We recently demonstrated the establishment of ectopic tuft cells in severely injured airways and alveolar areas post-IAV injury, which appear in locally large concentrations within dysplastic/remodeled Krt5⁺ regions (Figure 1A; Rane et al., 2019). To further investigate the nature of these cells, we utilized Trpm5-GFP reporter mice and sorted CD45^low/neg EpCAM^pos GFP^pos cells from post-IAV lungs followed by either ‘bulk’ RNA-Seq on the total population or single-cell RNA-Seq to reveal potential tuft cell heterogeneity (Figure 1B–C). Bulk RNA-Seq (Figure 1D–E) revealed a transcriptomic signature highly conserved with both intestinal and tracheal tuft cells (Nadjsombati et al., 2018), including enrichment in key tuft cell markers including Trpm5, Sox9, and Dclk1 (Figure 1D–E). In total, 898 genes were differentially expressed between purified tuft cells and the remaining lung epithelium (adjusted p-value <0.05). To directly compare influenza-induced lung tuft cells with tracheal tuft cells, we compared our bulk RNA-Seq data with published bulk RNA-Seq data from sorted tracheal tuft cells (Figure 1—figure supplement 1; Nadjsombati et al., 2018). Canonical tuft cell genes are expressed by post-influenza lung tuft cells at similar levels to tracheal tuft cells. An apparent distinction is reduced expression of Chat and Plcb2 by post-influenza lung tuft cells (Figure 1—figure supplement 1).

To evaluate tuft cell heterogeneity, we analyzed our single-cell RNA-Seq data on these cells (Figure 2A–D). To further enrich for tuft cells, we restricted analysis to cells expressing detectable Trpm5. Among these, we observed three to five populations of cells depending upon clustering variables (Figure 2A). One of these populations (cluster 2) co-expressed basal cell markers Krt5 and Trp63, suggesting that these were basal cells very early in their differentiation toward tuft cells, in agreement with our earlier demonstration that all post-IAV tuft cells are derived from intrapulmonary p63⁺ basal-like cells (Rane et al., 2019). To further confirm that tuft cells are derived from basal-like cells after flu, we performed lineage tracing using Krt5-CreERT2 and found that 90% of Dclk1⁺ cells were labeled in the alveoli, with tdTomato signal clearly visible in tuft cell nuclei (Figure 2E, n=323 Dclk1⁺ cells). Another population (merged clusters 0, 3, 4) bore relatively high mitochondrial gene reads, which we interpret as ‘stressed’ cells that nonetheless passed the quality control thresholds of data processing in Seurat. Whether this population is biologically relevant or a technical artifact is difficult to assess, though this ‘stressed’ population does contain subpopulations enriched for selected genes (Figure 2A, B).

The remaining population (largely cluster 1) bore neither stress-related genes nor basal cell genes and appeared to be heterogenous based on distribution in UMAP space (Figure 2A). Recent comprehensive single-cell analysis of the naïve murine tracheal airway epithelium suggested the existence of two distinct tuft cell subtypes denoted ‘Tuft-1’ (enriched in gustatory pathway genes) and ‘Tuft-2’ (enriched in leukotriene synthesis genes) (Montoro et al., 2018). Utilizing published gene sets for these two tuft cell subtypes, we performed gene module enrichment in Seurat, revealing a small subpopulation of ‘Tuft-1’ cells uniquely expressing known Tuft-1 genes, and a larger subpopulation of ‘Tuft-2’ cells, whose gene expression is relatively broad, but is expressed higher in this subpopulation than in ‘Tuft-1’ subpopulation (Figure 2D, Figure 2—figure supplement 1). To corroborate these as distinct identities in vivo, we performed immunostaining for ‘Tuft-1’ marker Gnb3 in Trpm5-GFP reporter mice, observing distinct Gnb3 expression in approximately 20% of Trpm5-GFP⁺ cells in situ (Figure 2F), though we did not observe any preferential localization of Gnb3⁺ cells.

