Advertisement

Succinate Produced by Intestinal Microbes Promotes Specification of Tuft Cells to Suppress Ileal Inflammation

  • Amrita Banerjee
    Affiliations
    Epithelial Biology Center, Vanderbilt University Medical Center, Nashville, Tennessee

    Department of Cell and Developmental Biology, Vanderbilt University School of Medicine, Nashville, Tennessee
    Search for articles by this author
  • Charles A. Herring
    Affiliations
    Epithelial Biology Center, Vanderbilt University Medical Center, Nashville, Tennessee

    Program in Chemical and Physical Biology, Vanderbilt University, Nashville, Tennessee
    Search for articles by this author
  • Bob Chen
    Affiliations
    Epithelial Biology Center, Vanderbilt University Medical Center, Nashville, Tennessee

    Program in Chemical and Physical Biology, Vanderbilt University, Nashville, Tennessee
    Search for articles by this author
  • Hyeyon Kim
    Affiliations
    Department of Cell and Developmental Biology, Vanderbilt University School of Medicine, Nashville, Tennessee
    Search for articles by this author
  • Alan J. Simmons
    Affiliations
    Epithelial Biology Center, Vanderbilt University Medical Center, Nashville, Tennessee

    Department of Cell and Developmental Biology, Vanderbilt University School of Medicine, Nashville, Tennessee
    Search for articles by this author
  • Austin N. Southard-Smith
    Affiliations
    Epithelial Biology Center, Vanderbilt University Medical Center, Nashville, Tennessee

    Department of Cell and Developmental Biology, Vanderbilt University School of Medicine, Nashville, Tennessee
    Search for articles by this author
  • Margaret M. Allaman
    Affiliations
    Division of Gastroenterology, Hepatology, and Nutrition, Department of Medicine, Vanderbilt University Medical Center, Nashville, Tennessee

    Center for Mucosal Inflammation and Cancer, Vanderbilt University Medical Center, Nashville, Tennessee
    Search for articles by this author
  • James R. White
    Affiliations
    Resphera Biosciences, Baltimore, Maryland
    Search for articles by this author
  • Mary C. Macedonia
    Affiliations
    Epithelial Biology Center, Vanderbilt University Medical Center, Nashville, Tennessee

    Department of Cell and Developmental Biology, Vanderbilt University School of Medicine, Nashville, Tennessee
    Search for articles by this author
  • Eliot T. Mckinley
    Affiliations
    Epithelial Biology Center, Vanderbilt University Medical Center, Nashville, Tennessee

    Division of Gastroenterology, Hepatology, and Nutrition, Department of Medicine, Vanderbilt University Medical Center, Nashville, Tennessee
    Search for articles by this author
  • Marisol A. Ramirez-Solano
    Affiliations
    Department of Biostatistics, Vanderbilt University Medical Center, Nashville, Tennessee

    Center for Quantitative Sciences, Vanderbilt University School of Medicine, Nashville, Tennessee
    Search for articles by this author
  • Elizabeth A. Scoville
    Affiliations
    Division of Gastroenterology, Hepatology, and Nutrition, Department of Medicine, Vanderbilt University Medical Center, Nashville, Tennessee

    Center for Mucosal Inflammation and Cancer, Vanderbilt University Medical Center, Nashville, Tennessee
    Search for articles by this author
  • Qi Liu
    Affiliations
    Department of Biostatistics, Vanderbilt University Medical Center, Nashville, Tennessee

    Center for Quantitative Sciences, Vanderbilt University School of Medicine, Nashville, Tennessee
    Search for articles by this author
  • Keith T. Wilson
    Affiliations
    Division of Gastroenterology, Hepatology, and Nutrition, Department of Medicine, Vanderbilt University Medical Center, Nashville, Tennessee

    Center for Mucosal Inflammation and Cancer, Vanderbilt University Medical Center, Nashville, Tennessee

    Veterans Affairs Medical Center, Tennessee Valley Healthcare System, Nashville, Tennessee

    Department of Pathology, Microbiology, and Immunology, Vanderbilt University Medical Center, Nashville, Tennessee
    Search for articles by this author
  • Robert J. Coffey
    Affiliations
    Epithelial Biology Center, Vanderbilt University Medical Center, Nashville, Tennessee

    Division of Gastroenterology, Hepatology, and Nutrition, Department of Medicine, Vanderbilt University Medical Center, Nashville, Tennessee
    Search for articles by this author
  • M. Kay Washington
    Affiliations
    Department of Pathology, Microbiology, and Immunology, Vanderbilt University Medical Center, Nashville, Tennessee
    Search for articles by this author
  • Jeremy A. Goettel
    Affiliations
    Division of Gastroenterology, Hepatology, and Nutrition, Department of Medicine, Vanderbilt University Medical Center, Nashville, Tennessee

    Center for Mucosal Inflammation and Cancer, Vanderbilt University Medical Center, Nashville, Tennessee

    Department of Pathology, Microbiology, and Immunology, Vanderbilt University Medical Center, Nashville, Tennessee
    Search for articles by this author
  • Ken S. Lau
    Correspondence
    Correspondence Address correspondence to: Ken S. Lau, PhD, Vanderbilt University Medical Center, 2213 Garland Avenue, 10475 Medical Research Building IV, Nashville, Tennessee 37232-0441.
    Affiliations
    Epithelial Biology Center, Vanderbilt University Medical Center, Nashville, Tennessee

    Department of Cell and Developmental Biology, Vanderbilt University School of Medicine, Nashville, Tennessee

    Program in Chemical and Physical Biology, Vanderbilt University, Nashville, Tennessee

    Center for Mucosal Inflammation and Cancer, Vanderbilt University Medical Center, Nashville, Tennessee

    Center for Quantitative Sciences, Vanderbilt University School of Medicine, Nashville, Tennessee
    Search for articles by this author
Open AccessPublished:August 20, 2020DOI:https://doi.org/10.1053/j.gastro.2020.08.029

      Background & Aims

      Countries endemic for parasitic infestations have a lower incidence of Crohn’s disease (CD) than nonendemic countries, and there have been anecdotal reports of the beneficial effects of helminths in CD patients. Tuft cells in the small intestine sense and direct the immune response against eukaryotic parasites. We investigated the activities of tuft cells in patients with CD and mouse models of intestinal inflammation.

      Methods

      We used microscopy to quantify tuft cells in intestinal specimens from patients with ileal CD (n = 19), healthy individuals (n = 14), and TNFΔARE/+ mice, which develop Crohn’s-like ileitis. We performed single-cell RNA sequencing, mass spectrometry, and microbiome profiling of intestinal tissues from wild-type and Atoh1-knockout mice, which have expansion of tuft cells, to study interactions between microbes and tuft cell populations. We assessed microbe dependence of tuft cell populations using microbiome depletion, organoids, and microbe transplant experiments. We used multiplex imaging and cytokine assays to assess alterations in inflammatory response following expansion of tuft cells with succinate administration in TNFΔARE/+ and anti-CD3E CD mouse models.

      Results

      Inflamed ileal tissues from patients and mice had reduced numbers of tuft cells, compared with healthy individuals or wild-type mice. Expansion of tuft cells was associated with increased expression of genes that regulate the tricarboxylic acid cycle, which resulted from microbe production of the metabolite succinate. Experiments in which we manipulated the intestinal microbiota of mice revealed the existence of an ATOH1-independent population of tuft cells that was sensitive to metabolites produced by microbes. Administration of succinate to mice expanded tuft cells and reduced intestinal inflammation in TNFΔARE/+ mice and anti-CD3E-treated mice, increased GATA3+ cells and type 2 cytokines (IL22, IL25, IL13), and decreased RORGT+ cells and type 17 cytokines (IL23) in a tuft cell-dependent manner.

      Conclusions

      We found that tuft cell expansion reduced chronic intestinal inflammation in mice. Strategies to expand tuft cells might be developed for treatment of CD.

      Graphical abstract

      Keywords

      Abbreviations used in this paper:

      CD (Crohn’s disease), COX2 (cyclooxygenase 2), DCLK1 (doublecortin-like kinase 1), IBD (inflammatory bowel disease), IL (interleukin), ILC2s (innate lymphoid type 2 cells), KO (knockout), LYZ (lysozyme), MBP (major basic protein), MPO (myeloperoxidase), MUC2 (mucin 2), pEGFR (phosphorylated epidermal growth factor receptor), scRNA-seq (single-cell RNA sequencing), TCA (tricarboxylic acid), Th (T-helper cell), TNF (tumor necrosis factor)
      See editorial on page 2025.

       Background And Context

      Inflammatory bowel disease incidence is inversely correlated with endemic parasite infestation. Clinical studies have suggested helminth therapy is beneficial in IBD treatment. Intestinal tuft cells modulate anti-parasite responses.

       New Findings

      At homeostasis, heterogeneous tuft cell populations exist and tuft cell frequency is decreased under inflammatory conditions. Tuft cell expansion in models of intestinal inflammation resolved disease.

       Limitations

      Therapeutic tuft cell expansion was only conducted in mouse models. Further studies are necessary to elucidate mechanisms by which tuft cells resolve disease.

       Impact

      Tuft cells are an understudied sector of epithelial biology and could potentially provide new biomarkers and/or therapeutic targets for IBD treatment.
      Global incidences of communicable disease are inversely correlated with rates of inflammatory bowel disease (IBD).
      • Molodecky N.A.
      • Soon I.S.
      • Rabi D.M.
      • et al.
      Increasing incidence and prevalence of the inflammatory bowel diseases with time, based on systematic review.
      Part of the “hygiene hypothesis,” this phenomenon is thought to result from improved hygiene practices associated with decreased tolerance to environmental antigens, such as those from the commensal microbiome.
      • de Silva P.
      • Korzenik J.
      The changing epidemiology of inflammatory bowel disease: identifying new high-risk populations.
      This paradoxical effect has led to emerging interest in the use of helminths, or parasitic worms, for the treatment of IBD.
      • Summers R.W.
      • Elliott D.E.
      • Urban J.F.
      • et al.
      Trichuris suis therapy in Crohn’s disease.
      Enhanced antiparasite immune responses characterized by type 2 cytokines, such as interleukin (IL)25 and IL13, have been shown to suppress T-helper (Th)1 and Th17 activities.
      • Su J.
      • Chen T.
      • Ji X.-Y.
      • et al.
      IL-25 downregulates Th1/Th17 immune response in an IL-10–dependent manner in inflammatory bowel disease.
      ,
      • Broadhurst M.J.
      • Leung J.M.
      • Kashyap V.
      • et al.
      IL-22+ CD4+ T cells are associated with therapeutic Trichuris trichiura infection in an ulcerative colitis patient.
      However, clinical trial data of this type of therapy for IBD have been inconclusive, with many discontinued due to lack of efficacy.
      • Summers R.W.
      • Elliott D.E.
      • Urban J.F.
      • et al.
      Trichuris suis therapy in Crohn’s disease.
      Moreover, helminth therapy has its drawbacks, given that prolonged infection causes complications.
      • Summers R.W.
      • Elliott D.E.
      • Urban J.F.
      • et al.
      Trichuris suis therapy in Crohn’s disease.
      Precision therapies using intermediary products may circumvent most of these issues.
      In acute mouse models of eukaryotic infection, small intestinal tuft cells were found to sense the parasites and orchestrate the antiparasite response.
      • Gerbe F.
      • Sidot E.
      • Smyth D.J.
      • et al.
      Intestinal epithelial tuft cells initiate type 2 mucosal immunity to helminth parasites.
      • Moltke J von
      • Ji M.
      • Liang H.-E.
      • et al.
      Tuft cell derived IL25 regulates an intestinal ILC2-epithelial response circuit.
      • Howitt M.R.
      • Lavoie S.
      • Michaud M.
      • et al.
      Tuft cells, taste-chemosensory cells, orchestrate parasite type 2 immunity in the gut.
      Through the release of IL25, tuft cells promote their own specification via a positive feedback loop mediated by type 2 immune cells, and tuft cell expansion is critical for parasite clearance. These studies examined tuft cells as a single population; however, recent work has revealed substantial heterogeneity within the tuft cell lineage,
      • Haber A.L.
      • Biton M.
      • Rogel N.
      • et al.
      A single-cell survey of the small intestinal epithelium.
      which led us to hypothesize that a distinct subpopulation of tuft cells is responsive to changes in the intestinal luminal environment.
      We investigated the heterogeneity of tuft cell function with regards to their origins of specification and the role of these cell populations in intestinal inflammation. Initial studies identified a secretory cell route of intestinal tuft cell specification, along with barrier-promoting goblet and Paneth cells, under the regulation of the master secretory transcription factor ATOH1.
      • Gerbe F.
      • Es JH Van
      • Makrini L.
      • et al.
      Distinct ATOH1 and Neurog3 requirements define tuft cells as a new secretory cell type in the intestinal epithelium.
      More recent studies have demonstrated that small intestinal tuft cells may have an alternative lineage specification route independent of the ATOH1-controlled secretory lineage.
      • Herring C.A.
      • Banerjee A.
      • Mckinley E.T.
      • et al.
      Unsupervised trajectory analysis of single-cell RNA-Seq and imaging data reveals alternative tuft cell origins in the gut.
      ,
      • Gracz A.D.
      • Fordham M.J.
      • Trotier D.C.
      • et al.
      Sox4 promotes Atoh1-independent intestinal secretory differentiation toward tuft and enteroendocrine fates.
      Here, we demonstrate that ATOH1-independent tuft cells expand upon luminal microbiome perturbation through a metabolic communication network and that this mechanism can be leveraged to suppress inflammation and restore epithelial architecture in small intestinal Crohn’s disease (CD).

