Bordetella pseudohinzii targets cilia and impairs tracheal cilia-driven transport in naturally acquired infection in mice

Several species of the Gram-negative genus Bordetella are the cause of respiratory infections in mammals and birds, including whooping cough (pertussis) in humans. Very recently, a novel atypical species, Bordetella pseudohinzii, was isolated from laboratory mice. These mice presented no obvious clinical symptoms but elevated numbers of neutrophils in bronchoalveolar lavage fluid and inflammatory signs in histopathology. We noted that this species can occur at high prevalence in a mouse facility despite regular pathogen testing according to the FELASA-recommendations. Affected C57BL/6 J mice had, in addition to the reported pulmonary alterations, tracheal inflammation with reduced numbers of ciliated cells, slower ciliary beat frequency, and largely (>50%) compromised cilia-driven particle transport speed on the mucosal surface, a primary innate defence mechanism. In an in vitro-model, Bordetella pseudohinzii attached to respiratory kinocilia, impaired ciliary function within 4 h and caused epithelial damage within 24 h. Regular testing for this ciliotropic Bordetella species and excluding it from colonies that provide mice for lung research shall be recommended. On the other hand, controlled colonization and infection with Bordetella pseudohinzii may serve as an experimental model to investigate mechanisms of mucociliary clearance and microbial strategies to escape from this primary innate defence response.

although the mice had no clinical signs of respiratory infection. Still, it was suggested to consider this species as a confounding organism in murine models of pulmonary disease with the consequence of colony screening 9 . The current FELASA (Federation for Laboratory Animal Science Associations) recommendations for the health monitoring of mouse colonies in experimental units do not list B. pseudohinzii in their health report form 13 , but the AALAS (The American Association for Laboratory Animal Science)/FELASA working group refers to B. hinzii as an "exotic agent" that "should be mentioned when found" 14 .
In the course of experiments designed to study the mechanisms that regulate mucociliary clearance, a mechanism that eliminates inhaled particles from the airways, we encountered single murine tracheas with unusually low cilia-driven particle transport speed (PTS). This was observed both at baseline and after stimulation with acetylcholine and ATP, known activators of ciliary beat frequency (CBF) 15 . Although these mice were derived from a specified-pathogen free (SPF) facility and had no overt clinical signs of respiratory illness, we subjected samples including oropharyngeal swabs, tissue samples from trachea and lung from all mice used for PTS experiments to routine microbiological diagnostics. As a common denominator, non-classical Bordetella, not listed in the current FELASA recommendations for the health monitoring of mice in breeding and experimental units 13 , were found in mice with reduced PTS. We then set out to identify this Bordetella species by next generation sequencing. The structural and functional impact on the lung and trachea was analysed by light and electron microscopy, by BAL analysis and measurements of cilia-driven transport and beat frequency in Bordetella carriers versus non-carriers. Once isolated and identified as B. pseudohinzii, we used explanted tracheas to identify the cellular target of this species. Our data demonstrate that B. pseudohinzii can occur at high incidence in mouse colonies kept according to current FELASA recommendations, despite displaying little to none overt clinical symptoms. Affected mice present extensive airway inflammation with epithelial remodelling and functional impairment of clearance mechanisms. Therefore, we feel it justified to recommend regular testing for this Bordetella species and aiming to exclude it from facilities that provide mice for experimental lung research. On the other hand, controlled colonization and infection with B. pseudohinzii may serve as a model to investigate mechanisms of mucociliary clearance.

Microbiology.
A high prevalence of disturbed tracheal PTS emerged in September 2016, and we began to subject samples of all animals used for experiments addressing pulmonary structure and function to microbiological diagnosis. Utilizing matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS), a high incidence of B. hinzii was detected. In the period between October 2016 and March 2017, 84 animals were analysed in total, 78 (93%) of which were Bordetella-positive, with high incidence in the pharynx (oropharyngeal swab; 80%), trachea (84%), and lungs (30%; Fig. 1a).
