Intramuscular enteric glia persist in Hirschsprung disease and undergo neurogenesis in response to GDNF-NCAM1 signaling

Abstract

Hirschsprung disease (HSCR) is a neurocristopathy, yet paradoxically, neural crest-derived EGCs are present within the muscle of the affected region. This study investigates the molecular identity, origins, and neurogenic potential of EGCs in the aganglionic mouse and human colon. We utilized single-cell RNA sequencing (scRNA-seq), immunohistochemistry, and in vitro culture of EGCs from aganglionic and ganglionated segments of Ednrb-null mice (Plp1-GFP;Baf-tdT;Ednrb^−/−) and human HSCR tissues. Neurogenic potential and network formation were assessed, and the effects of glial cell line-derived neurotrophic factor (GDNF) on neurogenesis were evaluated. scRNA-seq and immunohistochemistry revealed the absence of GFAP+ intraganglionic (IG) glia in aganglionic colon, while CAMK2b+ extraganglionic (EG) glia and Schwann-like cells (SLCs) were present. EG glia exhibited a transcriptional profile similar to SLCs, suggesting a possible shared embryonic origin. EGCs in the aganglionic segment (comprising EGs and SLCs) exhibited reduced neurogenic potential and network complexity compared to EGCs from the ganglionated region (comprising EGs and IGs). GDNF partially restored neurogenic capacity and enhanced network complexity of EGCs isolated from the aganglionic segment, acting through a non-canonical NCAM1-dependent pathway independent of RET signaling.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-17734-3.

Introduction

While glia were long considered ancillary cells with the sole function of providing structural support to their neighboring neurons, recent work has revealed their vital role in neuronal development and plasticity, modulation of neuronal signaling, and in the pathogenesis of nervous system disease¹. Glia are found in both the central nervous system and the enteric nervous system (ENS). The ENS is the largest subdivision of the autonomic nervous system² and is comprised of 200–600 million neurons³ and up to seven times as many glial cells⁴. These cells are distributed throughout the gut, predominantly arranged into two interconnected networks, the myenteric and submucosal plexuses. Together, the neurons and glia are responsible for controlling numerous gut functions, including digestion and absorption, intestinal transit, secretion of mucus, electrolytes, and signaling molecules, epithelial barrier function, and preservation of a healthy microbiome^1,4,5.

Like neurons of the ENS, enteric glial cells (EGCs) arise from neural crest-derived cells (NCCs) that migrate into and along the gut mesenchyme during embryologic development⁶. These enteric neural crest-derived cells (ENCCs) are bipotential, and ultimately give rise to both neurons and glia. Neural crest cell fate is partially determined by the specific axial level of the neural tube (cranial, cardiac, vagal, trunk, and sacral) from which they arise⁶. The ENS is primarily populated by vagal-derived NCCs, and to a lesser extent sacral NCCs^7,8. Recent studies have also identified Schwann cell precursors (SCPs) as an additional source of enteric neuronal precursors. SCPs are a neural crest-derived stem cell pool found on the peripheral nerves that invade the gut along extrinsic nerve fibers^9,10. After ENCCs reach their final destination, signals from the surrounding microenvironment influence their differentiation into either neurons or glia.

EGCs provide more than structural support to neurons. Work over the last two decades has revealed numerous unique functions of EGCs, including maintenance of epithelial barrier integrity^11,12, interactions with luminal microbes¹³, and regulation of gut transit^1,14–16. Ablation of EGCs is associated with gut dysmotility^14–16, and as shown in some studies, inflammation¹⁷. Deficits and alterations in EGCs are associated with chronic post-operative dysmotility and/or enterocolitis in patients with Hirschsprung disease¹⁸, idiopathic slow-transit constipation¹⁹, and post-operative ileus²⁰. EGCs also have been shown to help restore homeostasis via neurogenesis^21,22. In mouse models, there is increasing evidence that postnatal enteric neurogenesis occurs in response to injury²³, and EGCs represent a source of these neuronal progenitors²².

Glia have been classified into subtypes largely based on morphology and location. EGCs are widely distributed throughout the different layers of the gut, including within and outside the myenteric and submucosal plexuses. Pioneering work on guinea pigs proposed a classification scheme for EGCs based on morphology and location within the plexus²⁴. This was expanded upon more recently to include subsets of EGCs that are located outside of the plexus²⁵. Four types of enteric glia have been summarized and reviewed²⁶. Type I are the protoplasmic star-shaped glia, located within both myenteric (Type I_MP) and submucosal (Type I_SMP) plexuses. Type II are “fibrous” appearing cells with long processes running along interganglionic connectives. Type III lie at the level of the myenteric and submucosal plexuses, but are located outside the ganglia and interganglionic connectives. Type IV are located within the circular and longitudinal muscle along nerve fibers. The different morphologies and locations have led many to hypothesize that specific subtypes of EGCs have unique functions. EGCs have a distinct set of markers, including the transcription factors Sox10²⁷ and Sox2²², the myelin associated protein, proteolipid protein 1 (Plp1)²⁸, and the calcium binding protein, S100B²⁹.

ENS development depends on successful and well-timed neural crest cell proliferation, migration, differentiation, and survival⁵. Disruption of these processes can lead to incomplete neural crest-derived gut colonization and a resulting variable length of intestinal aganglionosis, the hallmark of Hirschsprung disease (HSCR) in humans. Since HSCR is due to a failure of neural crest-derived cells to fully populate the entire length of the gut, one would expect that the resulting aganglionic segment would lack both neurons and glia, since both are neural crest derived. However, it has been shown that glial cells, which reside along the hypertrophied extrinsic nerve fibers as Schwann-like cells, are present in the aganglionic gut region of HSCR^30–32. Further, a population of EGCs has been described as inhabiting the intramuscular space in HSCR²⁶. The glia along extrinsic fibers are thought to be Schwann-like cells and therefore derived from SCPs. However, the origin of the extraganglionic glia residing within the gut but separate from the extrinsic nerve fibers is unknown, and their presence in HSCR is paradoxical. We therefore sought to better resolve, describe, and further characterize this population of EGCs using single-cell transcriptomics, immunohistochemistry, and in vitro assays.

