Single-cell resolution of human airway epithelial cells exposed to bronchiolitis obliterans-associated chemicals

Abstract

Bronchiolitis obliterans (BO) is a fibrotic lung disease characterized by progressive luminal narrowing and obliteration of the small airways. In the nontransplant population, inhalation exposure to certain chemicals is associated with BO; however, the mechanisms contributing to disease induction remain poorly understood. This study’s objective was to use single-cell RNA sequencing for the identification of transcriptomic signatures common to primary human airway epithelial cells after chemical exposure to BO-associated chemicals—diacetyl or nitrogen mustard—to help explain BO induction. Primary airway epithelial cells were cultured at air-liquid interface and exposed to diacetyl, nitrogen mustard, or control vapors. Cultures were dissociated and sequenced for single-cell RNA. Differential gene expression and functional pathway analyses were compared across exposures. In total, 75,663 single cells were captured and sequenced from all exposure conditions. Unbiased clustering identified 11 discrete phenotypes, including 5 basal, 2 ciliated, and 2 secretory cell clusters. With chemical exposure, the proportion of cells assigned to keratin 5+ basal cells decreased, whereas the proportion of cells aligned to secretory cell clusters increased compared with control exposures. Functional pathway analysis identified interferon signaling and antigen processing/presentation as pathways commonly upregulated after diacetyl or nitrogen mustard exposure in a ciliated cell cluster. Conversely, the response of airway basal cells differed significantly with upregulation of the unfolded protein response in diacetyl-exposed basal cells, not seen in nitrogen mustard-exposed cultures. These new insights provide early identification of airway epithelial signatures common to BO-associated chemical exposures.

NEW & NOTEWORTHY Bronchiolitis obliterans (BO) is a devastating fibrotic lung disease of the small airways, or bronchioles. This original manuscript uses single-cell RNA sequencing for identifying common signatures of chemically exposed airway epithelial cells in BO induction. Chemical exposure reduced the proportion of keratin 5+ basal cells while increasing the proportion of keratin 4+ suprabasal cells. Functional pathways contributory to these shifts differed significantly across exposures. These new results highlight similarities and differences in BO induction across exposures.

INTRODUCTION

Bronchiolitis obliterans (BO) is a fibrotic lung disease characterized by progressive luminal narrowing and obliteration of the small airways, or bronchioles (1). BO occurs commonly after lung transplantation with nearly 50% of all lung transplant recipients experiencing some form of chronic lung allograft dysfunction at 5 years after transplant (2–4). In the nontransplant population, inhalation exposure to certain chemicals is also associated with BO development (1, 2, 5–7). Chemicals associated with BO after inhalation exposures include alpha-diketones such as diacetyl (DA; 2,3-butanedione) (6) and alkylating agents like sulfur mustard (SM; 2,2′-dichlorodiethyl sulfide) (8). Despite the growing list of chemicals associated with BO development, pathological mechanisms common to chemical inhalation exposures remain poorly understood.

One of the most well-known chemicals associated with BO development is diacetyl (6, 9, 10). Diacetyl (DA; 2,3-butanedione) is a reactive alpha-diketone added or naturally occurring in food products (11). Though previously classified as generally safe for consumption by the US Food and Drug Administration, DA is now known to cause BO when inhaled at occupationally relevant concentrations (6, 12). Workers exposed to high concentrations of DA develop severe shortness of breath and debilitating lung disease that infrequently progressed to lung transplantation (6). More recently, inhalation exposure to DA in other food processing occupations, such as coffee bean roasting, has been associated with similar airway pathology (11–13).

Sulfur mustard (SM; 2,2′-dichlorodiethyl sulfide) is a second chemical associated with BO. SM is classified as a serious chemical threat due to its known use as a warfare agent that remains the most utilized chemical weapon on the planet (14). During the Iran-Iraq War, SM use as a chemical weapon resulted in greater than 100,000 chemical casualties (14, 15). Like sulfur mustard, nitrogen mustard [NM; bis(2-chloroethyl)methylamine] is a chemical warfare agent stockpiled by many European countries during World War II (16). Both SM and NM cause acute and persistent injury to the respiratory tract when inhaled (17, 18). Acute pulmonary injury manifests as pulmonary edema, pulmonary hemorrhage, airway mucosal injury, and obstruction at high concentrations (19–21). In those individuals who survive an acute exposure, late pulmonary sequelae, including BO, often develop (7, 8, 22).

Chemical inhalation exposure results in injury to the airway epithelium (9, 21). The airway epithelium is the first line of defense against chemical inhalation exposures. Repair of the airway epithelium is dependent upon local stem cells, or resident progenitor cells, for an injury response (23). When the repair response of resident epithelial progenitor cells is impaired, a potent inflammatory response and subsequent scar tissue formation develop, and the end result is BO development (24). Considering BO as a histopathological diagnosis, understanding the initial response of the human airway epithelium to toxic chemical injury is essential to better understand the mechanisms contributing to end-stage disease development.