To further investigate the influenza-induced lung tuft cell lineage, we utilized Slingshot to identify trajectories (Figure 2—figure supplement 2). Consistent with lineage tracing results, a trajectory from the ‘basal ->tuft’ cluster to the ‘Tuft-2’ and ‘Tuft-1’ clusters was identified, with the trajectory ending on the ‘stressed’ cluster (Figure 2—figure supplement 2B–C). A second trajectory analysis program, Monocle, produced similar pseudotime results (Figure 2—figure supplement 2D). Taken together, we conclude that injury-induced ectopic lung tuft cells are, like their homeostatic counterparts in the trachea, transcriptionally heterogenous, though the biological relevance of this heterogeneity remains to be determined.

In the small intestine, tuft cells are present in small numbers during homeostasis, but their prevalence increases rapidly upon infection with various Th2-associated pathogens in a manner dependent upon ILC2-derived IL-13 and tuft cell-derived IL-25. These cytokines facilitate a feed-forward loop to promote increased differentiation of Lrg5⁺ stem cells into tuft cells, ultimately increasing tuft cell numbers alongside increased fractions of goblet cells to promote pathogen clearance (Gerbe et al., 2016; Howitt et al., 2016; von Moltke et al., 2016). We therefore performed IAV infection in Pou2f3^-/-, Trpm5^-/- (required for tuft cell chemosensing), Il4ra^-/- (co-receptor required for both IL-13 and IL-4 signaling) and Il25^-/- knockout animals. As expected, Pou2f3^-/- mice entirely failed to develop lung tuft cells (Figure 3A–C, Figure 3—figure supplement 1). In Trpm5^-/- mice, tuft cells still differentiate after IAV injury, and we did not observe a significant change in tuft cell number at 25 days post injury (Figure 3E–G, Figure 3—figure supplement 2A, C, Figure 3—figure supplement 3B). Though we anticipated a recapitulation of the circuit found in the small intestine, we observed no difference in total numbers of tuft cells in either Il4ra^-/- (Figure 3I–K, Figure 3—figure supplement 2B, D, Figure 3—figure supplement 3C) or Il25^-/- animals (Figure 3M–O). An alternative strategy to eliminate IL-4ra signaling, using an inducible, ubiquitously expressed Cre and floxed ll4ra allele, again resulted in no significant difference in tuft cell numbers compared with controls (Figure 3—figure supplement 4). Consistent with independence from Type 2 cytokines, in normal mice following infection, we did not observe significantly increased levels of Il4, Il5, or Il13, at timepoints when tuft cells are present in the lung – 12, 21, or 51 days post infection, although we did observe significant induction of Type 2 cytokines at earlier timepoints (Figure 3—figure supplement 5A). Additionally, we found that the loss of tuft cells in the Pou2f3^-/- mice had no effect on viral clearance (Figure 3—figure supplement 5B), in line with the timing that flu-induced tuft cells arise at timepoints well after viral clearance (Rane et al., 2019).

To address if the presence or absence of post-injury tuft cells in the various mutants altered the formation of Krt5⁺ dysplastic cells, we systematically assayed the percentage of Krt5⁺ cell area in total lung area by scanning and quantifying whole lobe sections to ensure representation of all regions of the lung (Figure 3—figure supplements 1–3). We observed the expected variation of the percentages due to variable response to infection from animal to animal in all the mutants and corresponding controls. In Trpm5^-/-, Il4ra^-/-, and Il25^-/- mutants where tuft cells are present, we noted no apparent change in overall Krt5⁺ area of the injured lung compared to corresponding controls (Figure 3H, L and P). There was also no statistically significant difference of overall Krt5⁺ area in Pou2f3^-/- mutant where tuft cells are not present compared to control (Figure 3D). Consistently, we did not detect a significant difference in the percent of proliferating (Ki67⁺) alveolar Krt5⁺ basal-like cells at 14 or 25 days post infection (Figure 3—figure supplement 5C-D). While we cannot rule out differences in the behavior of other reparative cells, these findings indicate that post-injury tuft cells and Th2 immune signals (at least IL-4, IL-13, and IL-25) are apparently dispensable for the formation of Krt5⁺ cells in injured lungs. We note that experiments utilizing Pou2f3^-/- and Il4ra^-/- mice were performed independently in two separate vivariums at different institutions with equivalent outcomes, indicating stability of the phenotype with no obvious impact of potentially distinct housing environments (Figure 3—figure supplement 3A).

