Bordetella pertussis-infected innate immune cells drive the anti-pertussis response of human airway epithelium

Pertussis is a severe respiratory tract infection caused by Bordetella pertussis. This bacterium infects the ciliated epithelium of the human airways. We investigated the epithelial cell response to B. pertussis infection in primary human airway epithelium (HAE) differentiated at air–liquid interface. Infection of the HAE cells mimicked several hallmarks of B. pertussis infection such as reduced epithelial barrier integrity and abrogation of mucociliary transport. Our data suggests mild immunological activation of HAE by B. pertussis indicated by secretion of IL-6 and CXCL8 and the enrichment of genes involved in bacterial recognition and innate immune processes. We identified IL-1β and IFNγ, present in conditioned media derived from B. pertussis-infected macrophage and NK cells, as essential immunological factors for inducing robust chemokine secretion by HAE in response to B. pertussis. In transwell migration assays, the chemokine-containing supernatants derived from this HAE induced monocyte migration. Our data suggests that the airway epithelium on its own has a limited immunological response to B. pertussis and that for a broad immune response communication with local innate immune cells is necessary. This highlights the importance of intercellular communication in the defense against B. pertussis infection and may assist in the rational design of improved pertussis vaccines.

The currently used aP vaccine protects well against pertussis disease, however, it does not prevent colonization by, or transmission of B. pertussis 14 . This allows circulation of the bacterium in the vaccinated population. For the rational design of next generation vaccines that prevent colonization by the bacterium, better knowledge on the local immune response against B. pertussis in the human respiratory tract is essential. Despite extensive research on immune responses to B. pertussis 10 , little is known about how the initial immune response at the human airway epithelium is orchestrated. In the present study, we investigate the immune response to B. pertussis at the human airway epithelium using primary human airway epithelial cells (HAE) differentiated at air-liquid interface. The HAE response to this pathogen was characterized by assessing mucociliary transport, barrier integrity, transcriptomics as well as cytokine and chemokine secretion. As during natural infection the airway epithelium closely interacts with immune cells, we also investigated the possible regulatory role of other innate immune cells on the HAE response to B. pertussis. Our results indicate that HAE requires communication with other B. pertussis-exposed innate immune cells. We identified IFNγ and IL-1β as critical immunological factors for the initiation of a robust immunological response to B. pertussis by HAE cells. Additionally, we show that the chemokines produced by these HAE effectively recruit monocytes.

Characteristics of the HAE infection model. Bordetella pertussis infects the human respiratory tract
and in this study we set out to investigate early immune responses of the human airway epithelium to this bacterium. To this end we set up an in vitro model that closely represents the natural infection niche of B. pertussis by using primary human bronchial/tracheal airway epithelium cells. Culture of these cells on transwells and exposing them to air on the apical side for 6-8 weeks yielded well differentiated human airway epithelium (HAE). HE staining of cross-sectional slices of the cells after 7 weeks of differentiation revealed several cell layers of epithelial cells with protruding cilia on the apical side (Fig. 1A). The HAE cultures also produced mucus, which was visible as a viscous layer on top of the cells. Fluorescent microscopy on the HAE cultures revealed the presence of several mucus producing (Muc5AC + , Pink) and a large amount of ciliated (β-tubulin IV + , Green) cells which were connected by tight junctions (ZO-1 + , Red) (Fig. 1B). To assess the barrier function of the HAE cultures we measured the transepithelial electronical resistance (TEER). Cultures from all donors used in this study showed significant increases in barrier function one week after air-liquid culture. Maximum TEER values were reached after three to four weeks and remained stable at ~ 200 Ω*m 2 for the rest of the culture period (Fig. 1C).

Disruption of barrier integrity and mucociliary clearance in B. pertussis-infected HAE cultures.
