Cystic Fibrosis airway epithelial cells have an exaggerated response to hypoxia that is partially reversible through transmembrane receptor modulation

Abstract

The pathophysiology of cystic fibrosis (CF) leads to epithelial cell hypoxia, which directly affects epithelial cells. CF is caused by genetic disruption of the CF transmembrane receptor that has important direct impacts on cell signaling and proteotoxic stress, and indirect impacts through microbiome alterations. How these alterations impact hypoxia signaling is not known. We collected primary human airway cells from explanted lungs of individuals with or without CF, differentiated them at air-liquid interface, and subjected them to short-term hypoxia. Differential gene expression was assessed by RNAseq, with findings validated by flow cytometry. We also assessed the impacts of modulator therapies on CF epithelial cells. While there was overlap in the transcriptomic response to hypoxia between CF and referent epithelial cells, CF cells activated additional pathways. In CF cells under hypoxia, activation of the hypoxia pathway was associated with HIF1α, EMT, and immune-related pathways, the latter not seen in referent cells. Among HIF1α related genes, VEGF was uniquely increased in cells from CF, and its expression was modulated through HIF1α signaling. We show that correction of CFTR blunts exaggerated response to hypoxia in CF cells. These results suggest CF airway cells have an exacerbated response to hypoxia, which may be alleviated through the correction of misfolded CFTR.

Introduction

Cystic fibrosis (CF) results from genetic defects in the cystic fibrosis transmembrane conductance regulator (CFTR) protein. Under physiological conditions, CFTR plays a crucial role in regulating the movement of chloride ions and water across cell membranes, modulating mucous viscosity¹. Thickened mucous contributes to recurrent pulmonary infections, bronchiectasis and severe hypoxemic respiratory failure².

In recent years, effective modulation via Elexacaftor/Tezacaftor/Ivacaftor (ETI) to improve CFTR conduction yielded significant improvements to quality of life, lung function, and surrogates of survival for patients with cystic fibrosis³. However, ETI is not a therapeutic option for 10% of the CF population, and has limited benefit in patients with existing advanced lung disease or contraindications to treatment⁴. Even in CF patients that are eligible at a young age for ETI, markers of neutrophilic inflammation, Pseudomonas burden and other infections remained unchanged up to 6 months after initiation of treatment^5–7. Therefore, there is a critical need to understand additional mechanisms of CF lung injury that may complement ETI or provide benefit to patients that are not candidates of ETI.

Sterile injury is a significant mediator of the pulmonary pathology of CF and is induced by a variety of non-infectious stimuli^8–10. Most importantly, hypoxia is a potent trigger of the sterile inflammatory cascade¹¹. As we previously demonstrated, hypoxia drives NK cell activation through upregulation of MICB and other NKG2D ligands^{12, 13}. In CF, tissue hypoxia is a key cellular feature of early lung disease and is present systemically in advanced CF lung disease¹⁴. Thickened mucous results from impaired epithelial Cl- transport and increased epithelium sodium channel (ENaC) activity. This results in dual mechanisms to lower CF epithelial cell intracellular oxygen (O₂) concentration: increased O₂ consumption fueled by ENaC activity and impaired oxygen diffusion through the thickened mucous layer^{2, 15}. Reduced intracellular oxygen has been shown to impair CFTR mRNA expression across organs, suggesting an amplification of underlying cell deficits¹⁶. In addition, hypoxia skewed expression of mesenchymal markers and transcription factors among epithelial cells in experimental models of lung disease^17–19. This process whereby epithelial cells lose lineage-defining proteins in the evolution of fibrosis is called Epithelial Mesenchymal Transition (EMT), which is conserved across lung diseases²⁰.

Hypoxia leads to several intracellular changes, including the activation of hypoxia-inducible factors (HIF)²¹. HIFs are a family of three tightly regulated nuclear transcription factors: HIF1α and HIF1β subunits, and HIF2α. In the absence of O₂, stabilized HIFα subunits translocate into the nucleus and participate in binding to hypoxia response elements at target genes. HIF activation leads to the transcription of over 120 individual genes implicated in inflammation, metabolism, angiogenesis, erythropoiesis, and other adaptations related to hypoxia²². There is emerging evidence that hypoxia causes cellular stress and accelerates epithelial to mesenchymal transition¹⁷. Several potential drug candidates have shown efficacy at modulating intracellular HIF pathways^{23, 24}.

It remains unclear how CFTR dysregulation affects the response of airway cells to hypoxia. We hypothesized that CF airway epithelial cells would have non-canonical hypoxia responses, including exaggerated inflammatory and EMT signaling.

Results

CF airway cells show dysregulated response to hypoxia.

