Abstract
Objectives
Examine the bronchial epithelium-derived gene expression signature of bronchiectasis (BE) in nasal epithelium.
Methods
We studied 220 participants from the Detection of Early Lung Cancer Among Military Personnel study with bulk RNA sequencing of nasal epithelium brushings. Gene set enrichment analysis (GSEA) examined whether genes previously identified as increased or decreased in the bronchial epithelium of individuals with radiologic BE are enriched among genes most altered in nasal epithelium. GSEA and cell-type specific signatures assessed changes in nasal epithelium cellular composition associated with radiologic BE.
Results
No genes were significantly differentially expressed in nasal epithelium between participants with and participants without widespread radiologic BE. However, genes previously found altered in bronchial epithelium are concordantly enriched among genes most increased or decreased in nasal epithelium (p=3×10−7, increased; p=6×10−8, decreased). GSEA using cell-type signatures shows that BE in nasal epithelium is associated with increased multiciliated and deuterosomal cell-related gene expression and decreased basal cell-related gene expression, consistent with bronchial findings.
Conclusion
This work validates our bronchial BE signature in an independent dataset and demonstrates that gene expression changes associated with radiologic BE in nasal epithelium are concordant but reduced in magnitude compared with bronchial epithelium. These findings support considering nasal brushings as a less invasive tool for BE screening and monitoring.
Shareable abstract
This study validates a previously reported bronchial signature of bronchiectasis in nasal epithelium. Changes in gene expression and cell-type composition associated with radiological bronchiectasis are similar across both bronchial and nasal epithelia. https://bit.ly/4qgq7SO
Introduction
Bronchiectasis (BE) is considered the final common pathway for numerous diseases, including rheumatologic conditions, primary ciliary dyskinesia, cystic fibrosis and pulmonary infections [1]. It is also associated with COPD [2–5]. Patients with BE experience chronic cough, sputum production, dyspnoea and haemoptysis [6]. They also have frequent exacerbations and chronic infection, significantly increasing their risk of morbidity and mortality [7]. On computed tomography (CT), BE is associated with imaging features, including an airway-to-arterial ratio >1, a lack of tapering of bronchi and airway visibility within 1 cm of a costal pleural surface or touching the mediastinal pleura [8–10]. Treatment of BE involves addressing the underlying aetiology, reducing infection risk with antibiotics and managing exacerbations [11].
Transcriptomic analysis of the airway epithelium has previously been employed in cancer [12, 13], COPD [14, 15], emphysema [16] and interstitial lung abnormalities [17]. The concept of an airway field of injury enables profiling the bronchial epithelium to assess molecular changes occurring distal to the disease site [18]. By identifying gene expression changes associated with CT features compatible with BE (radiologic BE) in normal-appearing bronchial epithelial cells from participants without physician-diagnosed disease, we identified a 655-gene signature consisting of five co-expression clusters based on hierarchical clustering [19]. Pathway enrichment analysis showed that clusters with decreased expression in BE were linked to cell adhesion and Wnt signalling, while those with increased expression were associated with endopeptidase activity and ciliogenesis. The gene expression pattern of the bronchial signature genes separated participants into three clusters: normal, intermediate and bronchiectatic. The bronchiectatic cluster was enriched for participants with more lobes of radiologic BE and more BE-associated symptoms reported, compared with the normal and intermediate participant clusters. Cell-type deconvolution suggested that radiologic BE in bronchial epithelium is associated with increased proportions of ciliated and deuterosomal cells and a decreased proportion of basal cells.
Our analysis suggests overall similarities between widespread radiologic BE and clinical BE, as supported by the prevalence of participants in cluster 1 meeting both radiologic and clinical criteria for BE. Moreover, in addition to describing gene expression related to biological processes previously associated with BE development (e.g., decrease in cell adhesion and increase of inflammatory processes), our results also highlighted novel mechanisms that may be associated with the initiation of BE, such as increased expression of genes involved in ciliogenesis and decreased expression of genes involved in Wnt signalling pathways [19].
In the present study, we tested the hypothesis that BE-associated molecular changes in bronchial epithelium might extend to nasal epithelium, and sought to validate the bronchial-derived signature of BE in independent samples from the Detection of Early Lung Cancer Among Military Personnel (DECAMP)-1 and DECAMP-2 study cohorts.
Methods
Study participants
Study participants were drawn from the DECAMP-1 and DECAMP-2 studies. The details of the DECAMP study protocols have been published previously [20]. The DECAMP studies recruited from 15 military treatment facilities, Veterans Affairs hospitals and academic centres and focused on researching molecular markers of lung cancer and lung cancer risk and collecting biospecimens such as nasal and bronchial brushings from normal-appearing airway epithelium. DECAMP-1 (NCT01785342) recruited current and former tobacco users with indeterminate pulmonary nodules (7–30 mm), while DECAMP-2 (NCT02504697) included current and former tobacco users at high risk of developing lung cancer (i.e., 20+ pack-year smoking history and either COPD or a family history of lung cancer) eligible for lung cancer screening by chest CT. None of the DECAMP participants reported physician-diagnosed BE at the time of enrolment. Per DECAMP protocol, each participant was asked to complete a Lung Health Questionnaire covering demographics, personal medical history, family medical history, medications, smoking history, alcohol and recreational drug history, and symptom history, such as cough, dyspnoea and sputum production. This study was approved by the Office for Human Research Protections for the United States Department of Defense and the institutional review board of every participating site. All participants provided written informed consent to participate in the study.
