Heterogeneity and phenotypic plasticity of glial cells in the mammalian enteric nervous system

Enteric glial cells are vital for the autonomic control of gastrointestinal homeostasis by the enteric nervous system. Several different functions have been assigned to enteric glial cells but whether these are performed by specialized subtypes with a distinctive phenotype and function remains elusive. We used Mosaic Analysis with Double Markers and inducible lineage tracing to characterize the morphology and dynamic molecular marker expression of enteric GLIA in the myenteric plexus. Functional analysis in individually identified enteric glia was performed by Ca2+ imaging. Our experiments have identified four morphologically distinct subpopulations of enteric glia in the gastrointestinal tract of adult mice. Marker expression analysis showed that the majority of glia in the myenteric plexus co‐express glial fibrillary acidic protein (GFAP), S100β, and Sox10. However, a considerable fraction (up to 80%) of glia outside the myenteric ganglia, did not label for these markers. Lineage tracing experiments suggest that these alternative combinations of markers reflect dynamic gene regulation rather than lineage restrictions. At the functional level, the three myenteric glia subtypes can be distinguished by their differential response to adenosine triphosphate. Together, our studies reveal extensive heterogeneity and phenotypic plasticity of enteric glial cells and set a framework for further investigations aimed at deciphering their role in digestive function and disease. GLIA 2015;63:229–241


Introduction
G lial cells are essential for the organization and function of the nervous system (Jessen, 2004). In addition to their "traditional" roles in providing nourishment and support for neurons, glial cells regulate synaptic transmission, maintain the blood-brain barrier, mediate communication between the nervous and immune systems and monitor the nutritional state of organisms (Allaman et al., 2011;Giaume et al., 2010;Perea et al., 2009). Highlighting their critical role during embryogenesis and in postnatal life, several developmental, degenerative, and inflammatory disorders of the nervous system have been associated with deficits in glial cell function (Barres, 2008;Lee et al., 2000;Lemke, 2001).
Enteric glial cells (EGCs) constitute a major population of peripheral glia that is located within the ganglia of the myenteric and submucosal plexus of the enteric nervous system (ENS) and in extraganglionic sites, such as the smooth muscle layers and the mucosa (Gershon and Rothman, 1991;Gulbransen and Sharkey, 2012;Ruhl et al., 2004). A series of studies has demonstrated that EGCs are critical for the regulation of gut motility, the secretory and absorptive functions of the intestinal epithelium and the host's defense against pathogens (De Giorgio et al., 2012;Gulbransen and Sharkey, 2012;Neunlist et al., 2013). Enteric neurons and EGCs are derived from neural crest cells which emerge during embryogenesis from the vagal and sacral level of the neural tube and colonize uniformly the entire length of the gastrointestinal tract (Heanue and Pachnis, 2007;Sasselli et al., 2012). Although diversity of enteric neuron subtypes has been studied intensively in mammals (Furness, 2000), our understanding of the extent, the origin and physiological significance of EGC diversification remains unclear. Pioneering work from Hanani and Reichenbach (1994) proposed a classification scheme for EGCs that was based on cell morphology and location within the plexus. However, the recognition that essential subsets of EGCs are located outside the ENS plexus, the advent of genetic tools that allow detailed characterization of EGCs and recent progress in understanding their physiological properties, warrant a reexamination of EGC subtype classification in a manner that takes into consideration morphological features, distribution within the gut wall as well as molecular and physiological attributes. Moreover, given that the gut constitutes one of the most dynamic organs of the body which adjusts constantly to the volatile microenvironment of the lumen (unpredictable availability of nutrients, changing populations of commensal bacteria, and exposure to pathogenic microorganisms), any classification scheme for EGCs must also consider the potential plasticity of subtypes identified at any given moment.