In view of the fact that this MALDI-TOF MS (database by BRUKER) approach does not allow for a discrimination between B. hinzii and the very recently discovered and closely related species B. pseudohinzii 7,8 , the genomes of four isolates (strains 3136, 3227, 3561 and 3562) were sequenced using Illumina's NextSeq. 500. In brief, the mean Phred quality score per sequence read was Q > 32 after adapter trimming and quality clipping, the number of contigs per strain ranged between 60-77 and the N50 coverage varied between 29-78 among the sequenced genomes (Supplementary Tables 2 and 3). Reference based SNP analysis indicated that all four isolates were highly related to B. pseudohinzii 8-296-03 (average 106.5 SNPs) compared to B. hinzii LMG 13501 (average 72546.75 SNPs) ( Supplementary Fig. 1a). Phylogenetic analysis by EDGAR 16 based on the amino acid sequences of 1649 core genes shared among 15 Bordetella genomes clustered the isolated strains to the B. pseudohinzii branch (Fig. 1b), which was also confirmed by ANI (Fig. 1c). Further characteristic features allowing discrimination from B. hinzii were the presence of the type II-C CRISPR Cas system reported in B. pseudohinzii 7 and the absence of the typical D-galactonate utilization gene region present in B. hinzii but not in B. pseudohinzii 8 (illustrated by GECO 17 , Supplementary Fig. 1b,c). Also, phylogenetic analysis of 16S rRNA gene sequences confirmed identity to B. pseudohinzii ( Supplementary Fig. 2).
Accordingly, genomic B. pseudohinzii DNA was identified by PCR in faecal and laryngeal samples from animals in which B. hinzii was diagnosed in the trachea by MALDI-TOF MS (Fig. 2a). Samples tested negative by MALDI-TOF MS were also Bordetella-negative in PCR (Fig. 2a). The PCR assay targets a 318 bp-region of the outer membrane protein A (ompA) gene of B. pseudohinzii. Nucleic acid sequencing of the PCR-product showed 100% identity to the ompA sequence of B. pseudohinzii (GenBank accession number CP016440) and 98%-99% homogeneity to available B. hinzii sequences (GenBank accession numbers CP012076, CP012077, AM748263, AM478264, AM478265, LT906461). From 6 animals with B. pseudohinzii-positive faecal PCR, CFU values were determined in homogenized trachea and lung. None of the animals used for this assay was positive for Pasteurella pneumotropica. Colonies from the agar plates were subjected to MALDI-TOF MS and diagnosed as B. hinzii, confirming that CFU counts indeed reflected Bordetella and that MALDI-TOF MS does not discriminate between B. pseudohinzii and B. hinzii. Colonies were obtained from all tracheas, and CFU ranged from 2.2 × 10 2 to 2 × 10 7 /100 mg tissue, and from 5.6 × 10 1 to 4.9 × 10 6 per entire organ (Fig. 2b, Supplementary Table 1). From two of the lungs, no bacteria were retrieved, and the values in the other 4 ranged from 0.7 × 10 1 to 1.5 × 10 5 CFU/100 mg tissue and from 1.3 × 10 1 to 1.9 × 10 5 CFU/lung (Fig. 2b, Supplementary Table 1).
Rod-shaped bacteria were seen at the surface of ciliated epithelial cells in Giemsa-stained sections (Fig. 3c). Scanning electron microscopy revealed single rod-shaped bacteria attached to kinocilia of ciliated cells (Fig. 3d). The relative frequency of tracheal ciliated epithelial cells was lower in Bordetella-positive animals compared to Bordetella-negative animals (30.9 ± 1.5%, n = 5, and 43.5 ± 5.3%, n = 7, respectively; p = 0.0009; Fig. 3e). None of the animals used for these cell counts was positive for Pasteurella pneumotropica.