Two hallmarks of pertussis infection are cellular damage of the airway epithelial lining and disruption of the mucociliary clearance 1 . To assess whether these hallmarks can be mimicked in our HAE cultures, we inoculated the cells with three different live naturally circulating B. pertussis strains at various MOI. After 22 h of incubation we observed that all B. pertussis strains reduced the TEER in a dose dependent manner as compared to the Mock HAE cultures ( Fig. 2A), indicating disruption of the epithelial barrier function by the pathogen. A significant difference in TEER disruption is observed between Bp1 and the other two strains at MOI 200. However, the basis for this variation remains unknown. Furthermore, we analyzed the ability of the HAE to transport beads over the apical surface of the cells by using fluorescent beads in combination with live imaging. Due to the damaging effect of a B. pertussis MOI 200 inoculation on the HAE cultures, only the lower MOIs were included in this analysis. Mock HAE cultures transported the beads with an average speed of 145 µm/s in a single direction ( Fig. 2B,C, Movie S1, S2). In stark contrast, incubation with MOI 100 of any of the three B. pertussis strains completely disrupted the ability of the HAE cells to transport the beads over their apical surface (Fig. 2B, Movie S3, S4). Incubation of HAE with B. pertussis at this MOI resulted mostly in immobile beads with some limited randomly-oriented movements compared to the longer trajectory and one directional movement of the beads in the Mock HAE cultures (Fig. 2C). At a lower MOI of 10, strains Bp2 and Bp3 significantly reduced the speed of bead movement whereas strain Bp1 did not affect the HAE ability to transport beads (Fig. 2B). These data show that B. pertussis infection of HAE cells results in a reduction of the barrier function and disruption of mucociliary transport, closely representing hallmarks of natural pertussis disease 1 .  (Fig. 3A, Fig S2), establishing a core DEG profile of these cells induced by B. pertussis. The 48 upregulated genes included several genes involved in cilia motility (FOXJ1, CFAP43, DNAH12, DNAH6 and HSPB11) indicating that the cells respond to the B. pertussis-induced disruption of mucociliary clearance by increasing expression of these genes. Also, genes involved in innate immune activation such as CXCL8, the AP-1 transcription factor subunit JUNB, and TBK1 involved in NF-κB activation were significantly upregulated after B. pertussis infection. No mucins or host defense peptides were significantly up-or downregulated after a 6 h B. pertussis infection.

Transcriptome analysis of B. pertussis-infected HAE cells.
To identify enriched pathways we performed GO enrichment analysis of the 48 core upregulated DEGs. This showed for B. pertussis-infected HAE an enrichment of various pathways associated with (hematopoietic) cell differentiation and tissue development (Pink), indicating active differentiation and development of the HAE cells (Fig. 3B). The pathway "Motile cilium assembly" is enriched (Blue), which is in accordance with the previously observed gene expression associated with cilia motility. The bacterium appears to actively interact with www.nature.com/scientificreports/ the epithelial cells (Orange, "Entry of bacterium in host cell", "Biological process involved in interaction with host"). GO-enrichment analysis also revealed innate immune activation of the HAE, indicated by the enrichment of pathways associated with innate immunity (Green, "(Cytoplasmic) pattern recognition receptor (PRR) signaling", "(Regulation of) chemokine production").

HAE immune activation.
