The airways are constantly exposed to potentially harmful agents including microbes and airway epithelial cells represent the first line of defense.1 Besides playing a role in maintaining the air-tissue barrier, airway epithelial cells have the ability to recognize bacterial components and secrete a wide range of mediators required to initiate and orchestrate a protective immune response against potentially harmful bacteria.2 Depending on the stimulus respiratory epithelial cells can secrete inflammatory mediators in a polarized manner, either to the basal or luminal side.3

In recent years much insight has been generated into the role of immunometabolism in host defense against pathogens; especially metabolic reprogramming of leukocytes upon interaction with bacterial components has received significant attention.4,5 Cells use distinct metabolic pathways to generate energy and biosynthetic intermediates to exert their function.6 Pro-inflammatory responses typically are associated with a shift toward glycolysis, while anti-inflammatory responses are usually associated with enhanced activity of the tricarboxylic acid (TCA) cycle and oxidative phosphorylation.6 Mammalian target of rapamycin (mTOR) has been implicated as a central regulator of intracellular metabolism in several immune cell types, and amongst other, stimulates glycolysis.7 However, studies investigating airway epithelial cell function are scarce and dominated by investigations using epithelial cell lines, in which metabolism is intrinsically altered,8 or in vivo mouse models.9,10 At present, it is unknown whether metabolic reprogramming mediates functional innate immune responses in primary human bronchial epithelial (HBE) cells upon activation by bacterial agonists. Here, we cultured primary HBE cells in an air–liquid interface mimicking the human airways11,12 to address this question. HBE cells highly express Toll-like receptor 5 (TLR5) that recognizes the bacterial component flagellin, one of the main constituents of flagella present in pneumonia-causing pathogens such as Pseudomonas aeruginosa.13 Upon sensing, TLR5 activation results in the production of various pro-inflammatory cytokines and chemokines by epithelial cells.14 Administration of flagellin via the airways has been suggested as an immune enhancing adjunctive therapy in pneumonia.14

Here, we sought to determine the metabolic adaptations in HBE cells upon activation by flagellin and their role in immune activation. We show that the mTOR pathway is a crucial regulatory node between the metabolic and immune status of HBE cells upon flagellin exposure, suggesting that modulation of the mTOR pathway might be an interesting therapeutic approach in airway inflammatory diseases caused by bacteria.


Human bronchial epithelial cells respond selectively to different bacterial stimuli

We first determined the ability of different Gram-negative bacteria to activate HBE cells. For this, we exposed HBE cells to the common human respiratory pathogens Klebsiella pneumoniae and Pseudomonas aeruginosa (UV-inactivated) and measured the release of the chemokines CXCL1, CXCL8, CCL20, and granulocyte colony-stimulating factor (G-CSF) as surrogate markers of HBE innate immune responses. Stimulation with whole inactivated pathogens strongly triggered polarized release of CXCL8, CCL20, and CXCL1 to the basolateral side (Figs. 1a and S1A) and G-CSF to the apical side (Figs. 1b and S1B). This was accompanied by an induction of the corresponding genes as well as upregulation of DEFB4A, the gene encoding the antimicrobial peptide β-defensin-2 (Fig. 1c).

Fig. 1: Human bronchial epithelial cells respond selectively to microbial stimuli.
figure 1

CXCL8, CCL20, and CXCL1 levels in the basolateral compartment (a) and G-CSF levels in the apical wash (b) of untreated HBE cells (PBS) or stimulated for 24 h with LPS (100 ng/ml), flagellin (FLA, 1 μg/ml), UV-inactivated Klebsiella pneumoniae (Kpneu) or Pseudomonas aeruginosa (Paer) (1:2 cell:bacteria). c Fold increase of mRNA expression for the indicated immune mediators (normalized to the housekeeping gene HPRT) in HBE cells from a relative to the unstimulated control (PBS). Data in ac are presented as box and whisker plots (n = 3–4 per group). p values were calculated using Student’s t test compared to unstimulated cells. *p < 0.05; **p < 0.01; ***p < 0.001, ns not significant.

Next, we sought to determine if either lipopolysaccharide (LPS) or flagellin alone, two components present in many Gram-negative bacteria, were able to activate HBE cells. The TLR4 ligand LPS was unable to induce the release of the prototypic HBE cell chemokines (Fig. 1a, b), which is in line with previous observations.15 The TLR5 ligand flagellin induced the polarized release of CXCL8, CCL20, and CXCL1 to the basolateral side (Fig. 1a) and G-CSF to the apical side (Fig. 1b) as well as enhanced expression of the corresponding genes (Fig. 1c). Besides TLR5, flagellin can activate the inflammasome component NLR family CARD domain-containing protein 4 (NLRC4).14 Cytokines that rely on inflammasomes for release of their mature forms include IL‐1β and IL‐18. NLRC4 can mediate the generation of mature IL-1β and IL-18 from their respective pro-forms through activation of caspase-1. While flagellin induced IL1B gene expression, IL-1β secretion did not occur to a significant extent; IL18 was not induced by flagellin (Fig. S2).

