Gut innervation and enteric nervous system development: a spatial, temporal and molecular tour de force

Proper function of the gastrointestinal (GI) tract relies on many specific cell types at precise locations working together in a tightly coordinated manner. Foremost among these cell types are enteric neurons and glia of the enteric nervous system (ENS), which are organized in two concentric and interconnected nerve plexuses embedded in the gut wall. Within each plexus, enteric neurons and glia are arranged in clusters termed ‘ganglia’, which are linked by interganglionic fiber tracts to form a characteristic mesh-like network. This extensive neural network supplies the full length of the GI tract, running from the esophagus, stomach, small intestine (i.e. duodenum, jejunum and ileum) and caecum (or appendix in humans), through to the end of the large intestine (i.e. colon). During embryonic development, the esophagus, stomach and initial part of the duodenum arises from the primordial foregut, whereas the rest of the small intestine and cecum arises from the midgut, and the colon derives from the midgut and hindgut (Rao and Gershon, 2018; Geesman et al., 2021). Crucially, the network controls muscular activity for the movement of luminal contents through the tract. The ENS is the largest, most complex division of the peripheral nervous system (PNS), comprising millions of nerve cells and glia that enable it to sense and integrate information, and exert appropriate motor outputs (Furness, 2012; Schneider et al., 2019; Fung and Vanden Berghe, 2020; Li et al., 2020). It is crucial that gut innervation develops correctly and, unsurprisingly, this is not a trivial task. In extreme cases, neural progenitors can fail to colonize the entire gut causing aganglionosis, in which the most distal part of the large intestine is left without innervation (Lake and Heuckeroth, 2013). This condition causes the life-threatening Hirschsprung's disease (HSCR), which occurs in about one in 5000 live births (Heanue and Pachnis, 2007; Lake and Heuckeroth, 2013). Nonetheless, even when enteric ganglia do develop, errors in their assembly can still have severe consequences, such as chronic idiopathic intestinal pseudo-obstruction (Gershon, 2009; Rao and Gershon, 2018). Furthermore, subtle developmental faults in circuit wiring can also have significant implications on GI functionality (Sasselli et al., 2013).

Compared with other components of gut innervation, the development of the myenteric plexus (MP) is by far the best-characterized. The MP mainly controls gut motility and is situated between the circular and longitudinal muscle layers of the GI tract. In mice, vagal neural crest cells first enter the foregut at approximately embryonic day (E) 9.5 and migrate caudally whilst proliferating and populating the gut at the level of the prospective MP (Young et al., 1998) (Fig. 1). As enteric neural crest-derived cells (ENCCs) begin to cluster and form ganglia in the myenteric layer, some of the progenitors undergo radial migration towards the gut lumen and give rise to the submucosal plexus (SMP), which is situated closer to the mucosa and is responsible for regulating fluid secretion, nutrient absorption and intestinal blood flow (Uesaka et al., 2015, 2016; Rao and Gershon, 2018) (Fig. 1). However, our understanding of how the SMP arises is relatively limited. How sensory projections from the MP and SMP are directed to the mucosa to sense information about luminal contents also remains largely unexplored. Furthermore, little is known about the molecular pathways that govern these essential developmental processes.

In this Review, we focus on the most recent advances related to enteric and extrinsic neuron development, highlighting current knowledge about how ENS architecture is established in three dimensions. We first briefly touch on how the ENS develops along the length of the gut and how regional differences in ENS wiring might arise. We then shift our focus to ENS development in the radial serosa-mucosa axis, with a particular emphasis on SMP formation. We also discuss recent findings on the extrinsic innervation of the developing gut. We address how neuroimmune interactions or external microbial and nutritional cues may influence ENS development. Finally, we summarize the signaling pathways and key molecules involved in these developmental processes. The Review is largely focused on mouse ENS development, due to the inherent flexibility of murine genetic models, but we do include other models such as zebrafish and chick where relevant.

During the early stages of development, vagal and some sacral neural crest-derived progenitors have to migrate and proliferate extensively to colonize the gut as it rapidly grows and elongates. These progenitors first disseminate throughout the gut, migrating in chains along the rostrocaudal axis (Fig. 1A). The process of gut colonization has been extensively discussed in a number of excellent reviews (Lake and Heuckeroth, 2013; Xu et al., 2014; Hao et al., 2016; Nagy and Goldstein, 2017; Rao and Gershon, 2018). Murine, chick and quail, and zebrafish models are commonly used to study ENS development (Hearn et al., 1998; Wallace et al., 2005; Goldstein and Nagy, 2008; Ganz, 2018; Hao et al., 2020), although zebrafish have a simpler ENS that does not have an SMP and lacks defined ganglia (Wallace et al., 2005; Ganz, 2018).