Computed tomography protocols
DECAMP-2 used a standardised protocol for image acquisition and reconstruction, while DECAMP-1 CT scans were acquired as part of routine clinical care. For DECAMP-2, volumetric scans were acquired with a low radiation dose helical technique on a minimum 16-slice multidetector scanner. Scans were acquired at 2.5–5 mm and reconstructed into 1.25-mm slice thickness using standard and high spatial frequency convolution kernels.
Radiologic bronchiectasis ascertainment
BE was detected visually by a pulmonologist (A.A. Diaz) with >10 years of experience in lung imaging, blinded to the gene expression profiles and participants’ clinical data. Radiologic BE was defined by the presence of one or more of the following criteria: 1) airway dilation (airway lumen diameter greater than adjacent pulmonary vessel diameter), 2) abnormal airway tapering of any extent (no decrease in or increase in lumen moving from proximal to distal airways) and 3) visualisation of a bronchus within 1 cm of the pleura. The lingula was considered a separate lobe. Widespread radiologic BE was defined as radiologic BE in ≥3 lobes [19].
RNA isolation, sequencing and data preprocessing
Nasal brushings were obtained from participants enrolled in DECAMP-1 and DECAMP-2. Per the DECAMP protocol, two nasal brushings from the inferior turbinate of the left nostril were collected from each participant with the assistance of a nasal speculum [20]. Total RNA was isolated using the miRNeasy Mini Kit (Qiagen, Valencia, CA, USA). RNA integrity was quantified by Agilent Bioanalyzer, and RNA purity was confirmed using a NanoDrop spectrophotometer. Transcript integrity was assessed using RSeQC to calculate the mean transcript integrity number (TIN) for each sample. Samples with mean TIN <60 were excluded from downstream analysis, with an average TIN of 72.6 (±6.1) across included samples. Libraries were generated using the Illumina TruSeq Stranded Total RNA kit and sequenced on Illumina NextSeq 500 instruments (with 50- or 75-base paired-end reads), Illumina HiSeq 2500 instruments (with 75-base paired-end reads) or Illumina NextSeq 2000 instruments (with 100-base paired-end reads) (Illumina, San Diego, CA, USA) to an average read depth of 70 million reads per sample. We developed an automatic pipeline (https://github.com/compbiomed/RNA_Seq) based on the Nextflow framework to obtain the expression levels for each gene [21]. Reads were aligned to the Genome Reference Consortium human build 37 using the Spliced Transcripts Alignment to a Reference (STAR) software [22]. Both gene- and transcript-level counts were calculated using RNA-Seq by Expectation Maximization (RSEM) with the Ensembl version 75 annotation [23].
Bulk RNA sequencing and sample analysis
We had previously deposited nasal gene expression data from 288 DECAMP participants in the US National Center for Biotechnology Information Gene Expression Omnibus (GEO) (series GSE210660) [24]. Of these, 136 were from individuals who had been scored for radiologic BE and who were not included in the BE signature discovery set. We identified 132 additional participants not included in the discovery set from whom we had generated nasal RNA sequencing (RNA-seq) data. We have deposited these data in GEO (GSE302095). Of these participants, 84 had also been scored for radiologic BE. Collectively, we refer to this combined dataset from participants not included in the BE signature discovery set who also had nasal gene expression data and radiologic BE assessment (n=220) as the nasal validation set (supplementary figure 1). The matrix of aligned counts per gene was filtered to remove lowly expressed genes that did not have a count per million of >1 in at least 10% of samples and non-protein-coding genes. Surrogate variable analysis was used to correct for heterogeneity in the data [25]. There were 17 surrogate variables, of which the first five were used as covariates [26].
Linear modelling was performed with limma [27] (version 3.10) in R (version 3.6.0) to assess associations between each gene's nasal expression level and the presence of widespread radiologic BE while controlling for the expression of the first five surrogate variables. We ranked genes by moderated t-statistic (from most increased in the nasal epithelium of individuals with widespread radiologic BE to most decreased) and examined the enrichment of different gene sets within the tails of this ranked list using gene set enrichment analysis (GSEA). Heat maps were used to visualise gene expression data. Hierarchical clustering was done using the “Ward.D2” algorithm. The average silhouette method [28] was used to identify the optimal number of sample clusters. For pathway enrichment analysis, GSEA was performed on ranked gene lists described above using Gene Ontology (GO) [29], Kyoto Encyclopedia of Genes and Genomes (KEGG) [30] and Reactome [31] gene sets.
To assess changes in the cellular composition of nasal epithelium associated with radiologic BE, we used GSEA to evaluate the enrichment of sets of cell-type marker genes reported by Deprez et al. [32] within the nasal BE ranked list described above. For cell-type deconvolution, reference gene expression profiles were obtained from a single-cell RNA-seq dataset of 12 534 nasal epithelial cells from four healthy volunteers [32]. Cells were filtered out if they met any of the following criteria: 1) bottom quartile for a total number of genes detected, 2) bottom quartile for total library size or 3) >30% of counts mapped to the mitochondrial genome. 7892 cells were kept for further analysis. Cell types were previously assigned and individually validated with reported cell markers [32]. The AutoGeneS package was used to identify the top 1000 most informative genes from the reference dataset before applying the deconvolve function to estimate cell proportion from the bulk RNA-seq nasal data with model parameter set to Nu Support Vector Regression (nusvr) [33].
