Intermittent exposure to whole cigarette smoke alters the differentiation of primary small airway epithelial cells in the air-liquid interface culture

Cigarette smoke (CS) is the leading risk factor to develop COPD. Therefore, the pathologic effects of whole CS on the differentiation of primary small airway epithelial cells (SAEC) were investigated, using cells from three healthy donors and three COPD patients, cultured under ALI (air-liquid interface) conditions. The analysis of the epithelial physiology demonstrated that CS impaired barrier formation and reduced cilia beat activity. Although, COPD-derived ALI cultures preserved some features known from COPD patients, CS-induced effects were similarly pronounced in ALI cultures from patients compared to healthy controls. RNA sequencing analyses revealed the deregulation of marker genes for basal and secretory cells upon CS exposure. The comparison between gene signatures obtained from the in vitro model (CS vs. air) with a published data set from human epithelial brushes (smoker vs. non-smoker) revealed a high degree of similarity between deregulated genes and pathways induced by CS. Taken together, whole cigarette smoke alters the differentiation of small airway basal cells in vitro. The established model showed a good translatability to the situation in vivo. Thus, the model can help to identify and test novel therapeutic approaches to restore the impaired epithelial repair mechanisms in COPD, which is still a high medical need.

A COPD-like phenotype is conserved in ALI cultures derived from donors with COPD. SAECs from three different COPD donors and three healthy controls were used to test whether pathologic changes in the epithelial histology are preserved in the differentiated ALI cultures. Of note, histological slices from ALI cultures obtained from COPD patients show a reduced number of ciliated cells and an increased number of secretory cells in comparison to control ALI cultures from healthy donors (Fig. 1a). To characterize the differential mRNA expression in detail, a Next Generation Sequencing (NGS) analysis was performed of ALI cultures obtained from healthy donors and COPD patients at day 28 upon air-lift. 248 transcripts were detected that were significantly changed in COPD-derived cells compared to healthy volunteer-derived cells. Hierarchical clustering of these transcripts reveals different mRNA signatures in healthy and COPD cultures (Fig. 1c). To assess the ability of the COPD ALI signature to distinguish between human small airway epithelial (SAE) samples of healthy non-smokers and COPD smokers as well as between healthy smokers and COPD smokers, the signature was used as feature vector in a random forest classification analysis 36 . The power of the signature to separate the respective samples was tested by determining the ROC (receiver operating characteristic) curves (Fig. 1d). Classifying samples from healthy non-smokers and COPD smokers resulted in an AUC of 1.0, indicating a nearly perfect separation of SAE samples by utilizing the ALI culture derived signature. To ensure, that the high classification power is not driven solely by very few well separating transcripts, but rather by a large proportion of the signature, the best correlating transcripts were identified. Subsequently, ten transcripts were removed from the signature, which showed the highest correlation. After reiterating the classification, an AUC of 0.98 was obtained. This indicates a clear relationship between the transcriptional changes within the ALI culture and the SAE samples, which is not dependent upon the top ten classifying transcripts. The high overlap between the in vitro derived signature and the clinical situation allows a good prediction of the health status. In the task of separating smokers from COPD smokers an AUC of 0.8 was obtained. Upon removal of the ten highest correlating transcripts, this drops to 0.65. Even though the classification power is lower in this case, it still indicates an overlap of COPD driven transcriptional changes in the ALI samples with the changes observed in SAE from human subjects.
Cigarette smoke exposure impairs epithelial barrier integrity. To investigate the harmful effects of CS on SAEC differentiation and physiology the ALI cultures were intermittently exposed to well-defined doses of whole CS during the differentiation period of 28 days. Unlike commonly used exposure models that use cigarette smoke extract (CSE) in which many volatile components are lost during the extraction process, using whole CS exposes the cells to all volatile toxins contained within CS. Furthermore, this is likely to focus CS exposure to the apical cell layer and reduce exposure to the basal cell layer, in contrast to a stimulation with CSE added to the basal medium. It is assumed that this setting is an optimal framework to mimic the direct toxic effects of smoking on the small airway epithelium.
