Flow Cytometric Analysis and Sorting of Enteric Nervous System Cells: From Human to Mouse, an Optimized Protocol

1. Introduction

The enteric nervous system (ENS) is a finely tuned network of neurons and glial cells, organized in enteric ganglia within the gastrointestinal (GI) tract. The ENS is located in the myenteric and submucosal plexuses. However, enteric glia are also located in the mucosa and within muscles [1]. The myenteric plexus is found between the circular and longitudinal muscle layers, and the submucosal plexus is situated between the muscle and mucosa [2,3]. The myenteric plexus predominantly regulates muscle contraction, while the submucosal plexus governs fluid secretion and absorption, blood flow, and senses stimuli from the epithelium and lumen, to maintain bowel function [4]. The critical role of the ENS is underscored by the array of enteric neuropathies arising from congenital abnormalities in its development, such as Hirschsprung disease (HSCR) and chronic intestinal pseudo-obstruction [5]. Precise mapping of the ENS and its embryonic development becomes necessary to understand these congenital conditions. However, isolating the ENS poses significant challenges, due to its deep embedding in the GI tract, making it difficult to extract without damaging the delicate ganglionic network [6]. Furthermore, enteric neurons comprise a small proportion of intestinal cells (about 0.01% of the total human colon [6]), and thus ENS enrichment becomes technically demanding.

Recently, we developed a reliable method for isolating human ENS cells, which combined an efficient dissociation protocol with an optimized fluorescent activated cell sorting (FACS) staining panel, yielding high-quality ENS cells suitable for downstream analysis [7]. Although using human tissue provides the most directly applicable insights into embryonic development and disease mechanisms, it is not always feasible to obtain. Often access to human biopsies is limited, especially from embryonic stages, and therefore, using alternative in vivo models, such as mice, allows for a more accessible and consistent source of material. Several mouse models have also been established for studying the ENS as well as enteric neuropathies [8,9,10], contributing to our understanding of genetic complexity and developmental origins of ENS malformations. Previous studies [11] have tried to map the cellular composition of the mouse ENS, using transgenic mice in which ENS cells are labelled. Although this approach allowed the characterization of numerous neuronal subsets, recapitulating ENS heterogeneity, creating transgenic mice is technically challenging, time-consuming and expensive. To overcome these issues, a previous study isolated and cultured the myenteric and submucosal plexuses from mouse, to obtain co-cultures of ENS cells with other non-neuronal intestinal cell types [12]. However, this method was limited in selectively isolating ENS cells or characterizing specific neuronal or glial cell subtypes.

In the current study, we developed a highly specific FACS-based method to efficiently isolate murine ENS cells, enabling detailed analysis and characterization of these cells. We found that CD56 (neural cell adhesion molecule 1 (NCAM1) and CD24 (HAS; a neuronal differentiation marker discriminating neurons from glia), are able to identify the ENS cluster, and enrich for neurons vs glial cells, similarly to what we have described for humans [7]. The best output was achieved using a Liberase-based protocol, allowing isolation of highly pure and viable enteric neurons and glial cells from murine colonic material, suitable for downstream applications.

2. Results

2.1. Mouse ENS Cells Are Successfully Isolated with the Human Protocol, but with Low Viability

Based on our previously developed FACS protocol, which allowed for isolation and subclustering of human ENS cells [7], we started by exploring its suitability for mouse tissue. Following dissociation of mouse colons with Collagenase II/Dispase, the single cell suspension obtained was stained with the same cell surface markers previously used for human cells, only using mouse-reactive antibodies (Table 1) [7]. CD45 (immune-hematopoietic cell marker) and CD31 (immune-hematopoietic and endothelial cell marker) were added as exclusion markers. Cells positive for these markers were referred to as Lin⁺. Red blood cells (RBC) were removed by RBC lysis and DAPI was used to discriminate live from dead cells on the same FACS Aria channel used for the Lin markers (i.e. BV421/DAPI channel, Table 1). Based on our previous results [7], CD56, CD24 and CD90 (Thy-1) were used as positive markers to identify and eventually subdivide the ENS cluster. In parallel, we also performed the assay on human colon biopsies, with the same antibodies previously used [7], and compared the outcome to the murine colon digestion.