In the present study, we have carried out high resolution morphological characterization of EGC subtypes by combining Mosaic Analysis with Double Markers (MADM) with immunostainings for glial markers. In addition, we have examined the physiological properties of different subtypes of myenteric glia using Ca 21 imaging on live tissue. Furthermore, using Cre/LoxP technology we have carried out in vivo fate mapping and inducible lineage tracing of defined subsets of EGCs with the aim of exploring their plasticity in the myenteric ganglia. Our experiments extend previous studies on the morphological and physiological characterization of EGCs and highlight a novel subtype of extraganglionic glia which is located within the plexus. Moreover, our experiments have revealed previously unrecognized plasticity in the molecular phenotype of defined glial subpopulations of the mammalian ENS.

Animals
All experimental procedures were carried under a UK Home Office Project license and approved by the local Animal Ethics Committees of the National Institute of Medical Research and the University of Leuven. Animals were housed at an ambient temperature of 22 C-24 C on a 12:12 hr light/dark cycle. Drinking water and laboratory chow were provided ad libitum. The Sox10::Cre (Matsuoka et al., 2005), MADM-6 GR and MADM-6 RG (Zong et al., 2005), hGFAP::-CreER T2 (Ganat et al., 2006), and the Rosa26R-YFP (Srinivas et al., 2001) transgenic mice have been described previously. In vivo labeling of EGCs in hGFAP::CreER T2 ; Rosa26R-YFP adult animals was achieved by intraperitoneal injections of 0.1-0.2 mg/g of 4-hydroxy tamoxifen (4-OHT; Sigma), which was dissolved in an ethanol/ sunflower oil (1:9) mixture at 10 mg/mL (Laranjeira et al., 2011). Control mice were not injected or injected with an ethanol/ sunflower oil (1:9) mixture.

Immunofluorescence of Adult Myenteric Plexus Preparations, Confocal Image Acquisition and Processing
Gut tissue was cut into 2 cm long pieces and mounted over a 1 mL pipette. An incision was made along the mesenteric border and a cotton swab soaked in PBS was used to roll out the longitudinal muscle with adherent myenteric plexus from the underlying tissue. Isolated adult myenteric plexus preparations were stretched and pinned onto Sylgard dishes (Sylgard 184 Elastomer, Dow Corning) using insect pins (0.2 mm, Agar Scientific) and fixed in ice-cold paraformaldehyde (4% in PBS) for 2 hr at 4 C. Fixed tissue samples were permeabilized in 0.3% TritonX-100 (Sigma) in PBS (PBT) and incubated in blocking solution (5% Normal Donkey Serum in PBT, Jacksons ImmunoResearch Laboratories). Primary antibodies (Table 1) were diluted in blocking solution and applied overnight at 4 C. Samples were washed in PBT and appropriate secondary antibodies were applied. The GFP antibodies were used to detect expression of the YFP reporter, and the two terms are used interchangeably. Secondary antibodies were goat or donkey hosted Alexa-Fluor 488, Alexa Fluor 568, Alexa Fluor 594, and Cy5 (all 1:500; Invitrogen and Jackson ImmunoResearch). Samples were mounted with Vectashield TM mounting medium (Vector Labs).
Confocal and 3-D Images were acquired with Leica TCS SP5 (CIAL facility, NIMR, UK) and Zeiss LSM780 2-photon (Cell Imaging Core, KU Leuven) confocal microscopes using standard excitation and emission filters, assisted with Leica Application Suite Advanced Fluorescence (LAS AF) and Zeiss ZEN software, respectively. All images were processed with either Adobe Photoshop CS4 (Adobe Systems) or ImageJ (Wayne Rasband, NIH). 3-D reconstructions and morphological quantifications were obtained with Imaris 7.7 (Bitplane AG) after deconvolution using Huygens Professional (Scientific Volume Imaging).