The purpose of this study was to compare and contrast the transcriptomic gene signatures from individual human airway epithelial cells exposed to DA or NM vapors using single-cell RNA sequencing (scRNA-seq) in order to identify pathways common to the airway epithelial after chemical exposures that contribute to BO development.

METHODS

Chemicals

Diacetyl (DA; 2,3-butanedione, 98.5% purity) and nitrogen mustard [NM; bis-(2-chloroethyl)-methylamine, 98.0% purity] were purchased (Sigma-Aldrich, Missouri) and approved by University of Rochester’s Environmental Health and Safety Committee.

Primary Human Airway Epithelial Cultures

Differentiated cultures were generated from primary human bronchial epithelial cells from healthy, nonsmoking donors (demographics in Supplemental Table S1). Detailed methods of culture technique are provided in the Supplemental Methods. In brief, airway epithelial cells were cultured at air-liquid interface (ALI) for 21 days before chemical exposure on a microporous Transwell (9 mm internal diameter) membrane in plastic inserts. Prior to chemical exposures, cultures were placed into 1 mL of culture medium (MatTek, Ashland, MA) in 6-well culture plates for 4 or 24 h. Three donors (3x) were used for each experiment with exposure replicates performed from each donor (2x). A schematic outline of the study design is also shown in Fig. 1.

Figure 1.

Schematic outline of the study design. Bronchial airway epithelial cells dissociated from the airways of 3 deidentified, deceased human lung donors were used for creating air-liquid interface (ALI) cultures. Following establishment at ALI, epithelial cells were exposed to chemical vapors, either DA, NM, or PBS. Exposures were performed in technical replicates (2x) for all 3 donors across the 3 exposures. Two exposure replicates from donor 1 (DA, NM, and PBS) were sent for single-cell RNA sequencing (scRNA-seq) at 4 h postexposure. Validation using Western blot, immunofluorescent staining (at 24 h), and real-time PCR (RT-PCR at 4 h) occurred for all 3 donors in replicates (3 × 2). [Image created with a licensed version of BioRender.com.]

In Vitro Vapor Cup Exposure

Exposure concentrations included 25 mM DA and 100 µM NM. Concentrations were based on prior dose-response experiments using the vapor cup exposure model (25, 26) and were comparable to occupational exposures for DA (27, 28) or chemical warfare exposures for NM (29, 30). Exposures occurred for 1 h at 37°C. Cells were collected at 4 h for RT-PCR or 24 h for histology (26, 28). Phosphate-buffered solution (PBS) and 10% hydrogen peroxide (H₂O₂) were used as negative and positive treatments, respectively.

Cell Viability via Lactate Dehydrogenase Activity

Lactate dehydrogenase (LDH) release was measured in apical washes of the airway epithelial cultures cultured at air-liquid interface (ALI) as a surrogate marker of cell viability using a commercially available calorimetric kit (490/680 nm; Thermo Scientific Pierce; Rockford, IL). The apical surface of the ALI tissues was gently rinsed with 0.4 mL of PBS before exposure, as well as at 4 or 24 h after exposure for LDH supernatant activity. Apical rinses were centrifuged to remove mucus/debris before LDH testing. A positive control was the absorbance of LDH release when lysed with 10% hydrogen peroxide (H₂O₂). The baseline LDH release was also measured in all cultures pretreatment.

Trans-Epithelial Electrical Resistance

Trans-epithelial electrical resistance (TEER) was measured in all tissue cultures before and at 4 or at 24 h after exposure using silver chloride electrodes (EVOM², World Precision Instruments, Sarasota, FL). Electrodes were connected to the volt-ohmmeter and equilibrated in balanced PBS solution (MatTek) for 15 min before use. Four hundred microliters of warm PBS solution (MatTek) were added to the apical surface of ALI cultures. TEER was measured by placing the longer electrode into the basal media and the shorter electrode into the apical Transwell insert. At least two measurements were taken from each insert at each time point and averaged. ALI cultures were not used for exposure if the TEER measurement was ≤300 Ω/cm² before exposure (26, 31).

Sample Collection, Library Preparation, and Single-Cell RNA Sequence Mapping

At 4 h postexposure, cultures were dissociated and processed using the single-cell RNA-seq kit (10x Genomics, California). Sequence data were generated using Illumina’s NovaSeq500 platform (Illumina, California), and cells were counted via 10x CellRanger (v3.0.2). Samples were aligned to 10x (GRCh38-2020-A) and the NCBI GenBank (MT644268) reference libraries. Donor 1 was used for single-cell RNA sequencing with validation experiments (RT-PCR and immunofluorescence) performed for all donors.

scRNA-Seq Quality Control Analyses, Normalization, and Data Integration

Quality control, cell subset analyses, and differential gene expression were performed using Seurat (v4.0.4) (Supplemental Fig. S1). Data were normalized using the SCTransform method (32) and integrated through Robust Principal Component Analysis (33). Cluster marker genes were determined from ToppCell Atlas (34, 35).

Differential Gene Expression and Functional Pathway Enrichment Analyses

Differential gene expression analysis was performed using the Model-based Analysis of Single-cell Transcriptomics (MAST) method (36). Genes were declared differentially expressed if the false discovery rate (FDR) adjusted P value <0.05 and the fold change >1.5. Canonical pathway, ontology, and phenotype functional enrichment analyses were performed using ToppGene and reported when the adjusted P value <0.05.

