Tuft cells – systemically dispersed sensory epithelia integrating immune and neural circuitry

Nasopharyngeal epithelia

The nasal passages represent a critical conduit to the outside world, lined by a complex epithelial layer extending from squamous and transitional zones at the anterior nares through the prominent anterior/inferior nasal, or respiratory, epithelium (NE), and posterior/superior olfactory epithelium (OE). OE constitutes a greater proportion of the epithelial surface in rodents (~40%) than in humans (~5%). The nasopharyngeal epithelia engage in numerous processes, including air humidification, particle filtration, odorant and toxin detection, and elaboration of mucus and antimicrobial peptides (19). The epithelia of both NE and OE includes TRPM5⁺, ChAT⁺, bitter taste receptor-expressing pear-shaped cells with an apical, luminal tuft, that are often designated solitary chemosensory cells in NE and microvillous cells in OE. In mice, these cells require POU2F3 for development (20), leading us to consider these bona fide tuft cells.

Application of T2R bitter ligands, including acyl-homoserine lactones associated with quorum sensing in Gram-negative bacteria, initiated α-gustducin- and TRPM5-dependent Ca²⁺ depolarization and ACh release from NE tuft cells, resulting in local neurogenic mast cell-mediated inflammation and activation of trigeminal sensory fibers, thereby initiating a protective apnea response (21, 22). Using human NE cell cultures, Lee et al. demonstrated that bitter taste receptor ligands initiated a propagating Ca²⁺ depolarization from tuft cells that stimulated robust antimicrobial peptide secretion from adjacent cells, suggesting a rapidly deployed method by which relatively few cells could control bacterial homeostasis across the epithelial surface (23). Conversely, activation of the sweet taste receptor T1R2/3 by bacterial D-amino acids from Staphylococcus aureus inhibited the pathway, leading to enhanced bacterial growth (24). These findings are consistent with a chemosensory role for nasal tuft cells in epithelial – microbial homeostasis, and possible activation of aversive behaviors, like coughing and sneezing.

The association of type 2 immunity with Group 2 innate lymphoid cells (ILC2s) prompted examination of patients with chronic rhinosinusitis with nasal polyps, which presents in an allergic form associated with asthma and eosinophilia, and more frequently among patients with worse disease (25). Elevated ILC2s, type 2 cytokines, and eosinophils were present in polyps (26), and blockade of IL-4Rα led to diminution in the polyp burden, implicating type 2 cytokines in the epithelial hyperplasia (27). Consistent with studies in the mouse, solitary chemosensory cells (tuft cells) were the major source of IL-25 in the NE from these patients (28), but whether IL-25 or additional epithelial cytokines, such as IL-33 (29), or both, drive the disease remains unknown. A recent single-cell RNA sequencing (scRNA-seq) analysis of human allergic nasal polyp tissues in NE revealed profound effects on epithelial diversity, with reductions in ciliated cells and basal cell hyperplasia (30).

Microvillous cells, here defined as tuft cells, are present within the complex OE, and are likely to arise from horizontal basal cells, which generate transit-amplifying-like globose basal cells competent to differentiate into olfactory sensory neurons, supporting sustentacular cells, and tuft cells (31). Computational analysis of scRNA-seq suggests the epithelial sustentacular and tuft cell populations diverge prior to cells committing to neural identity, although these trajectories were altered after epithelial injury (32). Tuft cells in OE express ChAT and the vesicular ACh transporter, VAChT, and modulate olfactory responses to xenobiotic challenges that activate Ca²⁺ depolarization and ACh release (33) and may play a role in maintaining olfactory function, in part via release of neuropeptides (34, 35). VAChT is notably missing from the tuft cell transcriptome in other tissues.

Eyes and ears

Tuft cells with typical morphology and expression of canonical markers of the taste receptor signaling cascade were present in the auditory tubes and nasal aspects of the conjunctival epithelia in mice (36, 37). In both tissues, tuft cells were in close proximity to cholinoreceptive sensory nerve fibers with the capacity for neuropeptide release, including CGRP, consistent with a tuft cell-mediated inflammatory circuit involved in restricting ingression of toxic substances into deeper tissues.