In order to assess effects of CS on epithelial barrier integrity, TEER was repeatedly measured in the ALI cultures during differentiation. Air control treated ALI cultures of healthy donors and COPD patients exhibited similar values at days 7, 14, 21 and 28 (Fig. 2a). However, the TEER values of CS exposed cultures were reduced compared to air controls (Fig. 2a). To quantify this effect the area under the curve (AUC) was calculated and standardized with respect to the time span. The adjusted AUC dropped in the healthy ALI cultures from 535 ± 91 Ω*cm 2 in air controls to 354 ± 36 Ω*cm 2 (p = 0.01) upon CS exposure, and in the COPD ALI cultures from 500 ± 143 Ω*cm 2 to 337 ± 50 Ω*cm 2 (p = 0,017), respectively (Fig. 2a). To investigate the underlying molecular mechanisms mRNA was extracted from the ALI cultures and the expression of junctional protein transcripts was analyzed. Some genes associated with cell adhesion, permeability and tight junction regulation were deregulated Absolute expression values were normalized to a range from 0 to 1 of SAEC ALI cultures. Numbers represent samples from same subjects (donor 1-3 HC and COPD, respectively). (d) ROC (receiver operating characteristic) curves and confusion matrices to characterize the classification power of the in vitro ALI COPD signature (predicted condition) towards a published data set from epithelial brushes of patients with COPD and healthy controls (true condition). A random forest classifier with 50-fold cross validation was used to classify COPD (n = 38) vs. non-smoker (n = 53) and COPD (n = 38) vs. healthy smoker (n = 50). SAECs from an independent cohort utilizing the ALI COPD signature as feature vector. Dashed regions indicate AUCs observed for the full signature, grey regions AUCs upon removal of the ten highest correlating transcripts. Numbers in parentheses represent values obtained for the reduced signature.
Cigarette smoke exposure induces squamous differentiation. One hallmark of COPD is squamous metaplasia 21 . Therefore, the expression of different squamous cell marker genes was investigated during the differentiation of smoked and control ALI cultures (RNA-seq, Fig. 3c). During the differentiation of the ALI cultures, the difference of threshold cycle values (∆Ct: Ct target gene -Ct endogenous control) for Keratin 5 (KRT5) increased from the day of airlift until day 14, where it reached a plateau, indicating that KRT5 mRNA expression is downregulated during the normal differentiation process. This was true for ALI cultures from healthy and COPD SAECs. However, the exposure to CS led to a decrease of the ∆Ct values which reflects an increase of KRT5 mRNA expression in healthy as well as in COPD ALI cultures (Fig. 3a). To quantify this effect over the observed time the AUC was calculated and standardized with respect to the time span. The exposure of ALI cultures to CS led to a decreased AUC which was similar in healthy (from −2.3 ± 0.4 ∆Ct to −2.8 ± 0.3 ∆Ct, p = 0.0004) and COPD derived cultures (from −2.6 ± 0.1 ∆Ct to −3.0 ± 0.1 ∆Ct, p = 0.0004) (Fig. 3b). The ∆Ct values for the squamous differentiation marker KRT14 increased during the ALI differentiation indicating that KRT14 mRNA expression is downregulated during the normal differentiation process. However, the ∆Ct values were not down regulated during the differentiation of CS exposed cultures (Fig. 3b) indicating that CS-exposure The exposure with CS led to a significant change of mRNA expression of genes related to various aspects of epithelial barrier integrity: JAM3, CLDN11, CLDN18, CLDN7 and CLDN10. Data obtained from NGS analysis. Significance of deregulation according to a linear model applied to the normalized data is indicated as * for q value <0.05; ** for q value <0.01; *** for q value <0.001. n = 3 HC and 3 donors with COPD. (c) Representative images of an immunohistochemical staining for Claudin-10 (CLDN10) from a healthy SAEC ALI culture section at day 28 after air-lift with intermittent CS exposure (smoke) and air for control and semiquantification using automated image analysis.