      Methods

      All human and animal studies were approved by the Vanderbilt Institutional Review Board and Institutional Animal Care and Use Committee, respectively, in accordance with National Institutes of Health guidelines. Additional information is available in the Supplementary Methods.

      Results

       Reduced Tuft Cell Numbers Are Correlated With Localized Inflammation in the Ileum

      Eukaryotic parasites largely colonize the small intestine, and up to 60% of CD patients have terminal ileal involvement. Thus, we assessed the correlation between intestinal tuft cell numbers and local tissue inflammation in ileal specimens from CD patients (n = 14) compared with normal ileal specimens (n = 11) from patients without a CD diagnosis (Supplementary Figure 1A). Because the canonical mouse tuft cell gene DCLK1 is not expressed in human TRPM5+ tuft cells (Supplementary Figure 1B), we used a previously published strategy for tuft cell identification in both human and mouse using phosphorylated epidermal growth factor receptor (pEGFR; Y1068) and cyclooxygenase 2 (COX2) coexpression.
      • Herring C.A.
      • Banerjee A.
      • Mckinley E.T.
      • et al.
      Unsupervised trajectory analysis of single-cell RNA-Seq and imaging data reveals alternative tuft cell origins in the gut.
      ,
      • McKinley E.T.
      • Sui Y.
      • Al-Kofahi Y.
      • et al.
      Optimized multiplex immunofluorescence single-cell analysis reveals tuft cell heterogeneity.
      Double-positive cells in the normal human ileal epithelium were distinct from single-positive pEGFR or COX2 cells in the lamina propria, and they possessed prominent pEGFR-positive apical “tufts” indicative of genuine tuft cells (Figure 1A).
      Figure thumbnail gr1
      Figure 1Tuft cell number is decreased in inflamed tissue with ileal inflammatory disease. Immunofluorescence (IF) of tuft cell staining in human ilea of (A) healthy and (B) CD patients. Arrows denote coexpression. (C) IF quantification for tuft cells per crypt/villus. SEM for n = 11 normal and n = 14 CD patients. (D) H&E and (EG) IF of cell markers of inflamed and uninflamed distal ilea from mice. (H) IF quantification for marker area LYZ and (I) MUC2 normalized to Hoechst area per crypt. SEM for n = 4 mice for LYZ and n = 3 for MUC2. (J) IF quantification of DCLK1+ tuft cells in wild-type (n = 146 villi) and TNFΔARE/+ villi, by low-grade (n = 11), middle-grade (n = 125), and high-grade (n = 62) inflammation. ∗P < .05, ∗∗P < .01, ∗∗∗P < .001.
      Compared with healthy controls, CD specimens had characteristic, heterogeneous inflammation, with areas of severe villus blunting, increased mucin 2-positive (MUC2+) granules, decreased lysozyme (LYZ+) Paneth cells with diffuse LYZ granule staining, and increased LYZ+ lamina propria immune cells (Supplementary Figure 1CF), consistent with previous reports.
      • Wehkamp J.
      • Wang G.
      • Ku I.
      • et al.
      The Paneth cell alpha-defensin deficiency of ileal Crohn’s disease is linked to Wnt/Tcf-4.
      In parallel, pEGFR+ COX2+ tuft cells were significantly reduced in CD ileal specimens (Figure 1B and C). In regions with less disease involvement and more organized tissue architecture in CD specimens, tuft cells could still be detected (Figure 1B, right). We speculate that suppression of tuft cell specification may contribute to the loss of inflammation control in CD and thus is associated with disease development or progression, or both.
      We then assessed tuft cells in the TNFΔARE/+ mouse model, which develops Crohn’s-like ileitis due to tumor necrosis factor (TNF)-α overexpression.
      • Kontoyiannis D.
      • Pasparakis M.
      • Pizarro T.T.
      • et al.
      Impaired on/off regulation of TNF biosynthesis in mice lacking TNF AU-rich elements: implications for joint and gut-associated immunopathologies.
      In the terminal ilea of 4-month-old animals, we observed distorted crypt structure and blunted villi, decreased number of LYZ1+ Paneth cells, increased MUC2+ granule staining, and increased myeloperoxidase-positive (MPO+) neutrophils infiltration into the lamina propria (Figure 1DI; Supplementary Figure 2A and B), similar to human CD.
      To address regional heterogeneity of inflammation, we performed spatially resolved analysis by quantifying doublecortin-like kinase 1-positive (DCLK1+) tuft cells in wild-type and TNFΔARE/+ villi, stratifying the latter by low-, mid-, and high-grade inflammation (Supplementary Figure 2C). Compared with wild-type, we observed increased tuft cells in low- and midgrade inflamed villi, while high-grade villi in TNFΔARE/+ animals were almost completely devoid of tuft cells (Figure 1J). We confirmed this regional heterogeneity by semiautomated image analysis based on the number of infiltrated MPO+ neutrophils as a surrogate for region-specific inflammation (Supplementary Figure 2DF). We conclude that tuft cell frequency is inversely correlated with severity of disease, raising the possibility that increasing tuft cell specification may reduce inflammation.

       ATOH1-Independent Tuft Cells Are an Inducible Cell Population Responsive to the Commensal Microbiota

      To potentially leverage tuft cells as a strategy for alleviating intestinal inflammation, we sought to understand how they are specified. Previous studies reported heterogeneous tuft cell populations.
      • Haber A.L.
      • Biton M.
      • Rogel N.
      • et al.
      A single-cell survey of the small intestinal epithelium.
      ,
      • Herring C.A.
      • Banerjee A.
      • Mckinley E.T.
      • et al.
      Unsupervised trajectory analysis of single-cell RNA-Seq and imaging data reveals alternative tuft cell origins in the gut.
      ,
      • Gracz A.D.
      • Fordham M.J.
      • Trotier D.C.
      • et al.
      Sox4 promotes Atoh1-independent intestinal secretory differentiation toward tuft and enteroendocrine fates.
      Our previously published Atoh1 knock-out (AtohKO) mouse model (Lrig1CreERT2/+; Atoh1fl/fl) revealed that while colonic tuft cells are ATOH1-dependent, AtohKO triggered significant small intestinal tuft cell expansion (Supplementary Figure 3A), in contrast to their ATOH1-dependence observed in prior studies.
      • Gerbe F.
      • Es JH Van
      • Makrini L.
      • et al.
      Distinct ATOH1 and Neurog3 requirements define tuft cells as a new secretory cell type in the intestinal epithelium.
      Thus, we focus on ATOH1-independent tuft cells as a flexible population that can be induced to expand, presenting a viable target that can be manipulated for the treatment of ileal CD.
      Consistent with the canonical intestinal differentiation hierarchy,
      • Shroyer N.F.
      • Helmrath M.A.
      • Wang V.Y.-C.
      • et al.
      Intestine-specific ablation of mouse atonal homolog 1 (Math1) reveals a role in cellular homeostasis.
      AtohKO intestines possessed normal crypt-villus architecture but lacked MUC2+ goblet cells and LYZ1+ Paneth cells (Supplementary Figure 3B and C). We observed increased MPO+ neutrophils in the villi of AtohKO animals accompanied by decreased weight gain (Supplementary Figure 3D and E), suggesting increased mucosal interaction with the microbiota due to the loss of barrier-forming secretory cell types. Despite increased neutrophils, overt inflammation, characterized by massive immune cell infiltration and villus blunting, was not observed. These observations led us to investigate whether ATOH1-independent tuft cell specification is driven by extrinsic cues originating from the luminal microbiota.
      To investigate the nature of luminal perturbation that results in ATOH1-independent tuft cell expansion in the intestine, we first determined that our mouse colony was devoid of large parasites known to trigger tuft cell expansion
      • Moltke J von
      • Ji M.
      • Liang H.-E.
      • et al.
      Tuft cell derived IL25 regulates an intestinal ILC2-epithelial response circuit.
      ,
      • Howitt M.R.
      • Lavoie S.
      • Michaud M.
      • et al.
      Tuft cells, taste-chemosensory cells, orchestrate parasite type 2 immunity in the gut.
      (Supplementary Figure 3F). To assess the necessity of the commensal microbiota in driving tuft cell expansion, we used a previously published antibiotic cocktail to deplete a broad range of gram-positive and gram-negative bacteria.
      • Meng D.
      • Newburg D.S.
      • Young C.
      • et al.
      Bacterial symbionts induce a FUT2-dependent fucosylated niche on colonic epithelium via ERK and JNK signaling.
      Microbiome depletion with loss of Atoh1 did not affect Paneth cells, but surprisingly suppressed ATOH1-independent tuft cell specification in a dose-dependent manner (Figure 2A and B; Supplementary Figure 3G). Antibiotic administration reduced MPO+ staining but not weight loss, suggesting the former but not the latter was due to microbiome regulation by barrier-forming cells (Supplementary Figure 3E and H). However, in wild-type mice that possess both ATOH1-dependent and -independent tuft cell populations, microbiome depletion with high-dose antibiotics or germ-free housing did not suppress tuft cell specification (Figure 2C and D; Supplementary Figure 3G). We reason that ATOH1-independent tuft cells are sensitive to luminal changes and that coexisting ATOH1-dependent tuft cells are largely insensitive to these changes.
      Figure thumbnail gr2
      Figure 2ATOH1-independent tuft cell expansion is microbiome-dependent. Immunofluorescence (IF) of cell markers in (A and B) AtohKO ilea under increasing doses of antibiotics (−, +, ++, +++), and (C and D) wild-type ilea under antibiotics and germ-free conditions. (E) IF of tuft cells in Lrig1CreERT2/+;Atoh1fl/fl small intestinal enteroids, with or without 4-hydroxytamoxifen (4-OHT) to ablate Atoh1, and with or without exogenous IL13. (F) Fraction of enteroids with 0, 1 to 3, 4 to 10, ≥11 tuft cells under different conditions.
      To verify the heterogeneity of tuft cells in a controlled manner, we turned to enteroids,
      • Sato T.
      • Vries R.G.
      • Snippert H.J.
      • et al.
      Single Lgr5 stem cells build crypt-villus structures in vitro without a mesenchymal niche.
      which are cultured in an environment devoid of microbiota. The sterility of the culturing conditions did not inhibit the specification of tuft cells, consistent with the independence of tuft cell specification and the microbiota when ATOH1 is present in vivo. However, tuft cell specification in sterile enteroids was suppressed when Atoh1 loss was induced via Cre recombination (Figure 2E and F), demonstrating that tuft cell specification observed in the control condition was ATOH1 dependent. As expected, loss of Atoh1 led to suppression of Paneth and goblet cell specification (Supplementary Figure 3I and J).
      To confirm that ATOH1-independent tuft cells respond to luminal perturbations, we perused the literature and identified that the type 2 cytokine IL13 drives tuft cell specification in response to luminal eukaryotic colonization.
      • Moltke J von
      • Ji M.
      • Liang H.-E.
      • et al.
      Tuft cell derived IL25 regulates an intestinal ILC2-epithelial response circuit.
      ,
      • Howitt M.R.
      • Lavoie S.
      • Michaud M.
      • et al.
      Tuft cells, taste-chemosensory cells, orchestrate parasite type 2 immunity in the gut.
      In wild-type culture, IL13 stimulated tuft cell specification, beyond the 0 to 3 number typically observed at baseline (Figure 2E and F). While AtohKO enteroids initially lacked tuft cells, IL13 administration induced tuft cell specification and expansion in an ATOH1-independent manner at levels similar to IL13-treated wild-type enteroids (Figure 2E and F; Supplementary Figure 3K). These results support the existence of ATOH1-independent and -dependent tuft cell populations, with the former being highly malleable to luminal cues.