To investigate the effect of hypoxia in cystic fibrosis (CF) airways, we exposed airway epithelial cells from de-identified CF (delF508/delF508) and non-CF donors (non-CF) to hypoxia (1% O₂) or normoxia after differentiation at air liquid interface. After 24 hours, whole RNA was extracted and sequenced (Figure 1A).

Figure 1. Airway hypoxia drives a transcriptomic shift in cystic fibrosis cells.

(A) Schematic overview of experimental design. Primary cells from patients with cystic fibrosis (CF, n = 5) or from non-CF donor controls (non-CF, n = 5) were grown at ALI for 28 days. After differentiation, cells were subjected to 24hrs of hypoxia (1% O₂) or normoxia. Cells were lysed directly on membrane and total RNA content was collected and sequenced. (B) Multi-dimensional scale (MSD) of differentially expressed genes (DEGs) from hypoxia (adjP < 0.1). (C) Volcano plot of the DEGs in CF airway epithelial cells comparing hypoxia to normoxia. (_D) Gene set enrichment analysis (GSEA) using MSigDB’s Hallmark gene set collection of the top 10 up- and down- regulated pathways in CF cells exposed to hypoxia.

Multidimensional scaling (MDS) analysis revealed shifts in overall transcriptomic profiles in hypoxic CF and non-CF cells (Figure 1B). Qualitatively, non-CF cells had a smaller and more uniform shift in MDS space as compared to hypoxic CF cells, suggesting greater heterogeneity in CF cells’ response to hypoxia. Indeed, the directionality of CF transcriptional responses varied across the first two principal components compared to non-CF cells (Supplemental Figure 1A).

In line with these observations, hypoxia induced differential expression of 259 (adjP < 0.1, log2FC > 0.6) genes in CF airway cells (Figure 1C) relative to normoxic cells. To contextualize these findings, we performed gene set enrichment analysis (GSEA) ²⁵ using the Hallmark gene set (Figure 1D) ²⁶. We found that genes within the hypoxia pathway were the most transcriptionally enriched in hypoxic CF cells; however, genes in epithelial-mesenchymal transition (EMT), inflammatory (TNFα signaling, inflammatory response) and metabolic (glycolysis) pathways were also increased. We also observed transcriptional downregulation in genes related to oxidative phosphorylation and proliferation (G2M, E2F targets) in CF airway epithelial cells.

In contrast, non-CF donor airway cells had a muted response to hypoxia. There, we observed induction of 73 (adjP < 0.1, log2FC > 0.6) differentially expressed genes (Supplemental Figure 1B). GSEA also identified that hypoxia induced differential transcription of genes in the hypoxia and EMT pathways (Supplemental Figure 1C), largely concordant with prior findings. This was further corroborated by significant increases in hypoxia (Supplemental Figure 1D) and EMT metagene scores for both groups (Supplemental Figure 1E). We observed a positive correlation between hypoxia and EMT pathway metagenes in CF cells (Spearman R = 1.0, p = 0.017, Supplemental Figure 1F) but not non-CF cells (Supplemental Figure 1G) subjected to hypoxia (no association was found for the normoxia conditions). This suggests a tighter relationship between hypoxia and EMT transition in CF cells. We observed other differences between CF and non-CF cells. Notably, the immune and inflammatory genes induced in CF airway cells were absent in non-CF cells. Globally, genes downregulated by hypoxia in CF cells, such as OXPHOS, were less impacted in non-CF cells. Key results were reproduced in CF cells exposed to hypoxia as measured by RT-PCR (Supplemental Figure 2). Together, these results show qualitative and quantitative differences in the transcriptional hypoxia response in CF cells relative to non-CF cells.

CF airway cells uniquely exhibit an inflammatory response to hypoxia.

To probe differences in the hypoxia response further between CF and non-CF, we directly contrasted the differentially expressed genes in both groups (Figure 2A). We observed that 90% of genes increased in hypoxia in non-CF were also upregulated in CF cells. There were an additional 193 upregulated genes in CF that were absent in non-CF cells, suggesting an exaggerated response to hypoxia.

Figure 2. Hypoxia induces a greater breadth of response in CF airway cells.

(A) Overlap of hypoxia-induced genes in CF and non-CF donors (adjP < 0.1 ; FC > 0.6) (B) Correlation between the log2 fold change of genes increased with hypoxia in non-CF (x-axis) with CF (y-axis) (padj <0.1 in at least one contrast). Solid black diagonal line represents a 1:1 association between fold changes of both groups. (C) Overrepresentation analysis on KEGG pathways for genes that were significantly induced by hypoxia exclusively in (C) non-CF (66 genes) or (D) CF (193 genes).

We observed a positive association between fold changes (FC) in response to hypoxia in CF and non-CF cells (Figure 2B). The FC were generally greater in cells from CF. Notable genes with greater upregulation in response to hypoxia (log2FC > 0.6) included genes coding for angiopoietin-4 (ANGPTL4), parathyroid hormone like hormone (PTHLH) and placental growth factor (PGF). Other genes uniquely upregulated in CF cells, such as colony-stimulating factor 3 (CSF3), play important roles in modulating granulocytes.