is blocking the physiologic downregulation of KRT14 expression. Accordingly, the AUC was downregulated upon CS exposure in ALI cultures from COPD patients (from 5.0 ± 0.3 ∆Ct to 3.3 ± 0.3 ∆Ct, p < 0.0001) and healthy controls (from 5.5 ± 0.8 ∆Ct to 3.1 ± 0.9 ∆Ct, p < 0.0001) (Fig. 3b). RNA sequencing analysis revealed deregulation of several cell marker genes, i.e. genes associated with basal cells, goblet and club cells as well as ciliated cells (Fig. 3c). KRT5 and KRT14 up-regulation was confirmed and the expression of other genes associated with squamous differentiation, namely Keratin 6 A (KRT6A), Keratin 13 (KRT13), and Stratifin (SFN) were found to be also up-regulated upon CS exposure, indicated by a positive fold-change (Fig. 3c). Besides evaluation on RNA Cigarette smoke exposure stimulates the differentiation of MUC5AC + cells and reduces SCGB1A1 + cells. To investigate the effect of CS on basal cell differentiation into secretory cells, the club cell marker SCGB1A1 and the goblet cell marker MUC5AC were analyzed on mRNA level at day 7, 14, 21 and 28 after air-lift. The ΔCt values decreased after the day of air-lift, indicating an induction of MUC5AC and SCGB1A1 mRNA expression, which goes along with the normal differentiation process. The CS exposed cultures exhibit lower ΔCt values for MUC5AC, indicating an increased MUC5AC mRNA expression. At the same time, ΔCt values for SCGB1A1 were increased during differentiation, indicating a reduced SCGB1A1 mRNA expression (Fig. 4a). The adjusted AUC was calculated to compare the differentiation into secretory cells of smoke exposed cultures and air controls. The AUC for MUC5AC was down-regulated upon CS exposure in ALI cultures from COPD patients (from −1.1 ± 0.9 ∆Ct to -1.5 ± 0.5 ∆Ct, p = 0.3252) and healthy controls (from −1.3 ± 0.6 ∆Ct to −2.0 ± 0.5 ∆Ct, p = 0.2080). This indicates an approximate increase of MUC5AC mRNA expression of 1.3 fold in the COPD cultures and 1.6 fold in the healthy ALI cultures. The AUC for SCGB1A1 was up-regulated in CS-exposed ALI cultures from COPD patients (from −10.3 ± 0.03 ∆Ct to −8.9 ± 0.8 ∆Ct, p = 0.0289) and healthy controls (from −9.9 ± 0.1 ∆Ct to −9.3 ± 0.1 ∆Ct, p = 0.1317). This indicates an approximate reduction of SCGB1A1 mRNA expression of 2.6 fold in the COPD cultures and 1.6 fold in the healthy ALI cultures. Furthermore, a RNA sequencing analysis indicated an increase of goblet cell associated genes (MUC5AC, MUC1) and a reduction of club cell genes (SCGB1A1, SCGB3A1, CYP2F1, TMEM45A) upon CS treatment (Fig. 3c).
To investigate if the changes on mRNA level led to altered protein expression and cell composition MUC5AC and SCGB1A1 protein expression were analyzed by means of FACS analysis. The analysis revealed for the majority of donors (five out of six) a decrease in SCGB1A1 + single positive cells after 28 days with intermittent CS treatment (Fig. 4b,c). The effect on MUC5AC + single positive and MUC5AC + /SCGB1A1 + double positive cells was not significant. Still half of the donors showed a clear increased expression of MUC5AC (however, two donors showed a decreased number of MUC5AC + /SCGB1A1 − cells). Next, it was assessed whether CS exposure induces histological differences of the ALI cultures by performing immunohistochemical stainings of sections. Unexpectedly, CS exposed cultures as well as air controls exhibited a differentiated pseudostratified epithelium without significant differences in the epithelial thickness. However, the analysis confirmed the previous findings, that the number of SCGB1A1 + cells decreased in the smoke exposed ALI cultures compared to air controls, whereas the number of MUC5AC + cells increased, respectively (Fig. 4d).

