Cytotoxic CD4⁺ tissue-resident memory T cells are associated with asthma severity

SUMMARY

Background

Patients with severe uncontrolled asthma represent a distinct endotype with persistent airway inflammation and remodeling that is refractory to corticosteroid treatment. CD4⁺ T_H2 cells play a central role in orchestrating asthma pathogenesis, and biologic therapies targeting their cytokine pathways have had promising outcomes. However, not all patients respond well to such treatment, and their effects are not always durable nor reverse airway remodeling. This observation raises the possibility that other CD4⁺ T cell subsets and their effector molecules may drive airway inflammation and remodeling.

Methods

We performed single-cell transcriptome analysis of >50,000 airway CD4⁺ T cells isolated from bronchoalveolar lavage (BAL) samples from 30 patients with mild and severe asthma.

Findings

We observed striking heterogeneity in the nature of CD4⁺ T cells present in asthmatics’ airways with tissue-resident memory (T_RM) cells making a dominant contribution. Notably, in severe asthmatics, a subset of CD4⁺ T_RM cells (CD103-expressing) was significantly increased, comprising nearly 65% of all CD4⁺ T cells in the airways of male patients with severe asthma when compared to mild asthma (13%). This subset was enriched for transcripts linked to T cell receptor (TCR) activation (HLA-DRB1, HLA-DPA1) and cytotoxicity (GZMB, GZMA) and, following stimulation, expressed high levels of transcripts encoding for pro-inflammatory non-T_H2 cytokines (CCL3, CCL4, CCL5, TNF, LIGHT) that could fuel persistent airway inflammation and remodeling.

Conclusions

Our findings indicate the need to look beyond the traditional T2 model of severe asthma to better understand the heterogeneity of this disease.

eTOC blurb:

Using single-cell transcriptomics, Herrera-De La Mata et al. report that CD103-expressing CD4⁺ T_RM cells with cytotoxic and pro-inflammatory features are increased in the airways of males with severe asthma and are likely to contribute to the pathogenesis of asthma severity.

Graphical Abstract

INTRODUCTION

Asthma is one of the most common chronic diseases affecting children and adults.^1–4 The classical symptoms of asthma like wheeze, cough, and breathlessness, are triggered by inflammation and remodeling of the airways that results in airway hyperreactivity and obstruction.^2,5,6 The airway inflammation in asthmatic patients is considered to be driven by type 2 cytokine-producing CD4⁺ T cells (T_H2) that play a central role in orchestrating recruitment and activation of innate immune cells such as eosinophils, basophils, and mast cells.^2,3,5–7 Suppressing airway inflammation with corticosteroids remains the mainstay of treatment for most patients with asthma.^8–10 However, many patients with severe forms of asthma fail to respond well to corticosteroids and suffer from persistent symptoms and recurrent life-threatening exacerbations.⁸ Moreover, the recently approved biological agents blocking type 2 cytokines or immunoglobulin (Ig) E are neither uniformly effective nor do they reverse disease pathogenesis in severe asthmatics.^3,11–14 These observations raise the possibility that other effector CD4⁺ T cells may contribute to airway inflammation and remodeling in severe asthmatics.

Functional studies in model organisms have implicated a wide-range of CD4⁺ T cell subsets such as T_H2, T_H9, T_H17, T_H1, follicular helper T (T_FH) cells, and tissue-resident memory T (T_RM) cells in the pathogenesis of allergic airway inflammation.^2,15–23 In murine models, T_RM cells have also been shown to have features that could support airway remodelling and fibrosis, important facets of severe asthma.^2,24–26 Because T_RM cells are mainly localized to the barrier sites²⁷ with only a very minor population recirculating in the blood, understanding the biology of T_RM cells in asthma would require analysis of airway specimens.^26,28,29 Targeted immunophenotyping studies performed in airway specimens from asthmatic patients have shown associations for certain CD4⁺ T cell subsets with asthma pathogenesis.^21,29–31 However, these studies do not provide an unbiased and precise characterization of CD4⁺ T cell subsets associated with asthma severity. The difficulty in obtaining airway specimens from severe asthmatics and limitations in the number of cells available for research have further compounded this problem. Single-cell genomic assays offer an attractive solution to address this problem in a hypothesis-free manner, and, importantly, such studies have yielded vital insights into the pathogenesis of human autoimmune diseases and cancer.^32–39 While a recent single-cell transcriptomic study reported significant increase in the proportion of T_H2 cells in patients with mild asthma compared to healthy controls,⁴⁰ it is not known whether this observation holds true for severe disease or if other CD4⁺ T cell subsets are involved as well.

Here, to define the CD4⁺ T cell subsets and their properties associated with severe asthma and corticosteroid resistance, we performed single-cell RNA-seq assays from purified CD4⁺ T cells isolated from the airways of patients with severe and mild asthma. Our unbiased approach found a significant association between the abundance of cytotoxic CD4⁺ T_RM cells and asthma severity in males that we hypothesize is uniquely relevant for driving their airway inflammation and remodeling.

RESULTS

Single-cell transcriptomic analysis reveals heterogeneity among airway CD4⁺ T cells in asthma

To understand why the disease phenotype shifts from an easily treatable and largely reversible condition in mild asthma to a permanent airway inflammation and impaired lung function in severe asthma, we focused our analysis to specifically identify differences in the properties of CD4⁺ T cells present in the airways of severe asthmatic patients. We performed bronchoscopy and collected bronchoalveolar lavage (BAL) from patients with severe asthma (n=16, 50% male) from the Wessex AsThma CoHort of difficult asthma (WATCH) study,⁴¹ which has extensively characterized patients with severe asthma requiring ‘high dose’ and/or ‘continuous or frequent use of oral corticosteroid’, in accordance with global initiative of asthma (GINA) management step 4 and 5^42,43 (Figure 1A; and Table S1A). To specifically identify differences that are associated with asthma severity, as controls, we enrolled patients with mild asthma (n=14, 64% male) (Figure 1A; and Table S1A). At the time of bronchoscopy, the majority of mild asthmatic patients (10/14) were on inhaled steroids, but at a significantly lower dose than the severe asthmatic patients, as it would be unethical to withhold them from regular asthma medications (Table S1A). By only using non-asthmatic patients as controls, we would have failed to discriminate changes that are linked to asthma versus asthma severity. However, we have included data from published studies^29,44–49 to draw relevant comparisons with non-asthmatic subjects.

Figure 1. Single-cell transcriptomic analysis reveals heterogeneity among airway CD4⁺ T cells.

(A) Study overview. (B) Uniform manifold approximation and projection (UMAP) visualization of Seurat-based clustering analysis of 27,771 single-cell transcriptomes of ex vivo sorted CD4⁺ T cells obtained from 9 mild and 16 severe asthmatic patients. Each dot represents a cell and is colored based on cluster type. Proportion of cells in each cluster is shown (parenthesis). (C) Heatmap of row-wise z-score-normalized mean expression of significantly enriched transcripts in each cluster. Adjusted P-value < 0.05 and log₂ (fold change) > 0.25. (D) Row-wise z-score-normalized mean expression (color scale) and percent of expressing cells (size scale) plot for a selection of marker genes in each cluster. (E) UMAP shows T_RM signature score (color scale) for each cell. Clusters are delineated by colored lines. (F) GSEA between CD103⁺ T_RM cluster (top) and CD103^– T_RM cluster (bottom) versus all non-T_RM clusters using published T_RM signature gene lists (Table S3A). NES, normalized enrichment score; q, false discovery rate. (G) Violin plot displays normalized expression of ITGAE (CD103) in T_RM clusters (CD103⁺ T_RM and CD103^– T_RM) compared to T_CM cluster. (H) Violin plots show normalized protein expression of CD103 and CD69 in T_RM clusters compared to T_CM cluster (analysis done for 6 severe asthmatic patients).

To define the airway CD4⁺ T cell subsets specifically associated with asthma disease severity and potentially contributing to persistent inflammation and unresponsiveness to current asthma treatment, we isolated and purified CD4⁺ T cells (n >40,000) from BAL specimens and performed single-cell RNA-seq assay using the oil droplet-based platform (10x Genomics). After exclusion of doublets and low-quality transcriptomes, clustering analysis of airway CD4⁺ T cells (n=27,771 cells; median ~1,304 cells/donor) revealed 8 transcriptionally distinct subsets (see STAR Methods; Figure 1B; Figures S1A-D; and Tables S2A-C).

The molecular identity for each cluster was determined based on enrichment analysis of canonical markers and signature gene lists linked to established CD4⁺ T cell subsets (Figures 1C and 1D; Figures S1E and S1F; and Tables S2C and S3A). The two largest clusters (cluster 1 and 2), representing 71% of all CD4⁺ T cells analyzed, were significantly enriched for T_RM signature genes^40,50,51 (Figures 1D-F; and Figure S1E), and thus were defined as T_RM cells. Notably, both T_RM clusters expressed high levels of the T_RM marker gene CD69⁵¹ (Figure 1D), whereas T_RM cells in cluster 1 expressed higher levels of another T_RM marker gene, ITGAE (Figure 1D and 1G), which encodes for the alpha chain of the integrin CD103, a transmembrane protein required for the adhesion of T cells to E-cadherin expressed by epithelial cells.^50,52,53 By labeling BAL cells with anti-CD69 and anti-CD103 DNA-barcoded antibodies prior to performing single-cell RNA-sequencing, we confirmed the enrichment of CD103-expressing T_RM cells in cluster 1 (Figure 1H); henceforth, for simplicity, we refer to cluster 1 and 2 as CD103⁺ T_RM and CD103^– T_RM clusters, respectively.