These data are consistent with previous studies that documented the potency of flagellin to activate respiratory epithelial cells14,16,17 and suggest an important role of TLR5 signaling in HBE cell activation.

Flagellin shifts the metabolic profile of HBE cells toward carbohydrates

Next, we investigated changes in cellular metabolism in HBE cells after flagellin activation aiming to find new targets to modulate the innate immune function of HBE cells. For this we assessed genome-wide gene expression by RNA sequencing in HBE cells stimulated with flagellin (or PBS control) for 24 h. Flagellin elicited profound changes in the HBE cell transcriptome with 1916 genes differentially expressed when compared to PBS control (Fig. 2a). Pathway analysis showed that flagellin as expected induced over-expression of genes encoding factors related to signaling pathways implicated in innate immunity; remarkably, flagellin treatment downregulated pathways included fatty acids and lipid metabolism-related pathways (Fig. 2b). Transcriptomic analysis of flagellin-induced genes involved in the main energy metabolism pathways revealed that the only metabolic pathway enriched with over-expressed genes was carbohydrate metabolism, while other energy metabolism pathways (aminoacids, lipids, and the TCA cycle and respiratory electron transport chain) were downregulated (Fig. 2c). Mitochondrial morphology has been suggested as an indicator of mitochondrial function.18 Using electron microscopy, we found no changes in mitochondrial morphology that can be attributed to flagellin stimulation (Fig. 2d). To further assess mitochondrial function, we examined the abundance of mitochondria and found no differences in mitochondrial mass attributed to flagellin stimulation (Fig. 2e). In addition, flagellin did not induce the generation of reactive oxygen species (ROS) compared to control (Fig. 2f). Supporting these findings, we found no differences in gene set enrichment analysis (GSEA)19 in the ROS generation pathway from RNAseq data (Fig. S3). These data show that flagellin produces robust alterations in the energy metabolism of HBE cells and suggest that flagellin induces a metabolic shift toward enhanced carbohydrate metabolism.

Fig. 2: Flagellin shifts the metabolic profile of HBE cells toward carbohydrates.
figure 2

a Volcano plot representation (integrating multiple-test adjusted p values and log2-fold changes) of genome-wide transcriptome differences in HBE cells after incubation with flagellin or PBS (control) for 24 h. In red up-regulated genes; in blue downregulated genes. Horizontal line represents the multiple-test adjusted significance threshold (Benjamini–Hochberg adjusted p < 0.05). b Pathway analysis from a showing pathways enriched for up-regulated (red) or downregulated genes (blue). c Pathway analysis of transcriptional response in energy metabolism pathways in HBE cells upon flagellin stimulation for 24 h relative to unstimulated control cells. In red pathways enriched with up-regulated genes; in blue pathways enriched with downregulated genes. Adjusted p values were calculated using Benjamini–Hochberg adjustment. d Mitochondrial morphology assessed by electron microscopy. e Mitochondrial mass as measured by mitochondrial (mt)-DNA/genomic (g)-DNA ratio. p value was calculated using Student’s t test compared to unstimulated cells. ns not significant. f Reactive oxygen species (ROS) levels as measured with carboxy-H2DCFDA. p value was calculated using Student’s t test compared to PBS control. ns not significant. Data in ac comprises three biological donors in duplicate from two independent experiments (n = 6). Data in e and f are displayed as box and whisker plots (n = 3–4 per group).

