Idiopathic subglottic stenosis arises at the epithelial interface of host and pathogen

Background Idiopathic subglottic stenosis (iSGS) is a rare fibrotic disease of the proximal airway affecting adult Caucasian women nearly exclusively. Life-threatening ventilatory obstruction occurs secondary to pernicious subglottic mucosal scar. Disease rarity and wide geographic patient distribution has previously limited substantive mechanistic investigation into iSGS pathogenesis. Result By harnessing pathogenic mucosa from an international iSGS patient cohort and single-cell RNA sequencing, we unbiasedly characterize the cell subsets in the proximal airway scar and detail their molecular phenotypes. Results show that the airway epithelium in iSGS patients is depleted of basal progenitor cells, and the residual epithelial cells acquire a mesenchymal phenotype. Observed displacement of bacteria beneath the lamina propria provides functional support for the molecular evidence of epithelial dysfunction. Matched tissue microbiomes support displacement of the native microbiome into the lamina propria of iSGS patients rather than disrupted bacterial community structure. However, animal models confirm that bacteria are necessary for pathologic proximal airway fibrosis and suggest an equally essential role for host adaptive immunity. Human samples from iSGS airway scar demonstrate adaptive immune activation in response to the proximal airway microbiome of both matched iSGS patients and healthy controls. Clinical outcome data from iSGS patients suggests surgical extirpation of airway scar and reconstitution with unaffected tracheal mucosa halts the progressive fibrosis. Conclusion Our data support an iSGS disease model where epithelial alterations facilitate microbiome displacement, dysregulated immune activation, and localized fibrosis. These results refine our understanding of iSGS and implicate shared pathogenic mechanisms with distal airway fibrotic diseases.


INTRODUCTION
Idiopathic subglottic stenosis (iSGS) is a rare (Maldonado et al., 2014) but devastating broin ammatory airway disease that occurs almost exclusively in adult Caucasian women (Gelbard et al., 2015). The disease is characterized by mucosal in ammation and localized brosis resulting in life-threatening blockage of the upper airway (Gelbard et al., 2016). Current treatments are limited by either their invasive nature or high recurrence rates, and the majority of iSGS patients require frequent procedural interventions following their initial diagnosis (Gelbard et al., 2019). Given the signi cant emotional, physical, and nancial costs associated with recurrent airway obstruction (Gnagi et al., 2015), most research efforts have focused on procedural techniques to improve airway patency .
However, highly focused scienti c approaches to identify key elements of iSGS disease pathophysiology are essential to developing less invasive and more durable treatments.
Histologically, iSGS cases show pronounced brosis restricted to the proximal airway mucosa (Mark et al., 2008). Diverse diseases in divergent organ systems are associated with brosis, suggesting common biologic triggers (Wynn and Ramalingam, 2012). In alternate pulmonary pathologies, airway brosis has been linked to structural and functional disruption of the respiratory epithelium. Alterations in apical cellcell junctions increase epithelial permeability and enhance translocation of pathogens into the subepithelial space, where they encounter local immune cells and generate airway in ammation culminating in tissue brosis (Rezaee and Georas, 2014). Recent data suggests iSGS may share these pathogenic mechanisms with lower airway brotic diseases. Altered epithelial morphology (Lina et al., 2022) and chronic in ammation in airway scar (Clark, 2015) provide preliminary support to the hypothesis that in iSGS, epithelial dysfunction promotes an aberrant immune response culminating in tissue brosis.
In iSGS the functional alterations underlying airway remodeling remain poorly understood. In this study, we harnessed tissue samples from an international iSGS patient cohort and cutting-edge molecular tools to de ne the molecular phenotype of the proximal airway mucosa at single-cell resolution. Our results provide an unbiased assessment of the cell types within the normal human subglottis and illuminate the changes accompanying iSGS. The results suggest epithelial barrier dysfunction and immune in ltration are key components of iSGS pathogenesis. Matched super cial and deep tissue microbiomes support displacement of the native microbiome into the lamina propria of iSGS patients rather than disruption of bacterial community structure. However, animal models con rm that bacteria are necessary for pathologic proximal airway brosis and suggest an equally essential role for host adaptive immunity in remodeling after mucosal injury. Clinical data support the role of epithelial dysfunction in treatment response. Our data suggests that in iSGS the native microbiome is displaced across a dysfunctional epithelial barrier leading to an adaptive immune response which drives obstructive airway brosis. These novel results reshape our understanding of iSGS, implicate shared pathogenic mechanisms with distal airway brotic diseases, and open new avenues for therapy.