Results
Participant demographics, pulmonary function and imaging measurements
Of the 220 participants screened in the nasal validation set, 52 had radiologic BE in at least one lobe and 16 had widespread radiologic BE (BE in ≥3 lobes; supplementary table 1). The participants in the discovery set used in Xu et al. [19] (n=173) were compared with those in the validation set; the average number of lobes with radiologic BE was lower (p<0.001) in the validation set, and fewer participants in the validation set had any radiologic BE reported (p<0.001). Among patients with any radiologic BE, the average number of affected lobes was similar in the discovery and validation sets (p=0.2) (table 1). Despite this, the prevalence of widespread BE in the validation set was not significantly different from that in the discovery set. The validation set had significantly lower rates of current smoking than the discovery set (p=0.001). Pulmonary function was similar between the discovery and validation sets.
TABLE 1.
Comparison of clinical features between participants in the nasal validation set (n=220) and bronchial discovery set (n=173)
| Bronchial discovery set (n=173) | Nasal validation set (n=220) | p-value | |
|---|---|---|---|
| Any BE | 96 (55) | 52 (24) | <0.001 |
| Widespread BE (≥3 lobes) | 20 (12) | 16 (7.3) | 0.14 |
| Average number of lobes with BE, mean (sd) | 1.06 (1.16) | 0.54 (1.18) | <0.001 |
| Male | 141 (82) | 154 (70) | 0.01 |
| Ethnicity | 0.08 | ||
| Asian | 5 (3) | 5 (2) | |
| Black | 21 (12) | 41 (19) | |
| Other/unknown | 18 (10) | 11 (5) | |
| White | 129 (75) | 163 (74) | |
| Smoking status | 0.089 | ||
| Current tobacco user | 75 (43) | 71 (37) | |
| Former tobacco user | 90 (52) | 123 (63) | |
| FEV1 % predicted, mean (sd) | 71 (18) | 66 (18) | 0.090 |
| FEV1/FVC, mean (sd) | 0.6 (0.13) | 0.59 (0.14) | 0.71 |
| Cough, yes | 77 (44) | 101 (46) | 0.74 |
| Phlegm, yes | 76 (44) | 99 (45) | 0.55 |
| Shortness of breath, yes | 98 (62) | 132 (60) | 0.29 |
| DECAMP cohort | <0.001 | ||
| DECAMP1 | 129 (75) | 84 (38) | |
| DECAMP2 | 44 (25) | 136 (62) | |
| COPD status | 0.1 | ||
| Yes | 119 (69) | 93 (42) | |
| No | 46 (26) | 55 (25) | |
| Unknown | 8 (5) | 72 (33) |
Data are presented as n (%) unless otherwise stated. BE: bronchiectasis; FEV1: forced expiratory volume in 1 s; FVC: forced vital capacity; DECAMP: Detection of Early Lung Cancer Among Military Personnel.
Genes significantly altered in the bronchial epithelium are among the genes most altered in the nasal epithelium
No genes were differentially expressed in the nasal epithelium between participants with and those without widespread radiologic BE at a false discovery rate threshold of <0.1. Similarly to our previously reported findings in the bronchial epithelium, pathway enrichment revealed upregulation of cilium-related pathways and immune dysregulation in widespread radiologic BE (supplementary table 2 and supplementary material). With the exception of olfactory-related pathways enriched in the nasal epithelium, we did not observe any strongly distinct pathways enriched in nasal epithelium that were not present in the bronchial epithelium.
To explore the relationship between nasal and bronchial epithelial gene expression alterations in BE, we performed GSEA using gene sets from the bronchial gene signature of BE previously described by Xu et al. [19]. We observed that genes in Xu et al.'s bronchial signature that had increased expression in individuals with widespread radiologic BE were significantly enriched among the genes most increased in the nasal epithelium of individuals with widespread radiologic BE (figure 1a; p<1×10−6). Similarly, genes whose expression was decreased in widespread radiologic BE in Xu et al.'s bronchial signature were significantly enriched among genes most decreased in the nasal epithelium of individuals with widespread radiologic BE (figure 1b; p<1×10−7). This pattern extended to each of the five gene clusters from Xu et al.'s bronchial BE signature (figure 1).
FIGURE 1.
Genes with bronchiectasis (BE)-associated gene expression in the bronchial epithelium are among the genes whose expression is most associated with BE in the nasal epithelium. a) Genes previously reported by Xu et al. [19] to have decreased expression in the bronchial epithelium of individuals with widespread radiologic BE are significantly enriched among the genes most decreased in the nasal epithelium of individuals with widespread radiologic BE. This is true both for the decreased genes as a whole and for the two decreased gene sub-clusters identified by Xu et al. (gene sets A and B). b) Genes previously reported by Xu et al. to have increased expression in the bronchial epithelium of individuals with widespread radiologic BE are significantly enriched among the genes most increased in the nasal epithelium of individuals with widespread radiologic BE. This is true both for the increased genes as a whole and for the three increased gene sub-clusters identified by Xu et al. (gene sets C, D and E). In both plots, genes are ranked (left to right) from the most positive moderated t-statistic for BE to the most negative. The vertical lines represent the position of the genes in each gene set within these ranked lists, and the y-axis represents the running gene set enrichment analysis enrichment score. NES: normalised enrichment score.
We further evaluated enrichment of bronchial BE signature genes in nasal epithelium across additional stratifications of BE severity. Compared with limited BE (one or two lobes with BE), participants with widespread radiologic BE showed no significant enrichment of up- or downregulated bronchial BE genes (supplementary figure 2A). In contrast, participants having any degree of radiologic BE (≥1 lobe) versus those without BE exhibited significant enrichment of both up- and downregulated bronchial BE genes (adjusted p<1×10−5), while participants with limited radiologic BE (one or two lobes with BE) versus those without BE showed significant enrichment of downregulated bronchial BE genes only (supplementary figure 2B and C).