To substantiate the RNA-seq data that suggest the induction of an innate immune response against B. pertussis, we analyzed cytokine and chemokine secretion into the basal medium of B. pertussis-infected HAE cells. IL-6 and CXCL8 increased dose-dependently after a 22 h incubation with any of the B. pertussis strains (Fig. 4). Since the human airway epithelium is able to secrete many more factors than the IL-6 and CXCL8 in response to respiratory pathogens as detected in our cultures 15,16 , we hypothesized that additional signals from immune cells may be required to induce the secretion of cytokines and chemokine by HAE cells in response to B. pertussis. To test this hypothesis we first added recombinant IFNγ, IL-1β and TNF to the HAE cells stimulated with Bp1 MOI 100. These cytokines are known to be secreted by immune cells in response to B. pertussis 5,10,17 . Addition of IFNγ induced CXCL9, CXCL10 and CXCL11 secretion into the basal medium of HAE cultures (Fig. 5A). Additionally, IL-1β induced low levels of CXCL10 and CXCL11 secretion (Fig. 5A), and high levels of CCL5, CCL20, CXCL5 and CXCL8 secretion (Fig. 5B). All cytokines were able to induce the secretion of CCL2 (Fig. 5C). Combination of all cytokines (IFNγ, IL-1β and TNF) induced secretion of all measured chemokines by the HAE. The additional presence of B. pertussis in these cultures did not significantly alter the secretion of any of the chemokines induced by IFNγ, IL-1β and TNF. Nevertheless, an increased trend in the levels of CCL5, CCL20, CXCL5 and CXCL8 was observed when TNF and B. pertussis were combined (Fig. 5B).  Bpstim-MΦ/NK-sup contained a wide variety of cytokines and chemokines, which were absent or present in lower concentrations in the Unstim-MΦ/NK-sup. CCL2 and CXCL8 were present in similar quantities in the Unstim-and Bpstim-MΦ/NK-sup ( Fig S2). HAE cell exposure to Bpstim-MΦ/NK-sup resulted in increased secretion of CXCL9, CXCL10, CXCL11, CCL2, CCL5, CCL20 and CXCL5 compared to these cells exposed to Unstim-MΦ/NK-sup (Fig. 6). Simultaneous stimulation with B. pertussis did not enhance nor diminish the chemokine secretion. CCL2 was also detected in the basal medium of Unstim-MΦ/NK-sup stimulated HAE cultures, however, this is likely due to the high concentrations of CCL2 present in the Unstim-MΦ/NK-sup.
To learn more about which factors secreted by immune cells mainly contributed to the HAE cell response, we blocked IFNγ in Bpstim-MΦ/NK-sup by neutralizing antibodies. This almost completely abrogated CXCL9,  www.nature.com/scientificreports/ CXCL10 and CXCL11 secretion by the HAE cells (Fig. 6A). On the other hand, IL-1β-neutralizing antibodies significantly inhibited CCL20 and CXCL5 secretion (Fig. 6B) and only when blocking both IL-1β and IFNγ activity, lower levels of CCL2 secretion were observed (Fig. 6C). These data strongly suggest that IFNγ and IL-1β are required for HAE cells to initiate the robust secretion of chemokines. This highlights the importance of macrophage and NK cell activation by B. pertussis for the induction of the HAE immune response to this pathogen.

HAE chemokine secretion induces monocyte migration.
To determine whether the factors secreted by HAE cells are able and sufficient to induce immune cell migration, we performed a PBMC migration assay. For this, we made use of HAE basal medium after stimulation with either Unstim-MΦ/NK-sup or Bpstim-MΦ/ NK-sup as described in Fig. 6. This basal medium was transferred to the basal compartment of an empty transwell system and PBMCs were added to the apical chamber. After a one-hour incubation the cells present in the apical and basal chamber were analyzed by flow cytometry (Fig. 7A). This revealed significant monocyte migration when basal medium of Bpstim-MΦ/NK-sup stimulated HAE cultures was used as chemoattractant (Fig. 7B). The monocyte migration was further enhanced when basal medium from the HAE cultures that had simultaneously been stimulated with B. pertussis and Bpstim-MΦ/NK-sup was used. When basal medium was used where both IL-1β and IFNγ activity were blocked in the Bpstim-MΦ/NK-sup prior to adding it to the HAE cells, monocyte migration was significantly reduced compared to the isotype control. These results indicate that immune factors produced by HAE cells in response to both IL-1β and IFNγ are capable to induce monocyte migration.   7,21 . Our finding that all tested B. pertussis strains completely abrogated the ability of the HAE cells to transport beads across their apical surface is consistent with the loss in mucociliary clearance by ciliated epithelium which is a hallmark of natural B. pertussis infection. Reduced mucociliary clearance has previously been attributed to a reduction of the amount of ciliated cells rather than a reduction in ciliary beat frequency 7 . Although we have not counted the amount of ciliated cells, we did observe that after B. pertussis infection the HAE cultures still contained beating ciliated cells but were nonetheless unable to transport beads (Movie S4).
