Distribution pattern and molecular signature of cholinergic tuft cells in human gastro-intestinal and pancreatic-biliary tract

Despite considerable recent insight into the molecular phenotypes and type 2 innate immune functions of tuft cells in rodents, there is sparse knowledge about the region-specific presence and molecular phenotypes of tuft cells in the human digestive tract. Here, we traced cholinergic tuft cells throughout the human alimentary tract with immunohistochemistry and deciphered their region-specific distribution and biomolecule coexistence patterns. While absent from the human stomach, cholinergic tuft cells localized to villi and crypts in the small and large intestines. In the biliary tract, they were present in the epithelium of extra-hepatic peribiliary glands, but not observed in the epithelia of the gall bladder and the common duct of the biliary tract. In the pancreas, solitary cholinergic tuft cells were frequently observed in the epithelia of small and medium-size intra- and inter-lobular ducts, while they were absent from acinar cells and from the main pancreatic duct. Double immunofluorescence revealed co-expression of choline acetyltransferase with structural (cytokeratin 18, villin, advillin) tuft cell markers and eicosanoid signaling (cyclooxygenase 1, hematopoietic prostaglandin D synthase, 5-lipoxygenase activating protein) biomolecules. Our results indicate that region-specific cholinergic signaling of tuft cells plays a role in mucosal immunity in health and disease, especially in infection and cancer.


Segment-specific distribution and shape of cholinergic tuft cells in the human gastro-intestinal
tract mucosal epithelia. We first analyzed by single brightfield immunohistochemistry the presence and distribution of non-neuronal cholinergic cells in the normal human gastro-intestinal tract, using the goat-anti-ChAT antiserum as marker. In all our stomach samples, we observed ChAT-immunoreactive (ir) nerve fibers innervating the mucosal glands (Fig. 1A). Also, neuronal cell bodies in the myenteric plexus were ChAT-ir (data not shown). In contrast, ChAT-ir epithelial cells were not present. In the duodenum (Fig. 1B) and in the jejunum (Fig. 1C), ChAT-immunoreactivity was seen in solitary epithelial cells scattered along the crypt-villus units (cvu). Single ChAT-ir cells were also present in the mucosal outflow tracts of duodenal Brunner´s glands, but were absent from deeper submucosal glandular epithelium (data not shown). In the ileum, ChAT-ir cells were predominantly found in the lower parts of the villi and in the crypt epithelium, but were largely absent from upper villus parts (Fig. 1D). Similarly, ChAT-ir epithelial cells were restricted to about the lower third of the colonic crypts (Fig. 1E). Such distribution patterns were true for all samples analyzed. ChAT-ir cells represented a rare subpopulation in the mucosal lining of the gut, with 1.6 ± 0.02 cells per cvu in duodenum, 1.5 ± 0.38 cells per cvu in the jejunum, 1.1 ± 0.54 cells per cvu in the ileum, and 1.1 ± 0.15 cells per cvu in the colon (n = 50-70 cvu analyzed per specimen).
ChAT-ir cell morphology was similar in all divisions of the intestine. In villus parts they appeared flat-oval with long and slim basal and apical protrusions, and spanned the entire epithelium (Fig. 1F). Mainly in crypts, protrusions were short and cells more flask-shaped (Fig. 1G). The cholinergic phenotype of these cells was restricted to the expression of ChAT, since immunoreactivity for the vesicular acetylcholine transporter, VAChT, was absent (Fig. 1H).

Distribution of cholinergic tuft cells in human extrahepatic bile duct and in
the pancreatic ductal system. In mice, numerous ChAT-ir cells reside in the epithelial lining of the gall bladder, the cystic duct, the common bile duct, and the papilla of Vater 21 . In all our human samples, however, ChAT-ir cells were absent from the simple columnar gall bladder epithelium ( Fig. 2A). Instead, solitary, slim or flask-shaped ChAT-ir cells were present in the extra-hepatic peribiliary gland columnar (Fig. 2B,C) and cuboidal (Fig. 2D) epithelia, but were absent from the common bile duct (Fig. 2E) Again, this distribution pattern was true for all specimens analyzed.
