsets of neurons, such as nitric oxide and VIP from inhibitory motor neurons, acetylcholine (ACh) and substance P from the excitatory motor neurons, and enkephalin from the ascending interneurons. Under normal circumstances, luminal distension activates IPANs which release ACh so that input signals are transmitted via interneurons to the ENS. Excitatory and inhibitory motor neurons are triggered so that interstitial cells of Cajal (ICCs) and smooth muscle respond in stereotypic patterns: circular muscle in front of the food bolus relaxes and the muscle behind it contracts. When the rectum or rectosigmoid becomes distended, the internal anal sphincter relaxes as the reflex arc runs in the myenteric plexus.

Nevertheless, cholinergic neurons, interneurons and the peptidergic inhibitory system are all demolished in the aganglionic segments. The accumulation of acetylcholinesterase (AChE) which consequently cleaves ACh, the neurotransmitter secreted by many enteric neurons is implicated in aganglionosis.58 The absence of cholinergic neurons and interneurons implies that the gut cannot relax after each propulsive contraction. Segmental contraction of the rectum mediated by the cholinergic and peptidergic neurons in the ENS is also disturbed. Rectal dilatation can no longer trigger the relaxation reflex and no defecation occurs, resulting in delayed passage of meconium, severe constipation, and bowel obstruction among HSCR patients.