Traditional studies of the anatomy and pathology of the ENS have relied either on whole-mount tissue staining (Lebouvier et al., 2010b) or on slicing, often combined with hematoxylin and eosin (H and E) staining and followed by imaging of selected single thin sections with bright-field microscopy. This approach makes it unlikely to capture the intricate three-dimensional (3-D) architecture of the ENS characterized by two interconnected types of ganglia in the intestinal wall. The myenteric plexus (also known as plexus of Auerbach) is located in between the circular and longitudinal muscularis, two orthogonal layers of smooth muscle cells that are responsible for the motility of the bowel. Neurons of the myenteric plexus mainly regulate the motor function of the gut. The submucosal plexus is located in vicinity of the gut lumen. It mainly regulates secretory functions (Kulkarni et al., 2018; Schneider et al., 2019; Figure 1). Both Auerbach’s and Meissner’s plexus are composed of different neuronal subtypes as well as enteric glia that interact with one another and they target different effectors (Boesmans et al., 2013; Schneider et al., 2019). This layered organization is best revealed on transverse intestinal sections. On the other hand, longitudinal sections better reveal the interconnections between these ganglia, which line the intestine like a fishnet (Costa et al., 2000). Imaging this 3-D structure comprising the dispersed neurites, fibers, and structural components is challenging, but it is of considerable interest.

Abnormalities of the ENS have been linked to many pathologies of the digestive tract. One drastic example is Hirschsprung’s disease where regions of the large intestine (colon) lack ganglia. Without these nerve cells stimulating gut muscles to move contents through the colon, the contents can back up and cause constipation. More surprisingly, ENS abnormalities are linked to central neuropathologies, too. For example, the analysis of the ENS by routine colonoscopy biopsies was shown to offer a pre-mortem diagnostic tool of Parkinson’s disease (PD), and studying ENS alterations provides insight into the progression of motor and non-motor symptoms, too (Lebouvier et al., 2010b; Ohlsson and Englund, 2019; Schneider et al., 2019). In addition, ENS dysfunction has been linked to other central pathologies, like autism (Israelyan and Margolis, 2018; Heiss and Olofsson, 2019), underpinning the need and interest of a detailed morphological characterization of alterations in its architecture with modern imaging techniques.

Detailed 3-D observations have been rare and they mainly comprise very limited tissue volumes (see, however, Cirillo et al., 2013; Fung et al., 2021). Typical 3-D reconstructions use serial thin sections, immunolabeling and their sequential observation with confocal microscopy, followed by image processing and 3-D rendering. Also, immunofluorescence protocols already require many hours on thin slices and days for the incubation of such thick samples. An additional bottleneck comes from the speed-limit of confocal laser-scanning microscopy (CLSM), which is inherently slow for reconstructing such large volumes, because the image is acquired and built up point-wise. In addition, as the entire sample volume is illuminated but fluorescence only collected only from a single spot, confocal imaging rapidly photobleaches the entire volume during volume scans. Thus, this serial two-dimensional (2-D) approach is extremely time-consuming for larger samples like the ENS that measures 3 mm for a transverse section of mouse colon and more for biopsies.

Two-photon (2P) excitation fluorescence microscopy (Denk et al., 1990) has become the gold standard for imaging thick tissue samples (Denk and Svoboda, 1997) because the better penetration of IR excitation and collection of scattered fluorescence (Oheim et al., 2001). Hundreds of fluorophores have action cross-sections that make them amenable to 2P imaging (Ricard et al., 2018) and even tissue autofluorescence (AF) can be used to generate intrinsic contrast and provide context information [see, e.g., Zipfel et al., 2003]. 2P microscopy and AF detection have been used for imaging various parts of the gut, including the stomach (Rogart et al., 2008), the small intestine (Orzekowsky-Schroeder et al., 2011; Ricard et al., 2012), and the colon (Aggarwal et al., 2013). Yet, 2P scanning–as CLSM–scales with the number of tissue voxels imaged and still requires 6–8 s/frame (Ricard et al., 2012) and several minutes for imaging a single block of intestinal tissue.