Live Calcium Imaging
Ca 21 imaging of EGCs and subsequent analysis was performed as previously described (Boesmans et al., 2013b;Vanden Berghe et al., 2008). The proximal colon of MADM-6 GR/RG ; Sox10-Cre mice was pinned flat in a Sylgard-lined dish filled with Krebs solution (bubbled with 95% O 2 to 5% CO 2 ) and the mucosa and longitudinal muscle layer were carefully removed. Myenteric plexus preparations were loaded with 1 lM Rhodamine 2 (Rhod2, RT, 20 min, TEFLabs) in Krebs with cremophor EL (0.01%, Fluka). During recordings, a constant flow (1 mL/min) of Krebs (with or without 10 lM ATP, Sigma) solution was maintained via a gravity-fed electronic valve system. 1 lM nifedipine (Sigma) was added to the Krebs solution to inhibit smooth muscle contractions. Images were recorded on an upright Axiovert 200M microscope (Zeiss) with a Poly V xenon monochromator (TILL Photonics) and water dipping lens (20X, 1.0 NA, Zeiss). MADM-induced GFP and Rhod2 were excited at 475 nm and 580 nm and captured at 525/50 nm (still) and 645/75 (at 2 Hz), respectively, on a Sensicam-QE CCD camera (PCO) using TillVisION (TILL Photonics). Calcium imaging analyses were performed with custom-written routines in Igor Pro (Wavematrics). Regions of interest were drawn, after which average Ca 21 signal intensity was calculated, normalized to the initial Rhod2 values and reported as DF/F 0 . Cells were considered as responders when the Rhod2 signal rose above baseline plus 3 times the intrinsic noise (standard deviation) during the recording.

Statistical Analysis
Analysis was performed on a total sample of cells (as reported) that were obtained from at least three animals per genotype or experimental condition. Differences between data sets were determined by ANOVA followed by Bonferroni post-hoc test (for multiple comparison tests). Chi-square or Fisher's exact tests were performed to compare proportions. A P value <0.05 was considered to be significant. Significant differences with P < 0.05, 0.01, and 0.001 were indicated with *, **, and ***, respectively. Statistical analysis was performed with Microsoft Excel (Microsoft) and GraphPad Prism (GraphPad Softwares). Error bars represent the standard error of the mean (SEM).

High Resolution Labeling of EGCs Using Mosaic Analysis With Double Markers (MADM)
Glia-specific molecular markers that are expressed widely in the central and peripheral nervous system have been used extensively for the identification and characterization of EGCs in mouse embryos and adult animals. Among such markers are the intermediate filament glial fibrillary acidic protein (GFAP) and the Ca 21 -binding protein S100b, which are expressed by committed EGC progenitors and differentiated EGCs (Ferri et al., 1982;Jessen and Mirsky, 1980;Young et al., 2003). In addition to these cytoplasmic proteins, the nuclear localized transcription factor Sox10 is expressed by multipotential ENS progenitors (Bondurand et al., 2003;Paratore et al., 2002) and is maintained in immature and mature EGCs (Hoff et al., 2008;Laranjeira et al., 2011;Young et al., 2003). These markers, which are usually identified by immunostaining on fixed gut preparations, are either unsuitable for morphological characterization of EGCs (Sox10) or are expressed by contiguous glial cells thus precluding studies that require single cell resolution. We reasoned therefore that sparse labeling of EGCs with genetically encoded fluorescent reporters would facilitate high resolution morphological and physiological studies on both fixed and live tissue. To achieve this, we employed MADM, a sparsely labeling genetic system in which Cre-dependent interchromosomal recombination restores expression of fluorescent lineage tracers encoded by reciprocally chimaeric nonfunctional cassettes targeted to the same locus on homologous chromosomes (Zong et al., 2005). For our current studies we used the original MADM alleles (MADM-6 GR and MADM-6 RG ) in which chimaeric GFP/RFP-Myc vectors have been inserted into the ubiquitously expressed Rosa26 locus on mouse chromosome 6 (Zong et al., 2005). Introduction of the