Type I (α/β) and type III (λ) interferons are highly expressed following influenza infection (Killip et al., 2015) and are crucial components of the innate anti-viral response (Pardy et al., 2019). As interferons play a crucial role early in flu infection and the action of type III interferon is restricted to the epithelium (Killip et al., 2015), we assessed whether they play a role in tuft cell development post IAV infection. Since IFNAR-deficient mice have been shown to have increased mortality when infected with flu (Arimori et al., 2013; Killip et al., 2015; Seo et al., 2011), we infected Ifnar1- and Il28r-deficient mice with a lower dose of PR8 than their BL/6 controls. Using this lower dose, the interferon receptor-deficient mice had an average weight loss of 22%, comparable morbidity to their BL/6 counterparts (Figure 4—figure supplement 1), and still bore abundant Krt5⁺ patches. In the absence of type III or type I interferon signaling, we observed no significant difference in tuft cell numbers following flu infection compared to their BL/6 comparators at 22 days post infection (Figure 4A-F). Therefore, type I and type III interferon signaling are dispensable in the development of tuft cells following PR8 infection.

Following the results that IFN signaling, which is commonly linked to viral infection, is not required for tuft cell formation, we sought to determine whether ectopic tuft cell development in the lung was specific to viral infection. To test if this is a general phenomenon occurring following lung injury, we treated mice with bleomycin, a chemotherapeutic agent that induces lung damage and subsequent Krt5⁺ areas in the lung (Vaughan et al., 2015). We identified tuft cells (Dclk1⁺) within Krt5⁺ patches in the lungs of bleomycin-treated mice by immunostaining, albeit sparser than in IAV-infected mice (Figure 4G, n=3 mice). These data demonstrated that tuft cell development can occur independently of infection and is a result of lung injury.

Amplification of tuft cells in the intestine promotes goblet cell metaplasia through Th2 cytokines (Gerbe et al., 2016; von Moltke et al., 2016). To address if the ectopic lung tuft cells and cytokines may perform a similar role, we stained for Agr2 (anterior gradient 2) as a cellular marker of goblet cells at 25 days after infection, a stage when goblet cells are robustly present (Chen et al., 2009; Di Valentin et al., 2009; Rane et al., 2019). To minimize variation due to the extent of injury, we normalized the area of Agr2⁺ cells to the area of Krt5⁺ cells. We found that in Pou2f3^-/-, Trpm5^-/-, and Il4ra^-/- animals, the percent of Agr2⁺ area within Krt5⁺ area was not significantly different from controls following influenza infection (Figure 5A–I). Consistent with this, Muc5b immunostaining in damaged regions of lung did not differ between Pou2f3^-/- or Il4ra^-/- lungs 25 days post infection (Figure 5J–M). In order to quantify mucus metaplasia in whole lungs of control and Pou2f3^-/- animals, we analyzed gene expression of goblet cell markers, Foxa3, Agr2, Muc5ac, and Muc5b, 51 days post infection. We did not observe a significant difference in relative expression of goblet cell markers when normalizing to Actb or to Krt5 as a marker of tissue damage (Figure 5N).

After IAV injury, basal-like cells are rarely observed to act as progenitors for AT1s or AT2s, though they do so at a higher frequency after bleomycin injury as indicated by lineage tracing with Krt5-creERT2 labeling AT2s (Vaughan et al., 2015; Yuan et al., 2019). As tuft cell differentiation is more heterogeneous in bleomycin compared with influenza, we investigated whether tuft cells are involved in preventing normal alveolar differentiation in Krt5⁺ areas. As shown above, we observed no change in total Krt5⁺ area of the of lung with Pou2f3 deletion (Figure 3A–D). To address if a relatively minor population of Krt5⁺ cells may be converting to AT2s, we double stained for Krt5 and the AT2 cell marker SPC⁺ in Pou2f3^-/- and control lungs at 25 days after infection to label cells in the process of conversion (Figure 6A). We did not observe any overlap of staining in either Pou2f3^-/- or control lungs. To more rigorously trace possible conversion, we bred Pou2f3^-/- mice to Krt5-CreERT2 fate mapping mice to lineage trace Krt5⁺ cells and assess whether any increased plasticity may occur in the absence of tuft cells (Figure 6B). Unlike tuft cells, which could be lineage traced from Krt5⁺ precursors, we did not observe any appreciable conversion of Krt5⁺ precursors into AT2 (SPC⁺) cells in the presence or absence of tuft cells (Figure 6C–D). Taken together, while ectopic lung tuft cells likely possess as-of-yet undiscovered function in post-IAV lungs, it does not appear that they play an important role in restricting Krt5⁺ epithelial cell plasticity, nor are they responsible for the inability of Krt5⁺ cells to efficiently differentiate into more regionally appropriate alveolar cell types.