In mice at E9.5, a small group of vagal ENCCs invades the foregut and starts to migrate rostrocaudally, with ENCCs proliferating and differentiating into neurons and glia behind the migratory wavefront (Fig. 1B). As the gut lengthens, it forms a single bend at ∼E11, such that the midgut is briefly closely apposed to the hindgut. Between E11 and E12, some vagal ENCCs directly cross the mesentery from the midgut to the hindgut, bypassing the caecum (Fig. 1C). These ‘transmesenteric’ vagal ENCCs give rise to at least the distal two-thirds of the hindgut ENS, with the distal half being almost exclusively derived from this population (Nishiyama et al., 2012). The transmesenteric and caecal (vagal) ENCC populations merge to form the ENS in the proximal hindgut. As vagal ENCCs arrive, the hindgut also grows in length. At E13.5, there is a smaller group of sacral ENCCs that enters the hindgut and migrates caudorostrally (Wang et al., 2011). Vagal ENCCs then reach the end of the gut by about E14.5. Around this time, myenteric neurites in the small intestine start to project radially towards the mucosa (Hao et al., 2020). A subpopulation of ENCCs from the MP then migrates inwards to begin forming the SMP from E15.5 onwards (Uesaka et al., 2013) (Fig. 1D). Although electrical activity and spontaneous calcium ion (Ca²⁺) activity is detected in the developing ENS from as early as E11.5 (Hao et al., 2011), neurogenic motor activity is only observed from E18.5 in the duodenum (Roberts et al., 2010) and by P10 in the colon (Roberts et al., 2007). The electrophysiological properties of developing enteric neurons continue to mature postnatally (Foong et al., 2012). The time between the onset of electrical activity and neurally-mediated gut function is likely an important period during which the formation of synaptic connections between different neurons and their various cellular targets occurs (Hao et al., 2013a).

The development of the human ENS occurs through a similar ENCC colonization process, in which ENCCs take a similar migratory route to that in mice and chick embryos between embryonic weeks (EW) 4-7 (Fu et al., 2003; Wallace and Burns, 2005; Obermayr et al., 2013). Subsequently, the development of myenteric ganglia occurs from EW7-EW14 and coincides with the differentiation of the smooth muscle. The SMP develops later than the MP and begins forming from EW8.5-EW14 (Fu et al., 2003, 2004). The human ENS displays electrically-evoked Ca²⁺ responses by EW16, and this timing coincides with the expression of key neurotransmitters and synaptic proteins (McCann et al., 2019).

Recent work by Li and colleagues has demonstrated that there are regional differences in the connectivity of the ENS in the proximal versus distal colon (Li et al., 2019). How such specializations are established is not well understood. One of the factors influencing the axonal projections of early enteric neurons is the directionality of migrating ENCCs (Young et al., 2002). The caecum is a specialized blind-ending pouch structure that develops by budding off the side of the gut tube at the junction between the ileum and colon. In line with its complex form, the pattern of ENCC migration in the caecum differs significantly from that of other intestinal regions (Druckenbrod and Epstein, 2005). Rather than continuing to advance as strands, the ENCC wavefront halts once it reaches the developing caecum at around E11.5. After a pause of ∼8-12 h, isolated ENCCs begin migrating into the caecum and form isolated cell aggregates. These then extend strands to other cell clusters and re-establish an interconnected network before resuming its caudal migration. It is during this time that transmesenteric ENCCs take a shortcut, crossing from the midgut to the hindgut as mentioned earlier (Nishiyama et al., 2012). It remains to be examined whether the distinct migration patterns of these early neuronal progenitors foreshadow differences in the configuration of the enteric circuitry in the caecum and colon.

ENCC migration in the developing gut is influenced by cell number, cell-cell interactions mediated by cell adhesion molecules, and factors secreted by the mesenchyme (Young et al., 2004a,b). It is well established that several signaling pathways are crucially important to establish a fully functional ENS. Two such pathways, the glial cell line-derived neurotrophic factor (GDNF) and its receptor/co-receptor RET/GDNF family receptor alpha-1 (GFRα1), and endothelin-3 (EDN3) and its receptor endothelin-receptor type B (EDNRB), are considered the most dominant in the developmental process (discussed below; Table 1) (Heuckeroth et al., 1998; Young et al., 2001; Gianino et al., 2003; Uesaka et al., 2013; Bondurand et al., 2018). Mutations in either of these pathways converge on an HSCR phenotype that leaves the last segment of the GI tract void of enteric neurons (Puffenberger et al., 1994; Romeo et al., 1994; Schuchardt et al., 1994; Amiel et al., 1996; Shimotake et al., 2001).