Glycolysis controls flagellin-induced activation of HBE cells

Considering the observed upregulation of genes involved in glucose metabolism we sought to investigate the functional role of glucose metabolism in HBE cell activation upon flagellin challenge. We first analyzed glucose uptake in HBE cells. After 24 h of flagellin stimulation, HBE cells showed increased glucose uptake compared to unstimulated control cells (Fig. 3a). To investigate whether the increased demand of glucose responded to a higher glycolytic metabolism, we assessed secretion of lactate, a hallmark of glycolysis. Indeed, 6 and 24 h after flagellin challenge, HBE cells secreted significantly more lactate than unstimulated control cells (Fig. 3b). The rapid induction of glycolysis upon flagellin stimulation prompted us to investigate how stable isotope-labeled 13C6-glucose is metabolized in HBE cells immediately following activation. In line with increased lactate accumulation, glucose-derived isotopologue of lactate (M + 3) was significantly increased 30 min after activation by flagellin (Fig. 3c). At this early time point the TCA intermediates malate (M + 2) and α-ketoglutarate (M + 2) were also increased in flagellin-stimulated cells, while pyruvate (M + 3) and fumarate (M + 2) were not affected (Fig. S4A). In total, 24 h after addition of flagellin, the intracellular levels of lactate were increased relative to medium control conditions, while the intracellular concentrations of pyruvate, fumarate, α-ketoglutarate, and malate were not altered (Fig. S4B). These results suggest that HBE cells display a rapid and sustained glycolytic response following flagellin activation. Additionally, supporting this observation, GSEA further confirmed the metabolic reliance on glycolysis in flagellin-stimulated HBE cells (Figs. 3d and S5).

Fig. 3: Glycolysis regulates flagellin-induced immune activation of HBE cells.
figure 3

Analysis of glucose uptake (a) and levels of lactate secreted to the culture medium (b) by HBE cells incubated with flagellin or PBS (control) for 6 or 24 h. c Glucose-derived carbon labeling of lactate in HBE cells 30 min after incubation with flagellin or PBS (control) in 13C-glucose-containing media. Bar graphs show mass isotopologue distribution of incorporated label. M + 0 bars represent the fraction of lactate with no 13C-glucose incorporation. d Gene set enrichment analysis (GSEA) of the glycolysis pathway in HBE cells activated with flagellin for 24 h compared to unstimulated control cells (PBS). The x-axis shows individual genes and the y-axis shows enrichment score. The 20 most enriched genes are listed. Red represents up-regulated genes and blue represents downregulated genes. Analysis of glucose uptake (e) and levels of lactate secreted to the culture medium (f) by HBE cells pretreated for 1 h with the glycolysis inhibitor 2-DG (5 mM) or vehicle and stimulated for 24 h with PBS or flagellin. CXCL8, CCL20, and CXCL1 levels in culture supernatant (g) and G-CSF in apical wash (h) of HBE cells incubated with flagellin or PBS (control) in the presence of 5 mM 2-DG of vehicle. Data in a, b, eh are presented as box and whisker plots. a, b, n = 5–7 per group from three donors in duplicate; c, n = 3 replicates; e, f, n = 3–6 replicates per group representative of two independent experiments; g, h, n = 3 biological donors representative of two independent experiments. p values were calculated using Student’s t test. *p < 0.05; **p < 0.01; ***p < 0.001, ns not significant. Data in d comprises three biological donors in duplicate from two independent experiments (n = 6).

To investigate the functional role of glycolysis in the flagellin response of HBE cells more directly, we next assessed the effect of glycolysis inhibition on glucose metabolism and cell activation after flagellin stimulation. As expected, glucose uptake after flagellin stimulation was reduced by the glycolysis inhibitor 2-deoxyglucose (2-DG) (Fig. 3e). The reduction in glucose uptake was accompanied by an impaired lactate production upon flagellin stimulation (Fig. 3f). This 2-DG induced inhibition of glycolysis limited the ability of HBE cells to secrete CCL20 and CXCL1 to the basolateral compartment (Fig. 3g) and G-CSF to the apical side (Fig. 3h). Taken together, these data demonstrate that glycolysis is triggered by flagellin and that this metabolic pathway is functionally required for flagellin-induced chemokine production by HBE cells.

Akt-mTOR axis is triggered by flagellin and controls the glycolytic function of HBE cells

The Akt-mTOR pathway is a master regulator of different cell functions, including glucose metabolism.20,21 Here, we aimed to determine whether the higher glycolytic rate in HBE cells induced by flagellin is under the control of the mTOR pathway. Flagellin triggered the Akt-mTOR pathway, as indicated by increased phosphorylation of Akt (Fig. 4a). Furthermore, the S6K1 substrate S6 ribosomal protein (S235/236) and 4E-BP1, two canonical downstream targets of mTOR,21 were activated upon flagellin stimulation (Fig. 4b). Treatment with rapamycin inhibited flagellin-mediated induction of the mTOR pathway, as shown by diminished levels of phosphorylated S6 and 4E-BP1 (Fig. 4b). Moreover, rapamycin treatment resulted in a significant reduction of lactate release after flagellin stimulation, indicating the involvement of the mTOR pathway in induction of glycolysis by flagellin (Fig. 4c). Likewise, two other mTOR inhibitors (torin-1, a mTORC1-2 inhibitor21; and BEZ235, a PI3K-mTOR dual inhibitor22) inhibited flagellin-induced lactate secretion, confirming the involvement of the mTOR pathway in the regulation of glycolysis in HBE cells and ruling out an off-target effect of rapamycin (Fig. 4d, e).