We next performed an overrepresentation analysis to investigate which KEGG pathways were commonly upregulated in both CF and non-CF cells in response to hypoxia. Upregulated genes shared between both groups revealed enrichment in two pathways, Hypoxia Inducible Factor 1 alpha (HIF1α) signaling and, to a lesser extent, carbon metabolism (Figure 2C). In contrast, a similar analysis on the genes uniquely upregulated in hypoxic CF cells showed 16 separate pathways. These included several immune-related pathways, including cytokine-associated responses (Figure 2D). These observations complement the Hallmark gene set enrichment analyses, demonstrating that in addition to a standard hypoxia response, CF airway cells display pro-inflammatory signatures relative to non-CF cells exposed to hypoxia.

Genes within the HIF1α pathway were over-represented in CF airway cells exposed to hypoxia relative to the non-CF airway cells (Figure 2D). We also found that genes related to fibrosis, angiogenesis and inflammation (PDGFB, TGFB1, TGFB3, SLC2A1 and VEGFA) were strongly induced in hypoxic CF cells compared to normoxic CF cells, though we observed high variability between samples (Figure S3A). In non-CF cells, SLC2A1, VEGFA and PDGFB were increased in hypoxia (Figure S3B). The leading edge (LE) genes common in CF and non-CF upon hypoxia relate mainly to the metabolic shift associated with reduced oxygen consumption (Figure S3C), while CF-specific LE are associated with angiogenesis. To understand the degree that findings attributed to CF basal cellular processes may impact these results, we investigated differences between CF normoxic and non-CF normoxic cells. We found CF cells uniquely expressed 263 upregulated genes and 79 downregulated genes during normoxia. Of these upregulated genes, 43 were common to the genes upregulated in CF cells in hypoxic conditions (Supplemental Table 1). Taken together, these results underscore a wider breadth of response upon exposure to hypoxia in CF cells compared to non-CF cells and suggest that HIF1α may link CF hypoxia and CF airway inflammation.

MICB and VEGFA are downstream of HIF1Α signaling in airway epithelial cells

We found that a HIF1α pathway metagene was increased in CF hypoxic cells relative to CF normoxic cells (p = 0.02) and that this response to hypoxia was lessened in non-CF cells (Figure 3A, p = 0.1). Notably, 4 of the 5 non-CF samples had increased HIF1α pathway expression following hypoxia. Concordant to these observations, gene counts of HIF1A showed wide variability in non-CF cells (F = 13.467, 95% CI 3.34 – 54.22, p = 0.00065) and were reduced in CF cells relative to non-CF cells (p = 0.06). Further, along the HIF1α signaling pathway (Supplemental Figure 4A), regulatory genes promoting reduced HIF1α protein were increased in CF cells at baseline relative to non-CF cells (Supplemental Figure 4B and 4C).

Figure 3. MICB is increased with HIF1α activation in vitro.

(A) Paired analysis of HIF1α metagene comparing hypoxia to normoxia for non-CF and CF airway epithelial cells. (B) To identify genes downstream of HIF1α activation in airway epithelial cells, 16HBE cells were subjected to hypoxia or normoxia for 24hrs in submerged culture. Total RNA was collected for qRT-PCR. (C) Baseline transcript expression (ΔCT) for 5 key genes are shown following 24 hours of normoxia in 16HBE cells. (D) Gene transcript expression, measured by fold change (2^ΔΔCT), is shown for VEGFA, MICB, and INFA1. Airway epithelial cells were exposed to hypoxia after treatment with HIF1α inducers (hypoxia, roxudustat, CoCl) or a HIF1α inhibitor (LW6). Fold change (2^ΔΔCT) in gene expression across conditions is shown for (E) MICB and (F) VEGFA genes. Each condition and assay used at least 5 biologic and 3 technical replicates and values for MICB and VEGFA shown in panel D are included in panels E and F as reference. P-value level of significance denoted by *, <0.05 and **, <0.01. Differences between groups were assessed by (A) paired Mann-Whitney U testing or unpaired Mann Whitney U testing (D,E,F).