Results

Two distinct subpopulations of EGCs remain present in the aganglionic segment of Ednrb KO mice

The longitudinal muscle/myenteric plexus layer was dissected from 10–14 day-old Plp1-GFP;Ednrb^−/− mice and their corresponding wild-type (Plp1-GFP;Ednrb^+/+) littermates. All enteric glia in these mice express GFP, allowing clear delineation of this cell type in the aganglionic segment under fluorescent microscopy. The distal aganglionic segment was isolated from KO mice (knock-out distal (KOD)) as well as the corresponding segment from WT mice (wild-type distal (WTD)). The transition zone was excluded, and the proximal ganglionic segments were isolated (knock-out proximal (KOP) and wild-type proximal (WTP); Fig. 1A). To resolve all cell populations, single-cell RNA sequencing (scRNA-seq) was performed on both the proximal and distal segments from KO and WT mice and the proportion of glial cells and neurons in each segment were evaluated (Fig. 1B). In the combined dataset, we observed distinct clusters of fibroblasts, smooth muscle cells (SMCs), glia, neurons, inflammatory cells (including macrophages, lymphocytes, and dendritic cells), and others (Fig. 1C). Notably, there were three populations of glia that clustered together. Based on gene signatures, we identified these as extraganglionic (EG) glia (Glia 1, cluster 2), intraganglionic (IG) glia (Glia 2, cluster 6), and Schwann-like cells (SLCs) (cluster 9). All three populations had high expression of known glial markers, including Plp1 and Sox10. The IG glia had higher expression of Gfap and Slc18a2, markers that have previously been identified as IG EGC markers^33,34, and the SLCs expressed myelin related genes including Mpz and Mbp. After stratifying the combined datasets based on location and disease (Fig. 1D), the KO distal segment was noted to have not only an absence of neurons (cluster 15), but also absence of IG glia (cluster 6). Interestingly, EGs and SLCs were still present, and the proportion of SLCs was increased (Fig. 1B). To validate this finding, we performed immunohistochemistry on similarly aged Plp1-GFP;Ednrb^−/− and Plp1-GFP;Ednrb^+/+ mice and evaluated both PLP1 and SOX10 expression (Fig. 1E). Consistent with the single-cell data, IG glia were absent in the KO mice (Fig. 1E, top row), while prominent and hypertrophied extrinsic nerve trunks were observed in the aganglionic segment, with SLCs present along these fibers (Fig. 1E, top row).

Fig. 1.

Two distinct subpopulations of EGCs remain present in the aganglionic segment of Ednrb KO mice. (A) Schematic representation of mouse tissue utilized in this study. (B) Pie chart representation of glial cells and neurons obtained from segments of wild-type proximal and distal and knock-out proximal and distal colon. (C) Clustering of cell populations reveals three populations of EGCs. (D) Intraganglionic (IG) glia and neurons are absent in the KO distal colon (gray), but populations of extraganglionic (EG) glia and Schwann-like cells (SLCs) remain (blue). (E) Immunohistochemistry of wholemount preparations of Edrnb KO mice reveals EGCs with co-expression of Plp1-GFP fluorescence and Sox10+ expression outside the myenteric ganglia (white arrows), confirming the presence of EG glia.

The missing EGC population in aganglionic colon corresponds to GFAP+ intraganglionic EGCs, while CAMK2b+ intramuscular EGCs and Dhh+ Schwann-like cells remain

To better resolve IG and EG glial populations, we sought to identify unique markers for each population for both gene and protein expression. We compared the gene expression of our clusters to the gene expression of the categories outlined by Windster et al. and found many similarities (Fig. 2A)³⁵. Our IG glia cluster exhibited high expression of Rlbp1 and Frzb, aligning with the Windster glia 1 category corresponding to canonical type I intraganglionic glia. Furthermore, our SLC cluster aligned closely with Windster’s Schwann-like 3 and 4 markers, expressing Reln, Mbp and Pllp. The EG cluster exhibited some transcriptional overlap with their Schwann-like 1 category expressing higher levels of Apod than other glia. Gfap, which has been shown to be predominantly a marker of IG glia³⁴, was validated in our dataset by both gene (Fig. 2A) and protein (Fig. 2B) expression. Camk2b, calcium/calmodulin dependent protein kinase II beta, is highly expressed by neurons and EG glia, but not IG glia or SLCs (Fig. 2A). Immunohistochemistry of Plp1-GFP;Ednrb^−/− tissues revealed co-expression of CAMK2B and PLP1, validating that CAMK2B is expressed by EG glia (Fig. 2C, top row). Immunohistochemistry of tissues from the corresponding WT mice showed CAMK2B expression within the ganglia but separate from the PLP1+ cells, signifying neuronal expression (Fig. 2C, bottom row). Finally, the marker MPZ was expressed along nerve fiber bundles, consistent with SLC expression (Fig. 2D).

Fig. 2.

The missing EGC population in aganglionic colon corresponds to GFAP+ intraganglionic EGCs, while CAMK2b+ intramuscular EGCs and Dhh+ Schwann-like cells remain. (A) scRNA-seq data reveals high Gfap expression amongst IGs and Camk2b expression by EGs (glia IM). Heatmap compares transcriptional markers of human glial and schwann-like cells categorized by Windster et al. to our cluster analysis. (B) Immunohistochemistry confirms high GFAP expression among IG glia in the ganglionic colon and (C) high CAMK2B expression amongst EG in the aganglionic colon and neurons within the ganglionic colon. (D) The Schwann-like cell (SLC) marker MPZ has high expression along nerve fiber bundles. (E) Correlation matrix of RNA gene expression reveals that EG glia share 99% of gene expression with SLCs, suggesting a possible Schwann cell precursor (SCP) origin of EG glia. (F) Wholemount staining of another Hirschsprung mouse model (Gfra1^flox-EGFP) with a Dhh::Cre reporter show a population of GFP+/Sox10+ EGs (indicating prior Dhh+ expression), further supporting a possible SCP origin. (G) Violin plot demonstrating that only the SLC population has ongoing Dhh expression.