Fig. 4: Akt-mTOR axis is triggered by flagellin and controls the glycolytic function of HBE cells.
figure 4

a Immunoblot analysis of phosphorylated (p-) Akt and total Erk (loading control), in HBE cells from two donors, incubated with 1 μg/ml flagellin or PBS (control) for 2 h. b Immunoblot analysis of p-S6 and p-4EBP1, as well as total actin (loading control), in HBE cells from two donors (a, b) incubated with flagellin or PBS (control) for 2 h with or without preincubation (1 h) with 10 nM rapamycin. Levels of lactate secreted to the culture medium by HBE cells pretreated for 1 h with the mTOR inhibitors rapamycin (10 nM) (c), torin-1 (100 nM) (d), BEZ235 (100 nM) (e), or vehicle and stimulated for 24 h with PBS or flagellin. For ce, data are presented as box and whisker plots (n = 4–7 replicates per group including two individual donors). p values were calculated using Student’s t test. *p < 0.05; **p < 0.01; ***p < 0.001; ns not significant.

mTOR regulates HBE cell-immune function in response to flagellin

Our findings that glycolysis is necessary for chemokine release by HBE cells and that this metabolic pathway is under the control of mTOR prompted us to study the role of the mTOR pathway during immune activation of HBE cells through RNA sequencing analysis. This revealed profound rapamycin induced changes in the transcriptome of HBE cells activated by flagellin (Fig. S6A, B). Of interest, GSEA showed that rapamycin downregulated glycolysis (Fig. 5a), further supporting the involvement of the mTOR pathway in the regulation of this energy pathway. A direct comparison of flagellin stimulation in the presence or absence of rapamycin showed that the mTOR pathway is crucial for the immune function of HBE cells. Most importantly, rapamycin suppressed the expression of genes encoding for immune mediators such as chemokines and antimicrobial peptides (Fig. 5b; validation of selected mRNA’s by RT-PCR in Fig. S7). In line with these results, inhibition of the mTOR pathway limited the immune response of HBE cells typified by decreased CXCL8, CCL20, and CXCL1 to the basolateral side (Fig. 5c) as well as reduced secretion of G-CSF to the apical compartment (Fig. 5d). Together these data demonstrate that flagellin triggers mTOR-dependent glycolysis in HBE cells to induce an immune response.

Fig. 5: mTOR regulates HBE cell-immune function in response to flagellin.
figure 5

a Gene set enrichment analysis (GSEA) of the glycolysis pathway in HBE cells activated with flagellin for 24 h in the presence of either vehicle or 10 nM rapamycin (pretreated for 1 h). Log2-fold expression values were ranked in descending order. The x-axis shows individual genes and the y-axis shows enrichment score. The 20 most enriched genes are listed. Red represents up-regulated genes and blue represents downregulated genes. b Heat-map of changes in the expression of selected flagellin-induced genes in HBE cells after 24 h in the presence of 10 nM rapamycin relative to vehicle. Red denotes high expression, blue indicates low expression. c, d Secretion of CXCL8, CCL20, and CXCL1 to the basolateral compartment (c) and G-CSF to the apical side (d) by HBE cells in the presence of 10 nM rapamycin or vehicle added 1 h prior to addition of flagellin and stimulated for 24 h. Data in a, b comprise three biological donors in duplicate from two independent experiments (n = 6). Data in c, d are presented as box and whisker plots (n = 6 from three biological donors in duplicate). p values were calculated using Student’s paired t test. *p < 0.05.

mTOR controls flagellin-induced inflammation in vivo

To study the consequences of mTOR inhibition during lung innate immune responses in vivo, we administered rapamycin or vehicle intraperitoneally both 24 and 1 h before challenging mice with flagellin via the airways. Flagellin triggered the glycolysis pathway in bronchial epithelial cells as reflected by a twofold increase in the expression of Hk2, the first enzyme of glycolysis, in bronchial brushes; this flagellin effect was prevented by treatment with rapamycin, indicating that mTOR also controls glycolysis in bronchial epithelial cells in vivo (Fig. 6a). Flagellin induced the upregulation of epithelial cell-specific immune mediators such as Cxcl1, Cxcl2, and Csf3 in bronchial brushes, responses that were almost abrogated by rapamycin (Fig. 6b). This rapamycin effect was accompanied by inhibited flagellin-induced CXCL1 and CXCL2 release into bronchoalveolar lavage (BAL) fluid (Fig. 6c). G-CSF remained undetectable in BAL fluid after flagellin administration.