The third largest cluster (cluster 3=13% of airway CD4⁺ T cells) was enriched for the expression of gene signatures linked to central memory T (T_CM) cells^50,54–56 (Figure 1D; and Figure S1E). Cluster 4 (4%) and 5 (4%) cells were enriched for gene signatures linked to regulatory T cells (T_REG)^57,58 and T_FH cells,⁵⁹ respectively (Figure 1D; and Figure S1E). Cluster 6 (3%) was highly enriched for type I and II interferon response gene signatures reminiscent of the recently described interferon response genes expressing helper T cell subset (T_IFNR) with a potential regulatory function in allergy²² (Figure 1D; and Figure S1E). Two other relatively small clusters of cells, cluster 7 (1%), enriched for cell cycle signature genes, represented proliferating cells,⁶⁰ and cluster 8 (1%), which expressed high levels of several transcripts encoding for cytotoxicity molecules (GNLY, PRF1, GZMB), represented cytotoxic CD4⁺ T cells⁶¹ (Figure 1D; and Figure S1E). Gene set enrichment analysis for signatures linked to canonical T helper (T_H) effector subsets (Figure S1E; and Table S3A) showed significant positive enrichment of T_H1 signature genes in CD103⁺ T_RM cluster, but T_H17 and T_H2 signature genes were not positively enriched in any of the CD4⁺ T cell clusters (Figures S1E and S1G). Overall, our results highlight a substantial level of heterogeneity in the profile of CD4⁺ T cells present in the asthmatic airways, with T_RM subsets making up a major contribution.

CD103⁺ T_RM cells are significantly increased in the airways of males with severe asthma

We next determined the CD4⁺ T cell subsets that were quantitatively increased in the airways of patients with severe asthma and assessed their association with clinical and physiological parameters of asthma severity. Because biological sex has been shown to stratify severe asthmatics into distinct endotypes, where males with severe asthma display poor lung function and high use of maintenance corticosteroids,^62,63 we also explored the influence of biological sex on the composition of airway CD4⁺ T cells in severe asthma. Among the CD4⁺ T cell subsets, the proportion of cells in the CD103⁺ T_RM subset was significantly increased in the airways of patients with severe asthma compared to mild asthma (46% versus 21%; P < 0.05), while the proportions of CD103^– T_RM subset were significantly decreased (22% versus 44%; P < 0.05) (Figure S2A). Interestingly, while the proportions of the CD103⁺ T_RM subset were not significantly different between males and females (Figure S2B), we observed that the increase in the proportion of CD103⁺ T_RM subset was significant only when comparing male patients with severe and mild asthma (Figures 2A-C; 64% versus 13% of all CD4⁺ T cells in severe versus mild asthma; P < 0.01). The proportions of the CD103^– T_RM cluster as well as the T_CM cluster were significantly lower in male severe asthmatics (Figures 2A-C; P < 0.01 and P < 0.05, respectively). All other airway CD4⁺ T cell subsets showed no significant differences between mild and severe asthmatics in both males and females (Figure 2C; and Figure S2C).

Figure 2. CD103⁺ T_RM cells are significantly increased in the airways of males with severe asthma.

(A) Pie charts represent average proportions of CD4⁺ T cell subsets in the 4 clinical groups: mild asthmatic (MA) and severe asthmatic (SA) patients separated by sex. Colors correspond to cluster type. (B) Normalized stacked bar charts represent the proportions of CD4⁺ T cell clusters per donor for the 4 clinical groups. Colors correspond to cluster type. (C) Dot plots show proportions of CD103⁺ T_RM, CD103^– T_RM, T_CM, T_REG, and T_FH clusters for the 4 clinical groups (*, P < 0.05; **, P < 0.01; Mann-Whitney U test). (D) Dot plots show proportions of CD103⁺ T_RM, CD103^– T_RM, non-T_RM, T_REG, and T_FH cells for the 4 clinical groups (*, P < 0.05; **, P < 0.01; ***, P < 0.001; Mann-Whitney U test). (E) Representative contour plot showing the expression of CD69 versus CD103 from CD4⁺ T cells by flow cytometry from two donors, one mild and one severe asthmatic. (F) Dot plots show proportions of CD103⁺ T_RM cluster in severe asthmatics off (–) or on (+) oral corticosteroids (OCS; left) or biologics (right) treatment separated by sex (Mann-Whitney U test). (G) Scatter correlation plots between proportions of cells in CD103⁺ T_RM (top) or CD103^– T_RM (bottom) cluster with clinical features (composite asthma severity score and 100% - post-bronchodilator FEV1/FVC %). Each dot represents data from a single patient and are colored and shaped based on the 4 clinical groups. Correlation coefficient r and P value were computed using Spearman correlation analysis (trendline black). (H) Dot plots show proportions of CD4⁺CD69⁺CD103⁺ T_RM cells (top) and CD4⁺CD69⁺CD103^– T_RM cells (bottom) in BAL samples measured by flow cytometry from healthy (green) donors and asthmatic (blue = mild to moderate asthma; red = severe asthma) donors (unspecified sex) (**, P < 0.01; Kruskal-Wallis one-way test followed by Dunn’s post-hoc test). Data obtained from published datasets #1,⁴⁴ #2,⁴⁵ and #3.²⁹ (C, D, F, H) Each dot represents data from a single subject, bars represent the mean, and t-lines represent SEM.

We validated the findings from single-cell transcriptome analysis by performing flow cytometric analysis of BAL cells from the same study participants. Based on the expression of cell surface markers, we classified airway CD4⁺ T cells into 5 subsets: two types of T_RM cells (CD69⁺CD103^–, CD69⁺CD103⁺), non-T_RM cells (CD69⁻), follicular helper T cells (T_FH, CXCR5⁺GITR^–), and regulatory T cells (T_REG, CXCR5^–CD127⁺CD25⁺) (Figures 2D and 2E; Figures S2D-G; and Table S1D). As expected, we found that the proportions of CD103⁺ T_RM cells were significantly increased in males with severe asthma (50%) compared to mild asthma (22%) (Figure 2D; P < 0.001). We observed no significant differences in the proportions of CD103⁺ T_RM cells between males with severe asthma who are currently on or off maintenance oral corticosteroids [OCS] or biologics (Figure 2F; and Tables S1A and S2A). Together, these findings establish a significative association between the frequency of CD103⁺ T_RM cells and asthma severity in male asthmatic patients that is not dependent on drug treatments.

We then performed unbiased Spearman correlation analysis in a sex-specific manner to investigate any association between the identified CD4⁺ T cell clusters with clinical and physiological parameters related to asthma severity (Figure S2H; and Table S1A and S1E). Among the significant associations (Figure S2H), we observed a positive correlation between the proportions of CD103⁺ T_RM cell subset and a composite asthma severity score (adapted from the Asthma Severity Scoring System [ASSESS])⁶⁴ in male but not female asthmatics (r_S = 0.8 and P < 0.01; Figure 2G; Figure S2H; and Table S1E). The composite asthma severity score is a score ranging from 0–20 from the Severe Asthma Research Program [SARP]⁶⁵ comprising of measurement of asthma symptoms, quality of life, degree of airflow obstruction, use of corticosteroids or biologics, and frequency of asthma exacerbations requiring oral corticosteroids and hospitalization (Figure 2G; Figure S2H; and Tables S1A and S1B). This asthma severity score was used at the time of bronchoscopy to classify these asthmatic patients (not at the time of diagnosis), and its reliability has been recently validated,⁶⁶ providing a more objective measure of the magnitude of treatment response in individual patients and phenotypic groups.⁶⁴ We also found a significant positive correlation with the severity of airflow obstruction (using post-bronchodilator FEV₁ and FEV₁/FVC) in males (r_S = 0.7 and P < 0.01; Figure 2G; Figure S2H; and Tables S1A and S1E). Notably, the proportions of airway T_REG cells were negatively correlated with the proportions of CD103⁺ T_RM cell subset in males (r_S = 0.7 and P < 0.05) and with the severity of airflow obstruction in males (Figures S2H and S2I; and Table S1E), suggesting a potential imbalance between CD103⁺ T_RM and T_REG subsets in the airways of males with severe asthma. Finally, we found that age was positively correlated with the proportions of CD103⁺ T_RM subset in the airways of male severe asthmatics (r_S = 0.7 and P < 0.05; Figures S2H and S2J; and Table S1E). The fact that 7 of the 8 male participants studied developed late-onset severe asthma (age >40 years), suggests that older age and late-onset of disease may also be linked to this specific immune profile observed in males with severe asthma (Figures S2H and S2J). To specifically address if older age as opposed to the male severe asthma phenotype was associated with increased frequency of CD103⁺ T_RM cells in the airways, we re-analyzed three published BAL T cell datasets from non-asthmatic healthy subjects (Table S1F)^29,44,45. We observed that the proportion of CD103⁺ T_RM cells in BAL samples from non-asthmatic healthy subjects was lesser (mean = 16%) when compared to our data in males with severe asthma (mean = 50%) (Figures 2D and 2H; and Table S1F). Notably, in the study by Tang et al., 2022 (#2),⁴⁵ 18 of the 20 study participants were older than 60 years (Table S1F), suggesting that older age per se is not associated with increased frequency of CD103⁺ T_RM cells in the airways. Furthermore, our re-analysis of BAL data from the recent study by Camiolo et al., 2021 (#3),²⁹ which also included patients with mild and severe asthma, showed high frequency of CD103⁺ T_RM cells only in some patients in the severe asthma group but not those in the mild asthma or healthy control groups, thus strengthening the association with asthma severity (Figure 2H; and Table S1F).

CD103⁺ T_RM subset displays qualitative features linked to TCR activation and cytotoxicity

To determine the molecular properties of cells in the CD103⁺ T_RM subset that may contribute to the pathogenesis of asthma severity, we first compared expression profiles of cells from the CD103⁺ T_RM subset (cluster 1) with those from the CD103⁻ T_RM subset (cluster 2) across all patient groups, using sex as a covariate. We observed major transcriptional differences between the two T_RM subsets with over 1,300 differentially expressed transcripts (Figure 3A; and Table S2D), and, importantly, these differences were mostly preserved when differential gene expression analysis (DGEA) between the two T_RM subsets was performed separately in males and females (Figure 3B; and Table S2E). Several genes involved in cytotoxic function (GZMB, GZMA, GZMH, FASLG)⁶¹ were increased in expression in the CD103⁺ T_RM subset (Figures 3A and 3B; and Table S2D and S2E). In addition, to evaluate the effect of treatment, we compared the expression profiles between both T_RM cell subsets across all patient groups using treatment (both OCS and biologics) as a covariate (Figure S3A; and Table S2F). We observed a positive correlation (R = 1, P < 0.0001) when comparing the results of DGEA between both T_RM subsets using sex or treatment as a covariate, with the vast majority of the genes being shared by both analyses (Figures S3B and S3C; and Table S2G), suggesting treatment is not a major confounding factor.