Multi-spot 2P-imaging techniques parallelize either fluorescence excitation and/or pixel readout to overcome the low frame rate of single-spot scanning schemes (Bewersdorf et al., 2006). The most common of these are 2-P multi-spot scanning systems (Kim et al., 1999; Nielsen et al., 2001) and 2-P spinning-disks microscopes (Egner and Hell, 2000; Straub et al., 2000; Kobayashi et al., 2002). Both geometries have in common that they associate a planar non-linear fluorescence excitation with a “one-shot” readout on a fast camera detector. By design, these systems compromise on optical sectioning and image contrast, as cross talk between neighboring pinholes and image blurring by scattered fluorescence becomes an issue, even at moderate imaging depths [see, e.g., Shimozawa et al. (2013) for discussion].

An alternative consists in generating a thin sheet of light orthogonal to the microscope axis, either by cylindrical optics and/or scanning the beam in a plane. The excited fluorescence is collected perpendicular to the illuminated plane in a separate collection optical path and imaged onto a large-format camera. This configuration combines improved contrast and axial sectioning with short acquisition and it reduces phototoxicity because only the imaged plane is illuminated at a given time (Huisken et al., 2004). However, even in its 2-P excitation variant (Truong et al., 2011; Mahou et al., 2014; Zong et al., 2015), light-sheet microscopes suffer from a reduced z-resolution compared to spot-scanning systems.

In response to these challenges, we recently introduced a 2P microscope produces a planar non-linear excitation but that retains the classical inline geometry with a single objective lens (Rakotoson et al., 2019). Its optical sectioning, high frame rate and low photograph damage are achieved through spatially and temporally multiplexing 2P excitation for planar illumination. When using a high-numerical aperture (NA) low-magnification objective lens (Oheim et al., 2001) some 50 excitation spots are scanned into the focal plane, with minimally 12 ms needed for a homogenously lit full-frame acquisition (and more for dimmer samples). Our instrument combines the speed, field-of-view and ease of a 2P light-sheet microscope with the resolution and image contrast of a scanning 2P microscope. We previously demonstrated simultaneous dual-color 3-D imaging of brain organoids derived from human inducible pluripotent stem cells (hIPSCs) with 300 nm lateral and 1.5 μm axial resolution over large fields of view. We reached an imaging depth exceeding 200 μm at a speed 10–100-times faster than a single-spot scanning microscope (Rakotoson et al., 2019). Under realistic (low mW) excitation powers and signal levels typical for AF imaging or the detection of faint biological labels, two 900 × 900 pixel dual-color images were simultaneously captured in 480 ms. These characteristics make our microscope attractive for 3-D imaging of healthy and pathological thick intestinal samples.

One major drawback common to all parallelized 2-P imaging systems is that they work best with transparent samples or in superficial tissue layers. The reason is that a major advantage of single-spot 2-P imaging is sacrificed by using an imaging detector rather than a “bucket” detector just collecting photons-the possibility of making efficient use of multiply scattered fluorescence (Oheim et al., 2001). Therefore, recent efforts in improving and speeding up volume imaging have typically gone along with improvements in tissue clearing [reviewed in Richardson and Lichtman (2015), Ueda et al. (2020), Tian et al. (2021), and Weiss et al. (2021)].

A plethora of different optical clearing approaches have been proposed to meet the demand for 3-D volume imaging [see, e.g., Neckel et al. (2016), Bossolani et al. (2019), Yuan et al. (2021) for work in the ENS]. Each technique has its own strengths but still suffers from either low transparency, long incubation time, processing complexity, tissue deformation, or fluorescence quenching (or a combination of these), and a single solution that best satisfies all aspects has been missing. In the present work, we combine a new patent-pending fast, universal tissue-clearing (Delhomme et al., in preparation) and volumetric-imaging strategies for imaging the mouse small intestine. We demonstrate a highly contrasted 3-D reconstruction of enteric ganglia entirely based on the AF-based identification of neurons and glial cells of the ENS. Our manuscript is organized as follows: (i) we first describe the fast clearing approach of intestine sample compatible with the conservation of AF signals; (ii) characterizing AF spectra and intensities at a cellular and sub-cellular scale, we reliably identify on label-free images neurons and enteric glial cells in both Auerbach’s and Meissner’s plexus; (iii) we validate this AF-based identification of by confocal immunofluorescence. Finally, (iv) using our fast 2P spinning-disk microscope, we demonstrate volumetric label-free imaging of the ENS throughout the entire depth of the intestinal wall. We expect the here proposed workflow associating fast clearing and rapid 3-D imaging to be valuable for future biomedical applications, e.g., during diagnostic imaging.

