Mycobacteria were grown in Difco Middlebrook 7H9 medium (BD Biosciences, Franklin Lakes, NJ, USA) containing 0.05% Tween-80 (Sigma-Aldrich, St. Louis, MO, USA), 0.5% glycerol (Sigma-Aldrich), and 10% Middlebrook ADC growth supplement (750 mM BSA + 110 mM dextrose + 130 µM catalase in H₂O, produced in our lab). A list of bacterial strains used, and their origins can be found in Table S1 . The M. avium strain expressing Wasabi, the M. bovis BCG strain expressing GFP, and the M. smegmatis strain expressing GFP were cultured with 100 µg/ml hygromycin B (Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA). The Mtb H37Rv strain expressed Venus constitutively. All strains were diluted to OD_600 = 0.25 the day before experiments were performed, to ensure exponential growth of the bacteria.

Batches of heat-killed mycobacteria were prepared; OD₆₀₀ was measured, and bacteria were heated to 56°C for 30 minutes. Heat-killed bacteria were then diluted to a multiplicity of infection (MOI) of 100 in epithelial cell culture medium (see below) without antibiotics. Aliquots of heat-killed bacteria were stored at -20°C.

Primary bronchial epithelial cells (PBEC) were isolated from macroscopically normal lung tissue obtained from patients undergoing resection surgery for lung cancer at the Leiden University Medical Center, the Netherlands. Patients from which this lung tissue was derived were enrolled in the biobank via a no-objection system for coded anonymous further use of such tissue (www.coreon.org). However, since 01-09-2022, patients are enrolled in the biobank using active informed consent in accordance with local regulations from the LUMC biobank with approval by the institutional medical ethical committee (B20.042/Ab/ab and B20.042/Kb/kb). Cells were cultured as previously described (Schrumpf et al., 2020). In brief, PBEC (passage 1) of individual donors were expanded for 5 days in T75 culture flasks pre-coated with a mixture of 30 μg/ml Purecol (Advanced BioMatrix, Carlsbad, CA, USA), 5 μg/ml human fibronectin (PromoCell, Heidelberg, Germany) and 10 μg/ml BSA (Fraction V; Thermo Fisher Scientific, Carlsbad, CA, USA). Per experiment, mixes of 4 unique donors were prepared (Thaler et al., 2023): PBEC of 2 male and 2 female donors, all non-COPD patients, were mixed, and seeded at equal concentrations and at a high density of 160.000 cells on pre-coated 0.4 µm membrane pore size Transwell inserts (Corning Costar, Corning, NY, USA) in 12-well plates to rapidly obtain confluence and prevent excessive growth of a certain donor in the mix. Cells were cultured in medium consisting of one part Bronchial Epithelial Cell Medium-basal (BEpiCM-b (ScienCell, Carlsbad, CA, USA)) and one part Dulbecco’s modified Eagle’s medium (DMEM (Gibco, Thermo Fisher Scientific)) supplemented with Bronchial Epithelial Cell Growth Supplement (BEpiCGS; ScienCell), 12.5 mM HEPES (Gibco, Thermo Fisher Scientific), 100 U/ml penicillin, 100 ug/ml streptomycin (all from ScienCell), 2 mM glutaMAX (Thermo Fisher Scientific), further referred to as complete BD-medium (cBD). In the submerged stage, cBD was supplemented with 1 nM EC-23 (Tocris, Bristol, UK). After reaching 100% confluence, apical medium was removed, and cells were differentiated at the air-liquid interface (ALI) in cBD supplemented with 50 nM EC 23. Culture medium was changed 3 times per week during which the apical side of the culture was carefully washed with 200 µl warm PBS for 10 min at 37°C to wash away mucus and cell debris. After 14 days of mucociliary differentiation at ALI, all dominant cell-types were present and ciliary activity was observed. Cultures were subsequently used for experiments.

