Transcriptomic signatures of IPF in ALI-cultured airway cells and their therapeutic implications

Idiopathic pulmonary fibrosis (IPF) is an interstitial lung disease characterised by progressive parenchymal scarring and impaired organ function. It carries a poor prognosis with a median survival of 3–5 years and exhibits extensive lung architecture remodelling like fibrotic foci and bronchiolisation. 1 2 IPF is currently understood as alveolar epithelial damage that triggers a dysregulated wound healing response, resulting in fibrotic remodelling. 1 However, emerging evidence indicates that human bronchial epithelial cells (HBECs) play a critical role in IPF pathogenesis. A known genetic risk factor for IPF is a promoter variant in MUC5B , a gene expressed by secretory cells of the airway. 3 In addition, airway-specific features such as bronchiolisation of distal regions and formation of honeycomb cysts underscore epithelial remodelling in IPF. 1 Recent single-cell RNA-sequencing (scRNA-seq) studies have also identified a novel KRT17 + basaloid cell type, with similar transcriptional features detected in bronchial basal cells in another study. 4 – 6 Together, these findings highlight the pathogenic relevance of HBECs in IPF. To further characterise the molecular and cellular roles of HBECs in IPF, we used air–liquid interface (ALI) cultures, a well-established preclinical model that recapitulates the airway mucosa. 7 Importantly, bronchial samples were obtained from treatment-naïve IPF patients at the point of diagnosis, enabling investigation of cellular phenotypes at earlier disease stages. This contrasts with existing scRNA-seq atlases derived from explant lungs, which primarily represent the latest stages of fibrotic disease. We also leveraged the practicality of ALI cultures to investigate the transcriptional effects of two FDA-approved drugs, nintedanib and pirfenidone, 1 alongside saracatinib, a Src kinase inhibitor under clinical evaluation for IPF ( NCT04598919 ). Here, we present a side-by-side comparison of these drugs at single-cell resolution in HBECs and fibroblasts, providing essential insights into their cellular effects and modes of action. Our analysis identifies transcriptional programmes defining primed fibroblasts, dysregulated basal cells and proinflammatory HBECs in IPF. Through comparative treatment analyses and experiments, we highlight the potential of saracatinib as an alternative therapeutic candidate. By understanding the early molecular phenotypes of the airway mucosa in IPF and evaluating their drug responses at single-cell resolution, this study provides insights into airway epithelial dysfunction that contributes to disease progression.

IPF patients were evaluated by the Interstitial Lung Disease Board of the Department of Pulmonary and Respiratory Critical Care Medicine, Thoraxklinik, Heidelberg University Hospital. Cases were enrolled immediately after referral and discussion by the ILD Board. At enrolment, the diagnosis of IPF was still under evaluation, and all patients were treatment naïve. When additional diagnostic work-up was required, patients were included in our study for sample collection, and in some cases, surgical lung biopsies (video-assisted thoracoscopic surgery) were needed to confirm the diagnosis. Controls composed of two sources: (1) patients who underwent diagnostic bronchoscopy to investigate indeterminate pulmonary nodules, with samples taken opposite the lesion site to minimise influence and (2) control lung biopsies obtained from the Tissue Bank of the National Center for Tumor Diseases Heidelberg and those purchased from Tissue Solutions to ensure sufficient sample numbers. No control patient was diagnosed for interstitial lung disease. Patient information is in table 1 . The study has two patient cohorts: (1) n=4 control and n=4 IPF patients for ALI cultures and scRNA-seq and (2) n=4 control and n=5 IPF patients whose lung biopsies were used for validation stainings. All treatment-naïve IPF samples were collected at diagnosis. Samples or tissues were obtained from the Tissue Bank of National Center for Tumor Diseases, Heidelberg (H) or purchased from TS. DLCO, diffusing capacity of lung for carbon monoxide; FEV, forced expiratory volume; FVC, forced vital capacity; IPF, idiopathic pulmonary fibrosis; PY, pack-years; scRNA-seq, single-cell RNA-sequencing; TS, tissue solutions. Primary HBECs were obtained from the endobronchial lining fluid (ELF) of patients by minimally invasive bronchoscopic microsampling from subsegmental bronchi, as previously described. 