All experiments were performed according to the French legislation and in compliance with the European Community Council Directive of November 24, 1986 (86/609/EEC) for the care and use of laboratory animals. The used protocols were approved by the local Ethics Committee. In total, this study used 12 C57BL/6 mice (Charles River) that were bred and housed in the local animal house on an inverted dark-light cycle with access to food and water ad libitum.

Due to the long incubation times in each protocol, all reagents used contained 0.02% NaN₃ as a preservative. Room temperature (RT) was 20–25°C.

In clinical neurogastroenterology, as well as in fundamental research, the ENS is mainly being studied using histological techniques that involve serial slicing, staining, and observation steps. As a consequence, on a typical transverse slice of the mouse intestine, it is very rare to find both Auerbach’s and Meissner’s plexus, and the connections in between them are lost (Figure 1A–left panel, inset).

An alternative is the “Swiss roll,” which increases the probability of finding both plexuses (Bialkowska et al., 2016; Figure 1A–left). The probability of finding both nervous plexuses within the same field of view increases dramatically (Figure 1A–zoom), however, despite its advantages, it is neither possible to observe the full 3-D architecture of the ENS, nor to study its connections with peripheral and central nerves.

Notwithstanding these limitations, we prepared a Swiss-roll ENS section and stained it with HES (Figure 1A–right panel)–an approach that can only be applied to thin slices–to provide a benchmark image against which we will have to hold our new method. On the other hand, in preliminary experiments, we observed a distinct tissue AF that produced sufficient contrast for identifying both plexuses (Figure 1B). We therefore set out to study if this label-free technique, with appropriate 3-D imaging techniques, would be amenable to a full volumetric reconstruction of intact (i.e., non-sliced) intestinal samples.

In intact intestinal samples the imaging depth is limited by the scattering of both excitation light and the generated fluorescence by tissue components—less by absorption that occurs mainly through pigments. Clearing techniques enhance tissue transparency and/or homogenize its RI [see, e.g., Avilov (2021), Richardson et al. (2021), Tian et al. (2021), and Weiss et al. (2021) for recent reviews]. Together with advances in 3-D microscopy, clearing techniques have permitted detailed volumetric reconstructions of large tissue volumes, also of the intestine (Liu et al., 2011; Bossolani et al., 2019; Graham et al., 2020).

To significantly speed up the workflow of volumetric tissue imaging, we developed BRG5, an ultra-fast, universal, non-toxic clearing protocol, based on a high-RI aqueous solution that was successfully tested on various organs (Delhomme et al., in preparation). Intrinsic imaging requires the clearing to have a minimally attenuating effect on the already faint AF signal. We compared two protocols, one of which involving a depigmentation step of 45 min. Using intact intestinal samples of 3 × 3 × 10 mm, we compared the efficacy of the clearing on depigmented and non-depigmented tissue. In either case, the whole intestine became rapidly transparent upon treatment with BRG5, and the depigmentation had little if any effect on the final transparency (Figure 2A). We substantiated this observation by the quantification of the evolution with time of Michelson’s contrast C_M (visibility) of a black and white stripe pattern placed underneath intestinal samples (Figure 2B). In either condition, a plateau was reached in less than 15 min and the final contrast attained was undistinguishable (71 vs. 73%; Figure 2C–right panel). The only parameter that was significantly different was the half-time for clearing (2 min 36 s for depigmented vs. 5 min 18 s for non-depigmented samples; Figure 2C–left panel). However, this advantage seems negligible in view of the extra 45 min added when using the depigmentation process. In addition, AF images of Auerbach’s plexus acquired upon 405-nm excitation revealed a stronger signal and higher contrast without depigmentation (Figure 2D). We conclude that depigmentation of the intestine is dispensable and that clearing of the whole intestinal wall can be performed in less than 15 min without destroying the AF signal.

We next confirmed that the AF structures were indeed neural cells by immunolabeling against HuC/D, a pan-neuronal marker (Mazzoni et al., 2021) and against S100β, an enteric-glia marker (Grundmann et al., 2019). As expected, we detected no specific labeling at the level of the muscularis longitudinalis (Figure 3–first line). Within Auerbach’s plexus (Figure 3–second row), neuronal cell bodies were located in the structures previously identified as ganglia on the AF images. Neurons exhibited strongly AF granules in their cell bodies. The colocalization between AF signals showing small cell bodies without granules and S100β immunolabeling allowed us to unambiguously pinpoint enteric glia on AF images of Auerbach’s plexus. Likewise, HuC/D reliably labeled neurons in the ganglia identified as Meissner’s plexus (Figure 3–third row), albeit at lower density compared to Auerbach’s plexus. Also, neurons in Meissner’s plexus exhibited less AF granules than those found in Auerbach’s plexus, and enteric glia were not detectable at this level. Finally, crypts exhibited solely AF signal (Figure 3–bottom row).

