All animal procedures were performed at Moredun Research Institute (MRI), UK under license as required by the UK Animals (Scientific Procedures) Act 1986, with ethical approval from MRI Animal Experiments Committee. Animals were raised at MRI under conditions designed to exclude accidental infection with helminth parasites and were helminth-naïve.

Abomasal tissue for IHC was collected over a time course of T. circumcincta infection with six animals per group, matched for weight and sex (Experiment 1, Supplementary Table 1 ). For IHC after H. contortus infection, abomasal tissue was collected at day 55 p.i. with 10,000 L3 (Experiment 2, Supplementary Table 1 ). Abomasal tissue for scRNA-seq (Experiment 3, Supplementary Table 1 ) was collected at day 21 p.i. from two lambs infected with 50,000 T. circumcincta L3. Small intestine tissue samples were collected from three lambs infected with 7,000 H. contortus L3 and tissue collected at day 56 post-infection. Sections of murine SI and ovine brain tissue were obtained from studies at MRI and were included in IHC to test specificity and species cross-reactivity of the DCLK-1 antibody.

Antibodies are detailed in Supplementary Table 2 . Abomasal tissue sections were fixed in 4% paraformaldehyde for 6 hours at room temperature (RT) and embedded in paraffin wax (PFPE). Following dewaxing, antibody staining was carried out as described previously (21). For double immunofluorescence, following antigen retrieval and blocking, slides were incubated overnight at 4°C with mouse anti-GFI1b. HRP labelled anti-mouse DAKO Envision™+ Polymer (Agilent) was applied for 30 min at RT followed by incubation with Alexa Fluor^® 488 Tyramide substrate (Invitrogen). Slides were then incubated with anti-POU2F3 antibody, washed and incubated with goat anti-rabbit IgG (H+L) conjugated to Alexa Fluor-546^® (Thermo Fisher Scientific). Hoechst stain (Invitrogen) was applied for 5 min and Prolong Gold (Invitrogen) added before viewing under a Zeiss Axiovision fluorescence microscope.

Periodic Acid Schiff (PAS) Stain kit (HS462, TCS Biosciences) was used to stain mucins. Tissue sections were incubated with 0.3% hydrogen peroxide-PBST, washed, then PAS added for 5 min. Following rinsing, Feulgen Stain (Schiff) was added for 15 min, slides washed, and stain left to develop for 10 min.

An Olympus BX50 Microscope (Model U-SD0, Olympus Optical Co., Ltd.) with an Olympus DP70 camera and Olympus U-CMAD3 adapter was used with the AnalySIS program for imaging. For enumeration of POU2F3⁺ cells, five images at x40 magnification were acquired for each tissue section, ensuring no overlap and samples scored blind. All epithelial cells were counted using Image J (22). Following counting of all POU2F3⁺ cells within the epithelial cells, the percentage of POU2F3⁺ cells/all epithelial cells was calculated. The mean percentage of POU2F3⁺ cells was calculated across the five images for each sheep.

R version 4.0.3 using the statistical package R Core Team (https://www.R-project.org/) was used unless otherwise stated. For T. circumcincta time-course infection, IHC cell count data were analyzed using a Kruskal-Wallis test followed by Dunn’s multiple comparisons test to compare median values where data were not normally distributed. Two-sample t-tests were used to compare mean values for normally distributed data. Generalised Linear Mixed Models (GLMM) were used to examine effects of abomasal region and infection status on percentage of tuft cells.