ALI-PBEC were cultured using cBD medium without antibiotics 2-4 days before infection experiments. Mtb Venus, BCG GFP, M. avium Wasabi and M. smegmatis GFP were grown as described above. Although the addition of 0.05% Tween-80 to bacterial cultures minimized bacterial aggregation, clumping cannot be completely ruled out. OD₆₀₀ of visually homogenous bacterial cultures was measured. Bacterial suspensions at predicted multiplicities of infection (MOI) of 10 and 100 were prepared accordingly in cBD medium supplemented with 50 nM EC-23. For these calculations, PBEC cell density was estimated at 1x10⁶ cells per insert. Bacterial suspensions were plated on 7H10 agar to retrospectively determine the actual infection load. Mucus was washed from the apical surface of ALI-PBEC cultures with warm PBS for 10 min at 37°C. Immediately following mucus removal, 50 µl bacterial suspension was carefully added to the apical side of the cells. Control cultures received medium without bacteria. For gene expression and Luminex assays, heat-killed bacteria were used in parallel with live bacteria. Next, cultures were centrifuged briefly at 300 x g for 2 min to spin the bacteria down onto the cell layer, and incubated for 24 h at 37°C. Then, 150 µl PBS was added to the apical side and cultures were incubated 15 min at 37°C. From each ALI-PBEC insert, 200 µl apical supernatant and 1 ml basal medium was collected and stored at -20°C for use in Luminex and neutrophil migration assays. The remaining cBD was changed with 1 ml new medium, and apical sides were treated with 30 µg/ml gentamicin (Sigma-Aldrich) for 15 min at 37°C to kill extracellular bacteria. Next, apical sides of cultures were washed with PBS, and cultures were incubated overnight at 37°C. At this point, cultures were used for the various assays described below.

Basal medium from ALI-PBEC cultures was replaced with calcium-free PBS and 200 µl was added to the apical side as well. Cultures were incubated 30 min at 37°C. PBS was removed, and 200 µl pre-warmed TrypLE Express (Gibco, Thermo Fisher Scientific) containing 10 µM Y-27632 (StemCell Technologies, Vancouver, BC, Canada) was added onto the cell layer. Cultures were incubated for 15 min at 37°C. After pipetting up and down to break up the cell layer, single cells were collected from the inserts and transferred to tubes containing 800 µl PBS. Of each sample, 100 µl of the total volume was transferred to a separate tube for the colony forming units assay (see below), the remaining cells were used for flow cytometry. Samples were stained with Violet Fixable Live/Dead dye (Invitrogen, Thermo Fisher Scientific), 1:800 in PBS for 30 min at RT, then washed with PBS. Samples infected with M. avium, BCG or M. smegmatis were fixed in 4% paraformaldehyde (PFA) (Alfa Aesar, Thermo Fisher Scientific) for 1 h, then resuspended in PBS with 0.1% w/v BSA. Flow cytometry was performed with a Cytek^®Aurora (Cytek Biosciences, Fremont, CA, USA) cytometer at the Flow cytometry Core Facility (FCF) of Leiden University Medical Center (LUMC) in Leiden, Netherlands (https://www.lumc.nl/research/facilities/fcf). Due to bio-safety level restrictions, samples infected with M. tuberculosis were fixed in 4% PFA for 10 minutes, then measured in fixative with a BD FACS Lyric 3L12C (BD Biosciences) cytometer in the BSL-3 lab.

Single cell samples collected in tubes during sample preparation for flow cytometry were treated with 30 µg/ml gentamicin (Sigma-Aldrich) for 15 min at 37°C to kill extracellular bacteria, then washed with PBS. Cells were lysed in H₂O + 0.05% SDS (Invitrogen, Thermo Fisher Scientific) for 15 min at RT. Then those lysates were washed with PBS, centrifuged at 800 x g and resuspended in 100 µl PBS. Serial dilutions of the lysates were plated on Difco 7H10 agar (BD Biosciences) plates and incubated at 37°C. The number of colonies formed was determined by manual count.