Thymus

Speculations regarding a role for cholinergic signaling in thymic function led Panneck et al. to investigate the sources of ACh using Chat-eGFP BAC transgenic mice (38). A subset of eGFP⁺ medullary thymic epithelial cells (mTECs) expressed α-gustducin, TRPM5 and PLCβ2, and was present as isolated cells and in loose aggregations associated with cornified, terminally-differentiated epithelia in Hassall’s corpuscle-like clusters. A taste receptor reporter mouse (Tas2r131) confirmed expression of several bitter taste-associated receptors (Tas2r105, Tas2r108, and Tas2r131) in these cells, although heterogeneity was evident (39). Thymic tuft cells were also labeled using BAC transgenic mice expressing CreERT2 from the Tas2r143 promoter, which is controlled by shared cis-regulatory elements in the bitter taste receptor cluster, which also includes Tas2r135 and Tas2r126 (40).

More recently, Miller et al, using inducible lineage-tracking of Aire expression in murine mTECs, identified a population of post-Aire cells that expressed the transcriptional signature of tuft cells: components of the taste receptor signal transduction cascade, ChAT, IL-25, DCLK1, and multiple bitter taste receptors. The latter were expressed heterogeneously among different cells (8). Post-Aire mTECs were absent in Pou2f3- and in Trpm5-deficient mice, corroborating their lineage specification as tuft cells and suggesting dependence on Ca²⁺-mediated depolarization. Mice deficient in post-Aire mTECs (tuft cells) had profound reductions in NKT2 cells that express IL-4 and Eomes-expressing CD8⁺ innate-like (“virtual memory”) thymocytes that require IL-4 for differentiation. Unlike peripheral tuft cells, thymic tuft cells expressed low levels of MHCII and CD74; transfer of Pou2f3-deficient thymi into athymic nude mice rendered animals non-tolerant to subsequent immunization with IL-25, suggesting that these cells enforce tolerance to tuft cell-derived antigens. A second scRNA-seq analysis of mouse thymic epithelia also found IL-25-expressing tuft cells and used ATAC-seq and ChIP-seq to define the landscape of POU2F3 binding sites in these cells (10).

Lower respiratory tract

The lower respiratory tract comprises the trachea and branching bronchi and bronchioles, which carry air to the distal alveoli for gas exchange. In humans, the trachea, bronchi, and larger bronchioles are lined by pseudostratified epithelium of multiciliated, secretory, and basal cells, followed by a simple cuboidal epithelium in the terminal bronchioles and squamous type 1 and cuboidal type 2 alveolar epithelial cells. In mice, only the trachea and main-stem bronchi are lined with pseudostratified epithelium; simple columnar epithelium lacking basal cells lines the smaller bronchi and bronchioles (reviewed in (41, 42)).

Despite early descriptions of tuft cells in respiratory epithelium, understanding remains incomplete beyond histological characterization. Tuft cells appear to comprise at least a subset of cells commonly termed “brush cells” due to their dense apical tuft morphology (43). Solitary α-gustducin, TRPM5, ChAT, villin, and bitter taste receptor-expressing pear-shaped cells with an apical, luminal tuft, were present in rodent airways with decreasing abundance from trachea to larger bronchioles, but not in smaller bronchioles and alveoli (44); a similar distribution was present in cattle (45). Some studies reported brush cells in the rat distal airways, including in alveolar epithelia, with increased frequencies under certain pathological conditions (reviewed in (43)).