RNAscope 2.5 HD Duplex Detection Kit (Advanced Cell Diagnostics (ACD), Bio-Techne, UK) was used to perform in situ hybridization on abomasum tissue collected from three sheep; two helminth-naïve and one sheep at 30 days post-H. contortus infection. Due to differences in fixation time (6 h for IHC, 24 h for RNAscope) we were unable to use the tissue collected following T. circumcincta infection for RNAscope. We had access to H. contortus-infected and control (uninfected) tissue fixed for the requisite time and we made use of these samples for validation of scRNA-seq data. Probes were designed by ACD to target ovine POU2F3 (green signal) and ovine IL17RB, TAS2R16 and DCLK-2 (fast red signal). Positive control probes were designed to target ovine beta actin (actb; green signal) and ovine peptidylprolyl isomerase (ppib; red signal). Negative control probe was designed to DapB from Bacillus subtilis. Ovine abomasum tissue was fixed in 10% formalin for 24 hours, moved to 70% ethanol, then processed and embedded into paraffin wax blocks. Sections (4 μM) were mounted onto Superfrost Plus glass slides. RNAscope was performed following manufacturer’s instructions, with optimization for ovine abomasum tissue and signal enhancement. Target retrieval was applied for 7 min, hydrogen peroxide for 10 min, followed by Protease Plus for 15 min. For IL17RB and TAS2R16 probes, the Amp 5 step was reduced to 10 min to optimize the signal. Slides were counterstained with Haematoxylin and imaged using light microscopy equipment described above.

Full-length Ova-POU2F3 gene (NCBI XM_015100962.2) with C-terminal enhanced yellow fluorescent protein (eYPF) tag was obtained from Eurofins. Ova-POU2F3 insert was ligated into XhoI-NotI digested pCI-neo vector (Promega) and plasmid DNA from the pCI-neo vector alone, pCI-neo containing Ova-POU2F3 or pCAG-neo containing ovine transcription factor FOXP3 gene (23) was transformed initially into Top10 cells, then into HEK293 cells. For western blot analysis, cells were lysed in NuPAGE sample buffer (Invitrogen) with 2% β-mercaptoethanol, proteins separated on a Mini-PROTEAN TGX Precast Protein Gel (Bio-Rad), then transferred to nitrocellulose membrane. Following Ponceau S staining, the membrane was cut between the 50 and 75 kDa marker proteins, blocked (5% milk powder in 0.1% PBS-Tween20, PBST), and probed overnight at 4°C with primary antibody: anti-POU2F3, 1:2500 (Sigma-Aldrich, HPA019652) or anti-actin, 1:1000 (Sigma, A-3853) before incubation for 1 hour at RT with secondary antibody: anti-mouse peroxidase, 1:20,000 (Sigma, A-2304) or anti-rabbit peroxidase, 1:10,000 (Sigma, A-0545). Following washing in PBST, membranes were re-joined and signal developed using SuperSignalTM West Pico PLUS Chemiluminescent Substrate (Thermo Scientific) and captured on x-ray film.

Ovine abomasum epithelial cells were isolated from six 3 cm² gut fold sections collected at post-mortem from two sheep 21-days post T. circumcincta infection, following 10x Genomics single cell protocol. Tissue sections were placed in ice-cold Hanks’ Balanced Salt Solution (HBSS) containing 2.5% fetal bovine serum (FBS), two sections of tissue per tube. Tissue was washed with HBSS, shaking gently to remove any stomach contents, then incubated with HBSS containing 2.5 mM EDTA (Sigma Aldrich), 1mM DTT (Sigma Aldrich) and 10 μg/ml DNase I (Roche) at 37°C for 20 min with agitation. Tubes were shaken vigorously, and the supernatant centrifuged at 330 x g for 5 min to pellet released cells, which were washed in HBSS, centrifuged again and re-suspended in HBSS containing 1 U/ml Dispase II (Stem Cell Technologies, 07913) and 10 μg/ml DNase I (Sigma, 10104159001) and incubated for 10 min at 37°C with agitation. FBS (250 μl) was added to each tube, which was then centrifuged. Following a final wash in HBSS, cells from the same sheep were pooled in 5 ml HBSS and strained through a 70 μm Nylon mesh sterile strainer (Fisherbrand). FBS (250 μl) was added, samples centrifuged and pellets re-suspended in 5 ml PBS. Cell count and viability assessment were performed with Trypan Blue staining.