44 ELF was obtained with a mucus collection probe (Olympus, BC-401C/BC402C). HBECs adhering to the cotton-like probe were released and used as the source material for primary culture. HBECs were cultured following previously published methods. 45 Briefly, cells were grown in bronchial expansion medium: DMEM/F12 (Gibco, 21331020) supplemented with Bovine Pituitary Extract (0.004 mL/mL), Epidermal Growth Factor (10 ng/mL), Insulin (5 µg/mL), Hydrocortisone (0.5 µg/mL), Triiodo‐L‐Thyronine (6.7 ng/mL) and Transferrin (10 µg/mL) (PromoCell, C-39160), 30 nM Sodium Selenite (Merck, S5261), 10 µM Ethanolamine (Merck, E0135), 10 µM Phosphorylethanolamine (Merck, P0503), 0.18 mM Adenine (Merck, A2786), 0.5 µM Sodium Pyruvate (Gibco, 11 360–070), 1X GlutaMAX (Gibco, 35050038), 15 mM HEPES (Gibco, 15 630–056), and in the presence of 10 µM ROCK Inhibitor (StemCell, 72308). After reaching confluence, cells were transferred into 12-well ThinCert cell culture inserts with 0.4 µm pores (Greiner Bio-One, 665641). They were airlifted after 2–3 days by removing the medium from the apical chamber and providing PneumaCult-ALI media (StemCell, 05021) in the basal chamber only. Differentiation into a pseudostratified epithelium was achieved after 24–27 days. For each patient line, three ALI inserts were pooled for each condition to reduce variability. All cultures were harvested at the same time for dissociation. Single-cell suspensions from ALI cultures in 0.05% BSA were passed through a 35 µm filter (Corning, 352235). Samples were prepared with Chromium Single Cell 3’ Library Kit v2 (10×Genomics; PN-120237, PN-120236, PN-120262) to generate cell and gel bead emulsions. The succeeding steps of reverse transcription, cDNA amplification and sequencing library generation were all performed according to the manufacturer’s instructions. Libraries were sequenced with NextSeq500 (Illumina; high-output mode, paired-end 26×49 bp) or HiSeq4000 (Illumina; high-output mode, paired-end 26×74 bp). See online supplemental file 1 for detailed methods and materials ( online supplemental figures S1–S8 ; online supplemental tables S1–S6 ).

To investigate the transcriptomic landscape of IPF airway mucosa and their responses to pharmacological inhibition, we performed scRNA-seq on ALI-cultured cells treated with one of three drugs or vehicle for 24 hours ( figure 1 A). A total of 129 986 transcriptomes were collected from 31 samples across 8 patients, identifying 16 cell types and states of predominantly HBECs ( figure 1 B,C; online supplemental figure S1 A). Although ALI cultures primarily model the airway epithelium, we were able to identify a smaller population of fibroblasts in the scRNA-seq data. Fibroblasts expressed mesenchymal markers PDGFRA and VIM , while HBECs expressed epithelial markers EPCAM and CDH1 ( online supplemental figure S1 B). To validate the identity of fibroblasts in the culture system, control and IPF HBEC lines were later expanded in 2D before undergoing EPCAM antibody-based selection. These EPCAM - fraction-expressed PDGFRA, a canonical fibroblast marker, confirming their identity as fibroblasts (henceforth ‘fibro’; figure 1 D). Further immunofluorescent stainings validated the presence of canonical HBECs like basal, goblet and ciliated cells in the culture system ( figure 1 E). After unsupervised clustering, cell type subclusters were annotated using marker genes and gene ontology (GO) term enrichment ( online supplemental figure S1 ; see online supplemental file 1 ). Fibroblast populations expressed markers like PDGFRA , COL1A2 and FGF7, and showcased states including ER stress, lipid metabolism and mitosis ( online supplemental figure S1 C), confirming the reported complexity of fibroblasts in fibrotic lungs. 8 9 On the other hand, basal cells were divided into three clusters: basal, basal extracellular matrix (ECM) and basal mitotic. Basal cells were involved in biogenesis, while basal ECM cells ( FN1 + ) were linked to wound healing responses ( online supplemental figure S1 C). While trends in cellular composition shifts were seen in IPF ALI cultures like fibroblast expansion, which reflects fibrotic progression, none reached statistical significance ( figure 2 A; online supplemental figure S2 A; online supplemental table S1 ). Epithelial-mesenchymal interactions were shown to be relevant in IPF 5 ; thus, we inferred how these cells from the airway communicated with one another ( figure 2 B,C). Basal ECM had the most numerous interactions in IPF, unlike basal mitotic cells and ionocytes. We prioritised key interaction pathways implicated in fibrosis, including SPP1, THBS and NOTCH, where fibroblast and basal populations were key players ( figure 2 C,D; online supplemental figure S2B). SPP1 (osteopontin) encodes a phosphoprotein involved in wound healing and inflammation that is overexpressed in IPF. 4 5 10 THBS1 (thrombospondin) produces a major activator of TGF-β, the master regulator of fibrosis. 