Cigarette smoke exposure impairs the development of ciliated cells. The development of ciliated
cells was monitored by measuring the area covered by beating cilia and their corresponding beat frequency at days 7, 14, 21 and 28. Moreover, the expression of the transcription factor FOXJ1 and the presence of acetylated tubulin was measured, since proper cilia function, changes in transcript levels and changes in protein levels may not necessarily correspond to each other. COPD-derived cultures exhibited a delayed and reduced differentiation of ciliated cells compared to healthy cultures, indicated by a reduced area covered by actively moving cilia, and reduced acetylated tubulin expression (data not shown). These parameters were further reduced by intermittent CS treatment (Fig. 5a-c). Interestingly, cigarette smoke exposure impaired the development of ciliated cells in healthy controls in almost the same manner. To quantify the reduction in ciliary beating, the AUC was calculated and standardized with respect to the time span. Both, the area covered by actively beating cilia and the cilia beat frequency were reduced by intermittent CS treatment (Fig. 5b). FOXJ1 mRNA increased during the normal differentiation process, indicated by a decrease in the ΔCt values (Fig. 5c). This induction was delayed in CS exposed cultures. To quantify the CS-induced effect on FOXJ1 mRNA expression the AUC was calculated and standardized with respect to the time span (Fig. 5c). Furthermore, FOXJ1 and DNAI1 expression was assessed by RNA sequencing on day 28. The analysis revealed a non-significant deregulation of the ciliated cell markers on RNA level upon 28 days of intermittent CS exposure (Fig. 3c). To investigate if CS exposure not only affects the cilia function and mRNA expression but also the cell composition acetylated tubulin was stained as a cilia marker in the histological slices. Air control cultures exhibited a well-differentiated ciliated epithelium with many acetylated tubulin-positive cilia. CS-exposed ALI cultures revealed less cilia (Fig. 5d).

CS exposure lead to the deregulation of a similar subset of genes in healthy and COPD derived
ALi cultures. To elucidate the transcriptional response of the small airway epithelium to CS a NGS analysis was performed. To address the question whether COPD derived cultures are differently affected by smoking compared to healthy cultures the transcriptional response upon CS exposure was compared by performing a differential expression analysis of the NGS data. CS modulated the expression of 2442 genes in healthy cultures and 3249 genes in COPD cultures. Notably, 1945 genes were regulated in healthy as well as in COPD derived cultures. That all but one of the 1945 overlapping transcripts were regulated the same way indicates that the main stimulus of altered gene expression in the study was CS and not the disease status of the donors, i.e. COPD vs. healthy.
To strengthen the translational link of the in vitro system with the situation in vivo, the CS-induced transcriptional changes of the healthy ALI cultures were compared with a published data set of epithelial brushes from healthy smokers and non-smokers (GSE11784) 38 to assess the overlap of deregulated genes. Therefore, the normalized gene expression data (range from 0 to 1) from SAE and ALI cultures were combined and filtered for those transcripts that were significantly deregulated in both comparisons: a) smoked ALI cultures versus air controls and b) SAE from healthy smokers versus SAE from non-smokers. Subsequently an unsupervised, hierarchical clustering was performed using correlation as distance measure on the combined and filtered data-set (Fig. 6a). This approach allowed the separation of different clusters according to their similarities. Two distinct clusters representing the in vitro samples from CS and air-treated ALI cultures were separated. Of note, most of the SAE  www.nature.com/scientificreports www.nature.com/scientificreports/ samples from smokers clustered with the smoked ALI cultures whereas the SAE from non-smokers clustered with the ALI air-controls. This unbiased approach (which is only focusing on transcripts that were significantly deregulated but not filtering for common directionality of deregulation) indicates that CS alters overlapping molecular pathways in SAEC ALI cultures and in SAE from smokers.
To quantify to what extent the smoked ALI culture resembles the SAE of smokers a correlation analysis was performed (Fig. 6b). More than 86% of the genes that were significantly deregulated in the CS-exposed ALI cultures and the SAE of healthy smokers showed a consistent direction of deregulation (Fig. 6c). Furthermore, a pathway analysis was performed to elucidate the common mechanisms affected by CS-treated ALI in vitro and SAE from smokers. In order to allow for identifying pathways that were altered in both data sets, the significantly deregulated transcripts that exhibit the same direction of change were combined. With this set of transcripts, a core analysis was performed utilizing Ingenuity Pathways Analysis (IPA). A large number of top scoring canonical pathways were categorized to "xenobiotic metabolism" (supplementary Table S3). Therefore, all transcripts that were reported to be associated to xenobiotic metabolism were extracted and used to generate a combined network representing the in vivo and in vitro data (Fig. 7). Among the top deregulated canonical pathways associated to the network and consistent with the in vivo SAE data were Nuclear factor erythroid 2-related factor 2 (NRF2)-mediated oxidative stress, xenobiotic metabolism and aryl hydrocarbon (AHR) signaling (Figs. 7 and 8).