Undoubtedly, the RET pathway plays the most crucial role in ENS development. RET, a member of the receptor tyrosine kinase superfamily, is a transmembrane receptor that is present in all ENCCs as they migrate through the gut. The activation of RET is crucial for precursor survival (Heuckeroth et al., 1998; Taraviras et al., 1999), proliferation (Heuckeroth et al., 1998; Gianino et al., 2003), migration (Natarajan et al., 2002), differentiation (Hearn et al., 1998) and neurite growth (Young et al., 2001). Loss of RET function causes intestinal aganglionosis within the entire gut of RET-null mice (Schuchardt et al., 1994; Shimotake et al., 2001), and similar mutations have been described in patients (Amiel et al., 2008). Apart from RET itself, its co-receptor GFRα1 and the ligand GDNF are also necessary to activate the RET pathway during fetal development (see Table 1 for references). Inactivating mutations in this pathway are responsible for the majority of familial and sporadic HSCR cases (Romeo et al., 1994; Angrist et al., 1996). Conversely, overactivation of this pathway, for example by gain-of-function mutations such as RET-C618F, leads to hyperganglionosis (Okamoto et al., 2019). Although the role of the RET signaling pathway is fairly well established, it remains unclear how specific RET gene mutations bring about some – but not other – pathological phenotypes (e.g. multiple endocrine neoplasia). It is also unclear why single mutations are insufficient to result in a phenotype in some circumstances and therefore require additional mutations in other genes (Nakatani et al., 2020). The importance of GDNF is further highlighted in vitro, where the ability of enteric neural progenitors to generate an ENS is enhanced by GDNF (McKeown et al., 2017). A number of transcription factors, including SOX10, RARB, GATA2 and PHOX2B, are all crucial to regulate the expression of RET (Chatterjee et al., 2017, 2019), and as such, also play a crucial role in ENS development. In addition, Sprouty2 (an inhibitor of receptor tyrosine kinase signaling) and the kinesin Kif26A serve as negative regulators of RET signaling (Taketomi et al., 2005; Zhou et al., 2009). Finally, perturbation of Hoxb5 causes Ret haploinsufficiency, impaired NCC migration and hypo/aganglionosis in mice (Lui et al., 2008; Carter et al., 2012; Kam et al., 2014; Kam and Lui, 2015), indicating a key role for HOXB5 in ENCC development.

The EDNRB signaling pathway is another crucial element in the control of ENCC proliferation, differentiation and migration (reviewed by Bondurand et al., 2018) (Table 1). Many known HSCR causative mutations are from members of the EDNRB pathway. EDNRB is a G protein-coupled receptor located on ENCCs that works together with the ligand-processing protease endothelin-converting enzyme (ECE) and the ligand EDN3, which are expressed in the gut mesenchyme during ENS development. EDN3 can also affect ENCCs indirectly by enhancing GDNF function, and it maintains ENCCs in an uncommitted and proliferative state (Stone et al., 1997; Hearn et al., 1998; Lahav et al., 1998; Bondurand et al., 2006; Nagy and Goldstein, 2006). EDN3 also modulates the extracellular matrix (ECM) to influence ENCC migration.

Intracellular c-Jun N-terminal kinase (JNK) and cAMP signaling have been proposed to act downstream of GDNF/RET and EDN3/EDNRB signaling (Hao et al., 2019). Recently, pharmacological agents that stimulate or inhibit intracellular JNK or cAMP signaling have been used to identify potential determinants of the speed and directionality of individual ENCCs in E12.5 mouse gut (Hao et al., 2019). Migration speed is modulated by JNK and cAMP signaling, whereas directionality and adhesion appear to be regulated by cAMP signaling, but not JNK, suggesting that these various migratory properties of ENCCs are differentially regulated. Apart from GDNF/RET and EDN3/EDNRB signaling pathways, a number of other molecules also play a crucial part, or have a refining role, in how the ENS is established (Tables 2 and 3). Notably, Sox10 is a key transcription factor with crucial roles in maintaining the progenitor state of multilineage ENCCs, and in glial development (Bondurand and Sham, 2013). Sox10 expression is maintained in glial cells, whereas it is downregulated as progenitors differentiate into enteric neurons (Hao et al., 2017a).

The microenvironment of the intestinal wall, which is composed of several different ECM components, has been proposed to be another determinant of ENCC migration. A recent study has shown that genes encoding ECM proteins are among the most enriched in mouse intestine at late developmental stages (i.e. at E15.5) (Nishida et al., 2018). Notably, over time, collagen VI (ColVI) expression is upregulated, whereas fibronectin (FN) expression is downregulated. Subsequent functional assays suggest a role for ColVI in inhibiting FN-induced ENCC migration. Another study has shown that both mouse and chick vagal- and sacral-derived ENCCs begin producing collagen type XVIII (Col18) close to the ENCC wavefront early in development and that Col18 is permissive for ENCC migration. In contrast, agrin, which impedes ENCC migration, starts to be expressed at a later stage (i.e. from E10) (Nagy et al., 2018). Collectively, these findings indicate that ENCCs migration is influenced by the microenvironment and ENCCs may be capable of actively shaping this environment through spatiotemporally controlled secretion of specific ECM proteins.

