Mouse intestinal tuft cells express advillin but not villin

Tuft (or brush) cells are solitary chemosensory cells scattered throughout the epithelia of the respiratory and alimentary tract. The actin-binding protein villin (Vil1) is used as a marker of tuft cells and the villin promoter is frequently used to drive expression of the Cre recombinase in tuft cells. While there is widespread agreement about the expression of villin in tuft cells there are several disagreements related to tuft cell lineage commitment and function. We now show that many of these inconsistencies could be resolved by our surprising finding that intestinal tuft cells, in fact, do not express villin protein. Furthermore, we show that a related actin-binding protein, advillin which shares 75% homology with villin, has a tuft cell restricted expression in the gastrointestinal epithelium. Our study identifies advillin as a marker of tuft cells and provides a mechanism for driving gene expression in tuft cells but not in other epithelial cells of the gastrointestinal tract. Our findings fundamentally change the way we identify and study intestinal tuft cells.


Results and Discussion
Based on a recent report, we noted that both villin and advillin expressing cells are present in the gastrointestinal epithelium 20 . Since villin and advillin share very significant (~75%) structural homology, our goal was to first identify antibodies that can distinguish between these two proteins 18 . Using recombinant human villin and advillin proteins we determined that most villin and advillin antibodies cross-react (Fig. 1a). We have previously shown that these antibodies do not cross-react with GST 36,37 . Careful characterization allowed us to identify two antibodies raised against the amino-terminus of villin (N-Villin and N20-Villin) that do not cross react with advillin (Fig. 1a). The region of human villin used to generate these antibodies shows some divergence when compared to human advillin gene. Furthermore, we noted that multiple advillin antibodies identify advillin much better than they cross-react with villin, this includes an advillin antibody raised against its amino-terminus (a.a. 440-526; N-advillin) and one raised against the carboxyl-terminus of advillin (a.a. 750-813; C-advillin). Using H & E (Hematoxylin and Eosin) staining of mouse distal ileum we identified less than 0.5% of the cells with a unique candle-like "tufted" morphology associated with intestinal tuft cells (Fig. 1b). Consistent with previous reports, these cells were determined to be approximately 10 μm long and approximately 5 μm in thickness. Based on the thickness of these cells, and the paucity of these cells in the normal mouse intestine, several attempts to use paraffin-embedded serial sections to correlate the morphological features of the same tuft cells with the expression of villin, advillin, and other known tuft cell markers, were unsuccessful. This may also be, one of the reasons, why such studies have not been performed previously. With that knowledge, we identified advillin expressing cells in the terminal ileum of C57/BL6 mice that resembled histomorphologically tuft cells, rather than enteroendocrine cells with a candle-like "tufted" morphology ( Fig. 1c) 20 . Using paraffin-embedded ileal tissue, we noted that cells with this tufted morphology do not express villin protein (Fig. 1c). Cryopreserved tissue sections (which preserve F-actin better than paraffin embedded tissue) from distal ileum also showed solitary tuft cell-like advillin expressing cells that do not express villin protein (Fig. 1d). The advillin expressing cells show protein localization at the apical tufts and at the basolateral surface in what appears to be vesicular structures (Fig. 1c,d). It may be noted that previous studies have identified villin in the apical tufts and along the basolateral membrane of tuft cells 38 . This is different from the apical brush border restricted distribution of villin in enterocytes. Using the villin knockout (VKO) mice we confirmed the absence of villin (as shown here and as reported previously; Fig. 1e) and the presence of advillin expressing cells that resembled morphologically, tuft cells (Fig. 1f) 39 . Please note the absence of advillin staining in the gastrointestinal epithelium of the VKO mice, demonstrating that in the absence of villin protein the advillin antibodies do not detect any other cells in the epithelium (Fig. 1f). Tuft cells are marked by DCLK1 and are enriched in PTGS1 9,40 . We confirmed our hypothesis that these advillin expressing cells are tuft cells by double labeling with antibodies against these commonly used markers of tuft/brush cells (Fig. 2a,b). Not all DCLK1 positive cells were positive for advillin and likely represent insulinoma-associated 1 positive enteroendocrine cells as has been suggested before 11,13 . Cryopreserved tissue sections show F-actin enriched in candle-like apical "tufts" of epithelial cells that are also positive for advillin expression (Fig. 2c). As described before, a unique morphological feature of tuft cells is the presence of axial bundles of actin filaments with a candle-like tuft or brush morphology 41 . This unique morphology identifies tuft cells and we find advillin but not villin enriched in these cells of the intestinal epithelium.