Sox10::Cre transgene, which drives expression of Cre recombinase in all neural crest cell lineages and peripheral glial cells (Laranjeira et al., 2011;Matsuoka et al., 2005), into MADM-6 GR/RG compound heterozygous mice (designated hereafter Sox10-CrejMADM), allowed us to mark with GFP or RFP-myc or both, isolated EGCs ( Fig. 1a) and enteric neurons (which originate from marked Sox10 1 progenitors, Supp. Info. Fig. 1a). Since detection of RFP in Sox10CrejMADM mice depends on immunostaining with anti-myc antibodies (Supp. Info. Fig. 1b), all studies reported here are based on the identification of GFP on fixed or live gut tissue preparations. The glial identity of all GFP 1 cells analyzed from Sox10CrejMADM mice has been confirmed by double immunostaining with S100b-specific antibodies (Fig. 1b-e). Characterization of GFP-expressing cells in wholemount preparations of the different layers of the gut wall from Sox10CrejMADM mice, allowed us to identify four EGC subtypes on the basis of morphology and location. Restricted within the ganglia of the myenteric and submucosal plexus were numerous GFP 1 cells bearing multiple irregular and highly branched processes which extended in all directions up to 42.5 6 2.5 lm away from the cell body (Fig. 1b,  these intraganglionic glial cells often terminate with "endfeet" like structures and were intermingled with processes from other neighboring EGCs and were contacting multiple neurons within the same ganglion ( Fig. 2a; Supp. Info. Movie 2). These astrocyte-like glial cells correspond to the "protoplasmic" or type-I EGCs described previously by Hanani and Reichenbach (1994). These authors provided also the first description of the "fibrous" or type-II EGC found within or at the edges of the interganglionic connectives. Using GFP expression as marker, we have confirmed that this EGC subtype has lanky processes that emerge mostly from one side of the cell and run for 53.9 6 6.1 lm in parallel to each other and to the neuronal fibers of the interganglionic strands ( Fig. 1c,g; Table 2; Supp. Info. Movie 3). These glial cell processes contact most neural fibers but they do not wrap around them in a manner analogous to the ensheathing of peripheral nerve fibers by Schwann cells (Fig. 2b; Supp. Info. Movie 4). In addition to the type-I and type-II EGCs we have uncovered a morphologically distinct glial subtype which is at the level of the myenteric and submucosal plexus but lies outside the ganglia and the interganglionic connectives (Fig. 1a). This subtype, designated hereafter as type-III, exhibits mainly four major processes (87.6 6 5.3 lm) with secondary branching (Fig. 1d,h; Table 2; Supp. Info. Movie 5). Type-III EGCs appear to form a matrix in the extraganglionic region, and remain in close association with neuronal fibers running along their processes (Fig. 2c). In addition, the processes of type-III EGCs are often seen wrapping around small blood vessels (Fig. 2d). EGCs with a similar morphology have been described previously as a distinct subtype within the lamina propria of the mucosa and have been termed "mucosal" or type-III EGCs (Gulbransen and Sharkey, 2012;Savidge et al., 2007). Given their morphological similarities but distinct location, we propose to discriminate between the mucosal and plexus positioned type-III EGCs by referring to them as type-III Mucosa and type-III MP/SMP , respectively. Finally, our MADM labeling system also revealed EGCs with a characteristic bipolar morphology which were located within the circular and longitudinal smooth muscle layers along nerve fibers (Fig. 1e,i; Table 2; Supp. Info. Movie 6). These cells correspond to the type-IV EGCs described previously (Gulbransen and Sharkey, Vanderwinden et al., 2003). Taken together, our experiments extend previous studies on the classification of EGCs and demonstrate that MADM-mediated fluorescent labeling can be used effectively for high resolution morphological analysis of individual glial cells throughout the mammalian ENS. All our subsequent studies were carried out on glial cells (types I-III) of the myenteric plexus.