       Trajectory Analysis of Single-Cell RNA Sequencing Data Supports an ATOH1-Independent Tuft Cell Specification Pathway

      To investigate tuft cell heterogeneity and associated pathways, we generated single-cell RNA sequencing (scRNA-seq) data from wild-type control, TNFΔARE/+, antibiotic-treated wild-type (ATOH1-dependent tuft cells only), and AtohKO (ATOH1-independent tuft cells only) ileal epithelium across multiple biological replicates (Supplementary Figure 4AD). t-distributed stochastic neighbor embedding (t-SNE) and clustering analyses performed on each of the combined data sets demonstrated that all clusters were represented in all replicates (Supplementary Figures 4EL and 5). Stem and progenitor cells, enterocytes, tuft cells, goblet cells, Paneth cells, and enteroendocrine cells (or the lack thereof in AtohKO) were present in the correct proportions in all conditions (Figure 3AD; Supplementary Figure 4IL). The exception here was the TNFΔARE/+ data sets due to the admixing of regionally heterogeneous inflamed and uninflamed areas, as well as protein products, such as secretory MUC2+ granules, not necessarily reflecting cell types. For tuft cells, their numbers were significantly reduced in the antibiotic-treated wild-type ileum, consistent with the persistence of few ATOH1-dependent tuft cells when ATOH1-independent tuft cells were eliminated (Figure 3C and E). In contrast, AtohKO tuft cells were significantly expanded (Figure 3D and E). These results were largely consistent with cell type distributions obtained by microscopy analysis.
      Figure thumbnail gr3
      Figure 3Trajectory analysis of scRNA-seq data supports alternative origins for ATOH1-dependent and ATOH1-independent tuft cells. (AD) T-distributed stochastic neighbor embedding analysis of scRNA-seq data generated from different murine models annotated with cell type cluster (n = 6 wild-type, n = 3 TNFΔARE/+, n = 2 antibiotic-treated wild-type, n = 3 AtohKO). (E) Tuft cell percentage quantification from scRNA-seq. SEM for replicates plotted. (FI) Top scoring p-Creode topologies of scRNA-seq data with lineage annotation. The node size represents cell state density. (JM) Quantification of n = 100 p-Creode topology maps for nonsecretory or secretory tuft cell placement. ∗P < .05, ∗∗P < .01.
      We next analyzed tuft cell specification pathways using the p-Creode algorithm (https://github.com/KenLauLab/pCreode) to produce lineage trajectory representations of scRNA-seq data (Figure 3FI; Supplementary Figures 6 and 7).
      • Herring C.A.
      • Banerjee A.
      • Mckinley E.T.
      • et al.
      Unsupervised trajectory analysis of single-cell RNA-Seq and imaging data reveals alternative tuft cell origins in the gut.
      The wild-type p-Creode map originated from stem cells and bifurcated into the absorptive and secretory lineages, with goblet and Paneth cells originating from a common secretory progenitor (Figure 3F). In contrast, the tuft cell lineage mainly shared a specification trajectory with absorptive cells, rather than secretory cells.
      To evaluate the robustness of this result, we generated bootstrapped p-Creode graphs by resampling the data set and quantified tuft cell placement into different lineages. Tuft cell placement was nonsecretory in 83% of wild-type trajectories and secretory in 17% (Figure 3J). Focused analysis of the tuft cell population in the wild-type ileum revealed 2 subclasses with divergent metabolism-related gene expression programs (Supplementary Figure 8A and B). These results are consistent with previous work documenting multiple tuft cell populations,
      • Haber A.L.
      • Biton M.
      • Rogel N.
      • et al.
      A single-cell survey of the small intestinal epithelium.
      which may account for the varying tuft cell lineage placement in p-Creode trajectories (Figure 3J).
      We repeated p-Creode analysis to include rare enteroendocrine cells, as well as using another data set by Haber et al,
      • Haber A.L.
      • Biton M.
      • Rogel N.
      • et al.
      A single-cell survey of the small intestinal epithelium.
      and observed similar tuft cell placement results (Supplementary Figures 9 and 10). p-Creode analysis of the TNFΔARE/+ data set generated similar results, illustrating that even under inflammatory conditions, tuft cells can share a trajectory with absorptive cells (Figure 3G and K). These results show that tuft cell lineage branching from a nonsecretory route was a robust and consistent feature of small intestinal cell differentiation across multiple data sets. Given the frequency of tuft cell placement outside the ATOH1-dependent secretory lineage, we surmise that ATOH1-independent tuft cells exist and in a microbiome-replete intestinal environment, account for most of the small intestinal tuft cells.
      Our p-Creode analysis of the antibiotic-treated wild-type data set, which we hypothesized to contain only ATOH1-dependent tuft cells, revealed that these tuft cells share a trajectory almost exclusively with secretory cells, with 99% placement with the secretory lineage (Figure 3H and L). Differential gene expression analysis of antibiotic-treated wild-type tuft cells revealed increased expression of genes associated with secretory cells, including Sox9, Muc2, and alpha-defensin Defa22 (Supplementary Figure 8CE).
      • Bastide P.
      • Darido C.
      • Pannequin J.
      • et al.
      Sox9 regulates cell proliferation and is required for Paneth cell differentiation in the intestinal epithelium.
      ,
      • Farin H.F.
      • Karthaus W.R.
      • Kujala P.
      • et al.
      Paneth cell extrusion and release of antimicrobial products is directly controlled by immune cell-derived IFN-γ.
      In addition, t-SNE analysis showed that these tuft cells are clustered with enteroendocrine cells, although their memberships do not overlap (Figure 3C), consistent with prior work demonstrating shared differentiation of enteroendocrine cells and a tuft cell subset from common Prox1+ progenitors.
      • Yan K.S.
      • Gevaert O.
      • Zheng G.X.Y.
      • et al.
      Intestinal enteroendocrine lineage cells possess homeostatic and injury-inducible stem cell activity.
      Similarly, ATOH1-dependent colonic tuft cells shared similar topologic relationships with colonic enteroendocrine cells (Supplementary Figure 8F and G).
      Finally, all bootstrapped p-Creode graphs generated from AtohKO data depicted tuft cells and absorptive cells as originating from a common progenitor (Figure 3I and M), again demonstrating their ATOH1-independence. Critically, these ATOH1-independent tuft cells expressed the canonical tuft cell gene signature derived from literature
      • Moltke J von
      • Ji M.
      • Liang H.-E.
      • et al.
      Tuft cell derived IL25 regulates an intestinal ILC2-epithelial response circuit.
      (Supplementary Figures 7D and 8H), showing that they are not a damage-induced stem cell lineage; however, they had decreased expression of secretory genes compared with ATOH1-dependent tuft cells (Supplementary Figure 8CE). These results further support the existence of ATOH1-dependent (secretory) and ATOH1-independent (nonsecretory) tuft cell populations, where the ATOH1-independent population is sensitive to luminal cues and can be induced to expand. We sought to leverage the AtohKO condition to identify signals that drive ATOH1-independent tuft cell expansion.

       Alterations in Tricarboxylic Acid Metabolic Pathways Are Associated With Tuft Cell Expansion in a Microbiome-Dependent Manner

      To identify molecular pathways altered in the course of ATOH1-independent tuft cell expansion, we focused our analysis on dynamic alterations in gene expression along the tuft cell lineage between the wild-type and AtohKO condition, with the latter having significant tuft cell expansion. This approach circumvents technical batch effects, because the dynamics of gene expression along a trajectory are self-contained within individual analyses. Examples illustrating lineage-specific marker expression dynamics are shown in Supplementary Figure 11A.
      We aimed to identify genes in the tuft cell lineage that switch their expression dynamics between the wild-type and AtohKO conditions. First, different gene dynamics trends were broadly classified into 4 categories, as well as a fifth category of unchanged or “flat” dynamics (Figure 4A). Group 1 genes, exemplified by Soux, trended upward along pseudotime of the stem-to-tuft cell trajectory and included genes known to be tuft cell markers (eg, Ptgs1 and Sox9) (Supplementary Figure 11B).
      • McKinley E.T.
      • Sui Y.
      • Al-Kofahi Y.
      • et al.
      Optimized multiplex immunofluorescence single-cell analysis reveals tuft cell heterogeneity.
      Group 4 genes, illustrated by Rps6, trended downward and included stem cell markers downregulated during differentiation. Intermediate genes that trended upward but returned to a lower baseline or those that initially trended downwards but returned to a higher baseline were categorized into groups 2 and 3, respectively.
      Figure thumbnail gr4
      Figure 4Analysis of AtohKO tuft cell gene expression identified upregulation in metabolic pathways. (A) Heat map of gene expression trends grouped into different dynamic trends as denoted in . (B) KEGG enrichment of genes that class switched from a lower order to a higher order in AtohKO, ordered by normalized enrichment score. (CH) Representative TCA cycle–related gene trends over the tuft cell lineage pseudotime. The solid lines represent mean expression trends, and the dashed lines represent confidence intervals fitted to raw data from 10 top-scoring p-Creode topologies. The data points are scaled expression data. (I) Top 20 gene set enrichment analysis (GSEA) gene sets enriched in AtohKO tuft cell transcriptomes, along with (JM) positive enrichment plots with highest normalized enrichment score. GO, gene ontology. (NS) Scaled TCA cycle gene expression in tuft cells. ∗P < .05, ∗∗P < .01, ∗∗∗∗P < .0001.
      When all expressed genes between the wild-type and AtohKO were visualized within these categories, a broad expansion of group 2 genes was observed upon loss of Atoh1 (Figure 4A). To identify gene expression changes accompanying tuft cell expansion in an unbiased manner, we extracted 1755 genes that were positively enriched in the AtohKO group, namely, those that switched categories from a lower category in the wild-type data to a higher category in AtohKO (Supplementary Figure 11C).
      Over-representation analysis of positively enriched genes in the AtohKO epithelium identified pathways related to the tricarboxylic acid (TCA) cycle and oxidative phosphorylation based on the Kyoto Encyclopedia of Genes and Genomes (KEGG), WikiPathways, and Reactome databases (Figure 4B; Supplementary Figure 11D and E). Simplifying the analysis by reclassifying genes as having only “upward” or “downward” dynamics generated similar enrichment of metabolism-associated processes (Supplementary Figure 11FH).
      We confirmed changes in TCA gene dynamics in the AtohKO tuft cell lineage by plotting expression trends fit to raw data from 10 representative p-Creode maps. Tuft (Dclk1, Trpm5) and progenitor (Myc, Pcna) signature genes served as controls, showing unaltered dynamics between conditions (Figure 4CH; Supplementary Figure 11I). As an example of genes with altered dynamics, the TCA cycle enzyme gene Mdh2 trended down along the wild-type tuft cell specification trajectory, while its expression remained constant along the AtohKO trajectory (Figure 4D). Similarly, other TCA enzymes, such as Idh3a, Sdhb, and Sdhd, and downstream electron transport chain genes, such as those coding for reduced nicotinamide adenine dinucleotide dehydrogenases and adenosine 5ʹ-triphosphate synthases, all switched to more positive dynamic trends along the AtohKO tuft cell trajectory compared with wild-type (Figures 4CH; Supplementary Figure 12).
      These results were confirmed by grouping cells within the tuft cell trajectory and performing standard differential expression analysis, followed by gene set enrichment analysis between wild-type and AtohKO cells, using a variety of databases such as Molecular Signatures Database (MSigDB), KEGG, PANTHER, and WikiPathways (Figure 4IM; Supplementary Figure 13AF). Similar to the dynamic trend analysis, TCA cycle–related genes Idh3b, Mdh2, Sdha, Sdhb, Sdhc, Sdhd, Citrate synthase (Cs), Idh3a, and Mdh1 were significantly increased in AtohKO tuft cells compared with wild-type (Figure 4NS; Supplementary Figure 13GL).
      To support the role of the microbiome in inducing these changes, we generated scRNA-seq data from AtohKO animals after antibiotic administration where microbiome depletion suppresses tuft cell expansion (Figure 2B). In the middle-dose condition (++), tuft cell expression of Idh3a, Idh3b, Idh3g, Sdha, and Sdhd was significantly reduced, whereas others, such as Mdh1, Mdh2, and Sdhc, trended downward (Supplementary Figure 13MX). These results indicate that ATOH1-independent tuft cell expansion is accompanied by increases in TCA cycle gene expression, which is driven by extrinsic signals from the intestinal microbiome.