The cell suspension for each sheep was processed separately using the droplet-based Chromium Controller microfluidic platform (10x Genomics) (24) with Chromium Single Cell 3’ Reagent Kit (v.3) according to the manufacturer’s protocol by University of Glasgow Polyomics facility using a total of 20,000 input cells in a volume of 46 μl. Briefly, cells were partitioned into Gel Beads in Emulsion, followed by lysis, and extracted RNA subject to barcoded reverse transcription. cDNA was amplified (12 cycles), libraries generated and sequenced on an Illumina NextSeq 500 system to a depth of 50,000 read pairs per cell. Sequences were mapped to the genomes for ovine (Ovis aries Oar_v3.1; https://www.ensembl.org/Ovis_aries/Info/Index) and bovine (Bos taurus ARS-UCD1.2; https://www.ensembl.org/Bos_taurus/Info/Index) separately and the respective gene counts matrices generated using Cell Ranger software (version 3.1.0). As only approximately 50% of transcripts had annotated 3’UTR, we extended the 3’UTR end of each transcript by 1500 bp.

Seurat package, version 3.1 (25, 26) was used with R version 4.0.3 (https://www.r-project.org). Other packages used for analysis were ggplot2, car, ggpubr (https://CRAN.R-project.org/package=ggpubr), dplyr (https://CRAN.R-project.org/package=dplyr) and cowplot (https://CRAN.R-project.org/package=cowplot).

Following the guidance of the Seurat Vignette (https://satijalab.org/seurat/v3.1/pbmc3k_tutorial.html), each sample was pre-processed, subject to QC and filters applied, based on the individual sample QC metrics. For sample 1, cells with <300 and >6500 genes were removed, to mitigate including potential doublets as well as cells with a mitochondrial gene percentage of >20%. For sample 2, cells with <300 and >5750 genes were excluded, and cells with mitochondrial gene percentage of >20%. Sample data were then transformed and normalized with the “Log Normalise” function. After processing, samples 1 and 2 contributed 7,963 and 8,928 cells, respectively, and >17,000 genes for downstream analysis. The “Find Integration Anchors” followed by the “Integrate Data” Seurat functions were used to combine the sample data. The number of significant principal components (PC = 35) was selected based on Elbow plots and applied when combining the data. Default Seurat methods (Findclusters) with resolution parameter 0.1 and UMAP were used to visualize the cells and clustering. The “Find Markers” function, with default parameters (log fold change (FC) >0.25), generated gene expression lists for each cluster based on differential expression, calculated by a non-parametric Wilcoxon rank sum test (26). Based on marker genes expressed by at least 25% of cells in the cluster, different cell populations were annotated. Use of the “RNA assay” was specified when generating plots of specific genes of interest.

To further analyze ovine tuft cells, we repeated the clustering step on these cells (PCA dimension of 15 and cluster resolution of 0.5). For trajectory inference, tuft cells were plotted using PHATE maps (27) (using the same genes as for PCA) and trajectories identified using slingshot (28), with sub-cluster 1 defined as the putative starting point. Genes with expression patterns associated with progression of the trajectory were identified using generalized additive models with tradeSeq package v1.3.18104 default parameters. Differential expression analysis was carried out using tradeSeq association Test function with default parameters, and significant genes (log FC > 0.25) clustered using tradeSeq clusterExpressionPattern.

To determine if cells characteristic of tuft cells are present in the ovine abomasal epithelium, antibodies to human POU2F3, GFI1b, TRPM5 and DCLK-1 (15) ( Supplementary Table 2 ) were tested by immunohistochemistry (IHC). Analysis of ovine genome data (https://www.ensembl.org/Ovis_aries/Info/Annotation) identified sufficient conservation ( Supplementary Table 3 ) to suggest antibody cross-reactivity.

We first examined samples at 21 days post-infection (p.i.) with T. circumcincta, a time-point at which adult parasites are present in the abomasum lumen. Clear labelling of POU2F3⁺ epithelial cells, consistent with the distribution of tuft cells in mice, was observed ( Figure 1A ). Western blot of transfected HEK cells expressing ovine POU2F3 confirmed specificity of the anti-POU2F3 antibody ( Supplementary Figure 1 ). Double-immunostaining with antibodies to tuft cell transcription factors POU2F3 and GFI1b showed co-localization to the cell nucleus, confirmed by Hoescht co-staining ( Figure 1B ), providing further evidence that POU2F3 positive cells were indeed ovine tuft cells. Murine tuft cells are characterized by an apical tuft, and this structure could be identified on ovine tuft cells using antibody to villin ( Figure 1C ). POU2F3⁺ cells were evenly distributed throughout the epithelial layer from the base of the gastric glands to the apical surface. This contrasted with the labelling observed following staining of mucous cells (goblet-like cells of the abomasum/stomach) and extracellular mucus with Periodic Acid-Schiff (PAS), which concentrated towards the apical surface of the epithelium where surface/neck mucous cells are located, and did not co-localize to POU2F3⁺ cells ( Figure 1D ).