Figure 3. CD103⁺ T_RM subset displays qualitative features linked to TCR activation and cytotoxicity.

(A) Volcano plot shows false discovery rate (FDR) (-log₁₀ adjusted P-value) and log₂ (fold change) in expression for genes differentially expressed in CD103⁺ T_RM versus CD103^– T_RM clusters using sex as covariate. Dots are colored according to the mean of expression (log₂) and sized based on the difference in the percentage of cells expressing the given gene, both derived from the group in which the gene is upregulated. Gray dotted lines represent the statistical threshold values: log₂(fold change) > 0.25 and -log₁₀(FDR > 1.3 (adjusted P-value < 0.05). (B) Heatmap of row-wise z-score-normalized mean expression of 1001 differentially expressed genes between CD103⁺ T_RM and CD103^– T_RM clusters in male and female patients separately. Adjusted P-value < 0.05 and log₂ (fold change) > 0.25. (C) IPA shows top 10 pathways enriched for genes with increased expression in CD103⁺ T_RM cluster compared to CD103^– T_RM cluster. Numbers show matching genes from dataset and IPA gene lists (Table S3B). (D, G) Violin plots show normalized expression for example genes up-regulated in CD103⁺ T_RM cluster linked to TCR signaling (D) or cytotoxicity (G). Color code represents the fraction of cells expressing the indicated gene in each cluster. (E, H) GSEA shows enrichment of genes linked to TCR signaling (E) or cytotoxicity (H) in CD103⁺ T_RM cluster compared to CD103^– T_RM cluster. NES, normalized enrichment score; q, false discovery rate. (F, I) GSVA shows TCR signaling (F) or cytotoxicity (I) enrichment scores per donor in CD103⁺ T_RM and CD103^– T_RM clusters (***, P < 0.001 and ****, P < 0.0001, respectively; Mann-Whitney U test). (E, F, H, I) Gene lists in Table S3A. (J) Single-cell pseudotime trajectory analysis of airway CD4⁺ T cell subsets. Trajectory constructed using the Monocle3 algorithm. (K) Dot plots show Shannon-Wiener (top) and Inverse Simpson (bottom) TCR diversity indexes in bulk samples from CD103⁺ T_RM, CD103^– T_RM, and non-T_RM cells (**, P < 0.01; ***, P < 0.001; Kruskal-Wallis one-way test followed by Dunn’s post-hoc test). (F, I, K) Each dot represents data from a single patient, bars represent the mean, and t-lines represent SEM. (L) Bar chart shows number of TCR clonotypes specific (left) or shared (right) between bulk samples from CD103⁺ T_RM, CD103^– T_RM, and non-T_RM cells. Total number of clones in each sample group (left bottom corner) is shown.

Unbiased ingenuity pathway analysis (IPA) of transcripts with increased expression in the CD103⁺ T_RM subset showed significant enrichment for multiple T-cell associated biological pathways (Figure 3C; and Table S3B). The most enriched pathway was the T_H1 pathway (CXCR3, IFNG), confirming that the CD103⁺ T_RM subset is enriched for T_H1 features. Secondly, the integrin signaling pathway (ITGAE, ITGA1, ITGB2), which plays an important role in mediating interactions of CD103⁺ T_RM subset with epithelial cells and extracellular matrix. Finally, we observed significant enrichment of genes involved in T cell receptor (TCR) signaling, CD28 and ICOS co-stimulation pathways (NFATC2, CSK, LAT, LCP2, CD40LG, CD52, HLA-DRB1, HLA-DRB5, HLA-DRA, MIR155HG), and survival pathways (ERK/MAPK signaling), and confirmed this finding by gene set enrichment analysis (GSEA) and gene set variation analysis (GSVA) (P < 0.001) (Figures 3C-F; Figure S3D; and Tables S2H, S3A, and S3B). Notably, HLA-DR expression has been reported to mark recently activated T cells following in vivo antigen-specific TCR engagement^67,68, suggesting that the CD103⁺ T_RM subset may be enriched for T cells that were recently activated in the asthmatic airways (Figure 3D). Furthermore, transcripts encoding for several cytokines (IFN-γ, TNF, LIGHT/TNFSF14) and chemokines (CCL4/MIP-1β, CCL5/RANTES), known to be involved in promoting airway inflammation and remodeling,^69–71 showed significant increased expression in the CD103⁺ T_RM subset (Figure 3D). Notably, despite treatment with high-dose corticosteroid and/or biological agents, the sustained expression of transcripts linked to TCR activation and cytokines in the CD103⁺ T_RM subset (Figure 3D) suggests that treatment fails to curtail the activation and functional responses of airway CD103⁺ T_RM cells in severe asthma.

GSEA analysis as well as single-sample GSVA (P < 0.0001) showed significant positive enrichment of cytotoxicity signature genes in the CD103⁺ T_RM subset (GZMB, GZMA, GZMH, FASLG)⁶¹ (Figures 3A, 3C, 3G-I; Figure S3D; and Tables S2D and S2H), indicating that cytotoxic CD4⁺ T cells were enriched in the CD103⁺ T_RM subset. In addition, treatment with either OCS or biologics in severe asthmatics was associated with increased expression of CCL4 and GZMH transcripts in the CD103⁺ T_RM cluster (Figures S3E and S3F; and Tables S2I and S2J). The CD103⁺ T_RM subset also showed increased expression of transcripts encoding for two transcription factors linked to cytotoxic function in T cells: HOBIT (ZNF683), which is linked to T_RM differentiation and persistence of cytotoxic effector T cells,^50,72,73 and HOPX, known to regulate GZMB expression⁷⁴ and to increase in vivo persistence of T_H1 cells, by reducing sensitivity to FAS-mediated apoptosis of T cells^75,76 (Figure 3G). Together, these findings suggest that the CD103⁺ T_RM subset is enriched for cells with increased cytotoxicity and effector properties, potentially driving airway inflammation and remodeling in severe asthma. Of note, bulk transcriptomic analysis of sorted CD103⁺ T_RM cells confirmed the increased expression of transcripts encoding for cytotoxicity-associated molecules like Granzyme B and the transcription factor HOBIT (P < 0.0001) (Figures S3G and S3H; and Table S4A). Using GZMB gene expression as a phenotypic marker for cytotoxic CD4⁺ T cells, we confirmed that the frequency of GZMB-expressing CD4⁺ T cells was significantly increased in male patients with severe compared to mild asthma (mean frequency 27% versus 12%; P < 0.05) (Figure S3I; and Table S2K). Notably, we analyzed published single-cell RNA-seq datasets^46–49 of BAL CD4⁺ T cells from healthy controls and observed that the mean frequency of GZMB-expressing CD4⁺ T cells in healthy BAL was decreased (9%) in comparison to our data (Figure S3I; and Table S2K). We also assessed the levels of Granzyme B, Granzyme A, CCL3 and CCL4 in BAL supernatants from matched patients by multiplex ELISA. However, we did not observe significant differences between disease groups (Figure S3J; and Table S1G), likely due to fact that other immune cell types like CD8⁺ T cells and NK cells may also release these molecules.

Next, to determine potential relationship between the CD103⁺ T_RM and CD103^– T_RM subsets, we performed single-cell trajectory analysis of all airway CD4⁺ T cells (Figure 3J) and a specific analysis of only T_CM and T_RM subsets (Figure S3K), which pointed to a transitional path originating from the T_CM cluster towards the CD103⁺ T_RM cluster through the CD103^– T_RM cluster as an intermediate population (Figure 3J; and Figure S3K). Additionally, TCR β-chain sequencing analysis of bulk populations of CD103⁺ T_RM, CD103^– T_RM, and non-T_RM subsets showed that the clonal repertoire of CD103⁺ T_RM cells was significantly less diverse when compared to CD103^– T_RM and non-T_RM clonotypes (P < 0.01 and P < 0.001) (Figure 3K; and Table S4B). Importantly, we observed that a large number of clonotypes from the CD103⁺ T_RM subset were shared with the CD103^– T_RM subset, suggesting that CD103⁺ T_RM cells are clonally related to CD103^– T_RM cells (Figure 3L; Figure S3L; and Table S4C). Overall, our single-cell transcriptomic analysis indicates that CD103⁺ T_RM cells in the airways of asthmatic patients exhibit a T_H1-like phenotype with increased TCR activation, cytotoxicity, and pro-inflammatory effector properties.

Molecules that restrain T cell activation and effector functions are reduced in severe asthma

To determine qualitative differences in airway CD4⁺ T cell subsets that are associated with asthma severity in both males and females, we performed independent single-cell differential gene expression analysis (scDGEA) of CD4⁺ T cells from severe versus mild asthmatics for each sex (Figure 4A; and Table S2L) and per subset (Figure S4A; and Tables S2L and S2M). Overall, we observed that the most differentially expressed genes increased in severe asthma were significantly enriched in male patients, especially those linked to cytotoxicity (GZMB, GZMH, ZNF683, HOPX, CCL4) (Figure 4A; and Table S2L). Cluster-specific analysis showed that these cytotoxicity genes were differentially expressed only in the CD103⁺ T_RM subset (Figure S4A; and Table S2L). While overall GSEA for cytotoxic signature genes showed significant enrichment in severe asthmatics, especially males, these observations were not confirmed at individual level when performing GSVA, likely due to a limited sample size and heterogeneity among the study participants (Figure S4B; and Table S2H). This finding suggests that although the CD103⁺ T_RM subset was enriched in genes linked to TCR signaling and cytotoxicity, these features were not significantly different when comparing the CD103⁺ T_RM subset from either males versus females or severe asthmatics versus mild asthmatics (Figures S4B and S4C; and Table S2H).