The gating strategy applied was the same as the one used for human tissue [7]. Briefly, following selection of the region of interest (FSC-A vs SSC-A) with preliminary exclusion of multicellular clusters as well as very small particles (Figure 1A), a large single-cell gate was applied (FSC-A vs FSC-W) and live/Lin^- events were selected (Figure 1A,B). CD56 expression was then examined to identify the ENS cluster. This marker was selected based on the evidence that it has the highest specificity for ENS cells in human tissue [7]. In mouse, a putative ENS cluster was detected with high CD56 expression (CD56^high) (Figure 1A), which matched well the human ENS cluster (Figure 1B). A lower CD56 expression (CD56^low) cluster was also identified, but we noticed that it was more pronounced in mouse than in human samples (Figure 1A,B). Next, we examined co-expression of CD56 with CD24 and CD90 (Figure 1C,D). The CD56^high cluster was positive for both CD90 and CD24 in mice, as well as in humans. This evidence supported the hypothesis that CD56^high corresponds to the ENS cluster in both systems. In contrast, the CD56^low cluster was mostly CD24 negative, suggesting that it is probably not composed by ENS cells. Using CD56, CD24 and CD90 in mouse, showed that CD56^high cells were sub-clustered in CD24 right (CD24R) and CD24 left (CD24L), and CD90 left (CD90L) and CD90 right (CD90R), respectively (Figure 1E). Human ENS cells showed the expected CD24^low, CD24^high, CD24^TIP, as well as CD90L and CD90R subclusters (Figure 1F). Since CD24 levels were higher in mouse than in human ENS, we preferred to use the definition of CD24R and CD24L subclusters to differentiate them from the CD24^low and CD24^high clusters defined in the human system (Figure 1F). Interestingly, comparison of the CD24L and CD24R, to, respectively, the CD90L and CD90R subclusters, did not show correspondence as found in the human ENS (Figure 1E,F). Interestingly, a CD24^TIP was also not immediately visible in the mouse ENS, whereas human samples showed, as expected [7], a small but clearly separate CD24^TIP subcluster, which roughly corresponded to the upper part of CD90R (Figure 1E,F).

In human colon preparations, we have showed that the ENS cluster is made by both cells and non-nucleated debris, largely overlapping in scattering levels and staining pattern [7]. Moreover, glial cells seemed to always carry neuronal remnants, and vice-versa, inevitably leading to a mixed neuronal/glial staining pattern. With these premises, we investigated the presence of real ENS cells vs debris in the mouse samples. Nuclear staining performed with DAPI on sorted and formalin fixed live/Lin^- fractions (Figure 1h), showed that about 35% of total ENS events belonging to the live fraction, were nucleated. Interestingly, this value overlapped with the percentage found in human colons (31±8%). However, further comparison of the percentage of live/Lin^- events in human and mouse samples, showed that they were significantly lower in mice (Figure 1G). The mean percentage of ENS events, relative to the total number of live/Lin^- cells, was also lower when compared with the human system, although this difference alone was not statistically significant. However, when comparing the number of ENS events in the live gate (i.e., the sortable fraction) relatively to the total raw number of events, the former were significantly lower in mouse than in human samples (Figure 1G). These results, indicated the need to optimize the dissociation protocol for better recovery of the murine ENS.

2.2. New Dissociation Protocols for Isolation of Murine ENS Cells

2.2.1. New Dissociation Protocols and Gating Strategy

To improve the dissociation protocol and increase the percentage of viable neural cells isolated, we adjusted its composition and duration. Replacing Collagenase II with Collagenase I was the first adjustment made, as different types of Collagenase breakdown different types of collagen. By replacing the type of Collagenase and decreasing its concentration, while increasing the concentration of Dispase, we aimed to improve the dissociation efficiency, as described before [13]. In parallel, we tested a Liberase-based protocol [14], that is a mixture of Collagenase I and II, where Thermolysin replaces Dispase. We also considered that ENS cells are fragile relatively to other intestinal cells types, and therefore decreased the dissociation time from 60 to 30 minutes (‘). This change was associated with the fact that at 60’, the samples seemed to be over-digested. In addition, the RBC lysis step was removed to further shorten the protocol and reduce unnecessary stress. As an alternative, an anti-TER119 antibody was added to the Lin antibodies, to stain for and gate out all red blood cells. Prior to any comparison, though, we decided to establish a new gating approach. This functioned as quality control for us, to better understand the reasons behind the different outputs potentially observed with different protocols, and examine the suitability of the alternative digestion approaches (Figure 2 and Figure S1). As a preliminary step, the threshold on FCS was lowered till the smallest DAPI⁺ cells were visible, to make sure all dead cells were included in the analysis (Figure S1a). Then, we first gated the CD56^high particles out of total (recorded) events, thus including all putative ENS particles that were present in suspension (Figure 2A). These raw events included live and dead ENS cells, debris, and multicellular aggregates. With this approach, we aimed to determine the total fraction of CD56^high particles extracted with each dissociation protocol, and identify to which region they corresponded (live, dead or debris) (Figure 2B). Eventual biases in ENS distribution introduced by each dissociation protocol, could be identified with this approach, e.g. we were able to visualize if missing ENS events in the live region were counterbalanced by an increase of dead cells or small debris. In other words, we could follow the viability and integrity of our putative ENS cells.