In the exocrine pancreas, solitary ChAT-ir cells lined cuboidal small- (Fig. 2F) and large-sized ( Fig. 2G) intralobular, and columnar interlobular (Fig. 2H,I) ducts, but were absent from acinar cells, from the common pancreatic duct, and from the islets of Langerhans, the endocrine part of the pancreas (data not shown). Also, ChAT-ir cells in bile and pancreatic ducts did not display VAChT expression in all specimens analyzed (data not shown).
Human cholinergic tuft cells exhibit region-specific combinatorial expression patterns of structural and biomolecule tuft cell markers. Having established the distinct presence of ChAT-ir epithelial cells in the human intestinal, in the extrahepatic biliary and in the exocrine pancreatic systems, we asked whether these cells, like in rodents, represent tuft cells. Murine tuft cells are discriminated from other cells in the respective epithelia by their strong expression of certain structural proteins, e.g. cytokeratin 18 (CK18) and villin. Both in rat 6 and in mouse 21 CK18 is expressed by a major subpopulation of mucosal epithelial cells, and labeling is more intense in tuft cells compared to non-tuft cells. In our samples taken from the human intestine (Fig. 3A), accessory bile ducts (Fig. 3B), and pancreatic ducts (Fig. 3C), CK18 antibodies strongly labeled the entire epithelia. Double immunofluorescent analysis revealed an overlap of ChAT-ir cells with the strongly labeled CK18 cells ( Fig. 3A-Ca-c).
Villin, an actin filament crosslinking protein, is highly expressed in the apex of tuft cells, including the apical brush border 4,14 . In the human intestine, villin expression was found on the luminal side of many, if not all cell types along the one-layered epithelium, exemplified here for the small intestine (Fig. 3D). Solitary cells displaying a strong apical accumulation of villin-ir were also ChAT-ir (Fig. 3Da-c). In both, peribiliary glands and pancreatic ducts, villin-ir was found restricted to solitary cells (Fig. 3E,F). In addition to the apical brush border, also the lateral cell aspects were stained. All these villin-ir cells were co-positive for ChAT-ir.
Advillin is another member of the gelsolin/villin family, and structurally related to villin 41 . In mice, advillin is expressed by many sensory and autonomic neurons 42 , but also found in taste cells in lingual taste buds 43 . In the human intestine, advillin-ir is believed to be present in endocrine cells (www.proteinatlas.org). We observed that advillin expression was not pan-epithelial, when compared to villin, but restricted to the solitary ChAT-ir cells in the human intestine (Fig. 3G). Likewise, advillin staining in the human peribiliary glands (Fig. 3H), and pancreatic duct epithelia (Fig. 3I) was found confined to ChAT-expressing tuft cells.