Using IHC, specific staining was not observed in ovine abomasal tissue with antibodies to DCLK-1 or TRPM5, in contrast to murine SI tuft cells (4–6), suggesting possible differences in tuft cells in mouse and sheep, or between gastric and SI tuft cells. Importantly, we established that the antibodies cross-react with the putative ovine homologues: specific labelling was observed with DCLK-1 antibody in ovine hippocampus tissue ( Supplementary Figure 2A ), consistent with expression of DCLK-1 in neurons (29). DCLK-1 antibody also showed reactivity with putative tuft cells in murine SI tissue, as expected ( Supplementary Figure 2C ), but specific antibody binding was not observed in ovine abomasum, duodenum or jejunum ( Supplementary Figures 2D–F ). TRPM5 antibody localized to tuft cells in the ovine duodenum and jejunum, but not in the abomasum ( Supplementary Figures 2A–I ). These results suggest that DCLK-1 is not expressed in ovine tuft cells, and that in sheep, TRPM5 may be present in intestinal but not gastric tuft cells.

The kinetics of the ovine tuft cell response to nematode infection was investigated using anti-POU2F3 antibody on abomasal tissue from T. circumcincta-infected sheep at days 5, 10 and 21 p.i. (Experiment 1, Supplementary Table 1 ). These time points capture key events in T. circumcincta development: at day 5 L4 stage larvae are present within the gastric glands, at day 10 immature adults are beginning to emerge from the glands, and at day 21 mature adults are present within the abomasal lumen (30). A significant increase in POU2F3⁺ cells was first observed at day 10 p.i. relative to helminth-naive counterparts, with a 5.2-fold increase at day 10 and 7.9-fold increase at day 21 compared to naïve animals (P < 0.05 and P < 0.01 for days 10 and 21 p.i., respectively) ( Figures 2A, B , and Supplementary Table 4 ). This is similar to data from mouse SI where an 8.5-fold increase in tuft cells was observed following N. brasiliensis infection (4).

We also examined, albeit at a later time point, tuft cells in sheep infected with H. contortus ( Supplementary Table 1 ), a highly pathogenic blood-feeding abomasal nematode related to human hookworms. Significantly more POU2F3⁺ cells were present in the abomasal epithelium at 55 days p.i. compared to uninfected sheep ( Supplementary Figure 3 ). This increase was lower than observed at day 21 of T. circumcincta infection which may reflect timing of the analysis, or variation in tuft cell responses to different nematode species. Nevertheless, significant increases in tuft cell frequency were observed after infection with both nematodes.

The tissue sections used in IHC were taken from the abomasal fundic region where T. circumcincta and H. contortus adult parasites predominantly localize. We also examined tuft cell numbers in different abomasal regions, by scoring sections from the cardiac (anterior), fundic (mid) and pyloric (posterior) regions following T. circumcincta infection. The percentage of POU2F3⁺ cells was greater in all three regions post-infection relative to helminth-naïve animals ( Supplementary Figure 4 ). Variation in the frequency of tuft cells in the different regions was observed between individual infected sheep, but across all animals there was no significant difference in the percentage of POU2F3⁺ cells across the different regions (P>0.05). In further studies we focused on the fundic region as representative of the general abomasal tuft cell response.