Fig. 6: mTOR inhibition impairs lung epithelial cell responses to flagellin in vivo.
figure 6

a, b Mice received rapamycin or vehicle 24 and 1 h prior to intranasal administration of flagellin or PBS and bronchial brushings were harvested 2 h after challenge for RNA extraction and qPCR analysis of the indicated genes. c CXCL1 and CXCL2 were measured in BAL fluid 2 h after the challenge with flagellin. Data are shown as box and whisker plots (n = 6 per group). p values were calculated using Student’s t test. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. ns not significant.


Immune cells can rewire their metabolic program to rapidly fulfill their energy requirements in order to sustain an effective immune response. Together with immune cells, the airway epithelium represents the first line of defense against inhaled pathogens. While knowledge on metabolic control of innate immune responses by different leukocyte subsets has rapidly expanded in recent years,4,5 data on the role of cellular metabolism in airway epithelium responses remains scarce. Here we used primary HBE cells to demonstrate that the common mucosal pathogen component flagellin triggers aerobic glycolysis that is required to sustain the production of inflammatory mediators. Specifically, we show that this metabolic adaptation is in part under the control of the Akt-mTOR pathway. Inhibition of the mTOR pathway prevented the induction of glycolysis and limited the secretory capacity of the HBE cells in response to flagellin. These results were recapitulated in a mouse model of airway inflammation induced by flagellin in vivo.

To the best of our knowledge our study is the first to report on energy metabolism pathways and mTOR function in primary human respiratory epithelial cells. The current results should be considered in the context of earlier studies on cellular metabolism in lymphocytes and myeloid cells.4,5,23,24 T-cell receptor mediated activation results in mTOR activation, and while naive and tolerant T cells primarily use oxidative phosphorylation to produce ATP, activation toward T-effector cells requires extensive metabolic rewiring and a shift toward glycolysis.23,24 Myeloid cells also experience metabolic reprogramming upon recognition of pathogens, which supports their activation and the initiation of an innate immune response. Whereas LPS elicits a similar metabolic shift toward glycolysis as observed in activated lymphocytes, other TLR ligands and intact pathogens may induce more complex alterations in energy metabolism pathways in monocytes.18 In myeloid cells mTOR is a master regulator of metabolism, which can be activated by a variety of stimuli relevant in the context of infection, including TLR ligands and cytokines.7 Intriguingly, activation of mTOR can have contrasting roles in distinct cells or models. The seemingly contrasting roles of mTOR in individual cells or models might have several explanations, including differences in experimental conditions, such as the analysis of primary cells versus cell lines, and differences between mice and humans concerning the importance of the arginine and nitric oxide system, which is regulated by mTOR.7 In addition, the mTOR pathway is characterized by a high degree of complexity. As an example, inhibition of mTORC1 activates the PI3K-AKT pathway; distinct AKT isoforms stimulate different macrophage polarization states and the phosphorylation status of AKT changes substrate specificity.25 Moreover, in different in vivo settings the relative importance of the mTOR pathway for an immune response versus other host mediators impacting this response determines the “inflammatory” contribution of mTOR to that response.7 In myeloid cells inhibition of mTOR stimulates IL-12 expression while reducing IL-10 expression, suggesting an anti-inflammatory role for mTOR.26 In contrast, tumor necrosis factor (TNF) and IL-6 production by myeloid cells are mostly blocked by rapamycin, indicating a pro-inflammatory role for mTOR.7 Similarly, mTOR positively contributes to the expression of pro-inflammatory CXCL8, IL-6, and TNF in neutrophils.27 We here report that mTOR clearly has a pro-inflammatory immune enhancing role in HBE cells stimulated with flagellin, as indicated by reduced expression of a large set of antimicrobial peptides and immune mediators in cells treated with rapamycin. In support of a pro-inflammatory role of mTOR in epithelial cells, a previous study using a respiratory epithelial cell line responsive to LPS reported reduced LPS-induced IL-6 and CXCL8 production after knockdown of mTOR by siRNA.28 In this latter study28 LPS was able to activate epithelial cells, contrasting with our data, which might be due to differences in the cells used (an HBE cell-line grown in submerged culture medium versus differentiated primary human HBE cells grown in air/liquid interface). One earlier study investigated the effects of flagellin on cellular metabolism: flagellin elicited mTOR activation in a mouse macrophage cell line and inhibition of mTOR by rapamycin diminished flagellin-induced TNF and IL-6 release by these cells.29

We did not find evidence for an effect of flagellin on mitochondrial function in primary HBE cells. In a previous study flagellin was reported to induce mitochondrial dysfunction and increased ROS production in IB3-1 cells, an immortalized cell line created from bronchial epithelial cells isolated from a patient with cystic fibrosis.30 Thus, flagellin effects on mitochondrial function in respiratory epithelial cells may differ between different experimental conditions.