The serine/threonine-protein kinase, doublecortin-like kinase 1 (Dclk1), is an established tuft cell marker in the mouse alimentary tract 7,44 , but recently was described to be restricted to a subset of enteroendocrine cells in the human intestine 45 . Using the same antiserum, we could confirm that DCLK1 is expressed by ChAT-ir tuft cells in the mouse gastro-intestinal system (data not shown). In our human gut samples, however, DCLK1-ir did not co-localize with ChAT-ir. Instead, DCLK1-ir co-localized with chromogranin A (CGA)-ir enteroendocrine cells. This staining, however, was interpreted to be non-specific since it could not be blocked when pre-incubating the primary DCLK1 antiserum with an excess of the peptide used for immunization (data not shown). www.nature.com/scientificreports www.nature.com/scientificreports/ Tuft cells in the mouse intestinal epithelium express the key enzymes needed for the production of prostaglandin D2 (PGD2), namely COX-1 and HPGDS 5,14,21 . In our study, COX1-ir was also present in solitary slim cells in our samples from the human small intestine (Fig. 4A), as well as in solitary cells in the epithelia of human peribiliary glands (Fig. 4B) and pancreatic ducts (Fig. 4C). All these cells co-labeled for ChAT-ir ( Fig. 4A-Ca-c).  www.nature.com/scientificreports www.nature.com/scientificreports/ Immunohistochemistry for HPGDS resulted in an equivalent staining pattern in small intestine (Fig. 4D), peribiliary glands (Fig. 4E), and pancreatic ducts (Fig. 4F). Again, HPGDS-positive cells were co-ir for ChAT ( Fig. 4D-Fa-c). Lipocalin-type PGDS-ir was not detectable in our human samples (data not shown). Only recently, mouse airway and intestinal brush cells were identified as a source of leukotriene production 5,46 . Using antibodies against the 5-lipoxygenase-activating protein, FLAP, we detected solitary cells with a tuft cell morphology in the mucosal linings of the human intestine (Fig. 4G), and in pancreatic ducts (Fig. 4I). Within the peribiliary gland epithelium FLAP-ir cells did not localize to the epithelium, but were flat-associated to the basal membrane (Fig. 4H). Co-staining of FLAP with villin revealed that FLAP-ir cells represent tuft cells in the small intestine (Fig. 4Ga-c), and in pancreatic ducts (Fig. 4Ia-c), but staining was mutually exclusive in peribiliary glands (Fig. 4Ha-c).
cholinergic tuft cells are separate from the enteroendocrine cell population. Finally, we asked if also enteroendocrine cells are a source of ChAT expression, as previously suggested 14 . However, we found that CGA, a marker protein of the gut enteroendocrine cell lineage 47 , was absent from ChAT-ir cells in the intestinal (Fig. 5A-D), accessory bile duct (Fig. 5E-H), and pancreatic duct (Fig. 5I-L) epithelia.

Discussion
The present study provides the first characterization of the cholinergic tuft cell system in the human gastro-entero-pancreatic and extrahepatic biliary tract. Immunohistochemistry for the cholinergic marker enzyme, ChAT, revealed that ChAT not only serves as a reliable tuft cell marker throughout the human entero-pancreatic and extrahepatic biliary systems, but also highlights the central role ACh may play in the diverse functions that recently have been assigned to tuft cells. Co-expression of ChAT with additional tuft cell biomolecule marker (COX1, HPGDS, FLAP) provides a basis for the presumed immune-regulatory role these cells play in health and disease.
The distribution patterns of ChAT-positive tuft cells in the human small and large intestines as demonstrated here essentially correspond to that reported for rodent intestines 21 . Hence, solitary ChAT-positive tuft cells reside to both, crypts and villi of the small and large intestine. A restriction of cholinergic tuft cells to the lower villus and crypt compartments in human ileum and colon is suggestive of a faster turnover of these cells in these regions as compared to duodenum and jejunum. ChAT-positive tuft cells were not present in our samples from human stomach. This is in contrast to the reported abundance of ChAT-positive tuft cells in the columnar epithelium at the squamo-columnar junction, and in other parts of the gastric mucosa of rodents 21,48,49 . The absence of a limiting ridge region in the human stomach may explain the lack of a bona fide cholinergic tuft cell population in the human gastric mucosa, albeit human gastric epithelial cells exhibiting immunohistochemical staining for DCLK1 or COX2 in the context of inflammation, hyperplasia, and cancerogenesis have been reported 31,48,49 . Thus, the existence of a human gastric tuft cell subpopulation that lacks a cholinergic phenotype has to be considered. Also, our observations of a presence (small and large intestine), or absence (stomach) of cholinergic tuft cells may not account for the entire organ, and sub-regional variations may exist.