The specific effects of HIF1α activation in airway epithelial cells remain poorly elucidated. To investigate targets downstream of the HIF1α pathway in the airway epithelium, we subjected 16HBE airway epithelial cells to hypoxia for 24 hours, and in the presence of HIF1α modulators (Figure 3B). 16HBE cells were used in submerged culture to improve HIF1α modulator delivery and limit heterogeneity in primary cells. We measured gene transcripts reported by our group¹² and in other cell types to be downstream of HIF1α (Figure 3C). We observed a high baseline expression for VEGFA (5.53 median dCT, IQR 5.47 – 5.54), MICB (5.38 median dCT, IQR 5.27 – 5.72) and ALDH3B1 (8.3 median dCT, IQR 7.42 – 8.31), while EPO (17.8 median dCT, IQR 17.8 2– 17.95) and INFA1(16.98 median dCT, IQR 16.3 – 17.47) were present at lower levels. We found that hypoxia induced increased expression of MICB (1.38 median IQR 1.23 – 1.5 FC, p = 0.01) and VEGF (1.47 median IQR 1.45 – 1.54 FC, p = 0.05, Figure 3D). We observed no changes in EPO (0.9 median IQR 0.7 –0.99 FC, p = 0.4), and ALDH3B (1.1 median IQR 1.06 – 1.26 FC, p = 0.1) expression and reduced INFA1 expression (0.43 median IQR 0.42 – 0.59 FC, p = 0.02) upon hypoxia (Figure 3D).

To understand if MICB (Figure 3E) and VEGFA (Figure 3F) gene transcription are potentially downstream of HIF1α activity, we treated submerged 16HBE airway epithelial cells with modulators known to directly induce HIF1α (cobalt chloride), prevent HIF1α inhibition (roxudustat) or inhibit HIF1α (LW6) in non-airway epithelial cells²³. We found that LW6 reduced MICB expression (0.85 median log FC IQR 0.84 – 0.85 p <0.0001) and but produced no change in VEGFA expression (1.05 median log FC IQR 0.98 – 1.11 p = 0.35) during hypoxia. In contrast, hypoxia, roxudustat and cobalt chloride increased expression of both MICB and VEGFA. These findings show that MICB and VEGFA are potentially downstream of HIF1α signaling in airway epithelial cells; though, additional experiments directly measuring HIF1α protein are necessary to confirm these results. As MICB is an externally expressed protein ²⁷, and our prior work has shown that hypoxia induces surface MICB protein, this may be an indicator of HIF1α pathway activation in airway epithelial cells.

Hypoxia induces cell death and stress in subsets of CF airway epithelial cells.

To determine if hypoxia differentially impacts subsets of CF airway epithelial cells, we differentiated CF and non-CF airway epithelial cells at air liquid interface (Figure 4A, Supplemental Figure 5). There were differences in distributions of basal cells, club cells, ciliated cells, or secretory cells between CF and non-CF across conditions (PERMANOVA p = 0.02, Figure 4B). However, hypoxia caused disproportionate cell death, as assesed by viability dye uptake, in the distributions of CF cells relative to non-CF cells (Figure 4C, Χ² p = 0.02). We found no differences in viability dye-positive basal cells (p = 0.1), club cells (p = 0.1), or secretory cells (p = 0.2) between CF and non-CF (Figure 4D). However, 73% IQR 66 - 74% of dead (viability dye positive) epithelial cells were ciliated cells in hypoxic CF conditions relative to 54% IQR 44 – 55% of non-CF donor cells (Figure 4D, p = 0.008).

Figure 4. Expression of MICB on CF airway epithelial cells.

(A) Primary CF or non-CF cells were grown at ALI for 28 days. Cells were subjected to hypoxia (1% O2) or normoxia for 24 hours, then collected for flow cytometry analysis. (B) Relative frequencies of epithelial cell subsets across conditions for non-CF and CF samples. (C) Relative frequencies of epithelial subsets among dead (viability dye positive) epithelial cell subsets in non-CF and CF samples following hypoxia. (E) Representative histograms showing MICB expression on epithelial cell subsets in fluorescent minus one (FMO) control, normoxia, and hypoxia samples. (F) Frequency of MICB+ cells among dead epithelial cells. (G) Median fluorescent intensity (MFI) of MICB on dead (viability dye positive) epithelial cell subsets. Differences across populations were assessed by (B) PERMANOVA or (C) χ² testing. P-value for other comparisons reflect Mann-Whitney U testing.

As we found the MICB gene was downstream of HIF1α signaling in airway epithelial cells, we hypothesized that surface MICB may be an indicator of HIF1α activation and that hypoxic stress and cell death would be correlated. We measured MICB on the surfaces of cell subsets (Figure 4E). Across the dead (viability dye positive) cells, we found that hypoxic CF epithelial cells had increased MICB relative to non-CF (Figure 4F, p = 0.03). Notably, MICB was increased on CF basal cells (p = 0.003) and club cells (p = 0.04) but not ciliated cells (p = 0.4) or secretory cells (p = 0.2) (Figure 4G).

CFTR modulation may alter the response to hypoxia

Our results have shown that cells from CF patients have a skewed response to hypoxia through increased involvement of the HIF1α pathway. We hypothesized that this response may be accentuated by dysfunctional CFTR. To test this, we treated airway cells from CF patients with ETI modulator therapy²⁸ or vehicle control for 14 days prior to 24 hours of exposure to hypoxia or normoxia (Figure 5A). We measured several target genes in HIF1α pathways (VEGFA, MICB) and EMT pathways (PTHLH, LOXL2, VIM), selected for differential expression in CF airway cells.