In the normal mouse intestine the number of tuft cells is very low (~0.5%) although this number can be increased 10-fold in the gut following parasitic (enteric metazoan and protozoan) colonization and infection 42,43 . Alternatively, this can be achieved by exogenous addition of interleukins (IL) IL-4 and IL-13 to isolated mouse enteroids, as reported before and as shown here (Fig. 3a) 43 . Similar to the intestinal tissue, ex vivo in the enteroids, advillin labeled the tuft cells as noted by co-localization of DCLK1 and PTGS1 (Fig. 3b). Using Virtual Channels to acquire multichannel confocal images, we show that all three proteins, PTGS1, DCLK1 and advillin are localized to the same cells (Fig. 4a). As expected, in mouse enteroids advillin co-localizes with the cytoskeletal proteins F-actin and tubulin (Fig. 4b). F-actin and advillin localized primarily to the eponymous apical tuft consisting of actin microfilaments that terminate at the perinuclear region. In contrast, tubulin localizes to the upper half of the cell and the basolateral surface of tuft cells where advillin co-localizes with the cytoplasmic tubulin. As reported before and as shown here, tubulin expression in the upper half of the cell is also unique to tuft cells and is never seen in other gastrointestinal or respiratory epithelial cells 3 . Similar to data shown in Figs. 1c,d,f, 2a-c, 3a,b, the expression of advillin in tuft cells appears to be associated with vesicular structures (Fig. 4b,c). More notably, in mouse enteroids like in the mouse intestine, advillin expressing cells do not express villin protein (Fig. 4c). Although advillin was identified over two decades ago as a new member of the gelsolin/villin family of proteins, the advillin protein has not been studied to determine either its actin binding ability or its ability to modify actin dynamics 18 . Based on that we elected to study, for the first time, the actin regulatory functions of advillin. All in vitro studies were performed with comparable levels of recombinant villin and advillin proteins. Using a standard in vitro assay for the measurement of actin binding by purified recombinant villin and advillin, we show that like villin, advillin is an actin binding protein (Fig. 5a) 44 . Advillin contains a villin-like carboxyl-terminal headpiece domain that is associated with villin's bundling function 18 . Using a standard sedimentation assay for actin bundling we now show that this domain in advillin is functional and that advillin bundles actin similar to the villin protein (Fig. 5b). Advillin also shares the six domain structure of gelsolin and villin that is responsible for the actin depolymerizing functions of both proteins 6 . We now show, for the first time, that like its family members villin and gelsolin, advillin can nucleate, cap and sever actin filaments (Fig. 5c-e). These data demonstrate, for the first time, that advillin shares structural but also functional homology with other members of its family. Both villin and advillin are expressed in the gastrointestinal epithelium but are restricted to distinct cell types. While villin expression is restricted to differentiated intestinal epithelial cells, advillin expression is restricted to the chemosensory tuft cells. This lack of villin from tuft cells may also explain the unique ultrastructural features of tuft cells not shared by enterocytes namely, an apical tuft of stiff microvilli with long microvillar actin rootlets and no terminal web 45 . Furthermore, we hypothesize that unlike the restricted apical brush border localization www.nature.com/scientificreports www.nature.com/scientificreports/ the number of tuft cells 11,12 . However, a subsequent study using the Lgr5-Cre/ERT2 mice crossed to Atoh1 flox/flox mice also demonstrated that loss of Atoh1 from Lgr5 + cells (which includes tuft cells) results in an increase in the number of tuft cells in the intestinal epithelium 12 . Despite that confirmation, the role of Atoh1 in tuft cell development remains controversial, leading some to suggest an Atoh1-dependent and Atoh1-independent lineage for intestinal tuft cells 47 . Similarly, a role for the regulatory associated protein of MTOR complex 1 (RPTOR) in tuft cell lineage has been reported using the Vil1-Cre/ERT2 and the Raptor flox/flox mice even though ex vivo enteroids derived from these mice have normal tuft cells 48 . The role of the Wnt target gene Sox9 is likewise debatable. While several groups have demonstrated that DCLK1 + tuft cells express Sox9 the Vil1-Cre Sox9 flox/flox mice show no effect of Sox9 loss on tuft cell distribution 46,49,50 . The role of tuft cells in the regulation of intestinal repair is also discordant. Deleting Dclk1 using the Vil1-Cre Dclk1 flox/flox model showed that animals fail to recover normal crypt-villus architecture following genotoxic insult or treatment with dextran sodium sulfate, suggesting a loss of epithelial regeneration accompanied by a dramatic loss of self-renewal pathways regulated by Notch and mammalian target of rapamycin (mTOR) 51,52 . Consistent with that, two independent studies using the Vil1-Cre Dclk1 flox/flox model demonstrated that tuft cells regulate DNA damage response and that the crypts of these mice have significantly higher number of apoptotic cells following radiation-induced injury 51,53 . In contrast, McKinley et al. demonstrated that when compared to most intestinal epithelial cells, tuft cells are extremely resistant to mucosal atrophy in response to acute fasting 33 . Additionally, using Dclk1-Cre/ERT2 Rosa26-LacZ reporter mice it was shown that tuft cells do not function as reserve stem cells further supporting the idea that tuft cells play no role in epithelial regeneration, a finding that was supported by studies done by Nakanishi et al. 12,54 Our study underscores that the most likely explanation for many of these and similar conflicting findings related to tuft cell lineage commitment and function namely, the absence of villin protein from tuft cells. While the villin-Cre and villin-Cre/ERT2 mice are used to target gene expression or gene loss from a limited number of adult gastrointestinal epithelial cells that endogenously express villin, namely the enterocytes, unexpected expression of Cre recombinase in cells that do not endogenously express villin such as goblet cells is known. This 'ectopic' expression of villin-Cre in goblet cells and potentially tuft cells should be approached with caution. It is also known that a given Cre-expressing mouse can have different recombination efficiencies for different floxed genes in different cell types. Our findings www.nature.com/scientificreports www.nature.com/scientificreports/ underscore the importance of testing the Cre expression outside the expected cell type and highlight the need for a greater degree of caution that must be used when empolying the Vil1-Cre or villin-Cre/ERT2 mice to study gastrointestinal tuft cells.