       Nonparasite-Derived Sources of Succinate in the Luminal Environment Drive Tuft Cell Expansion

      Given the microbiome dependency and metabolic gene upregulation observed in ATOH1-independent tuft cell expansion, we sought to characterize the differences in (1) metabolites and (2) microbiota composition in the intestines of wild-type and AtohKO animals. Mass spectrometry analysis with O-benzylhydroxylamine revealed that the relative concentration of succinate was significantly increased in the AtohKO cecal luminal contents but not in tissue, whereas levels of malate and butyrate were not altered (Figure 5AC).
      • Tan B.
      • Lu Z.
      • Dong S.
      • et al.
      Derivatization of the tricarboxylic acid intermediates with O-benzylhydroxylamine for liquid chromatography–tandem mass spectrometry detection.
      Succinate is a metabolic intermediate in the TCA cycle and is converted by the enzyme succinate dehydrogenase into fumarate, which trended upward in the AtohKO tissue, suggestive of host metabolic processing of luminal succinate (Figure 5D). Moreover, the disparity in succinate concentration between luminal contents and whole tissue was indicative of a commensal microbiota-derived rather than a host-derived origin for succinate (Figure 5A). Thus, we repeated the analysis in the AtohKO condition after microbiome depletion, which resulted in a significant decrease in luminal succinate, confirming that the commensal microbiota was primarily responsible for increased succinate production in the AtohKO model (Figure 5E).
      Figure thumbnail gr5
      Figure 5Succinate production in the AtohKO small intestine drives ATOH1-independent tuft cell expansion. (AE) Mass spectrometric measurements of metabolites from the cecal lumen and tissue. SEM for n = 5 wild-type, n = 3 AtohKO, and n = 3 antibiotic-treated AtohKO animals. (F) Heat map of z-score normalized PICRUSt category scores between wild-type and AtohKO. P < .05 for all categories. (G) Relative abundance of genus contributing to “chlorocyclohexane and chlorobenzene degradation” category. SEM for n = 4 wild-type and n = 3 AtohKO. (H) IF for cell markers of germ-free animals gavaged with wild-type or AtohKO contents at 3 or 7 days. (I) Raw and (J) normalized (to nucleus count) tuft cell number in untreated and oral-gavaged animals. Con, control; SPF, specific pathogen free; WT, wild-type. Data points represent fields of view and SEM across multiple biological replicates calculated. ∗∗P < .01, ∗∗∗∗P < .0001.
      We verified that exogenous succinate administration in wild-type animals induced tuft cell expansion, whereas major basic protein (MBP)-positive eosinophils and GATA3+ cells, components of the antiparasite response, were increased in both succinate-treated wild-type and AtohKO intestinal tissues (Supplementary Figure 14AC).
      • Schneider C.
      • O’Leary C.E.
      • Moltke J Von
      • et al.
      A metabolite-triggered tuft cell-ILC2 circuit drives small intestinal remodeling.
      • Lei W.
      • Ren W.
      • Ohmoto M.
      • et al.
      Activation of intestinal tuft cell-expressed Sucnr1 triggers type 2 immunity in the mouse small intestine.
      • Nadjsombati M.S.
      • McGinty J.W.
      • Lyons-Cohen M.R.
      • et al.
      Detection of succinate by intestinal tuft cells triggers a type 2 innate immune circuit.
      • Yamashita M.
      • Ukai-Tadenuma M.
      • Miyamoto T.
      • et al.
      Essential role of GATA3 for the maintenance of type 2 helper T (Th2) cytokine production and chromatin remodeling at the Th2 cytokine gene loci.
      To confirm that succinate is the intermediary downstream of the microbiome, we evaluated antibiotic-treated AtohKO animals, which are devoid of tuft cells, for the ability of succinate alone to rescue tuft cell expansion in a depleted microbiome environment. Succinate administration restored tuft cell expansion to microbiome replete levels in the small intestine of these animals (Supplementary Figure 14D), demonstrating that this TCA cycle metabolite alone is sufficient to drive specification of ATOH1-independent tuft cells.
      To probe the response of tuft cells in response to succinate, we performed scRNA-seq in the ilea of succinate-treated wild-type mice. Consistent with immunofluorescent imaging, quantification of scRNA-seq data showed induced tuft cell expansion by exogenous succinate administration (Supplementary Figure 15AC). Consistent with TCA cycle gene induction in other conditions of tuft cell expansion, expression of Idh3g, Mdh2, Sdhb, Sdhc, Ogdh, and Rpl18a were significantly increased in succinate-treated tuft cells, whereas Cs and Idh3b trended upward (Supplementary Figure 15DL). This analysis provided an opportunity to investigate lineage-specific induction of these genes. Expression of Idh3g, Sdhb, Sdhc, Rpl18a, and Idh3b was similarly increased in cells of the absorptive lineage by exogenous succinate administration, again confirming that these cells may share similar biology (Supplementary Figure 15DO). In contrast, secretory lineage-specific expression of most of these genes was unchanged or even decreased. Differentiation trajectory and tuft cell lineage placement were similar to untreated wild-type (Supplementary Figure 15P and Q). These results are consistent with the ability of ATOH1-independent tuft cells outside of the secretory lineage to respond to luminal succinate.
      We also queried the effect of succinate administration in the colon, where tuft cells were shown to be ATOH1-dependent. Wild-type colon was not responsive to succinate, regarding both tuft cell expansion or downstream MBP+ type 2 cell infiltration (Supplementary Figure 16A and B). Colonic tuft cells do not express the succinate receptor (Sucnr1) (Supplementary Figure 16C).
      • Moltke J von
      • Ji M.
      • Liang H.-E.
      • et al.
      Tuft cell derived IL25 regulates an intestinal ILC2-epithelial response circuit.
      ,
      • Nadjsombati M.S.
      • McGinty J.W.
      • Lyons-Cohen M.R.
      • et al.
      Detection of succinate by intestinal tuft cells triggers a type 2 innate immune circuit.
      In antibiotic-treated AtohKO animals, where ATOH1-dependent colonic tuft cells are absent, succinate administration did not restore tuft cell specification (Supplementary Figure 16D). Succinate administration to wild-type mice did not trigger TCA cycle genes in colonic tuft cells (Supplementary Figure 16EP). These results demonstrate that commensal-derived metabolic signals, specifically succinate, are necessary and sufficient to induce ATOH1-independent tuft cell expansion in the small intestine but do not affect ATOH1-dependent tuft cells in the colon.
      Because the microbiome was necessary for in vivo tuft cell expansion, we used sequencing of the V4 region of the 16S ribosomal RNA gene to investigate altered microbiome distribution in the ileal luminal contents of cohoused wild-type and AtohKO littermates. Quality control analyses of sequencing data demonstrated that wild-type and AtohKO replicates clustered together due to biological variation rather than cage effects (Supplementary Figure 17AF). Analysis of microbiome composition revealed a decrease in genus Barnesiella within AtohKO replicates compared with wild-type, whereas the relative abundance of Parasutterella and Bifidobacterium was increased (Supplementary Figure 17GI). Bifidobacterium infantis, B breve, and B pseudolongum are components of the VSL#3 (Sigma-Tau Pharmaceuticals, Inc, Gaithersburg, MD) probiotic that can induce remission in a subset of patients with active IBD.
      • Bibiloni R.
      • Fedorak R.N.
      • Tannock G.W.
      • et al.
      VSL#3 probiotic-mixture induces remission in patients with active ulcerative colitis.
      We performed PICRUSt (Phylogenetic Investigation of Communities by Reconstruction of Unobserved States) analysis on our 16S gene sequence data and identified 8 functional categories that were positively enriched in the AtohKO microbiome, including “chlorocyclohexane and chlorobenzene degradation” and “retinol metabolism” (Figure 5F).
      • Langille M.G.I.
      • Zaneveld J.
      • Caporaso J.G.
      • et al.
      Predictive functional profiling of microbial communities using 16S rRNA marker gene sequences.
      A simplified diagram of the chlorocyclohexane and chlorobenzene degradation pathway (ko00361) shows that this process is associated with succinate production (Supplementary Figure 17J).
      Further analysis revealed that Bifidobacterium, Lactobacillus, Sutterellla, Acinetobacter, and Akkermansia contributed to the positive enrichment of this pathway in the AtohKO microbiome; however, only Bifidobacterium was significantly increased compared with the wild-type microbiome (Figure 5G). Specifically, we observed that B pseudolongum, a known producer of succinic acid,
      • Meulen R Van der
      • Adriany T.
      • Verbrugghe K.
      • et al.
      Kinetic analysis of bifidobacterial metabolism reveals a minor role for succinic acid in the regeneration of NAD+ through its growth-associated production.
      was increased 6-fold in the AtohKO microbiome (Supplementary Figure 17K).
      To confirm the contribution of the microbiome to tuft cell expansion, we transferred the cecal microbiome from AtohKO or wild-type animals into germ-free wild-type animals for a short-term (3-day) and long-term (7-day) period, with the caveat that the intact antimicrobial functions of Paneth and goblet cells may counteract effects of the inoculum. DCLK1+ tuft cells were significantly increased in both the duodenum and the ileum 3 days post-inoculation with AtohKO contents compared with controls (uncolonized germ free or specific pathogen free) or germ-free mice colonized with wild-type contents, although the increase was not as pronounced as other tuft cell expansion conditions (Figure 5HJ; Supplementary Figure 18AF). After 7 days, however, tuft cell numbers in the duodenum or the ileum were not significantly different between colonized wild-type and AtohKO animals, consistent with previous results.
      • McKinley E.T.
      • Sui Y.
      • Al-Kofahi Y.
      • et al.
      Optimized multiplex immunofluorescence single-cell analysis reveals tuft cell heterogeneity.
      In contrast, colonic tuft cells did not increase from AtohKO content gavage under any conditions (Supplementary Figure 18GL). 16S sequencing of the gavage inocula and intestinal lumen contents in postgavage animals showed that B pseudolongum trended upward in the AtohKO inoculum compared with the wild-type inoculum and was enriched at 3 days but not 7 days postgavage in the small intestine of AtohKO-gavaged animals compared with wild-type–gavaged controls (Supplementary Figure 18M). These findings reveal the metabolic potential of certain commensal communities to drive ATOH1-independent tuft cell expansion.