To identify the origin of the persistent EG glia in Ednrb^−/− mice, we performed a correlation matrix (Fig. 2E) and found that EG glia are most highly correlated with SLCs, sharing 99% gene expression. To validate this further, we utilized another Hirschsprung mouse model (Gfra1^flox-EGFP) with a Dhh::Cre reporter¹⁰. In this model, SCPs, which express Dhh, and their progeny retain GFP expression¹⁰. Staining aganglionic colon from these mice with the glial marker SOX10 and neuronal marker PGP9.5 showed GFP+/Sox10+ EG glia (Fig. 2F, arrows). We assessed Dhh expression in our scRNA-seq database and found Dhh exclusively expressed by SLCs (Fig. 2G), supporting the SCP-derived origin of EG glia and suggesting that EG glia lose their Dhh expression over time.

We next sought to validate the presence of these glial subtypes in mouse and human colon. Cross-sections of gut tissues from Plp1-GFP;Ednrb^−/− mice (Fig. 3A) showed absence of TUJ+ enteric ganglia, as expected for aganglionic bowel, with PLP1+ cells (white arrows) present within the muscularis propria, consistent with EG glia. Plp1-GFP;Ednrb^+/+ mice (Fig. 3B) exhibited TUJ+/PLP1+ ganglia with a similar population of PLP1+ cells (white arrows) within the muscle layer. Cross-section of human ganglionated colon from a patient with Hirschsprung disease (Fig. 3C, white arrows) showed similar S100B+ glia within the smooth muscle and outside of the myenteric plexus, consistent with EG glia. High-power images (Fig. 3C, right panels) demonstrated that EG glia were consistently adjacent to TUJ+ nerve fibers. In human aganglionic colon from the same patient, S100B+ cells were observed in the hypertrophic nerve bundles as well as in individual cells closely apposed to TUJ+ nerve fibers (Fig. 3D), consistent with the presence of EG enteric glia in Hirschsprung disease.

Fig. 3.

A population of intramuscular EGCs persists in mouse and human Hirschsprung tissue. (A) Cross-sections of tissues from Plp1-GFP;Ednrb^−/− mice reveal an absence of TUJ+ enteric ganglia, with PLP1+ cells (white arrows) within the muscularis propria, consistent with EG glia. (B) Wild-type Plp1-GFP;Ednrb^+/+ mice exhibit TUJ+/PLP1+ ganglia, with a similar population of PLP1+ cells (white arrows) within the muscle layer. (C) Cross-section of human ganglionated colon reveals similar S100B+ glia outside of the myenteric plexus within the smooth muscle, consistent with EG glia. High-power images (right panels) demonstrated that EG glia were consistently adjacent to TUJ+ nerve fibers. (D) Aganglionic colon from the same patient shows a population of S100B+ cells separate from the hypertrophic nerve bundles, consistent with EG glia.

EGCs isolated from aganglionic colon are less neurogenic than their ganglionated counterparts

To compare the properties of EGCs from aganglionic and ganglionated colon, cells were isolated from the distal and proximal segments of Plp1-GFP;Baf-tdT;Ednrb^−/−dual EGC (GFP) and enteric neuron (tdT) reporter mice as previously described^36,37, carrying the homozygous Ednrb mutation (Fig. 4A,B). Intestinal segments were visually inspected to identify aganglionic versus ganglionated regions, with BAF53b-tdT+ neuronal soma observed in the ganglionated region (Fig. 4A, higher power 4A′) and BAF53b-tdT+ hypertrophic nerve fiber bundles in the aganglionic segment (Fig. 4B, higher power 4B′). Tissues were dissected to isolate the muscularis propria from aganglionic and ganglionated regions, and then enzymatically digested and filtered through a 10 µm filter to obtain cell suspensions for subsequent culture (Fig. 4C). This technique results in a suspension that is predominantly EGs, but may contain some IGs, SLCs, and a fraction of neurons. Importantly, the cell suspension from the distal aganglionic colon is devoid of IGs and neurons given their absence in Hirschsprung disease. After the initial cell seeding (Fig. 4D), cell cultures were monitored over time (Fig. 4E–G). By day 9, all cultures formed enteric neurospheres, suggesting the presence of neuronal progenitors in each of the EGC populations.

Fig. 4.

EGCs isolated from both ganglionic and aganglionic colon form neurospheres in culture. (A–A’) Tissue from the proximal colon of Plp1-GFP;Baf-tdT;Ednrb^−/− dual EGC (GFP) and enteric neuron (tdT) reporter mice shows BAF53b-tdT+ neuronal soma in the ganglionated region and (B–B’) Baf53b-tdT+ hypertrophic nerve fiber bundles in the aganglionic region. (C) Schematic representation of methods utilized to generate cell suspension cultures. (D) Cells were seeded in neuroproliferation media and monitored over time, (E–G) revealing that by day 9 all cultures had formed enteric neurospheres.