Identification of two participant clusters based on gene expression profiles in the nasal validation set
From the signatures of genes increased in the bronchial epithelium of individuals with widespread BE or the decreased signature, we used the expression of the 25 genes that were the most significantly differentially expressed from each signature with respect to widespread radiologic BE in the nasal epithelial samples to divide the research participants into two clusters (figure 2a). We chose two clusters based on the average silhouette method [28] (supplementary figure 3). We compared the expression of these genes to their expression in the bronchial discovery set, providing a direct comparison of gene expression patterns between the bronchial and nasal samples (figure 2b).
FIGURE 2.
Clustering based on nasal bronchiectasis (BE)-associated gene expression identifies two clusters of participants that differ with regard to radiologic BE prevalence. a) From the bronchial BE up and down signatures, we selected the 25 genes most significantly differentially expressed with respect to widespread radiologic BE in nasal epithelium in each signature to divide the research participants into two groups using hierarchical clustering. The smaller of these participant clusters (cluster 1, n=68), which is characterised by increased nasal expression of bronchial BE “up” genes and decreased expression of bronchial BE “down” genes, is significantly enriched for participants with bronchiectasis (p<0.001; Fisher's exact test). Bronchial BE “down” genes were previously divided into a cluster enriched for genes involved in cell adhesion (gene set A) and another enriched for genes related to Wnt signalling pathways (gene set B). Bronchial BE “up” genes were previously into three clusters: two enriched for genes involved in endopeptidase activity (gene sets C and E) and another enriched for genes related to cilium biology (gene set D). The expression values are z-score normalised, with blue, white and red indicating z-scores of ≤−3, 0 and ≥3, respectively. b) Expression of the same 50 genes in the bronchial discovery set (n=173). Bronchial participant clusters 1, 2 and 3 correspond to the bronchiectatic, normal and intermediate clusters reported in Xu et al. [19] c) Principal component analysis of the participants in the nasal validation set, based on the nasal BE-associated gene expression shown in figure 2a, demonstrates the separation of the two participant clusters. Ellipses indicate 95% confidence intervals for the centroid of each cluster. PC: participant cluster.
To assess the concordance of gene expression changes between bronchial and nasal epithelia, we compared the t-statistics of these 50 genes for the comparison of expression levels in participants with and those without widespread BE in nasal epithelium and the t-statistics from the comparison in bronchial epithelium. There was a high degree of correlation (Pearson's ρ=0.6, p<5×10−5), and a high degree of concordance in the sign of the t-statistics (p<1×10−9; Fisher's exact test), indicating that expression changes tend to occur in the same direction between bronchial and nasal sites (supplementary figure 4A). We next examined the correlation of log2 fold change values for these genes (supplementary figure 4B), which were positively correlated between bronchial and nasal samples (Pearson's ρ=0.4; p<0.005), suggesting that nasal expression changes mirror bronchial changes but with smaller effect sizes. Finally, comparison of the standard deviation for these genes across tissues revealed no significant difference (supplementary figure 4C).
We termed the clusters generated in the nasal validation set as “cluster 1” and “cluster 2” (figure 2a). The predominant cluster (n=152, “cluster 2”) was primarily composed of patients without widespread radiologic BE. The other cluster (n=68, “cluster 1”) had a pattern of relative gene expression similar to that of the bronchial bronchiectatic cluster (increased expression of bronchial “up” genes and decreased expression of bronchial “down” genes) and included most of the patients with widespread radiologic BE (p<0.001; Fisher's exact test). Cluster 1 had a higher average number of lobes with BE (p<0.001), a higher prevalence of any radiologic BE (p<0.001) and a higher prevalence of widespread BE (p<0.001) (table 2). The likelihood of having BE-associated symptoms such as cough, dyspnoea or phlegm did not differ significantly between the groups. Pulmonary function and smoking status were also similar between the participant clusters defined by nasal gene expression.
TABLE 2.
Clinical features of participants in the nasal validation set across expression-derived clusters
| Participant cluster 1 (n=69) | Participant cluster 2 (n=151) | p-value | |
|---|---|---|---|
| Any BE | 26 (38) | 26 (17) | <0.001 |
| Widespread BE (≥3 lobes) | 12 (17) | 4 (2.6) | <0.001 |
| Average number of lobes with BE, mean (sd) | 0.99 (1.63) | 0.34 (0.83) | <0.001 |
| Male | 48 (70) | 106 (70) | 0.64 |
| Ethnicity | 0.48 | ||
| Asian | 1 (1) | 4 (3) | |
| Black | 17 (25) | 24 (16) | |
| Other/unknown | 3 (4) | 8 (5) | |
| White | 48 (70) | 115 (76) | |
| Smoking status | 0.99 | ||
| Current tobacco user | 23 (37) | 48 (37) | |
| Former tobacco user | 40 (63) | 83 (63) | |
| FEV1 % predicted, mean (sd) | 68 (18) | 65 (19) | 0.5 |
| FEV1/FVC, mean (sd) | 0.60 (0.14) | 0.59 (0.13) | 0.7 |
| Cough, yes | 30 (43) | 71 (47) | 0.80 |
| Phlegm, yes | 27 (39) | 72 (48) | 0.39 |
| Shortness of breath, yes | 41 (59) | 91 (60) | 0.48 |
| DECAMP cohort | 0.43 | ||
| DECAMP1 | 29 (42) | 55 (36) | |
| DECAMP2 | 40 (58) | 96 (64) | |
| COPD status | 0.65 | ||
| Yes | 28 (41) | 65 (43) | |
| No | 20 (29) | 35 (23) | |
| Unknown | 21 (30) | 51 (34) |
Data are presented as n (%) unless otherwise stated. BE: bronchiectasis; FEV1: forced expiratory volume in 1 s; FVC: forced vital capacity; DECAMP: Detection of Early Lung Cancer Among Military Personnel.