Once a gut region is populated by ENCCs, myenteric neural precursors begin clustering to form ganglia. It is not clear which factors govern this particular organization, nor whether mechanical forces generated by early onset of neural activity play a role. There is indeed some evidence suggesting that gangliogenesis is an activity-dependent process. One study has shown that ganglion formation correlates with intracellular Ca²⁺ transients in the sympathetic nervous system (McKinney and Kulesa, 2011). Similarly, spontaneous propagating Ca²⁺ waves occur between adjacent enteric neural progenitors during embryonic development. These waves involve purinergic signaling, and specific antagonism of purinergic P2 receptors causes defects in ENS network formation (Hao et al., 2017b). Apart from intrinsic activity in the forming ENS, theoretical modeling also suggests that changes in adhesive capacity of the cells may underlie such grouping behavior (Hackett-Jones et al., 2011). The relationship between ganglion formation, their size or the numbers of neurons they contain and their later functionality is completely elusive; several different reports show that both hyper- and hypoganglionosis are associated with ENS functionality defects (Wedel et al., 2002; Breau, 2006; Meier-Ruge et al., 2006; Yin et al., 2006; Hendershot et al., 2007).

As enteric ganglia form during development, progenitors differentiate into glia and diverse subtypes of enteric neurons (Wallace and Burns, 2005), including sensory neurons, various types of interneurons, and excitatory and inhibitory motor neurons (Fung and Vanden Berghe, 2020). Different functional subsets of enteric neurons can be inferred based on their axonal projection patterns and expression of primary transmitters, their synthesizing enzymes, calcium-binding proteins or cytoskeletal proteins (Qu et al., 2008; Mongardi Fantaguzzi et al., 2009). Based on these properties, 16 different neuronal subtypes have been identified in the mature ENS (Hao and Young, 2009). Different neurochemical subtypes of enteric neurons are born (i.e. exit the cell cycle) over specific developmental time points (Fig. 2). For example, serotonin neurons, which are anally-projecting interneurons, are one of the earliest born neuronal subtypes. They have been shown to exit the cell cycle from as early as E8, with a peak time of cell cycle exit at around E10-E11.5 (Pham et al., 1991; Hao and Young, 2009; Bergner et al., 2014). On the other hand, calcitonin gene-related peptide (CGRP) neurons (i.e. putative sensory neurons; Qu et al., 2008) are born later, from around E10-E11.5 (Pham et al., 1991; Bergner et al., 2014). Recent work using human fetal gut samples has shown that the development of several key enteric neuronal subtypes occurs between EW12 and EW14 (McCann et al., 2019). The mechanisms responsible for neuronal differentiation and specification of different neuronal subtypes remain to be elucidated.

Although the pool of progenitors is considered fairly homogenous, the mesh-like pattern of the myenteric ganglia, as well as the size of the ganglia, can vary considerably between gut regions (Furness, 2006). The neurochemical coding of enteric neurons can also differ along the gut: for example, calretinin-immunoreactive neurons in the mouse small intestine are predominantly cholinergic, whereas in the colon, ∼25% of calretinin⁺ neurons are non-cholinergic (Sang and Young, 1998). How regional differences in the proportions of specific neuronal subtypes arise requires further investigation. It is likely that these differences reflect the intrinsic wiring and/or composition of the enteric circuitry in a given region, and that this in turn mirrors the specialized functions of different gut regions, such as storage or emptying (Rao, 2020).

A recent study has compared the transcriptional profiles of all ENS cells (isolated using a Wnt1:Cre line), ENS progenitors (isolated using a Sox10:Cre line) and non-ENS cells of the mouse gut at E11.5 and E15.5, to reveal that a substantial number of novel as well as previously identified transcription and signaling factors vary in spatial and temporal expression in the ENS between these developmental ages (Memic et al., 2018). This comprehensive dataset will undoubtedly provide important clues to understanding the transcription factors and signaling mechanisms that contribute to the development of neuronal and glia diversity (Boesmans et al., 2019).