Figure 5. Effect of CFTR modulation on key HIF1α-associated gene expression.

(A) Primary airway epithelial cells from patients with cystic fibrosis were grown in submerged culture in the presence of ETI or vehicle control. After 14 days, the cells were subjected to 24 hours of hypoxia before whole cell lysis. RNA was collected for qRT-PCR. (B) Fold change (2^ΔΔCT) in gene transcription for key inflammatory and EMT genes following hypoxia are shown for ETI and vehicle control cells. Data were analyzed using generalized estimating equation adjusted linear regression.

We found significant heterogeneity in the response to hypoxia across individuals. CFTR modulation with ETI did not alter CF airway cell expression of IL-33 (p = 0.35, VEGFA (p = 0.74), or PTHLH (not shown) genes (Figure 5B). However, we found increased transcription of MICB genes (p < 0.0001) and decreased transcription of the EMT gene VIM in cells treated with CF modulation preceding hypoxic challenge (p = 0.02). These results suggest that CFTR may potentiate some of the dysregulated CF response to hypoxia in airway epithelial cells, but future investigation is needed to probe the processes specific to airway epithelial cell subsets.

Discussion

We observed that hypoxia solicits a distinct inflammatory transcriptional profile in CF airway epithelial cells using primary cells from explanted CF lungs grown at air-liquid interface. Our results suggest a maladaptive response to hypoxia in CF patients, with HIF1α playing a central role. Finally, components of this response may be rescued using highly effective CFTR modulator therapy.

Notably, we found that airway cells from CF individuals had increased transcription of genes in HIF1α, epithelial-to-mesenchymal transition, inflammatory signaling, and metabolic pathways in response to hypoxia. Several factors can explain this altered response, including differences in cell composition¹⁹. Prior work using single cell RNA-sequencing identified transcriptomic shifts in key populations of airway cells from CF patients, such as an expanded ciliated cell gene expression in ex vivo CF cells, as well as decreases in a proliferative basal subset and club subset in favor for an expansion of an inflammatory ciliated population in CF patients compared to non-CF individuals²⁹. It would follow that different subpopulations of cells have separate responses to hypoxia. While we did not observe differences in cell composition between CF and non-CF recipients in our study, our flow cytometry panel did not have the same resolution as other single cell technologies. Nonetheless, there is a possibility that other factors may explain these differences such as pre-existing cellular stress or inflammatory priming from the CF pulmonary milieu^30–32.

Here, we used a cell culture model of hypoxia to replicate conditions within the airways of advanced CF lung disease. While in vivo tissue and cellular oxygen content is challenging to measure, CF lung tissue experiences hypoxia through a variety of mechanisms. In contrast to other epithelium, airway cells receive oxygen primarily through luminal oxygen as compared to vascular sources³³. Thus, obstructive mucous plugging is a feature of CF, which has been shown to induce cellular hypoxia in mouse models of emphysema^{34, 35}. Furthermore, CF epithelial cells have been shown to consume more oxygen due to the increased metabolic demands of ENaC-mediated Na+ absorption³⁶. CF patients are commonly anemic with low iron stores, exacerbating systemic hypoxia³⁷. Thus, cell culture hypoxia models the extremes of airway hypoxia experienced in advanced CF lung disease. This may drive some of the observed differences in cell populations by altering cell cycle entry and cell differentiation³⁸. Investigating these drivers will form the basis of critical future work.

Importantly, this study adds to the existing literature demonstrating the significant inter-dependence between CFTR and hypoxia signaling^{39, 40}. Prior studies have demonstrated that CFTR inhibition impairs HIF-1α stabilization, likely through modulation of reactive oxygen species³⁹. In CF airway cells treated with hypoxia mimetics, HIF1α protein was reduced relative to CFTR-corrected cells⁴¹. Notably, while we did not measure HIF1α protein directly, our findings were largely concordant to this prior work as HIF1A transcripts and HIF1α pathway genes trended lower and HIF1α regulatory gene transcripts were higher in CF cells. The relatively smaller effect observed here may be attributed to differences in sample size, non-CF cell variability, and experimental cell types.