Because we did not observe significant differences in the likelihood of BE-associated symptoms between the two clusters identified from the nasal validation set, we next investigated the bronchial BE signature using all available samples. We expanded the dataset to include both the nasal validation set and those participants from the Xu et al. bronchial discovery set having matched nasal RNA-seq data (n=299). From this larger sample, we identified two participant clusters (n=101 and n=198) using the same clustering strategy described previously (supplementary figure 5). Similarly to the nasal validation set, the participant clusters differed in whether or not they had any BE reported (p<0.05), the average number of lobes with BE (p<0.005) and the presence of widespread BE (p<0.05) (supplementary table 3). However, as with the validation set, in the expanded dataset, pulmonary function measurements, smoking status and the likelihood of having BE-associated symptoms were similar between participant clusters.
Characterising differences in the cellular composition of nasal epithelium, as compared to bronchial epithelium, in bronchiectasis
Xu et al. reported significant differences in the estimated cell-type proportions of basal, deuterosomal and multiciliated cells when comparing participant clusters defined by bronchial BE signature expression [19]. To test whether these same cell-type composition changes are present in nasal epithelium, we leveraged the same signatures from Deprez et al. [32] for basal, deuterosomal and multiciliated cells that were used in Xu et al. [19]. Using the nasal gene expression data from both the validation set (n=220) and expanded dataset (n=299), we performed GSEA to assess whether these cell-type signatures were enriched among genes most altered in the nasal epithelium of individuals with widespread radiologic BE. Deuterosomal and multiciliated cell signature genes are significantly enriched among the genes most increased in the nasal epithelium of individuals with widespread radiologic BE, while the basal cell signature genes are significantly enriched among genes most decreased (figure 3a and b).
FIGURE 3.
Cell-type marker gene expression and deconvolution suggest similar bronchiectasis (BE)-associated changes in the composition of bronchial and nasal epithelia. a) Marker genes for multiciliated and deuterosomal cells (as defined by Deprez et al. [32]) are enriched among the genes whose expression is most increased in the nasal epithelium of individuals with widespread radiologic BE in the nasal validation set (n=220). In contrast, marker genes for basal cells are enriched among the genes most decreased. This is similar to the enrichment patterns seen in BE-associated bronchial epithelial gene expression and suggests that the proportion of multiciliated and deuterosomal cells is increased in the nasal epithelium of individuals with widespread radiologic BE while the proportion of basal cells is decreased. Genes are ranked, from left to right, from the most positive moderated t-statistic for BE to the most negative, as in figure 1. The vertical lines represent the position of the genes in each gene set within these ranked lists, and the y-axis represents the running gene set enrichment analysis enrichment score. b) Marker genes for multiciliated and deuterosomal cells are similarly enriched among the genes whose expression is most increased with widespread radiologic BE in the expanded nasal dataset (n=299), while basal cell marker genes are enriched among the genes most decreased. These enrichment patterns mirror those observed in the nasal validation set (n=220). c) Box plots of cell-type proportions estimated by AutoGeneS [33] in bulk RNA-seq data from the expanded nasal set (red; n=299) and bronchial discovery set (blue; n=173) epithelium. For each dataset, participants were divided into clusters on the basis of BE-associated gene expression. The bronchial clustering and cell-type proportion estimations are the same as what was previously reported by the Xu et al. [19]. Cell proportion differences between the clusters were determined by Mann–Whitney (nasal) or Kruskal–Wallis (bronchial) test. NES: normalised enrichment score. *: p<0.05; **: p<0.01; ***: p<0.001; ****: p<0.0001.
We also estimated the cellular composition of each sample using the cell-type deconvolution strategy described in Xu et al. [19]. When we compared the estimated cell-type proportions between clusters in the nasal validation set (n=220), we did not observe significant differences for all of the same cell-types (supplementary figure 6). For this reason, we expanded our sample to also include those participants from the Xu et al. bronchial discovery set that had matched nasal RNA-seq data (n=299). We next compared the estimated cell-type proportions for the clusters identified in the expanded nasal validation set. The nasal epithelium from individuals in the expanded participant cluster 1 demonstrates an increase in the predicted proportions of deuterosomal and multiciliated cells (p<0.05). In contrast, the predicted proportion of basal cells decreases (p<0.05). These predicted changes in the cellular composition of the nasal epithelium are similar to the differences observed in the bronchial epithelium [19] (figure 3c).
Discussion
We were interested in determining the effect of BE on nasal epithelial gene expression, as these samples are noninvasive and easy to obtain. Our group previously described a signature of 655 genes whose expression is altered in the bronchial epithelium of individuals with widespread radiologic BE [19]. We hypothesised that these transcriptomic alterations might be shared between the bronchial and nasal epithelia due to an airway-wide “field of injury” effect [18].
In our present study, we validated the bronchial BE signature in nasal epithelium using an independent dataset of 220 participants drawn from the same clinical cohorts as the discovery set. Participants in the nasal validation set had rates of widespread radiologic BE (≥3 affected lobes) that were similar to those of participants in the bronchial discovery set, but they had significantly lower rates of any radiologic BE (≥1 affected lobe). We also found no association between widespread radiologic BE and shortness of breath in the validation set, in contrast to the discovery set, where these variables were positively correlated.