ENS development continues postnatally, where the proportions of different enteric neuronal subtypes are still being established and their connections are plastic (Hao et al., 2013b). Postnatal development is considered a crucial period for establishing the gut microbiome (Box 1) and shaping gut health, which can have significant implications for later life (Foong, 2016; Hao et al., 2016). For example, neonatal antibiotic administration is associated with increased susceptibility to various diseases, such as inflammatory bowel disease (Ananthakrishnan, 2013). Although not well understood, this is inevitably linked to the immune system (Box 2). Recent work shows that the administration of vancomycin during the neonatal period, as well as during the post weaning period, leads to structural and functional alterations in the enteric circuitry and disrupted colonic motility (Hung et al., 2019, 2020). The transition from a liquid to a solid diet during weaning presents another substantial environmental change that corresponds with a period of ENS maturation, whereby significant changes in synaptic contacts and neurochemistry were observed over this time (Parathan et al., 2020). The extent to which alterations in diet (Box 3) and/or microbiota (Box 1) may influence developmental changes and the involvement of neuroimmune interactions (Box 2) is yet to be unraveled.

It remains contended whether adult neurogenesis occurs in the ENS, the circumstances under which it may occur and the source of such potential progenitors (Joseph et al., 2011; Kulkarni et al., 2017; El-Nachef and Bronner, 2020; McCallum et al., 2020). It has been proposed that enteric neurogenesis is maintained in the mature ENS by enteric glia (Laranjeira et al., 2011; McCallum et al., 2020) or Sox10-expressing multipotent progenitors (Kulkarni et al., 2017), with convincing evidence for the ability of glial cells to serve as a source of new neurons when lesioned areas need to be repopulated (Laranjeira et al., 2011). In addition, Uesaka and colleagues have shown that, in mice, Schwann cell precursors (SCPs) can contribute to adult neurogenesis in the ENS. SCPs likely enter the gut via the extrinsic nerves that act as a scaffold onto which they migrate towards the gut (Uesaka et al., 2015). Recent work in mouse models of HSCR show that SCPs situated within extrinsic nerves play a role in GDNF-induced neurogenesis, providing a source of enteric neurons and glia (Soret et al., 2020). In zebrafish, neural crest-derived cells at the trunk level, likely to be SCPs, can also contribute to postnatal neurogenesis during development and following injury (El-Nachef and Bronner, 2020). Yet another source of enteric neurons, based on pancreatic duodenum homeobox 1 (Pdx1), has been proposed by Brokhman and colleagues. Pdx1 expression is first detected in the fore-midgut junction at ∼E8.5 and later in all the mucosa cells in the duodenum (Offield et al., 1996). Pdx1::Cre recombination experiments suggest that a fraction of enteric neurons is of mesoderm origin derived from Pdx⁺ progenitors that start migrating at E11.5 (Brokhman et al., 2019).

Although the number of additional pathways or molecules revealed in recent years remains limited, some interesting advances have been made in relation to mitogen-activated protein kinase (MAPK) activity, and distal-less homeobox (Dlx)/vasoactive intestinal peptide (VIP) signaling in relation to ENS development. A recent in vitro study has shown that miR-4516, a cell migration suppressor (Chowdhari et al., 2017), directly targets Mapk10 and alters HSCR susceptibility (Wang et al., 2020). Dysfunction of Mapk10, which is highly expressed in the mammalian ENS, also causes migration delay in zebrafish (Heanue et al., 2016). Another recent study links ENS development-related gene Dlx to the expression of the VIP neurotransmitter (Wright et al., 2020). The transcription factors Dlx1 and Dlx2 are expressed during ENS development and, although Dlx1/2^−/− mice show no defects on the topography of the ENS, they do show severe intestinal motility dysfunction at postnatal day (P) 0. RNA sequencing of E14.5/P0 Dlx1/2^+/+ and Dlx1/2^−/− enteric neural progenitors has identified dozens of dysregulated genes, including a downregulation of Vip. The Dlx1/2^−/− ENS shows a significant decrease in VIP-lineage neurons and neuronal Vip expression, which may contribute to gut dysmotility.

As we have discussed, in mice the full length of the gut is colonized by vagal neural crest cells at ∼E14.5. Only after this process is complete does the radial migration of a subpopulation of cells from the myenteric region into the submucosa to form the SMP begin. In the proximal small intestine, ENCCs are observed in the submucosal region from E15.5 (Hao et al., 2020) (Fig. 1D). The entry of SCPs associated with ingrowing extrinsic fibers coincides with the inward migration of submucosal ganglia precursors from the myenteric region. SCPs also contribute to a small population (less than 5%) of submucosal, but not myenteric, neurons and glia in the small intestine (Uesaka et al., 2015, 2016).