In parallel to this unbiased analysis, we followed a standard gating approach to refine our populations (Figure 2C). An initial gate was applied on FSC-A vs SSC-A to exclude the biggest clusters, but include all detectable debris. This was followed by an inclusive “single cell gating” on FSC-H vs FSC-W, as previously reported [7] and also applied in Figure 1. Notably, the “single cell gate” was calibrated not on all cells, but on the final ENS population, thus including singlets, but also multicellular aggregates until, approximately, the scattering level of triplets/quadruplets. This was necessary to prevent eventual loss of big and irregular shaped cells, as dissociated neurons are predicted to be. More stringent doublet exclusion gating, when necessary, was performed at a later stage based on DNA staining [7]. In addition, DAPI⁺ (dead) and Lin⁺ cells were gated separately, as opposed to what was done before, leading to a clear evaluation of Lin⁺ and dead cells. To this purpose, we plotted the DAPI/BV421 channel vs the free BV510 channel (Figure 2D). Due to a more red-shifted emission spectrum, DAPI showed stronger relative fluorescence in the BV510 than in the BV421 channel, when compared with the BV421-conjugated Lin antibodies, making possible a clear-cut separation of the two fluorochromes. Lin^- events were further gated in a DAPI/FSC-A plot and subdivided in three regions: live, dead and low FSC particles or debris. For the sake of clarity, Lin⁺ cells were also shown in parallel to visualize the presence of the expected Lin⁺ clusters (Figure 2E). Overall, this approach allowed to distinguish low viability from an excess of Lin⁺ cells, and also to draw cleaner live/dead gates. Finally, the live/Lin^- events were further analyzed with respect to their CD56, CD24, and CD90 staining pattern to detect the ENS cluster. All this information, including qualitative and quantitative elements, was taken into account for proper comparison of the protocols. Representative plots were obtained using a sample dissociated with Collagenase I/Dispase for 30’.

2.2.2. The Liberase-Based Protocol Increases Overall Viability and ENS Recovery from Mouse Colons

Once the strategy was defined, we compared the new Collagenase I/Dispase- and Liberase-based protocols. For a more linear data flow, we describe in parallel, FACS plots relative to Collagenase I/Dispase (Figure 3A–D) and Liberase samples (Figure 3E–H).

Following preliminary definition of Lin^- and live cells, we gated our putative ENS events (Figure 3A,E). The pattern was as expected, with both protocols showing a C56^high and a CD56^lowcluster. As a second step, we analyzed the expression of CD24 and CD90 (Figure 3B,C,F,G). CD24 was highly expressed by the CD56^high cluster and not by the CD56^low cluster. The subdivision in CD24R and CD24L was as previously observed, but the CD24R subcluster was less pronounced in the Liberase samples (Figure 3F vs 1B). With both protocols we could detect some ENS particles at very high CD56/CD24 levels, which we defined as CD56/CD24^TIP (Figure 3B,F). Although not as well separated as in the human ENS, a density plot confirmed the mouse CD56/CD24^TIP to be a small cluster with its own identity. Notably, although the CD24R cluster was smaller in the Liberase samples, the CD56/CD24^TIP was comparable with that of the Collagenase samples. Since we expected only the CD24^TIP to be enriched for neurons, we did not consider a lower amount of CD24R to be a major issue. CD90 was, as expected, 100% highly expressed by the CD56^high cluster (Figure 3C,G). Subdivision in CD90L and CD90R was hardly detectable with the CollagenaseI/Dispase protocol, but well visible with the Liberase protocol. Interestingly, with both protocols there was no overlap of CD24L and CD24R with CD90L and CD90R, respectively (Figure 3D,F), confirming what we observed with the human dissociation protocol (Figure 1E,F). Due to this, and considering that CD90 is not a neuronal selective marker and, also, that common anti-CD90 antibodies do not work with all mouse strains (Table 2), we decided to base our selection and sub-clustering of the ENS population in mouse, with only CD56 and CD24.