Our results demonstrate that the genes in the previously identified bronchial BE signature are significantly and concordantly enriched among the genes whose expression in the nasal epithelium is most strongly associated with widespread radiologic BE. While Xu et al. found 655 genes whose expression was significantly associated with widespread radiologic BE (false discovery rate q<0.1, |log2 fold change|>0.25) in bronchial epithelium in a discovery set from 173 participants, we did not detect genes whose nasal expression pattern is significantly associated with widespread radiologic BE in our larger dataset derived from 220 participants using the same criteria (data not shown). Our findings suggest that the absence of individually significant nasal genes reflects not only greater variability within the nasal epithelium but also attenuated BE-associated changes. We propose two hypotheses for why BE-associated genes may be less affected in the nasal epithelium than in the bronchus: 1) the nasal epithelium is anatomically more distant from the site of pathology, leading to weaker transcriptional effects, and 2) if BE-associated gene expression turns out to be enriched in specific cell types, there is a potential that these cell types might be less prevalent in nasal epithelium than in bronchial epithelium. Additional studies examining single-cell RNA-seq of both tissues in individuals with and individuals without BE may be helpful to explore these hypotheses.
Xu et al. [19] demonstrated that bronchial gene expression changes associated with BE align with current models of its pathogenesis. These findings revealed reduced expression of cell adhesion genes and elevated expression of inflammatory genes, consistent with the airway dilation and recurrent inflammation seen in early BE. Increased expression of genes related to ciliogenesis was also observed, suggesting a compensatory response to epithelial injury. In contrast, genes enriched for Wnt signalling pathways showed decreased expression, potentially indicating biological ageing and an associated increased risk of BE. In the present study, pathway enrichment analysis of the nasal epithelium using GO, KEGG and Reactome gene sets supported these findings, demonstrating increased expression of genes related to ciliogenesis and downregulation of immune-related pathways associated with widespread radiologic BE.
Based on the nasal expression levels of the top bronchial BE-associated genes, the participants were divided into two clusters. Participants in the cluster with the pattern of relative gene expression like the “bronchiectatic” cluster defined by bronchial gene expression had a higher radiologic BE burden, but we did not observe significant differences in clinical symptoms such as cough, phlegm or dyspnoea between the two clusters. This is in contrast to differences in symptom frequency between the participant clusters identified in Xu et al. [19] based on bronchial BE-associated gene expression. In addition, an intermediate BE participant cluster was identified in the previous bronchial BE study. Importantly, this intermediate cluster differed from the normal participant cluster by its higher proportion of current tobacco users (p<0.01), but otherwise was noted to have clinical characteristics similar to those of the normal participant cluster, potentially highlighting the elevated risk of BE among those who smoke. In our analysis of the nasal validation dataset, we did not identify an intermediate BE cluster. One possible explanation for the lack of association between BE-related expression and clinical features as well as the absence of an intermediate BE cluster in the nasal validation set is the lower burden of radiologic BE in the nasal dataset, compared with the bronchial dataset. Future studies with larger cohorts and greater representation of individuals with radiologic BE may help validate these observed differences between nasal and bronchial BE-associated gene expression.
The enrichment of cell-type specific signatures derived from single-cell sequencing experiments among the genes whose expression is altered by radiologic BE in nasal epithelium suggests that similar changes in the proportion of deuterosomal, multiciliated and basal cells occur in both bronchial and nasal epithelia. These patterns were consistent in the nasal validation set, which served as an independent cohort, and also in the expanded nasal dataset, which included a larger sample and provided greater statistical power. These results were also supported by our cell-type deconvolution analysis, suggesting that BE-associated changes in these cell populations extend into the upper airways. We previously hypothesised that an increase in multiciliated and deuterosomal cell-related gene expression and a decrease in basal cell-related gene expression in the bronchial epithelium might reflect a response to ciliated cell damage; if this hypothesis is correct, our present finding suggests that such damage extends into the upper airway. Future studies that integrate transcriptomic profiling alongside histological immunostaining of nasal biopsy samples will be needed to validate these cell-type associations and strengthen the biological interpretation of these findings.
An important limitation of the current study is that, while it validates BE-related gene expression in independent samples, these validation samples are from the same cohort in which the bronchial BE-related gene expression signature was originally derived. Additional validation studies in independent cohorts are therefore required to demonstrate that these gene expression differences can be detected more broadly. Furthermore, our study is also limited by the relatively low number of radiologic BE cases in the validation set and did not include ascertainment of variables required to determine BE severity index scores. Moreover, we did not observe statistically significant differential nasal epithelial gene expression between radiologic BE and non-BE participants. This limited our ability to explore the relationship between smoking and other comorbid respiratory disease and BE-related gene expression and to understand the relationship between BE-related gene expression and BE severity. It will be especially interesting to examine BE-related gene expression in individuals capturing the broader spectrum of BE, including those with clinically diagnosed BE, those with severe clinical BE and individuals with less extensive history of tobacco smoke exposure.
Although no individual genes reached significance for BE-related differences in the nasal epithelium, and the two sites are not interchangeable, genes associated with widespread radiologic BE in the bronchial epithelium were among those most significantly altered in the nasal epithelium. This suggests that the pattern of bronchial epithelial disease-associated gene expression extends to nasal epithelium, supporting the concept of an airway-wide “field of injury”, whereby the upper airway mirrors pathophysiologic changes occurring in the lower airway. This phenomenon has been similarly observed in COPD [34] and lung cancer [35]. While our data suggest that the impact of radiologic BE on nasal gene expression is less dramatic than it is for bronchial gene expression, we predict that radiologic BE alters the cellular composition of both compartments in a similar manner. While the gene expression signal for BE is stronger in the bronchial epithelium than in nasal brush samples, bronchial brush samples are more difficult to obtain than nasal brush samples. Our findings leave open the question of whether the gene expression changes observed in nasal epithelium can be combined to create a biomarker that would have utility for noninvasive screening or longitudinal monitoring of BE.