RESULTS
Single-cell sequencing reveals epithelial cell loss and a pronounced immune cell in ltrate in subglottic mucosal scar of iSGS patients. iSGS patients possess obstructive mucosal scars in the proximal airway below the vocal cords (Fig. 1A). Study patients were diagnosed according to standard clinical criteria (Gelbard et al., 2015). Tissue biopsies obtained under sterile conditions in the OR, were immediately digested into single-cell suspensions for single-cell RNA sequencing (Fig. 1B). To determine the distribution and phenotype of the cellular populations present in iSGS airway scar, we generated singlecell suspensions from tissue biopsies of both airway scar (n = 7). We matched unaffected airway mucosa (n = 3) (Supplemental Table T1) and performed scRNAseq using the 10x Genomics Chromium platform (see supplementary Materials and Methods). The samples were collected and processed at two different sites (Supplemental Table T2); however, both sites collected cases and controls. To maximize our ability to identify rare cell populations, we jointly analyzed data from all samples. We de ned inclusion criteria   for cells based on observations from the entire dataset, removed low-quality cells accordingly, applied  normalization and variance stabilization of the 25,974 recovered cells using Seurat(Hafemeister and   Satija, 2019; Stuart et al., 2019), integrated the data using the harmony(Korsunsky et al., 2019), performed unsupervised clustering using Seurat, and classi ed the cell type of each cluster based on PanglaoDB (Franzen et al., 2019) followed by manual annotation based on canonical markers to annotate clusters. We de ned 22 cell types/states in the subglottis (Supplemental Figure S1). Initially one small CD8 effector T cell population was grouped together with a larger CD8 T effector population due to observations that cell cycle activity was driving distinct cluster identity. All cell types were identi ed both in airway scar and healthy mucosa. Notably, we did not observe overt batch effects driven by processing site or sequencing batch in our dimensionality reduction and visualization (Supplemental Figure S2 & S3). Cell types/states were also manually grouped into 4 broad tissue classes (Immune/Epithelial/Endothelial/Mesenchymal) based on their identity (Fig. 1C) and con rmed with canonical lineage markers (Supplemental Figure S4). Quanti cation of cell types demonstrated signi cant differences between iSGS airway scar and matched healthy mucosa controls. Airway scar showed signi cantly more Immune cells (cell count per 1000 cells: scar vs healthy control: 636 vs. 238, P = 0.018) and signi cantly fewer epithelial cells (scar vs healthy control: 155 vs. 685, P < 0.001) (Fig. 1D). In addition to quantitative differences in cell types/states, we compared phenotypic alterations between scar and healthy mucosa by examining the number of differentially expressed genes using  Figure S5). When grouping cells into their tissue layer, the epithelium demonstrated signi cantly more DEG than immune cells (mean number of genes P < 0.05 and fold change > |1.5|, Epithelium = 252 vs 11 in immune, P = 0.007) (Fig. 1E). The difference between epithelial cells and broblasts or endothelium was not signi cant. These results suggest that in addition to a quantitative reduction in cell numbers, the residual epithelium in iSGS scar is also phenotypically distinct. The cell type quanti cation based on transcriptional data was con rmed at the protein level with ow cytometry (Fig. 1F).