The effects of rapamycin in HBE cells were partially reproduced in mouse respiratory epithelial cells in vivo, as reflected by strongly reduced Cxcl1, Cxcl2, and Csf3 expression in epithelial brushes and attenuated CXCL1 and CXCL2 protein release in BAL fluid from mice administered with flagellin via the airways. These results are in agreement with an earlier study showing that selective disruption of Mtor in the respiratory epithelium of mice was associated with reduced release of CXCL1 and CXCL2 in BAL fluid upon intratracheal injection of LPS28; it would be of great interest to study the effects of flagellin in these genetically modified mice. Furthermore, rapamycin exerted protective anti-inflammatory effects in mouse models of LPS-induced acute lung injury.28,31

Our investigation only studied the metabolic consequences of stimulation of HBE cells with flagellin, which has relevance for respiratory pathogens expressing this TLR5 ligand (most notably P. aeruginosa) as well as in the context of studies evaluating flagellin as a potential immune stimulating adjuvant in the treatment of airway infection.14 We here showed that besides UV-inactivated P. aeruginosa also K. pneumoniae (an un-flagellated bacterium) activates primary HBE cells. Further research is needed to assess whether intact pathogens induce similar or more complex changes in energy metabolism pathways, such as has been documented in monocytes exposed to whole bacteria.18 Our data do not proof that flagellin effects are mediated by TLR5. A role for the alternative flagellin receptor NLRC4 seems less likely considering the absence of IL-1β protein release in the presence of clear IL1B gene induction. Notably, primary human airway epithelial cells express high levels of TLR5 at their surface32,33 and in mice flagellin promoted epithelial cell responses in a TLR5-dependent manner that did not require NLRC4 signaling.16

mTOR consists of two complexes: mTORC1 and mTORC2.7 Of these mTORC1 has been implicated in the regulation of metabolism and immune functions, while mTORC2 regulates cytoskeletal dynamics and activates AKT, which in turn can activate mTORC1. Rapamycin is a selective mTORC1 inhibitor and thus our findings are in agreement with a central role for mTORC1 in cellular metabolism and immunity.7

The respiratory epithelium represents a first line of defense in the lungs and plays an essential role in mounting an innate immune response to pathogens that invade the lower airways. We here establish the mTOR-glycolysis axis as a key metabolic regulator of the host immune function in the airway epithelium in response to flagellin. These results increase current knowledge on mechanisms underlying activation of the respiratory epithelium during airway inflammation and infection and may guide selection of therapeutic targets in these conditions.


Primary HBE cell culture

HBE cells were obtained anonymously from three patients undergoing a lobectomy for lung cancer at the Academic Medical Center, the Netherlands. Healthy tracheobronchial tissue was isolated distant from tumorous tissue by a pathologist and the absence of malignant cells was checked by microscopy. The Institutional Review Board of Amsterdam University Medical Center approved the study protocol (2015-122#A2301550) and written informed consent was obtained from the patients before sampling. Epithelial cells were isolated following the Fulcher’s protocol.11 Briefly, passage 2 (P2) to P4 passaged HBE cells were expanded in human type IV placental collagen coated porous support 24-well Transwell inserts (Corning, Corning, NY, USA) for differentiation in PneumaCult-Ex Plus media (StemCell Technologies, Vancouver, Canada). When the cells were confluent in the inserts, the media was changed to PneumaCult-ALI medium (StemCell Technologies) containing penicillin and streptomycin on the basolateral side and the media on the apical side was removed, forming an air–liquid interface. The media were renewed every 2 or 3 days until full differentiation. Well-differentiated HBE cells were used on day 30 for stimulation in PneumaCult-ALI medium.

Cell stimulation

HBE cells were stimulated in PneumaCult-ALI medium with 1 μg/ml flagellin from Pseudomonas aeruginosa (Invivogen, Toulouse, France), 100 ng/ml LPS from Klebsiella pneumoniae (L1519; Sigma, St. Louis, MO, USA), 1:2 (cell:bacteria) of UV-inactivated Klebsiella pneumoniae (ATCC 43816), Pseudomonas aeruginosa (ATCC 47085), or PBS added to the apical compartment. Unless stated otherwise, HBE cells were stimulated for 24 h. HBE cells were preincubated for 1 h prior to stimulation with vehicle, the glucose analog 2-DG (5 mM; Sigma) to inhibit glycolysis or the mTOR inhibitors rapamycin (10 nM; Cayman Chemical, Ann Arbor, MI, USA), BEZ235 (100 nM; BioVision, Milpitas, CA, USA) or torin-1 (100 nM; Tocris, Bristol, UK) added to the culture medium. After 1 h the bacterial stimuli (or PBS) were added to the apical side. Cell culture supernatants were stored at −80 °C until being analyzed. Cells were lysed in Lysis/Binding Buffer (Roche, Basel, Switzerland) and stored at −80 °C for RNA isolation or directly stored at −80 °C until processed for western blot or metabolite measurements as described below.