Real-Time PCR of Whole Airway Epithelial Culture Homogenates

Total RNA was extracted from a subset of exposed cultures at 4 h postexposure using RNeasy and one-step real-time (RT) PCR kits (Qiagen, Hilden, Germany). After generation, RNA integrity was assessed using an Agilent Bioanlayzer (Agilent Technologies, Palo Alto, CA), and then converted to cDNA using the cDNA Synthesis Kit (Qiagen, Hilden, Germany). Real-time PCR was performed using the SYBR Green PCR Master Mix on Bio-Rad C100 Touch Thermal Cycler with associated software technologies (Bio-Rad, Hercules, CA). Peptidylprolyl isomerase A (PPIA) was used as the endogenous control primer. Select primers for basal cell genes (KRT5, KRT14, SOX2, and TP63) and ubiquitin-protein stress genes [UBC and sequestome-1 (SQSTM1)] were used for validation (Bio-Rad, Hercules, CA). Ct values was normalized to PPIA, and relative quantification was calculated using the 2^−ΔΔCt method. Samples from all three donors were used for RT-PCR, and the mean was reported for each donor for the two exposure replicates.

Trajectory Analysis

Monocle 3 (version 1.2.7) was used for trajectory analysis (37). Basal cell clusters included “proximal basal” (cluster 2), “differentiating basal” (cluster 5), “basal” (cluster 6), and “proliferating basal” (cluster 8). Chemical exposure was used to build trajectories, and trajectories were measured via pseudotime.

Histological and Immunofluorescent Evaluation of Primary Airway Epithelial Cultures

Immunohistochemistry and immunofluorescent staining were performed at 24 h postchemical exposure as changes in protein expression often lag changes in transcriptional changes. For immunofluorescent staining, cultures were fixed in 10% neutral buffered formalin, embedded, and sectioned. Sections were stained with hematoxylin and eosin (H&E) or primary antibodies against keratin 5 (Krt5), keratin 4 (Krt4), acetylated tubulin (α-Tub), sequestome-1 (p62), or major histocompatibility antigen class II (MHC-II). IgG antibodies were used as negative controls. Slides were counterstained and mounted with DAPI Fluoromount-G. Staining was performed for all three donors with two exposure replicates (6 total slides/group) and for each exposure group (3 groups: DA, NM, or PBS; see Fig. 1). For semiquantitation of immunofluorescent-stained slides, two images at ×100 were obtained for each stained section at 25% and 75% length of the total embedded section.

Statistical Analyses

Exposure replicates (2x) were performed for all three donor samples. Single-cell RNA sequencing data were analyzed as described earlier. For validation methods, means ± standard deviation (SD) were reported and analyzed using a one-way analysis of variance (ANOVA) with Dunnett’s post hoc analysis in Prism 8.0 (GraphPad, California).

RESULTS

Identification and Clustering of Chemically Exposed Airway Epithelial Cells

A total of 75,663 dissociated single cells were captured and sequenced from primary human airway epithelial cells exposed to DA, NM, or PBS. Average reads per cell was 52,802 with an average of 4,126 genes identified/cell and an average read fraction of 82.9%. Unbiased clustering identified 11 discrete clusters, including 5 basal cells, 3 ciliated cells including “FOXN4+” cells, 2 secretory cell clusters, as well as the unique “ionocytes” cluster (Fig. 2A). The top five most abundant genes associated with each cluster are listed in Fig. 2B with genes identified within each cluster across each exposure and capture.

Figure 2.

A: uniform manifold approximation and projection (UMAP) visualization of the primary human airway epithelial cell dataset. Each distinct cell cluster is defined by a specific color and with each respective title (far right). B: top 5 most abundant genes associated for each cluster. C: UMAP plots demonstrating the relative distribution of various basal cell genes including keratin 4 (KRT4), keratin 5 (KRT5), keratin 13 (KRT13), tumor protein transcription factor p63 (TP63), topoisomerase II A (TOP2A), marker of proliferation Ki-67 (MKI67), vimentin (VIM), fibronectin (FN1), small proline-rich region 1B and 2D (SPRR1B and SPRR2D). D: UMAP plots demonstrating the relative distribution of ciliated cell markers (FOXJ1), dynein axonemal heavy chain 11 (DNAH11), and cystic fibrosis transmembrane receptor (CFTR). E: UMAP plots of secretory cell genes including mucin 5B and 5AC (MUC5B and MUC5AC) and secretoglobin 1A1 (SCGB1A1).