Footnotes
Provenance: Submitted article, peer reviewed.
Conflict of interest: W. Souery reports support for the present study from the National Institutes of Health (NIH) National Heart, Lung, and Blood Institute (NHLBI) (T32HL007035). J. Wong, K. Xu, X. Xiao, H. Liu, G. Liu, A.C. Gower and Y.O. Alekseyev have nothing to disclose. A.A. Diaz reports support for the present study from NHLBI (R01-HL149861, R01-HL164824 and R01-HL173017); payment or honoraria for lectures, presentations, manuscript writing or educational events from Zambon Pharmaceutical; participation in a data safety monitoring board or advisory board with Sanofi–Regeneron and Verona Pharma; a leadership role with Sanofi; and patents planned, issued or pending (USPTO patent number 11,946,928 B2 “Methods and compositions relating to airway dysfunction”). E. Billatos reports support for the present study from the Department of Defense (W81XWH-11-2-0161), the National Cancer Institute (U01CA196408), Johnson & Johnson Services, Inc., NIH/NHLBI (R01HL149861) and Novartis Biomedical Research; and grants from Janssen Research & Development, LLC, Johnson & Johnson Enterprise Innovation, Inc. (project number: 1258521), the NIH National Cancer Institute (5U2CCA271898) and Veracyte (DHF 009-050P). M.E. Lenburg reports support for the present study from Johnson & Johnson, Novartis Biomedical Research, NIH (U01CA196408, U01HL146408 and R01HL122477) and the Department of Defense (W81XWH-11-2-0161); grants from AstraZeneca and NIH (U2CCA271898); and patents planned, issued or pending (Boston University – gene expression signature of bronchiectasis and gene expression signature of lung function decline).
Support statement: This work was supported by the Department of Defense (W81XWH-11-2-0161), the National Institutes of Health (T32HL007035, U01CA196408, U01HL146408, R01HL122477 and R01HL149861), Johnson & Johnson and Novartis Biomedical Research. Funding information for this article has been deposited with the Open Funder Registry.
Supplementary material
Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.
Figure S1
00814-2025.SUPPLEMENT
Figure S2
00814-2025.SUPPLEMENT2
Figure S3
00814-2025.SUPPLEMENT3
Figure S4
00814-2025.SUPPLEMENT4
Figure S5
00814-2025.SUPPLEMENT5
Figure S6
00814-2025.SUPPLEMENT6
Supplementary tables
00814-2025.SUPPLEMENT7
Supplementary material
00814-2025.SUPPLEMENT
References
- 1.Chalmers JD, Aliberti S, Blasi F. Management of bronchiectasis in adults. Eur Respir J 2015; 45: 1446–1462. doi: 10.1183/09031936.00119114 [DOI] [PubMed] [Google Scholar]
- 2.Polverino E, Dimakou K, Hurst J, et al. The overlap between bronchiectasis and chronic airway diseases: state of the art and future directions. Eur Respir J 2018; 52: 1800328. doi: 10.1183/13993003.00328-2018 [DOI] [PubMed] [Google Scholar]
- 3.Traversi L, Miravitlles M, Martinez-Garcia MA, et al. ROSE: radiology, obstruction, symptoms and exposure – a Delphi consensus definition of the association of COPD and bronchiectasis by the EMBARC Airways Working Group. ERJ Open Res 2021; 7: 00399-2021. doi: 10.1183/23120541.00399-2021 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Diaz AA, Nardelli P, Wang W, et al. Artificial intelligence-based CT assessment of bronchiectasis: the COPDGene study. Radiology 2022; 307: e221109. doi: 10.1148/radiol.221109 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Diaz AA, Wang W, Orejas JL, et al. Suspected bronchiectasis and mortality in adults with a history of smoking who have normal and impaired lung function: a cohort study. Ann Intern Med 2023; 176: 1340–1348. doi: 10.7326/M23-1125 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.King PT, Holdsworth SR, Freezer NJ, et al. Characterisation of the onset and presenting clinical features of adult bronchiectasis. Respir Med 2006; 100: 2183–2189. doi: 10.1016/j.rmed.2006.03.012 [DOI] [PubMed] [Google Scholar]
- 7.Morrissey BM, Evans SJ. Severe bronchiectasis. Clin Rev Allergy Immunol 2003; 25: 233–247. doi: 10.1385/CRIAI:25:3:233 [DOI] [PubMed] [Google Scholar]
- 8.Tiddens HAWM, Meerburg JJ, van der Eerden MM, et al. The radiological diagnosis of bronchiectasis: what's in a name? Eur Respir Rev 2020; 29: 190120. doi: 10.1183/16000617.0120-2019 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Hill AT, Sullivan AL, Chalmers JD, et al. British Thoracic Society Guideline for bronchiectasis in adults. Thorax 2019; 74: 1–69. doi: 10.1136/thoraxjnl-2018-212463 [DOI] [PubMed] [Google Scholar]
- 10.Weycker D, Hansen GL, Seifer FD. Prevalence and incidence of noncystic fibrosis bronchiectasis among US adults in 2013. Chron Respir Dis 2017; 14: 377–384. doi: 10.1177/1479972317709649 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.ten Hacken NH, Wijkstra PJ, Kerstjens HA. Treatment of bronchiectasis in adults. BMJ 2007; 335: 1089–1093. doi: 10.1136/bmj.39384.657118.80 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Beane J, Vick J, Schembri F, et al. Characterizing the impact of smoking and lung cancer on the airway transcriptome using RNA-Seq. Cancer Prev Res (Phila) 2011; 4: 803–817. doi: 10.1158/1940-6207.CAPR-11-0212 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Billatos E, Vick JL, Lenburg ME, et al. The airway transcriptome as a biomarker for early lung