Following infection, ALI-PBEC cultures were washed with PBS on both sides to remove mucus and cellular debris, and remove any remnants of culture medium. Cultures infected with M. avium, BCG or M. smegmatis were fixed in 4% PFA for 1 h at 4°C, while cultures infected with M. tuberculosis were fixed in 4% PFA for 24 h at 4°C due to safety regulations. Cultures were washed with and stored in PBS at 4°C until staining. Cell culture inserts were blocked with PBS containing 1% BSA and 0.3% Triton X-100 (Fluka Chemie, Buchs, Switzerland) for 10 min at RT. The PET membranes were removed from the plastic holders with a scalpel, and cut in two. The membrane halves were submerged in primary antibody solution ( Table S2 ) for overnight incubation at 4°C. Membranes were washed 3 times with PBS before incubation with secondary antibodies ( Table S2 ) and DAPI for 2 h at 4°C. Membranes were washed 3 times in PBS and 3 times in demi water, then placed on a glass slide, and mounted in ProLong Diamond Antifade mountant (Invitrogen, Thermo Fisher Scientific). All samples were imaged with an SP8 confocal microscope (Leica, Wetzlar, Germany) at the Light and Electron Microscopy Facility of the Leiden University Medical Center in Leiden, Netherlands (https://www.lumc.nl/research/facilities/light-and-electron-microscopy-facility/services/). Image processing was performed in Leica Application Suite X (LAS X) software.

ALI-PBEC cultures were stimulated with live or heat-killed bacteria. At the relevant time point, cultures were washed with PBS to remove mucus and cellular debris. The cell layer was lysed with 350 µl TRIzol (Ambion, Foster City, CA, USA) for 5 min at RT. All liquid was collected into microtubes and stored at -80°C awaiting processing. To isolate RNA, samples were thawed on ice, incubated with 70 µl chloroform (EMSURE, Merck, Rahway, NJ, USA) for 3 min, and centrifuged at 12.000 x g for 15 min at 4°C. The aqueous phase of each sample was transferred to microtubes with 175 µl isopropanol (Supelco, Bellefonte, PA, USA), incubated 10 min at RT, and centrifuged. Supernatant was removed and the pellet was resuspended in 350 µl 75% ethanol (EMSURE, Merck). Samples were centrifuged and all ethanol was removed. Isolated RNA was then dissolved in nuclease-free H₂O (Ambion). RNA was stored at -80°C until use in the gene expression assay.

RNA was treated with DNA-free DNA Removal Kit (Invitrogen, Thermo Fisher Scientific) according to manufacturer’s instructions. RNA concentration was measured with a DS-11 spectrophotometer (DeNovix, Wilmington, DE, USA) and samples were diluted to 50 ng/µl. cDNA was synthesized using Reverse Transcription Master Mix (Fluidigm, South San Francisco, CA, USA), according to manufacturer’s instructions. The cDNA was stored at -20°C. A set of 94 genes was selected to broadly analyse the host response. These included genes involved in pathogen sensing, inflammation, epithelial-to-mesenchymal transition (EMT), and also genes encoding antimicrobials and cell type markers. A pool of these 94 primer pairs ( Table S3 ) was created in DNA suspension buffer (10 mM Tris, pH 8.0, 0.1 mM EDTA). For pre-amplification reactions, PreAmp Master Mix (Fluidigm) was combined with the primer pool and the cDNA previously obtained, and thermal cycled according to manufacturer’s instructions. Reactions were cleaned with Exonuclease I (Thermo Fisher Scientific) treatment. Cleaned pre-amplified cDNA was diluted 5 times in DNA suspension buffer. Target assay and sample pre-mixes were prepared to load into the Integrated Fluidic Circuit (IFC) chip (Fluidigm). Assay pre-mixes each contained 0.4 µl of respective primer pairs, 1.6 µl DNA suspension buffer and 2 µl Assay Loading Reagent (Fluidigm). Sample pre-mixes each contained 2 µl SsoFast Evagreen Supermix (BioRad, Hercules, CA, USA), 0.2 µl Sample Loading Reagent (Fluidigm) and 1.8 µl pre-amplified ExoI-cleaned cDNA. A 96x96 IFC chip (Fluidigm) was primed and assay and sample pre-mixes were loaded. The chip was pressurized using the HX IFC controller and run on the Biomark HD (qPCR) system according to manufacturer’s instructions. During analysis, data points with a cycle threshold (Ct) value over 35 were considered at the cutoff for reliable detection and were set to Ct 35. Relative target gene expression (ΔCt) was calculated by subtracting the geometric mean of three most stable reference genes (determined using geNorm from the ctrlGene r package): RPL13A, ATP5B and B2M.