Full statistical analysis of the different populations and subpopulations, clearly showed that the 30’ Liberase protocol significantly increased the percentage of live/Lin^- cells, as well as the number of sortable ENS events in the live region (Figure 3I,J). Also, the total amount of raw and unbiased ENS particles extracted with the Liberase protocol, was significantly higher than with other protocols (Figure S1). Based on these results, we considered the 30’ Liberase protocol to be the most efficient for mouse ENS isolation.

2.3. Validation and Subdivision of the ENS Cluster

To validate the identity of the ENS cluster following dissociation with the Liberase protocol, we first performed live nuclear staining to discriminate real ENS cells from ENS debris. This step was essential to define the effective suitability of the dissociation protocol for ENS cell enrichment, as we wanted these cells to be suitable for downstream analysis, such as single cell RNA sequencing or reverse transcriptase quantitative PCR (RT-qPCR). For this purpose, we added the cell permeant nuclear dye Dye Cycle Green, to the standard staining panel. Due to this, CD24-BV605 was used instead of CD24-PE, to avoid spillover from the Dye Cycle Green to the PE channel. This set-up, allowed us to confirm that the ENS cluster obtained with the Liberase protocol was made of nucleated cells and non-nucleated particles of different sizes (Figure 4A and Figure S2a,b), as previously found with the human protocol. Notably, also the average fraction of nucleated ENS cells (34 ±11%), perfectly overlapped with the one obtained with the human protocol (Figure 1H). This indicates that the fraction of nucleated ENS events out of those gated in the live/Lin^- fraction, is not significantly influenced by the digestion protocol or the percentage of live cells. On the other hand, prolonged handling or even simple prolonged storage of ENS samples on ice, can significantly and selectively reduce ENS cells viability (Figure S2c,d) and thus, the percentage of nucleated ENS events in the live region. Interestingly, we also observed that the CD56/CD24^TIP, considered to be enriched for neurons [7], was more pronounced in the nucleated than in the non-nucleated fraction (Figure 4A–D).

To further validate the identity of the CD56^high population, selective neuronal and glial markers were used. In parallel, the CD56^low population was also analyzed. We started by using CD24 as a cell-surface/neuron-selective marker within the ENS cluster. Previous analysis (Figure 3E–H and Figure 4A) clearly showed that CD24 expression is associated with the ENS cluster. To better address this point, we made in depth statistical analysis of those samples (Figure 4D). All ENS particles (cells and debris) were confirmed to be 100% CD24⁺, vs. 8 % of CD56^low and 22% of CD56^- (Figure 4D). The presence of CD24⁺ cells in the CD56^- fraction was expected, as epithelial crypt bottom cells are CD24⁺. We then performed a more stringent validation of the ENS population using intracellular staining on formalin fixed cells, as previously reported [7]. Tubulin Beta 3 (TUBB3/Tuj1) was used as a neuronal selective marker, and SRY-Box Transcription factor 10 (SOX10) as a glial selective marker (Figure 4E,F) [11,15]. To discriminate nucleated from non-nucleated events, and single vs non-single cells, DAPI was added to the samples. As expected, all putative ENS cells (CD56^high) were positive for both markers (Figure 4F,G). On the other hand, CD56^low and CD56^- cells were negative for Sox10, showing the same background noise (around 10%), as expected when gating positive vs negative events in the context of a low staining index. Regarding Tubb3 expression, CD56^- cells were totally negative, while CD56^low showed a small fraction of cells with dim Tubb3 staining. Based on these results, CD56^high were confirmed to be ENS cells, whereas the staining pattern of CD56^low was not compatible with either glial or neuronal cells. As described for the human samples [7], we observed the presence of a mixed neuronal and glial staining pattern in all ENS cells, with all the markers tested. This confirmed that neuronal and glial cells cannot be extracted pure by any of the dissociation protocols tested by us until now, as they seem to always carry, respectively, glial or neuronal fragments attached.