As a first step to elucidate the HAE response to B. pertussis infection, we performed RNA-seq and compared the transcriptomes of B. pertussis-infected and non-infected HAE cells. We observed a large variation in the amount of DEGs induced by the different B. pertussis strains (Fig S1). However, this was most likely due to the low amount of donors available for the RNA-seq and does not necessarily reflect actual differences between the strains. Whether these differences are due to the variation among the stains which may differently express www.nature.com/scientificreports/ was significantly upregulated after a 6 h incubation. Another study, which analyzed the transcriptional profile of B. pertussis-infected undifferentiated BEAS-2B cells, a transformed human bronchial epithelial cell line, after a 3 h incubation did find significant upregulation of both IL6 and IL8 30 . This could indicate that IL6 upregulation already occurred in HAE cells by the time we performed RNA-seq, which was at 6 h after incubation with B. pertussis. Additionally, while they found IL1B to be upregulated in BEAS-2B cells, we did not observe any increased expression or secretion of IL-1β (data not shown) by HAE cells after B. pertussis infection. As IL-1β secretion is regulated post-transcriptionally 31 , the lack of IL-1β secretion does not exclude the upregulation of IL1B before our transcriptome analysis was performed. We were surprised to find only IL-6 and CXCL8 secretion by B. pertussis-infected HAE because it is known that B. pertussis infection leads to massive immune cell infiltration into the lungs 1,10 , which requires the production of chemokines. Additionally, our findings indicate that B. pertussis did not have any enhancing or inhibitory effect on the cytokine-induced chemokine secretion by HAE cells. The lack of strong immunological activation of the HAE by B. pertussis may be due to strong inhibitory properties of B. pertussis virulence factors. Using murine models for pertussis, the virulence factor pertussis toxin has been shown to inhibit neutrophil recruitment to the lungs and to actively suppress chemokine expression 32,33 . However, in those studies the authors analyzed the effect of pertussis toxin on chemokine expression in whole lungs of B. pertussis-infected mice and on the MH-S murine alveolar macrophage cell line but have not looked at the effect on the airway epithelium itself. Future studies should elucidate whether the inhibitory effect of pertussis toxin on chemokine production and cell recruitment is due to inhibition of other immune cells present in the lungs or by also targeting the airway epithelial cells. Another study showed the ability of recombinant adenylate cyclase toxin (ACT), a virulence factor produced by B. pertussis, to inhibit IL-17A induced CXCL8 secretion by HAE cells 34 . We showed that the live naturally circulating B. pertussis strain used in this study, which has previously been shown to produce ACT 22 , was unable to inhibit IL-1β-induced CXCL8 secretion. Future studies should investigate whether live B. pertussis is able to inhibit IL-17A-induced chemokine secretion by the HAE and whether this is dependent on B. pertussis-derived ACT.
In the natural situation, HAE closely interacts with immune cells 8 , therefore, we hypothesized that cytokines secreted by immune cells present at the epithelial lining were essential to fully activate the airway epithelial cells. Indeed, our results indicate that intercellular communication between HAE and local innate immune cells namely, macrophages and NK cells, is essential for the secretion of a wide variety of chemokines by the HAE. Macrophages and dendritic cells are the first innate immune cells that sense and respond to B. pertussis 10 by secreting cytokines such as, IL-1β, TNF and IL-6 5,10 , and together with NK cells also produce IFNγ 17 . Exposure of HAE to recombinant IL-1β and IFNγ resulted in robust chemokine secretion. However, exposure to the more complex supernatant of B. pertussis-infected macrophages and NK cells also resulted in the secretion of a wide variety of chemokines by the HAE. Especially IL-1β and IFNγ derived, among others, from B. pertussisinfected macrophages and NK cells, induced the production of chemokines by HAE cells. As both early IL-1β and IFNγ secretion in response to B. pertussis were shown to depend on NLRP3 inflammasome activation in macrophages 17,35 this also highlights the important role of inflammasome activation in shaping an inflammatory response to B. pertussis in the HAE.