Figure 4. Molecules that restrain T cell activation and effector functions are reduced in severe asthma.

(A) Crater plot shows the log₂ (fold change) expression of genes between severe and mild asthma in males (x-axis) and females (y-axis). Dotted lines indicate the statistical threshold value of fold change for gene filtering (adjusted P-value < 0.05 and log₂ (fold change) > 0.25). (B) Plot shows row-wise z-score-normalized mean expression (color scale) and percent of expressing cells (size scale) for indicated genes in each cluster per disease. (C) GSEA plot shows enrichment of genes linked to cAMP immunoregulation pathway in cells from severe compared to mild asthmatics, in males (left) and females (right). NES, normalized enrichment score; q, false discovery rate. (D, E) GSVA shows cAMP signaling enrichment scores per donor grouped by disease per cluster (D) or per donor in CD103^– T_RM (left) and CD103⁺ T_RM (right) clusters for the 4 clinical groups (E) (*, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; Mann-Whitney U test). (C, D, E) Gene lists in Table S3A. (F) Violin plots show normalized expression for genes down-regulated in severe asthma in the T_FH (top) and T_REG (bottom) clusters. Color code represents the fraction of cells expressing the indicated gene in each cluster.

Interestingly, in both males and females with severe asthma, most CD4⁺ T cell subsets including the T_RM subsets showed the highest significant reduction in the expression of several transcripts encoding for molecules known to dampen TCR signaling and effector functions in T cells (CREM, DUSP1, DUSP2, DUSP4, TNFAIP3)^77–81 (Figure 4B). One of the most downregulated differentially expressed genes is CREM (cyclic AMP responsive element modulatory), encoding for a transcription factor known to repress promoters of inflammatory cytokine genes like IL2, IL13, IL4 and to dampen type 2 inflammation in murine asthma models.⁷⁹ Other transcripts with reduced expression in severe asthma encode for several members of the dual specificity phosphatase (DUSP) family proteins: DUSP1, a glucocorticoid responsive molecule that is known to inhibit activity of mitogen-activated protein kinases (MAPKs) that trigger cytokine production in immune cells;^77,82–84 DUSP2, which catalyzes dephosphorylation of STAT3 and inhibits T_H17 differentiation and inflammation;^77,85 and DUSP4, which dephosphorylates STAT5 and negatively regulates IL-2 signaling and T cell proliferation.^77,86 Lastly, TNFAIP3 is known to negatively regulate NFkB signaling, which is involved in T cell activation and effector function.^87,88 GSEA as well as single-sample GSVA confirmed negative enrichment for the immunoregulatory cAMP signaling pathway⁸⁹ in multiple CD4⁺ T cell subsets from both male and female patients with severe asthma (Figure 4C-E; and Tables S2H and S3A).

Finally, one of the most significantly increased transcripts in many clusters in severe asthmatics, in both males and females, was FKBP5, a steroid-responsive gene, which encodes for the FK506 binding protein 5 (FKBP5)^90–92 (Figure 4B). As expected, this gene was significantly increased in the CD103⁺ T_RM subset from severe asthmatics on OCS and biologics (Figures S3E and S3F). In severe asthmatics, cells in the T_FH cluster showed reduced expression of transcripts encoding for molecules linked to inhibitory function like PD-1, TIM-3, LAG-3, and TIGIT, which suggested the potential for unrestrained activity of airway T_FH cells in severe asthma⁹³ (Figure 4F). T_REG cells from severe asthmatics had significantly reduced levels of transcripts encoding for AP-1 family of transcription factors (JUNB, JUN, FOS, FOSB) that have been shown to be important for its suppressive functions^94,95 (Figure 4F). Overall, our analysis highlights the molecular properties of airway CD4⁺ T cell subsets that can potentially trigger their unrestrained activation as well as confer resistance to corticosteroids in severe asthma.

Non-T_H2 pro-inflammatory cytokines are expressed by airway CD4⁺ T cells from severe asthmatics

To examine the effector potential of airway CD4⁺ T cells from patients with mild and severe asthma, we stimulated, ex vivo, a fraction of BAL cells with phorbol 12-myristate 13-acetate (PMA) and Ionomycin for 2 hours, and performed single-cell RNA-sequencing on over 35,000 sorted CD4⁺ T cells (Figure S5A; and Table S2B). scDGEA between resting and stimulated CD4⁺ T cells (Table S2N), as well as between stimulated cells from mild and severe asthmatic patients for each sex (Figure 5A; and Tables S2O and S2P), revealed several hundred transcripts linked to T cell effector function (Tables S2N and S3C). Transcripts linked to T_RM markers (ITGAE, ITGA1) (Figure 5B) were downregulated following stimulation, which prevented reliable identification of the T_RM subsets in stimulated datasets (Figure 5B). As expected, transcripts encoding for chemokines known to be released by cytotoxic CD4⁺ T cells like CCL3, CCL4, and CCL5 ^61,96–98 showed increased expression in stimulated airway CD4⁺ T cells from patients with severe asthma compared to mild asthma (Figures 5A and 5B). These chemokines play key roles in the recruitment of several immune cell types expressing C-C chemokine receptor 1(CCR1), CCR3, CCR5, like neutrophils, monocytes, macrophages, NK cells, and T cell subsets,⁹⁹ which have the potential to drive airway inflammation and remodeling.^100,101

Figure 5. Pro-inflammatory cytokines are expressed by airway CD4⁺ T cells from severe asthmatics.

(A) Crater plot shows the log₂ (fold change) expression of genes between severe and mild asthma in males (x-axis) and females (y-axis). Dotted lines indicate the statistical threshold value of fold change for gene filtering (adjusted P-value < 0.05 and log₂ (fold change) > 0.25). (B) Plot shows row-wise z-score-normalized mean expression (color scale) and percent of expressing cells (size scale) for indicated genes in resting and stimulation conditions per disease. (C) Volcano plot shows false discovery rate (FDR) (-log₁₀ adjusted P-value) and log₂ (fold change) in expression for genes differentially expressed in GZMB⁺ cells versus GZMB^– cells. Dots are colored according to the mean of expression (log₂) and sized based on the difference in the percentage of cells expressing the given gene, both derived from the group in which the gene is up-regulated. Gray dotted lines represent the statistical threshold values: log₂(fold change) > 0.25 and -log₁₀(FDR > 1.3 (adjusted P-value < 0.05). (D) Plot shows row-wise z-score-normalized mean expression (color scale) and percent of expressing cells (size scale) for indicated genes in all GZMB^–, mild asthma (MA) GZMB⁺, and severe asthma (SA) GZMB⁺ cells. (E) Scatter plots show co-expression of CCL3 with other cytokine genes transcripts in GZMB-expressing CD4⁺ T cells in stimulation condition. Percentage of co-expressing cells is indicated (top right). Each dot represents one cell. Cells are colored based on density value. Dotted lines indicate threshold of Seurat normalized gene expression (log2FC) (> 0.5).

Other transcripts upregulated by stimulation encode for molecules associated with T_H1 effector (IFNG, TNF, FASLG, XCL1, XCL2), and pro-fibrotic (LIGHT, AREG, TGFB1) properties (Figures 5A and 5B; and Table S2N). Although T_H2 and T_H17 cytokine transcripts were observed in stimulated airway CD4⁺ T cells, only a relatively small fraction of cells was expressing IL4, IL5, IL13, and IL17A transcripts (3%, 2%, 16%, and 8%, respectively) and they were significantly reduced in patients with severe asthma compared to mild asthma (Figure 5B; and Table S2O). However, several other pro-inflammatory non-T_H2 cytokine transcripts were expressed by a larger fraction of airway CD4⁺ T cells (TNF (80%), CSF2 (40%), CCL20 (50%) and IL21 (23%)) (Figure 5B; and Tables S2N and S2O). Together, these data from stimulated cells suggests that the effector potential of airway CD4⁺ T cells is not fully curtailed by high-dose corticosteroid treatment and that non-T_H2 cytokines may contribute to the pathogenesis of severe asthma.

To further explore the effector molecules expressed by cytotoxic CD4⁺ T cells present in the airways, we examined the co-expression of cytokine transcripts specifically in GZMB-expressing cells, as GZMB is a canonical cytotoxicity-associated marker gene (Figures S5B and S5C; and Tables S2N, S2O, and S2Q). scDGEA between GZMB⁺ versus GZMB^– cells revealed over 1000 differentially expressed genes, among which we found transcripts encoding for cytotoxic molecules (GZMA, GZMH), and several pro-inflammatory cytokines and chemokines (CCL3, CCL4, CCL5, CCL20, TNF, IFN-γ, CSF2, IL21, LIGHT, TGFB1), known to contribute to airway inflammation, fibrosis, and remodeling; in part through regulating activity of fibroblasts and smooth muscle cells^{50,70,71,102–111} (Figures 5C-E; Figure S5D; and Tables S2Q, S2R and S2S). These data suggest that cytotoxic CD4⁺ T_RM cells, besides their potential for direct killing of target cells, can also express pro-inflammatory molecules, which may play an important role in sustaining airway inflammation and remodeling. To address the technical limitation in accurately classifying T_RM cells in stimulated CD4⁺ T cells and to confirm their effector potential, we isolated specific CD4⁺ T_RM cell populations from ex vivo stimulated BAL cells by flow cytometry and examined their bulk RNA-seq profile. As expected, stimulated cells from the CD103⁺ T_RM subset expressed high levels of transcripts encoding for cytotoxicity-associated molecules (GZMB, GZMA, GZMH), pro-inflammatory chemokines (CCL3, CCL4, CCL5), and cytokines (TNF, IFN-γ, CSF-2, IL-21, IL-17A, IL-23A, IL-2, IL-13, LIGHT) (Figures S5E and S5F; and Tables S4D and S4E). Overall, these findings supported the existence of a population of CD4⁺ T_RM cells with features of cytotoxicity and polyfunctionality in BAL samples collected from severe asthmatics.