A cholinergic tuft cell phenotype has not been reported yet for the human biliary tract, in spite of ultrastructural evidence that tuft cells are present 39 . We found ChAT immunoreactive tuft cells scattered throughout extra-hepatic peribiliary glands, while they were absent from the gall bladder and common bile duct epithelium. Such a rather limited distribution pattern is in contrast to rodents, where numerous cholinergic tuft cells are present throughout the gall bladder, especially in its neck region, and in the main epithelium of the extra-hepatic bile duct system 21 . Hence, the non-existence of cholinergic tuft cells in our samples from human gall bladder and common bile duct does not exclude a focal presence in distinct microcompartments of these organs. To determine the specific roles of cholinergic tuft cell signaling in health and disease at the particular sites of the human extra-hepatic biliary tree presented in our report warrant further functional and clinical investigations, especially in the context of a potential impact of pancreatic adenocarcinoma on the local tuft cell occurrence (see below).
In the human pancreas, we frequently observed solitary cholinergic tuft cells residing in intra-and interlobular ducts, which extends observations by others 40 . Cholinergic tuft cells were absent from acini, and from the main pancreatic duct. Again, such a distribution pattern seems to be a specialty for humans, since in rats a vast number of tuft cells is present in the main pancreatic duct 11 , while this cell type is completely absent from the normal mouse pancreas 40 . Recently, presence of DCLK1 in the main pancreatic duct of the mouse has been reported 50 , and increased expression of DCLK1 was a feature of human pancreatic adenocarcinoma 51,52 . In addition to being a tuft cell marker in rodents 7 , DCLK1 marks certain tumor-initiating cancer cells 44,53 . DCLK1-positive epithelial cells in normal and diseased human pancreatic tissue 54 have not been analyzed with regards to a cholinergic co-phenotype, and their origin is under debate. One explanation is that these cells are not bona fide tuft cells, but arise by a process called transdifferentiation from pancreatic acinar cells, and phenocopy tuft cells from the pancreaticobiliary duct 40 . Our detection of cholinergic tuft cells in the normal human pancreatic duct system, together with their biomolecule co-phenotypes (see below), is suggestive of a role in immunological defense rather than cancer.
In our study we identified advillin as a structural tuft cell marker that is superior to the established markers CK18 and villin. The complete and selective co-localization of advillin with ChAT, and the complete segregation of ChAT staining from CGA-positive enteroendocrine cells rules out the possibility that either advillin or ChAT are expressed by enteroendocrine cells, as suggested in other reports 35 . In addition, given the high sequence Ia-c). The single antibody labeling (a,b) is shown in greyscale, the composite (c) in green (ChAT) and in red (CK18, VIL, AVIL) color coding, respectively. For all double-immunofluorescent analyses of ChAT, the goatanti-ChAT antiserum was used. The bar in A equals 20 µm and applies to all bright-field images. The bar in Aa equals 10 µm and applies to all images from immunofluorescence analysis. www.nature.com/scientificreports www.nature.com/scientificreports/ homology of human villin and advillin it is conceivable that villin antibodies cross-reacted with advillin in tuft cells, rather than monitoring true presence of villin. Nevertheless, since advillin is also a tuft cell-specific marker in the mouse intestinal mucosal epithelium 5 , experimental targeting of tuft cells with the aid of the advillin promotor will be a useful addendum to the existing genetic tools 16 .
We found that intestinal, biliary, and pancreatic cholinergic tuft cells constitutively express two enzymes necessary for prostaglandin D2 production, namely COX1 and HPGDS. This agrees with observations in mice, where a minimal tuft cell core gene signature was defined that included these markers 14,16 , and with an observation in human intestine, where HPGDS immunoreactive tuft cells were identified in adenoma and in adenocarcinoma 14 . Furthermore, using anti-FLAP antibodies in our studies, we found that intestinal and pancreatic cholinergic tuft cells are endowed with leukotriene biosynthesis capacity, while biliary tract tuft cells are not. Alox5 had already been shown previously to be enriched in mouse intestinal tuft cells 5 , and FLAP was detected specifically in mouse airway brush cells 46 . Thus, some tuft cells from the human alimentary tract may secrete a 'soup' of bioactive molecules, including ACh, prostaglandins and leukotrienes into their local microenvironment. However, specific functions for these effector molecules derived from tuft cells have not yet been proposed, and the target cells expressing the cognate receptors not been identified.