In the trachea of mice, tuft cells express IL-25 (11), and share a core gene expression program with tuft cells in other organs (9). Tracheal tuft cells are absent in Pou2f3^−/− mice, whereas a cell population with similar morphology – marked by a villin-bright apical tuft, but of otherwise unknown identity and negative for ChAT and TRPM5 – remained intact (46), revealing heterogeneity among airway brush cells as previously noted (16). The tuft cell developmental trajectory and relationship with other lineages in the lower airways remain unknown. Lineage tracing in mice revealed regionally distinct epithelial progenitor pools (41, 42), including Trp63⁺ NGFR⁺ basal cells, Club cells, and type 2 alveolar epithelial cells. Using BrdU pulse-chase experiments, Saunders et al. determined that tracheal tuft cells, like those in other tissues, are post-mitotic. During post-natal tracheal growth, tuft cells differentiated with slow kinetics from BrdU-labelled proliferating precursors that expressed basal cell-associated cytokeratin 14 (47). However, precise lineage tracing during development, in adulthood, and in the context of tissue damage is lacking.

In addition to taste signaling components, tracheal tuft cells express various taste receptors, consistent with chemosensory functions (9, 16, 40, 44, 46). Tracheal tuft cells were in contact with peptidergic cholinoceptive vagal sensory neurons, and stimulation with the bitter ligands cycloheximide or 3-OxoC12-homoserine lactone, a P. aeruginosa-associated quorum sensing molecule, resulted in decreased respiratory rate that was inhibited by a nicotinic acetylcholine receptor antagonist (16, 48). Bitter taste receptors were also present on cultured human airway epithelia, which responded to agonist stimulation with increased ciliary beat frequency (49), and on human airway smooth muscle cells, which responded to bitter tastants with relaxation and airway dilation (50). Single cell analysis of airway tuft cells will clarify whether taste receptors are expressed individually or combinatorially (e.g.: T1Rs with T2Rs), distinct from the classical concept of taste segregation among taste bud cells (51), but consistent with the observed co-expression in nasal tuft cells (52) with antagonistic signaling characteristics (23, 24). Elucidating as-yet unexplored immunologic roles for airway tuft cells will be important to identify roles in chronic diseases like asthma and eosinophilic lung vasculitic syndromes (e.g.: eosinophilic granulomatosis with polyangiitis (EGPA), formerly Churg-Strauss syndrome).

The Gastrointestinal Tract

During embryogenesis, the gastrointestinal tract develops from the endodermal primitive gut tube, subsequently forming the esophagus, stomach, intestines, and their associated organs, the liver, gall bladder and pancreas (53). The specialized mucosal epithelia in these tissues differ in microanatomical organization, cellular composition and turnover rates, and can be broadly separated into stratified squamous non-keratinizing epithelia in the esophagus and anal canal, and simple columnar epithelia in the remaining gastrointestinal tract, organized into gastric pits and glands in the stomach, crypts in the colon and cecum, and crypt-villus units in the small intestine (54). Tuft cells are absent from the stratified squamous epithelium of the anal canal and esophagus but may increase following replacement with a metaplastic, intestinal-like columnar epithelium (55, 56). In contrast, tuft cells are present throughout the simple columnar epithelia of the pancreato-biliary system, stomach, small and large intestine, and cecum.

Pancreato-biliary system

The pancreato-biliary system includes the gallbladder, extrahepatic bile ducts (EHBDs), pancreas and pancreatic ducts (57). The liver, pancreas and biliary tree develop together from the foregut endoderm, with the pancreas and extrahepatic biliary tree sharing a precursor distinct from the intrahepatic ducts and the liver (58). The biliary tissue consists of simple columnar epithelium, a thin lamina propria, and smooth muscle, and is innervated at several levels (59). The morphologically similar epithelial cells of the intrahepatic and extrahepatic ducts are commonly termed cholangiocytes (59). In the pancreas, the larger interlobular ducts and the major pancreatic duct are lined by simple columnar epithelium; small intralobular pancreatic duct epithelium is cuboidal (60). During feeding, cholecystokinin stimulates gallbladder emptying and the release of pancreatic secretions into the duodenum (57, 61). Bile is a key route for excretion of xenobiotics, and both gallbladder and EHBDs can be sites of long-term bacterial colonization and persistent bacterial shedding (62). Thus, similar to the small intestine and the colon, the epithelium of the pancreato-biliary system, and particularly the biliary tree, is exposed to nutrients and hormones, as well as micro-organisms and soluble xenobiotics.