As in the small intestine, development of the SMP in the large intestine also follows that of the MP and continues postnatally in mice (McKeown et al., 2001). Exceptionally, in the avian hindgut the SMP develops before the MP (Burns and Douarin, 1998). Unlike the small intestine, in which neural progenitors are predominantly derived from vagal ENCCs, sacral-level ENCCs also contribute to progenitors in the large intestine in mammalian and avian systems (Fig. 1D). In mice, one population of these enters the hindgut between E13.5 and E15.5, following the arrival of the vagal NCCs. From E16.5, a separate population of SCP cells enters the large intestine alongside extrinsic fibers and first settles in the MP. From E18.5, individual sparsely distributed neuronal cell bodies are observed in the presumptive submucosa and the arrangement of these cells into ganglia begins postnatally at ∼P3 (McKeown et al., 2001). Ultimately, SCPs contribute to ∼20% of myenteric and submucosal neurons in the colon (Uesaka et al., 2015, 2016). Hence, a subpopulation of SCPs must also migrate later perinatally, from the MP towards the submucosa. Although SCPs also give rise to enteric glia, it is unclear whether they differentiate into specific morphological-defined subtypes (Boesmans et al., 2015; Uesaka et al., 2016).

In the developing avian hindgut, endothelial cells can be seen to populate the gut in two concentric rings, seemingly delineating the presumptive myenteric and submucosal layers, even before the actual arrival of ENCCs (Goldstein and Nagy, 2008). It has been suggested that the established vasculature serves as a route for ENCC migration and that the interaction between ENCCs and endothelial cells is mediated by β1 integrin signaling (Nagy et al., 2009). However, later studies in embryonic chick and mouse hindgut have indicated that ENCC migration, at least in the rostrocaudal axis, occurs independently of vascularization (Delalande et al., 2014; Hatch and Mukouyama, 2015). Furthermore, the perturbation of intestinal vasculature formation in a Vegfa^120/120 mouse model does not prevent colonization of the full length of the intestine by ENCCs (Delalande et al., 2014). Nonetheless, these mice do show a notable disruption in the organization of ENS architecture at E16.5, suggesting that the vasculature may play a role in forming myenteric ganglia. The potential involvement of vascular cues in the clustering of enteric ganglia and patterning in the radial axis requires further investigation.

A number of signaling pathways are important in SMP development (Uesaka et al., 2016) (Fig. 3). In mice, netrin signaling plays a role in radial migration; netrins are expressed in the mucosal epithelium and act as chemo-attractants. Ablating the netrin receptor deleted in colorectal cancer (DCC) in mice leads to a failure in the formation of the SMP (Jiang et al., 2003). Similarly, GDNF signaling is also required for radial migration, as ablating the GDNF receptors (GFRα1 or RET) once the rostrocaudal colonization of the gut is complete (E15.5) severely disrupts the development of submucosal ganglia (Uesaka et al., 2013). Furthermore, conditional knockdown of RET at E13.5 results in a retention of GFRα1-expressing ENS precursors in the MP and these cells fail to migrate to the submucosal region by E18.5 (Uesaka et al., 2013). On the other hand, sonic hedgehog (SHH) signaling in the epithelium appears to play an important role in demarcating the SMP by restricting inward-migrating ENCCs from advancing beyond the submucosa (Sukegawa et al., 2000). There is some suggestion that this may involve the hedgehog cell surface receptor growth arrest specific 1 (Gas1), expressed on ENCCs (Biau et al., 2013). Indeed, both SHH^−/− and Gas1^−/− mutant mice show ectopic enteric neurons in the mucosa (Ramalho-Santos et al., 2000; Biau et al., 2013). However, it is difficult to delineate whether these ENS abnormalities are direct consequences of the loss of SHH signaling or secondary to the additional defects in the smooth muscle and mesenchyme (Jin et al., 2015). SHH may also inhibit GDNF-induced ENCC migration via the induction of BMP4 expression in the mucosal mesenchyme (Fu et al., 2004, 2006). This BMP4 signal is refined by a BMP antagonist, noggin, which acts to both shield inward-migrating ENCCs in the submucosa from the inhibitory effect and promote GDNF-induced migration in mice (Fu et al., 2006). A recent study has also identified cerebral dopamine neurotrophic factor (CDNF) as a key factor for keeping submucosal neurons viable, because submucosal neurons of CDNF-knockout animals are vulnerable to aging effects more than myenteric neurons (Lindahl et al., 2020). Furthermore, avian studies indicate that SHH can induce the expression of specific ECM proteins in the mesenchyme, such as collagen IX, to impede ENCC migration (Nagy et al., 2016; Nagy and Goldstein, 2017). Thus, development of the ENS in the radial axis requires an intricate coordination of numerous chemoattractive and inhibitory factors.