Discussion
COPD is a disease that strongly affects the small airways 17,39 . To investigate the underlying pathophysiological and molecular mechanisms in vitro, a new protocol for SAEC ALI cultures and novel methods for their analysis were established. In contrast to multi-layered bronchial epithelial cell cultures, the SAEC ALI culture exhibit a two-layered pseudostratified epithelium closely resembling the human small airway in vivo 40 . This model allows culturing SAEC from healthy donors and COPD patients with the latter displaying some conserved pathological characteristics, such as reduced numbers of ciliated cells, increased numbers of secretory cells and a distinct gene signature. The analysis of the gene signature of the ALI cultures derived from healthy or COPD donors with in vivo samples from human SAE revealed a high classification power and therefore a high predictability of the health status. www.nature.com/scientificreports www.nature.com/scientificreports/ A pathologic hallmark of COPD is a reduced airway epithelial barrier function leading to an increased susceptibility to infections [18][19][20][24][25][26] . In this model, CS exposure during SAEC differentiation impaired epithelial barrier formation characterized by a decreased TEER. This is consistent with previous studies investigating the effect of cigarette smoke extract on HBEC ALI cultures 27 . The observed functional decline of epithelial barrier function may be linked to the deregulation of several genes that are associated with the maintenance and control of cell-cell adhesion, permeability and tight junctions. An important molecular role in these aspects play the Claudin family members, which are tight junction membrane proteins that are conceptually divided into barrier-forming (e.g. Claudin-11, -18) and pore-forming (e.g. Claudin-7, -10) proteins [41][42][43] . Notably, Claudin-11 and -18, which are known to seal paracellular permeability 41 , were down-regulated in the CS exposed cultures. Claudin-18 deficiency was also described to be associated with airway epithelial barrier dysfunction and asthma 44 . Furthermore, Claudin-10, which has been unequivocally defined as pore-forming claudin 41,45 was up-regulated. Claudin-10 and Claudin-7 were also shown to be up-regulated in smokers and COPD smokers, respectively 46 . Apart from claudins, the junctional adhesion molecule C (JAM3), a component of desmosomes, which mediates cell-cell adhesion 47 was described to be down-regulated in COPD patients. JAM3 was also decreased in the ALI cultures after 28 days of CS exposure. Taken together, different genes associated with the regulation of epithelial barrier function were deregulated in the CS-treated ALI cultures that may be associated with the pathogenesis in COPD.
Basal cell hyperplasia and squamous metaplasia play an essential role in the emergence of airway obstruction in COPD 17 and increase the risk to develop squamous cell carcinoma [48][49][50] . Previous studies have found increased squamous metaplasia in the bronchial epithelium of smokers with COPD compared to healthy smokers 21 . In this study, an increase in several genes associated with squamous metaplasia, namely KRT5, KRT6A, KRT13, KRT14 and SFN was observed indicating that CS plays a role in squamous differentiation. These in vitro findings are in accordance with histological (known for a long time 51 ) and expression analyses (more recent ones 52 ) of COPD patients indicating that basal cells are the first cell population affected in smokers and leading to an altered differentiation pattern and architecture of the small airway epithelium. Taken together, the model can be used to investigate the molecular mechanisms underlying the early effects of smoking on basal cell proliferation.  www.nature.com/scientificreports www.nature.com/scientificreports/ findings indicate a shift in the differentiation pattern of the small airways of smokers and COPD patients towards a more proximal airway epithelium 35,55 ; an effect recently described as smoking-dependent distal-to-proximal repatterning 56 . Goblet cells for example were found in the small airways of smokers 6 , where they are normally absent. This aspect of the disease was found in the smoked SAEC ALI cultures. CS exposure led to a reduction of club cell markers (SCGB1A1, CYP2F1) and a downward trend of MUC5AC − /SCGB1A1 + cells in the ALI cultures. This may reflect the pathophysiological decline in club cells observed in COPD. The reduced club cell numbers in the airways of smokers and COPD patients, respectively 7,15 , are associated with reduced uteroglobin (SCGB1A1) serum levels, which were correlated to a decline in FEV1 (Forced Expiratory Volume in 1 Second) and COPD disease progression 14,52 . In contrast to our findings, the exposure to CSE did not show a reduction in club cell numbers 27 . Therefore, we think that the apical exposure of the SAEC ALI culture with whole CS leads to a better representation of the pathophysiological processes induced in human smokers. The model may help to dissect the molecular mechanisms, leading to the impaired differentiation of basal cells in COPD patients 14,53,[57][58][59] .