The presence of tuft cells in biliary epithelium was noted decades ago by EM of the mouse gallbladder and the rat bile duct, where the striking microvillus structure was noted among the homogeneous principal cells (63–65). Tuft cells in gallbladder and EHBDs are far more abundant than in small intestine (63). Similar cells are present in human tissues (66). More recent work using reporter mice for ChAT, TRPM5, and IL-25, and staining for DCLK1 has confirmed the abundance of tuft cells in the gallbladder and EHBDs (11, 55). IL-25⁺ tuft cells are not present in intrahepatic bile ducts (IHBDs) (our unpublished observations), consistent with their development from distinct epithelial precursors (58). Despite their abundance, tuft cell function in the gall bladder and EHBDs remains unknown. Given the involvement of type 2 cytokines in fibrotic diseases (67), it will be important to evaluate potential roles for tuft cells in this constellation of hepatobiliary diseases. Biliary epithelial cells are found in close proximity to neurons, which directly innervate biliary epithelium, and smooth muscle, both potential cell signaling partners (68, 69).

Tuft cells in the pancreatic duct were also noted in early studies (70, 71), but little information has since accrued. The high abundance of tuft cells – 22% in specific ductal locations – reported in rat pancreatic duct (72) has not been quantified in other studies, although tuft cells in the major pancreatic duct are POU2F3-dependent in mouse (46). Tuft-like cells have also been documented in some pancreatic cancers (73, 74), and may have a biliary phenotype (73). Although additional work is needed, this suggests that pancreatic ductal tissue may acquire a biliary phenotype during oncogenesis, marked by appearance of tuft-like cells in a Sox17-dependent fashion (73). However, the alternative hypothesis, that this represents expansion of an existing tuft cell population in the pancreatic ducts, as suggested by older literature, has not been tested. Staining of normal pancreas using DCLK1 revealed ductal tuft cells, but also peripheral islet cells suggested to be stem-like cells, perhaps complicating tuft cell studies in this tissue (75).

Stomach

Jawed vertebrates have a muscular stomach, derived from the foregut endoderm, used for food storage, mechanical disruption, and preliminary digestion by acid and proteases (76). The stomach is adapted to food intake, and thus varies substantially from species to species: rodents have an anatomically distinct forestomach and glandular stomach (the corpus and the antrum) while in humans the corpus is expanded (76). The simple columnar epithelium of the gastric mucosa in humans begins abruptly after the esophagus and continues throughout the stomach, while in rodents the forestomach epithelium, like the esophagus, is squamous.

Tuft cells are found in the gastric mucosa of the corpus and antrum in rodents and humans as solitary cells (6, 77). In mice, however, tuft cells are enriched along the limiting ridge at the transition from squamous to simple columnar epithelium separating the forestomach and the corpus (78, 79). Tuft cells appear in the stomach of mice at birth, but are rare and lack any enrichment at the limiting ridge (80). At this time, murine corpus and antrum tuft cells appear morphologically distinct. As the glandular stomach develops, tuft cells increase to achieve the densely packed organization of the limiting ridge observed after weaning (80). Although this stereotyped appearance of gastric tuft cells suggests that tuft cells are linked to digestive function of the stomach during postnatal feeding – perhaps delimiting a barrier between the acidic gastric juices and the storage function of the forestomach – the role for tuft cells remains unclear.