All animal procedures were approved by the Institutional Animal Care and Use Committee of the University of Pennsylvania, the University of California San Diego (UCSD), and the University of California San Francisco (UCSF): Il25^-/- (Fallon et al., 2006), Il4ra^-/- (Noben-Trauth et al., 1997), Pou2f3^-/- (Matsumoto et al., 2011), Trpm5^-/- (Damak et al., 2006), Trpm5-GFP (Clapp et al., 2006), Krt5-creERT2 (Van Keymeulen et al., 2011), Ai14 (Madisen et al., 2010). Il4ra^fl (Herbert et al., 2004) and Rosa26-ERT2-Cre (Ventura et al., 2007) were used for inducible knockout of Il4ra, animals received 2 mg tamoxifen by i.p. injection every other day from 7 to 14 days post infection. The Ifnar1^-/- and Il28r^-/- mice were received from the Fuchs and Striepen laboratories, respectively (University of Pennsylvania). For experiments at University of Pennsylvania, adult mice of both sexes were used in relatively equal proportions, for UCSD, 8- to 10-week-old mice (<25 g) of both sexes were used in equal proportions and all mice are on a C57BL6/J background unless otherwise noted. The protocol number associated with the ethical approval of this work is 806262 (University of Pennsylvania) and S16187 (UCSD). For all animal studies, no statistical method was used to predetermine sample size. The experiments were not randomized, and the investigators were not blinded to allocation during experiments and outcome assessment.

All viral infections utilized influenza strain A/H1N1/PR/8 obtained from Dr Carolina Lopez (Garcia et al., 2020). For influenza infection at the University of Pennsylvania, virus was administered intranasally. Mice ranging between 15 and 20 g in weight were infected with 30 tissue culture infectious dose (TCID)50 units of PR8, mice weighing between 20 and 25 g were given 40 TCID50 units, and mice ranging between 25 and 30 g in weight were given 50 TCID50 units. Due to increased mortality in the Ifnar-deficient mice (Arimori et al., 2013; Seo et al., 2011), interferon receptor-deficient mice were specifically infected with 10 TCID50 units less than their BL/6 controls. Briefly, mice were anesthetized with 3.5% isoflurane for 5 min until bradypnea was observed. Virus dissolved in 30 µl of PBS was pipetted onto the nostrils of anesthetized mice, whereupon they aspirated the fluid directly into their lungs. Infections performed at the UCSF (Figure 3—figure supplement 3A) were performed nearly identically, as previously described (Vaughan et al., 2015; Xi et al., 2017). A/H1N1/PR/8 infection was also performed independently at the UCSD, with virus obtained initially from ATCC (VR-95PQ). In all cases, control and experimental groups were infected simultaneously in the same cohort by the same investigator so that direct comparison between groups was justified and appropriate.

For the bleomycin injury model, mice were anesthetized as above and treated intranasally with a single dose of bleomycin (Cayman Chemicals) at 2.25 mg/kg and harvested 22 days post treatment.

Lungs were harvested from mice and single-cell suspensions were prepared as previously described (Zhao et al., 2020). Briefly, the lungs were thoroughly perfused with cold PBS via the left atrium to remove residual blood in the vasculature. Lung lobes were separated, collected, and digested with 15 U/ml dispase II (Thermo Fisher Scientific, #17105041) in PBS for 45 min at room temperature (RT) and mechanically dissociated by pipetting in sort buffer (DMEM + 2% CC + 1% P/S, referred to as ‘SB’). Next, cell suspensions were filtered by the 40 μm cell strainer (Thermo Fisher Scientific, #352340) and treated by Red Blood Cell Lysis Buffer (Thermo Fisher Scientific, A1049201) for 5 min, and the cell suspension was incubated in SB containing 1:1000 DNase I (Millipore Sigma, #D4527) for 45 min at 37°C. Whole-lung cell suspensions were then used for subsequent experiments.