After 24h infection, 200 µl apical supernatant and 1 ml basal medium were collected from each cell culture and stored at -20°C until processing. These were filtered in FiltrEX 96-wells filter plates (Corning Costar) with pore size 0.2 µm to remove bacteria. Supernatants and medium were diluted 4 times with Luminex Assay buffer (Bio-Rad, Hercules, CA, USA). Following manufacturer’s instructions, the Bio-Plex Pro Human Cytokine 27-plex Assay (Bio-Rad) was performed. Samples were measured on a Bio-Plex 200 System (Bio-Rad). Per analyte, a lower and upper limit of detection was determined with standard curves. Concentrations measured below the assays’ detection limit were set to 1 pg/ml, and those measured over the detection limit were set to the maximum quantifiable pg/ml per analyte. As apical wash was collected in a 5 times smaller volume than basal medium, concentrations retrieved in the groups were corrected for differences in sample volume before further analysis.

Apical and basal supernatants from infected ALI-PBEC (see Infection of airway epithelial cell cultures above) were pooled to obtain sufficient volume to assess functional neutrophil recruitment: supernatants from live- and heat-killed-stimulated ALI-PBEC were combined together, and then respective samples from all 5 donor pools were combined as well. A schematic of how samples were pooled can be found in Supplementary Figure 4A . Pooled apical supernatants were then diluted 1:1 in PBS to obtain a sufficient volume for this assay.

Whole blood from 3 healthy donors enrolled in the LUMC Voluntary Donor Service (LuVDS) was collected in two 10 ml heparin blood collection tubes. All donors gave broad informed consent. The biomaterial of these donors was released for research purposes only, and its use was approved by the Institutional Review Board (LuVDS23.007). Each tube of whole blood was transferred to a 50 ml tube and diluted with 15 ml PBS. Next, 10 ml Ficoll (LUMC Hospital Pharmacy, Leiden, Netherlands) was added at the bottom of each tube, which were then centrifuged for 20 min at 650 x g. Supernatant was discarded until only a pellet containing neutrophils and erythrocytes was left. To lyse erythrocytes, pellets were resuspended in 40 ml lysis buffer (0.1 mM EDTA, 180 mM NH₄Cl, 10 mM KHCO_{3 in} H₂O, pH 7.4). The moment the suspension turned dark red after approx. 2-5 minutes, tubes were immediately centrifuged for 10 min at 290 x g. Supernatant was discarded, pellets were combined into 1 tube and resuspended in 45 ml PBS, then centrifuged 10 min at 290 x g. This washing step was repeated once. Cell pellets were resuspended in 1-2 ml RPMI supplemented with 20 mM HEPES (Gibco), and cells were counted manually. Cell suspension was diluted to 7.5x10⁵ cells/ml in RPMI with HEPES. Then, 3.0 µm inserts (Greiner Bio-one, Alphen aan den Rijn, Netherlands) were placed in a 12-well culture plate and treated with 400 µl blocking solution (PBS containing 0.1% w/v BSA) for 15 min at RT. Blocking solution was removed from the inserts, and 1 ml pooled supernatant from infected ALI-PBEC was added to each basal compartment. As a positive control, 10 nM fMLP (Sigma-Aldrich) and 10% fetal bovine serum (Corning) in cBD medium (to compare to basal supernatants) or in PBS (to compare to apical supernatants) was used. Of the neutrophil cell suspension, 400 µl was added on top of each blocked insert. Neutrophils were allowed to migrate through inserts for 1.5 hours at 37°C. After incubation, 50 µl of 50 mM EDTA was added to basal compartments. After 2 min, inserts were discarded. From basal compartments 500 µl liquid was transferred to FACS tubes. From each sample, the number of neutrophils in 100 µl was measured on an Accuri C6 Plus (BD Biosciences) flow cytometer.

Data analysis was performed in Graphpad Prism version 9.3.1, and in R Studio version 4.1.2. R packages used were: caret, tidyverse, tidyr, ggplot2, reshape2, viridis, plotrix, ggrepel, mixOmics, scales, dplyr, and ctrlGene. All data were considered non-parametric. Differences in infection percentages were tested with the Friedman test + Dunn correction for multiple testing. Differentially expressed genes (DEGs) were determined by calculating log2 fold change by -ΔΔCt-method, subtracting the median relative expression (ΔCt) of the reference group from that of the target group and multiplying by -1. Threshold for statistical significance of DEGs was p < 0.05. Statistics were performed using the Mann-Whitney U test. Due to the small sample size and the resulting low statistical power, no corrections for multiple testing were performed on gene expression data. Differentially secreted cytokines were determined with the Mann-Whitney U test with FDR correction (Benjamini-Hochberg) for multiple testing. Clustering of gene expression and cytokine secretion profiles was done with the one minus Spearman rank correlation method.