One unique trait of the CF airway epithelial cell response to hypoxia was the activation of immune-related pathways. CF-specific co-morbidities may underpin this inflammatory cascade. For instance, lung microbial communities change in CF as the lung disease progresses⁴², with increases in pathogenic bacteria such as Pseudomonas aeruginosa ⁴³, and decreases in overall microbial diversity ⁴⁴. Both alterations are associated with disease progression ⁴⁵ and increased inflammation. One key finding we identified is that hypoxia induced transcription of MICB, a major ligand to the NK cell NKG2D damage receptor. NKG2D ligands have been reported as an important factor in the clearance of Pseudomonas aeruginosa, thus have added significance in CF cells. We found MICB to be highly expressed on CF airway epithelial cells with markers of cellular death. Interestingly, we found higher proportions of dead ciliated cells in CF cultures relative to non-CF cultures. At the same time, we found more surface MICB in basal and club cells within the dead cell populations. These findings may signify differences in susceptibility to hypoxia-mediated cell-death among CF airway cell subtypes. Further, it raises the important questions of how correlated hypoxic stress and cell death may be across airway subtypes, and whether MICB signals hypoxia exposure in all cell types.

We also found that ETI treatment resulted in increased MICB gene expression in response to hypoxia. These results may reveal a regulatory mechanism to suppress cellular stress in CF airway cells that is potentially rescued with ETI. Notably, CFTR also promotes the regulation of antioxidant defenses, such as glutathione transport, which protects cells against oxidative stress induced by hypoxia ⁴⁶. Thus, while not contributing to the UPR, CFTR dysfunction may result in lower compensatory mechanisms to reduce inflammation, particularly during episodes of hypoxia. Taken together, these observations suggest different underlying pathways contribute to the altered inflammatory response to hypoxia in CF airway epithelial cells.

While both CF and non-CF cells had increased epithelial mesenchymal transition (EMT) gene transcription upon exposure to hypoxia, there were several prominent EMT genes only enriched in hypoxic CF cells. In line with this observation, transcriptional skewing towards EMT has been found broadly in CF cells¹⁹. Given the association between hypoxia and EMT pathways in CF cells only, it is possible that the exacerbated response to hypoxia is at least in part related to EMT for these patients. In vivo, increased EMT signaling aligns with altered barrier integrity due to disorganization of tight junctions observed in airway epithelia ⁴⁷. In support of this, Vimentin (VIM), central to airway epithelial EMT⁴⁸, was uniquely increased in hypoxic CF cells. Moreover, widespread upregulation of VIM is associated with changes in morphology, increased expression of mesenchymal markers, and loss of epithelial markers ⁴⁸. Here, we found that in vitro correction of CFTR through ETI reduced the upregulation of VIM. Thus, VIM may serve as a surrogate marker of the hypoxic EMT response in CF.

In contrast to VIM, many genes which were increased in both CF and non-CF cells upon hypoxia were not affected by ETI. Moreover, we found trends for increased IL-33 signaling and found increased MICB transcripts in CF cells treated with ETI relative to vehicle following hypoxia. This suggests that ETI does not abrogate all the response to hypoxia, only the exacerbated response to EMT signaling in CF, and potentially at the expense of inflammatory signaling. This may reflect the relatively short cell culture time as a study in adolescents with cystic fibrosis showed that 1 month of ETI resulted in a decrease of multiple systemic markers of inflammation alongside improved lung function⁴⁹.

Some limitations of the study included the limited number of samples, which was in part due to a decrease in number of lung transplantations in CF patients since the introduction of ETI. Furthermore, the non-CF and CF samples differed slightly in anatomic site as we were limited in our collection of non-CF airway samples to the trachea. However, flow cytometry findings showed that differences in transcriptomic profiles are unlikely to result from underlying airway epithelial cell composition at baseline. Some experiments were performed in 16HBE cells in submerged culture, which can differ from ALI conditions and those of primary cells. We opted to select patients with the delF508/delF508 mutation to remove CFTR mutation as a confounding factor given our small sample size. How other CFTR mutations may impact the response to hypoxia will be an important future question. In addition, all CF samples were obtained from explanted lungs, where the deteriorating state of the lung may skew the response to hypoxia. While obtaining samples from CF patients with healthier lungs may help identify links not affected by these covariables, ETI has greatly reduced the need for bronchoscopies, making sample collection rarer. Future studies comparing response to hypoxia pre and post ETI will be key in narrowing the link between defective CFTR and hypoxic responses. Finally, we did not directly measure HIF1α protein across these experiments, so it is possible that experiments using HIF1α modulators may have had pleiotropic effects. As such, findings may also result from hypoxia-induced but non-HIF1α pathways.

Our study highlights an exacerbated response to hypoxia in epithelial cells from CF patients. These findings may have important clinical implications. CFTR modulation is not available for all patients with CF and components of the hypoxia signaling cascade may represent viable alternative therapeutic targets to reduce chronic lung injury. In addition, we found that hypoxia induced a robust cellular stress response even after ETI, suggesting that airway injury resulting from cellular stress may be a hidden challenge for CF patients on ETI. Finally, this work has relevance outside of CF, especially in other patients with obstructive airway disease and immediately after lung transplantation during airway ischemia. Thus, future work could leverage these findings to advance the development of additional therapies for hypoxia as a common mechanism in CF lung disease and beyond.