Basal cell clusters identified included “proximal basal,” “keratinocytic basal,” “differentiating basal,” “basal,” and “proliferating basal.” Genes common to airway basal cells, such as keratin 5 (KRT5) and transcription factor p63 (TP63), were expressed most abundantly in the “keratinocytic basal,” “basal,” and “proliferating basal” clusters (Fig. 2C). The “proliferating basal” cell cluster was marked by the cell cycle control genes TOP2A and MKI67. Similar to the “basal” cell cluster, the “keratinocytic basal” cluster expressed keratin 14 (KRT14), the intermediate partner filament of KRT5. In contrast, multiple genes associated with mesenchymal cells, including fibronectin 1 (FN1) and vimentin (VIM), were also expressed in the “keratinocytic basal” cluster suggestive of an epithelial-mesenchymal transition (EMT) cell type. The other basal cell clusters, “proximal basal” and “differentiating basal,” expressed the intermediate filaments (IFs) keratin 4 (KRT4) and keratin 13 (KRT13). The “proximal basal” cell cluster was most similar to the transitional, suprabasal cell cluster identified previously within the mouse trachea (38, 39) and proximal human airway epithelium (40), also known as “hillock” cells. In contrast, the “differentiating basal” cell cluster genes commonly associated with squamous cell metaplasia and cornification, such as small proline-rich genes [small proline-rich region 1B (SPRR1B), SPRR2B, and SPRR2D; Fig. 2C], were more commonly expressed than in the “proximal basal” cell cluster. These five distinct basal cell clusters align with other single-cell RNA sequencing datasets from both mouse trachea (38, 41) and human airways (40, 42), identifying multiple and distinct basal cell clusters within our ALI culture.

The three ciliated cell clusters (ciliated I, ciliated II, and FOXN4+) were enriched for genes associated with microtubules and dynein arms. Using forkhead box J1 (FOXJ1) as one of the most common genes associated with ciliated cells (43), expression was strongest in “FOXN4+ cells” followed by “ciliated II” and then “ciliated I” cell clusters (Fig. 2D). This same expression pattern was common for multiple other cilia-associated genes, suggesting that the “ciliated II” cluster was more differentiated than the “ciliated I” cluster. Genes affiliated with secretory cells were expressed most abundantly in the “mucus cell” and “goblet cell” clusters (Fig. 2E). The common club cell gene secretoglobin 1A1 (SCG1A1) was of greatest abundance in the “mucus cells” cluster, whereas the “goblet cell” cluster expressed the common mucin genes MUC5AC and MUC5B, albeit not mutually exclusively. Collectively, these gene markers highlight the richness and diversity of this single-cell RNA sequencing dataset.

Chemical Inhalation Exposures Prime Airway Epithelial Cells toward Proximal Basal and Secretory Cell Phenotypes

Airway epithelial cell cultures were then exposed to chemical vapors, either diacetyl (DA) or nitrogen mustard (NM), and compared with PBS-exposed cultures. Supernatant lactate dehydrogenase (LDH) (Fig. 3A) and trans-epithelial electrical resistance (TEER) measurements (Fig. 3B) were obtained at 4 and 24 h after exposure as functional outcomes of epithelial cell death and permeability, respectively. Increased cell death was developed in epithelial cultures exposed to NM at 24 h compared with PBS control cultures. There was no significant rise in supernatant LDH levels in DA-exposed cultures. For both DA- and NM-exposed cultures, TEER measures decreased significantly at 4 h relative to PBS controls. By 24 h, TEER measurements did not differ significantly in DA-exposed cultures compared with that of PBS controls but remained significantly lower in NM-exposed cultures in comparison with PBS-exposed samples. These results highlight the similarities and differences in functional responses of the airway epithelial following chemical vapor exposures.

Figure 3.

Supernatant lactate dehydrogenase (LDH; A) and transepithelial electrical resistance (TEER; B) measurements from exposed epithelial cultures. Preexposure measurements (black bar) occurred in all epithelial cultures collected immediately prior to chemical exposure (n = 24). Hydrogen peroxide (10%, gray bar) was used as a positive control for both functional assays. Apical washes were collected in DA-exposed cultures (red bars; n = 3 donors × 2 replicates/group and time point) or NM-exposed cultures (blue bars; n = 3 donors × 2 replicates/group and time point) at 4 or 24 h after exposure, respectively. Supernatant LDH increased significantly in NM-exposed cultures at 24 h compared with PBS-exposed controls—24-h cultures (*P < 0.05). TEER decreased significantly at 4 h in DA-exposed (***P < 0.001) and NM-exposed cultures (**P < 0.01) compared with PBS-exposed controls. TEER measurements remained significantly decreased in NM-exposed cultures at 24 h compared with PBS-exposed controls (****P < 0.0001). C: bar plot with the proportion of each cell cluster with respect to exposure. D: representative images of cross-sectional airway epithelial cultures exposed to PBS, DA, or NM, embedded at 24 h after exposure and stained with hematoxylin and eosin (H&E) (bar: 100 µm). E: representative immunofluorescent images of exposed airway epithelial cultures stained for keratin 5 (Krt5; red), keratin 4 (Krt4; green), and 4′,6-diamidino-2-phenylindole (DAPI; blue) (bar: 50 µm). F: percent (%) Krt5+ cells per 100 nucleated (DAPI+) epithelial counted along the membrane cells assessed across each exposure (PBS: black; DA: red; NM: blue; n = 3 donors × 2 replicates/group). Average with standard deviation reported (ANOVA with Tukey’s correction: *P = 0.012 and *P = 0.039). G: percent (%) keratin 4+ cells per 100 nucleated (DAPI+) suprabasal epithelial cells assessed across each exposure (PBS: black; DA: red; NM: blue; n = 3 donors × 2 replicates/group). Average with standard deviation reported (ANOVA with Tukey’s correction: *P = 0.019 and *P = 0.049). DA, diacetyl (2,3-butanedione); NM, nitrogen mustard (mechlorethamine); PBS, phosphate-buffered solution. H: UMAP plots of clusters with keratin 4 (KRT4; red), keratin 5 (KRT5; green), and colocalization of KRT4 and KRT5 (yellow). UMAP, uniform manifold approximation and projection.