For the sake of completeness, we have also analyzed the expression of these ENS markers among non-nucleated events. Interestingly, ENS debris already shown to be CD24⁺ (Figure 4D), were also totally positive for Tubb3, indicating that all of them were pure neuronal or mixed neuronal/glial fragments, mostly resulting from broken terminations. On the other hand, Sox10 positivity was not universal, indicating the presence of debris that only contain neuronal fragments but no glial ones at all (or not enough to stain positive).

Additional validation of the ENS cluster was also performed by RT-qPCR using neuronal and glial markers (Elavl4; Tubb3; Sox10; Cd56/Ncam). Our results showed that the CD56^high population is indeed enriched for Cd56 as expected, but also for neuronal and glial markers (Figure 4G).

2.4. CD56/CD24^TIP Is Enriched in Neurons

In an attempt to subdivide neurons from glial cells, we tested the hypothesis that murine neurons as human neurons, should have higher levels of CD24 and be enriched in the CD56/CD24^TIP. This hypothesis was tested, as a first approach, by staining for ELAVL4, a selective intracellular neuronal marker concentrated in the neuronal soma and rapidly decreasing to the terminations [7]. As an alternative to the BD FACS Aria III platform, we used the new flow cytometer BD S8 FACS Discover, which combines spectral flow cytometry with cell view technology, providing spatial and morphological insights on the events under analysis. Since imaging with FACS Discover is only possible on three blue-laser dependent channels (green, red, and far red), DAPI was not suitable as a nuclear probe, but was still used as a reference nuclear dye. Also, being the green channel dedicated to ELAVL4, we decided to visualize the nuclei using the red emitting Doxorubicin. The latter binds DNA and its excitation and emission spectra, with peaks at 480 and 590 nm respectively, matches well the second imaging channel (Table 2), with little spillover to the first (green) channel. Doxorubicin was further calibrated to minimize its green spillover (Table 2). In addition, to minimize spectral interference of CD24 with the red imaging channel, a near-infrared version of the anti-CD24 antibody (BV786 conjugate) was used.

The gating strategy on FACS Discover is presented in Figure 5A–D. Briefly, following a preliminary gate on FSC vs SSC, and a standard, preliminary doublet exclusion strategy, CD56^high and CD56^- events were selected (Figure 5A). These events were then divided in nucleated and non-nucleated, by means of DAPI staining (Figure 5A). The nucleated fraction was further subdivided to have single G1/G0 cells, i.e. the region of interest for post-mitotic ENS cells, and a fraction containing a few proliferating and non-single cells (Figure 5B). ENS cells were further subdivided in TIP vs non-TIP region (Figure 5C). TIP cells showed high ELAVL4 staining (Figure 5D) and also high SSC levels, as expected with potential neurons. Statistical analysis of ELAVL4 median fluorescent intensity (MFI) performed vs a control antibody, showed significantly higher ELAVL4 levels in the TIP (Figure 5E), followed by the rest of the ENS cells. Non-ENS cells had, as expected, the lowest FMI level.

Although statistical analysis was only performed on single G1/G0 cells, to correct for different cell sizes, we used a ratio approach vs control Ab, instead of the most common subtraction of control Ab noise levels. Imaging of single G1/G0 cells clearly showed the prevalence of ELAVL4 bright cellular shapes compatible with neurons in the TIP, and ELAVL4 dim cells in the ENS/non TIP region and non-ENS cells (Figure 5I). The region near the TIP (border between TIP and non TIP) was characterized by intermediate patterns, among them some that resembled ELAVL4^- glial cells carrying neuronal fragments (ELAVL4⁺) on their putative proximal part. Imaging with Doxorubicin confirmed that all single cells selected by DAPI, had a single red nucleus. On the other hand, the non-nucleated fraction gated with DAPI, was confirmed to be devoid of Doxorubicin-positive nuclear-like structures. Interestingly, the CD56/CD24^TIP of the non-nucleated fraction was enriched with structures resembling broken neurons containing part of the soma and/or the proximal part of their terminations. Moving to the doublet/triplet regions, it was possible to observe different combinations, among them some ELAVL4^- glial cells attached to ELAVL4⁺ neurons. These results confirmed the selectivity of the staining for putative neurons.