Molecular and functional evidence of epithelial dysfunction in iSGS scar. We further analyzed the epithelial clusters ( Fig. 2A) and identi ed conserved transcriptional programs in basal (four clusters), ciliated (three clusters), secretory (one cluster), and a proliferating cell subset (one cluster). Based on our observation of differential cluster abundance between scar and healthy mucosa (Fig. 2B), we quanti ed the number of cell types/states from both scar and healthy mucosa. The clusters comprising basal, secretory, and ciliated cells showed signi cant reductions in scar samples (Boxes depict median and interquartile range, whiskers show min to max, *P < 0.05 by Mann-Whitney U.; Fig. 2C). In addition to the dramatic loss of basal cells within airway scar, geneset enrichment analysis demonstrated that residual proliferating epithelial cells expressed a molecular program for epithelial-mesenchymal transition (EMT) (Fig. 2D). Additional upregulated genesets included oxidative phosphorylation and mTOR signaling consistent with observed proliferation markers Regulatory Subunit 105 (Ki67) and Cyclin-Dependent Kinase 1 (CDK1) (Supplemental Figure S4). Gene ontology (GO) pathway analysis supported the Hallmark geneset EMT ndings; iSGS airway scar showed enrichment for mitochondrial matrix genes (along with aerobic respiration and electron transfer activity). In parallel, proliferating epithelial cells in iSGS airway scar showed down-regulated glycosylation and junctional protein complexes (both apical and tight). Both aerobic respiration and loss of cell-cell adhesion are consistent with EMT.
To provide functional evidence of epithelial barrier dysfunction, we investigated if native bacterial displacement into the deeper lamina propria was a unique feature of iSGS. Employing uorescence in situ hybridization (FISH) with the pan-bacterial probe, Eub338, we investigated if mucosal biopsies from iSGS and healthy controls evidenced bacteria in the deep layers of the proximal airway mucosa. Representative FISH stains show iSGS mucosa possessed signal for bacteria in the deeper lamina propria while healthy control did not (Fig. 2E). In a separate biopsy derived from iSGS mucosal scar, transmission electron microscopy demonstrated numerous forms consistent with the size and shape of bacteria in the cell cytoplasm, supporting the FISH staining (Fig. 2F).
Characterization of the mucosal microbiome in iSGS. Given the observed bacterial displacement into the lamina propria of the airway mucosa, we investigated if alterations in microbiome community structure were also observed in iSGS. We rst quanti ed the number of 16S rRNA gene copies in deep tissue biopsies via qPCR (Fig. 3A). We detected consistent signals in both iSGS patients and disease controls (patients that developed subglottic stenosis following prolonged intubation: iLTS). iLTS patients had a signi cantly higher bacterial load than iSGS samples (mean iSGS copy number: 520,000 vs. iLTS: 1,370,000; P < 0.0001). We then performed 16S rRNA sequencing for insight into bacterial community structure. For further analysis, 37/50 (74%) iSGS and 18/27 (67%) iLTS samples were retained after implementing a cutoff of 500 high-quality 16S reads (2 standard deviations above the maximum number of reads in any of the negative controls).
The top 20 most abundant bacterial families and genera are shown per each sample for iSGS, iLTS and healthy subglottic controls (Supplemental Figure S6). The majority of bacterial species present in healthy subglottis were consistent with the established healthy lung microbiome composed of supraglottic predominant taxa (e.g., Prevotella, Streptococcus)(Kitsios et al., 2018). Using principal coordinates analysis (PCoA), we compared the overall microbial community structure between iSGS, iLTS, and healthy subglottic controls. As seen in Fig. 3B, there was no signi cant differences in the centroids between the three groups (PerMANOVA adonis2 testing p = 0.06). To validate these ndings, we next utilized Bray-Curtis dissimilarities to make binary comparisons between healthy and iLTS samples and between healthy and iSGS samples. iSGS samples more closely resembled healthy controls than iLTS samples resembled healthy subglottic controls (Wilcoxon rank sum test with continuity correction, p-value = 0.001 Additional microbial community structure testing using established diversity and richness indices con rmed no detectable differences between iSGS, iLTS and healthy controls. ANOVA testing of alpha diversity (mean Shannon index -iSGS: 15 Anatomic location of bacterial species in iSGS mucosal scar. To validate our ndings and explore if super cial and deep tissue sampling methods produced unique bacterial communities, we next compared published 16S rRNA sequencing data of super cial swabs of iSGS scar (n = 5)(Hillel et al., 2019) with our deep tissue biopsies (Fig. 3E). The top 20 most abundant genera were highly concordant between the super cial and deep sampling methods, despite differences in patient populations and lab processing protocols; offering support for our ndings (the one signi cantly different genus was Halomonas which was abundant is swab samples and absent in tissue).