It has been proposed that bile duct and gastric tuft cells may play a role in regulating levels of bicarbonate and/or sensing acidic conditions; indeed, the high proportion of tuft cells that are observed in acidic locations suggests a potential specialized role in dealing with electrolyte-rich, acidic luminal content, similar to the role for chloride cells in fish (81). Ultrastructural analysis of rat bile duct and gastric groove following in vivo exposure to electrolytes induced visual morphological changes in tuft cells (81). Although previously suggested that tuft cells in rat gastric mucosa and bile duct express components of the bicarbonate secretion machinery (82, 83), this was not apparent in RNAseq data from gallbladder tuft cells (6). Eberle et al. proposed that secretion of prostaglandins from gastric tuft cells downstream of acid-sensitive GPCRs might act on neighboring bicarbonate-secreting cells (84). Although correlative, the relationship between tuft cells and acid sensing is consistent with the impact of acid on TRPM5 signaling (85), and further study is warranted.

Small and large intestine

Early studies, recently reviewed (5), described tuft cells in the small and large intestines as rare isolated cells similar to other secretory cell lineages, including goblet and enteroendocrine cells, interspersed between the more abundant enterocytes. Epithelial progenitors and signals promoting the differentiation of epithelial lineages are best characterized in small intestine, where dividing LGR5⁺ stem cells at the crypt base give rise to rapidly proliferating transit-amplifying cells and continuously renew the differentiated epithelium on the villus with a complete turnover rate of approximately 4-5 days (86). Despite increasing clarity regarding stem cell-epithelial cell differentiation in the small intestine, tuft cell differentiation remains controversial.

The fate of intestinal progenitor cells, i.e., specification into absorptive (enterocyte) vs. secretory (tuft, goblet, enteroendocrine, Paneth) lineages, is controlled in stem and progenitor cells by Notch signaling, which is activated by delta-like 1 and 4 expressed in intestinal secretory cells, including Paneth cells and crypt base goblet cells (87). Proteolytically-released Notch intracellular domain translocates into the nucleus resulting in transcription of target genes, including the transcriptional repressor Hes1. HES1 promotes enterocyte fate, in part through repression of Atoh1 (88), the transcriptional activator essential for intestinal secretory cell commitment (89, 90). However, the requirement for ATOH1 in tuft cell differentiation is unclear. Studies report both absence of tuft cells in the small intestine of mice with conditionally deleted Atoh1 alleles (91), and increased abundance of tuft cells, possibly because of the deficiency in other secretory lineages (92). The latter study reported only minor frequencies of fate-mapped tuft cells using Atoh1-Cre and Atoh1-CreER lineage tracing in adult mice. However, Bjerkenes et al did note absence of tuft cells from the small intestine, but not from the gastric epithelium, of global Atoh1 knockout mice on embryonic day E18.5. While the authors suggested this was related to developmental abnormalities in Atoh1 deficiency, an alternative interpretation is that fetal small intestinal tuft cell development is ATOH1-dependent (92). Another study verified intact small intestinal tuft cells in adult mice with conditional deletion of Atoh1 using Lgr5-CreERT2 (93). A third study used Vil1-CreERT2 and Lrig-CreERT2 to delete Atoh1 from the intestinal epithelium and observed increased tuft cells in the proximal and distal small intestine consistent with the results of Bjerknes et al.; in contrast, colonic tuft cells were strongly reduced after Atoh1 deletion, suggesting differential requirements for ATOH1 in small and large intestinal tuft cells (94). Using p-Creode, a novel computational trajectory-mapping algorithm for high-dimensional single-cell data, these authors also provided evidence for a tuft cell developmental trajectory that was distinct from ATOH1-dependent goblet and Paneth cells.