Considering TEER measurements decreased in both chemical exposure groups, 4 h postexposure was chosen for performing single-cell RNA sequencing. First, the proportion of single cells aligned to each cluster was assessed for shifts in proportions after chemical exposure (Fig. 3C). For both exposures, the relative proportion of cells aligned to secretory cells (“mucus” and “goblet cell” clusters) and “proximal basal” cells increased, whereas the proportion of cells assigned to basal cell clusters decreased compared with PBS-exposed controls (number and relative proportions of cells found in Supplemental Tables S2 and S3). To validate this shift in airway epithelial cell phenotypes after chemical exposure, a subset of exposed cultures were monitored for 24 h and embedded for histological (Fig. 3D) and immunofluorescent evaluation (Fig. 3E). The number of keratin 5 positive (Krt5+; red) cells adjacent to the basement membrane decreased in both DA-exposed and NM-exposed cultures at 24 h compared with PBS control samples (Fig. 3, E and F). Conversely, the proportion of suprabasal epithelial cells positive for keratin 4 (Krt4; green) increased in both exposure groups compared with control (Fig. 3, E and G). Few cells coexpressed Krt4 and Krt5 by staining or by KRT4 + KRT5 coexpression via UMAP plots among the various clusters (Fig. 3H). Collectively, these findings support that chemical exposure to DA or NM shifts the proportions of airway epithelial cells assigned to each cluster, with reduced cells aligned to basal cell clusters and increased proximal basal and secretory cell phenotypes following chemical exposure.

Functional Pathway Analysis Identifies Commonality in Ciliated Cells but Significant Variation in Airway Basal Cell Responses following Chemical Exposures

Next, differentially expressed genes (DEGs) were identified for each cluster by comparing DA-exposed or NM-exposed cultures with PBS-exposed samples. In total, across all exposure conditions and cell clusters, 15,526 DEGs were identified. The number and proportion of DEGs for each cluster are shown in Fig. 4A and Supplemental Table S4. The top 10 upregulated or downregulated genes associated with each cluster and exposure are listed in Fig. 4, B and C, respectively. Relative fold change for each gene is included in Supplemental Table S5. Upregulated genes with the greatest fold change and common to most clusters after DA exposure included C11ORF96, CXCL8, DDIT3, GDF15, KLF6, and PPP1R15A. Pathways associated with these genes include the unfolded protein response and endoplasmic reticulum (ER) stress response. In NM-exposed cultures, ALDH1A3, CYP1B1, and DHRS9 were commonly upregulated genes across clusters. Pathways associated with these upregulated genes include retinol and drug metabolism. Few of the top 10 upregulated genes from each cluster were common across chemical exposures (DA vs. NM).

Figure 4.

A: total number of differentially expressed genes (DEGs) within each cluster for diacetyl (DA; red) exposed and nitrogen mustard (NM; blue) exposed. B: top 10 upregulated differentially expressed genes within each cluster for diacetyl (DA; top row) exposed and nitrogen mustard (NM; bottom row) exposed for fold change relative to control samples. The size of a word represents the absolute value of expression change. C: top 10 downregulated differentially expressed genes within each cluster for diacetyl (DA; top row) exposed and nitrogen mustard (NM; bottom row) exposed for fold change relative to control samples. The size of a word represents the absolute value of expression change. D: relative expression of common basal cell genes TP63, KRT5, KRT14, and SOX2 across exposures. Average of exposure replicates with standard deviation graphed and relative to PPIA (PBS: white bars; DA: red bars; NM: blue bars; two-way ANOVA with Dunnett’s correction: KRT14, *P = 0.019; SOX2, *P = 0.042; n = 3 donors/group—average of exposure replicates shown). E: unsupervised hierarchical clustering of functional pathway analyses upregulated in DA-exposed (red boxes) and NM-exposed (blue boxes) airway epithelial cultures and compared across cell clusters. Of note, color coding associated with each cell cluster matches that seen in Figs. 2A and 3C. Partition 3 identifies similar functional pathways upregulated in both DA- and NM-exposed cells and common to “ciliated II” (cluster 7). F: unsupervised hierarchical clustering of functional pathway analyses downregulated in DA-exposed (red boxes) and NM-exposed (blue boxes) airway epithelial cultures and compared across cell clusters. Colors associated with each cell cluster match that seen in Figs. 2A and 3C.