3. Discussion

In this study, our aim was to develop an optimized protocol for enrichment and sorting of mouse ENS cells. Our initial approach was to apply the human protocol that we previously developed, based on the use of CD56, CD90 and CD24 markers [7]. Although these markers appeared functional in the identification of putative mouse ENS cells, which were qualitatively similar to the human ones, there were key differences observed. Specifically, CD90 appeared to be non-essential, highlighting the unique features of each specie, which require tailored approaches. In addition, the viability of the ENS cluster was insufficient for downstream applications. A series of adjustments were, therefore, made to the dissociation protocol, including testing different enzyme combinations and incubation times. Alternating Collagenase II with Collagenase I improved the output, however not significantly. On the other hand, Liberase increased significantly the number of ENS events. However, this was only the case when the dissociation was performed for 30’. Extending this process to 60‘, led to over-digestion of the tissue, resulting in no ENS yield. Based on these results, we were able to show the importance of fine-tuning protocols to ensure the best outcomes for cell sorting, particularly in compartments as complex as the ENS.

Further characterization of the ENS cluster obtained, was performed to validate the neural nature of these cells. With this purpose, selective markers for neurons (Cd24, Tubb3 and Elavl4) and glial cells (Sox10) were used. Interestingly, while RT-qPCRs on sorted samples, further confirmed the ENS identity of these cells, the staining pattern obtained was characterized by co-expression of glial and neuronal markers on the same cells, mirroring our findings in the human ENS [7]. This observation points to the intricate relationship between enteric neurons and glial cells, which makes it difficult to extract them as pure populations, at least with the digestion protocols tested until now, since they inevitably carry membrane remnants of each other. As a consequence, defining if a single cell is a neuron or a glial cell carrying proximal or distal neuronal terminations is quite challenging. Nonetheless, by using advanced analysis techniques, such as imaging flow cytometry with the S8 FACS discover, we were able to show that our gating strategy with the selection of CD56/CD24^TIP can indeed enrich for neurons vs glial cells. Taken together, we have developed a FACS–based protocol for isolation of viable mouse ENS cells, which also allows preliminary discrimination of neurons from glia.

4. Materials and Methods

4.1. Animals and Intestinal Isolation

For this study, intestines were removed from C57/BL6 or FVB mice and separated into small intestine and colon. The later was placed in a petri dish with cold 1X PBS. Fat was removed and colon was cut open with a blunt end scissor. The tissue was cleared from faeces and placed on a pre-cooled chopping board and gently scraped with a glass slide to remove debris on the luminal side. Finally, the colon was cut into 1mm pieces with a razorblade.

4.2. Dissociation of Mouse Colon Tissue

The pieces were transferred to a gentleMACS C-tube (Miltenyi Biotec, 130-093-237) with 5 ml/mouse of the selected dissociation medium (Table 1) and placed in a gentleMACS Octo Dissociator (Miltenyi Biotec, 130-095-937). The tissue was dissociated for 30’ or 60’ at 37°C. After dissociation, the suspension was mixed with 5 ml Hanks' balanced salt solution (HBSS) (14175095, Thermo Fisher Scientific) containing 10% fetal bovine serum (FBS) (FBS-12A, Capricorn Scientific), and gently triturated with a 19G needle and syringe. If needle occlusion was detected, the occluded piece of tissue was removed with a wet wipe and discarded. Then the suspension was filtered through a 70 μm cell strainer (352350, Falcon®) and mixed with 10 ml HBSS-10% FBS. Subsequently, the suspension was centrifuged at 4°C, 400 RCF for 9’, re-suspended in 2 ml HBSS-10% FBS.

4.3. Intracellular and Extracellular Staining

Primary and secondary antibodies, and the corresponding dilutions or final concentrations are listed in Table 2. Cell suspension was first stained for the selected extracellular markers. The antibody mix was diluted in HBSS 8% FBS, and incubated on ice for 30‘. Cells were then washed twice using HBSS 2% FBS, followed by centrifugation and resuspension of the pellet in HBSS 8% FBS containing DAPI (1 μg/ml; D9542, Sigma-Aldrich) for dead cell exclusion during FACS analysis. For intracellular staining, live cells were incubated with a LIVE/DEAD violet fixable viability dye (130-109-816, Miltenyi Biotech), for 10’ at room temperature (RT). Then cells were fixed in 4% paraformaldehyde for 30’ at RT, washed twice with PBS, permeabilized and blocked with PBS-0.2% Triton X-100, 10% FBS 4h to O/N. For intracellular staining, both primary and secondary antibodies were diluted in PBS, 0.2% Triton X-100, 10% FBS. For conjugated antibodies, cells were stained on ice for 3h, washed twice with 1 ml PBS-2% FBS, centrifuged at 4°C, 540 RCF for 3’ and resuspended in 100 μl PBS. For unconjugated primary antibodies, cells were washed and stained for an extra 2 hours with secondary antibodies, then washed again 3 times.