DISCUSSION

Here, we report on the single-cell transcriptomes from purified CD4⁺ T cells isolated from the airways of patients with severe and mild asthma. This unbiased approach led to the discovery of a cytotoxic CD4⁺ T_RM cell subset in severe asthma that we hypothesize is critical in driving airway inflammation and remodeling in a specific subgroup i.e., males with severe asthma, where we found a striking increase in the proportions of a CD4⁺ T_RM subset (CD103⁺ T_RM cells) with cytotoxic properties in the airways.

Asthma is now acknowledged to be a heterogeneous state with numerous clinical phenotypes that show considerable diversity. The current prevailing paradigm is to stratify asthma endotypes as either type 2 (T2)-high or T2-low.¹¹² However, recent findings from the analysis of clinical and pathophysiological characteristics between patients classified as T2-high and T2-low from the WATCH study demonstrated that when properly characterized, asthma is an overwhelmingly T2 state (93%),¹⁴ supporting that the T2 paradigm by itself cannot explain the diversity of asthma at a pathophysiological level, instead other pathways driven by for example cytotoxic CD103⁺CD4⁺ T_RM cells may define endotypes underlying diverse phenotypes of severe asthma.^14,113 For instance, the WATCH cohort study defined distinct clinical phenotypes of severe asthma stratified by age of asthma onset and sex.⁶² These findings confirmed a previously poorly recognized phenotype of severe asthma, adult-onset male severe asthmatics, characterized by numerous adverse clinical features including worse lung function despite short disease duration, higher peripheral blood eosinophilia, higher exhaled nitric oxide, greater oral corticosteroid dependency, and little of the typical psychophysiological comorbidities seen in difficult-to-treat asthma.⁶² This clinical phenotype has been previously shown in other studies though attracted little focus hitherto.^114–117

In the present study, we identified potential endotypic features, i.e. cytotoxic CD103⁺CD4⁺ T_RM cells, that underpin such clinical phenotype. The CD103⁺CD4⁺ T_RM cells also displayed a unique pro-inflammatory cytokine and chemokine signature, highly enriched in transcripts encoding molecules (e.g. Granzymes, CCL3, CCL4, LIGHT, TNF, IL-21) that drive inflammation, cell death, and fibrosis.^{71,108,118,119} Because T_RM cells are a long-term resident population in the airways, they have the potential for sustained interaction with airway structural cells and thus the products they release are likely to promote persistent airway inflammation and remodeling in severe asthma.

While T_H2 cells and, to a lesser extent, T_H17 and T_H1 cells have been implicated in asthma pathogenesis and therapies targeting T_H2 cytokines are beneficial for some patients with asthma, the role of cytotoxic CD4⁺ T cells in severe asthma pathogenesis has not been previously described. Cytotoxic CD4⁺ T cell responses have been reported in certain viral infections such as human cytomegalovirus, human immunodeficiency virus, dengue virus, hepatitis C virus, influenza virus, and, more recently, with SARS-CoV2 virus.^96,120–130 Outside the context of viral infections, reports from studies employing single-cell genomics have shown an increased number of cytotoxic CD4⁺ T cells in patients with autoimmune diseases such as rheumatoid arthritis¹³¹ and multiple sclerosis.¹³² These cells expressed high levels of pro-inflammatory cytokines and are hypothesized to drive disease pathogenesis. Most importantly, an increased number of cytotoxic CD4⁺ T cells has been observed in several steroid-resistant diseases with pronounced organ fibrosis such as systemic sclerosis,¹⁰⁰ idiopathic pulmonary fibrosis,¹³³ IgG4 disease,¹⁰¹ and graft versus host disease,¹³⁴ suggesting that cytokines and other currently uncharacterized factors released by cytotoxic CD4⁺ T_RM cells are likely to play a significant role in promoting clinical hallmarks of severe asthma such as persistent inflammation, fibrosis, and airway remodeling. Based on previous research showing that sustained T cell activation through TCR engagement drives the differentiation of CD4⁺ T cells into CD4 cytotoxic T lymphocytes (CD4 CTLs),¹³⁵ we hypothesize that, in the airways of severe asthmatics, CD103⁺CD4⁺ T_RM cells with cytotoxic features are likely to represent cells that received chronic TCR stimulation.

Furthermore, our findings support that airway CD4⁺ T_RM cells display a T_H1-like pro-inflammatory potential namely by secreting CCL3, CCL4, and CCL5, pro-inflammatory chemokines that are also linked to viral airway infections and virus-triggered asthma exacerbations.^136,137 Their cognitive receptor, mainly CCR5, has been recently associated with CD4⁺ T_RM cells and associated with worsening of lung function in asthmatic patients.²⁹ CCR5 inhibition in a murine model of type-1 allergic lung inflammation showed a drastic reduction of airway hyperresponsiveness and inflammation.^138,139 Based on this finding, we speculate that, in severe asthma, especially males, the sustained release of CCL3, CCL4, and CCL5 by activated CD4⁺ T_RM cells triggers a pro-inflammatory milieu that can worsen airway inflammation and remodeling in severe asthma.

In summary, our findings support a potentially important role for cytotoxic CD4⁺ T_RM cells in driving disease pathogenesis and, more specifically, in a subgroup of severe asthmatics, and thus may represent an attractive target for therapeutic development. Our work also potentially expands the portfolio of treatable traits that have gained increasing attention in severe asthma management,¹⁴⁰ with the intriguing concept that patient sex and airway CD4⁺ T cell profile may help guide disease assessment, prognosis and treatment.

Limitations of the study

Future studies are required to address the nature of antigens recognized by cytotoxic CD4⁺ T_RM cells in severe asthma, as it is unclear if these antigens are allergens, pathogens, or potential autoantigens. Determining the specificity of cytotoxic CD4⁺ T_RM is likely to provide insights into mechanisms that induce and sustain their generation and maintenance in the airways of severe asthmatics. Furthermore, studies in model organisms are required to functionally establish whether cytotoxic CD4⁺ T cells and the pro-inflammatory molecules they produce are capable of initiating and maintaining persistent airway inflammation and remodeling.

In addition, as severe asthma is a heterogeneous disease, it will be essential to confirm the association between cytotoxic CD4⁺ T_RM cells and specific phenotypes of severe asthma in large-scale studies of airway CD4⁺ T cells from well-characterized cohorts of patients with severe and mild asthma across different age groups, ethnicities, and diverse geographical locations. Longitudinal studies are required to define whether severe asthmatics with high frequency of cytotoxic CD103⁺CD4⁺ T_RM cells are associated with more adverse long-term outcomes. In particular, future studies should assess associations of CD103⁺CD4⁺ T_RM cells with biologics treatment outcomes to establish their potential as biomarkers of biologics responsiveness. Such studies should also monitor the frequency of CD103⁺CD4⁺ T_RM cells in the airways before and after biologics treatment to determine if their frequency can be modulated by currently available biologics therapies.

Finally, as sex hormones fluctuate across an individuals’ lifespan with important consequences on severe asthma pathogenesis as well as on treatment efficacy^141,142, a larger study that examines the influence of sex and sex hormones on the immune landscape of airway T cells in severe asthmatics is required.

STAR METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by Lead Contact, Grégory Seumois (gregory@lji.org).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• Sequencing data are available from NCBI as Gene Expression Omnibus, SuperSeries accession number GSE181711 (Subseries: GSE181709 for bulk RNA-seq data and GSE181710 for single-cell RNA-seq data).

• Scripts are available in our repository on GitHub (https://github.com/vijaybioinfo/ASTHMA_AIRWAYS_2021).

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Patient recruitment, ethical approval, and characteristics

Study participants diagnosed with asthma were recruited into the National Institutes of Health Epigenetics of Severe Asthma study (n=193) from established cohorts of patients (United Kingdom (UK)): the Wessex AsThma CoHort of difficult asthma (WATCH) at University Hospital Southampton Foundation Trust UK⁴¹ (n=501) consisting of patients with severe/difficult-to treat asthma (GINA management steps 4 and 5⁴³) and the Isle of Wight Whole Population Birth Cohort (IOWBC) at the David Hide Asthma and Allergy Research Centre, Isle of Wight, UK¹⁴³ (n=1,456), along with clinic/community recruitment on the Isle of Wight, UK, of mild asthma patients (GINA management steps 1 and 2, with a small proportion with step 3, see Table S1A). Adherence to treatment in these patients was clinically assessed using a Medicines Possession Ratio (MPR) threshold of acceptable. This study received approval from the South Central Hampshire B – Southampton Research Ethics Committee, UK (REC reference: 18/SC/0105) and from the La Jolla Institute for Immunology Institutional Review Board (IRB VD-156–1118, La Jolla Institute for Immunology, La Jolla, USA). Written informed consent was obtained from all study participants. Thirty patients were included in this study based on the identification of clinically severe airways disease and on the safety criteria for bronchoscopy being satisfied. Briefly, study participants underwent an extensive clinical characterization process including detailed clinical, health and disease-related questionnaires, anthropometry, and lung function testing (Table S1A). Patients with mild (GINA 1 to 3, n=14) and severe asthma (GINA 4 and 5, n=16) underwent fiberoptic bronchoscopy for collection of BAL samples. Mild asthmatic patients were treated with inhaled bronchodilators alone (Salbutamol 200 μg as required) (n=4) and/or with low to medium dose of inhaled corticosteroids (400 to 800 μg/day beclomethasone dipropionate (BDP) equivalent, n=10). Conversely, all severe asthmatic patients were treated with high dose inhaled corticosteroids (1,200 to 2,000 μg/day BDP equivalent) and second controller medication (n=14), and/or on daily maintenance oral steroids (n=5) and/or biological monoclonal antibody treatment (Omalizumab (anti-IgE), n=2; Mepolizumab (anti-IL-5), n=7) (details provided in Table S1A).