In the proximal small intestine, the projection of neurites from the MP towards the mucosa begins as early as E13.5, preceding the development of the SMP and even villi (Hao et al., 2020). By E15.5, these neurites in the developing villi are already able to transmit information back to the MP (Hao et al., 2020). This also overlaps with the beginning of the differentiation of different epithelial cell types, including enteroendocrine cells (specialized sensor cells in the mucosa), which occurs from ∼E14-E15 in the proximal small intestine (Desai et al., 2008). Following this, from E16.5, mature ENS synapses can be detected at the ultrastructural level (Vannucchi and Faussone-Pellegrini, 2000). Thus, the prenatal gut may already be equipped to begin detecting luminal contents at this stage. Enteroendocrine cells possess specialized processes termed ‘neuropods’ that form synaptic connections with extrinsic sensory nerves that project into the mucosa and enable the communication of luminal information to the central nervous system (CNS) (Kaelberer et al., 2018). It remains to be determined whether such neuro-epithelial connections between the neurites projecting into the mucosa and the neuropods of enteroendocrine cells are established at this point in development, as well as the mechanisms underlying the formation of this specialized innervation (Bohórquez et al., 2015).

Developing submucosal ganglia precursors, scouting intrinsic axons and extrinsic fibers (see below) appear to follow similar guidance cues as they migrate towards the mucosa, although this is yet to be fully established. It is unclear whether they may interact throughout this process (Hao et al., 2016; Uesaka et al., 2016). For example, netrin/DCC signaling is involved in the radial migration of ENCCs and also attracts vagal sensory afferent fibers towards the submucosa (Jiang et al., 2003; Ratcliffe et al., 2006). However, it remains to be shown whether netrins similarly guide enteric neurites towards the mucosa. In culture, netrin-1 promotes enteric neurite outgrowth, but this effect required the presence of non-crest derived cells of the gut wall, suggesting that co-factors produced from these cells are necessary (Jiang et al., 2003). SHH signaling, via Gas1, is also involved in directing chemorepulsion of myenteric neurites away from the mucosa (Jin et al., 2015) (Fig. 3). Whether the development of enteric mucosal innervation and the migration of ENCCs towards the submucosa are indeed governed by similar chemoattractant and chemorepellent factors needs to be further examined.

The myenteric and submucosal plexuses are interconnected, but little is known about how this connectivity is formed (Rao and Gershon, 2018). Neurites in the MP are seen to begin projecting towards the colonic mucosa at ∼E16.5, before the appearance of submucosal neurons (McKeown et al., 2001). Therefore, connections between myenteric and submucosal neurons must develop at a later stage. In an elegant study by Lasrado and colleagues, various genetic tools have been employed to map the fate of individual Sox10⁺ ENCCs in mouse embryos from E12.5 (Lasrado et al., 2017). Labeled ENCCs gave rise to clones that consisted of only neurons, only glia, or both neurons and glia. Lineally related cells of glial and bipotential neuroglial progenitors are distributed in columns along the radial axis. Furthermore, clonally related myenteric neurons display a greater degree of synchrony in their calcium responses to electrical stimulation, compared with unrelated neurons (Lasrado et al., 2017). It will be interesting to further investigate whether lineally related myenteric and submucosal neurons within a radial column show similar synchronization, as this may provide insights into how the wiring of these different elements is established.

Extrinsic neurons innervate the ENS via projections from the CNS to the gut. They include vagal afferent neurons, spinal afferent neurons, pre-/postganglionic sympathetic neurons and parasympathetic motor neurons (Fig. 4). Vagal afferent cell bodies reside in the nodose or jugular ganglia (i.e. the inferior or superior ganglia of the vagus nerve) and extend their projections centrally to the brainstem, whereas spinal afferent cell bodies reside in dorsal root ganglia and extend their projections to the spinal cord. Both vagal and spinal afferents direct their other projections towards the gut, penetrating and ramifying into the multiple layers of intestinal wall (Brookes et al., 2016). Previous studies have shown that vagal nerve fibers share the same pathway with migrating vagal NCCs in mice, although they move slower than vagal NCCs from the hindbrain to the foregut (Baetge and Gershon, 1989; Anderson et al., 2006b).

Latest work by Niu and colleagues provides an extensive account of the timing and trajectories of the development of extrinsic nerves that supply the murine GI tract (Niu et al., 2020). Briefly, vagal afferent axons project along the esophagus at E10.5. Following a short delay, parasympathetic neurons in the dorsal motor nucleus of the hindbrain project their axons along pioneering vagal afferent fibers and together they enter the stomach by E11.5. Vagal afferent axons then converge with spinal afferent axons at E12.5, while sensory afferent and parasympathetic axons also split into a gastric branch to the stomach and a celiac branch to the intestine. Post-ganglionic sympathetic neurons in the prevertebral ganglia begin projecting towards the gut at E12.5 and divide into a gastric and a celiac trajectory at E14.5. From E14.5 to E16.5, vagal afferent, spinal afferent, post-ganglionic sympathetic and parasympathetic axons project in close association along the gut mesentery and innervate the full length of the gut by E16.5.