One explanation for the conflicting results from distinct laboratories and different developmental timepoints could be the existence of separate tuft cell differentiation pathways, one ATOH1-dependent and one ATOH1-independent, perhaps equivalent to the IL-4Rα-independent and IL-4Rα-/STAT6-dependent populations of tuft cells found in small intestine (11–13, 95). Indeed, using scRNA-seq analysis of small intestinal epithelial cells, Haber et al. suggested the existence of two tuft cell populations; tuft cell heterogeneity was also described based on protein expression analysis using multiplex immunofluorescence (96). This heterogeneity could be driven by the variable presence of type 2 immune stimuli in different mouse colonies caused by Tritrichomonas (described below). Notably, IL-13 can cause upregulation of SOX4 in crypt progenitors, and a recent study showed increased SOX4 promotes tuft and enteroendocrine fate in an ATOH1-independent manner (97). Vil1-Cre x Sox4^fl/fl mice displayed a minor baseline decrease in tuft cells, but a more pronounced defect in tuft cell expansion after infection with the intestinal helminth N. brasiliensis, which potently activates the IL-13 – IL-4Rα axis. In vitro deletion of Atoh1 in organoid cultures, unexposed to exogenous IL-4/IL-13 and perhaps more analogous to the fetal state, suppressed tuft cell differentiation. This defect could be overcome by overexpression of Sox4. Goblet cell fate remained ATOH1-dependent.

Notably, SOX family proteins are well-characterized partners for heterodimerization with transcription factors of the POU homeodomain family (SIDEBAR POU2F3) (98), of which POU2F3 has emerged as the only lineage-defining transcription factor identified that is required for tuft cell specification. POU2F3 is expressed in TRPM5⁺ chemosensory cells, including taste receptor cells and tuft cells, which are all absent in Pou2f3-deficient mice (13, 20, 99, 100), with the only exception so far being certain TRPM5⁻ Villin^bright brush cells outside the intestinal tract, whose identity remains unknown (46). How Pou2f3 expression is regulated and its target genes in tuft cell fate determination are unknown. The role of the transcription factor GFI1B, which in the intestinal epithelium is expressed specifically by tuft cells, remains to be determined (92).

Based on existing literature, we propose the following model for tuft cell lineage specification in the small intestine (Figure 3a and b): under germfree/fetal conditions, epithelial progenitors acquire tuft cell fate in an ATOH1-dependent manner, like other secretory lineages; alternatively, upon activation of IL-4Rα signaling and SOX4, an ATOH1-independent pathway is engaged, possibly involving different progenitor cells, enabling a rapid and dynamic increase in small intestinal tuft cell differentiation in the context of colonization with helminths or certain protists.

While the roles of small intestinal tuft cells in type 2 immune responses have recently been elucidated (see below), both small and large intestinal tuft cells share a “neuronal” gene signature suggestive of roles in both immune and neuronal circuits. The emerging concept that tuft cells in the intestine are part of a diffuse cholinergic chemosensory system is supported by the expression of taste transduction signaling components and ChAT by intestinal tuft cells. However, intestinal tuft cells lack the high-affinity choline transporter (CHT) and the vesicular acetylcholine transporter (VAChT), responsible for loading acetylcholine into secretory organelles, with the exception of a VAChT+ subset of tuft cells in the ascending colon (55). The bitter ligand denatonium stimulates intracellular calcium flux in colon tuft cells (55). A Tas2r143/Tas2r135/Tas2r126 cluster fate-mapping mouse model marked tuft cell in the gastroinstinal tract and revealed Tas2r126 expression by gastric tuft cells (40). Further, tuft cell – neuronal cross talk has been suggested based on the co-localization of tuft cells with nerves (reviewed in (101)). Nonetheless, many proposed roles of intestinal tuft cells, including a receptive function—until recently (discussed below)—lacked supporting experimental evidence.

Urogenital tissues

Chat-eGFP reporter mice were used to identify urogenital tuft cells in the urethral duct epithelia and adjoining excretory gland ducts (15, 102). Urethral tuft cells are POU2F3-dependent (46) and express typical taste receptor sensing pathway constituents (15, 40); further, these cells fluxed Ca²⁺ in response to monosodium glutamate and denatonium, as well as heat-killed Escherichia coli, a common urinary tract pathogen, and ACh concentration in the media increased following denatonium-challenge of urethral cells in vitro (15). In vivo, denatonium induced activation of the bladder detrusor muscle, suggesting a role for tuft cells in restricting reflux of microbes into urogenital organs (15). Similar cells were present in humans and other placental mammals, consistent with an ancient evolutionary origin (103).