The native proximal airway microbiome generates an antigen-speci c immune response in in ltrating CD4 + and CD8 + T cells. In order to probe the function of the observed immune in ltrate, tissue biopsies acquired during the operative endoscopy of 5 unique iSGS patients were used to create fresh single-cell suspensions as described. Suspensions were rested for 6 hours, then cultured in the presence of a matched iSGS airway microbiome, the microbiome from an unrelated healthy subject, or left untreated.
Clinical data suggest that restoring epithelial barrier function is associated with a more durable treatment response. Treatment outcomes of iSGS patients in an ongoing cohort study (Gelbard et al., 2018) demonstrate that the most durable treatment responses are seen with surgical resection of the affected airway scar and replacement with healthy proximal airway mucosa (an operation termed cricotracheal resection, CTR (Grillo, 1982) (Fig. 5A). 38 iSGS patients undergoing CTR were propensity score matched to 38 iSGS patients undergoing endoscopic dilation (patient characteristics described in supplemental table 4). CTR had a 5% recurrence rate (2/38), compared with 47% for endoscopic dilation (18/38). (Fig. 5B). Kaplan-Meier analysis and log rank testing con rm a signi cantly lower rate of disease recurrence in patients that underwent CTR (P < 0.0001). These data suggest restoring the epithelial barrier with healthy mucosa minimizes disease recurrence.

DISSCUSSION
Idiopathic subglottic stenosis (iSGS) is a debilitating localized brosis of the proximal airway. Affected patients possess tightly conserved clinical demographics, histopathology, and physiologic impairment (Gelbard et al., 2019). Our human data suggest that defects in epithelial barrier function can allow displacement of the native microbial community deep into the airway mucosa and contribute to dysregulated immune activation. Our complementary animal data support the necessity of bacteria and adaptive immune response to the observed tissue remodeling after proximal airway epithelial injury. Clinical treatment outcome data from an international iSGS patient cohort suggests that restoring the epithelial barrier with healthy mucosa minimizes disease recurrence.
The pseudostrati ed epithelium lining the human airway comprises several distinct populations of cells with specialized effector functions. Airway epithelium and the overlying mucociliary layer maintain a physical barrier against environmental insults (pathogens, allergens, and toxins). Many primary respiratory diseases, including chronic obstructive pulmonary disease (COPD), asthma, and idiopathic pulmonary brosis (IPF), display substantial pathological alterations in the airway epithelium. Evidence suggests that impairment of the epithelial barrier allows bacteria to penetrate the overlying mucociliary layer, intercalate within the epithelium, and activate host immunity. When sustained inappropriately, this in ammation can culminate in brotic tissue remodeling and physiologic impairment.
In COPD secondary to cigarette exposure, respiratory mucosal in ammation and brotic remodeling contribute to small airway obstruction and clinical symptoms (McDonough et al., 2011). Yet even after smoking cessation, many patients with COPD have sustained in ammation and disease progression (Rutgers et al., 2000). Published data demonstrate that persistent epithelial dysfunction in COPD leads to bacterial invasion deep into the mucosa and brotic airway remodeling (Polosukhin et al., 2017). Animal models suggest that endogenous bacteria orchestrate a persistent and pathologic adaptive immune response that drives tissue remodeling (Richmond et al., 2021). Similarly, in idiopathic pulmonary brosis, animal models, in vitro human data and genetic evidence suggest that the airway epithelium plays a central role in disease susceptibility ( Interestingly, in addition to the observed basal cell depletion in iSGS airway scar, geneset enrichment analysis (GSEA) of the residual epithelial cells showed a mechanistic target of rapamycin (mTORC1) pathway activation and enhanced aerobic metabolism. mTOR is a master sensor that integrates environmental factors to regulate cell growth. In general, activation of mTOR stimulates proliferation, mitochondrial biogenesis, and oxidative phosphorylation (Laplante and Sabatini, 2012). The GESA ndings were consistent with the observed KI67 and CDK1 expression (marking cellular proliferation) in the epithelium of iSGS airway scar. Additionally, GSEA showed that the residual epithelium in iSGS airway scar acquired a mesenchymal phenotype with epithelial-mesenchymal transition (EMT) pathway enrichment. EMT allows the disassembly of cell-cell junctions, actin cytoskeleton reorganization, and induction of contractile proteins as non-motile epithelial surfaces convert into individual, motile mesenchymal phenotypic. These phenotypic alterations may result from localized pathogen-driven in ammation(Hofman and Vouret-Craviari, 2012) or can result from mucosal injury secondary to physiologically relevant bile acid exposure (Aldhahrani et al., 2018). Integrating the basal cell depletion with GSEA ndings supports the hypothesis that mucosal barrier dysfunction participates in iSGS disease pathogenesis.