In the present study, we showed that immune factors produced by cytokine-stimulated HAE cells were able to induce monocyte migration and that this was dependent on both IL-1β and IFNγ stimulation of the HAE cells. Since CCL2 in known to attract monocytes 36 and both IL-1β and IFNγ induced the secretion of CCL2, this chemokine is likely responsible for the observed monocyte migration. However, since monocytes express a wide variety of chemokine receptors which are targeted by a broad range of chemokines 37 , future studies should investigate the contribution of different chemokines to monocyte migration. Monocyte migration induced by basal medium derived from Bpstim-MΦ/NK-sup stimulated HAE cultures was increased if the HAE cells were simultaneously stimulated with B. pertussis. Since we did not find any increased chemokine production by cytokine-stimulated HAE cells in the presence of B. pertussis it appears that unmeasured chemokines or other undetected factors, such as host defense peptides 38 , are responsible for the increased monocyte migration. Here, we focused on monocyte migration because these innate cells are among the first to be recruited to the lungs during B. pertussis infection 10 and after their differentiation to macrophages they secrete IL-1β and IFNγ both required to enhance the immune response to this pathogen.
Altogether, we show that primary HAE cultures are an excellent model to investigate the early events of B. pertussis infection. Using this in vitro model we could mimic disruption of epithelial barrier integrity, reduction of mucociliary clearance, immune activation and immune cell infiltration, which are all hallmarks of pertussis disease 1 . We identified IL-1β and IFNγ as essential immunological factors for a robust HAE response against B. pertussis. Our data suggests that the airway epithelium on its own has a limited immunological response to B. pertussis and that for a broad response, local innate immune cells are necessary (Fig. 8). In other words, the cytokines secreted by these local immune cells upon infection with this pathogen stimulate the airway epithelium to produce different chemokines which are required for the recruitment of additional immune cells enhancing local inflammation. This highlights the importance of local intercellular communication in the defense against B. pertussis infection and may add to the design of next generation pertussis vaccines that aim to prevent not only disease but also airway colonization by the bacterium.

Materials and methods
Bacterial strains and growth conditions. In this study we used three B. pertussis strains isolated from Dutch pertussis patients: two isolated in 2016 (Bp1 and Bp2) and another isolated in 1998 (Bp3), Bp2 was deficient for the virulence factor pertactin. All strains were selected based on the grouping described by Kroes et al. 22 . In short, all strains have comparable genotypes (pertussis toxin (ptx) promoter 3, pertactin type 2, ptxA www.nature.com/scientificreports/ type 1, fimbriae 2/3 type 1) and only differ in pertactin expression or year of isolation. To ensure the consistent use of the B. pertussis strains at a logarithmic growth phase, flash frozen cultures were prepared as previously described 17 and used for all experiments.
Human airway epithelium air-liquid culture. Primary human bronchial/tracheal epithelial cells from four healthy donors (Lonza, catalog#: CC-2540) were expanded in serum-free PneumaCult-Ex medium supplemented with hydrocortisone (480 ng/ml), penicillin (100 U/ml) and streptomycin (100 μg/ml) at 37 °C and 5% CO 2 . ReagentPack Subculture Reagents (Lonza) were used to passage the cells. Before reaching passage 4, a total of 50,000 cells were seeded into 6.5 mm transwell plates with 0.4 μm pore polyester membrane inserts (Corning Inc) that were pre-coated with Collagen I Rat Tail Protein (30 μg/ml) for 45 min at 37 °C. After 2-4 days of submerged culture in complete PneumaCult-Ex medium (500 μl in basal and 100 μl in apical compartment) the apical medium was removed and the basal medium was replaced with 500 μl serum-free PneumaCult-ALI medium supplemented with hydrocortisone (480 ng/ml), Heparin (4 μg/ml), penicillin (100 U/ml) and streptomycin (100 μg/ml), creating an air-liquid interface. These HAE air-liquid cultures were differentiated for 6-8 weeks towards a pseudostratified epithelial cell layer. PneumaCult-ALI medium in the basal compartment was refreshed every 2-3 days and after 2 weeks the apical surface was washed every week with 200 μl PBS for 5 min at 37 °C.