Differential Expression of Glial Markers in Subtype
Populations of Myenteric Glia So far, the markers GFAP, S100b, and Sox10 have been used interchangeably for the identification of EGCs. To determine whether the EGC subtypes within the myenteric plexus can be distinguished by the differential expression of these commonly used glial markers, we performed double immunostaining of myenteric plexus preparations isolated from the intestine of P90 mice using combinations of marker-specific antibodies. First, we quantified the expression of GFAP and S100b. Whereas, the majority of type-I EGCs residing in the myenteric ganglia co-expressed GFAP and S100 (GFAP 1 /S100b 1 ; 55.8 6 5.2%), the remainder cells of this subtype population expressed either S100b (GFAP 2 /S100b 1 ; 29.9 6 4.4%) or GFAP (GFAP 1 /S100b 2 ; 14.3 6 3.9%) only (Fig. 3a,b). Interestingly, the percentage of GFAP 2 /S100b 1 cells among the type-II and type-III MP EGCs was significantly higher (59.6 6 1.6% and 78.9 6 0.7%, respectively) relative to that in type-I glia. Notably, a relatively high percentage of GFAP 1 / S100b 2 cells (50%) displayed remarkably high immunoreactivity for GFAP (Fig. 3a, arrow). Next, we compared the expression of GFAP and Sox10. Similar to our findings with the GFAP/S100b combination, in all three EGC subtypes we identified cells that expressed either one or both of these markers. Thus, 74.4 6 2.4% of the type-I glia were GFAP 1 / Sox10 1 but only 19.3 6 1.5% were GFAP 2 /Sox10 1 and 6.2 6 0.9% expressed only GFAP (Fig. 3c,d). In type-II EGCs the percentage of GFAP 2 /Sox10 1 cells increased to 46.4 6 3.7% at the expense mostly of the double positive population, whereas in type-III MP glia it increased further to 67.6 6 5.1% at the expense of both the double positive and the GFAP-only population (Fig. 3d). The relative percentage of GFAP 2 / Sox10 1 cells observed in the three types of glial cells mirrors the relative percentage of GFAP 2 /S100b 1 glia (Fig. 3b) suggesting that the majority of S100b 1 glial cells co-express Sox10. This was confirmed by double immunostaining with a combination of S100b and Sox10 antibodies which demonstrated that virtually all S100b 1 cells co-expressed Sox10 (Fig.  3e,f). Nevertheless, a small population of type-I (5.5 6 1.5%) and type-II (7.4 6 1.4%) glia expressed only Sox10 (S100b 2 / Sox10 1 ) and this fraction increased slightly in type-III MP EGCs (10.7 6 1.9%, Fig. 3f). Taken together, these data indi-cate that none of the markers examined here (and commonly used for the identification of EGCs) can uniquely identify any of the morphologically distinct glial subtypes within the myenteric plexus. Nevertheless, the differential abundance of cells expressing combinations of GFAP, S100b, and Sox10 within the glial subtypes I-III of the myenteric plexus suggests that glia-specific gene expression in the mammalian gut is highly dynamic and is influenced at the population level by spatially and temporally controlled factors.

Differential Response of EGC Subtypes to Purinergic Activation
To investigate whether, in addition to the differential marker expression, EGC subtypes are also characterized by distinct physiological properties, we performed Ca 21 imaging experiments on fresh preparations of myenteric plexus. Given the established paradigm of purinergic neuron-to-glia communication in the ENS (Boesmans et al., 2013b;Gomes et al., 2009;Gulbransen and Sharkey, 2009), we decided to test whether the three myenteric glia subtypes respond differently to purinergic receptor stimulation. For this, we loaded myenteric plexus preparations dissected from Sox10CrejMADM mouse colon with the red Ca 21 indicator Rhodamine-2. This enabled us to unambiguously measure the changes in intracellular Ca 21 concentration