Neurospheres generated from ganglionated and aganglionic segments were cultured on fibronectin and their neurogenic potential quantified. After 1 week, cells were fixed and stained for Hu (Fig. 5A,B). The enteric neuron (EN) to EGC ratio was over threefold lower in cultures derived from the aganglionic segment compared to those from the ganglionated proximal region (0.06 vs 0.21, p = 0.03) (Fig. 5C). In vitro network formation was assessed using BAF53b-tdT fluorescence as this marks terminally differentiated ENs derived from the distal and proximal colon (Fig. 5D,E). In cultures from proximal colon, EN networks contained more neurite branches (112 vs 45, p = 0.02) and network junctions (56 vs 19, p = 0.01), including triple (29 vs 2, p = 0.002) and quadruple (4 vs 0, p = 0.002) junction points, indicating a more complex network than in cultures from the distal colon (Fig. 5F–I). The sum of neurite branches (4136 vs 1934 µm, p = 0.02) and maximum branch length (424.2 vs 174.2 µm, p = 0.003) were greater in proximal colon, with no difference in average branch length or maximum direct walk-path across the network (Fig. 5J–M). These findings suggest that EGCs present in the aganglionic colon (EGs and SLCs) possess neurogenic potential, but at significantly lower levels than that of EGCs in the normoganglionic region (containing EGs, SLCs, and IGs).

Fig. 5.

EGCs isolated from aganglionic colon are less neurogenic than their ganglionated counterparts. (a) Neurospheres generated from cell suspensions from the aganglionic (n = 7) and (B) ganglionic colon (n = 21) were cultured on fibronectin for 1 week and then fixed and stained for Hu. (C) Quantification of neurons revealed a higher ratio of enteric neurons (ENs) to EGCs in the proximal colon Mann–Whitney test, *p < 0.05. (D) BAF53b-tdT fluorescence marks terminally differentiated ENs revealing network formation in the distal and (E) proximal colon. (F) EN networks from the proximal colon contain more neurite branches, (G) network junctions, (H) triple junction points, and (I) quadruple junction points than those derived from the distal aganglionic colon. Mann Whitney test, *p < 0.05, **p < 0.01, ***p < 0.001. (J) Networks from the proximal colon also revealed higher total sum of neurite branches and (K) maximum length branches than networks from the distal colon, with no differences observed in (L) average branch length or (M) maximum direct walk-path. Mann Whitney test, *p < 0.05, **p < 0.01, ***p < 0.001.

GDNF promotes neurogenesis and network formation in EGCs from the aganglionic colon via non-canonical NCAM1-dependent signaling

The results above indicate that progenitors isolated from the distal aganglionic colonic region have a substantially reduced ability to generate ENs and form neural networks even in favorable culture conditions. Interestingly, previous studies have indicated that GDNF promotes neurogenesis in the aganglionic gut³⁰. Our scRNA-seq analysis did not identify expression of the GDNF receptor, Ret in any of the glial cell populations, although they did express Ncam1 (Figure S1), a known alternate receptor for GDNF³⁸. Therefore, we examined the neurogenic effect of GDNF on EGCs derived from the aganglionic colon (EGs and SLCs) and tested whether NCAM1 signaling is required. Quantification of the ENs and EGCs indicated that the addition of GDNF to cell cultures doubled the EN to EGC ratio in the distal aganglionic colon (0.12 vs 0.06, p = 0.04) (Fig. 6A–D). These effects were reversed by the addition of an NCAM1 blocking antibody in combination with GDNF (0.07, p = 0.04 when compared to GDNF alone) (Fig. 6A–D), supporting the idea that GDNF promotes neurogenesis in EG glia and SLCs via NCAM1 signaling. Further, the addition of GDNF induced prominent network forming effects, (Fig. 6E–G) including increased neurite branching (1121 vs 70, p = 0.02), network junctions (496 vs 28, p = 0.02) and complex triple/quadruple junction points (triple 132 vs 6, p = 0.004; quadruple 26 vs 1, p = 0.006) in aganglionic colon EGCs, all of which were reversed in the presence of anti-NCAM1 (Fig. 6H–K). GDNF also increased the sum of neurite branches (18,121 vs 2584, p = 0.04), decreased average neurite length (25 vs 36 µm, p = 0.05), and produced networks with longer maximum direct walk-paths (1593 vs 410 µm, p = 0.05), the latter being reversed by anti-NCAM1 (Fig. 6L–O). These results indicate that in addition to promoting neurogenesis, GDNF, acting via NCAM1, also supports the formation of neural networks with larger coverage and finer network connections in enteric neurons derived from a subpopulation of EGCs present in the aganglionic bowel.

Fig. 6.

GDNF promotes neurogenesis and network formation in EGCs from the aganglionic colon via non-canonical NCAM1-dependent signaling. (A) Neurospheres generated from cell suspensions from the aganglionic distal colon were cultured on fibronectin for 1 week and fixed and stained for Hu in media alone (n = 7), (B) media supplemented with GDNF (n = 5), or (C) media supplemented with GDNF and an anti-NCAM1 antibody (n = 7). (D) Treatment with GDNF doubled the ratio of ENs to EGCs, which was reversed by the addition of an NCAM1 blocking antibody. Welch ANOVA, *p < 0.05. (E) EN network formation was assessed using BAF53b-tdT fluorescence in the control, (F) GDNF, and (G) GDNF + anti-NCAM1 treatment groups. (H) Treatment with GDNF increased the cells’ ability to form neurite branches, (I) network junctions, (J) triple junction points, and (K) quadruple junction points, effects that were again reversed by blocking NCAM1 signaling. Welch ANOVA, *p < 0.05, **p < 0.01, ***p < 0.001. (L) GDNF treatment also increased the sum of total neurite branches, (M) decreased average branch length, and (N) increased length of maximum direct-walk paths, (O) with no effect on maximum branch length. Blocking NCAM1 reversed the effect on maximum direct-walk path. Welch ANOVA, *p < 0.05, **p < 0.01, ***p < 0.001.