HAE infection assays. Two days prior to use in experiment, medium of the HAE cells was replaced by
PneumaCult-ALI medium without antibiotics. HAE cells were inoculated with live B. pertussis at the indicated multiplicity of infection (MOI) which was calculated relative to the amount of cells seeded onto the insert. www.nature.com/scientificreports/ Additionally, HAE cells were treated with either recombinant human IFNγ, IL-1β, TNF-α (10 ng/ml), and antihuman IFNγ, anti-human IL-1β or the isotype control mouse IgG1 (10 μg/ml). All suspensions were prepared in PneumaCult-ALI medium without antibiotics and HAE cells were exposed in a total of 50 μl in the apical compartment for 22 h at 37 °C and 5% CO 2 . Polymyxin B was added to a concentration of 100 μg/ml to all collected supernatants and incubated for 30 min at 37 °C to kill any live bacteria present before storage at − 80 °C. Conditioned macrophage/NK co-culture supernatants were prepared as previously described 17 . Briefly, monocytes were isolated from blood of healthy donors and differentiated to macrophages using GM-CSF. Autologous NK cells were added to the monocyte-derived macrophages in a 1:1 ratio and these co-cultures were stimulated with B. pertussis. After a 20 h stimulation supernatants were collected and filtered using a 0.22 μm filter to remove any bacteria.
Transepithelial electronical resistance (TEER) measurement. To (25 μl) were added to the apical compartment and beads were allowed to settle on the epithelial cells for 10 min at 37 °C. Excess PBS was removed from the apical compartment and bead movement was recorded at 37 °C in the Y5 channel of a Leica DMi8 microscope with a Leica DFC7000 GT camera by making a 30 s movie containing 149 frames. Two or three movies were recorded per well. Because bead speed in the center of the insert can differ from the speed near the edges of the insert all movies were recorded at several locations exactly one field of view from the edge. Leica LAS X 3.4.2 software with LAS X time-lapse extension was used for image analysis. Beads were automatically detected with a minimal size of 10 pixels. Particle tracks were generated by the software using a link range of 2 and a max displacement of 20-40 depending on the speed of the beads. Tracks of less than 8 concessive frames were removed from the analysis. The location of each bead at a certain time point was defined by a pixel centroid X axis (cX) and a pixel centroid Y axis (cY) position. Per group, all the locations were centered around 0 (to be able to overlap them) and were plotted with the aid of ggplot2 R package 39  www.nature.com/scientificreports/ RNA isolation, sequencing, analysis and data processing. Total RNA was extracted from HAE cells exposed to medium only, and after a 6 h incubation with B. pertussis. Library preparation and RNA sequencing were performed as previously described 22 . FASTQ files were aligned to the Homo sapiens reference genome (GRCh38) and most recent transcript annotations using kallisto (v0.46.1) 40 . Raw counts were filtered for low count genes, eliminating those targets with less than 1 count in at least 5 samples. Filtered count distribution was normalized with the EDAseq R package 41 by the use of betweenLaneNormalization() function and the full quantile method. The suitability of this global normalization method was assessed and confirmed by quantro R package 42 .
Remove Unwanted Variation using Residuals (RUVr) from RUVseq R package 43 was employed to obtain the covariates that correct for Donor and other technical bias in the downstream analysis. Two parameters were needed to be provided for that purpose: residuals and parameter "K". Deviance residuals were obtained from a fitted Genewise Negative Binomial Generalized Linear Model (glmFit() function from edgeR R package 44 ). The model design was ~ Cell type, corresponding to the different experimental groups: unstimulated HAE (Mock) and HAE stimulated with any of the three strains used in the study. Parameter "K" was set at 4, since it is the number of significant surrogate variables detected by svaseq() function of sva R package 45 , when using the same model design as for glmFit.