To establish how ovine tuft cells may be activated and respond to nematode infection, RNA profiling was carried out. As surface-expressed markers are not currently available for ovine tuft cells, fluorescent-activated cell sorting (FACS) using surface markers (5), was not possible. We therefore performed scRNA-seq to determine the gene expression profile of tuft and other epithelial and immune cells in the abomasal mucosa following nematode infection. Informed by our IHC data, tissue from the abomasal fundic region was collected at day 21 post-T. circumcincta infection from two sheep ( Supplementary Table 1 ) and subjected to high throughput droplet-mediated scRNA-seq using the 10x Genomics Chromium platform (https://www.10xgenomics.com/products/single-cell/) (24). Cell Ranger analysis showed that approximately 9,500 cells were sequenced from each sample. Sheep sample 1 produced over 503 million reads that mapped to the ovine (92%) and bovine (72%) genomes. More reads mapped to the bovine transcriptome (43%) compared to the ovine transcriptome (33%), due to better annotation of the bovine genome. Sheep 2 sample produced over 287 million reads, of which 45% and 40% mapped to the bovine and ovine transcriptomes, respectively. Approximately 17,000 genes were detected from each sample with median unique molecular identifier (UMI) counts of approximately 4,000 per cell.

Detailed analysis and clustering of scRNA-seq data was carried out using R toolkit Seurat 3.0 (http://satijalab.org/seurat/) (25). Ovine and bovine mapped data identified similar numbers of cell clusters; due to the higher number of mapped reads and better annotation, we focused on the bovine data. Datasets from the two samples were integrated using the canonical correlation analysis (CCA) method (26) and hierarchical clustering generated 15 distinct groups, (see Figure 3A ). Based on POU2F3 expression, a tuft cell cluster was identified and all clusters were annotated as far as possible ( Figure 3A ) based on expression of marker genes identified here ( Supplementary Table 5 ) and in previous studies of epithelial (31), tuft (9, 32) and immune cell genes. Clusters were enriched for epithelial cell adhesin EPCAM or immune cell marker PTPRC (CD45) ( Figures 3B, C ), except for clusters 4 and 10, identified as pre-B cells and granulocytes, respectively, in which few cells expressed either marker.

Most clusters showed overlap between samples from the two animals ( Figure 3D ) however fewer epithelial cells and more immune cell populations, particularly tissue macrophages, pre-B cells and neutrophils were contributed by sheep 2 ( Figures 3D, E ). At necropsy, sheep 2 showed gross inflammation of the abomasal mucosa, not observed in sheep 1, and had a higher adult worm burden and fecal egg count than sheep 1 (10,880 vs 6,910 adult worms and 432 vs 45 eggs per gram feces, for sheep 2 vs sheep 1, respectively). While these observations may explain the greater proportion of immune cell types sequenced in sheep 2, gene expression profiles were similar within the common cell clusters from the two sheep.

Four clusters showed strong expression of EPCAM and, based on expression of previously reported marker genes (9, 31, 32), were putatively identified as parietal, mucous, tuft and enterochromaffin-like (ECL) cells ( Figure 3C and Supplementary Table 6 ). The tuft cell cluster represented 1.1% of the total cells and 3.6% of epithelial cells ( Supplementary Table 7 ), and showed enriched expression of canonical tuft cell genes POU2F3, GFI1b, TRPM5 and IL-25 receptor IL17RB ( Figure 4A ). TRPM5 was expressed by a relatively low number of tuft cells (19 out of 189 tuft cells by scRNA-seq) which may, in part, explain the failure to detect TRPM5 by IHC. POU2F3 and GFI1B were also not detected in all cells of the putative tuft cell cluster, which likely reflects low expression of these transcription factors. Haber et al. (32) also reported a low percentage of cells expressing some murine canonical tuft cell genes. The identity of the ovine tuft cell cluster was supported by highly enriched expression of additional tuft cell marker genes (see below).