Discussion

HSCR is, by definition, intestinal aganglionosis. Since a ganglion represents a collection of neurons, the disease is often considered primarily an intestinal neuropathy, with a lack of intrinsic neuronal innervation leading to associated sequelae. Interestingly, we found that while all enteric neurons are absent in HSCR, the same is not true for the enteric glia. Rather, the aganglionic colon lacks a distinct subset of EGCs, and these are the IG glia, a subpopulation we have previously shown to be highly neurogenic^34,36. This absence of the IG subset may therefore be an underlying contributor to the failure of persisting EGCs to replenish the absent neurons in HSCR. While other EGC populations that reside in the aganglionic region (EGs and SLCs) retain some neurogenic potential, their capacity is significantly limited compared to IG glia. Importantly, our data reveal that neurogenesis in these residual EGCs can be enhanced by glial cell-derived neurotrophic factor (GDNF) acting through a non-canonical GDNF-NCAM1 signaling pathway, as EGCs express Ncam1 but not Ret.

The heterogeneity of EGCs has gained increasing attention recently, particularly through advances in single-cell RNA sequencing (scRNA-seq)^34–36. Traditionally, EGCs have been classified based on their anatomical location within the intestine and cellular morphology with four main types of EGC described²⁵. Type I glia are the canonical intraganglionic glia, type II run along interganglionic connectives, type III lie at the level of the myenteric and submucosal plexuses but reside outside the ganglia and interganglionic connectives, and type IV reside within the smooth muscle.

Dependent on species and current knowledge, the ENS may arise from one (lamprey)³⁹, two (zebrafish⁴⁰ and avian⁴¹) or three (mice^8,10,42) major sources of progenitor cells: the vagal neural crest, later-arriving Schwann cell precursors (SCPs), and sacral neural crest cells. SCPs contribute to ENS development and constitute approximately 20% of enteric neurons in the mouse colon^9,10. Recently, two broad subpopulations of EGC in the human intestine have been defined using scRNA-seq by Windster et al., canonical EGCs and a larger population of Schwann-like EGCs, termed due to their expression of markers associated with Schwann cells and the known contribution of SCPs to the colon³⁵.

In their classification system, a distinct transcriptional EGC subset was confirmed to align closely with the classical type I glial morphology via spatial validation of RLBP1 expression, Notably, this population also shares transcriptionally overlap with our GFAP+ IG subset in the mouse. The expression of GFAP as a marker of type I EGCs was previously shown by Rao et al.²⁸. Similar results were later confirmed by RNAscope after we observed high levels of GFAP in a transcriptionally unique subset of EGCs, with elevated levels seen in the type I EGCs compared to the type III and IV EGs³⁴. Windster et al. also identified six categories of Schwann-like enteric glia based on unsupervised clustering. They indicate that their Schwann-like 1 category aligns with the conventional type IV intramuscular glia. Comparing to our dataset, Apod is highly expressed by their Schwann-like 1 cells and our EGs, which may suggest some population overlap. Their Schwann-like 2 cluster had an undefined anatomical locality. Schwann-like 3 and 4 cells highly expressed markers including Reln, Mbp, and Pllp, and have significant transcriptional similarity to our population of SLCs in the mouse. Windster et al. suggested this population corresponds to type III glia. In their spatial validation of RELN expression, this population localized to mesenteric nerve fiber bundles and the hypertrophic extrinsic nerve fibers observed in Hirschsprung disease. Importantly, the classical taxonomy defined by anatomic location and morphology is based on the developmentally normal ENS, however, such prominent extrinsic nerve bundles are not abundant in the healthy ganglionated gut. Moreover, type III and IV glia are traditionally associated with the thinner unbundled extraganglionic nerve fibers within the plexus layer and between smooth muscle cells. In our imaging of the aganglionic segment, we observe singular glia associated with these smaller nerve fibers and other glia within thickened hypertrophic nerve bundles. Given the high proportions of SLCs in the aganglionic segment corresponding to the presence of hypertrophic nerves, and their low abundance in the healthy ENS lacking them, these data suggest that SLCs constitute a transcriptionally distinct glial population which warrants its own morphological taxonomy. Schwann-like 5 and 6 glial cells are located within the mucosa and therefore not evaluated in this study. While scRNA-seq has facilitated the identification of transcriptional markers for EGC subpopulations, there is still uncertainty amongst categorization and type, and definitive markers with robust empirical validation are lacking. Overcoming this gap will aid in deeper investigations into the distinct biological roles and functional attributes of these glial subsets in both physiological and pathological states.

In our study, we report three broad groups of EGCs in the muscularis propria, namely the IGs, EGs and SLCs. GFAP was predominately observed in the type I IGs, whereas CAMK2B was observed in type IV intramuscular bipolar cells, and MPZ in cells originating from extrinsic fibers, which we therefore consider to be Schwann-like cells. While we were unable to identify markers of type II EGCs in the interganglionic connectives, as Rao demonstrated that GFAP can be expressed by type II and type I IGs, these populations may exhibit a degree of uniformity on the transcriptional level²⁸. Correlation analysis was conducted for transcriptional similarities between the IG, EG and SLC clusters, revealing that EGs and SLCs share more similarities to each other than the IGs, consistent with the observations of Windster et al.³⁵. This is suggestive of a potential Schwann cell origin of extraganglionic glia, however further studies are necessary to confirm their origins, including the possibility of a mixed vagal enteric neural crest and Schwann cell contribution, especially in the healthy gut.