Downregulated genes common across clusters after DA exposure included DDIT4, FOS, KRT5, KRT14, KRT15, MIR205HG, S100A2, and SOX2. With NM exposure, downregulated genes with the greatest fold change common to multiple clusters included FOS, JUN, KRT5, KRT14, KRT15, and S100A2. Considering multiple intermediate filaments commonly associated with airway basal cells were downregulated across clusters for both chemical exposures, RT-PCR was performed as validation for changes in basal cell gene expression of TP63, SOX2, KRT5, and KRT14 (Fig. 4D). As a whole, the relative expression of the selected basal cell genes down trended in both chemical exposures; however, significant variation occurred across DA and NM exposures. SOX2 gene expression decreased significantly in DA-exposed cultures, whereas KRT14 expression decreased significantly in NM-exposed cultures by RT-PCR compared with PBS controls.

Next, using the DEGs for each cluster, functional pathways analyses were compared across chemical exposures for commonly (Fig. 4E) upregulated or downregulated pathways (Fig. 4F). Similar to the variation seen at the individual gene level, only a small number of pathways partitioned together across exposures. More specifically, upregulated pathways common to both DA- and NM-exposed cells were associated with the “ciliated II” cell cluster and, to a smaller extent, “keratinocytic basal” cell clusters (Fig. 4E). No common partitions occurred across chemical exposures for pathways downregulated (Fig. 4F).

Looking more closely at the upregulated pathways common to both exposures in the “ciliated II” cell cluster, the top five (>10 log fold) upregulated pathways included “interferon signaling,” “interferon alpha/beta,” “antigen processing and presentation,” “interferon gamma,” and “cytokine signaling in immune system” (Fig. 5A). At the individual gene level, multiple genes associated with major histocompatibility class II (MHC-II) presentation, including HLA-DPA1, HLA-DQA1, and HLA-DRA, increased within this cluster for both DA and NM exposures (Fig. 5B). Airway epithelial cultures were costained for the pan-MHC-II antibody with the ciliated cell marker acetylated tubulin as further validation (Fig. 5C). Indeed, the percentage of cells positive for MHC-II increased in DA- or NM-exposed epithelial cultures compared with PBS controls (Fig. 5C). The increased expression of MHC-II occurred primarily in apically oriented epithelial cells, with less protein expression detected in the basolateral epithelial cell layer (Fig. 5C). Collectively, these changes in differential gene expression, functional pathway analysis, and immunofluorescent staining support commonality across chemical exposures in apically oriented airway epithelial cells, but with differences in basal cell responses.

Figure 5.

A: biological pathways significantly upregulated from the “ciliated II” cell cluster (cluster 7) in DA-exposed (red) and NM-exposed (blue) epithelial cells (P value <0.05; expressed as P value −log₁₀). B: bubble plot for relative expression of genes upregulated and associated with the top biological pathways identified in the ciliated II cluster with DA or NM exposure. Size of bubble proportional to percent expression with upregulation (red) and downregulation (blue). C: representative immunofluorescent images of exposed airway epithelial cultures stained for acetylated tubulin (α-Tub; red), major histocompatibility class II (MHC-II; green) and 4′,6-diamidino-2-phenylindole (DAPI; blue) in PBS (i), DA (ii), and NM (iii) samples (bar: 50 µm). iv: percent (%) MHC-II+ cells per 100 nucleated (DAPI+) apical epithelial cells assessed across each exposure (PBS: black; DA: red; NM: blue; n = 3 donors × 2 replicates/group). Average with standard deviation reported (ANOVA with Dunnett’s correction: *P = 0.024 and *P = 0.011). DA, diacetyl; NM, nitrogen mustard.

Trajectory Analysis Suggests a Differential Responses of Airway Basal Cells Exposed to Diacetyl Exposures, Not Seen with Nitrogen Mustard Exposures

To help explain the variation in airway basal cell gene signatures associated with each chemical exposure, basal cell clusters were selected for single-cell trajectory analysis using Monocle (Supplemental Fig. S2). UMAP plots of trajectory analyses identified a clear separation in basal cells exposed to DA compared with basal cells exposed to NM or PBS (Fig. 6A). Pseudotime expression allowed for tracking of basal cell trajectories based on chemical exposure (Fig. 6B). Gene clustering for trajectory analysis identified the biological pathways of “hypertrophy,” “unfolded protein response,” “ATF2 pathway,” “AP1 pathway,” and “nuclear receptors pathways” as the top five upregulated pathways in DA-exposed compared to NM- or PBS-exposed basal cells (Fig. 6C).

Figure 6.