upon ATP (10 lM, 20 s) stimulation in individual cells that belong to the three EGC subtypes of the myenteric plexus (Fig. 4). We found that the majority (86.8%) of intraganglionic type-I EGCs respond with a Ca 21 transient (DF/F 0 : 7.0 6 0.9%) when challenged with FIGURE 3: Differential expression of GFAP, S100b, and Sox10 in enteric glia of the myenteric plexus of the ileum. (a) Double immunostaining for GFAP (green) and S100b (gray). Although most EGCs express both markers, EGCs only positive for GFAP (arrow) or S100b (arrowheads) can be observed. (c) Double immunostaining for GFAP (green) and Sox10 (red). Note the presence of EGCs expressing only GFAP (arrows) or Sox10 (arrowhead). (e) Double immunostaining for S100b (gray) and Sox10 (red). The majority of cells were found positive for both S100b and Sox10. A small proportion of cells positive for Sox10 did not express S100b (arrows). EGCs positive for S100b but not for Sox10 were not observed. ATP ( Fig. 4d-g). Although purinergic receptor stimulation induced Ca 21 rises in a similar proportion (73.9%) of type-II EGCs in the interconnectives fiber tracts, Ca 21 transient amplitudes were significantly lower (DF/F 0 : 3.1 6 0.6%) (Fig. 4f,g). Among type-III MP EGCs located in the intergan-glionic space, only 34.6% displayed Ca 21 rises (DF/F 0 : 1.9 6 0.3%) upon ATP stimulation (Fig. 4f,g). Thus, compared to type-I and type-II EGCs, extraganglionic type-III MP cells display a decreased responsiveness to purinergic receptor stimulation. These data are consistent with our

Sox10 Negative EGCs are Derived From Sox10-Expressing Progenitors
Since Sox10 is known to mark all glia of the peripheral nervous system (Herbarth et al., 1998;Pusch et al., 1998), we were surprised to identify a small but readily detectable population of GFAP-expressing EGCs that were negative for this transcription factor (Fig. 3c,d). This subset of EGCs could be derived either from a previously uncharacterized Sox10 2 cell lineage or from Sox10 1 precursors that down-regulate expres-sion of Sox10 upon terminal differentiation. To distinguish between these possibilities, we used Cre/LoxP technology to fate map Sox10-expressing ENS progenitors in vivo. More specifically, we combined the Sox10::Cre transgene, which drives expression of Cre recombinase in all neural crest cell lineages, with the Cre-dependent reporter Rosa26-ReYFP (Laranjeira et al., 2011;Matsuoka et al., 2005). Expression of Sox10 in these mice (called hereafter Sox10CrejYFP) would be accompanied by induction of Cre and permanent activation of the Rosa26-ReYFP reporter irrespective of the continued activity of the Sox10 locus. By analyzing preparations of the myenteric plexus from P90 Sox10CrejYFP mice, we found that the majority (95.7%, 47 cells) of Sox10 2 /GFAP 1 cells expressed YFP, indicating that they are derived from Sox10 1 progenitors (Fig. 5). These data confirm that all populations of EGCs in the myenteric plexus are derived from Sox10-expressing progenitors and suggest that the expression of characteristic marker genes in EGC lineages can be highly dynamic. The latter was confirmed when double immunostaining for GFAP, S100b, and Sox10 was performed on primary cultured EGCs derived from the myenteric plexus (Supp. Info. Fig. 2). Although we found that the levels of GFAP immunoreactivity did vary between individual cells in culture, EGCs expressing only one of the markers were not observed after 4 days in vitro. Thus, when EGCs are isolated from their natural environment, and their spatial interactions with other cell types in the gut wall are lost, they can no longer be distinguished based on their expression of GFAP, S100b, or Sox10. Taken together, our experiments support the phenotypic plasticity of EGC subtypes.