To assess whether Mtb, M. avium, BCG, and M. smegmatis could infect human bronchial airway epithelial cell cultures and compare potential differences in infection percentages between these strains, well-differentiated PBEC were infected with the four different fluorescent mycobacterial strains for 24 h, followed by a short gentamicin treatment of 15 min to kill remaining extracellular bacteria. ALI-PBEC were then incubated for an additional 24 h before further analyses ( Figure 1A ). Following these 48 h, cultures were harvested, and the percentage of infected cells was measured by flow cytometry. The gating strategy is depicted in Figure S1A . The various mycobacterial strains infected epithelial cell cultures to a different extent ( Figure 1B ). The percentage infected cells depended on both the multiplicity of infection (MOI) and the bacterial strain. Due to BSL-3 restrictions, samples infected with Mtb were analysed on a different type of flow cytometer. To confirm that differences between pathogens were not the result of alternative flow cytometers, we also analysed samples from BCG, M. avium and M. smegmatis-infected ALI-PBEC models in the flow cytometer in the BSL-3 laboratory and obtained identical results ( Figures S1B–D ).

Combining data from all independent donor mixes, the highest infection percentage was observed for M. avium, followed by BCG and M. smegmatis, whereas the percentage of epithelial cells infected with Mtb was lowest ( Figure 1C ). Interestingly, the efficiency of M. smegmatis and M. avium infection in epithelial cells seemed to be at least partly dependent on the donor pool ( Figures S1E, F ). In the case of M. smegmatis infections, this donor mix-dependent variation in infection percentage was only observed at a high MOI of 100, whereas with M. avium this was seen at high and low MOIs.

The total intracellular mycobacterial burden per ALI-PBEC model, independent of bacterial fluorescence, was determined by colony forming unit (CFU) assays ( Figure 1D ). Similar to the flow cytometry read out, intracellular M. avium content was highest when comparing all tested mycobacterial strains. Intracellular M. smegmatis CFUs were highly dependent on MOI, while intracellular Mtb and BCG content was relatively low overall.

To confirm that mycobacteria were present intracellularly, infected ALI-PBEC cultures were stained with an antibody against the epithelial cell membrane marker EpCAM. Fluorescent bacteria of all four species were observed intracellularly by confocal microscopy ( Figure 1E ). M. avium-infected cultures again displayed the highest bacterial burden, with most infected cells per area, and often multiple bacteria per individual cell. In contrast, Mtb and BCG-infected cultures contained far fewer infected cells per area. Cultures infected with M. smegmatis generally showed more infected cells than Mtb- and BCG-infected cultures, but fewer than M. avium-infected ones. Infected cells were detected mostly on the apical side of ALI-PBEC cultures, and bacteria had not yet penetrated deeper layers. Finally, infection of ALI-PBEC cultures was rather heterogeneous, with not all areas of the epithelial culture having the same number of infected cells. Images shown in Figure 1E are of areas with high numbers of infected cells, from the top plane of each model which was defined as the first apical 1-5 µm of the epithelial culture.

Taken together, we demonstrate that well-differentiated epithelial cell cultures can be infected intracellularly by four mycobacterial species, albeit with varying efficiency.

To explore the transcriptional response of airway epithelial cells induced by mycobacteria, ALI-PBEC were infected with Mtb, BCG, M. avium or M. smegmatis at MOI 100 (or the heat-killed equivalent thereof), after which changes in gene expression were assessed. Differential gene expression was assessed using microfluidic multiplex qPCR, including 94 genes involved in pathogen sensing, cytokine and antimicrobial mediator production, cell type-related genes, and genes involved in inflammation and EMT ( Table S3 ).