Numerous experimental studies in rodents have delineated the emerging concept that cholinergic signaling may be an integral part of the functioning of tuft cells in the context of chemosensation along epithelial barriers 23,24,26,48 . Chemosensation by tuft cells requires presence of G-protein-coupled transmembrane receptors at the luminal surface of tuft cells. Binding of proper ligands, e.g. quorum sensing molecules 55 , is assumed to activate the canonical taste transduction cascade, which includes α-gustducin, PLCβ2, and TRPM5. Ultimately, this intracellular signaling cascade may result in a release of ACh at the abluminal, and perhaps lateral surfaces, into the nearby complex microenvironment, which consists of a heterogeneous population of epithelial cells, lamina propria immune and stromal cells, as well as subepithelial nerve endings of sensory afferent and autonomic efferent fibers. If human intestinal, biliary, and pancreatic cholinergic tuft cells express components of the taste transduction cascade, as determined already for human vallate taste buds 56 , requires further investigation. Nevertheless, the presence of markers for immune-regulatory biomolecule production in human alimentary tract tuft cells (see above) suggests that these cells are involved in chemosensation and regulation of a local immune circuitry. Given the demonstrated absence of VAChT from tuft cells in the intestinal, extrahepatic biliary, and pancreatic tracts, the exact mechanism of the presumed release of ACh from tuft cells cannot be by regulated vesicular exocytosis and, therefore, remains enigmatic. Choline transporter-like proteins are alternative candidates, since they have been shown to be involved in non-neuronal ACh metabolism in humans 57 .
The physiological and pathophysiological roles of tuft cell-derived cholinergic signaling in the human entero-pancreatic and biliary systems is a matter of speculation. In any case, ACh liberated from tuft cells upon sensing luminal chemosignals is likely to have actions on the various nearby cellular and neural targets bearing diverse nicotinic and/or muscarinic receptors. In this context, it has to be considered that ACh released from tuft cells would compete with ACh released by vesicular exocytosis from cholinergic nerve endings. The submucous plexus neurons and subepithelial endings have been shown to contain VAChT and to target the epithelial and lamina propria tuft cell microenvironment (Fig. 1) 38 . Another possible source of ACh are subpopulations of T cells that have been shown to code for ChAT 58 . ACh released from tuft cells, alone or together with ACh from local nerve endings and local T cells, can be expected to influence in a paracrine manner diverse epithelial cells, i.e. enterocytes, goblet cells, Paneth cells, enteroendocrine cells, intestinal stem cells, and transit-amplifying progenitor cells. In addition, a wide variety of cells in the lamina propria, i.e. mesenchymal cells, macrophages, innate lymphoid cells type 2 (ILC2s), mast cells, dendritic cells, T cells, B cells, Tregs, mast cells, basophils and eosinophils, all shown to be endowed with nicotinic and/or muscarinic cholinergic receptors 59 , are possible targets of tuft cell derived ACh. Thus, ACh from tuft cells is likely to contribute to cholinergic modulation of diverse modalities such as barrier function, type 2 innate and adaptive immune responses against e.g. helminths, allergic effector cells, and cancerogenesis 44,60 . Whether cholinergic human tuft cells, like cholinergic tuft cells in mice, represent the host reservoir for chronic infection by norovirus 61 is not yet known. In any case, the precise outcome of a tuft cell-derived cholinergic signaling will definitely depend on the cellular composition of the local microenvironment, and on the disease modality.