Dynamic Expression of GFAP in EGCs
Up-regulation of GFAP is one of the main features of reactive astrogliosis, a hallmark of several central nervous system pathologies (Sofroniew, 2009). In the ENS, increased GFAP expression has been demonstrated in inflammatory bowel disease patients, in mouse EGCs following ischemia and in cytokine-treated cultures of rat EGCs (Cornet et al., 2001;Thacker et al., 2011;von Boyen et al., 2004). Variation in the levels of GFAP expression is also illustrated by the different grades of GFAP immunoreactivity that we have observed mostly in type-I EGCs (Fig. 3a,b). Although these studies suggest a certain degree of plasticity in the molecular phenotype of EGCs, direct evidence for a temporal regulation of glial-specific cell markers at the single-cell level in the ENS of adult animals is currently not available. In addition, it remains unclear whether the differential expression of GFAP marks distinct sublineages of EGCs or represent mere fluctuations in the levels of this protein in response to cell intrinsic or environmental changes. To address these issues, we used Cre/LoxP technology to follow over a period of several days the potential change in GFAP expression in a defined population of GFAP 1 EGCs. More specifically, mice heterozygous for the hGFAP::CreER T2 transgene, which drives expression of tamoxifen-inducible Cre recombinase under the control of the human GFAP promoter (Ganat et al., 2006), and the Cre-dependent Rosa26-ReYFP reporter (mice called GFAPCreERjYFP), were administered 4-OHT and expression of GFAP in YFP 1 cells was analyzed in myenteric plexus preparations 2 and 7 days later. As expected, 2 days after 4-OHT administration, the vast majority of YFP 1 EGCs (91%) was also positive for GFAP (Fig. 6a). Interestingly, similar analysis 7 days following 4-OHT injection showed that only 77% of YFP 1 cells co-expressed GFAP (Fig. 6b). Since the percentage of GFAP 1 cells in the myenteric ganglia of adult animals remains constant (Joseph et al., 2011;Stenkamp-Strahm et al., 2013), these experiments suggest that under physiological conditions expression of GFAP in individual EGCs varies considerably over time.

Discussion
Traditionally, EGCs were thought to contribute primarily to the structural integrity and nourishment of the ENS. However, during the last decades several studies have confuted the concept of a merely supportive function of EGCs and ascribed to them a wide variety of roles that are essential for proper gastrointestinal function. Thus, in addition to being a scaffold for neurons, EGCs function in mucosal integrity, neuroprotection, adult neurogenesis, neuro-immune interactions, and synaptic transmission (De Giorgio et al., 2012;Gulbransen and Sharkey, 2012;Neunlist et al., 2013;Ruhl et al., 2004). Although different morphological subtypes of EGCs have been described previously (Hanani and Reichenbach, 1994), they are generally thought to represent manifestations of a molecularly and physiologically homogeneous multitasking cell population. Some differences in the expression of receptors, channels, and enzymes have been reported (Boesmans et al., 2013b(Boesmans et al., , 2014Costagliola et al., 2009;Nasser et al., 2007Nasser et al., , 2006bTjwa et al., 2003), but so far no studies have addressed the intrinsic heterogeneity of the recognized subtypes of EGCs. Recent in vivo ablation models have advanced considerably our understanding of EGC function (Aube et al., 2006;Bush et al., 1998;Cornet et al., 2001;Nasser et al., 2006a), but tissue disruption that encompasses relatively large cell populations is likely to lead to general effects on gastrointestinal homeostasis thus making it very difficult to pinpoint the relative contribution of EGC subtypes to the impaired epithelial barrier, inflammation, and nerve defects that have been observed. Here, we have employed a high resolution genetic marking technique which allowed us to label individual glial cells of the gastrointestinal tract in situ and characterize their morphological and physiological properties. Using this approach, we have extended previous morphological studies and propose that the complement of EGCs is subdivided into four subtypes which correspond to unique locations within the plexus and extraganglionic spaces. Although these subtypes share many of the molecular markers that have been used to identify glial cells throughout the nervous system, they can be distinguished by the prevalence of marker expression across different populations. In particular, GFAP, a marker used widely for the identification of peripheral glial cells including EGCs, is expressed primarily by type-I intraganglionic glial cells but the majority of type-II and type-III EGCs are negative for this marker. Although GFAP has been implicated in the morphological characteristics and adhesive properties of astrocytoma cells (Sofroniew and Vinters, 2010), it is currently unclear whether the differential distribution of this intermediate filament has a functional relevance or simply reflects the sensitivity of type-I EGCs to GFAP-inducing signals. Irrespective of the physiological relevance of GFAP expression, our data suggest that the morphologically distinct subsets of EGCs express unique combinations of glial cell markers in response to signals from their immediate microenvironment and/or as a result of functional specialization. Our molecular studies so far have employed standard glial cell markers that are widely expressed in other parts of the peripheral and central nervous system and which have been used extensively in the ENS field. Systematic characterization of the molecular profile of EGC subtypes is likely to uncover unique combinations of additional molecules that can be used for the further classification and study of EGCs. Combinatorial codes of marker molecules emerging from such studies will assist considerably further investigations on the roles of EGCs in normal gut physiology and pathophysiology.