METHOD DETAILS

BAL samples processing

BAL fluid was obtained by instilling a total volume of 120 ml of warm 0.9% saline in small aliquots (initially 40 ml followed by 20 ml, holding for 10 seconds each time) into the right upper lobe segments using a fiberoptic bronchoscope procedure (n = 30). Aliquots were pooled together and collected as 1 sample with immediate storage on ice. The median recovery of BAL volume was 53 ml (Inter-quartile range: 24–60 ml). RNAse inhibitor (v:v 1:1000, Takara Bio) and protease inhibitor (v:v 1:50; Sigma Aldrich) were immediately added to BAL collected. BAL was then filtered within 30 minutes with a 100 μm BD cell strainer and centrifugated at 300 x g for 10 minutes at 4°C. Cellular fractions were resuspended in 1 ml of phosphate buffer solution (PBS) with RNAse inhibitor (v:v 1:100). Two Cytospin slides were generated with 70 μl of cell suspension using a Shandon Cytospin 2 and stained using rapid Romanowsky (Diff-Quick) stain to obtain differential cell counts and to ascertain the volume of squamous cell contamination.⁴¹ Samples were centrifugated once more at 300 x g for 10 minutes at 4˚C. Supernatants were discarded and cell pellets resuspended into freezing media (50% human decomplemented AB Serum (Sigma Aldrich), 40% complete Gibco Roswell Park Memorial Institute (RPMI) medium (ThermoFisher Scientific) complemented with 10% heat-inactivated fetal bovine serum (FBS) (Sigma Aldrich, 10% DMSO (Sigma Aldrich), and 5 μl of RNAse inhibitor, before slow cryopreserving at −80°C as described previously.⁵⁸ All BAL samples were collected and cryopreserved in the UK, stored in liquid nitrogen tanks, and sent later to La Jolla Institute for Immunology for processing.

Flow cytometry of cryopreserved BAL cellular samples

Cryopreserved samples were thawed, and cells were transferred first to 1 ml of cold heat-inactivated FBS and then quickly diluted up to 10 ml with complete TCM medium (Gibco Iscove’s Modified Dulbecco’s Medium (IMDM) with 5% FBS and 2% human serum; ThermoFisher Scientific). Samples were centrifugated at 250 x g for 5 minutes at room temperature. Supernatants were discarded and cells resuspended in appropriate volume of MACS buffer (PBS, 2 mM EDTA, 2% heat-inactivated FBS) to reach 2 million cells per ml. Around 25% of the sample or maximum of 500,000 cells were separated for stimulation assays. Remaining cells were centrifuged at 400 x g for 5 minutes at room temperature, supernatant was discarded, and cells resuspended in 200 μl of MACS buffer complemented with 1% RNAse inhibitor. All samples (resting or after stimulation) were stained following a standard procedure previously described.⁵⁸ Briefly, 200 μl cell suspensions were first incubated with 20 μl of FcgR blocking solution (Miltenyi Biotec) for 15 minutes on ice, and subsequently stained with the following combination of fluorescently-conjugated antibodies: anti-CD45-Alexa Fluor 700 (2D1; BioLegend), anti-CD3-APC-Cy7 (SK7; BioLegend), anti-CD8a-BV570 (RPA-T8; BioLegend), anti-CD4-BV510 (RPA-T4; BioLegend), anti-CD357(GITR)-BV711 (108–17; BioLegend), anti-CD185(CXCR5)-BV421 (RF8B2; BD Biosciences), anti-CD25-BB515 (2A3; BD Biosciences), anti-CD127-APC (eBioRDR5; eBioscience), anti-CD69-BV605 (FN50; BioLegend) and anti-CD103-PE-Cy7 (Ber-ACT8; BioLegend). The Brilliant Stain Buffer Plus (BD Biosciences) was also added to the antibody mix as recommended. For a fraction of the samples, the DNA-oligonucleotide conjugated anti-CD103 (TotalSeq-A0145; Ber-ACT8; BioLegend) and anti-CD69 antibodies (TotalSeq-A0146; FN50; BioLegend) were also added. After 20 minutes incubation in the dark, on ice, cells were washed once with 5 mL of ice-cold MACS buffer, centrifugated at 400 x g for 5 minutes at room temperature (RT), resuspended in 250 μl MACS buffer with RNAse inhibitor (10%), and brought to flow cytometry for immunophenotyping analyses and sorting. Live and dead cells were discriminated using propidium iodide (PI, 1:200 vol:vol). All stained samples were analyzed using BD FACSAria Fusion Cell Sorter (BD Biosciences) and FlowJo software (v10.7.1).

Stimulation assays

Maximum 25% of BAL samples and no more than 500,000 cells from each BAL sample were stimulated ex vivo in 1 ml of complete TCM medium complemented with PMA (final 20 nM, phorbol-12-myristate-13-acetate) and ionomycin (final 1 μM; Sigma Aldrich) for 2 hours in a cell culture incubator at 37˚C and 5% CO₂. Samples were then processed and stained for flow-cytometry analysis and sorting as described here above.

Cell isolation for bulk and single-cell RNA-seq assay

For bulk RNA-seq assays, cells of interest were directly collected by sorting 400 cells into 0.2 ml PCR tubes (low-retention, Axygen) containing 8 μl of ice-cold lysis buffer (Triton X-100 [0.1%, Sigma-Aldrich] and 1% RNase inhibitor [Takara Bio]). Once collected, tubes were vortexed for 10 seconds, spun for 1 minute at 3000 x g and directly stored at −80˚C. For single-cell RNA-seq assays (10x Genomics), 1,000 to 2,000 airway CD4⁺ T cells were sorted per BAL sample directly in low retention and sterile ice-cold 1.5 ml collection tubes containing 500 μl of PBS:FBS (1:1 vol:vol) with RNAse inhibitor (1:100). Samples were batched in groups of 5 to 6 donors with similar disease status. Samples were also separated based on stimulation. In total, we performed 6 sorting experiments (see Table S1C). Collection tubes with ~10,000 to 20,000 sorted CD4⁺ T cells were inverted a few times, ice-cold PBS was added to reach a volume of 1,400 μl, and tubes were centrifuged for 5 minutes at 600 x g and 4˚C. Supernatant was removed with caution, leaving a volume of around 10 μl. Pellets were then resuspended with 35 μl of 10X Genomics resuspension buffer (0.22 μm filtered ice-cold PBS supplemented with ultra-pure bovine serum albumin (0.04%, Sigma-Aldrich). 40 μl of cell suspension were transferred to an 8 PCR-tube strip for downstream steps as per manufacturer’s instructions (10x Genomics).

Bulk RNA library preparation for sequencing

For full-length bulk transcriptome analyses, we used the Smart-seq2 protocol (adapted for samples with small cell numbers).^144–146 Briefly, RNA was captured using oligo-poly(dT)-3’ primers and reverse transcription was performed using 5’-template switching oligos (LNA technologies, Exicon). cDNA was pre-amplified by PCR cycle for 20 cycles. Amplified cDNA was cleaned by applying a double size purification (0.6 vol:vol and 0.8 vol:vol with Ampure-XP magnetic beads (Beckman Coulter)). After quantification and quality assessment using capillary electrophoresis (Fragment analyzer, Advance analytical), 0.5 ng of pre-amplified cDNA was used to generate indexed Illumina libraries (Nextera XT library preparation kit, Illumina). Every sample was quality checked for fragment size by capillary electrophoresis (Fragment analyzer, Advance analytical) and quantified (Picogreen, Thermofisher). No libraries failed our quality control check steps and therefore were pooled at equal molar concentration before loading on the NovaSeq 6000 Illumina sequencing platform. Every library was sequenced to reach a minimum of 15 million 100 × 100 bp pair-ended sequencing reads (S4 flowcell 200 cycle v1.0, Xp workflow; Illumina).

10x Genomics single-cell RNA library preparation for sequencing

Samples were processed using 10X Genomics 3v3.0 single cell gene expression profiling chemistry as per manufacturer’s recommendations; after droplet generation, and in-droplet based reverse transcription, cDNA was amplified by PCR for 11 cycles and gene expression library preparation followed. After quantification, equal molar concentration of each library was pooled and sequenced using the NovaSeq6000 Illumina sequencing platform to obtain 28- and 100-bp paired-end reads using the following read length: read 1, 100 cycles; read 2, 100 cycles; i7 index, 8 cycles and i5 index 8 cycles. Each library was sequenced aiming at a minimum mean sequencing depth of 87,000 reads per cell. For samples stained with DNA-oligo-conjugated-cell-surface antibodies (TotalSeq-A, Biolegend), amplified DNA generated from antibody-DNA oligos was separated from transcriptomic cDNA based on size-selection following amplification. Antibody-DNA amplified fragments are less than 300 bp. Library preparation was followed in accordance with the manufacturer’s recommendations. TotalSeq libraries were quantified and sequenced in the same manner as the gene expression libraries, as described above. Each library was sequenced aiming for 5,000 reads per cell.

Genotyping

For each patient, genomic DNA was isolated from PBMC using the DNeasy Blood and Tissue Kit (Qiagen) and utilized for genotyping using the Infinium Multi-Ethnic Global-8 Kit (Illumina) following the manufacturer’s instructions. Chip-arrays were run on an Illumina iScan System using the University of California - San Diego, Institute of Genomic Medicine. Raw data from the genotyping analysis, data quality assessment and SNPs identification were performed as previously described.⁵⁸

Bulk RNA-seq analysis

Bulk RNA-seq data were mapped against the hg19 genome reference using our in-house pipeline (https://github.com/ndu-UCSD/LJI_RNA_SEQ_PIPELINE_V2). Briefly, FASTQ data from sequencing was merged and filtered using fastp (v0.20.1), reads were aligned with the STAR aligner (v2.7.3a), followed by further processing with samtools (v0.1.19–44428cd), bamCoverage (v3.3.1), and Qualimap (v.2.2.2-dev). Raw and transcripts per million reads (TPM) counts were taken from STAR’s BAM aligned output.

To identify genes expressed differentially between groups, we performed negative binomial tests for paired comparisons by employing DESeq2¹⁴⁷ (v1.16.1) with default parameters and batch and sex as a covariate. We considered genes to be expressed differentially by any comparison when the DESeq2 analysis resulted in a Benjamini-Hochberg–adjusted P-value of less than 0.01 and a log₂ fold change of at least 1.