Coordinated and efficient ciliary beating is essential to clear the airways from inhaled particles and pathogens and is therefore an important defense mechanism 60 . COPD patients often suffer from severe exacerbations caused by bacterial or viral infections, due to insufficient clearance 26,61 . We show that CS exposure during SAEC differentiation impairs the development and function of ciliated cells. These findings are consistent with studies, investigating the effect of cigarette smoke extract on human bronchial epithelial cell differentiation 27 . To evaluate to which extent the CS ALI in vitro model reflects smoking induced effects on the lung epithelium of small airways in living humans, a transcriptomic comparison analysis of SAEC ALI cultures and human SAE brushings was conducted. Of note, SAEC ALI cultures stimulated with CS exhibit a gene expression pattern overlapping to epithelial samples from healthy smokers. When combining both datasets and performing a hierarchical clustering on the overlapping de-regulated transcripts, it would be expected to see similar patterns but not to find the in vivo data and the in vitro data forming separate clusters. Remarkably, CS stimulated SAEC ALI cultures clustered with samples from smokers, whereas air treated SAEC ALI cultures clustered with healthy control samples in an unexpected fashion. By this means, we did not include mechanisms that were only affected in vitro and not in vivo or vice versa, but we focused on processes that were modulated in both situations. Our results highlight the presence of shared processes that show a corresponding de-regulation, i.e. a similar up or downregulation in vitro and in vivo. Thus, we think that the CS treated ALI culture from SAEC is a valuable in vitro method to investigate smoke induced effects on epithelial differentiation. However, our data does not allow distinguishing CS-induced effects in ALI cultures from healthy donors compared to COPD patients, since the CS is a very dominant driver of differential changes. The dissection of potential differences in smoked healthy versus smoked COPD ALI cultures would require a much higher donor number. Here we see a clear limitation of our complex experimental in vitro setup; nevertheless, ALI cultures from COPD patients may be extremely helpful to test whether potential therapeutic concepts work also in a diseased setting.
Among the top deregulated canonical pathways were Nuclear factor erythroid 2-related factor 2 (NRF2)-mediated oxidative stress, xenobiotic metabolism and aryl hydrocarbon receptor (AHR) signaling [62][63][64][65] . It is known that CS induces reactive oxygen species, which activates the NRF2 pathway 63,66 . Moreover, CS comprises a number of toxic chemicals including polycyclic aromatic hydrocarbons (PAH) 67 . A prominent member of PAH is benzo[a]pyrene, which activates the AHR pathway, inducing the expression of several metabolic enzymes, including CYP1A1, CYP1A2 and CYP1B1 62,68 . The model reflects these CS induced pathways in close resemblance to the in vivo situation and offers novel opportunities to study new therapeutic targets addressing oxidative stress responses and xenobiotic metabolism.
Inconsistencies with the in vivo data were strongly associated with lipid metabolism (e.g. synthesis of cholesterol).