11 Lastly, Notch signalling (represented by its ligand JAG1 ) is a critical developmental pathway influencing cell fate decisions within the airway epithelium. 12 SPP1 was highly upregulated in most IPF fibroblasts and primarily interacted with basal cells, while THBS1 and JAG1 were upregulated in IPF basal and basal ECM cells ( online supplemental figure S2 B). We validated these expression profiles with immunofluorescence staining of lung tissue in an independent cohort of newly diagnosed IPF patients ( figure 2 E–I; online supplemental figure S3 ). SPP1 localised predominantly around fibroblasts, with lower levels in collagen-rich regions but higher expression in surrounding areas ( figure 2 F,G). JAG1 expression was not only increased in IPF but also distinctly marked basal pod structures in tissue ( figure 2 H). Finally, THBS1 expression was detected in basal cells as well as other HBECs in IPF tissues, indicating a broader role in facilitating TGF-β activation and promoting fibrosis ( figure 2 I). To further understand the influence of HBECs on fibroblasts, we performed co-culture experiments and assessed the activation of fibroblasts in vitro ( figure 2 J). In the presence of HBECs, the fibroblasts, regardless of disease, showed a tendency to increase their proliferation rates ( figure 2 K). In terms of activation, the fibroblasts in co-culture conditions showed significantly higher expression of the contractility marker α-smooth muscle actin (α-SMA; figure 2 L). In contrast, collagen production was significantly reduced under co-culture conditions ( figure 2 M). These functional assays reflect that the presence of HBECs can greatly influence the phenotype of fibroblasts, indicating that their presence in our ALI cultures may be more biologically representative of the in vivo state. To identify transcriptional changes across cell types in IPF, differentially expressed genes for each cellular population were calculated ( online supplemental figure S2 C; online supplemental table S2 ). Among basal cells, reparative epithelial KRT6 genes 13 and TGF-β activation mediators, such as THBS1 and integrins, 14 were significantly upregulated, particularly in the basal ECM subset ( online supplemental figure S2 D). Beyond basal cells, we found several genes consistently upregulated in fibroblasts and HBECs, prompting an overlap analysis to determine shared transcriptional programmes between these populations ( online supplemental figure S2 C,E–F). In fibroblasts, the shared signature comprised of genes involved in oxidative phosphorylation (OXPHOS) and a priming signature, which included cytoskeletal genes ACTG2 and ACTBL2 , along with AEBP1 , 15 a known myofibroblast differentiation enhancer ( online supplemental figure S2 E). Conversely, HBECs shared an immune modulation signature composed of proinflammatory chemokines ( CXCL6 and IL8 ) and immunomodulatory genes (SAA and S100A family; online supplemental figure S2 F). Notably, SAA and S100 proteins play a key role in inflammation by recruiting macrophages and neutrophils, thereby potentially contributing to the immune dysregulation seen in IPF. 16 – 18 SAA family proteins were shown to promote the transition of macrophages from their proinflammatory M1 to the tissue reparative M2 state, which is highly involved in fibroproliferative processes. 16 18 We validated the expression of these genes in IPF tissues ( online supplemental figure S2 G–I, online supplemental figure S4 ). Stainings confirmed that S100A8/9 was upregulated in HBECs ( online supplemental figure S2 G). Similarly, while SAA1/2 expression was limited to airways in controls, its expression expanded across IPF tissues ( online supplemental figure S2 H). Furthermore, we observed the previously reported increase in bronchiolised structures in IPF lungs, 19 which further emphasises the relevance of HBECs, as their expansion likely amplifies their effects in the disease ( online supplemental figure S2 I). To characterise this immunomodulatory nature, we measured the levels of proinflammatory cytokines in the ALI cultures and found significantly elevated levels of CXCL8, CCL20 and CXCL5 in IPF ( online supplemental figure S2 J). In contrast, CCL17 and CCL2 showed slight downregulation in IPF. These findings suggest a functional convergence between fibroblast and HBEC populations in IPF, where fibroblasts engage in metabolic reprogramming and priming, while HBECs contribute to a highly inflammatory environment, typically observed to precede progressive fibrosis. 