TCR-seq analysis

To profile the bulk data’s TCR repertoire, FASTQ files per bulk RNA-seq libraries were fed to the MiXCR algorithm¹⁴⁸ (v2.1.10, RepSeq.IO v1.2.11, MiLib v1.8.3). The TCR sharing was determined using the β chain from MiXCR’s clonalSequence per donor when comparing cell types. This TCR sharing was displayed using the ComplexUpset package (v1.3.3).

TCR diversity indexes calculation.

Inverse Simpson and Shannon-Wiener diversity indexes were calculated with CalcDiversityStats module of vdjtools software (v1.2.1, default settings)¹⁴⁹ taking TCR β amino acid and nucleotide sequences, TRB clone counts, and VDJ gene information per donor of sorted T_RM bulk TCR samples previously analyzed with MiXCR software (Table S4B).

Integration analysis of published BAL healthy datasets

To determine the proportions of CD4⁺CD69⁺CD103⁺ T_RM and CD4⁺CD69⁺CD103^– T_RM cells by flow cytometry in BAL healthy donors (Figure 2H), we used published data from Diniz, et al.,⁴⁴ Tang, et al.,⁴⁵ and Camiolo, et al. (GSE136587)²⁹. We retrieved 29 healthy donors from these published datasets.

To determine the frequency of GZMB-expressing CD4⁺ T cells in single-cell RNA-seq BAL healthy datasets (Figure S3I), healthy donors were directly collected from Liao, et al. (GSE145926);⁴⁶ Grant, et al. (GSE155249);⁴⁷ Morse, et al. (GSE128033);⁴⁸ and Mould, et al. (GSE151928)⁴⁹ datasets. Author’s labels were used to analyze CD4⁺ cells from these datasets, except Morse, et al. for which we used CD3D⁺ cells. We retrieved 17 healthy donors from these published datasets.

Single-cell RNA-Seq analysis

Analysis of 3’ transcriptome of single-cell from 10x Genomics platform.

Raw data was processed as previously described,^22,33,61 merging multiple sequencing runs using Cell Ranger’s count function (Table S2B), then aggregating multiple cell types with aggr (v3.1.0).¹⁵⁰

Doublet cell filtering and donor labeling.

Barcoded single-cell RNA-seq was demultiplexed patient-wise using Demuxlet¹⁵¹ with the following parameters: alpha=0, 0.5 and --geno-error=0.05. Each cell was assigned a donor ID or marked as a doublet. Cells called as doublet by Demuxlet were removed from downstream analyses. We did not observe major changes in singlets/doublets proportions between the different 10x Genomics libraries, suggesting optimal processing of cells during 10x (Gel Bead-In emulsions) GEM generation and downstream steps. All downstream analyses were performed using cells labelled as singlets.

CD4⁺ cells selection from stimulated CD3⁺ library.

Differential gene expression analysis was performed between filtered CD4⁺ CD8B⁻ and CD4⁻ CD8B⁺ cells from the CD3⁺ library, and differentially expressed genes were used for clustering. Genes were selected with a Benjamini-Hochberg adjusted P-value less than 0.05, log₂ |Fold Change| higher than 2 and sex-related genes were excluded (RPS4Y1, XIST, SPRY1, DDX3Y). Two clusters were identified correlating with the CD4 and CD8 T cell types from which only the CD4⁺ was taken for downstream analysis of stimulated data.

Transcriptome-based clustering analysis.

The merged data was transferred to the R statistical environment for analysis. Unbiased clustering analysis was performed using Seurat (v3.0.2).¹⁵² A first round of analysis was run and single-cell transcriptomes not meeting quality control thresholds (see below) as well as a cluster of contaminating cells characterized by a strong monocyte/macrophage signature were eliminated from the second round of analysis. Only cells expressing between 200 and 6,000 genes, less than 30,000 total unique molecule identifier (UMI) content, and less than 15% of reads mapping to mitochondria genome were included. Only genes expressed in at least 0.1% of the cells were included in the analysis. Expression counts were then log-normalized and scaled (by a factor of 10,000) per cell. Variable genes were detected with the VST method and the top highly expressed (UMI mean greater than 0.01) genes representing 15% of the cumulative variance were selected for cluster analysis. Transcriptomic data from each cell was then further scaled by regressing the number of UMI-detected and the percentage of mitochondrial reads. Principal component analysis (PCA) was then run on the variable genes, and the first 20 principal components were selected for downstream analyses based on the standard deviation of PCs (“elbow plot”) (Figure S1C). Cells were clustered using Seurat’s functions FindNeighbors and FindClusters with a resolution of 0.4. We performed downstream analyses excluding a cluster (T_APOPTOSIS) (< 2% of airway CD4⁺ T cells) enriched in apoptosis signature genes (as reported by GSEA) and interpreted as a technical artefact (Figure S1F).

Gene-set score calculation and gene set enrichment analysis (GSEA).

Signature scores were calculated with AddModuleScore function from Seurat with default settings. The score is derived from the mean of the gene list after subtracting a background expression calculated from a random list of genes (same size as the gene set). The normalized GSEA enrichment score was calculated using fgsea (v 1.10.1) in R with the signal-to-noise ratio as a metric.¹⁵³ Default parameters were used except minSize = 3 and maxSize = 500. Gene set lists are in Table S3A.

Gene-set variation analysis (GSVA).

To determine enrichment scores for predefined lists of genes in individual samples, we performed GSVA.¹⁵⁴ Single-cell GSVA was performed using the scgsva function with default parameters, utilizing the scGSVA package (v0.0.14, available at https://github.com/guokai8/scGSVA). To calculate the enrichment scores per donor for a specific cluster of cells, i.e. CD103⁺ T_RM cells, the arithmetic mean of all single-cell enrichment scores was computed.

Pathway analysis.

For unbiased pathway enrichment analysis of differentially expressed genes as well as network and upstream regulator analysis shown in Figure 3C, Figure S3D, and Table S3B, we used the IPA software (IPA, QIAGEN Redwood City, www.qiagen.com/ingenuity). Relevant pathways were selected using a filter based on significance index (P value [-log10]) > 2 and ordered based on z score. For the figure, any multiple pathways with similar enriched genes were summarized into one representative pathway labelled based on gene functions. For Table S3C, we used the clusterProfiler package¹⁵⁵ (v4.6.2) in R and performed enrichment analysis of gene ontology (GO) categories using the enrichGO function taking as “genes” highly differentially expressed genes (log2FC >= 2, padj < 0.05, stimulation versus resting condition comparison) and all differentially expressed genes (log2FC >= 0.25, padj < 0.05, stimulation versus resting condition comparison) as background (Table S3C). Default parameters were used except ont = “MF”.

Protein expression analysis using DNA-oligo-antibody single-cell sequencing.

TotalSeq-A reads were analyzed based on recommendation provided by manufacturer (BioLegend).

Single-cell differential gene expression analysis.

Pairwise analyses were performed using the R package MAST (v 1.8.2)¹⁵⁶ with cellular detection rate (CDR) as a covariate, after normalizing the data to log₂ counts per million (log₂(CPM+1)). A gene was considered as differentially expressed if its Benjamini-Hochberg adjusted P-value was < 0.05 and log₂ (|fold change|) was >0.25 (otherwise noted in legends). Cluster specific markers were determined by MAST using the Seurat function FindAllMarkers with default parameters. Violin plots represent the distribution of expression (based on a Gaussian Kernel density estimation model) of cells including cells with no expression. Violins are colored according to the percentage of cells expressing the transcript of interest.

Volcano plots represent the differentially expressed transcripts with the color showing the average expression (log₂) derived from the group in which the gene is up-regulated and the size showing the difference in percentage of expressing cells between groups. Only transcripts present in > 25% of the total cells are shown in the volcano, but all the differentially expressed genes are shown in the corresponding supplementary tables. Crater plots are scatter plots that depict log₂ (|fold change|) of expression for all transcripts from two distinct scDGEA comparisons, each comparison representing one axis. Every dot is a given transcript, with the size representing the average of both significance values [-log₁₀ (adjusted P-value)] and the color representing the average level of expression for all cells analyzed.

Single-cell trajectory analysis.

Single-cell trajectory analysis was performed using Monocle 3 (v1.0.0, default settings).¹⁵⁷ Briefly, the method employed the number of unique molecular identifiers (UMIs) and the percentage of mitochondrial UMIs as the model formula, in addition to the most variable genes identified through Seurat analysis, to ensure consistency. Two different partitions were obtained using the cluster_cells function, followed by trajectory learning with learn_graph (minimal_branch_len=30). To visualize the cell pseudotime, the trajectory was projected onto the principal component analysis (PCA) and unified manifold approximation and projection (UMAP) generated from Seurat analysis. The T_CM cluster was selected as the root based on prior biological knowledge.

ELISA for protein quantification

Cytotoxic molecules (GZMA, GZMB) and pro-inflammatory cytokines (CCL3, CCL4) from BAL supernatants (> 100 mL) from mild and severe asthmatic patients were quantified using a multiplex ELISA assay (U-PLEX assay platform, Meso Scale Discovery), following the manufacturer’s instructions. Total protein in the BAL supernatants was measured using the Pierce BCA technique (Pierce BCA Protein Assay Kit, Thermo Scientific) for normalization of protein concentrations.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analysis

For RNA-seq data analysis, statistical methods have been described here above. We used unpaired nonparametric T test (Mann-Whitney) for analysis between two groups, and unpaired non-parametric Kruskal-Wallis test for multiple comparisons (more than two groups) followed by Dunn’s post-hoc test for correction. For correlation analysis with clinical features, as the data used was either ordinal and/or non-linearly distributed, we used Spearman correlation coefficients followed by Bonferroni-Hochberg correction. Correlation trendlines were drawn by simple linear regression. We used GraphPad Prism 9.0.1. All source data is detailed in Supplementary Tables.

Supplementary Material

2

Table S1. Study design and experimental details (Related to Figures 1, 2, and 5; and Figures S2 and S3).

(A) Clinical details per donor. (B) Development of the adapted asthma severity score. (C) Experimental and sample details. (D) Proportions of CD4⁺ T cell subsets sorted per donor from gating strategy in resting conditions. (E) Clinical associations. (F) Proportions of CD4⁺CD69⁺CD103⁺ T_RM and CD4⁺CD69⁺CD103^– T_RM cells in BAL samples from healthy donors, data obtained from published datasets.^29,44,45 (G) Soluble protein levels measured by multiplex ELISA from BAL supernatants.