The lungs presented with multifocal bronchointerstitial pneumonia with dominating neutrophils and macrophages in major airways and alveolar spaces (Fig. 4e). Accordingly, neutrophils made up between 9% and 21% (mean: 14%; n = 6; Fig. 4d) of cells in the BAL fluid from Bordetella-positive animals, whereas they accounted for less than 0.6% throughout in Bordetella-negative animals (mean: 0.2%; n = 6). None of the animals used for cell counts in BAL was positive for Pasteurella pneumotropica, one was positive for Klebsiella oxytoca (this was the only Klebsiella-positive animal identified in the entire course of the study). Lymphocytes, plasma cells and macrophages formed peribronchial cuffs. Prominent aggregates of bronchus-associated lymphoid tissue (BALT), not seen in Bordetella-negative animals, were present close to the main bronchi in the hilar regions (Fig. 4a,b and c). Furthermore, a lymphoid hyperplasia of lymph nodes could be observed (Fig. 3a). Three of the 6 Bordetella-positive animals used for histopathological assessment were also positive for Pasteurella pneumotropica. There were no criteria identified that distinguished these animals from those carrying only B. pseudohinzii. Specifically, BALT formation was seen in both Pasteurella-positive ( Fig. 4a-c) and Pasteurella-negative (Fig. 3a) samples.
In vitro-infection of tracheal explants. Ciliated cells maintained intact cilia and normal ciliary beat frequency (CBF) in non-infected tracheal explants for up to 52 h, the longest time point examined ( Supplementary  Fig. 4). In these controls, bacteria were detected neither in semithin sections nor by scanning or transmission electron microscopy (Figs 5a,d,g and j and 6a,d,g,j,m and p). B. pseudohinzii was added to explanted tracheas at 5 different doses for 4 h and 24 h, respectively. With respect to CFU/entire trachea, the 3 lower doses were in the range of what we had recovered from infected animals. After 4 h, no bacteria were detected at the lowest dose, and only very few, single rod-shaped bacteria adhered to cilia at the two next higher doses (Figs 5b,e and 6b,e and h; Supplementary Fig. 5a,c; Fig. 6a,b,d,e,g and h). At a dose of 2.3 × 10 8 CFU/ml, bacteria regularly adhered to and intermingled with cilia (Fig. 6c,f and i), and bacterial chains, indicative of replication, were attached to cilia (Fig. 5c,f). At the highest dose (4.6 × 10 9 CFU/ml), microcolonies were observed at ciliated cells, cilia were shortened, distorted, and irregularly arranged (Supplementary Fig. 5d; Fig. 6c Cilia-driven transport and beat frequency. Baseline PTS was significantly lower in explanted tracheas taken from Bordetella-positive compared to Bordetella-negative animals (21.2 ± 8.3 µm/s, n = 11, and 52.7 ± 4.3 µm/s, n = 8, respectively; p < 0.0001; Fig. 7a,b). ATP-induced increase in PTS, recorded as absolute values of micrometer per second, was smaller in Bordetella-positive than in Bordetella-negative animals (24.4 ± 5.4 and 53.9 ± 4.8 µm/s; p = 0.0011; Fig. 7c). Expressed in percent of baseline value before ATP-stimulation, however, no significant difference was noted (104.7 ± 9.8% versus 130.8 ± 31.5%; p = 0.5022; Fig. 7d). Among the 11 Bordetella-positive animals analysed in this assay, 4 were also positive for Pasteurella pneumotropica. This had no  Supplementary Table 1. additional negative impact on PTS, as it was even slightly higher (33.4 ± 3.3 µm/s) in tracheas taken from these animals than in those from Bordetella-positive/Pasteurella-negative animals (19.5 ± 3.3 µm/s).
Next, the acute effects of Bordetella pseudohinzii on cilia-driven transport and beat frequency were investigated in explanted tracheas from Bordetella-negative animals exposed for 4 h to 3 different concentrations of bacteria. With respect to CFU/entire trachea, the 2 lower doses were in the range of what we had recovered from infected animals, and the highest dose was one order of magnitude higher. In control explants kept for 4 h without bacteria, baseline PTS (55.8 ± 3.7 µm/s) and CBF (20.6 ± 2.5 Hz) did not differ significantly from values obtained from freshly used tracheas. In both readouts, a CFU-dependent decline of function was observed with about 50% reduction at 1.6 × 10 5 CFU and practically complete loss of ciliary function and particle transport at 2.66 × 10 7 CFU (Fig. 8). In the videos taken for analysis of these parameters, attachment of bacteria to cilia was seen in a CFU-dependent manner, beginning at 0.4 × 10 4 CFU.