A prominent area of interest in EGC biology includes their unique capacity to transdifferentiate into enteric neurons, a process with significant implications for understanding homeostasis of the ENS and for developing novel therapeutic approaches for enteric neuropathies. Several studies have estabdemonstrated that GDNF promoteslished that EGCs serve as progenitors for “newborn” enteric neurons in the postnatal intestine. While it is widely accepted that EGCs have neurogenic potential, the baseline rates of neurogenesis are notably low under homeostatic conditions^37,43. However, inflammation has been shown to enhance neurogenesis in EGCs, presumably as part of a regenerative response to restore normative functions to the intestine^22,44,45. Additionally, high rates of neurogenesis were observed in EGCs derived from postnatal intestine when cultured under specific in vitro conditions^34,36,43. Notably, our recent findings indicate that IG EGCs exhibit intrinsic properties that render them more responsive to neurogenic induction compared to EG EGCs. This increased neurogenic potential in IG EGCs aligns with their retention of open chromatin at loci associated with neuronal gene expression³⁴, highlighting that neurogenesis is not a uniform EGC function but a more distinct characteristic of the IG subpopulation. The findings presented in this study corroborate previous observations^34,36 that intraganglionic EGCs exhibit superior neurogenic capacity than extraganglionic glial and Schwann cell populations. This is evident in our in vitro cultures where EGCs isolated from the proximal colon, containing intraganglionic EGCs, produced cultures with a higher neuron-to-EGC ratio and more complex, extensive neural networks, emphasizing the pivotal role of IG EGCs in neurogenesis and their unique contribution to ENS maintenance and repair.

Notably, our mouse studies and findings from HSCR samples reveal that the absent EGC population corresponds to IG glia. Given the robust neurogenic capacity of this subset, this underscores how a lack of post-developmental regeneration could contribute to the disease. Specifically, in HSCR, the role of IG EGCs in formation and maintenance of the ganglia may represent a unique function in an EGC subset that cannot be compensated for by other EGC subpopulations. As distal intestinal aganglionosis has historically been attributed to impaired migration of vagal neural crest cells in mice, our findings support the idea that the residual EG and SLC populations in Ednrb mutants predominantly arise from SCPs. Likewise, the absence of the highly neurogenic IGs in Ednrb mutants suggest that this subpopulation in particular is derived specifically from vagal neural crest cells. However, it remains unclear whether progenitor cells from separate origins generate EGCs with distinct functional and transcriptional properties or whether these properties are acquired. Intriguingly, there is some evidence of postnatal neurogenesis in HSCR, particularly within the transition zone. In this region, SCPs may enhance their neurogenic potential in response to the reduced pool of vagal-derived neural crest cells⁹. While SCPs are capable of neurogenesis, especially under the influence of GDNF, the lack of significant neuronal regeneration in HSCR raises important questions. The interplay between environmental factors, genetic mutations, and the origin of progenitor cells likely determines their neurogenic and network-forming potential within the ENS. Elucidating these mechanisms is essential for understanding the pathogenesis of aganglionosis in HSCR and may offer new insights into regenerative strategies for treating the condition.

The lack of neurogenesis in HSCR has prompted efforts to develop therapeutic strategies aimed at restoring neuronal populations and ganglia formation. Soret et al. demonstrated that GDNF promotes neurogenesis in EGCs from the aganglionic distal colon, leading to restoration of neuronal numbers in mouse models of HSCR³⁰. We found that that EGCs in HSCR do not express the RET receptor but instead express NCAM1, which mediates the neurogenic effects of GDNF. This is of particular importance as RET mutations are the most common cause of HSCR. The ability of GDNF to bypass RET signaling and still promote neurogenesis holds therapeutic promise for HSCR. Targeting the NCAM1-dependent pathway may offer a potential avenue for restoring neurogenic potential in the aganglionic colon. In addition to pharmacological interventions, cell therapies derived from isolated glia including EGCs and Schwann cells have shown promise in preclinical models of HSCR^37,46–48. Optimizing the neurogenic potential of these cells, particularly through targeted activation of NCAM1 signaling, could enhance the efficacy of such approaches.

Glial cells do more than merely support the neurons, actively participating in neurogenesis and network formation during normal ENS development. The absence of an important neurogenic glial population in HSCR suggests that glial cells might play a more central role in the pathogenesis of HSCR than previously thought, where the remaining EGC subtypes with decreased neurogenic potential are unable to rescue the aganglionosis. While the origin of the remaining EGCs in HSCR is linked to SCPs, further exploration of the molecular environment in the aganglionic region could provide insights into how neurogenesis is regulated in the remaining EGCs. Understanding the molecular cues that guide these glial cells toward neurogenesis could offer new avenues for intervention. Additionally, genetic variation in glial progenitor cells, as well as how these interact with the mutations commonly seen in HSCR, warrants further exploration.

Methods

Single-cell analysis

Two-week-old Plp1-GFP;Ednrb^−/− mice and their wild-type littermates were euthanized and their colon was removed from cecum to rectum. The longitudinal muscle-myenteric plexus (LMMP) layer was dissected from underlying tissue in PBS. Using a fluorescent microscope, the tissue was segmented into aganglionic and ganglionic segments, and the transition zone was excluded. Segments were labeled knock-out distal (KOD), knock-out proximal (KOP), wild-type distal (WTD), and wild-type proximal (WTP). The tissue was digested for 60 min at 37 °C in dispase (250 μg/ml; STEMCELL Technologies, Vancouver, BC) and collagenase XI (1 mg/ml; Sigma-Aldrich, St. Louis, MO). Samples were filtered through a 40 μm filter, stained with DAPI (Invitrogen), and underwent FACS sorting for DAPI negativity to obtain live cells as described below. Cells were manually counted with Trypan blue to assess number and viability. 10X Genetics (Pleasanton, CA) v3.1 kits, along with a 10X Chromium Controller located within our laboratory, were utilized to generate gel bead emulsions (GEMs). cDNA libraries were then prepared according to the manufacturer’s protocol.