       Succinate Administration Ameliorates Inflammation in the TNFΔARE/+ Model

      Given the hypothesis that expanding tuft cells can suppress inflammatory disease, we administered succinate in the drinking water of adult TNFΔARE/+ mice after disease onset to activate the microbiome-responsive ATOH1-independent population. Succinate administration in TNFΔARE/+ animals markedly improved intestinal tissue organization compared with age-matched untreated TNFΔARE/+ controls, based on restored crypt-villus architecture, minimized villus distortion, inflammation, and injury scored by a pathologist (Figure 6AC). Inflammation-associated immune cell subsets enhanced in the TNFΔARE/+ model,
      • Kontoyiannis D.
      • Pasparakis M.
      • Pizarro T.T.
      • et al.
      Impaired on/off regulation of TNF biosynthesis in mice lacking TNF AU-rich elements: implications for joint and gut-associated immunopathologies.
      including MPO+ neutrophils and forkhead box P3-positive (FOXP3+) T-regulatory cells, were significantly reduced in long-term succinate-treated animals, indicative of decreased infiltrative disease (Figure 6DG).
      Figure thumbnail gr6
      Figure 6Succinate administration enhances antiparasitic immune response to counteract inflammation in TNFΔARE/+ animals. (A) H&E of TNFΔARE/+ ilea with succinate. (B and C) Histopathologic scoring of control (n = 5) and long-term succinate-treated (n = 7) TNFΔARE/+ mice. SEM plotted. (D and E) Immunofluorescence (IF) of immune cell markers. (F and G) IF quantification of immune cell types normalized to nuclei area. Each dot represents a field of view and SEM across n = 4 animals. (HN) Luminex cytokine measurements from ileal tissues. SEM across multiple biological replicates calculated (circles, males; triangles, females). (O and P) IF of type 2 immune markers. (QS) IF quantification of nuclear GATA3+ cells in the lamina propria. (T and U) IF of type 17 immune markers. (VX) IF quantification of nuclear RORGT+ cells in the lamina propria. Data points represent fields of view and SEM from multiple biological replicates calculated. ∗P < .05, ∗∗P < .01, ∗∗∗P < .001. ∗∗∗∗P < .0001.
      We used multiplex Luminex (Austin, TX) to assay cytokine levels of short-term succinate-treated animals to capture the initial changes to immune responses. Succinate-induced decrease in inflammation was not a result of decreased TNF-α levels in the TNFΔARE/+ model (Supplementary Figure 19A). Short-term succinate administration induced IL27, which inhibits type 17 responses,
      • Stumhofer J.S.
      • Laurence A.
      • Wilson E.H.
      • et al.
      Interleukin 27 negatively regulates the development of interleukin 17-producing T helper cells during chronic inflammation of the central nervous system.
      and suppressed IL23 expression, which drives differentiation of Th17 cells
      • Aggarwal S.
      • Ghilardi N.
      • Xie M.H.
      • et al.
      Interleukin-23 promotes a distinct CD4 T cell activation state characterized by the production of interleukin-17.
      (Figure 6H and I). Similar to past reports that showed IL17+ IL22+ cell increases in helminth treatment of IBD,
      • Broadhurst M.J.
      • Leung J.M.
      • Kashyap V.
      • et al.
      IL-22+ CD4+ T cells are associated with therapeutic Trichuris trichiura infection in an ulcerative colitis patient.
      we observed significant increases in IL17 and IL22 in succinate-treated animals, with the latter cytokine also shown to enhance mucosal regeneration and worm clearance (Figure 6J and K; Supplementary Figure 19B).
      • Turner J.-E.
      • Stockinger B.
      • Helmby H.
      IL-22 mediates goblet cell hyperplasia and worm expulsion in intestinal helminth infection.
      Canonically, host response to eukaryotic infection is facilitated by type 2 cytokines. IL25 is released by epithelial tuft cells, and IL13 and IL4 are released by innate lymphoid type 2 cells (ILC2s).
      • Moltke J von
      • Ji M.
      • Liang H.-E.
      • et al.
      Tuft cell derived IL25 regulates an intestinal ILC2-epithelial response circuit.
      Short-term succinate administration in TNFΔARE/+ animals significantly increased levels of IL25, IL4, and IL13, with other antiparasite cytokines trending upward (Figures 6LN; Supplementary Figure 19CE). Finally, IL21, which is enhanced in patients with active CD, trended downward with succinate administration (Supplementary Figure 19F).
      • Holm T.L.
      • Tornehave D.
      • Søndergaard H.
      • et al.
      Evaluating IL-21 as a potential therapeutic target in Crohn’s disease.
      Cytokine analysis suggested that succinate-driven suppression of inflammation occurs due to a decrease in proinflammatory type 17 immunity and an activation of an anti-inflammatory profile analogous to an antiparasite immune response.
      To complement the cytokine analysis, we used imaging cytometry to examine cells involved in antiparasite type 2 responses in spatially heterogeneous inflamed tissues. Directing our analysis to lamina propria immune cells (Supplementary Figure 19G),
      • McKinley E.T.
      • Sui Y.
      • Al-Kofahi Y.
      • et al.
      Optimized multiplex immunofluorescence single-cell analysis reveals tuft cell heterogeneity.
      we focused on type 2 cells that are CD3+ (Th2) and CD3 (ILC2), first using a lineage panel to exclude B cells, erythrocytes, and myeloid/dendritic cells (Supplementary Figure 19HJ) as well as non-nuclear GATA3 in the inflammation-activated submucosal stroma (Supplementary Figure 19K).
      • Yamashita M.
      • Ukai-Tadenuma M.
      • Miyamoto T.
      • et al.
      Essential role of GATA3 for the maintenance of type 2 helper T (Th2) cytokine production and chromatin remodeling at the Th2 cytokine gene loci.
      While both inflamed ilea of untreated TNFΔARE/+ animals and uninflamed ilea of wild-type animals had low numbers of CD3+/GATA3+ Th2 cells and CD3/GATA3+ ILC2s, short- and long-term succinate treated animals both had increased levels of these cell types, with ILC2s significantly increased (Figure 6OS; Supplementary Figure 19LM). We applied a similar methodology to examine inflammatory type 17 cells through the transcription factor RORGT. RORGT+ cells, including both CD3+/RORGT+ Th17 cells and CD3/RORGT+ ILC3s, which were rare in uninflamed tissues, were increased in inflamed TNFΔARE/+ animals and were subsequently reduced and returned to baseline with short- and long-term succinate administration (Figure 6TX; Supplementary Figure 19N and O). Thus, enhanced tuft cell specification and a commensurate increase in antiparasite responses is implicated in the mechanism of succinate-mediated suppression of a type 17 inflammatory response.

       Tuft Cells Are Necessary for Succinate-Induced Suppression of Inflammation

      Succinate administration to the TNFARE/+ model completely restored epithelial organization. LYZ1-expressing cells were restored to the base of the crypts and exhibited the typical Paneth cell morphology, while lamina propria LYZ1 expression was absent (Figure 7A). OLFM4+ stem cells were expanded beyond the +4 crypt position upon short-term succinate administration, consistent with ongoing restitution of inflammation-induced epithelial damage, and returned to their normal positions long-term (Supplementary Figure 19P and Q). As expected, DCLK1+ tuft cells were increased by succinate administration in TNFARE/+ animals (Figures 7B). To demonstrate universality, we evaluated another model of intestinal inflammation whereby the administration in an anti-CD3E antibody results in infiltrative disease of the small intestine (Figure 7C; Supplementary Figure 20).
      • Miura N.
      • Yamamoto M.
      • Fukutake M.
      • et al.
      Anti-CD3 induces bi-phasic apoptosis in murine intestinal epithelial cells: possible involvement of the Fas/Fas ligand system in different T cell compartments.
      Figure thumbnail gr7
      Figure 7Tuft cells are necessary for succinate-mediated inflammation suppression. (A and B) Immunofluorescence (IF) of cell markers of TNFΔARE/+ ilea with succinate. The white arrows mark DCLK1+ tuft cells. (C) Percentage body weight change in succinate-treated animals in the anti-CD3E model. SEM across multiple biological replicates calculated. (DG) H&E and IF of cell markers in anti–CD3E-treated mice, with or without tuft cells and succinate. (H) IF quantification of MPO+ cells normalized to nuclei area. Data points represent fields of view and SEM across multiple biological replicates calculated. (I) Summary diagram. ∗∗P < .01, ∗∗∗P < .001.
      Succinate administration had protective effects in animals treated with the anti-CD3E agent because these animals had significantly less weight loss, tissue destruction based on histology, immune cell infiltration based on the reduced presence of MPO+ neutrophils, and restored tuft cell numbers (Figure 7CE and H; Supplementary Figure 20A and B). To probe the necessity of tuft cells in succinate-induced inflammation suppression, we repeated these experiments in a Pou2f3-null mouse model, where intestinal and colonic tuft cells were absent and could not be induced by succinate (Supplementary Figure 20E and F).
      • Yamashita J.
      • Ohmoto M.
      • Yamaguchi T.
      • et al.
      Skn-1a/Pou2f3 functions as a master regulator to generate Trpm5-expressing chemosensory cells in mice.
      Succinate failed to rescue anti–CD3E-driven tissue destruction, weight loss, and neutrophil infiltration in Pou2f3-null animals (Figure 7C and FH; Supplementary Figure 20C and D). These findings implicate a causal role for the succinate-tuft cell specification axis in modulating small intestinal inflammation.

      Discussion

      Chemosensory tuft cells have recently been identified to be necessary and sufficient to drive type 2 immune responses against eukaryotic colonization, possibly through parasite-derived succinate.
      • Gerbe F.
      • Sidot E.
      • Smyth D.J.
      • et al.
      Intestinal epithelial tuft cells initiate type 2 mucosal immunity to helminth parasites.
      ,
      • Schneider C.
      • O’Leary C.E.
      • Moltke J Von
      • et al.
      A metabolite-triggered tuft cell-ILC2 circuit drives small intestinal remodeling.
      • Lei W.
      • Ren W.
      • Ohmoto M.
      • et al.
      Activation of intestinal tuft cell-expressed Sucnr1 triggers type 2 immunity in the mouse small intestine.
      • Nadjsombati M.S.
      • McGinty J.W.
      • Lyons-Cohen M.R.
      • et al.
      Detection of succinate by intestinal tuft cells triggers a type 2 innate immune circuit.
      Despite being a global health concern, helminths have been proposed as a therapeutic option for CD.
      • Summers R.W.
      • Elliott D.E.
      • Urban J.F.
      • et al.
      Trichuris suis therapy in Crohn’s disease.
      ,
      • Broadhurst M.J.
      • Leung J.M.
      • Kashyap V.
      • et al.
      IL-22+ CD4+ T cells are associated with therapeutic Trichuris trichiura infection in an ulcerative colitis patient.
      Prevailing thought postulates that the antiparasite immune response may counteract proinflammatory signaling driving CD.
      • Summers R.W.
      • Elliott D.E.
      • Urban J.F.
      • et al.
      Trichuris suis therapy in Crohn’s disease.
      We observed decreased numbers of tuft cells in inflamed ileal tissues from CD patients and mouse models, and thus, we hypothesized that tuft cells may be the conduit between parasite and host that can be leveraged for counteracting proinflammatory signals in the intestine.
      We identified ATOH1-dependent and ATOH1-independent routes of specification that result in heterogeneous tuft cell populations (Figure 7I). ATOH1-dependent and ATOH1-independent tuft cells both exist in the small intestine, whereas only ATOH1-dependent tuft cells exist in the colon. Surprisingly, we found that ATOH1-independent tuft cells are a malleable cell population that can be expanded in the context of luminal perturbation, whereas the ATOH1-dependent population is invariant.
      Specifically, succinate derived from the commensal microbiome drives ATOH1-independent tuft cell gene expression and expansion; these may include several succinic acid producers of the Bifidobacterium genus known to maintain intestinal health.
      • Langille M.G.I.
      • Zaneveld J.
      • Caporaso J.G.
      • et al.
      Predictive functional profiling of microbial communities using 16S rRNA marker gene sequences.
      ,
      • Stumhofer J.S.
      • Laurence A.
      • Wilson E.H.
      • et al.
      Interleukin 27 negatively regulates the development of interleukin 17-producing T helper cells during chronic inflammation of the central nervous system.
      Given the potential transmissibility of the tuft cell expansion phenotype by microbiome transfer, our results further revealed that the downstream metabolite succinate from a nonparasitic source is necessary and sufficient to drive ATOH1-independent tuft cell specification. Administration of succinate after disease onset in 2 mouse models suppressed inflammation and restored epithelial architecture, accompanied by enhanced antiparasite and reduced type 17 response. How succinate induces cytokine gene expression to initiate downstream responses remains an area of study, with possible hypotheses related to signaling pathway- or metabolite-regulation of the epithelial epigenome.
      Our findings shed light on the heterogeneity of small intestinal tuft cells and their functional role in reducing intestinal inflammation through luminal succinate. Beyond succinate, our findings suggest that enhanced tuft cell specification could be a viable therapeutic strategy, where exploring the mechanisms linking tuft cells to anti-inflammatory signaling could lead to clinically-viable targets for CD therapy.

      Acknowledgments

      The authors would like to thank Cherie’ Scurrah and Paige Vega in the Lau laboratory for support in mouse experiments, Drs Seth Bordenstein, Andrew Brooks, Nicholas Markham, Oliver McDonald, Izumi Kaji, James Goldenring, Eunyoung Choi, Joseph Roland, Damian Maseda, and Mariana Byndloss, the Vanderbilt Digestive Disease Research Center, and the Vanderbilt Epithelial Biology Center for helpful discussions centered around the microbiome, tuft cells, and metabolism. The authors would also like to thank various cores and centers at Vanderbilt for making this work possible, including the Vanderbilt Mass Spectrometry Research Center, Cooperative Human Tissue Network, VANTAGE Genomics Core, the Translational Pathology Shared Resource, and the Digital Histology Shared Resource. The authors would like to acknowledge the UNC Chapel Hill CGIBD Gnotobiotic Core and Dr Jakob von Moltke for their assistance in the acquisition of mouse lines.