Although an often-overlooked anatomic subsite, the subglottis is uniquely enriched in antigen-presenting dendritic cells and T lymphocytes (Friedman et al., 2007;Jecker et al., 1999). Additionally, it functions as a transition zone from the ciliated lining of the trachea to the squamous epithelium of the larynx. As a consequence, the subglottis has increased exposure to pathogens as the cilia-driven upward movement of the airway mucus layer temporarily stalls (Lee et al., 2000). While the inciting event for barrier dysfunction in iSGS is unclear, both genetic predisposition and environmental insults can impair epithelial function. The genetic foundations of iSGS remain obscure (Drake et al., 2020). Neither an association of iSGS with HLA subtypes nor alternate genetic risk alleles have yet been explored.
Prior small case series examining environmental factors contributing to iSGS have implicated disruption of the proximal airway microbiome, one using pathogen-speci c molecular approaches , and another employing 16S rRNA sequencing of mucosal swabs (Hillel et al., 2019). While these studies implicated the presence of microbial species in iSGS, our new data from a larger, and more diverse patient cohort provides an unbiased characterization of the mucosal tissue microbiome. Our current results suggest that rather than microbiome disruption, in iSGS a "healthy microbiome" is displaced across a dysfunctional epithelium. This displacement is associated with adaptive immune in ltration and activation in response to bacterial species. This is supported by the nding that both a matched host microbiome, as well as the microbiome from a healthy control can drive adaptive immune activation in the proximal airway scar.
However, the displaced microbiome may not be the sole target of the adaptive immune response observed in iSGS mucosal scar. A feed-forward in ammatory loop may become established when peptides from microbial proteins share su cient structural similarity with self-peptides and activate autoreactive T cells, termed "molecular mimicry" (Bachmaier et al., 1999). In ammation resulting from bacterial infection can also activate local antigen-presenting cells and enhance the processing and presentation of self-antigens, referred to as "epitope spreading". Alternatively, T cell populations can be activated to mount effector responses in an antigen-independent, in ammation-dependent manner (termed "bystander activation"). Sustained bystander activation of CD8 T cells has been shown to generate spatially restricted tissue damage in the human disease (Groh et al., 2003;Meresse et al., 2004).
Although we acknowledge the limited ability to assign causality in pathologic studies involving human tissue, single-cell transcriptional pro ling provides an unbiased assessment of the cell types within the human subglottis and illuminates the epithelial and immune alterations accompanying disease. Histologic localization of pathogens deep in the lamina propria and bacterial community pro ling support the functional impact of the observed epithelial dysfunction rather than suggest microbiome disruption is a primary driver of disease. Animal models con rm the importance of both bacteria and host immunity in the pathogenesis of airway brosis after epithelial injury and robust clinical data suggests restoring the epithelial barrier with healthy mucosa minimizes disease recurrence.
Despite the inherent limits involved in rare disease research, our ndings dramatically shift our concept of iSGS disease pathogenesis (Fig. 6). Unbiased transcriptional data, functional studies and clinical data support the concept that epithelial barrier dysfunction allows translocation of native bacteria deep into the airway mucosa and drives dysregulated immune activation leading to brotic remodeling. The disease model establishes a new direction for future studies in iSGS. Treatments that noninvasively promote epithelial barrier integrity and blunt the local adaptive immune response may bene t patients and warrant rigorous future study.

CONSENT FOR PUBLICTION
All authors consent to publish the data.

AVAILABILITY OF DATA AND MATERIAL
The accession number for the 16S sequencing data reported in this paper Bioproject number PRJNA784956, and the accession number for the single-cell RNA-sequencing data is GSE191128.

COMPETING INTERESTS
The authors of this manuscript declare no nancial or other con icts of interest to disclose as described by the journal Microbiome.