Differential gene expression analysis and pathway enrichment analysis. The DEG lists was obtained with DeSeq2 R package 46 under standard parameters, for the comparisons Mock vs Bp1, Mock vs Bp2 and Mock vs Bp3. The model design was composed by the 4 covariates obtained in the RUVr correction (function pData()) and the ~ Cell type parameter described above, using as input the filtered normalized counts. Genes were considered significantly expressed if they showed an adjusted p-value (Bonferroni-Hochberg multiple testing (padj)) lower than 0.05 and a Fold Change > 1.2.
Venn diagrams were produced with VennDiagram R package 47 , giving as input all the DEG lists from the different comparisons.
Heatmap to represent the selected DEGs were generated with the aid of ComplexHeatmap R package 48 . Dendrogram trees of the heatmap were obtained by hierarchical clustering (Ward's D2 method) of spearman correlation gene distances.
Gene Ontology (GO) enrichment of the selected DEGs was performed and plotted by using the functions enrichGO() and barplot() respectively of the clusterProfiler R package 49 . As input, only the common up-regulated and down-regulated genes in the comparison of the 3 strains vs Mock were used.
Monocyte migration assay. Monocyte migration was determined by seeding 500,000 PBMC, isolated from blood of healthy donors using lymphoprep gradient centrifugation, in the apical chamber of a Corning HTS Transwell 96 well permeable support with 3.0 μm pore polycarbonate membrane. One hundred microliters of undiluted basal medium derived from HAE cells was added to the basal compartment of the transwell system and the plates were incubated for 1 h at 37 °C and 5% CO 2 . After incubation, cells from the apical and basal compartment were collected separately and both compartments were washed extensively with PBS. Cells attached to the bottom of the membrane were detached by incubating for 10 min at 37 °C with Accutase (100 μl) in the basal compartment. Detached cells were added to the cells collected from the basal compartment. Marker expression was determined with a LSRFortessa X-20 (BD) and data were analyzed using FlowJo software version 10.6.2. The percentage of monocyte migration was calculated by determining the abundance of CD3 − CD14 + CD16 +/− cells in the basal compartment relative to the total amount of CD3 − CD14 + CD16 +/− cells in the apical and basal compartment.
Ethics statement. This study was conducted according to the principles described in the Declaration of Helsinki. Buffy coats were provided by the Sanquin blood supply. For the collection of blood samples and subsequent analyses, all blood donors provided written informed consent. Blood samples were processed anonymously and the research goal, primary cell isolation, required no review by an accredited Medical Research Ethics Committee, as determined by the Dutch Central Committee on Research involving human subjects. Statistical analysis. Data were subjected to a Shapiro-Wilk normality test and when passed subjected to a two-tailed paired t-test using the GraphPad Prism version 9.1.0 software and corrected for multiple testing using the Bonferroni-Holm method in RStudio.
For the ciliary movement, µm/s were used for representation and statistics. In the rest of the analysis of cytokine secretion, the raw cytokine concentrations were divided by the triplicate average in the medium condition of the corresponding donor, to normalize for the donor effect and to obtain a relative change to that condition (Fold Induction). Having these values as input, the permutation based non-parametric Exact Wilcoxon-Mann-Whitney test was applied for the different comparisons of interest. All the p-values were corrected for multiple testing by the Bonferroni-Hochberg method (padj). For representation purposes, the triplicate average per donor of the normalized data was used.
Effect sizes (ES) were calculated, providing Cliff 's delta 50 (function orddom() from orddom R package 51 ). The descriptors for Cliff 's delta are: small, < 0.28; medium, < 0.43; large, > = 0.43 52 . All comparisons with both a padj < 0.05 and ES medium or higher are considered as biologically relevant and highlighted in the present work.

Data availability
All RNA-seq data is available in the GEO database, accession number: GSE182807 (reviewer token: ehwbyyukjxgdnyx).