Discussion
Until now, isolates of the recently identified non-classical Bordetella species B. pseudohinzii that were unequivocally identified by whole genome sequencing had been obtained from mice in various facilities in the United States 7,8,18 and Malaysia 10 . In addition, there have been reports on the widespread occurrence of B. hinzii in mice in experimental facilities in Japan 11,12 , which, based upon the published partial sequences, most likely have monitored B. pseudohinzii instead 8,9 . In Europe, so far only one case has been reported. One mouse imported from Australia to the University of Düsseldorf in Germany had first been diagnosed as being infected with B. hinzii, and subsequent resequencing of this strain identified it as B. pseudohinzii 8,19 . In our study we obtained 4 isolates of B. pseudohinzii from laboratory mice bred and kept at another German SPF facility, showing that the occurrence of this species is not restricted to East Asia, Australia and North America.
At least in the colonies used for experimental work of our group, the incidence of B. pseudohinzii was high over a period of 6 months (93% of all animals tested; n = 84), which is far above the incidence reported for B. hinzii, probably representing B. pseudohinzii as discussed above 8,9 , in experimental facilities in Japan 12 . There, 1.6% (195 out of 12,192) of all tested mice were positive, and 2.8% of facilities (44 out of 1572) in universities and research institutes were affected. Notably, no B. hinzii-positive mice were found in 127 pharmaceutical companies surveyed. Taking only those facilities with positive findings, 32% of mice were positive (195 out of 613) 12 . While this is still less than half of the frequency we have observed, it has to be considered that a similarly high prevalence might have occurred in single facilities which cannot be extracted from the pooled data. Most importantly, the numbers on prevalence in Japan are derived from a systematic study testing regardless of clinical or experimental evidence for pulmonary problems, whereas our analysis was biased, because it was triggered by findings of impaired mucociliary clearance. Similar to our counts, 12 out of 15 animals (representing 7 mouse strains) taken from a facility where analysis was initiated due to high neutrophil counts in BAL fluid, were positive for B. pseudohinzii in faecal PCR 9 . Altogether, it appears that spread of B. pseudohinzii in mouse facilities may be focal but not regionally restricted, and can reach peaks of prevalence of > 80% in mice with C57BL/6 J genetic background, as reported here.
Colonization with B. pseudohinzii may predispose for infection with other lung pathogens, which may render interpretation of data difficult. Thus, samples from all animals used for histopathology or functional readouts were analysed by MALDI-TOF MS. Among potential candidates, Klebsiella oxytoca was identified in only one animal, so that it could be neglected in this context. Pasteurella pneumotropica was either entirely absent from cohorts used in a certain assays, or no significant differences between animals additionally carrying this bacterium and those being positive for B. pseudohinzii alone were noted. In line with this observation, experimental inoculation with Pasteurella pneumotropica caused histopathological pulmonary changes only in immunodeficient (NOD/ShiJic-scid/Jcl), but not in immunocompetent Crlj:CD1 (ICR) mice in another study 20 . Other potential lung pathogens were not identified in our study. Thus, we ascribe the histopathological and functional changes that we have observed to B. pseudohinzii.