In contrast to murine SI tuft cells (4, 32), DCLK-1 was not identified in the ovine tuft cell cluster, despite being present in the ovine and bovine genomes. Tuft cell expression of DCLK-2, a paralog of DCLK-1, was observed ( Figure 4A ), and confirmed by in situ hybridisation using RNAscope (see below). Expression of CHAT, which encodes choline acetyltransferase and is highly enriched in murine SI tuft cells (5, 32), was almost exclusive to ovine tuft cells, but detected in only a small number (7/189) ( Figure 4A ). IL25, a key Th2 alarmin produced by tuft cells in mice, was not identified in the tuft cell cluster, possibly due to low levels of gene expression and/or the timing of analysis. However, as in the mouse, ovine tuft cells express high levels of the IL-25 receptor, IL17RB ( Figure 4A ). Figure 4B demonstrates enriched expression of selected marker genes from a previous study of gastric epithelial cells (31) and the top 10 marker genes of the four EPCAM-enriched cell clusters are shown by heat map in Figure 4C .

Clusters enriched for immune marker PTPRC (encoding CD45) could be putatively identified as T cells expressing CD4 or CD8, B or pre-B cells expressing CD19, CD40, CD79a/b, J chain and/or VPREB, or myeloid cells ( Supplementary Figure 5 and Supplementary Table 6 ) and the top five enriched genes expressed in each cluster are shown in Figure 4D . The complete list of marker genes is shown in Supplementary Table 5 , resulting in the first gene expression atlas of the ovine abomasal mucosa at the single cell level, available at http://cellatlas.mvls.gla.ac.uk/Ovine_Abomasum.

Among the top tuft cell-enriched genes were those encoding leukotriene C4 synthase (LTC4S), prostaglandin-endoperoxide synthase 1 (PTGS1), arachidonate 5-lipoxygenase (ALOX5) and arachidonate 5-lipoxygenase activating protein (ALOX5AP), all of which are involved in the synthesis of leukotriene and prostaglandin inflammatory mediators ( Figures 4C and 5A ). In addition, advillin (AVIL) and keratin 23 (KRT23) were highly enriched in ovine tuft cells, while keratin 18 (KRT18) was abundant, but not specific to tuft cells ( Figure 5A ). These have been identified as markers of murine SI tuft cells (14, 32), with advillin localizing to the apical tuft structure (33). Genes encoding regulators and mediators of signal transduction were also significantly enriched in ovine tuft cells, including Grp1-associated scaffold protein GRASP, inositol 1,4,5-triphosphate receptor-associated protein LRMP, guanine nucleotide-binding subunit gamma-13 GNG13, and Regulator of G Protein Signalling RGS13 ( Figure 5A ). Importantly, these ovine abomasal tuft cell enriched genes are among the most highly expressed genes in tuft cells from mouse SI (32) and are conserved in human tuft cells from different tissues (9). Figure 5B shows the overlapping set of tuft cell genes from the three species, with four genes (AVIL, ALOX, ALOX5AP, PTGS1) conserved in all three species, and 11 genes (GNG13, LRMP, LTC4S, RGS13, KRT18, KRT23, BMX, SPIB, SH2D7, TSPAN6, TMEM45B) conserved in ovine and murine tuft cells. This documents the conservation of signaling and inflammatory pathways in tuft cells across species and in different tissue locations.

In contrast, we found that G-protein coupled receptors (GPCRs), which may enable tuft cells to ‘sense’ parasites, diverge between tissues and species. Murine SI tuft cells are enriched for free fatty acid receptor FFAR3 (GPR41), succinate receptor SUCNR1 (GPR91) and orphan receptor GPRC5C (32, 34). However expression of SUCNR1 was very low in ovine abomasal tuft cells ( Figure 5C ), and was expressed mostly by cells in the basophil/mast cell cluster. GPRC5C was highly, but not specifically expressed in ovine tuft cells, while FFAR3 was detected in very few abomasal cells overall (3/16,891 total cells), as was taste receptor TAS1R3 ( Figure 5C ). This receptor is highly expressed in murine gastric tuft cells, particularly at the “gastric groove”, a tissue fold between the fundus and corpus of the rodent stomach (35), and in murine tuft cells in the distal SI (36). Notably, taste receptor TAS2R16 and adhesion receptor ADGRG6 (GPR126) were specifically enriched in ovine tuft cells ( Figure 5C ). Our scRNA-seq data are important in showing that while there is conserved expression of some GPCRs (TAS2R16, GPRC5C) in the SI and abomasum between species, others are expressed only in distinct GI locations, and suggest both tissue- and species-specific sensing. Therefore, while the downstream mediators of tuft cell function have been retained across species and GI regions, the ‘sensing’ receptors have diverged.