Methods

Sample selection and cell lines

The Committee on Human Research at the University of California San Francisco approved the use of human tissues for these studies. CF samples and non-CF airway tissue were available from a biobank where samples lacked identifying information beyond condition and collection date. Non-diseased human airway cells were derived from excess donor tissue following lung transplantation⁵⁰. We obtained tracheas and mainstem bronchi from CF patients from autopsies performed within 24 hours after death, as previously described⁵¹. CF samples were selected based on the presence of the deltaF508 mutation. For select experiments, we used 16HBEs14o- (#SCC150, Millipore Sigma, MA, USA) in submerged culture conditions.

Cell collection, expansion, and air-liquid interface culturing

After tissue extraction, primary airway epithelial cells (AEC) were re-suspended in Pneumacult Expansion Plus (P-Ex+, #05040, Stemcell Technologies, BC, Canada). Cells were expanded in submerged culture seeded at a density of 5000 cells/cm², with media changes every 3 days. When confluence above 80% was reached, cells were trypsinized using tryspin-EDTA (0.25%) (#25200056, ThermoFisher, CA, USA). Trypsin was neutralized with 5% FBS in DMEM, and cells were strained through a 100 μm filter and centrifuged before experimentation.

For air liquid interface, 24-well transwell plates (#3470, Corning , CA) were coated with fibronectin (#ACT1-0407-100, AssayCell Technologies, NJ, USA). The apical layer of the transwell was seeded with airway epithelial cells at a density of 50,000 cells/200μL/well. 500μL of P-Ex+ was added to the basal layer. Cells expanded for 7 days before air-lift, whereby apical media was removed and media in the basal compartment was replaced with Pneumacult-ALI-S media (P-ALI-S, #05050, Stemcell Technologies, BC, Canada). Media changes were performed 3 times a week, and cells were allowed to differentiate for at least 28 days.

For samples subjected to hypoxia, hypoxic media was generated by bubbling media with 1% O₂ (5%CO₂) for 3 minutes. Media was exchanged for submerged and ALI cultures, and plates were subsequently added into a hypoxia chamber (#27310, Stemcell Technologies, BC, Canada). The chamber was flushed with hypoxic gas for 2 minutes, then sealed and inserted into the incubator for 24hrs. For samples not subjected to hypoxia, normoxic media change was performed, and the plates were returned to the incubator for 24 hrs.

RNA extraction and sequencing

Whole RNA was extracted at the end of the 24h incubation. Plates were removed from the incubator or the hypoxia chamber, and all media was removed. 350uL of Qiazol was added to the apical layer of the ALI transwell. Qiazol (#79306, Qiagen, Hilden, Germany) was pipet-mixed on membrane, then directly transferred into labeled collection tubes followed by 30 seconds of vortexing. All samples were store at −80°C until RNA isolation.

RNA extraction was performed using the Qiagen RNEasy Mini kit (74104, Qiagen, Hilden, Germany), following manufacturer’s instructions. Briefly, samples were thawed at 4°C. 350uL of fresh 70% ETOH was added to each sample, then 700uL was transferred into an RNEasy Mini spin column placed in a collection tube. Samples was centrifuged at 10,000g for 15 seconds. We added 10uL of DNAse I (79254, Qiagen) to the column membrane before incubation at room temperature for 15 minutes. Samples were washed twice with buffer before flow-through material was discarded. The collection column was placed in a 1.5mL collection tube, and 30uL of RNAse-free water was directly added to the membrane before 1 minute of centrifugation at 10,000g. DNA contamination was assessed via Nanodrop. RNA concentration was measured with Qubit, and RNA integrity was assessed with a BioAnalyzer.

RNA sequencing was performed by SeqPal (Santa Clara, CA, USA). Library preparation for bulk RNA sequencing used the Illumina Stranded mRNA Preparation kit, per manufacturer’s recommendations. In summary, 100 ng of total RNA was used for each library preparation. Oligo(dT) magnetic beads were used to purify and capture mRNA molecules containing polyA tails. The purified mRNA was fragmented and copied into first strand complimentary DNA (cDNA) using reverse transcriptase and random primers. Adenine (A) and thymine (T) bases were added to fragment ends and adapters were ligated. The resulting products were purified with magnetic beads and indexes were added to the anchor-ligated DNA fragments with PCR. The resulting products were purified and selectively amplified for sequencing on an Illumina system. The quality of the libraries was checked with an Agilent 2100 Bioanalyzer with a High Sensitivity DNA kit (catalog #, vendor name, vendor location). The libraries were sequenced on a NovaSeq 6000 with 100 bp single-end reads at a depth of 20 million reads per sample.