cancer detection. Clin Cancer Res 2018; 24: 2984–2992. doi: 10.1158/1078-0432.CCR-16-3187 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Gower AC, Steiling K, Brothers JF, et al. Transcriptomic studies of the airway field of injury associated with smoking-related lung disease. Proc Am Thorac Soc 2011; 8: 173–179. doi: 10.1513/pats.201011-066MS [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Steiling K, Lenburg ME, Spira A. Personalized management of chronic obstructive pulmonary disease via transcriptomic profiling of the airway and lung. Ann Am Thorac Soc 2013; 10 Suppl: S190–S196. doi: 10.1513/AnnalsATS.201306-190AW [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Campbell JD, McDonough JE, Zeskind JE, et al. A gene expression signature of emphysema-related lung destruction and its reversal by the tripeptide GHK. Genome Med 2012; 4: 67. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Menon AA, Lee M, Ke X, et al. Bronchial epithelial gene expression and interstitial lung abnormalities. Respir Res 2023; 24: 245. doi: 10.1186/s12931-023-02536-w [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Steiling K, Ryan J, Brody JS, et al. The field of tissue injury in the lung and airway. Cancer Prev Res (Phila) 2008; 1: 396–403. doi: 10.1158/1940-6207.CAPR-08-0174 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Xu K, Diaz AA, Duan F, et al. Bronchial gene expression alterations associated with radiological bronchiectasis. Eur Respir J 2023; 61: 2200120. doi: 10.1183/13993003.00120-2022 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Billatos E, Duan F, Moses E, et al. Detection of Early Lung Cancer Among Military Personnel (DECAMP) consortium: study protocols. BMC Pulm Med 2019; 19: 59. doi: 10.1186/s12890-019-0825-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Di Tommaso P, Chatzou M, Floden EW, et al. Nextflow enables reproducible computational workflows. Nat Biotechnol 2017; 35: 316–319. doi: 10.1038/nbt.3820 [DOI] [PubMed] [Google Scholar]
- 22.Dobin A, Davis CA, Schlesinger F, et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 2013; 29: 15–21. doi: 10.1093/bioinformatics/bts635 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Li B, Dewey CN. RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome. BMC Bioinformatics 2011; 12: 323. doi: 10.1186/1471-2105-12-323 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Xu K, Shi X, Husted C, et al. Smoking modulates different secretory subpopulations expressing SARS-CoV-2 entry genes in the nasal and bronchial airways. Sci Rep 2022; 12: 18168. doi: 10.1038/s41598-022-17832-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Leek JT, Storey JD. Capturing heterogeneity in gene expression studies by surrogate variable analysis. PLoS Genet 2007; 3: 1724–1735. doi: 10.1371/journal.pgen.0030161 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Wang L, Nie J, Sicotte H, et al. Measure transcript integrity using RNA-seq data. BMC Bioinformatics 2016; 17: 58. doi: 10.1186/s12859-016-0922-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Ritchie ME, Phipson B, Wu D, et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 2015; 43: e47. doi: 10.1093/nar/gkv007 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Lengyel A, Botta-Dukát Z. Silhouette width using generalized mean – a flexible method for assessing clustering efficiency. Ecol Evol 2019; 9: 13231–13243. doi: 10.1002/ece3.5774 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Ashburner M, Ball CA, Blake JA, et al. Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nat Genet 2000; 25: 25–29. doi: 10.1038/75556 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Kanehisa M, Sato Y, Furumichi M, et al. New approach for understanding genome variations in KEGG. Nucleic Acids Res 2019; 47: D590–D595. doi: 10.1093/nar/gky962 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Milacic M, Beavers D, Conley P, et al. The Reactome pathway knowledgebase 2024. Nucleic Acids Res 2024; 52: D672–D678. doi: 10.1093/nar/gkad1025 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Deprez M, Zaragosi LE, Truchi M, et al. A single-cell atlas of the human healthy airways. Am J Respir Crit Care Med 2020; 202: 1636–1645. doi: 10.1164/rccm.201911-2199OC [DOI] [PubMed] [Google Scholar]
- 33.Aliee H, Theis FJ. AutoGeneS: automatic gene selection using multi-objective optimization for RNA-seq deconvolution. Cell Syst 2021; 12: 706–715.e4. doi: 10.1016/j.cels.2021.05.006 [DOI] [PubMed] [Google Scholar]
- 34.Boudewijn IM, Faiz A, Steiling K, et al. Nasal gene expression differentiates COPD from controls and overlaps bronchial gene expression. Respir Res 2017; 18: 213. doi: 10.1186/s12931-017-0696-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.AEGIS Study Team . Shared gene expression alterations in nasal and bronchial epithelium for lung cancer detection. J Natl Cancer Inst 2017; 109: djw327. doi: 10.1093/jnci/djw327 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.
Figure S1
00814-2025.SUPPLEMENT
Figure S2
00814-2025.SUPPLEMENT2
Figure S3
00814-2025.SUPPLEMENT3
Figure S4
00814-2025.SUPPLEMENT4
Figure S5
00814-2025.SUPPLEMENT5
Figure S6
00814-2025.SUPPLEMENT6
Supplementary tables
00814-2025.SUPPLEMENT7
Supplementary material
00814-2025.SUPPLEMENT