18 Although nintedanib and pirfenidone are currently used to treat IPF, these therapies only slow disease progression. 1 We therefore investigated saracatinib, a Src kinase inhibitor originally developed as an experimental anticancer drug, as a potential alternative therapeutic for IPF. 20 The Src kinase family has been implicated in cellular events relevant to IPF, including cell proliferation, myofibroblast differentiation and even epithelial-to-mesenchymal transition (EMT). 20 – 22 The effects of all three drugs were assessed side by side by treating fully differentiated ALI cultures for 24 hours prior to scRNA-seq. Cluster distributions of cell types remained relatively uniform across treatments, as expected given the short treatment duration of differentiated ALI cultures ( figure 3 A; online supplemental table S1 ). To systematically compare drug responses across cell types, we evaluated the effects of each treatment using a cell-specific IPF signature. Scores for each cell type were generated under different treatments and disease conditions ( figure 3 B; online supplemental tables S3 and S 4 ; see online supplemental file 1 ). Mean score differences were used to assess treatment efficacy, with treatments considered effective when scores shifted toward control levels. The fibroblast populations, exemplified by the fibro subcluster, showed the most significant score shift after treatment. Basal cells also showed significant improvement as seen with the basal ECM subcluster. We next examined the effects of treatments on previously identified disease signatures. In fibroblasts, SPP1 was the only cell-cell interaction gene significantly affected by all three drugs, in contrast to JAG1 and THBS1 in basal cells, which were unaffected ( figure 3 C; online supplemental figure S5 A,C). Similarly, reparative epithelial KRT6 and TGF-β activation genes in basal cells were largely unaffected by the treatments ( online supplemental figure S5 C). In contrast, several shared IPF signatures of fibroblasts and HBECs exhibited significant drug responses with variations in magnitudes ( online supplemental figure S5 B,D). In fibroblasts, the OXPHOS signature was downregulated, most effectively by pirfenidone, while the priming signature was also similarly downregulated, apart from AEBP1 , which remained elevated after pirfenidone or saracatinib treatment. Meanwhile, in HBECs, CXCL6 and SAA4 were mostly unaffected. Together, these results suggest that while current antifibrotic therapies can attenuate specific fibrotic signatures, none of the three drugs fully reversed the IPF signatures we have observed. To identify treatment-specific effects, we first visualised all downregulated genes after treatment, revealing distinct compartmentalised responses in fibroblasts and HBECs ( online supplemental figure S5 E). Overlap analyses showed partially shared responses between treatments, with GO term enrichment of these shared genes showing that fibroblasts responded to antibiotics, heat and unfolded proteins, indicative of a drug response. Meanwhile, HBECs exhibited processes related to RNA splicing and protein translation ( figure 3 D,E). Despite the overlaps, each drug still induced distinct transcriptional effects in fibroblasts and HBECs ( figure 3 F). In fibroblasts, nintedanib downregulated fibrosis-associated genes such as microtubules TUBA1A , matrix-associated protein SPARC and myofibroblast enhancer AEBP1 , which is consistent with its antifibrotic effects. Pirfenidone affected MAP2K2 , a key component of the ERK cascade implicated in lung fibrosis, 23 potentially attenuating FGFR or PDGFR signalling. In contrast, saracatinib broadly suppressed genes involved in core cellular functions like ACTB , NDUFB6 and RPL31 . In HBECs, nintedanib particularly downregulated the signalling factors SFN , CTTN and MDK , while pirfenidone impacted ribosomal proteins RPS12 and RPL29 . Saracatinib uniquely downregulated the expression of integrin ITGB1 and senescence factor CDKN2A , indicating its potential effects on cell-matrix interactions and cellular senescence, respectively. To further characterise the functional effects of the