PBEC displayed transcriptomic changes at 48 h post infection by all mycobacterial strains. Four genes were consistently found to be significantly differentially expressed in PBEC infected with all four strains of live mycobacteria in comparison to uninfected cells ( Figure 2A ). DEFB4, S100A8, and BIRC3 were significantly upregulated, while BPIFB1 was significantly downregulated compared to uninfected control PBEC ( Figure 2B ). Expression of other genes that were strongly upregulated after infection with most mycobacteria included SERPINB1, CXCL2, CXCL8, CXCL10, and LCN2 ( Figure 2C ). Expression of WNT5B on the other hand, was reduced in response to all tested mycobacteria, though this did not reach statistical significance for all species of mycobacteria ( Figure 2C ). Gene expression patterns in PBEC exposed to heat-killed bacteria followed largely the same pattern, albeit with smaller gene expression changes compared to live bacilli. Upon comparing live Mtb to live BCG ( Figure 2D ), the expression of several genes was increased by Mtb. These included MMP9, S100A4, TLR4, EGF and CCL18. On the other hand, BCG increased expression of IL6, DEFB1, DEFB103, and RNASE7. A similar comparative analysis of M. avium and M. smegmatis revealed that M. smegmatis increased several pro-inflammatory genes such as CXCL10, IL6, CXCL8 and TNFA, while M. avium increased genes related to cell adhesion such as VIM and TGFB3.

Hierarchical clustering revealed that both live bacteria and heat-killed bacteria induced gene expression changes in PBEC that were most similar for BCG and M. smegmatis, followed by M. avium, while Mtb-induced gene expression patterns were most different from those induced by the other strains ( Figure 3A ). Direct comparison of transcriptome changes induced by live and heat-killed bacteria of each pathogen revealed that more genes are significantly upregulated in response to live bacteria ( Figure 3B ). For Mtb, this included mostly antimicrobials such as DEFB4 and S100A8. Interestingly, changes between live and heat-killed bacteria were different for each mycobacterium.

Of note, for many of the analysed genes, the magnitude of the epithelial response did not seem to be dependent on infection percentage. For example, induction of S100A8 expression was nearly equal for all four mycobacteria ( Figure 2B ), while infection percentages between the four strains were variable ( Figure 1C ).

Besides the gene expression signature described above, pathogen-specific induction of gene expression was also observed. For example, expression of MMP9 was increased by Mtb but strongly reduced by the other mycobacteria ( Figure S2 ). Differences between pathogens were also detectable in the EMT regulatory factor SNAI2. M. smegmatis significantly upregulated this transcription factor, while the other species did not. Furthermore, several cell lineage gene markers were influenced by infection. Club cell marker SCGB1A1 was significantly downregulated in epithelium by BCG and M. smegmatis, and Mtb and M. avium induced a trend similar to the other pathogens. Furthermore, expression of ciliated cell marker TUBA1A was significantly increased only by M. smegmatis. DCLK1, which is considered a marker for tuft cells, was slightly upregulated by all pathogens, though only for BCG this reached statistical significance.

To summarize, we have shown that the four species of mycobacteria induce a common gene expression pattern in PBEC, which consists mainly of antimicrobials and pro-inflammatory cytokines, though species-specific induction of differential gene expression was also observed.

Epithelial cells secrete a plethora of cytokines and chemokines during homeostasis and upon infection, to interact with surrounding cells and tissue, including the immune system. We determined cytokine profiles of well-differentiated PBEC cultures in response to heat-killed and live mycobacteria. As ALI-PBEC cultures are polarized, samples were collected from both the apical (apical wash) and basal (medium) side.

First, secreted cytokine profiles were analysed by principal component analysis (PCA) and hierarchical clustering. Both methods separated apical and basal samples, indicating a polarized secretion of molecules ( Figures 4A, B ). In agreement with the similarity in patterns found in the gene expression analysis, also the cytokines secreted in response to live, or heat-killed mycobacteria did not seem to differ notably ( Figure 4C and Figure S3 ). Clustering identified a group of inflammation-related cytokines secreted on the apical side in response to infection, including IL-5, IL-17, IL-10 and IL-1RA. In contrast, the basal production contained mainly tissue growth factors and chemokines, such as VEGF, FGF, GM-CSF, CXCL10 and CCL4 ( Figure 4A ).