Cytokine measurements

Human CXCL8, CCL20, CXCL1 and G-CSF, and mouse CXCL1 and CXCL2 were measured by ELISA according to manufacturer’s instructions (R&D Systems, Minneapolis, MN, USA).

Reactive oxygen species and cell viability

For measurement of ROS, cells were stimulated for 24 h with vehicle (PBS) or flagellin and for the last 30 min, 10 μM of Carboxy-H2DCFDA (Thermofisher, Waltham, MA, USA) was added. After stimulation, HBE cells were digested for 30 min at 37 °C with 5 mg/ml Liberase TM (Sigma, St. Louis, MO, USA) and 10 mg/ml DNase (Roche). Data were acquired on a FACS Canto II cytometer (Becton Dickinson, Franklin Lakes, NJ, USA) and analyzed using FlowJo software (Treestaar Inc, Ashland, OR, USA). Fixable Viability Dye kit (eBioscience) was used to assess cell viability (>90% in all conditions).

Glucose and lactate measurements

Glucose and lactate were quantified using an enzymatic assay in which glucose and lactate were oxidized and the resulting H2O2 was coupled to the conversion of Amplex Red reagent to fluorescent resorufin by horseradish peroxidase (HRP) as described in ref. 18

13C-glucose labeling

For glucose labeling experiments, glucose-free RPMI media was supplemented with 2 mM [U-13C]-glucose (Cambridge Isotope Laboratories, Tewksbury, MA, USA). Samples were separated in two phases as described previously.34 Briefly, the aqueous phase was evaporated and the metabolite residue was processed on a SeQuant ZIC-cHILIC column (100 × 2.1 mm, 3 mm particle size, Merck, Darmstadt, Germany) coupled to a Bruker Impact II mass spectrometer (Bruker Daltonic, Bremen, Germany). Data were analyzed using Bruker TASQ 2.0 software (Thermo Scientific, Waltham, MA, USA). 13C enrichment was calculated based on mass distribution isotopomer analysis, results were corrected for their natural 13C abundance by solving the associated set of linear equations for lactate using non-negative least squares.35

Steady-state metabolomics

Metabolomics was performed as described previously.36 For details see Supplementary Information.

mRNA extraction and RT-PCR

Total RNA was extracted from HBE cells using the High Pure RNA Isolation Kit (Roche) following the manufacturer’s instructions. Reverse transcription was performed using the M-MLV Reverse Transcriptase (Promega, Madison, WI, USA) in the presence of RNase inhibitor (ThermoFisher) with 300 ng of DNase I (Roche) treated total RNA. RT-PCR was performed on LightCycler 480 (Roche) using the SensiFAST SYBR No-ROX Kit (Bioline, London, UK). Data are expressed as fold increases in expression, relative to the non-stimulated condition after normalization to HPRT as a housekeeping gene. Mitochondrial mass was determined as described in ref. 37 Briefly, gDNA was isolated using the AllPrep DNA/RNA mini kit (Qiagen, Hilden, Germany) and the ratio mtDNA/gDNA was calculated.

Total RNA was extracted from mouse epithelial brushes with Nucleospin RNA kits (Macherey-Nagel, Düren, Germany). RNA was reverse-transcribed with the High-Capacity cDNA Archive Kit (Applied Biosystems, Foster City, CA, USA). Gene expression was quantified by cDNA amplification in a QuantStudio 12 K real-time PCR system (Applied Biosystems), using Power SYBR® Green PCR Master Mix (Applied Biosystems). Relative mRNA levels were determined by comparing (1) the cycle thresholds (Ct) for the gene of interest and three calibrators Actb, B2m, and Rpl13a (∆Ct), and (2) the 2−∆Ct values for treated and control groups. The upper boundary for Ct was set to 35 cycles. Primers used in this study are listed in Table S1.