Our data demonstrate that CS specifically alters the small airway epithelium by affecting basal cell differentiation. Thereby, important pathways are consistently modulated compared to the small airways of human smokers. Taken together, critical pathophysiologic and molecular effects could be mimicked in vitro that were described in small airways of smokers, which might lead to novel essential therapeutic approaches to restore the impaired epithelial repair mechanisms in a smoke-induced chronic disease like COPD. cigarette smoke exposure. SAECs were exposed to whole cigarette smoke of K3R4F reference cigarettes (University of Kentucky, Lexington, KY, USA) using an automated cigarette smoking machine (In-Expose smoking robot, Scireq Montreal, QC, Canada) and the P.R.I.T. (Professional In-Vitro Technologies) ExpoCube in the experimental setup developed by Fraunhofer Institute for Toxicology and Experimental Medicine. Four cigarettes were smoked in parallel in compliance to ISO 3308, drawing every 15 seconds a puff from the sequentially www.nature.com/scientificreports www.nature.com/scientificreports/ smoked cigarettes (9 puffs per cigarette). The smoke was diluted with ambient air (dilution rate: 0.5 l/min). The cells were exposed to CS three times a week during the differentiation phase (28 days) beginning on day 0 (= day of air-lift). Cells were cultured on 12-well plates with the upper four wells exposed to smoke and the lower four wells exposed to air 69 . All wells were uniformly exposed to CS due to computational fluid dynamics-optimized geometry of the flow paths. This was previously tested and described in detail by Ritter et al., Fraunhofer Institute for Toxicology and Experimental Medicine 70 . Readouts were performed 24 hours post last exposure.

SAec
Cytotoxicity of smoke-exposed cells was assessed on day 28, based on the measurement of lactate dehydrogenase (LDH) activity released from the cytosol of damaged cells into the supernatant (compare Supplementary Methods). There was no significant difference in LDH release between smoke-exposed and air-exposed cells (Supplementary Fig. S4).
transepithelial electrical Resistance (teeR). Epithelial barrier integrity was monitored by measuring Transepithelial Electrical Resistance (TEER) using an EVOM2 epithelial volt-ohmmeter (World precision Instruments, Sarasota, FL, USA) according to the manufacturer's instructions. To perform the TEER measurement, phosphate buffered saline (PBS, 300 µl) was added to the apical compartment of the insert. The long stem of the electrode was inserted through the gap of the transwell to be in contact with the basolateral medium and the short stem of the electrode was placed above the apical surface to be in contact with the apical PBS. The measured resistance was multiplied by the surface area of the epithelium (1.12 cm 2 ) to obtain TEER (Ω*cm 2 ).
Histology and Immunohistochemistry. Cells, growing on transwell membranes (pore size of 0.4 µm), were fixed in 4% paraformaldehyde over night and embedded in paraffin. After sectioning (section slice thickness approximately 3-4 µm), slides were deparaffinized and hematoxylin and eosin (H&E) stainings were performed according to standard procedures. For immunohistochemical (IHC) investigation, paraffin-embedded tissue slides were subjected to heat induced antigen retrieval in citrate buffer (pH 6.0, H.I.E.R, BioLegend San Diego, CA, USA) for 30 minutes. Antibodies used for immunofluorescence or chromogenic antigen detection are listed in Supplementary Information Germany) in case of chromogenic IHC. Stainings for KRT5, MUC5AC, SCGB1A1 and acetylated Tubulin were used to assess protein expression qualitatively and representatively. Claudin-10 protein expression was semi-quantitatively analyzed using digital image analysis (compare section "Image Analysis"). Image Analysis. ALI cultures with Claudin-10 IHC staining (ab52234, Rabbit polyclonal to Claudin-10, Abcam, Cambridge, MA, USA) were scanned with an Axio Scan.Z1 scanner (Carl Zeiss Microscopy GmbH, Jena, Germany) using a 20x objective (0.22 µm/px) in bright field illumination. Semi-quantitative image analysis was performed with the Halo 3.0 software using the CytoNuclear 1.6 module (Indicalabs, Corales, NM, USA). In preparation for the automated analysis, regions with damaged ALIs or insufficiently stained cells were manually excluded. For automated analysis, a random forest tissue classifier was trained to recognize the cellular layer of the ALIs to limit the following analysis to this region and exclude the supporting material and background. Subsequently, the CytoNuclear module was applied within this region. As first step, the channels of nuclear (hematoxylin) and IHC stain (3'Diaminobenzidine (DAB) signal of Claudin-10) were unmixed using color deconvolution. Cell segmentation used the nuclear stain and empirical parameters on size and shape of the cells. Positive cells were identified using a manually optimized threshold value on the DAB channel, within the segmented region of a cell. The generated readout per ALI culture was the number of positive cells n per analyzed area of the cell layer A, i.e. n/A. Cilia beat measurement of SAEC ALI cultures. Ciliary function was analyzed using video-microscopy as described in Supplementary Information. Briefly, cilia beat frequency was calculated from image stacks of 2D + time. For each pixel in the field of view, the signal in time domain is searched for a meaningful frequency. The results were mapped back to image space forming an image, where each pixel represents the beat frequency of the cilia covering that area.