All cytokines that were significantly differentially secreted in response to live infection were compiled and plotted as a Venn diagram ( Figure 4D ) to examine the common cytokine response to mycobacteria. The 8 cytokines induced by live bacteria of all four species in all different donor mixes are listed in Tables 1 , 2 . Overlapping cytokines produced both apically and basally included chemotactic factors G-CSF, CCL2, CCL4 and CCL11. Furthermore, pro-inflammatory cytokines IL-6 and IL-9 were highly upregulated in response to all mycobacteria. Finally, apical secretion of IL-1β and IL-5, and basal secretion of CXCL10 and TNF-α were strongly upregulated in response to all mycobacteria.

In conclusion, we illustrated that mycobacterial infection induced secretion of a polarized common cytokine profile by PBEC, which consisted of pro-inflammatory cytokines apically and growth factors and chemokines basally.

To investigate whether mycobacteria also induce cytokine profiles that are species-specific in addition to the common profile described above, PCA was performed ( Figure 5A ). Even though epithelial cytokine production induced by the pathogens showed considerable overlap with each other (see above), on the apical side Mtb and M. avium were displaying more similarities to each other than to the other species in their induced cytokine expression profile. BCG and M. smegmatis-induced cytokine profiles also grouped together on the basis of being more similar to each other than to other pathogens. On the basal side, however, the four pathogens induced profiles that were similar. Clustering analysis of apical secretions showed that Mtb- and BCG-induced cytokine profiles were more comparable ( Figure 5B ), contradicting results found with PCA analysis.

Detailed analysis of the subtle differences between mycobacterial strains that were observed by PCA showed that apical IL-1β release was significantly induced by all pathogens apically, whereas only Mtb significantly triggered its basal release ( Figure 5C ). In contrast, Mtb-infected PBEC secreted similar levels of IL-1RA as uninfected cells, whereas the other mycobacteria significantly upregulated apical IL-1RA secretion. Finally, M. smegmatis infection significantly downregulated VEGF secretion apically, while levels secreted upon infection with Mtb and M. avium remained similar to those of uninfected PBEC. Three out of five BCG-infected PBEC pools showed drastic downregulation of VEGF as well, but this did not reach statistical significance. Finally, GM-CSF was significantly upregulated in M. avium-infected PBEC and not by other species, albeit only apically.

In short, we have established that infection of PBEC by different mycobacterial species induced different up- and downregulation of several cytokines.

Cytokines secreted by epithelial cells may act as chemoattractants for various immune cell subsets. To assess whether secretions from infected ALI-PBEC could attract neutrophils, which are generally first responders during bacterial infections, a neutrophil migration assay was performed. Basal medium and apical supernatants from five infected ALI-PBEC donor mixes per mycobacterial species were pooled ( Supplementary Figure S4A ), and used to attract neutrophils isolated from three independent healthy donors. N-formyl-met-leu-phe (fMLP) with 10% fetal bovine serum (FBS) which induces chemotaxis and degranulation of neutrophils, was used as a positive control. The number of migrated neutrophils was determined by flow cytometry (gating strategy depicted in Supplementary Figure S4B ).

Firstly, a much higher number of neutrophils migrated in response to apical wash compared to basal medium. This difference is evident in the increased chemotactic capacity of the respective uninfected control. In addition, the number of neutrophils attracted by apical secretions of Mtb, BCG and M. smegmatis even surpassed those attracted by the fMLP + 10% FBS control. Although apical wash was 2.5 times more concentrated than basal medium, apical wash from infected PBEC attracted approximately 4 times as many neutrophils compared to basal medium ( Figure 6 ). This was true for both the uninfected as well as infected conditions, with exception of M. avium basal medium. Furthermore, we observed several differences between the four species. Apical secretions of Mtb and BCG-infected ALI-PBEC attracted neutrophils to a similar extent as fMLP with 10% FBS, while secretions from M. smegmatis-infected ALI-PBEC even surpassed those numbers. Surprisingly, apical secretions from cells infected with M. avium barely attracted any neutrophils. In fact, those numbers were close to the number of neutrophils that spontaneously migrated towards uninfected ALI-PBEC secretions. On the other hand, the numbers of neutrophils attracted by basal secretions from infected ALI-PBEC were very similar across all pathogens, though still higher than the uninfected condition.

Taken together, cytokines secreted by PBEC infected with mycobacteria, except M. avium, were able to attract neutrophils from healthy independent donors.