Pulmonary histopathology and cellular composition of BAL fluid have been previously reported for mice with B. pseudohinzii-positive faecal PCR, with elevated neutrophils in the BAL fluid up to 20% in 10/12 animals and peribronchiolar infiltrates exceeding that seen in negative animals in 5/8 cases 9 . This is principally consistent with our current findings, although signs of inflammation were generally stronger in our cohort and we observed no overlap with negative animals in that none of the positive animals had neutrophil counts lower than 8%. Also,   pulmonary histopathology was not restricted to peribronchiolar infiltrates but presented as BALT formation and bronchointerstitial pneumonia. Similarly, bronchopneumonia with an increase in lymphoid tissue has been reported for mice taken from Japanese facilities, at that time classified as carrying B. hinzii 11,12 , now reconsidered as B. pseudohinzii 8,9 . Most Bordetella species affecting the respiratory tract, however, do not primarily target the peripheral lung but ciliated cells of the airway epithelium (e.g. B. pertussis 21 , B. bronchoseptica 22 , B. avium 23 ). Our data on explanted mouse tracheal rings demonstrate that kinocilia are also the primary site of attachment of B. pseudohinzii, with formation of bacterial chains indicative of replication. At bacterial CFU numbers corresponding to those which we determined from tracheas of infected mice, attachment to kinocilia was seen as early as 4 h of incubation. This was more pronounced in live videos taken in the course of analysis of ciliary function than in samples processed for electron microscopy, which may be due to the numerous washing steps included in the latter procedure. Dependent on time of incubation and bacterial load, extensive damage and loss of ciliated cells occurred, consistent with the observation of a reduced frequency of ciliated cells in the tracheas of B. pseudohinzii-positive mice.
Importantly, not only the number of ciliated cells was decreased in B. pseudohinzii-positive mice, but those remaining were functionally impaired so that cilia beat slower. Collectively, this resulted in more than 50% reduction in cilia-driven PTS, a readout parameter for mucociliary clearance. Our experiments on isolated tracheas from B. pseudohinzii-negative mice exposed for 4 h to this bacterium in culture revealed similar reduction in PTS and CBF, so that direct effects on ciliated cells appear to contribute more to the overall effect than epithelium remodelling.
Mucociliary clearance is a primary innate defence mechanism that eliminates inhaled particles, including bacteria, from the airways. It is driven by the coordinated beating of cilia that generate a vectorial fluid transport and propel the surface liquid layer consisting of a periciliary fluid and mucus towards the larynx 24,25 . Viscosity of the airway surface liquid, mainly determined by mucins, is another independent determinant of overall efficacy of mucociliary clearance. Our data were generated in mucus-free explanted tracheas, where solely the cilia-driven component is assessed 15,26,27 . While this offers the advantage to analyse in detail the mechanisms underlying the cilia-driven component, the PTS measured therein is not necessarily in direct linear relation to overall mucus or bacterial clearance from the trachea. Still, defects in ciliary function generally do have a negative effect on overall clearance, as it is exemplified in various forms of inherited ciliopathies 28 .
The closest relative of B. pseudohinzii, i.e. B. hinzii, has been isolated from respiratory secretions of cystic fibrosis patients which had impaired mucociliary clearance because of increased mucus viscosity due to their underlying disease 29,30 . Data on direct ciliotropic effects of B. hinzii, however, have yet not been reported in affected patients or in the natural hosts. However, the closely related species B. avium also attaches to kinocilia in explanted turkey tracheal rings and causes apoptosis of ciliated cells 31 . Intranasal inoculation of B. avium leads to extensive tracheitis with loss of ciliated cells 2,32 and, again, this is linked to impaired mucociliary clearance. Not in the early phase, but at 21 days after inoculation, overall clearance of inhaled 99m technetium-sulfur colloids was significantly reduced as assessed by scintigraphy 33 , indicating that epithelium remodelling rather than acute toxic effects is the underlying cause in this condition.
Despite the extensive airway inflammation with epithelial remodelling and a drastic functional impairment of an important clearance mechanism, the affected animals had no overt clinical symptoms in our facility. Microbiological diagnosis was initiated because of suspicious functional data sets rather than by compromised general condition of the animals. This is in accordance with a recent report from a facility in the United States, where increased neutrophil counts in BAL fluid in otherwise inconspicuous mice triggered diagnosis of B. pseudohinzii, revealing a prevalence of 80% (12/15) 9 . The index case for isolation of B. (pseudo)hinzii from mice (a female C57BL/6 mouse), however, presented sneezing with a chattering sound, and these clinical signs together with mild dyspnea were observed in all Jcl:ICR mice from day 3 up to at least 28 days after nasal inoculation with this bacterial isolate 11 . Among inoculated immunocompromised mice (NOD/ShiJic-scidJcl), 2 out of 10 were euthanised because of severe dyspnea and additional two died spontaneously in the experimental period (up to day 28) without prior signs of severe symptoms 11 . Thus, mice carrying this non-classical Bordetella strain present a broad range from subclinical colonization over modest respiratory infection to even lethal infection in immunocompromised animals.