Next generation sequencing and detection of differentially expressed genes. RNA extraction, Illumina library preparation and sequencing was performed as previously described 71 . Comparative analysis was done using the limma R-package 72 by applying a the following linear model: anno.sub < -data.frame(group = pre.design$group,donorID = pre. design$donorID) dmat < -model.matrix(~0+group+donorID,data = anno.sub) Benjamini-Hochberg correction was used to adjust for multiple testing. Expression data have been deposited to Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/; accession no. GSE135188). The previously published data set GSE11784 38 was used for comparative analysis.
Flow cytometry. Single cell suspensions of SAEC ALI cultures (see Supplementary Information) were fixed in 4% paraformaldehyde for 20 minutes at room temperature. Cells were washed twice using Perm/Wash Buffer (BD Biosciences, San Jose, CA, USA) and blocked to prevent non-specific binding using Human BD Fc Block (BD Biosciences, San Jose, CA, USA). Cells were stained in Perm/Wash Buffer for 1 hour at room temperature. Anti-human antibodies used for staining are listed in the supplement. Flow cytometry was performed using a BD LSRFortessa X-20 cytometer equipped with DIVA-software (BD Biosciences, San Jose, CA, USA).
Statistics. Statistical testing used for RNA sequencing analysis is described in section "Next generation sequencing and detection of differentially expressed genes". All the other analyses were performed as followed. Instead of modelling the longitudinal data, the analyses were performed with adjusted areas under the curve (AUC) or the logarithmized adjusted AUC. The adjusted AUC was obtained by dividing the AUC by the length of the considered time interval. Each of the endpoints was analyzed separately. The data was analyzed with a repeated measurement model, as both treatments (CS/Air) were applied to cells of each donor. Compound symmetry was assumed as underlying covariance structure in order to account for the repeated measurement within a donor. The fixed factors status of the cells (HC/COPD), treatment (CS/Air) and the interaction of status of the cells and treatment were included in the model. The degrees of freedom were calculated by dividing the residual degrees of freedom into between-subject and within-subject portions. If a fixed effect changed within any subject, the within-subject degrees of freedom were assigned to the considered fixed effect (cell status, treatment, interaction effect). Otherwise, the between-subject degrees of freedom were used. The applied repeated measurement model assumes that the errors follow a normal distribution.
Using linear combinations of the parameter estimates obtained from the repeated measurement model, it was tested whether treatment with CS compared to Air had an impact on the considered endpoint. This analysis was performed separately for each status of cells (HC/COPD). Depending on the expected change of the endpoint under the treatment with smoke, the following hypotheses were tested: For the following endpoints the above stated hypotheses were tested as a decrease was assumed when cells were treated with smoke: Logarithmized AUC of Area covered by active cilia, AUC of Cilia beat frequency, logarithmized AUC of TEER, AUC of KRT5 mRNA (Ct), AUC of KRT14 mRNA (Ct), AUC of MUC5AC mRNA (Ct). For the following endpoints the above stated hypotheses were tested as an increase was assumed when cells were treated with smoke: Claudin-10 protein expression, AUC of SCGB1A1 mRNA (Ct), AUC of FOXJ1 mRNA (Ct).
All tests are t-test like tests based on the linear combinations of the parameter estimates from the repeated measurement model and were performed post-hoc. The p-values were not adjusted for multiplicity induced by the number of tested endpoints and hypotheses and the selection of the finally applied statistical model.
The statistical evaluation was prepared using the software package SAS Version 9.4 (SAS Institute Inc., Cary, North Carolina, USA).