drugs on primary fibroblasts, cells were treated for 48–96 hours and tested for proliferation and activation ( figure 3 G–J; online supplemental figure S5 F,G). Fibroblasts were also stimulated with TGF-β, a master regulator of fibrosis, to test if the drugs could inhibit the activated state. TGF-β stimulation increased cellular proliferation and expression of α-SMA (ACTA2) and COL1A1, showing that these fibroblasts can be activated. Co-treatment of fibroblasts across all assays showed that saracatinib significantly suppressed fibroblast proliferation and activation despite TGF-β stimulation ( figure 3 I,J; online supplemental figure S5 F,G). Strikingly, saracatinib outperforms nintedanib and pirfenidone in reducing cellular proliferation and contraction, but not in COL1A1 production. All three also reduced COL1A1, although to a lesser extent by saracatinib. This superiority of saracatinib in inhibiting fibroblast proliferation and contractility may be attributed to its unique capacity to downregulate contractile actin ACTB . We next assessed the sensitivity of HBECs to the three drugs, as a previous study reported that saracatinib can inhibit bronchial basal cell responses in IPF. 6 Consistent with these findings, HBEC proliferation was significantly reduced by saracatinib, which again demonstrated greater efficacy than nintedanib or pirfenidone ( online supplemental figure S5 H). Together, these results position saracatinib as the most potent inhibitor of fibroblast activation and epithelial proliferation among the drugs tested. Here, we describe the potential mode of action of saracatinib, for the first time to our knowledge, on HBECs and fibroblasts by their expression of genes encoding Src and Abl kinases. We found that FYN was uniquely expressed by fibroblasts, while HBECs expressed SRC , YES1 , LYN , FRK and ABL2 ( figure 3 K). In addition, these genes in HBECs showed higher expression in IPF, most especially ABL2 in basal cells. This may help explain the compartmentalised responses of HBECs and fibroblasts, and the heightened sensitivity of fibroblasts and HBECs to saracatinib than to nintedanib and pirfenidone. To compare the transcriptional programmes of the culture system, which we purport to represent earlier disease states, with later stages, we integrated the ALI culture dataset with two IPF atlases derived from explant lungs, resulting in a dataset of 444 106 transcriptomes ( figure 4 A). 4 5 The combined embeddings of each cellular compartment show an overlap of general cell type identities across datasets ( figure 4 B,C). To account for differences in dataset annotations, we conducted analyses at a broader cell type level ( online supplemental table S5 ). We then interrogated potential cell–cell interactions between the ALI culture HBECs and fibroblasts with other cells from IPF atlases that are absent in the culture system by integrating the missing cell types ( figure 4 D). Fibroblasts had more interactions, while basal cells had less in the IPF atlases. The ALI culture dataset revealed a stronger emphasis on cytokine and chemokine interactions, including SAA, SPP1, CXCL, CCL and IL1, suggesting an inflammatory-driven microenvironment at this stage of the disease ( figure 4 E). In contrast, IPF atlases highlighted growth factor signalling pathways such as FGF, HGF, EGF and IGF. These growth factors are known to be closely linked to fibroblast activation and tissue remodelling. 24 25 We further investigated the expression profile of the cytokine and chemokine ligands between datasets and found their significant upregulation in ALI culture IPF HBECs—again emphasising their previously observed inflammatory nature ( figures 4 F; online supplemental figure S2 F). Notably, IPF monocytes and macrophages from IPF atlases expressed higher levels of SPP1 as compared with ALI culture IPF fibroblasts ( figure 4 F). To further characterise the cellular phenotypes captured in the ALI culture dataset, we compared their transcriptional signatures with those reported in IPF atlases. ALI culture fibroblasts expressed fibroblast markers like S100A4 and SERPINF1, 4 5 while also weakly expressing the contractile marker ACTA2 ( figure 4 G). While most signatures were segregated by dataset or disease stage, a subset of genes that was shared across both datasets, like AEBP1 , PCOLCE and VEGFA ( figure 4 H). Habermann et al , reported the expression of PLIN2 in lipofibroblasts, which we also interestingly observe in the ALI culture dataset. In contrast, EMT markers like TWIST1 and SNAI1 were restricted to the IPF atlases, suggesting a stronger EMT signature at later disease stages. The IPF atlases also identified a basaloid cell type expressing ECM genes characterised as KRT5 − and KRT17 + . 