In conclusion, our data demonstrate that B. pseudohinzii physically and functionally targets ciliated cells of the airway epithelium in C57BL/6 J mice. The affected animals present extensive airway inflammation with epithelial remodelling and functional impairment of clearance mechanisms, despite showing little to none overt clinical symptoms. Therefore, we feel it justified to recommend regular testing for this Bordetella species in animal facilities and aiming to exclude it from colonies that provide mice for experimental lung research. Otherwise, it may occur at high incidence in mouse colonies kept according to current FELASA recommendations. On the other hand, controlled colonization and infection with B. pseudohinzii may serve as a model to investigate mechanisms of compromised mucociliary clearance and microbial strategies to escape from this primary innate defence response.

Materials and Methods
Animals. All animals were housed under standard laboratory SPF conditions (10 h dark, 14 h light). In this facility, health monitoring of mice takes place quarterly following the FELASA recommendations for health monitoring of rodents 13 . Animals examined are aged stock animals. Serological monitoring via ELISA is performed for all viruses of the FELASA recommendations, and for Clostridium piliforme and Mycoplasma pulmonis. Examination for and differentiation of Helicobacter species is done via faecal PCR. Swab samples of trachea, caecum and vagina/prepuce are used for bacteriological cultivation and subsequent MALDI-TOF MS identification of cultivated agents. Fur, gross intestine and feces are examined microscopically for endo-and ectoparasites. Pathological examination of animals is performed yearly. By this monitoring the facility is identified as positive for Mouse norovirus, Helicobacter spp. (including H. hepaticus, H. rodentium, H. typhlonius), Bordetella hinzii, Klebsiella oxytoca, Pasteurella pneumotropica, Staphylococcus aureus, Entamoeba muris, Flagellates and Tritrichomonas muris.
All samples were taken after anaesthesia by isoflurane (5%) (Abbott, Wiesbaden, Germany) and killing of mice by cervical dislocation. Wild-type C57BL/6 J mice and five different transgenic mouse strains with C57BL/6 J background were tested for the incidence of B. pseudohinzii in our animal colonies.
For all functional experiments, wild-type C57BL/6 J mice of both sexes and at least 12 weeks of age were used. They were either bred in our animal facility or purchased from Janvier (St. Berthevin, France), and tested for B. pseudohinzii by bacteriological investigations including MALDI-TOF MS. Among the potentially pneumotropic bacteria listed above, Klebsiella oxytoca was identified in only one animal used for cell counts in BAL. Numbers of animals positive for Pasteurella pneumotropica are specified in detail for each experimental readout in the respective paragraph of the results chapter. All animals were held according to the European Community guidelines for the care and use of laboratory animals, and breeding and use of samples of sacrificed mice for further in vitro-experiments were registered by the responsible authorities at the Regierungspräsidium Giessen, Hesse, Germany.

Bacteriological investigations.
To examine bacterial colonization, the mice were dissected under sterile conditions and samples were taken from larynx, trachea and lung, around 20-40 mg each. In addition, samples from the pharynx were taken by wiping with a swab (Nerbe plus, Winsen, Germany).
To determine the total B. pseudohinzii CFU number per organ, tissue of animals tested positive for B. pseudohinzii by PCR was homogenized using a ball mill (MM 300; Retsch, Haan, Germany). The homogenate was then filtered using a cell strainer (100 µm pore size; Corning, Tewksbury, USA) and serial dilutions were plated onto selective agar plates (blood agar containing cephalexin hydrate [20 mg/ml; Sigma Aldrich]) and cultured for 48 h at 37 °C and 10% CO 2 . Bacterial colonies were counted and analyzed by MALDI-TOF MS.