A: uniform manifold approximation and projection (UMAP) visualization of selected basal cell clusters across exposures [green: diacetyl (DA), blue: nitrogen mustard (NM), and pink: control (CTL)]. B: UMAP visualization of basal cells with representative pseudotime trajectories. Starting position designated as “basal” cell cluster (cluster 6) with significant deviations in cell paths noted at each node (numbers present on graph). C: biological pathways significantly upregulated by functional pathway analysis in DA-exposed samples and differentially expressed in cells traveling from nodes 13 to 5 in Fig. 5B. D: relative gene expression of sequestome-1 (SQSTM-1) and polyubiquitin-C (UBC) across exposures. Average of exposure replicates with standard deviation graphed and relative to PPIA (PBS: white triangles; DA: red circles; NM: blue squares; 2-way ANOVA with Dunnett’s correction: SQSTM1—*P = 0.012, n = 3 donors/group—average of exposure replicates shown). E: representative immunofluorescent image of exposed airway epithelial cultures stained for keratin 4 (Krt4; green), sequestome-1 (p62; red), and 4′,6-diamidino-2-phenylindole (DAPI; blue) (bar: 50 µm).) PPIA, peptidylprolyl isomerase A.

Two genes common to these upregulated pathways with DA exposure included sequestome-1 (SQSTM1) and the ubiquitin stress gene polyubiquitin-C (UBC). To validate these responses, gene expression for SQSTM1 and UBC was assessed by RT-PCR for all three exposures. The relative gene expression of SQSTM1 and UBC increased in DA-exposed cultures compared with PBS controls, but not in NM-exposed cultures (Fig. 5Civ). By immunofluorescent staining, the expression of p62 (the surrogate protein of SQSTM1) colocalized with Krt4+ cells in DA-exposed cells but was not seen in NM-exposed cells (Fig. 6E and Supplemental Fig. S3). These results suggest that airway epithelial cells upregulate genes associated with the unfolded protein and ubiquitin stress responses after DA exposures with improved cell survival (Fig. 3A) and improved epithelial barrier integrity at 24 h postexposure (Fig. 3B). In contrast, airway epithelial cells fail to respond similarly in NM-exposed cultures with increased cell death and increased airway epithelial permeability.

DISCUSSION

Mechanisms of bronchiolitis obliterans (BO) development are highly complex and remain poorly understood. Central to most BO development pathways is injury to the airway epithelium. Proper repair of an injured airway epithelium is dependent upon resident progenitor cells. In humans, airway basal cells, marked by dual positivity for p63 and Krt5, function as the primary airway epithelial progenitor cell (23, 44, 45). Following most inhalation exposures, airway basal cells respond to injury with proliferation, differentiation, and reestablishment of barrier integrity (46, 47). In contrast, when challenged with certain BO-associated chemicals, significant damage occurs, not only to the airway epithelium but also to the airway basal cell population (25, 48). In support of this hypothesis, Swatek et al. (48) showed depletion of p63 + krt5+ basal cells in a ferret model of transplant-associated BO. Similarly, O’Koren et al. (46) showed reduced numbers of Krt5+ basal cells preceding BO-like pathology in mice exposed to chlorine gas. These preclinical models of BO highlight the relevance of airway basal cells in BO development.

Here, the use of single-cell RNA sequencing provides unprecedented resolution into the differential responses of primary human airway epithelial cells to toxic inhalation exposures. Two chemicals commonly associated with BO development were used for modeling epithelial injury: diacetyl (DA) and nitrogen mustard (NM). For both chemical exposures, the proportion of cells associated with Krt5+ basal cell clusters decreased with the proportion of cells associated with secretory and proximal basal cell clusters, marked by Krt4+ expression, increasing. By functional pathway analysis, pathways associated with antigen processing/presentation and interferon signaling were upregulated with localization of MHC class II expression in apically oriented airway epithelial cells. In contrast, the response of airway basal cells differed significantly between chemical exposures, with induction of the unfolded protein response in DA-exposed basal cells not seen in NM-exposed basal cells.

In prior investigations using DA exposures, global proteomics identified decreased ciliated cell proteins with increased squamous cell metaplasia in primary human airway epithelial cells (49). Similarly, transcriptomic evaluation also identified induction of the proximal basal cell genes keratins 4 and 13 (50). Findings from our studies are directly in line with these prior investigations. Specifically, the proportion of airway basal cells decreased, whereas the number of Krt4+ suprabasal cells increased after DA exposure. With NM exposures, similar results occurred with decreased proportions of Krt5+ cells and increased proportions of Krt4+ suprabasal cells. Interestingly, functional pathways and trajectory analyses found greater divergence than convergence in airway basal cell responses across DA versus NM exposures. With DA, airway epithelial cells upregulate the unfolded protein and ubiquitin stress response, not seen in NM-exposed cultures. One potential explanation for these dichotomous responses is differences in chemical structure and function of DA versus NM. DA is a reactive alpha-diketone, whereas NM is an alkylating agent. Previous literature on NM supports preference of NM to react with guanine and other DNA bases via alkylation (51). In contrast, DA nonenzymatically reacts with arginine residues common to intermediate filaments and other cytoskeletal proteins (52). Hence, differences in chemical tropism are one potential explanation for the dichotomous response of airway basal cells to these two BO-associated chemical inhalation exposures.