4 5 In comparison, the ALI culture basal cells were KRT5 + and expressed KRT17 at lower levels ( figure 4 G). Similar to fibroblasts, most basal and basaloid signature genes were dataset or disease stage specific. Notably, basal ECM cells expressed high levels of VIM , FN1 , LAMB3 , SOX4 and SOX9 ( figure 4 H), highlighting their aberrant phenotype. To place our ALI culture data in the context of recent scRNA-seq studies of the airway epithelium and fibrosis models, we next integrated it with datasets from Jaeger et al , who profiled bronchial epithelial cells from IPF patients, and Lang et al , who modelled early fibrotic responses ex vivo ( online supplemental figure S6 ). 6 26 Integration with the Jaeger et al dataset showed substantial overlap of the HBECs, particularly among basal cells ( online supplemental figure S6 A,B). Shared overexpressed genes, including KRT6A , KRT17 , LAMC2 , SOX4 , JAG1 , THBS1 , ITGAV and ITGB6 , revealed a consistent aberrant basal cell phenotype across both datasets ( online supplemental figure S6 C). Similarly, integration with the Lang et al dataset showed a general overlap of fibroblast and HBEC populations ( online supplemental figure S6 D E). To quantitatively assess transcriptional similarity, we calculated a cell-specific fibrotic score using the top differentially expressed genes from each cell type in the Lang et al dataset ( online supplemental figure S6 F–H, see online supplemental file 1 ). In this analysis, ALI culture fibroblasts showed substantial divergence, with large effect sizes for fibroblast and myofibroblast fibrotic scores, suggesting an inactive fibroblast state ( online supplemental figure S6 F). In contrast, the HBECs exhibited comparable fibrotic scores and small effect sizes relative to Lang et al , reflecting an overlap in transcriptional signatures during ‘early stage’ fibrosis ( online supplemental figure S6 G,H). Together, these integrative analyses further validate the cellular phenotypes we have captured in the ALI culture system as being at an early stage of IPF disease. To further validate the relevance of these fibroblast and basal cell phenotypes we identified in the ALI culture model, we used RNA fluorescence in situ hybridisation (RNA-FISH) to identify these cells in newly diagnosed IPF patient tissues. We first investigated the presence of SPP1 + fibroblasts ( DCN + ) in relation to the activation marker COL1A1 in tissue, as SPP1 exhibited the strongest signature in the ALI culture IPF fibroblasts ( figure 4 I). While we confirmed the existence of SPP1 + fibroblasts in IPF tissues, we also noticed an inverse relationship between SPP1 and COL1A1 on the transcript level within fibroblasts ( figures 4 I; online supplemental figure S7 A,B). This is akin to the previous observation with immunostainings from figure 2 G. Next, FN1 + basal ECM cells ( TP63 + , KRT5 + ) were identified in IPF tissues neighbouring other cells highly expressing FN1 , which are most likely fibroblasts ( figures 4 J; online supplemental figure S7 C). Interestingly, these cells also formed basal pod structures as presented in figure 2 H. We further investigated this inverse relationship of SPP1 and COL1A1 . Using the scRNA-seq integrated fibroblast subset, a significant negative correlation between SPP1 and COL1A1 expression was observed (Pearson’s r =−0.4319, p=2.2e16). This negative correlation was notably below the fifth percentile of correlation coefficients of SPP1 across all expressed genes in fibroblasts, and could additionally be supported by the ranked SPP1 expression in these cells ( figure 4 K,L). In summary, single-cell profiling of diagnostic-stage IPF ALI cultures revealed epithelial and mesenchymal programmes defined by profibrotic basal cells, primed fibroblasts and inflammatory HBECs, showing both conserved and stage-specific features when compared with explant-derived IPF atlases. All three drug treatments partially normalised IPF signatures across cell types, with saracatinib showing the strongest inhibition of fibroblast activation and epithelial proliferation in vitro. The expression of Src and Abl kinase family members further points to a mechanistic basis for the observed cell type specific responses. Together, we provide transcriptional phenotypes characteristic of early stages of IPF and highlight their potential as therapeutic targets.