      Supplementary Methods

       Human Tissue

      Formalin-fixed, paraffin-embedded (FPPE) blocks of ileum surgical resections were obtained from the Vanderbilt Cooperative Human Tissue Network Western Division, along with deidentified patient data and pathology reports. Pathologic examination was used to classify samples as “normal” (n = 14) or “diseased” (n = 19). Samples from patients with CD were included only if inflammation was evident in the distal ileum. Tissue samples were prepared for immunofluorescence imaging, as described below.

       Mouse Experiments

      Lrig1CreERT2 and Atoh1flox/flox strains, each in a C57BL/6 background, were purchased from The Jackson Laboratory (Bar Harbor, ME) to generate Lrig1CreERT2/+;Atoh1fl/fl (AtohKO) animals. Cre recombinase activity was induced in cohoused 2- or 3-month-old Lrig1CreERT2/+;Atoh1fl/fl and Lrig1+/+;Atoh1fl/fl littermate males via intraperitoneal administration of 2 mg of tamoxifen (Sigma-Aldrich, St. Louis, MO) for 4 consecutive days. Mice were humanely killed 21 days after the initial injection, and tissues were harvested. Cohoused TNFΔARE/+ and wild-type littermates were humanely killed after disease onset (age 4 months), followed by tissue collection. Animal weights were recorded at initiation of experiment and at the time of euthanasia.
      Microbiome depletion was performed by pretreatment of 2- or 3-month-old Lrig1CreERT2/+;Atoh1fl/fl animals with a broad-spectrum antibiotic cocktail containing kanamycin (4.0 mg/mL), metronidazole (2.15 mg/mL), gentamicin (0.35 mg/mL), colistin sulfate (8500 U/mL), and vancomycin (0.45 mg/mL) in their drinking water for 7 days before tamoxifen treatment. Mid-dose antibiotics and low-dose antibiotics were 0.75× and 0.25× of the original 1× concentration, respectively. After tamoxifen administration, male Lrig1CreERT2/+;Atoh1fl/fl littermates received standard or antibiotic-supplemented drinking water for an additional 14 days. Microbiome analysis was performed on cohoused male Lrig1CreERT2/+;Atoh1fl/fl littermates, which received vehicle (corn oil) or tamoxifen for 1 month before euthanasia.
      For colonization experiments, cecal contents were collected from wild-type and Lrig1CreERT2/+;Atoh1fl/fl animals, resuspended in 2 mL of sterile PBS, and vortexed at high speed for 10 minutes to homogenize. Samples were centrifuged to allow for sedimentation of larger particles, and the supernatant was mixed with an equal volume of 50% sterile glycerol before storage at −20°C. We purchased 4-week-old germ-free animals in a C57BL/6 background from the Center for Gastrointestinal Biology and Disease Gnotobiotic Core, University of North Carolina Chapel Hill. Littermates were inoculated twice with 100 μL wild-type or AtohKO contents through oral gavage and humanely killed at 3 and 7 days postinoculation, followed by tissue fixation and microbiome analysis.
      For succinate experiments, antibiotic-treated Lrig1CreERT2/+;Atoh1fl/fl mice received sodium succinate hexahydrate (120 mmol/L; Alfa Aesar, Ward Hill, MA) in addition to the antibiotic cocktail. Male TNFΔARE/+ mice received sodium succinate hexahydrate (120 mmol/L; Alfa Aesar) or standard drinking water after disease onset (3-4 months old) for 1 week (short-term) or 1 month (long-term).
      • Lei W.
      • Ren W.
      • Ohmoto M.
      • et al.
      Activation of intestinal tuft cell-expressed Sucnr1 triggers type 2 immunity in the mouse small intestine.
      ,
      • Schneider C.
      • O ’leary C.E.
      • Moltke J Von
      • et al.
      A Metabolite-Triggered Tuft Cell-ILC2 Circuit Drives Small Intestinal Remodeling.
      Anti–CD3E-driven inflammation was induced in 2- or 3-month-old C57BL/6 males via retro-orbital administration of 2 mg of anti-mouse CD3E (145-2C11; BD Biosciences, San Jose, CA) at 0, 48, and 96 hours. Succinate treatment was performed as described above. The source of C57BL/6N-Pou2f3<tm1.1(KOMP)Vlcg>/Tcp (Pou2f3-null) animals was the Canadian Mouse Mutant Repository (Toronto, ON, Canada). Anti-CD3E experiments were repeated in 2- or 3-month-old Pou2f3-null male littermates.
      Male and female mice of the C57BL/6 background were both used in experiments, and littermate controls were used when possible. All animals were housed 2 to 5 per cage in a controlled environment in standard bedding with a standard 12-hour daylight cycle, cessation of light at 7 PM, and free access to standard chow diet and water. Experiments were conducted during the light cycle, excluding continuous dietary interventions.

       Immunofluorescence Staining and Imaging

      FFPE tissues were sectioned (5 μm) before deparaffinization, rehydration, and antigen retrieval using a citrate buffer (pH 6.0) for 20 minutes in a pressure cooker at 105°C, followed by a 20-minute cooldown at room temperature (RT). Endogenous background signal was quenched by incubating tissue slides in 3% hydrogen peroxide for 10 minutes at RT. Tissue sections were blocked in staining buffer (3% bovine serum albumin/10% normal donkey serum) for 1 hour at RT before incubation with primary antibody overnight at RT. If secondary detection was needed, it was performed for 1 hour at RT. Hoechst (1:10,000; Life Technologies, Waltham, MA) was used for staining nuclei for 10 minutes at RT. Standard microscopy imaging was performed using a Zeiss Axio Imager M2 microscope with Axiovision digital imaging system (Carl Zeiss GmbH, Jena, Germany). Imaging cytometry was performed by using an multiplex iterative staining and fluorescence-inactivation protocol, as previously described,
      • McKinley E.T.
      • Sui Y.
      • Al-Kofahi Y.
      • et al.
      Optimized multiplex immunofluorescence single-cell analysis reveals tuft cell heterogeneity.
      and imaged on an Olympus X81 inverted microscope (20× magnification) with a motorized stage (Olympus, Melville, NY).

       Antibodies (Structured, Transparent, Accessible Reporting Methods)

      Human: Alexa Fluor (AF) 488-conjugated pEGFR(Y1068) (1:100, EP774Y; Abcam, Cambridge, MA), COX2 (1:100, D5H5; Cell Signaling Technology, Danvers, MA), and anti-rabbit AF-647–conjugated antibody for secondary detection.
      Mouse: anti-LYZ1 (1:100, polyclonal; DAKO, Glostrup, Denmark), anti-MUC2 (1:100, N-300; Santa Cruz Biotechnology, Dallas, TX), anti-MPO (1:100, A0398; DAKO), anti-DCLK1 (1:100, N-14; Santa Cruz Biotechnology), anti-FOXP3 (1:50, FJK-16S; eBioscience, San Diego, CA), anti-GATA3 (1:50, EPR16651; Abcam), anti-CD3E (1:100, M-20; Santa Cruz Biotechnology), anti-RORGT (1:50, AFKJS-9; eBioscience), anti-Ly6G/Ly6C (1:50, RB6-8C5; BioLegend, San Diego, CA), anti-B220/CD45R-AF594 (1:50, RA3-6B2; BD Pharmingen, San Diego, CA), anti–smooth muscle actin-AF488 (1:200; Invitrogen, Carlsbad, CA), anti-CD11c (1:50, N418; BioLegend), anti–CD45-AF488 (1:100, 30-F11; BioLegend), anti–β-catenin–AF550, anti–TER-119-AF647 (1:100, TER-119; BioLegend), anti-DCLK1-AF488 (1:100, EPR6085; Abcam), anti-MBP (Mayo Clinic Arizona, Phoenix, AZ), anti-OLFM4 (1:100, D6Y5A; Cell Signaling Technology), And AF-555– or AF-647–conjugated antibodies (1:500, Life Technologies) for secondary immunofluorescence detection. Anti-rat horseradish peroxidase was used for secondary immunohistochemistry detection.

       Image Quantification

      To quantify human tuft cells, the number of crypt and villus structures was counted per field of view (FOV) for each individual (approximately 15-20 FOVs per sample). Tuft cell number, as identified by pEGFR and COX colabeling, was manually counted per FOV in a blinded fashion. Quantification of tuft cell number per epithelial structure (crypt and villus) was generated for each individual and then stratified by disease state (“normal” or “CD”) based on the pathology report. For MUC2 quantification, manual demarcation of the epithelial villi was performed in MATLAB software (MathWorks, Natick, MA) to generate a MUC2 and nuclear mask; total area of both was calculated to generate a normalized ratio of MUC2 intensity to nuclear staining. This process was repeated for LYZ1 quantification on a per crypt basis.
      For DCLK1 quantification in wild-type and TNFΔARE/+ animals, histology was used to stratify villi from the latter into low-, mid-, and high-grade inflammation. The number of tuft cells in each category was manually counted. For MPO and DCLK1 quantification in TNFΔARE/+ animals, each villus was considered a separate unit, and the number of MPO+ neutrophils and DCLK1+ tuft cells was quantified by threshold gating in MATLAB. Tuft cell number per villi was stratified based on MPO staining as “low inflammation” (<35 neutrophils per villi) or “high inflammation” (≥35 neutrophils per villi). For both quantification methods, numbers of villi counted and used for significant testing can be found in Supplementary Figure 2. DCLK1, FOXP3, and MPO staining were quantified using threshold gating in MATLAB. Multiplexed imaging cytometry analysis was performed, as described previously, by performing epithelial and muscularis mask subtraction, cell segmentation, and calculating median intensity for each analyte with respect to the whole cell, plasma membrane, cytoplasm, and nucleus.

       Immunohistochemistry and Histologic Scoring

      FFPE ileal tissue from wild-type and succinate-treated slides were retrieved as described above and standard H&E staining was performed for histology. Standard immunohistochemistry with hematoxylin counterstain was performed by the Vanderbilt Translational Pathology Shared Resource, and 20× bright-field scanning of immunohistochemistry slides was performed by a Leica SCN400 Slide Scanner (Leica Biosystems, Buffalo Grove, IL) in the Vanderbilt Digital Histology Shared Resource. The degree of inflammation in the distal ileum was assessed in a blinded fashion by a trained pathologist. Damage to the epithelium was assessed using depth of inflammation on a 5-point scale (0 = no damage, 1 = mucosal damage only, 2 = submucosal infiltration, 3 = infiltration into muscularis propria, and 4 = infiltration into the peri-intestinal fat), whereas extent of inflammatory injury in the tissue Swiss roll was also scored on a 5-point scale (0 = <10%, 1 = 10%-25%, 2 = 25%-50%, 3 = 50%-75%, and 4 = >75%).

       Enteroid Culture

      Ileal tissue was dissected and incubated in chelation buffer (3 mmol/L EDTA/ethylene glycol-bis[β-aminoethyl ether]-N,N,N′,N′-tetraacetic acid, 0.5 mmol/L dithiothreitol , 1% penicillin/streptomycin) at 4°C for 45 minutes before shaking in PBS and filtration through a 100-μm filter to isolate individual ileal crypts. The crypt suspension was centrifuged at 2.8 × 1000 rpm for 1.5 minutes at 4°C, after which 10 μL of crypt pellet was resuspended in 300 μL of reduced growth factor Matrigel and embedded in a 24-well dish. Enteroids were cultured initially in IntestiCult Organoid Growth Medium (StemCell Technologies, Vancouver, BC, Canada) supplemented with Primocin antimicrobial reagent (1:1000; InvivoGen, San Diego, CA) for 4 days before being changed to Primocin-supplemented differentiation media, as previously described.
      • Sato T.
      • Vries R.G.
      • Snippert H.J.
      • et al.
      Single Lgr5 stem cells build crypt-villus structures in vitro without a mesenchymal niche.
      For ex vivo CRE-activation, Lrig1CreERT2/+;Atoh1fl/fl enteroids were treated for 24 hours at 37°C with 1 μmol/L 4-hydroxytamoxifen (Sigma-Aldrich) or vehicle (ethanol) in differentiation media.
      Enteroids were dissociated and passaged the following day into 8-well chamber slides and, after 5 days, were fixed at RT using 4% paraformaldehyde for immunofluorescence staining. To induce tuft cell expansion, ileal enteroids were treated with 800 ng/mL murine recombinant IL13 (BioLegend) or vehicle (PBS) for 48 hours before fixation. Manual quantification of DCLK1+ tuft cells was performed on ETOH/PBS (n = 41), 4-OHT/PBS (n = 26), ETOH/IL13 (n = 29), and 4-OHT/IL13 (n = 25) enteroids across biological and technical replicates.