Another recent study has documented the development of the sympathetic innervation of the gut in human embryos and fetuses between EW4 and EW9.5 (Kruepunga et al., 2020): NCCs migrate to the para-aortic region and then the pre-aortic region from EW5 to EW6, in which they begin forming the sympathetic trunk and prevertebral ganglia, respectively. Extrinsic nerve fibers then migrate along the intestinal arteries to merge with the intrinsic innervation between EW7 and EW9. By EW9.5, the extrinsic innervation has yet to reach the distal midgut. After arriving in the gut, extrinsic nerves serve as an important route of information exchange between the GI tract and the CNS. Vagal and spinal afferents are responsible for communicating different aspects of sensation. Vagal afferents convey sensations of fullness and satiety, whereas spinal afferents signal high-threshold sensations such as bloating, pain and urgency.

In addition to the netrin/DCC pathway described earlier, a number of other pathways and molecules have been shown to be involved in the development of extrinsic gut innervation (Tables 2 and 3; Fig. 4). For example, Tbx1 is a transcription factor required for the vagal projections into the gut wall and for the development of vagal ganglia (Uesaka et al., 2016). Once vagal fibers reach the gut, Robo/Slit signaling is important for determining which cell types within the GI tract are innervated by vagal nerve fibers (Goldberg et al., 2013) (Fig. 4). Other important molecules that have been identified relate to the survival and maintenance of extrinsic neurons. For vagal afferents, these molecules include brain-derived neurotrophic factor (BDNF) (Jones et al., 1994; Erickson et al., 1996; Murphy and Fox, 2010), neurotrophin-3 (NT-3) (Ernfors et al., 1994; Fariñas et al., 1994; Liebl et al., 1997; Tessarollo et al., 1997), and their receptors tropomyosin receptor kinase B (TrkB) and tropomyosin receptor kinase C (TrkC), respectively. On the other hand, nerve growth factor (NGF) has a crucial role in the survival of small nociceptive sensory neurons, as well as sympathetic neurons (Fariñas, 1999). Espinosa-Medina and colleagues have shown that parasympathetic ganglia derived from SCPs make their way along already established preganglionic fibers (Espinosa-Medina et al., 2014). Further studies are necessary to better characterize the molecules involved.

Despite the striking advances made over the past 30 years in our knowledge of the development and function of the ENS, a number of key aspects remain to be addressed. For example, we still have a limited grasp on the development of innervation in the radial axis – how the SMP arises, how sensory projections from the MP and/or SMP sense information about luminal contents, how these sensory circuits develop and how extrinsic nerves are integrated all remain elusive. Future efforts towards a better understanding of ENS development are multifaceted, ranging from further investigations into gangliogenesis to studying how regional differences in ENS connectivity may be established, as well as microbiota-neuroimmune crosstalk and more extensive characterization of the underlying molecular pathways. Importantly, given that the majority of ENS development studies still heavily rely on murine models owing to the limited accessibility of human fetal gut, efforts towards better understanding the developing human ENS are crucial. There has been substantial progress in this area, with recent work conducted by McCann and colleagues providing key insights into the development of the functional enteric circuitry and the emergence of electrically-evoked neural activity in the human fetal intestine (McCann et al., 2019). In addition, human pluripotent stem cells and induced pluripotent stem cells (iPSCs) have proven to be valuable alternatives for studying human neural development, circumventing the limited availability of human samples (Frith et al., 2018). Such work will be informative for identifying novel approaches to repairing the ENS in disease and also reciprocally feeds back into advancing iPSC technology (Burns et al., 2016; McCann et al., 2017). Identifying the molecules involved in orchestrating ENS development, their concentrations and timing of expression, will facilitate iPSC-based and stem cell-based regenerative therapies for enteric neuropathies, such as HSCR (Hotta et al., 2013; Fattahi et al., 2016; McCann et al., 2017; Frith et al., 2018; Barber et al., 2019). Theoretically, achieving an optimal ENS differentiation protocol for patient-derived iPSCs would be an ideal outcome, which could pave the way to better therapies for many GI neuropathies. However, for any such approach to be effective and efficient, research into iPSC and repair approaches will have to go hand in hand with continuing basic research to refine our knowledge about molecular factors (and their timing) involved in the differentiation of different subtypes of enteric neurons and glia, the generation of the typical concentric organization and interaction with the extrinsic nerves.

We thank the team members of the Laboratory for Enteric NeuroScience (LENS) for their contributions and fruitful discussions.