When staining with secondary Fab monovalent antibodies, 0.25 µg of the primary mouse or rabbit antibody were mixed with 0.75 µg of the secondary Fab at a ratio 1:3, and incubated for 5’ at RT. Pre-dilutions from the antibody stocks were eventually used. The excess of secondary antibody was then blocked with 4 µl of mouse (015-000-120, Jacksonimmuno) or rabbit (NS01L-1ML, Sigma-Aldrich) serum, depending on the secondary antibody used. The mix was immediately diluted in PSS 10% FBS and used for staining, keeping the final concentration of the primary antibody as described in Table 2. Mouse and rabbit control IgGs were pre-incubated with Fab secondary antibodies, and then incubated with cells at the same concentrations used for the corresponding primary antibodies.

4.4. Fluorescence-Activated Cell Sorting

FACS analysis and sorting were performed with a BD FACS Aria III or BD S8 FACS Discover (BD Biosciences, New Jersey, USA) using a 100 micron nozzle for standard analysis and sorting. All the details about FACS Aria settings and reagents used are reported in Table 2. With FACS Discover, the standard available configuration was used, with 5 lasers and 78 detectors for fluorescence detection. Forward and side scattering were available on both blue and violet laser. Light loss was also available, to detect the loss of laser light produced by intercepted particles. Imaging was available on blue laser only, with 3 channels (green, red, and far red) and additional parameters, among them forward and side scattering, and light loss. Imaging does not allow compensations between fluorescent channels. Thus, for the sake of simplicity, we limited our FACS Discover usage to simple imaging and small antibody panels to minimize spillover between channels. Also, complex imaging parameters were not used to gate cells, keeping our gating structure close to the one used on FACS Aria III. Prior to analysis or sorting, the cell suspension was filtered through a 40 μm cell strainer (352340, Falcon®). Initially, samples were gated by size and granularity using a Side Scatter (SSC-A) versus a Forward Scatter (FSC-A) plot, to allow exclusion of the smallest debris and bigger cell clusters. Non-single cells were partially excluded using an FSC-H vs FSC-W gating. The latter gate was not applied stringently, to ensure inclusion of neurons with high FSC. Dead cells were excluded using DAPI, plotted against FSC-A. CD31 (endothelial cells), CD45 (immune cells), and TER119 (red blood cells) positive cells, were also gated out in the same channel as dead cells, to keep other channels free for additional markers. All unstained and compensation controls were applied as previously reported [7].

4.5. Gene Expression Analysis

Total RNA was extracted from sorted ENS cells using the RNeasy Micro Kit (QIAGEN, cat. no. 74034), following the manufacturer’s protocol. Concentration of isolated RNA was determined using the DeNovix DS-11 (DeNovix), and complementary DNA (cDNA) was prepared using the iScript™ cDNA Synthesis Kit (Bio-rad, cat. no. 1708890 and 1708891), following manufacturer’s instructions. For quantification of gene expression, RT-qPCR was performed using iTaq SYBR Green Supermix (Bio-rad, cat. no. 1725120). For normalization of gene expression levels, Gapdh and β-Actin were used as housekeeping genes and the 2^−∆∆Ct method was used for calculations. Primer sets used are listed in Table 3. Three independent experiments were performed to ensure robustness and reliability of the results.

4.6. Quantification and Statistical Analysis

Data was analysed using GraphPad Prism version 9 (Graphpad Software®, La Jolla, CA, USA). Data were tested for normality, using the Shapiro–Wilk test. The Mann–Whitney U test was used if data were not normally distributed. Differences were tested for statistical significance using either unpaired Student’s t-tests or one-way ANOVA. Data are presented as means with standard deviations. Significance thresholds (p-value) were set as follows: p < 0.05 (*), p < 0.01 (**), p < 0.001 (***), p < 0.0001 (****).