Haber et al. (32) identified two sub-clusters of differentiated tuft cells in murine SI: tuft-1, neuronal like, and tuft-2, expressing greater levels of immune genes (PTPRC (CD45) and TSLP). In the airways, two tuft cell sub-clusters were also defined, based on expression of genes associated with taste transduction (tuft-1) or leukotriene biosynthesis (tuft-2) (37). From ovine abomasal tuft cell scRNA-seq data, we identified three sub-clusters ( Figure 6A ), but these were not defined on expression of neuronal or immune-related genes. Sub-clusters 2 and 3 showed greater levels of overall gene expression (higher UMI) compared to sub-cluster 1 ( Figure 6B ). Most tuft cells from sheep 2 were in sub-cluster 2 ( Figure 6C ); whether sub-cluster type may influence infection outcome will require analysis from additional animals. All three sub-clusters were enriched for canonical tuft cell genes, however sub-cluster 1 showed lower expression of ALOX5AP, G-protein signaling genes GNG13 and RGS13, and GPCRs than sub-clusters 2 and 3, but greater expression of CHAT ( Figure 6D ). The top 10 marker genes of the three sub-clusters are shown Figure 6E and all sub-cluster marker genes are listed in Supplementary Table 8 . The top genes in sub-cluster 1 are involved in regulation of gene expression and cell proliferation (7SK.RNA, BTG2, AHNAK, EHD4), while those of sub-clusters 2 and 3 are associated with cell activation, proliferation, cellular interaction and immunity (CLCA1, C3, ENPP4, DAAM1, SPIB, S100A10, LGALS3, TFF3).

Trajectory inference and pseudotime analysis identified an expression trajectory, which we propose may start from sub-cluster 1 ( Figure 6F ), and gene expression across pseudotime showed dynamic patterns of up or down regulation ( Supplementary Figure 6 ). Gene Ontology (GO) analysis identified enrichment of genes associated with regulation of cell proliferation, extracellular space, cell surface and extracellular exosome in sub-cluster 1 (False Discovery Rate (FDR) <0.05)(e.g. GKN1, GKN2, AHNAK, ABL2, TSC22D2, NFKB1A), while GO term extracellular exosome (e.g. S100A10, LGALS3, TSPAN1, TFF3) was enriched for sub-cluster 3 genes. From this trajectory analysis, albeit from a low number of cells, we suggest that sub-clusters 1, 2 and 3 may represent early, intermediate and mature tuft cells, respectively.

RNAscope in situ hybridisation was carried out to confirm expression of selected tuft cell genes identified by scRNA-seq. Tissue fixation, probe hybridisation and signal detection were optimized using positive control probes to ovine beta actin (actb) and ovine peptidylprolyl isomerase (ppib) Figure 7D ), while negative control probe (DapB) showed no hybridisation, as expected ( Figure 7E ). Probes specific to POU2F3 localized to numerous cells in ovine abomasal tissue sampled following H. contortus infection, and co-localized with probes to IL17RB, TAS2R16 and DCLK-2 ( Figures 7A–C ). This supported co-expression of tuft cell genes identified by scRNA-seq. The signal was more abundant for IL17RB, than for DCLK-1 and TAS2R16, consistent with scRNA-seq data. To confirm expression and co-localization of tuft cell genes in additional samples, we also tested the probes on available tissue from naive sheep. Similar to the tissue sampled post-helminth infection, probes to IL17RB, TAS2R16 and DCLK-2 co-localized with POU2F3 probe, but in a lower number of cells, as expected from our IHC data ( Supplementary Figure 7 ). Most tuft cells were labelled with co-localized probes, while some singly labelled cells were also observed, suggesting possible heterogeneity between tuft cells, expression in additional cell types and/or differences in orientation of the cells during sectioning. RNAscope validated the scRNA-seq data and confirmed expression and co-localization of tuft cell genes.