All sequencing was performed at the Harvard University Bauer Core Facility, where libraries were sequenced on Illumina NextSeq instruments. Genome alignment and feature-barcode matrix generation was performed with the 10X Genetics Cell Ranger software pipeline⁴⁹ using the “cellranger count” command on the Mass General Brigham ERISOne Research Computing Cluster. Sequencing data can be accessed at https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE293138 (GEO accession GSE293138, token: gbudoeoenjufzml). Subsequent analysis was performed in the R environment (RStudio Version 2023.03.0+386) with Seurat. To achieve high-quality data, datasets were thresholded, removing cells with low numbers of unique molecular identifiers (UMIs), cells with low number of genes, and cells with greater than 10% mitochondrial genes. This filtration removes low quality or dying cells, empty droplets, and doublets or multiplets from the dataset.

Different segments were integrated using the “FindIntegrationAnchors” command in Seurat⁵⁰. After integration, principle component analysis (PCA) was performed. Neighbors were identified and uniform manifold approximation and projection (UMAP) dimensional reduction was performed using the first 30 principal components. Clusters were identified using the “FindClusters” command with resolution = 0.5 using the Louvain algorithm⁵¹. Differential gene expression analysis was performed using the R package DESeq2⁵². Genes with log2 fold changes > 0.5, percent expression > 25%, and adjusted p value < 0.001 were considered significant.

Animals

This study was performed according to experimental protocols approved by the Institutional Animal Care and Use Committees of Massachusetts General Hospital. All methods were performed in accordance with the relevant guidelines and regulations this study is reported in accordance with ARRIVE guidelines. Mice were euthanized using CO₂ asphyxiation. Plp1::GFP mice were gifted to the Goldstein laboratory by Wendy Macklin⁵³ or purchased from Jackson Laboratories (Bar Harbor, ME, USA) (JAX Stock #033357). Animals homozygous for GFP expression were crossed with B6;129-Ednrb^tm1Ywa/J (JAX Stock #003295). Ednrb heterozygotes were bred to generate Ednrb knockout mice (Ednrb^−/−) which have distal colonic aganglionosis akin to human Hirschsprung disease and where EGCs are labeled with GFP fluorescence. Additionally, Plp1-GFP;Baf-tdT dual reporter mice were generated as previously described (BAF53b-Cre, JAX Stock #027826)^34,54, and mice homozygous for both tdT and GFP expression were similarly crossed with B6;129-Ednrb^tm1Ywa/J (JAX Stock #003295). Ednrb heterozygotes were bred to generate Ednrb knockout mice with EGCs labeled with GFP fluorescence and mature enteric neurons labeled with tdT fluorescence. All mice were housed and bred at the Center for Comparative Medicine animal facility at Massachusetts General Hospital under specific pathogen-free conditions. Additionally, wholemount tissue from Gfra1^fl-EGFP;Dhh::Cre mice was kindly gifted from Dr. Enomoto¹⁰.

Human tissue

This study was approved by the research ethics committees of Baylor College of Medicine (IRB protocol #H-50144). Written informed consent was collected for all subjects and/or their legal guardians prior to participation in this study. All methods were performed in accordance with the relevant guidelines and regulations. Specimens of ganglionated and aganglionic Hirschsprung disease (HSCR) colon were taken from patients during pull-through surgery at Texas Children’s Hospital.

Isolation of cells and generation of neurospheres

Cells were isolated from mice and grown in neuroproliferation media to generate neurospheres as described previously⁵⁵. Briefly, mice were euthanized and their colon was removed from cecum to rectum. The longitudinal muscle-myenteric plexus (LMMP) layer was dissected from underlying tissue in PBS. Using a fluorescent microscope, the tissue was segmented into aganglionic and ganglionic segments, and the transition zone was excluded. The tissue was digested for 60 min at 37 °C in dispase (250 μg/ml; STEMCELL Technologies, Vancouver, BC) and collagenase XI (1 mg/ml; Sigma-Aldrich, St. Louis, MO). Following digestion, the cells were filtered through a 10 µm filter. Anything larger than 10 µm was discarded. For the ganglionic segment, the flow through < 10 µm was considered a single cell suspension, which primarily contained EGs, but also likely contained a smaller proportion of IGs, SLCs, and neurons. For the aganglionic segment, the flow through < 10 µm was considered a single cell suspension, which primarily contained EGs, but also likely contained a smaller proportion of SLCs. Cells were resuspended with a total volume of 3 mL of neuroproliferation media placed in ultra-low attachment 6-well plates (Corning Inc, Corning, NY). Neuroproliferation media contained Antibiotic–Antimycotic (1%; Gibco, Thermo Fisher Scientific)B27 Supplement (1x; Gibco, Thermo Fisher Scientific), N-2 Supplement (1x; Gibco, Thermo Fisher Scientific), basic fibroblast growth factor (20 ng/mL; Stemcell Technologies), insulin-like growth factor 1 (20 ng/mL; Thermo Fisher Scientific), retinoic acid (75 ng/mL; Sigma Aldrich), and 2-mercaptoethanol (50 μmol/L; Gibco, Thermo Fisher Scientific) in equal parts Neurocult Mouse Basal Medium (Stemcell Technologies) and Dulbecco’s Modified Eagle Medium (DMEM; Gibco, Thermo Fisher Scientific). These primary neurospheres were cultured at 37 °C and 5% CO₂ for 10–14 days.

Immunohistochemistry and image processing

Immunohistochemistry was performed as previously described⁵⁶. For cryosections, tissue was fixed in 4% PFA for 1 h, then incubated with 15% sucrose overnight at 4 °C, infiltrated with 7.5% gelatin/15% sucrose in PBS for 1 h at 37 °C, then rapidly frozen at − 50 °C in methylbutane. Twelve µm-thick cryosections were stained using the following primary antibodies: mouse anti-tubulin β3 (Tuj1, 1:400, conjugated to Alexa Fluor 594, (BioLegend #801208). Primary antibodies were applied for 45–90 min, followed by Alexa-conjugated fluorescent secondary antibodies: Alexa Fluor 546 donkey anti-rabbit (1:400, Invitrogen, A10040). Cell nuclei were stained with DAPI (Invitrogen). Human tissue was stained with fluorescently Alexa Fluor 594 conjugated mouse anti-TUJ1 (1:400, Biolegend, 801208) and rabbit anti-S100B (1:200, Abcam, ab52642). Secondary antibody included the fluorescent secondary antibody Alexa Fluor 488 (donkey anti-rabbit, 1:400, Invitrogen, A-21206). The sections were covered by aqueous Poly/Mount (Polyscience Inc. Warrington, PA, 18606). Section images were recorded using a Zeiss LSM 710 confocal microscope.