Whole-lung single-cell suspensions were prepared as above and then blocked in SB containing 1:50 TruStain FcX (anti-mouse CD16/32) Antibody (BioLegend, #101319) for 10 min at 37°C. The cell suspension was stained using allophycocyanin/Cy7-conjugated rat anti-mouse CD45 antibody (1:200, BioLegend, #101319), PE-conjugated rat anti-mouse EpCam antibody (1:500, BioLegend, G8.8, #118206) for 45 min at 4°C. Stained cells and ‘fluorescence minus one’ controls were then resuspended in SB + 1:1000 Dnase + 1:1000 Draq7 (BioLegend, #424001) as a live/dead stain. All FACS sorting was done on a BD FACSAria Fusion Sorter (BD Biosciences).

One-thousand to 3000 EpCam + Trpm5-GFP⁺ cells from mice at day 43 post influenza were sorted directly into lysis buffer from Takara SMART-Seq v4 and RNA/cDNA was amplified according to the manufacturer’s instructions. All downstream library preparation and sequencing was performed by the Next-Generation Sequencing Core at the Perelman School of Medicine, University of Pennsylvania. In brief, libraries were sequenced on a NovaSeq sequencer at 100SR, raw FASTQ files were imported into R and reads mapped with Kallisto, and differential expression performed by Limma. Raw data is deposited at GEO.

Single-cell RNA-Seq was performed using the Chromium System (×10 Genomics) and the Chromium Single Cell 3’ Reagent Kits v2 (×10 Genomics) at the Children’s Hospital of Philadelphia Center for Applied Genomics. As with bulk RNA-Seq, ~3000 EpCam + Trpm5-GFP⁺ cells were sorted into PBS + 0.1% BSA from mice at day 28 post influenza and loaded onto the ×10 Chromium system. After sequencing, initial data processing was be performed using Cellranger (v.3.1.0). Cellranger mkfastq was used to generate demultiplexed FASTQ files from the raw sequencing data. Next, Cellranger count was used to align sequencing reads to the mouse reference genome (GRCm38) and generate single-cell gene barcode matrices. Post-processing and secondary analysis was performed using the Seurat package (v.4.0). First, variable features across single cells in the dataset will be identified by mean expression and dispersion. Identified variable features was then be used to perform a PCA. The dimensionally reduced data was used to cluster cells and visualize using a UMAP plot. Contaminating non-tuft cells were removed by sub-setting data, requiring counts for Trpm5 > 1. Slingshot and Monocle were used for trajectory analysis. High-throughput sequence data is available at GEO.

Each lung was thoroughly perfused with PBS via the left atrium and then perfused with 1 ml of 3.2% paraformaldehyde (PFA, Thermo Fisher Scientific) and placed in a 50 ml tube with 25 ml of PFA to shake at RT for 1 hr. Following the 1 hr incubation, the PFA was replaced with 25 ml of PBS every 20 min for the next hour while shaking at RT. The lungs were then placed in 30% sucrose (Sigma-Aldrich) overnight shaking at 4°C. The following day, the tissues were placed in 15% sucrose-50% optimal cutting temperature compound (OCT, Fisher Healthcare) shaking for 2 hr at RT. The fixed lungs were then embedded in OCT, flash frozen using ethanol and dry ice and stored at –80°C. Using a cryostat, the lungs were then section (6 µm) and stored at –20°C.

Lung sections were fixed with 3.2% PFA for 5 min at RT and then washed three times with PBS for 5 min at RT while gently shaking. Slides were then blocked for 1 hr at RT in a humid chamber with blocking buffer ([1% BSA, Gold Bio], 5% donkey serum [Sigma], 0.1% Triton X-100 [Fisher BioReagents] and 0.02% sodium azide [Sigma-Aldrich] in PBS). The slides were then stained in blocking buffer overnight at 4°C with a combination of primary antibodies. The following day, the slides were washed three times for 5 min while gently shaking at RT with PBS + 0.1% Tween (Sigma) and stained for 90 min in blocking buffer with a combination of secondary antibodies. Slides were then washed three times for 5 min while gently shaking at RT with PBS + 0.1% Tween and stained with DAPI (1:10,000 dilution; catalog no. D21490, Thermo Fisher Scientific), for 7 min and washed in PBS + 0.1% Tween as mentioned above. Slides were then mounted with Fluoroshield (Sigma) and imaged using a Leica inverted fluorescent microscope Dmi8 and analyzed using Las X and Fiji softwares.