Genome sequencing and bioinformatics analysis. Bordetella  Sequencing reads were quality clipped as described before and mapped onto aforementioned Bordetella genomes with bowtie2 37 . Afterwards, SNPs were detected via Samtools and BCFtools toolkits 38  Histopathology. Mice were sacrificed by cervical dislocation and the lungs were removed after ligation of the trachea to avoid alveolar collapse as described 41 . Lungs were immersion fixed for at least 12 h in 4% paraformaldehyde (Carl Roth) in 0.1 M phosphate buffer, washed in 0.1 M phosphate buffer and embedded in paraffin, cut in 2 µm thick horizontal sections and stained with hematoxylin and eosin (H&E) after dewaxing in xylene and rehydration in decreasing ethanol concentrations. An additional serial section was Giemsa-stained for visualization of bacterial DNA. Two pathologists (CG, KD) evaluated three evenly distributed sections per lung microscopically to assess the distribution and quality of pathological alterations using specified inflammation parameters such as tracheitis, total lung area affected, distribution of lung lesions, peribronchial, interstitial and intra-alveolar inflammation, infiltration by neutrophils, macrophages and the presence of BALT and lymphoid hyperplasia of lymph nodes.

Counting of ciliated tracheal epithelial cells.
To quantify the number of ciliated cells, the tracheas of C57BL/6 J mice were removed and fixed by immersion for at least 12 h in Zamboni fixative (2% formaldehyde [Roth] and 15% picric acid [Merck] in 0.1 M phosphate buffer). Afterwards, specimens were washed, paraffin embedded and 7 µm thick coronal sections were cut and stained with H&E. At least 400 nucleated epithelial cell profiles per animal were classified as belonging to either ciliated or non-ciliated cells; images were taken using a LEICA DM750 light microscope (Bensheim, Germany).
Cell counting in BAL fluid. Animals were sacrificed, the trachea was exposed and cannulated. Lungs were flushed twice with 800 µl of sterile PBS (Thermo Fisher Scientific, Waltham, USA) each. Cytospins (Cellspin ® I, Cellspin diagnostics, Waldsolms, Germany) were prepared from BAL fluid and Pappenheim-stained (Merck, Darmstadt, Germany). Differential cell counts were performed from at least 250 cells per sample using a Neubauer chamber.

Cultivation of B. pseudohinzii isolates.
Isolates of B. pseudohinzii were cultured overnight at 37 °C in BHI medium (Thermo Fisher Scientific). Thereafter, bacteria were diluted and cultured until they reached a phase of logarithmic growth, followed by three times washing with minimal essential medium (MEM; Thermo Fisher Scientific) without antibiotics. Bacterial infection dose was determined by plating bacterial suspension on BHI agar plates and counting CFU after 12 h of incubation at 37 °C.
In vitro-infection of tracheal explants. Tracheas of B. pseudohinzii-negative C57BL/6 J mice were explanted aseptically, surrounding tissue was removed and tracheas were cut into approximately 2 mm long pieces consisting of 2-3 cartilage rings. For electron microscopy, the rings were opened by cutting the trachealis muscle, washed in sterile PBS (Thermo Fisher Scientific) supplemented with penicillin (100 U/ml; PAA, Etobicoke, Canada) and streptomycin (0.1 mg/ml; PAA) and transferred into 6-well culture plates (Greiner bio-one, Kremsmünster, Austria). The rings were then submerged in phenol red-free MEM (Thermo Fisher Scientific) supplemented with penicillin, streptomycin and L-glutamine (2 mM) (Thermo Fischer Scientific) and cultured at 37 °C and 5% CO 2 for 24 h. After 24 h, viability of explants was validated by visualizing the beating of ciliated cells. Vital explants were then transferred into fresh plates and washed with subsequent addition of B. pseudohinzii cultures (strain 3227). Throughout, explants were kept in 1 ml of medium. For measurements of PTS and CBF of in vitro-infected tracheal explants, each trachea was divided into 2 halves. They were transferred into a delta T-dish, washed with PBS, bacteria were added and the samples were incubated for 4 h at 37 °C. PTS and CBF were examined as described below. Throughout, explants were kept in 1 ml of medium.