Understanding the cellular mechanisms driving IPF progression is essential for identifying therapeutic targets. Here, we provide a comprehensive characterisation of the IPF bronchial airway mucosa and its response to treatments using the preclinical ALI culture model. Based on the sampling from the subsegmental bronchi, the fibroblasts captured in our ALI cultures are likely peribronchial in origin, residing directly below the airway epithelium. 27 28 These cells exhibited features of early activation defined by SPP1 overexpression together with OXPHOS and priming signatures. SPP1 is well-documented in IPF and has been reported as overexpressed by macrophages based on scRNA-seq analyses. 10 29 In our validation stainings, we did not note an influx of SPP1 + macrophages in the tissue as observed in explant lungs, suggestive that this is another difference brought about by disease progression. Meanwhile, more recent lineage tracing studies in murine fibrosis models have uncovered the trajectory of activation of fibroblasts, characterised by an intermediate Spp1 + stage and a fully activated Cthrc1 + myofibroblast state associated with collagen production and contractility. 28 30 The IPF fibroblasts in our study barely expressed CTHRC1 and only low levels of ACTA2 , along with other myofibroblast-associated genes such as AEBP1 , SPARC and FN1 , 15 30 indicating that they are not in the fully activated state. Integrative analyses with scRNA-seq IPF atlases further support this, revealing an inverse relationship between SPP1 and COL1A1 expression, and a marked difference in fibrotic scores relative to the Lang et al dataset. Functionally, these cells remained responsive to TGF-β stimulation, demonstrating their retained capacity for activation. In the studies by Tsukui et al , peribronchial fibroblasts expressed Acta2 the highest but Col1a1 the lowest levels when compared with alveolar and adventitial fibroblasts. 27 28 Consistent with this, our fibroblasts were indeed highly contractile yet produced collagen following TGF-β stimulation, suggesting that peribronchiolar fibroblasts can also contribute to ECM deposition. Together, these findings contextualise the primed state of fibroblasts we identified in newly diagnosed IPF patients using ALI cultures, capturing a snapshot of the fibroblast phenotype along this spectrum of activation at this stage of disease. As stem cells of the airway mucosa, basal cells can drive respiratory pathologies when aberrant. This is evident in IPF basal cells with their increased Notch pathway signalling, as shown here and in other studies. 12 29 In murine lung injury models, persistent Krt5 + pods with bronchiolised identities were sustained through Notch signalling. 31 32 Consistent with these findings, we observed different basal pod structures expressing JAG1 and FN1 in IPF tissues. The basal ECM cells identified in this study also expressed cytokeratins KRT6A and KRT17 , reminiscent of the KRT5 - distal basaloid cells described in IPF atlases, 4 13 as well as transcription factors SOX4 and SOX9 , key regulators of lung development and regeneration. 33 34 In the study by Jaeger et al , pathogenic KRT17 + basal cells were found in bronchial regions, resembling the basal ECM cells described here. 6 Integrative analyses revealed highly concordant cellular phenotypes, especially among basal cells, which is further exemplified by their shared overexpression of KRT6A , KRT17 , SOX4 , JAG1 , THBS1 , ITGAV and ITGB8 . These cells likely represent the same disease-associated basal subtype given their shared bronchial origin, and our dataset further provides insight into their drug responses. Comparisons with HBECs from the Lang et al dataset revealed similar fibrotic scores with their model system, consistent with the earlier disease stage represented by our samples. Functionally, co-culture experiments demonstrated the influence of HBECs on fibroblast activation as they became more proliferative and contractile, consistent with previous findings. 6 Interestingly, fibroblast collagen production decreased in the presence of HBECs, which may reflect a proliferative state after 72 hours of co-culturing. Longer-term culture and measurement of mature collagen fibrils may reveal the inductive effect reported by others. 6 Recent studies also identified transdifferentiated alveolar cells with bronchiolar identities that develop into honeycomb cysts, emphasising the importance of HBEC phenotypes in IPF pathology. 35 36 Here, we identified a distinct basal cell phenotype using ALI cultures ( TP63 + , KRT5 + , KRT17 + ), representing a separate aberrant population of basal cells at the bronchi implicated in disease progression. Collectively, these findings suggest the presence of diverse pathogenic basal cell phenotypes across tissue regions and disease stages, each likely contributing to the dysregulated repair that is rampant in IPF. All three drug treatments induced significant transcriptional changes in IPF, with fibroblasts showing the strongest response—highlighting the known antifibrotic effects of nintedanib and pirfenidone. 1 More importantly, saracatinib demonstrated a strong potential as a treatment candidate. It more effectively suppressed fibroblast proliferation and contraction even in the presence of TGF-β, likely through the downregulation of ACTB , a key component of the cellular contractile apparatus during wound healing and cellular division. 37 This observation aligns with Ahangari et al , who similarly reported superior inhibition of fibroblast activation by saracatinib compared with nintedanib or pirfenidone. 38 However, their study did not employ a single-cell approach, leaving open whether distinct fibroblast populations exhibit differential drug sensitivities. Further studies addressing this will uncover these responses more comprehensively. In HBECs, saracatinib downregulated CDKN2A , a senescence marker linked to apoptosis-resistant fibroblasts, aberrant basaloid cells and the senescence-associated secretory phenotype of epithelial cells. 