Primary antibodies used: rabbit anti-Dclk1 (1:500 dilution; catalog no. ab37994 or ab31704, Abcam), chicken anti-Krt5 (1:1000 dilution; catalog no. 905901, BioLegend) sheep anti-eGFP (1:500, Thermo Fisher Scientific, OSE00001G), rabbit anti-Gnb3 (1:200, 10081-1-AP, Proteintech), rabbit anti-POU2F3 (1:500, Sigma, HPA019652), rabbit anti-Agr2 (1:200, Cell Signaling Technology, 13062), rabbit anti-Muc5b (1:400, Cloud-Clone Corp, PAA684Mu01), rabbit anti-Pro-SPC (Seven Hills Bioreagents, WRAB-9337), rabbit anti-Ki67 (Abcam, ab15580). Secondary antibodies used: donkey anti-rabbit AF568 (1:1000 dilution; catalog no. A10042, Thermo Fisher Scientific), donkey anti-rabbit AF647 (1:1000 dilution; catalog no. A31573, Thermo Fisher Scientific), donkey anti-chicken AF488 (1:1000 dilution; catalog no. 703-545-155, Jackson ImmunoResearch), donkey anti-sheep AF488 (1:1000 dilution, catalog no. A11015, Thermo Fisher Scientific).

For experiments performed at the UCSD (Figures 3A–L–5A–I, 6C–D): lung sections were imaged on a Nikon A1 microscope using a 20×0.75 NA objective. Nikon Elements Jobs Module was used for automating multiple slide imaging, batch stitching, and maximum intensity projection of images. An analysis pipeline using Nikon Elements General Analysis 3 was used to analyze images in batch. Imaging processing split channels for individual analysis. For total tissue area, DAPI signal area was recorded. For area quantifications, a Gaussian filter was applied, an intensity threshold and size threshold was set manually, and the area of the resulting binary mask was recorded. For quantification of Dclk1⁺ cells, a Gaussian filter was applied, and intensity and minimum size thresholds were set manually. The Dclk1⁺ count was verified and adjusted manually. For Krt5 and SPC staining (Figure 6A), Krt5 lineage tracing (Figure 6B), and Muc5b staining (Figure 5J–M), left lobe lung sections were scanned on an Olympus VS200 slide scanner using a ×20 objective.

For experiments performed at the University of Pennsylvania, sections were imaged on a Leica Dmi8 as described above. Image quantification was performed by manual area measurements utilizing ImageJ/FIJI for Krt5 area and manual quantification of Dclk1⁺ cells as at UCSD (Figure 3M-P) (Figure 3—figure supplement 3A). For interferon signaling deficient experiments (Figure 4A-F), the Krt5 area quantification was performed utilizing the ImageJ/FIJI software, setting the auto threshold, using the Otsu method, and manual quantifying Dclk1⁺ cells.

Tamoxifen dissolved in corn oil was administered by i.p. injection at 25 mg/kg.

Lungs from Pou2f3^+/- and Pou2f3^-/- mice infected with PR8 were homogenized in serum-free media. Virus titer was determined by TCID50 assay using MDCK cells (ATCC, CCL34), and the presence of virus in the supernatant was determined by hemagglutination assay. Hemagglutination assay and TCID50 calculations were performed as previously published (Xue et al., 2016).

Whole-lung lobes were dissected into Trizol (Invitrogen) and Rneasy Mini RNA Extraction Kit (Qiagen) was used to extract total RNA. RT-PCR was performed using iScript Select cDNA Synthesis Kit (Bio-Rad). qPCR was performed on a CFX Connect system (Bio-Rad) using SYBR Green (Bio-Rad). Three technical replicates were performed for each target gene.

All statistical calculations were performed using GraphPad Prism. p-Values were calculated from unpaired two-tailed t-tests with Welch’s correction or ANOVA for multivariate comparisons. Variance was analyzed at the time of t-test analysis. This data is not included in the manuscript but is available upon reasonable request.