3

Table S2. Single-cell analysis (Related to Figures 1–5; and Figures S1-S5).

(A) CD4⁺ cluster cell counts and proportions for each donor. (B) Post-sequencing and aggregation quality controls from Cell Ranger (v3.1.0). (C) Single cell differential gene expression analysis (scDGEA) between each cluster [CD103⁺ T_RM, CD103^– T_RM, T_CM, T_REG, T_FH, T_IFNR, T_CYCLE, T_CTL] and all other cells. (D) scDGEA between CD103⁺ T_RM and CD103^– T_RM clusters, using sex as covariate. (E) scDGEA between CD103⁺ T_RM and CD103⁻ T_RM clusters in male and female asthmatic patients separately. (F) scDGEA between CD103⁺ T_RM and CD103⁻ T_RM clusters, using treatment as covariate. (G) Gene lists of DGEA between CD103⁺ T_RM and CD103⁻ T_RM clusters using sex or treatment as covariate. (H) Gene set variation analysis (GSVA) per donor and cluster for TCR signaling, cytotoxicity, and cAMP signaling lists (Lists in Table S3A). (I) scDGEA between CD103⁺ T_RM cluster on OCS versus No OCS treatment, using sex as covariate. (J) scDGEA between CD103⁺ T_RM cluster on Biologics versus No Biologics treatment, using sex as covariate. (K) Integration analysis table showing proportions of GZMB-expressing cells in all CD4⁺ T cells from healthy subjects, mild asthmatic and severe asthmatic patients.^46–49 (L) scDGEA between disease groups separated by sex for each cluster [CD103⁺ T_RM, CD103⁻ T_RM, T_CM, T_REG, T_FH, T_IFNR, T_CYCLE, and T_CTL]. (M) scDGEA between disease groups regardless of sex for each cluster. (N) scDGEA between stimulation and resting conditions. (O) scDGEA between CD4⁺ T cells stimulated from severe asthma versus mild asthma. (P) scDGEA between disease groups separated by sex for stimulation condition. (Q) Proportions of GZMB-expressing cells in all stimulated CD4⁺ T cells from mild and severe asthmatic patients. (R) scDGEA between GZMB-positive and GZMB-negative cells in stimulation condition, using sex as covariate. (S) scDGEA between GZMB-positive cells from severe versus mild asthma in stimulation condition, using sex as covariate. (T) Co-expression analysis of a selection of genes in GZMB-positive cells in stimulation condition. Percentage of GZMB-expressing cells expressing given transcripts are shown. (C-F, I, J, L-P, R and S) For each gene, fold change (log₂), P value, Bonferroni-Hochberg adjusted P-value, mean expression, and fraction of expressing cells (CPM >0) are shown.

4

Supplementary Table 3. Gene signatures (Related to Figures 1 and 3-5; and Figures S1, S3, and S4).

(A) Gene signature lists used for Gene Set Enrichment Analysis. (B) IPA analysis for genes upregulated in the CD103⁺ T_RM cluster. (C) Pathway enrichment analysis for genes upregulated post-stimulation.

5

Supplementary Table 4. Differential gene expression analysis in bulk populations (Related to Figure 3; and Figures S3 and S5).

Per gene, log₂ fold change, adjusted P-value, mean expression. Statistical tests between 3 cell types (CD103⁺ T_RM, CD103⁻ T_RM, and non-T_RM) in both resting (A) and stimulation (D) conditions. (B) TCR clonotype diversity indexes for the 3 cell types. (C) TCR clonotype sharing between the 3 cell types. (E) Differential gene expression analysis between CD103⁺ T_RM cells in stimulation versus resting condition.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-Human CD45 (2D1) - Alexa Fluor 700	BioLegend	Cat# 368514; RRID: AB_2566374
Anti-Human CD3 (SK7) - APC-Cy7	BioLegend	Cat# 344818; RRID: AB_10645474
Anti-Human CD8a (RPA-T8) - BV570	BioLegend	Cat# 301038; RRID: AB_2563213
Anti-Human CD4 (RPA-T4) – BV510	BioLegend	Cat# 300546; RRID: AB_2563314
Anti-Human CD357/GITR (108–17) – BV711	BioLegend	Cat# 371212; RRID: AB_2687161
Anti-Human CD185/CXCR5 (RF8B2) – BV421	BD Biosciences	Cat# 562747; RRID: AB_2737766
Anti-Human CD25 (2A3) – BB515	BD Biosciences	Cat# 564467; RRID: AB_2744340
Anti-Human CD127 (eBioRDR5) - APC	eBioscience	Cat# 17–1278-42; RRID: AB_1659670
Anti-Human CD69 (FN50) – BV605	BioLegend	Cat# 310938; RRID: AB_2562307
Anti-Human CD103 (Ber-ACT8) – PE-Cy7	BioLegend	Cat# 350212; RRID: AB_2561599
Anti-Human FcR Blocking Reagent	Miltenyi Biotec	Cat# 130–059-901
Anti-Human CD103 (Ber-ACT8) - TotalSeq-A	BioLegend	Cat# 350231; RRID: AB_2749996
Anti-Human CD69 (FN50) - TotalSeq-A	BioLegend	Cat# 310947; RRID: AB_2749997
Biological samples
Human bronchoalveolar lavage (BAL) samples	This paper	Protocol number IRB VD-156–1118
Chemicals, peptides, and recombinant proteins
Recombinant RNAse inhibitor	Takara Bio	Cat# 2313B
PMA (Phorbol 12-myristate 13-acetate)	Sigma-Aldrich	Cat# P8139–1MG
Ionomycin (calcium salt)	Sigma-Aldrich	Cat# I0634–1MG
Brilliant Stain Buffer Plus	BD Biosciences	Cat# 566385
Propidium Iodide solution	Sigma-Aldrich	Cat# P4864
Critical commercial assays
Chromium Single Cell 3’GEM Kit v3	10x Genomics	Cat# PN-1000077
Chromium Single Cell 3ʹ Library Kit v3	10x Genomics	Cat# PN-1000078
Chromium Single Cell 3ʹ Gel Bead Kit v3	10x Genomics	Cat# PN-1000076
Chromium Chip B Single Cell Kit	10x Genomics	Cat# PN-1000073
Chromium i7 Multiplex Kit	10x Genomics	Cat# PN-120262
Chromium Single Cell 3ʹ Feature Barcode Library Kit	10x Genomics	Cat# PN-1000079
Nextera XT DNA Library Preparation Kit	Illumina	Cat# FC-131–1096
DNeasy Blood and Tissue Kit	Qiagen	Cat# 69506
Infinium Multi-Ethnic Global-8 v1.0 Kit	Illumina	Cat# WG-316–1001
Deposited data
Bulk RNA-sequencing data	This paper	GEO: GSE181709
Single-cell RNA-sequencing data	This paper	GEO: GSE181710
Combined bulk and single-cell RNA-sequencing data contained in GSE181709 and GSE181710	This paper	GEO: GSE181711
Original code	Zenodo https://doi.org/10.5281/zenodo.8342757	N/A
Software and algorithms
FlowJo v10.7.1	FlowJo	https://www.flowjo.com/
Prism 9	Graphpad	https://www.graphpad.com
IPA (Ingenuity Pathway Analysis)	Qiagen	www.qiagen.com/ingenuity
Cell Ranger v3.1.0	10x Genomics	https://www.10xgenomics.com
Seurat v3.0.2	(Stuart et al., 2019)	https://www.satijalab.org/seurat
R v4.0.1	R Core team	www.R-project.org
DESeq2 v1.16.1	(Love et al., 2014)	http://www.bioconductor.org/packages/release/bioc/html/DESeq2.html
ComplexUpset v1.3.3	(Lex et al., 2014)	https://github.com/krassowski/complex-upset
Monocle3 v1.0.0	(Trapnell et al., 2014)	https://github.com/cole-trapnell-lab/monocle3
MAST v1.8.2	(Finak et al., 2015)	https://github.com/RGLab/MAST
MiXCR v2.1.10	(Bolotin et al., 2015)	http://mixcr.milaboratory.com/
vdjtools software v1.2.1	(Shugay et al., 2015)	https://github.com/mikessh/vdjtools
FGSEA v1.10.1	(Korotkevich et al., 2021)	https://bioconductor.org/packages/release/bioc/html/fgsea.html
clusterProfiler v4.6.2	(Wu et al., 2021)	https://github.com/GuangchuangYu/enrichment4GTEx_clusterProfiler
SCGSVA v0.0.14	(Hänzelmann et al., 2013)	https://github.com/guokai8/scGSVA
Other
Smart-seq2 protocol	(Picelli et al., 2013) (Rosales et al., 2018)	https://www.nature.com/articles/nmeth.2639 https://link.springer.com/protocol/10.1007/978–1-4939–7896-0_21

Highlights:

CD103⁺CD4⁺ T_RM cell frequencies are increased in a subset of severe asthmatics

CD103⁺CD4⁺ T_RM cells in airways display T_H1, cytotoxic, and pro-inflammatory features

The CD103⁺CD4⁺ T_RM^HIGH endotype is associated with male late-onset severe asthma phenotype

Context and Significance:

Severe asthma is a heterogeneous chronic disease, characterized by persistent airway inflammation and remodeling refractory to current treatments. The current paradigm suggests that CD4⁺ T_H2 cells play a central role in asthma pathogenesis. Although therapies targeting type 2 cytokines have shown promise, less than 50% of patients respond to treatments, and effects are neither durable nor reverse airway remodeling, which progressively lead to permanent airflow obstruction. Researchers from the La Jolla Institute for Immunology and the University of Southampton profiled the single-cell transcriptomes of airway T cells from severe asthmatics and discovered the presence of airway tissue-resident cytotoxic CD4 T cells that may play an important role in asthma pathogenesis. This important discovery opens a new avenue for a better stratification of severe asthma patients and new therapies.