       Enteroid Immunofluorescence Staining

      Fixed enteroids were permeabilized with Triton X-100 for 30 minutes and blocked with 1% normal donkey serum (PBS) for 30 minutes at RT. Enteroids were stained with primary antibodies overnight, followed by conjugated secondary antibodies (1:500, Life Technologies) and Hoechst (1:10,000, Life Technologies) 1 hour at RT. Vectashield Antifade Mounting Medium (Vector Laboratories, Burlingame, CA) was applied to enteroids before imaging with a Nikon A1R Resonance Scanning Confocal microscope (Nikon, Melville, NY).

       Droplet-Based Single-Cell RNA Sequencing

      Ileal crypts and villi from human and mouse tissue were isolated as described above. Tissues were dissociated into single cells using a cold-activated protease (1 mg/mL)/DNAseI (2.5 mg/mL) enzymatic cocktail in a modified protocol that maintains high cell viability.
      • Adam M.
      • Potter A.S.
      • Potter S.S.
      Psychrophilic proteases dramatically reduce single-cell RNA-seq artifacts: a molecular atlas of kidney development.
      Dissociation was performed at 4°C for 15 minutes, followed by trituration to mechanically disaggregate cell clusters. Cell viability was assessed by counting trypan blue-positive cells. The cell suspension was enriched for live cells with a MACS dead cell removal kit (Miltenyi Biotec Inc, San Diego, CA) before encapsulation. Single cells were encapsulated and barcoded using the inDrop platform (1CellBio, Watertown, MA) with an in vitro transcription library preparation protocol.
      • Herring C.A.
      • Banerjee A.
      • McKinley E.T.
      • et al.
      Unsupervised Trajectory Analysis of Single-Cell RNA-Seq and Imaging Data Reveals Alternative Tuft Cell Origins in the Gut.
      Briefly, the Cell Expression by Linear Amplification Sequencing (CEL-Seq) work flow entailed (1) reverse transcription (RT), (2) ExoI digestion, (3) solid-phase reversible immobilization purification (SPRIP), (4) second strand synthesis, (5) T7 in vitro transcription linear amplification, (7) SPRIP, (8) RNA fragmentation, (9) SPRIP, (10) primer ligation, (11) RT, (12) SPRIP, (13) library enrichment polymerase chain reaction, and (14) SPRIP.
      • Herring C.A.
      • Banerjee A.
      • McKinley E.T.
      • et al.
      Unsupervised Trajectory Analysis of Single-Cell RNA-Seq and Imaging Data Reveals Alternative Tuft Cell Origins in the Gut.
      Each sample was estimated to contain approximately 2500 encapsulated cells.
      After library preparation, the samples were sequenced using NextSeq 500 (Illumina, San Diego, CA) using a 150-base pair paired-end sequencing kit in a customized sequencing run.
      • Herring C.A.
      • Banerjee A.
      • McKinley E.T.
      • et al.
      Unsupervised Trajectory Analysis of Single-Cell RNA-Seq and Imaging Data Reveals Alternative Tuft Cell Origins in the Gut.
      After sequencing, reads were filtered, sorted by their barcode of origin, and aligned to the reference transcriptome using the inDrop pipeline. Mapped reads were quantified into unique molecular identifier-filtered counts per gene, and barcodes that corresponded to cells were retrieved based on previously established methods. Overall, from approximately 2500 encapsulated cells, approximately 1800 to 2000 cells were retrieved per sample.

       Preprocessing and Batch Correction of Single-Cell RNA Sequencing Data

      Data sets were filtered for cells with low library size or high mitochondrial gene expression.
      • Herring C.A.
      • Banerjee A.
      • McKinley E.T.
      • et al.
      Unsupervised Trajectory Analysis of Single-Cell RNA-Seq and Imaging Data Reveals Alternative Tuft Cell Origins in the Gut.
      Filtered data sets for each replicate were analyzed using the Seurat pipeline.
      • Stuart T.
      • Butler A.
      • Hoffman P.
      • et al.
      Comprehensive Integration of Single-Cell Data.
      Briefly, count matrices were log scale normalized, followed by feature selection of highly variable genes. Canonical correlation analysis was used to align replicates based on biological condition using dynamic time warping. Visual assessment of alignment between replicates was performed using t-distributed stochastic neighbor embedding analysis.
      • Maaten L Van Der
      • Hinton G.
      Visualizing data using t-SNE.

       p-Creode Mapping and Trajectory Analysis

      Filtered scRNA-seq data sets were feature selected using the binned variance method
      • Butler A.
      • Hoffman P.
      • Smibert P.
      • et al.
      Integrating single-cell transcriptomic data across different conditions, technologies, and species.
      and then analyzed using the p-Creode algorithm (https://github.com/KenLauLab/pCreode).
      • Herring C.A.
      • Banerjee A.
      • McKinley E.T.
      • et al.
      Unsupervised Trajectory Analysis of Single-Cell RNA-Seq and Imaging Data Reveals Alternative Tuft Cell Origins in the Gut.
      For graph scoring, 100 independent runs were generated from each combined data set. Overlay of ArcSinh-normalized expression data was used to identify cell lineages and quantify tuft cell placement as “secretory” or “nonsecretory.”

       Trend Analysis Overview

      Trend analysis was performed to identify gene expression changes in the AtohKO tuft cell lineage compared with the wild-type tuft cell lineage. Ten p-Creode resampled runs were used from the wild-type and AtohKO data set. The top 2500 genes (ranked by variance) over the tuft cell trajectory were selected from each of the wild-type and AtohKO data sets. The union of these gene sets (3420 genes) was used for downstream analysis. The dynamic trend of gene expression for each gene over the tuft cell trajectory was obtained by fitting a linear generalized additive model with a normal distribution and an identity link function using 10 splines. The fitted curves were then normalized between 0 and 1 for comparison between data sets. Classification of the dynamics for each gene trend was performed by calculating its dynamic time warping distances to 12 reference trends. These categories were then broadly combined into 5 classes: (1) upward, (2) upward transitory, (3) downward transitory, (4) downward, and (5) flat. For the coarse grain analysis, 3 trend classes were formed by combining (1) group 1 and 2 genes into “upward,” (2) group 3 and 4 genes into “downward,” and (3) flat. Trend classification was scored by consensus over 10 resampled runs
      • Herring C.A.
      • Banerjee A.
      • McKinley E.T.
      • et al.
      Unsupervised Trajectory Analysis of Single-Cell RNA-Seq and Imaging Data Reveals Alternative Tuft Cell Origins in the Gut.
      for both the (A) 5-trend and (B) 3-trend analysis. Genes with high consensus are those with a cumulative sum of 16 between the 2 classifications; for instance, a gene being grouped in the same trend in 8 of 10 p-Creode replicates in the 5-trend analysis and in 8 of 10 p-Creode replicates in the 3-trend analysis. This resulted in 2004 high-consensus genes being used for downstream over-representation analysis. From the list of high-consensus genes, we identified genes that switched from (A) wild-type group 4 to AtohKO group 1, 2, or 3, (B) wild-type group 3 to AtohKO group 1 or 2, or (C) wild-type group 2 to AtohKO group 1. Over-representation analysis, based on KEGG, Reactome pathway, and WikiPathway data sets of upregulated genes was performed in WebGestalt (http://www.webgestalt.org/).
      • Wang J.
      • Vasaikar S.
      • Shi Z.
      • et al.
      WebGestalt 2017: a more comprehensive, powerful, flexible and interactive gene set enrichment analysis toolkit.
      Using 3-trend analysis, we identified genes that switched from wild-type group 2 (downward) to AtohKO group 1 (upward) and performed over-representation analysis.

       Visualization and Significance Testing of Trend Analysis

      Visualization of trend dynamics was performed using MATLAB software for enriched genes from the Reactome Pathway “citric acid cycle (TCA cycle)” (https://reactome.org/content/detail/R-HSA-71403) gene list. Each gene plot includes raw expression data from each wild-type or AtohKO p-Creode replicate and the trend line as an average of raw expression data aligned by dynamic time warping across all 10 resampled runs for each respective condition.

       Gene Set Enrichment Analysis of Differential Expression

      Median difference in gene expression was calculated between wild-type and AtohKO tuft cells. Gene set enrichment analysis (http://software.broadinstitute.org/gsea/index.jsp) of differential gene expression was performed to identify positively enriched pathways based on the normalized enrichment score, a metric used to identify an “enriched gene set” or a set of genes whose members are highly expressed in the data set being analyzed. We used 20 gene sets with the highest normalized enrichment score and most significant P value for further analysis. Gene set enrichment analysis and over-representation analysis for specific gene sets (KEGG, Reactome pathway, and WikiPathways) was performed using WebGestalt (http://www.webgestalt.org/).
      • Wang J.
      • Vasaikar S.
      • Shi Z.
      • et al.
      WebGestalt 2017: a more comprehensive, powerful, flexible and interactive gene set enrichment analysis toolkit.

       DNA Extraction and 16S Ribosomal RNA Gene Sequencing

      Ileal luminal contents were collected fresh from tamoxifen- or vehicle-treated Lrig1CreERT2/+;Atoh1fl/fl animals, as described above. All samples were collected on the same day and frozen in 2 mL Eppendorf tubes (DNAse and RNAse free) at −80°C. Microbial genomic DNA extraction was performed using the PowerSoil DNA Isolation Kit (MO BIO Laboratories, Carlsbad, CA).
      • Zackular J.P.
      • Moore J.L.
      • Jordan A.T.
      • et al.
      Dietary zinc alters the microbiota and decreases resistance to Clostridium difficile infection.
      Briefly, luminal contents were added to PowerBead Tubes and homogenized twice in a Bead Beater machine for 3 minutes with a 2-minute cooldown in between homogenization. Samples were then processed as according to the kit instructions, and DNA was eluted into a sterile buffer. The V4 region of the 16S ribosomal (r)RNA gene from each sample was amplified and sequenced by Georgia Genomics and Bioinformatics Core (http://dna.uga.edu) using the MiSeq (Illumina) platform.
      • Zackular J.P.
      • Rogers M.A.M.
      • Ruffin M.T.
      • et al.
      The human gut microbiome as a screening tool for colorectal cancer.
      Raw 16S rRNA sequences were filtered for quality (target error rate <0.5%) and length (minimum final length, 225 base pairs) using Trimmomatic
      • Bolger A.M.
      • Lohse M.
      • Usadel B.
      Genome analysis Trimmomatic: a flexible trimmer for Illumina sequence data.
      and Quantitative Insights Into Microbial Ecology (QIIME).
      • Caporaso J.G.
      • Kuczynski J.
      • Stombaugh J.
      • et al.
      QIIME allows analysis of high-throughput community sequencing data.
      Spurious hits to the PhiX (Illumina) control genome were identified using nucleotide-nucleotide basic local alignment search tool (BLASTN) and removed. Passing sequences were trimmed of forward and reverse primers, evaluated for chimeras with UCLUST
      • Edgar R.C.
      • Bateman A.
      Search and clustering orders of magnitude faster than BLAST.
      (de novo mode), and screened for mouse-associated contaminant using Bowtie2, followed by a more sensitive BLASTN search against the GreenGenes 16S rRNA gene sequence database. Chloroplast and mitochondrial contaminants were detected and filtered using the Ribosomal Database Project classifier with a confidence threshold of 80%. High-quality 16S rRNA gene sequences were assigned to a high-resolution taxonomic lineage using Resphera Insight.
      • Drewes J.L.
      • White J.R.
      • Dejea C.M.
      • et al.
      High-resolution bacterial 16S rRNA gene profile meta-analysis and biofilm status reveal common colorectal cancer consortia.
      Functional gene content was inferred using PICRUSt.
      • Langille M.G.I.
      • Zaneveld J.
      • Caporaso J.G.
      • et al.
      Predictive functional profiling of microbial communities using 16S rRNA marker gene sequences.
      Statistical analyses between groups was performed using R 3.5 software (http://cran.r-project.org/).