Electron microscopy. Tracheal explants were fixed for at least 4 h in a fixative mixture consisting of 1.5% glutardialdehyde (Merck) and 2.5% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4). After fixation, tissues were washed in HEPES, osmicated for 2 h in aqueous 1% osmium tetroxide (Sigma Aldrich) and washed in distilled water. For scanning electron microscopy, samples were dehydrated, critical point dried and sputtered with gold. Samples were examined by using Phillips XL30 and Zeiss EM9DSM982 scanning electron microscopes. For transmission electron microscopy, specimens were contrasted in 1% uranyl acetate (Merck) overnight, dehydrated and embedded in epon (Sigma Aldrich, St. Louis, USA). Sections of 80 nm thickness were cut using an ultramicrotome (Reichert Ultracut E, Leica, Bensheim, Germany), stained with alkaline lead citrate and viewed using a EM 902 (Zeiss, Wetzlar, Germany) transmission electron microscope equipped with a slow scan 2 K CCD camera (TRS; Tröndle, Moorenweis, Germany). Semithin sections of 0.8 µm thickness were cut using the same ultramicrotome and stained with methylene blue (Carl Roth, Karlsruhe, Germany) for light microscopic evaluation (LEICA DM750, Bensheim, Germany).
Negative staining of B. pseudohinzii. B. pseudohinzii (strain 3227) was cultured in BHI medium, fixed with 4% paraformaldehyde solution and washed in 0.1 M phosphate buffer. This bacterial suspension was placed on a Formvar-coated grid (Plano GmbH, Wetzlar, Germany), stained with 1% ammonium heptamolybdate (Merck) for 30 s, and analysed by transmission electron microscopy (EM 902).

Measurement of PTS.
A modified version of the previously published method was used 15 . Briefly, the explanted trachea was transferred in a delta T-dish with a thin layer of Sylgard polymer (Dow Corning, Wiesbaden, Germany) on its glass bottom. After the surrounding tissue was removed, tracheas were oriented with the m. trachealis facing upward, fixed with insect needles and submerged in HEPES (pH 7.4). The temperature of the buffer was constantly held at 31 °C. The muscle was cut and the trachea was fixed flat on the T-dish. Polysterene dynabeads with a diameter of 2.8 µm (Invitrogen, Carlsbad, USA) were added before measurements.
For the freshly explanted tracheas, the first measurement was taken 30 min after the mouse was sacrificed, followed by measurements every 5 min until 55 min after start, then every 2 min. ATP (100 µM) was added at minute 59. Drugs and dynabeads were well mixed before every measurement. For measuring the baseline PTS of in vitro-infected tracheas, three videos of three different areas were taken after 2 min of equilibration time. For each measurement, videos consisting of 200 images (640 × 512 pixels; 12 bit; 1 image per 84 ms) were taken with using UMPLFL20xW/0.5 water immersion objective (Olympus, Shinjuku, Japan) and a SMX-150M (EHD imaging GmbH, Damme, Germany) camera. Dynabeads were tracked and visualized by using Image-Pro Plus (MediaCybernetics, Rockville, USA) software, which calculated the average speed of all tracked particles.

Measurement of CBF.
Preparation of tracheas and microscope setup were the same as for PTS measurements, except that images were taken at higher magnification (UMPLFL40xW/0.8 water immersion objective; Olympus, Shinjuku, Japan). Videos were taken before adding the dynabeads for the PTS measurements. For each animal, one video was taken. To investigate the baseline CBF of in vitro-infected tracheas, two videos of two different areas were taken after 2 min of equilibration time. The videos consisted of 1000 images with a frame rate of 105 images/s. During the measurements, no drugs were added, in order to determine a baseline CBF. For calculation of mean CBF, the dominating beat frequency of each single ciliated cell (30-45 cells per video) was analysed by fast Fourier transformation using a Graphical User Interface (GUI) for MATLAB R2016b which was programmed by Peter König (University of Lübeck, Germany).
Statistical Analysis. Data are presented as individual data points in figures and as mean ± standard error of the mean (SEM). Statistical analyses of data were performed with GraphPad Prism version 6 (La Jolla, USA). Data were analysed for normal distribution by the Kolmogorov-Smirnov test followed by two tailed Student's unpaired t-test.