For whole-mount immunofluorescent staining of mouse tissue, colons were fixed overnight in 4% PFA at 4 °C, permeabilized with 10% donkey serum (Sigma, #D9663) and 0.1% Triton X-100 in PBS (Sigma Aldrich, 9036-19-5) for 1 h. Specimens were then labeled overnight with primary antibodies that included rabbit anti-SOX10 (1:250, Abcam, ab155279) in a solution of 2% donkey serum and 0.1% Triton X-100 in PBS at 4 °C. After washing in PBS, fluorescent secondary antibody (Alexa Fluor 546, donkey anti-rabbit, 1:400, Invitrogen, A10040) in a solution of 2% donkey serum and 0.1% Triton X-100 in PBS was applied for 2 h at room temperature, and then subsequently covered with diluted DAPI solution (Invitrogen). Whole-mount images were recorded using a Keyence BZX-700 all-in-one microscope (Keyence America, Itasca, IL, USA).

For staining, cells were fixed for 30 min in 4% PFA at 4 °C, permeabilized in 0.1% Triton X-100 for 20 min, washed with PBS, and blocked with 10% donkey serum for 1 h at room temperature. Primary antibodies were diluted in 10% donkey serum, which included anti-Hu (ANNA-1) (1:10,000, gifted from Dr. Vanda Lennon; Mayo Clinic) and incubated overnight at 4 °C. Secondary antibodies were diluted in 10% donkey serum, which included Alexa Fluor 594 donkey anti-human (1:400, Jackson ImmunoResearch, 709–585-098) and incubated at room temperature for 2 h. Cell nuclei were counterstained with DAPI solution (Invitrogen). Images were obtained with a Keyence BZX-700 all-in-one microscope (Keyence America, Itasca, IL, USA) and analyzed using Image J software v1.53t (National Institutes of Health, Bethesda, MD, USA).

Image processing and in vitro analysis

Image processing was performed using ImageJ software v1.53t (National Institutes of Health, Bethesda, MD, USA, http://rsb.info.nih.gov/ij/). Immunohistochemically labelled cells and transgenic fluorescence were imaged on a Keyence BZX-700 All-In-One Microscopy system (Keyence America, Itasca). Primary neurospheres generated from Plp1-GFP;Baf-tdT;Ednrb^−/− mice were plated on fibronectin in neuroproliferation media and cultured for 1 week at 37 °C. Cells were grown in media alone, supplemented with GDNF (50 ng/mL), or supplemented with GDNF (50 ng/mL) and anti-NCAM1 antibody (ab220360, 1:250). After 1 week cells were fixed and stained as described above. For quantification of Plp1-GFP+ EGCs and Hu+ ENs a custom Fiji macro was utilized to quantify cells within ganglia-like clusters. To obtain regions of interest (ROI) of ganglia-like clusters, the Plp1-GFP channel was converted to 8-bit, and Gaussian blur (sigma = 5) was applied to remove borders between individual cells and capture the larger network anatomy. Local thresholding using the ‘Phansalkar’ method was performed, followed by binary mask conversion and hole filling. ROIs of the cellular network were extracted from the binary image and were utilized for the area of cell quantification in subsequent steps. Within the ROI, the Hu and Plp1-GFP channels were processed from separate images using local thresholding with the ‘Phansalkar’ method as described above without a Gaussian blur applied to focus the analysis on individual cells, and applying the ‘Fill Holes’ and ‘Watershed’ functions to separate individual objects. Cell counts for Plp1-GFP and Hu channels were determined using the ‘Analyze Particles’ function (size = 20–∞ µm², circularity = 0.30–1.00). For analysis of neural network skeletons images of BAF53b-tdT fluorescence were processed by enhancement of local contrast using the CLAHE algorithm (block size = 20, histogram bins = 256, maximum slope = 12) to improve neurite detection and background subtracting using a rolling ball radius approach. Local thresholding was applied using the ‘Phansalkar’ method, followed by despeckling and binary conversion. The dilation function was performed on binarized images to prevent gaps in pixels within the fine neurite branches and the images were skeletonized, and skeleton analysis was performed using the ‘Analyze Skeleton (2D/3D)’ plugin without pruning to obtain results, including branch and node junction properties.

Statistical analysis

Analyses were performed using GraphPad Prism software (GraphPad, CA, USA). Depending on the number of comparisons, EGC/EN ratios and network formation metrics were assessed using a Mann Whitney test or Brown-Forsythe ANOVA test with an Unpaired t test with Welch’s correction performed posthoc. p values < 0.05 were considered statistically significant. Bar graphs are presented with the mean and standard error of the mean (SEM) bars.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

JLM and RS—conceptualization, acquisition of data, analysis, and writing the main text. RS and AMG—conception and design of project, project oversight, reviewed manuscript. CH and AL—data acquisition and analysis. VC, LO, RG, TU, HE, and LC—data acquisition. RH and AB – conception and design, review of manuscript.

Funding

Grant Support: JLM: F32DK131792; RS: Charles H. Hood Foundation Child Health Research Award; AMG: 5R01DK119210.

Data availability

Sequencing data can be accessed at [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE293138] (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE293138) (GEO accession GSE293138, token: gbudoeoenjufzml). All other datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.