       Pathogen Testing

      Fresh fecal samples were collected from non-cohoused animals (n = 5) and frozen at −20°C. Samples were shipped to IDEXX BioAnalytics (Columbia, MO) for pathogen polymerase chain reaction panel and helminth float testing.

       O-Benzylhydroxylamine Derivatization of Cecal Contents and Tissue

      O-Benzylhydroxylamine (O-BHA) derivatization of common tricarboxylic acid intermediates was performed from both cecal luminal contents and tissues
      • Tan B.
      • Lu Z.
      • Dong S.
      • et al.
      Derivatization of the tricarboxylic acid intermediates with O-benzylhydroxylamine for liquid chromatography–tandem mass spectrometry detection.
      by the Vanderbilt Mass Spectrometry Service Laboratory. Briefly, luminal filtrate was mixed in MeOH/H2O with 0.1% formic acid and added to pyruvate 13C3 (327 μg/mL), 1-ethyl-3-(3-dimethylamino) propyl carbodiimide (EDC), hydrochloride, and O-BHA, as described in the Sherwood protocol.
      • Fensterheim B.A.
      • Young J.D.
      • Luan L.
      • et al.
      The TLR4 Agonist Monophosphoryl Lipid A Drives Broad Resistance to Infection via Dynamic Reprogramming of Macrophage Metabolism.
      Tissue homogenates were processed similarly, and both were incubated at RT for 1 hour before to extraction with ethyl acetate. Then, 100 μL of luminal content sample (200 mg/mL) or tissue sample (250 mg/mL) was analyzed for specified metabolites, and analyte response ratios were calculated using validated standards.
      • Zhu J.
      • Djukovic D.
      • Deng L.
      • et al.
      Colorectal Cancer Detection Using Targeted Serum Metabolic Profiling.
      An aliquot of PBS was processed and reconstituted as a negative control.

       Tissue Cytokine and Chemokine Analysis

      Mouse ileum were dissected as described above, and 1-mm segments were collected before fixation and frozen at −80°C. Tissues were lysed in CelLytic MT lysis extraction reagent (Sigma-Aldrich) with a mortar and pestle–type rotary homogenizer and used for multiplex assay.
      • Singh K.
      • Coburn L.A.
      • Barry D.P.
      • et al.
      Deletion of cationic amino acid transporter 2 exacerbates dextran sulfate sodium colitis and leads to an IL-17-predominant T cell response.
      Premixed MILLIPLEX MAP 32-plex Mouse Cytokine/Chemokine Magnetic Bead Panel (MCYTMAG-70K-PX32; EMD Millipore, Burlington, MA) and 12-plex Mouse Th17 Magnetic Bead Panel (MTH17MAG-47K-12, EMD Millipore) kits were used according to the manufacturer’s instructions before analysis with a FlexMAP 3D Instrument System (Luminex). Data were standardized to tissue protein concentrations measured by the bicinchoninic acid method.

       Statistical Methods

      Statistical analysis between conditions was conducted using Student’s t test when 2 groups were compared (GraphPad Prism; GraphPad Software, San Diego, CA). For multiple comparison testing, analysis of variance, followed by Tukey’s post test was used. Results are expressed as mean ± SEM, unless otherwise stated. A P value of at least .05 was considered statistically significant.
      For significance testing between wild-type and AtohKO p-Creode trends, randomized classifications were generated for each gene in the wild-type and AtohKO condition by resampling from the same distribution for each condition. The null hypothesis stated that there was no consensus across the wild-type classifications or that there was no upward class switching from wild-type to AtohKO trajectories. Permutation testing comparing the randomized and observed classifications was performed 10,000 times to obtain a P value.
      For microbiome analysis, heat map clustering visualizations were generated using Excel software (Microsoft, Redmond, WA), selecting microbiome features with false discovery rate– adjusted P values <.10, and log-transforming values for color scaling.

      Supplementary Material

      References

        • Molodecky N.A.
        • Soon I.S.
        • Rabi D.M.
        • et al.
        Increasing incidence and prevalence of the inflammatory bowel diseases with time, based on systematic review.
        Gastroenterology. 2012; 142: 46-54.e42
        • de Silva P.
        • Korzenik J.
        The changing epidemiology of inflammatory bowel disease: identifying new high-risk populations.
        Clin Gastroenterol Heptaol. 2015; 4: 690-692
        • Summers R.W.
        • Elliott D.E.
        • Urban J.F.
        • et al.
        Trichuris suis therapy in Crohn’s disease.
        Gut. 2005; 54: 87-90
        • Su J.
        • Chen T.
        • Ji X.-Y.
        • et al.
        IL-25 downregulates Th1/Th17 immune response in an IL-10–dependent manner in inflammatory bowel disease.
        Inflamm Bowel Dis. 2013; 19: 720-728
        • Broadhurst M.J.
        • Leung J.M.
        • Kashyap V.
        • et al.
        IL-22+ CD4+ T cells are associated with therapeutic Trichuris trichiura infection in an ulcerative colitis patient.
        Sci Transl Med. 2010; 2: 60ra88
        • Gerbe F.
        • Sidot E.
        • Smyth D.J.
        • et al.
        Intestinal epithelial tuft cells initiate type 2 mucosal immunity to helminth parasites.
        Nature. 2016; 529: 226-230
        • Moltke J von
        • Ji M.
        • Liang H.-E.
        • et al.
        Tuft cell derived IL25 regulates an intestinal ILC2-epithelial response circuit.
        Nature. 2016; 259: 221-225
        • Howitt M.R.
        • Lavoie S.
        • Michaud M.
        • et al.
        Tuft cells, taste-chemosensory cells, orchestrate parasite type 2 immunity in the gut.
        Science. 2016; 351: 1329-1333
        • Haber A.L.
        • Biton M.
        • Rogel N.
        • et al.
        A single-cell survey of the small intestinal epithelium.
        Nature. 2017; 551: 333-339
        • Gerbe F.
        • Es JH Van
        • Makrini L.
        • et al.
        Distinct ATOH1 and Neurog3 requirements define tuft cells as a new secretory cell type in the intestinal epithelium.
        J Cell Biol. 2011; 192: 767-780
        • Herring C.A.
        • Banerjee A.
        • Mckinley E.T.
        • et al.
        Unsupervised trajectory analysis of single-cell RNA-Seq and imaging data reveals alternative tuft cell origins in the gut.
        Cell Syst. 2018; 6: 37-51.e9
        • Gracz A.D.
        • Fordham M.J.
        • Trotier D.C.
        • et al.
        Sox4 promotes Atoh1-independent intestinal secretory differentiation toward tuft and enteroendocrine fates.
        Gastroenterology. 2018; 155: 1508-1523.e10
        • McKinley E.T.
        • Sui Y.
        • Al-Kofahi Y.
        • et al.
        Optimized multiplex immunofluorescence single-cell analysis reveals tuft cell heterogeneity.
        JCI Insight. 2017; 2e93487
        • Wehkamp J.
        • Wang G.
        • Ku I.
        • et al.
        The Paneth cell alpha-defensin deficiency of ileal Crohn’s disease is linked to Wnt/Tcf-4.
        J Immunol. 2007; 179: 3109-3118
        • Kontoyiannis D.
        • Pasparakis M.
        • Pizarro T.T.
        • et al.
        Impaired on/off regulation of TNF biosynthesis in mice lacking TNF AU-rich elements: implications for joint and gut-associated immunopathologies.
        Immunity. 1999; 10: 387-398
        • Shroyer N.F.
        • Helmrath M.A.
        • Wang V.Y.-C.
        • et al.
        Intestine-specific ablation of mouse atonal homolog 1 (Math1) reveals a role in cellular homeostasis.
        Gastroenterology. 2007; 132: 2478-2488
        • Meng D.
        • Newburg D.S.
        • Young C.
        • et al.
        Bacterial symbionts induce a FUT2-dependent fucosylated niche on colonic epithelium via ERK and JNK signaling.
        Am J Physiol Gastrointest Liver Physiol. 2007; 293: 780-787
        • Sato T.
        • Vries R.G.
        • Snippert H.J.
        • et al.
        Single Lgr5 stem cells build crypt-villus structures in vitro without a mesenchymal niche.
        Nature. 2009; 459: 262-265
        • Bastide P.
        • Darido C.
        • Pannequin J.
        • et al.
        Sox9 regulates cell proliferation and is required for Paneth cell differentiation in the intestinal epithelium.
        J Cell Biol. 2007; 178: 635-648
        • Farin H.F.
        • Karthaus W.R.
        • Kujala P.
        • et al.
        Paneth cell extrusion and release of antimicrobial products is directly controlled by immune cell-derived IFN-γ.
        J Exp Med. 2014; 211: 1393-1405
        • Yan K.S.
        • Gevaert O.
        • Zheng G.X.Y.
        • et al.
        Intestinal enteroendocrine lineage cells possess homeostatic and injury-inducible stem cell activity.
        Cell Stem Cell. 2017; 21: 78-90.e6
        • Tan B.
        • Lu Z.
        • Dong S.
        • et al.
        Derivatization of the tricarboxylic acid intermediates with O-benzylhydroxylamine for liquid chromatography–tandem mass spectrometry detection.
        Anal Biochem. 2014; 465: 134-147
        • Schneider C.
        • O’Leary C.E.
        • Moltke J Von
        • et al.
        A metabolite-triggered tuft cell-ILC2 circuit drives small intestinal remodeling.
        Cell. 2018; 174: 1-14
        • Lei W.
        • Ren W.
        • Ohmoto M.
        • et al.
        Activation of intestinal tuft cell-expressed Sucnr1 triggers type 2 immunity in the mouse small intestine.
        PNAS. 2018; 115: 5552-5557
        • Nadjsombati M.S.
        • McGinty J.W.
        • Lyons-Cohen M.R.
        • et al.
        Detection of succinate by intestinal tuft cells triggers a type 2 innate immune circuit.
        Immunity. 2018; 49: 33-41.e7
        • Yamashita M.
        • Ukai-Tadenuma M.
        • Miyamoto T.
        • et al.
        Essential role of GATA3 for the maintenance of type 2 helper T (Th2) cytokine production and chromatin remodeling at the Th2 cytokine gene loci.
        J Biol Chem. 2004; 279: 26983-26990
        • Bibiloni R.
        • Fedorak R.N.
        • Tannock G.W.
        • et al.
        VSL#3 probiotic-mixture induces remission in patients with active ulcerative colitis.
        Am J Gastroenterol. 2005; 100: 1539-1546
        • Langille M.G.I.
        • Zaneveld J.
        • Caporaso J.G.
        • et al.
        Predictive functional profiling of microbial communities using 16S rRNA marker gene sequences.
        Nat Biotechnol. 2013; 31: 814-821
        • Meulen R Van der
        • Adriany T.
        • Verbrugghe K.
        • et al.
        Kinetic analysis of bifidobacterial metabolism reveals a minor role for succinic acid in the regeneration of NAD+ through its growth-associated production.
        Appl Environ Microbiol. 2006; 72: 5204-5210
        • Stumhofer J.S.
        • Laurence A.
        • Wilson E.H.
        • et al.
        Interleukin 27 negatively regulates the development of interleukin 17-producing T helper cells during chronic inflammation of the central nervous system.
        Nat Immunol. 2006; 7: 937-945
        • Aggarwal S.
        • Ghilardi N.
        • Xie M.H.
        • et al.
        Interleukin-23 promotes a distinct CD4 T cell activation state characterized by the production of interleukin-17.
        J Biol Chem. 2003; 278: 1910-1914
        • Turner J.-E.
        • Stockinger B.
        • Helmby H.
        IL-22 mediates goblet cell hyperplasia and worm expulsion in intestinal helminth infection.
        PLoS Pathog. 2013; 10e1003698
        • Holm T.L.
        • Tornehave D.
        • Søndergaard H.
        • et al.
        Evaluating IL-21 as a potential therapeutic target in Crohn’s disease.
        Gastroenterol Res Pract. 2018; 2018: 5962624
        • Miura N.
        • Yamamoto M.
        • Fukutake M.
        • et al.
        Anti-CD3 induces bi-phasic apoptosis in murine intestinal epithelial cells: possible involvement of the Fas/Fas ligand system in different T cell compartments.
        Int Immunol. 2005; 17: 513-522
        • Yamashita J.
        • Ohmoto M.
        • Yamaguchi T.
        • et al.
        Skn-1a/Pou2f3 functions as a master regulator to generate Trpm5-expressing chemosensory cells in mice.
        PLoS One. 2017; 12e0189340