pancreatic ducts (Ia-c), but not in peribiliary glands (Ha-c). The single antibody labeling (a,b) is shown in greyscale, the composite (c) in green (ChAT, Villin) and in red (COX1, HPGDS, FLAP) color coding, respectively. The bar in A equals 20 µm and applies to all bright-field images. The bar in Aa equals 10 µm and applies to all images from immunofluorescence analysis. www.nature.com/scientificreports www.nature.com/scientificreports/ The procurement of human material during surgery was done in accordance with local regulations, and approved by the Ethics Committees of the Medical Faculties of the University of Bern, Switzerland or Heidelberg, Germany. Informed consent was obtained from all participants and/or their legal guardian/s. For subsequent immunohistochemical analyses, all human tissues were removed rapidly during surgery and fixed overnight in buffered 4% paraformaldehyde solution and embedded into paraffin after dehydration. Paraffin-embedded specimens were cut in 4 µm sections using a microtome and mounted on silanized glass slides. www.nature.com/scientificreports www.nature.com/scientificreports/ immunohistochemistry. Single brightfield immunohistochemistry was performed as described previously 21 .
In brief, tissue sections were deparaffinized in xylene and rehydrated through a graded series of isopropanol, including 30 min incubation in methanol/0.3% H 2 O 2 to block endogenous peroxidase activity. Subsequently, antigen retrieval was achieved by incubation in 10 mM sodium citrate buffer (pH 6.0) at 92-95 °C for 15 min. Non-specific binding sites were blocked with 5% bovine serum albumin (BSA) in 50 mM phosphate buffered saline (PBS, pH 7.45) for 30 min, followed by an avidin-biotin blocking step (Avidin-Biotin Blocking Kit, Boehringer, Ingelheim, Germany) for 20 min each. Primary antibodies (see Table 1 for details) were applied in PBS/1% BSA over night at 16 °C followed by 2 h at 37 °C. After several washes in PBS, the sections were incubated for 45 min at 37 °C with species-specific biotinylated secondary antibodies (see Table 1 for details), diluted 1:200 in PBS/1% BSA, washed, and incubated for 30 min with avidin-biotin-peroxidase complex (Vectastain Elite ABC kit; Vector Laboratories, Burlingame, CA). Immunoreactions were then visualized by 8 min incubation in 3,3-diaminobenzidine (DAB, Sigma Aldrich, Deisenhofen, Germany), enhanced by the addition of 0.08% ammonium nickel sulfate (Fluka, Bucks, Switzerland). After three 5 min washes in distilled water, some sections were counterstained with hemalaun solution, dehydrated through a graded series of isopropanol, cleared in xylene and finally mounted under coverslips. Digital bright-field pictures were taken with an Olympus AX70 microscope, equipped with an Olympus UC90 camera and Olympus cellSens analyses software.  www.nature.com/scientificreports www.nature.com/scientificreports/ Double immunofluorescence analysis was performed on paraffin sections as follows: After deparaffinization and blocking procedures (see above), two primary antibodies from different donor species (see Table 1 for details) were co-applied in PBS/1% BSA and incubated over night at 16 °C, followed by 2 h at 37 °C. After extensive washing in distilled water followed by PBS, immunoreactions for one antibody were visualized with a secondary antibody labeled with Alexa Fluor 647. The other antibody was visualized by a two-step procedure using a species-specific biotinylated secondary antibody, followed by streptavidin conjugated with Alexa Fluor 488 (see Table for details). Incubation times were 45 min with the biotinylated secondary antibody only, followed by 2 h incubation with a mixture of fluorochrome-conjugated secondary antibody and streptavidin.
Immunofluorescence signals were documented in a surface scan using a BX50WI confocal laser scanning microscope (Olympus Optical, Hamburg, Germany) and Olympus Fluoview 2.1 software, and stored as false color images (8-bit tiff format) without modifications in brightness and contrast.