1 4 5 However, none of the treatments significantly reduced the dysregulated repair and TGF-β processing signatures in basal ECM cells, suggesting that these cells may require alternative therapeutic targeting. Consistent with Jaeger et al , who showed that saracatinib reduced bronchosphere formation of KRT17 + bronchial basal cells in vitro, we observed a similar suppression of HBEC proliferation. Despite this, basal cells appear to retain aspects of their aberrant phenotype, including enhanced TGF-β processing, which may still contribute to tissue remodelling. While saracatinib’s ability to fully reverse all pathological processes in vivo remains uncertain, the NCT04598919 clinical trial will help determine its efficacy in improving the outcomes of IPF patients. Mechanistically, saracatinib inhibits Src family kinases, including Src, Yes, Lyn, Fyn as well as Abl kinases. 39 These kinases play crucial roles in TGF-β-driven processes such as myofibroblast differentiation, EMT and other fibrogenic responses. 40 Our findings highlight the complexity of drug effects with the use of ALI cultures, revealing both overlapping and distinct properties across cellular populations of the airway mucosa. On the transcriptomic level, saracatinib exhibited similar responses to nintedanib and pirfenidone, which already hints at its antifibrotic potential. However, on the functional assays, saracatinib often outperformed nintedanib and pirfenidone, likely reflecting modulation of genes uniquely affected by saracatinib, including ACTB , ITGB1 and CDKN2A , which are central to fibroblast and epithelial disease phenotypes. Additional gene-level changes not discussed here were also modulated, indicating broader cellular effects. Future therapeutic strategies may build on these findings and data by combining saracatinib with complementary agents targeting residual dysregulated pathways to achieve more comprehensive antifibrotic responses. Here, we speculate the following cascade of events that contextualises our observed pathogenic signatures in IPF ( figure 5 ). The process is initiated by HBECs releasing proinflammatory cytokines/chemokines, triggering leucocyte recruitment to the airway mucosa. Macrophages then release latent TGF-β that will be processed by basal cells and other HBECs. The presence of bronchiolised structures intensifies these effects due to increased HBEC numbers. Activated TGF-β transforms primed fibroblasts within the basal membrane, leading to their proliferation and differentiation into myofibroblasts. We also hypothesise that basal cells contribute to bronchiolisation in IPF through persistent basal pods, driven by increased Notch signalling and impaired regeneration. A variety of preclinical models are available to study airway biology, ranging from 2D cultures, ALI cultures, spheroids/organoids, precision-cut lung slices and organ-on-a-chip systems. 7 41 Each method presents a trade-off between cost-effectiveness and biological complexity. For example, while 2D cultures are very accessible, they are limited to the cell type they represent and lack the structural and functional complexity of the airway mucosa. Given the goals of this study, ALI cultures provided an optimal balance and sufficient complexity to capture the airway mucosa biology at single-cell resolution, 42 while also providing the possibility for drug treatment studies. 43 A key strength of ALI cultures is the relative accessibility of samples via bronchial microsampling, unlike those from more distal lung regions. We further validated the effectiveness of ALI cultures for perturbation experiments, reinforcing their value in drug discovery and mechanistic studies. Moreover, our study did not need hard-to-reach samples and complex experimental systems to identify complex transcriptomic phenotypes of the airway mucosa and their drug responses by leveraging the use of ALI cultures. Other studies have also used other preclinical models to investigate the effects of nintedanib and pirfenidone at single-cell resolution in IPF. 26 These complementary approaches highlight the diversity of available models and show the importance of choosing the appropriate system depending on the research question. Our study also has its limitations. First, the scRNA-seq dataset sample size is relatively small. Second, we did not enrich for HBECs during procurement, leading to the inclusion of fibroblasts in ALI cultures, which we nevertheless validated in both identity and function. Third, drug treatments were conducted over different time frames: scRNA-seq samples for 24 hours and cellular assays for 48–96 hours. Thus, we cannot rule out that longer treatments might reveal further transcriptomic differences, nor those brought about by the culture method. Fourth, we acknowledge that the diagnostic timepoint may capture varying stages of the disease. Nevertheless, it represents one of the earliest feasible opportunities to obtain patient samples and thus still provides critical insights into IPF disease progression. In this study, we offer a comprehensive dataset from ALI cultures and lung tissues, highlighting earlier IPF transcriptional programmes involving primed fibroblasts, dysregulated basal cells and proinflammatory HBECs. We compared treatment effects of nintedanib, pirfenidone and saracatinib with single-cell transcriptomics and cellular assays. Saracatinib exhibits similar and unique effects to nintedanib and pirfenidone on the transcriptomic level and outperformed the two drugs in cellular assays, showcasing its potential utility against IPF. This study provides valuable insights into IPF mechanisms and serves as a valuable resource for investigating drug-related effects and their efficacy in inhibiting fibrotic responses in newly diagnosed IPF patients.