In addition to reduced proportions of Krt5+ basal cells, the number of proximal Krt4+ suprabasal cells increased in proportion with exposure to DA or NM. Keratin 4 is a type II intermediate filament (IF) that heterodimerizes with its IF partner keratin 13 (Krt13) (53). These two intermediate filaments have gained greater attention from single-cell RNA sequencing datasets isolated from mouse airways (38, 39) and human airway biopsies (40). Keratin 13+ epithelial cells are found within the intercartilaginous regions of the mouse trachea (38) with trajectory analysis inferring KRT13+ cells as transitional cells, termed “hillock cells,” between KRT5+ basal cells and SCGB1A1+ club cells, and that are independent of KRT8+ epithelial cells. Lineage tracing experiments further support early priming of basal cells toward a club cell phenotype, independent of goblet or ciliated cell differentiation (39). Hillock cells have also been identified in healthy human airway epithelial cells isolated from the proximal airways (40). Our results suggest that the proportion of Krt4+ basal cells increases after chemical inhalation exposures. One explanation is that hillock cells are more resistant to chemical inhalation injury than prototypical Krt5+ basal cells. Future work is underway to evaluate the similarities and differences of these responses after chemical inhalation exposures and any associations between hillock cells and BO development.

Historical teaching is that immune cells, and more specifically antigen-presenting cells, are the primary source of major histocompatibility class II processing and presentation for the lung (54). More recently, a greater appreciation has been given to lung-specific epithelial cells in antigenic presentation (55–57). In healthy human airway samples, staining for MHC II expression is conflicting with initial reports, suggesting that MHC II expression is present in ciliated bronchial epithelial cells (58, 59), while human lung biopsies failed to identify similar airway epithelial expression (60, 61). With autoimmune and allergic lung disease, MHC II expression has been seen in human bronchial epithelial samples (60, 61). Similarly, MHC II expression markedly increases in airway epithelial cells following viral infection but decreases with bacterial infection (62, 63). In the current study, we identified increased expression of MHC class II in ciliated and other apically oriented airway epithelial cells following chemical exposure. Further studies are required to investigate the functional role of increased MHC class II expression in apically oriented airway epithelial cells after chemical exposure.

The current study is not without some limitations. Related to toxicology, single concentrations of DA and NM were used for all exposures and were based on prior publications using primary human airway epithelial cell cultures (25, 26). Timing of cell collection for single-cell RNA sequencing occurred at 4 h postexposure with minimal cell death by supernatant LDH in either chemical exposure (25, 26). A single donor was used for single-cell RNA sequencing, but results from scRNA-seq were validated using RT-PCR as well as immunofluorescent staining from multiple donors. Future studies with larger and more diverse samples are required to further validate findings. Third, BO development is not due to changes in epithelial cells alone. Coculture with other cell types, such as antigen-presenting cells and/or mesenchymal cells, may provide greater functional relevance to changes in MHC II expression after chemical exposure. In addition, the proportion of basal cells in the small airway epithelium is likely less than the proportion of basal cells represented here at ALI. Finally, sequestome-1 (SQSTM1) and polyubiquitin-C (UBC) were chosen as two genes important for protein folding and stress responses. Other genes relevant to the unfolded protein responses, including PPP1R15A and DDIT3, were identified as upregulated in DA-exposed cultures by scRNA-seq and warrant further validation, especially considering the protective role of PPP1R15A in fibrotic lung disease (64).

In conclusion, this study provides novel single-cell resolution to the effects of chemical inhalation exposures on primary human airway epithelial cells. BO-associated chemicals reduced the proportion of Krt5+ basal cells while increasing the proportion of proximal basal cells marked by Krt4 positivity, but the functional pathways contributing to these shifts differed significantly across chemical exposure. Future investigations into the functional consequences of shifting proportions of airway epithelial cells are required to better explain their contribution to BO pathogenesis, yet these new insights provide early identification of airway epithelial signatures unique to BO-associated chemical exposures.

DATA AVAILABILITY

Data are readily available via the supplemental data DOI listed below, and further data are available upon reasonable request.

SUPPLEMENTAL DATA

Supplemental Methods, Supplemental Tables S1–S5, and Supplemental Figs. S1–S3: https://doi.org/10.6084/m9.figshare.24185508.v1.

GRANTS

This work was supported by NIH National Institute of Environmental Health Sciences (NIEHS) Grants P30-ES001247 (to M.D.M.) and K08-ES033290 (to M.D.M.) and by NIH National Heart, Lung, and Blood Institute (NHLBI) Grant U01HL148861-02S2 (to G.S.P.).

DISCLAIMERS

The content is solely the authors’ responsibility and does not necessarily represent the official views of the University of Rochester.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

C.C., S.K., G.S.P., T.J.M., and M.D.M. conceived and designed research; C.C., S.K., T.J.M., and M.D.M. performed experiments; C.C., G.S.P., T.J.M., and M.D.M. analyzed data; C.C., S.K., G.S.P., T.J.M., and M.D.M. interpreted results of experiments; C.C., S.K., G.S.P., T.J.M., and M.D.M. prepared figures; C.C., S.K., G.S.P., T.J.M., and M.D.M. drafted manuscript; C.C., S.K., G.S.P., T.J.M., and M.D.M. edited and revised manuscript; C.C., S.K., G.S.P., T.J.M., and M.D.M. approved final version of manuscript.