Bezencon et al. used the Trpm5 promoter to express enhanced green fluorescent protein (EGFP) in intestinal tuft cells followed by RNA-Seq of the isolated EGFP expressing cells 29 . Their study reported very high expression of advillin but not villin mRNA in tuft cells and they identified no advillin mRNA in other EGFP null cells of the gastrointestinal epithelium 29 . Based on that the authors even suggested that it was possible that previously published immunostaining of brush cells with villin antibodies resulted from cross-reactivity with advillin 29 . We note that multiple single-cell RNA-Seq studies published after that have also identified advillin but none have identified villin mRNA in tuft/brush cells and a few have also shown the absence of advillin mRNA from other intestinal epithelial cells including enterocytes 9,30-32 . Nevertheless, we also note that none of these studies address these discrepancies related to villin/advillin expression in tuft/brush cells. Using the advillin promoter driven EGFP expression and a Human Protein Atlas advillin antibody, Hunter and colleagues reported that advillin is expressed in the duodenum in solitary endocrine cells even though these cells appear to morphologically resemble the candle-like "tufted" cells 20 . Their study also did not include any specific marker of enteroendocrine cells. Additionally, the authors reported advillin expression in enteric neurons 20 . We note that while Hunter et al. identify cells in the Meissner's plexus that are enriched in EGFP-advillin, these cells do not appear to be the same as those identified by the advillin antibody. More significantly, the same antibody does not identify any enteric neurons in the duodenum or any other sections of the gastrointestinal tract but identifies solitary cells in the gastrointestinal epithelium with a candle-like "tufted" morphology (https://www.proteinatlas.org/ENSG00000 135407-AVIL/tissue). Our own studies with these antibodies (N-advillin) identify advillin positive tuft cells. It is known that tuft cells communicate with afferent nerves and that PGP9.5-positive nerves make direct contact with duodenal tuft cells 29 . We suggest that one possibility is that Hunter et al. have identified enteric neurons (identified by EGFP-advillin) in close proximity of advillin expressing tuft cells 20 . Alternatively, this discrepancy could be attributed to the ectopic expression of EGFP advillin in enteric neurons. There are a few single cell RNA-Seq studies on enteric neurons that do not identify advillin mRNA in enteric neurons or glia cells (Public databases: PanglaoDB; Single Cell Expression Atlas -EMBL-EBI) 34,35,55,56 , Based on that we suggest that advillin expression in the gastrointestinal tract may be restricted to the tuft cells. It is generally agreed that only tuft cell specific deletion of genes will provide the most definitive understanding of tuft cell lineage and function in vivo 7,31 . Unlike the Vil1-Cre or Vil1-Cre/ERT2 mice the availability of the advillin Cre mouse allows, for the first time, the targeting of this subpopulation of gastrointestinal epithelial cells, rather than all epithelial cells. Such approaches are more likely to provide a detailed investigation of the molecular mechanisms and functions of intestinal tuft cells.
The exact function of advillin has not been investigated although its distribution to intestinal tuft cells (and by extension respiratory brush cells) indicates a role in chemosensing and initiation of immune type 2 response. The fact that tuft cells express advillin and not villin suggests that despite the significant structural and functional homology, advillin may be uniquely adapted to regulate chemosensory and mechanosensory functions of tuft cells. Advillin also differs from villin in significant ways such as advillin interacts with the scavenger receptor SREC-1 and contributes to neurite outgrowth in vitro [57][58][59] . Since tuft cells are found in contact with nerve fibers, advillin may have a unique function in regulating the enteric nervous system. That would also suggest that advillin plays a role in transferring sensory signals to the enteric nervous system.
Mice. All experimental protocols were approved by Institutional Animal Care and Use Committee (IACUC) of the University of Houston. Villin-1 knock-out (VKO) mice and their wild-type (WT) littermates have been described by us previously 60 . C57BL/6 J mice were purchased from the Jackson laboratory.
Enteroid 3D culture and cytokines treatment. All methods were performed in accordance with the relevant guidelines and regulations. Crypt isolation and enteroid 3D cultures were derived from distal ileum of mice as previously described 61 . Organoid lines were passaged up to 10 times before experiments to ensure pure