Analysis of bulk RNA Sequencing data

Data was processed using nf-core/rnaseq v3.12.0 of the nf-core collection of workflows⁵². RNA data were mapped to the GRCh38 (2020A) human genome. The pipeline was executed with Nextflow v23.10.0⁵³. Aligned gene counts were normalized using DESeq2 and transformed using the “vst” function on R where hypoxia was contrasted to normoxia datasets. Metagene scores were calculated as the sum of all counts of genes of interest, centered and scaled to a mean of 0 with a standard deviation of 1. A complete notebook of the script is available at: https://github.com/elsabrunetratnasingham.

CFTR correction using ETI

For assays investigating the effect of highly effective modulatory therapy, primary CF airway epithelial cells, were seeded in submerged culture in fibronectin-coated 24-well plates at a density of 50 000 cells/well in P-Ex+ (#05040, Stemcell Technologies). Upon cell adhesion, media was exchanged with P-Ex+ containing ETI (3 uM Elexacaftor, S8851; 10 uM Tezacaftor, #S7059; 5 uM Ivacaftor, #S1144;; Selleck Chemicals, TX, USA). Cells were allowed to reach 80% confluency before experimentation.

In vitro modulation of HIF and CFTR signaling

We modulated in vitro HIF1α activity in 16HBEs14o- (#SCC150, Millipore Sigma, MA, USA) cells using several small molecules. The HIF stabilizer cobalt chloride (CoCl₂) (200uM, #7646-79-9, Millipore Sigma, MA, USA) was added to cultures for non-hypoxic HIF induction. Similarly, Roxudustat (50 uM, #S1007-10MM/1ML, Selleck Chemicals, TX, USA), an inhibitor of HIF prolyl-hydroxlyase that prevents HIF1α degradation, was aslo used as an alternate inducer of HIF. To inhibit HIF activity LW6 (5 uM, #HY-13671, Medchemexpress LLC, NJ, USA), an inducer of von Hippel-Linau (VHL) that increases HIF1α degradation, was added to culture²³. Cells were treated for 24 hours in submerged culture. Media change was performed, and cells were exposed to 24 hours of hypoxia or normoxia. Cells were collected, lysed with Qiazol and RNA was extracted (Qiagen RNEasy kit, #74104, Qiagen), as previously described.

Quantitative real-time PCR

From extracted RNA, cDNA was generated using the SuperScript III First-Strand Synthesis Mix (#18080400, ThermoFisher, CA, USA). Taqman qPCR (ThermoFisher) was performed to quantify EPO, VEGFA, MICB, INFA1, ALDH3B, LOXL2, MICB, VEGFA, PTHLH, and VIM gene expression. We used the housekeeping gene PPIA to to compute differences in cycle threshold (CT, ΔCT or ΔΔCT), as previously described⁵⁴.

Airway Epithelial Cell Flow Cytometry

For select experiments, primary CF airway epithelial cells and non-CF airway epithelial cells were grown to air liquid interface. Cells were exposed to hypoxia or normoxia for 24 hours, as previously described. Cells were trypsinized, washed, centrifuged and aliquoted at 0.5M cells per 200 uL. A flow cytometry panel was designed to identify airway epithelial cell subsets based on previously published work⁵⁵. At 4°C, cells were stained with anti-human CC10 AF488 (clone E-11, sc-365992 AF488, Santa Cruz Biotech, Santa Cruz CA), anti-human MICB AF750 (clone 236511, FAB1599S, R&D Systems, Minneapolis, MN), Ghost Dye Violet 450 Fixable Viability Dye (49826S, Cell Signaling Technology, Danvers, MA), anti-human CEACAM6 BV510 (clone B6.2/CD66, 742684, BD Biosciences, Milpitas, CA), anti-human alpha Tubulin AF647 (clone 6-11B-1, sc-5286 AF647, Santa Cruz Biotechnology, Santa Cruz, CA), and anti-human NGFR PE/Cy7 (clone ME20.4, 345110, BioLegend, San Diego, CA). Data were collected on a FACSAria Fusion flow cytometer (Becton Dickinson, Franklin Lakes, NJ, USA). Cell analysis and data visualization were performed with FCS Express (De Novo Software, Los Angeles, CA, USA). The gating strategy for these experiments are shown in Supplemental Figure 3.

Statistical analyses

Data are reported as means with interquartile ranges (IQRs). All figure legends detail statistical testing. Comparisons of means across multiple groups were made with two-way ANOVA with post-hoc testing performed using Sidak’s multiple comparisons test. Comparisons between 2 groups were made with Mann-Whitney U testing. Overall differences in cellular distribution between groups was assessed with the PERMANOVA test. Finally, to analyse the effect of CFTR modulation on EMT and hypoxia gene expression, we used generalized estimating equation regression to account for repeat measures across individuals. P values less than 0.05 were considered significant. Statistical analyses were performed in R (version 4.2, R Foundation for Statistical Computing).