Heterogeneity among EGC subtypes could be developmentally hard-wired by the differentiation of distinct glial cell lineages emerging during embryogenesis. Alternatively, the different subtypes of EGCs could originate from the same glial cell precursors which, depending on their final destination, could give rise to morphologically and functionally distinct cell progeny. Developmental studies so far have failed to identify distinct progenitors of EGCs and the markers used so far are thought to label interchangeably a common progenitor (Heanue and Pachnis, 2007;Sasselli et al., 2012). Moreover, our lineage studies here argue for a considerable degree of phenotypic plasticity among EGC subtypes in adult animals. Taken together, these findings support a model according to which all subtypes of EGCs originate from a common progenitor which, depending on its final location and physiological context is capable of generating different glial subtypes. Probably, the ability of glial cells to acquire different properties is not restricted to embryonic or early postnatal stages but is maintained throughout adult life pointing to a previously unrecognized degree of phenotypic plasticity of the ENS. Given the high correlation between location of EGCs within the plexus and phenotypic properties, it is currently unclear whether the acquisition of new properties by glial cells in adult animals is associated with cell translocation between the different subdomains of the neuroglial networks of the gut. Irrespective of the underlying cellular mechanisms, our data indicate that EGCs are capable of adjusting their phenotypic and molecular characteristics in response to continuous challenges associated with nutrition, microbiota, mechanical factors, or disease.
Although the exact role of EGCs in the mechanisms controlling the physiological output of the gut wall such as peristalsis or secretion remains elusive (Broadhead et al., 2012;MacEachern et al., 2011), previous studies have shown that purines arising from both intrinsic enteric and sympathetic neurons are the principal mediators of neuron-to-glia communication in the ENS (Gomes et al., 2009;Gulbransen and Sharkey, 2009;Gulbransen et al., 2010). Our experiments provide further support to this model and highlight the plausible role of type-I EGCs in integrating paracrine neurogenic information and modulating neuronal activity within the myenteric plexus (Boesmans et al., 2013b;Gulbransen and Sharkey, 2009). Although Ca 21 imaging techniques convincingly prove the activation of glial cells, the functional consequences of intracellular Ca 21 rises in EGCs have yet to be established. Purinergic enteric neuron-to-glia signaling is believed to be mediated via the purinergic receptor subtype P 2 Y 1 (Boesmans et al., 2013a;Gomes et al., 2009;Gulbransen et al., 2012) but it is currently unclear whether this receptor is not expressed by type-II and type-III EGCs. Irrespective of the molecular mechanisms, the lack of response to ATP by EGC subtypes II and III may reflect developmental immaturity of the latter cell populations. This idea is consistent with the reduced expression of GFAP in these subpopulations, a marker thought to be expressed generally by mature EGCs (Jessen and Mirsky, 1980). Further lineage tracing studies will be necessary to resolve these issues.
In conclusion, by combining genetic single-cell labeling, in vivo fate mapping and live Ca 21 imaging we provide evidence that EGCs can not only be discriminated on the basis of morphology and location but also display distinct molecular signatures and functional characteristics. Our studies set the basis for further exploring the heterogeneity among EGCs, which is a prerequisite to fully understand their role in gastrointestinal function and disease.

Author Contributions
Experiments, interpretation of data, drafting and editing of the manuscript were done by WB and RL. Study design, supervision, interpretation of data, editing of manuscript and acquisition of funding was performed by PVB and VP.