Immunoblot analysis

HBE cells were lysed in RIPA buffer supplemented with HALT protease and phosphatase inhibitor (ThermoFisher) and stored at −20 °C until processing. Samples were heat-denatured in reducing sample buffer (Laemmli buffer). Samples were loaded on either 10% or 4–20% polyacrylamide precast gels (Bio-Rad, Hercules, CA) and transferred to PVDF membranes. After incubation for 1 h with blocking buffer, immunoblotting was performed using antibodies to β-Actin (4967L; Cell Signaling, Danvers, MA), Erk (9107S; Cell Signaling), phospho S6 Ribosomal Protein (Ser235/236) (2211S; Cell Signaling), phospho-Akt (4060S; Cell Signaling), or phospho-4EBP1 (9459S; Cell Signaling). Goat anti-rabbit (7074S; Cell Signaling) and goat anti-mouse (7076S; Cell Signaling) HRP antibodies were used as secondary antibodies. Blots were incubated with the Lumi-Light detection kit (Roche) and pictures were taken using ImageQuant LAS-4000 (GE Healthcare, Chicago, IL, USA).

RNA library preparation and sequencing

RNA was isolated from HBE cells (from three distinct donors in duplicates for each condition) as described above. RNA quality was assessed by bioanalysis (Agilent, Santa Clara, CA), with all samples having RNA integrity numbers >7. Total RNA concentrations were determined by Qubit® 2.0 Fluorometer (Life Technologies, Carlsbad, CA, USA). Sequencing libraries were prepared by means of the KAPA RNA HyperPrep with RiboErase (Roche) as per manufacturer’s instructions. Libraries were sequenced using the Illumina HiSeq4000 (Illumina, San Diego, CA, USA) to generate 50 bp single-end reads.

Bioinformatics analysis of RNAseq

The sequence read quality was assessed using FastQC methods (version 0.11.5; Babraham Institute, Babraham, Cambridgeshire, UK). Trimmomatic version 0.3238 was used to trim the Illumina adapters and filter low quality reads and ambiguous nucleotide-containing sequences. Low quality leading (three nucleotides) and trailing (three nucleotides) bases were removed from each read. A sliding window trimming using a window of four and a phred score threshold of 15 nucleotides was used to access the quality of the reads. After preprocessing, the remaining high-quality reads were aligned against the Genome Reference Consortium Human Genome Build 38 patch release 7 (GRCh38.p7) using Bowtie2 version with default parameters. Count data were generated by means of the FeatureCounts method,40 and differential expression analyzed using the DESeq2 method41 in the R statistical computing environment (R Core Team 2014. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria).

Throughout significance was calculated using Benjamini–Hochberg adjusted p values.42 Using the Reactome pathway knowledgebase,43 the GSEA approach19 was applied to determine, for any given pathway, whether a predefined set of genes (Reactome pathways) showed statistically significant differences between two conditions.

Electron microscopy

To assess mitochondrial morphology HBE cells were fixed in Mc Dowell containing paraformaldehyde and glutaraldehyde (Polyzciences, Inc., Warrington, PA) embedded in gelatin and small blocks of gelatin embedded cells were postfixed with 1% osmium tetroxide (OsO4, Electron microscopy sciences, Hatfield, PA; in cacodylate buffer). Subsequently, the samples were dehydrated in an alcohol series and embedded into Epon (LX-112 resin Ladd research, Wiliston, VT) and 70 nm thin sections were made with an ultramicrotome (Leica UC6). Sections were analyzed with a FEI Tecnai-12 G2 Spirit Biotwin electron microscope (Fei/ThermoFisher, Eindhoven, The Netherlands), and images were made with a Veleta camera using Radius Software (EMSIS, Münster, Germany).


Six- to eight-week-old female C57BL/6JRj mice (Janvier Labs, Saint-Berthevin, France) were housed in a pathogen-free facility in compliance with institutional guidelines (#C59-350009, Institut Pasteur de Lille). All experimental procedures were approved by the local animal care and use committee and by the French Ministry of Research (authorization APAFIS#5165—2015121722376405_v2). Mice were administered with rapamycin intraperitoneally (4 mg/kg; Cayman Chemical) or vehicle (2% DMSO in PBS) 24 and 1 h prior to intranasal administration of 0.5 μg ultrapure flagellin from P. aeruginosa (Invivogen). After 2 h, BAL fluid samples were obtained as described before.17 Briefly, the trachea was cannulated via a midline incision and 2 × 0.3 ml of sterile PBS was instilled to retrieve the BAL fluid. Thereafter, bronchial brushes were performed as described by Chen et al.44 and qPCR analyses were performed as described above. In short, sanded PE50 tubing (P240 sandpaper) treated with RNaseZap were inserted into the main bronchus via the trachea with gentle brushing and immediately processed for RNA extraction.

Quantification and statistical analysis

All the analyses were done using GraphPad Prism 7.03. The number of replicates and the statistical tests used for each data set are described in the figure legends. In most cases, Student’s t test was used. A p value <0.05 was considered statistically significant.