Tissue AF is a complex multi-component signal characterized by broad fluorescence excitation and emission spectra. We investigated if spectral information could be used to further refine AF segmentation into different ENS structures. To this end, we selected four ROIs in Auerbach’s plexus: (i) neuronal cell bodies that exhibited a strong granular AF (red ROI); (ii) cytoplasmic regions of enteric glia (green); (iii) neuronal nuclei (blue) and, (iv) ROIs within the muscularis (yellow) (Figure 4A–left panel). Although we observed no significant spectral differences (Figure 4B–left) we could discriminate different structures based on their intensity that was strongest in neurons, followed by enteric glia, muscularis and finally nuclei, which showed the lowest AF and uniform signal. Moreover, neuronal but not glial cells bodies showed numerous brightly AF granules in their cytoplasm (Figure 4C–left).

At the level of Meissner’s plexus, we selected ROIs in 3 structures: neuronal cell bodies (red), neuronal nuclei (blue) and within the submucosa (purple) (Figure 4A–right). As previously observed in Auerbach’s plexus, these regions had indistinct AF emission spectra (Figure 4B–right) but they showed similar differences in terms of AF intensity (Figure 4C–right).

We conclude that the AF arising from the ENS does not exhibit a sufficient spectral variability upon 405-nm excitation. This is not necessarily a bad thing, because it allows the broad-band detection for maximizing AF collection. The strong intensity contrast observed between different structures of the ENS is sufficient and–together with morphological features–allows for a straightforward identification of cellular and sub-cellular structures.

We used a modified Nipkow-Petráň-type 2P microscope (Rakotoson et al., 2019) for deep AF imaging. Our instrument produces a 1.5-μm thin, planar non-linear excitation, and with the here used × 25/NA1.1 water objective and detector, we could acquire a 150 × 150 μm diffraction-limited AF image in 480 ms. With a spatial resolution close to the one obtained on a confocal microscope (330 nm laterally) and a temporal resolution at least 10 times faster, our microscope is particularly adapted to studying ENS volumes.

Figure 5 and Supplementary Videos 1, 2 show AF imaging across the entire cleared intestinal wall, starting from the external face down to the crypts. Muscularis longitudinalis (Figure 5A) and circularis (Figure 5C) were recognized clearly by their perpendicular arrangement of muscle fibers. In between, Auerbach’s plexus (Figure 5B) exhibited a strong AF contrast and ganglia are readily identified. As observed before upon 1P-excitation, neurons (arrow) but not glia (arrowhead) displayed intense AF granules in their cell bodies upon fs-pulsed 800 nm excitation. Likewise, Meissner’s plexus was characterized by ganglia of systematically fewer neurons compared to Auerbach’s plexus (Figure 5D–arrow). As a consequence of the non-linear excitation, a Second-harmonic Generation (SHG) signal can also be detected in the submucosa, arising from fibrous structures. This narrow signal around 400 nm can be separated from the broader AF by appropriate filtering (not shown). Crypts appeared as contrasted structures with sparse AF granules that have earlier be identified as zymogen-rich granules in Paneth cells (Ricard et al., 2012; Figure 5E). The orthogonal (xz) reconstruction (virtual transversal slice) illustrates the cross-sectional reconstruction of the entire intestinal wall, demonstrating the force of label-free imaging in rapidly cleared intestinal tissue (Figure 5F). Of note, using the OASIS 2P microscope, we were able to image AF across 140 μm of the entire mouse intestinal wall (1 μm z-step, from the serosa to the crypts) in less than 90 s, compared to 15 min–60-fold longer–than on a confocal laser scanning microscope (see Supplementary Video 3 for a direct comparison).

All the work was carried out exclusively on small intestine. However, our technique performs very well on colon (Supplementary Video 4), which is particularly interesting for the analysis of colon biopsies for clinical research and diagnosis.