miR449 Protects Airway Regeneration by Controlling AURKA/HDAC6-Mediated Ciliary Disassembly

Airway mucociliary regeneration and function are key players for airway defense and are impaired in chronic obstructive pulmonary disease (COPD). Using transcriptome analysis in COPD-derived bronchial biopsies, we observed a positive correlation between cilia-related genes and microRNA-449 (miR449). In vitro, miR449 was strongly increased during airway epithelial mucociliary differentiation. In vivo, miR449 was upregulated during recovery from chemical or infective insults. miR0449−/− mice (both alleles are deleted) showed impaired ciliated epithelial regeneration after naphthalene and Haemophilus influenzae exposure, accompanied by more intense inflammation and emphysematous manifestations of COPD. The latter occurred spontaneously in aged miR449−/− mice. We identified Aurora kinase A and its effector target HDAC6 as key mediators in miR449-regulated ciliary homeostasis and epithelial regeneration. Aurora kinase A is downregulated upon miR449 overexpression in vitro and upregulated in miR449−/− mouse lungs. Accordingly, imaging studies showed profoundly altered cilia length and morphology accompanied by reduced mucociliary clearance. Pharmacological inhibition of HDAC6 rescued cilia length and coverage in miR449−/− cells, consistent with its tubulin-deacetylating function. Altogether, our study establishes a link between miR449, ciliary dysfunction, and COPD pathogenesis.


Introduction
Cilia are highly conserved, microtubule-based surface organelles with important mechanical and sensory functions. Almost all mammalian cells build up a single, immotile primary cilium, while specialized epithelial cells lining the airways, reproductive ducts, and brain ventricles form multiple motile cilia per cell that promote directional fluid flow and particle transport along the multiciliated surfaces by coordinated beating. Each cilium on a multiciliated cell (MCC) is nucleated by a basal body that is derived from a centriole; hence, cell cycle exit is a prerequisite for cilia formation in most of the cells [1]. However, upon injury, cilia are disassembled to permit cell proliferation and epithelial repopulation prior to differentiation [2]. Motile cilia are composed of a characteristic axonemal 9 + 2 microtubule architecture that polymerizes from αand β-tubulin dimer. Another important class of components within the complex ciliary nanomachine are the axonemal dyneins, generating the ciliary beat [1].
In the respiratory system vigorous beating of multiple cilia transports inhaled pollutants and pathogens out of the airways, thereby acting as a primary innate defense mechanism. Functional defects in airway MCCs impair mucociliary clearance and are associated with human pathologies, e.g., primary ciliary dyskinesia (PCD) [3]. As well as MCCs, the pseudostratified epithelium of the conducting airways is lined by club, goblet, and basal cells. All these epithelial cell types are repeatedly subjected to injury by inhaled environmental toxicants and pathogens as well as systemically administered xenobiotics. Hence, rapid and balanced regeneration of the epithelial surfaces is essential to maintain respiratory function [4] and healthy lung aging [5].
The events that trigger multiciliogenesis and lead to its dysfunction in respiratory disease depend on a hierarchical network of transcriptional and post-transcriptional mRNA regulators [6]. Such post-transcriptional regulators include microRNAs (miRNAs), a class of small regulatory noncoding RNAs. miRNAs exert their function by binding to the three prime untranslated regions (3 UTR) of their target mRNAs, thereby causing mRNA degradation or translational repression [7]. In mammalian cells, in most cases, miRNA bind their target with imperfect complementarity. One miRNA family involved in the regulation of multiciliogenesis is the miR34/449 family derived from three genomic loci encoding six homologous miRNAs (miR34a, miR34b,c, and miR449a/b/c) [8][9][10][11][12][13]. The miR449 cluster is expressed from the second intron of cell division cycle 20B (CDC20B) on chromosome 13 D2.2 in mice (5q11.2 in humans) [14]. This locus contains other key regulators of multiciliogenesis including Cyclin O (CCNO) and multiciliate differentiation and DNA synthesis-associated cell cycle protein (MCIDAS) [15][16][17]. The importance of this locus in multiciliogenesis is highlighted by the fact that patients harboring mutations in CCNO or MCIDAS develop congenital mucociliary clearance disorders with the reduced generation of multiple motile cilia [16,18]. Moreover, miR449 is by far the most strongly induced miRNA during mucociliary differentiation [8,9]. The depletion of all members of the miR34/449 family in mice demonstrated their essential function in multiciliogenesis, resulting in infertility, severe respiratory dysfunctions caused by defective mucociliary clearance, and early death from respiratory failure [10][11][12][13].
Increasing air pollution [19] and the demographic shift towards an aging population [20] are associated with the increased prevalence of chronic respiratory diseases worldwide, notably COPD. The prevalence, morbidity, and mortality of COPD are high,

Cilia-Related Genes Positively Correlate with miR449 Expression in COPD Patients
To elucidate whether miR449 plays a role in cilia-associated processes, we analyzed genome-wide miRNAs and mRNAs of bronchial biopsies from 57 COPD patients included in the GLUCOLD study [24] (Figure 1a and Supplementary Table S1, for details of GLUCOLD study, see Method section). We found 1444 genes to be positively correlated with miR449a (FDR < 0.05). In order to determine the biological processes in which the 1444 miR449a-correlated genes are involved, gene set enrichment analysis (GSEA) [25] was conducted using a list of genes ranked on the basis of the strength of their correlation to miR449a. Several gene sets associated with ciliogenesis showed an enrichment among the genes positively correlated to miR449a, including the first five processes shown in Figure 1b. As expected from previous data [8,[10][11][12], genes positively correlated with other miR34/449 family members, except for miR34a, were also enriched for cilia-associated processes (Supplementary Tables S2 and S3). To identify genes through which miR449a regulates motile ciliogenesis, we compared miR449a-positively correlated genes (1444) with a published list containing 581 cilia-associated genes [26] and identified 135 overlapping genes related to the structure, maintenance, and function of cilia (Figure 1c). Among the most significant genes was the dynein intermediate chain 1 (DNAI1), encoding an outer dynein arm protein involved in cilia beating [27] (Figure 1c). In conclusion, miR449a is positively correlated with ciliary genes in COPD patients.   . miR449 expression is positively correlated with genes associated with ciliogenesis in COPD patients. (a) Schematic overview of the GLUCOLD study design. Microarray analysis was performed for mRNAs and miRNAs using bronchial biopsies, which were collected from 57 COPD patients diagnosed "with moderate to severe symptoms". (b) Enrichment of the top 10 biological processes among miR449a-positively correlated genes. The GSEA was conducted using the Gene Ontology gene set for biological processes and the list of genes ordered by the strength of their correlation to miR449a. Green bars mark cilia-associated processes. All displayed gene sets for biological processes are significant with FDR < 0.001. (c) Comparison between miR449a-positively correlated genes in COPD patients and cilia-associated genes identifies 135 overlapping ciliary genes [26]. Left: table displays the top 20 significantly overlapped ciliary genes (out of 135). Right: illustration of the intraflagellar transport, cross-section of a motile cilium, and of a microtubule doublet highlights the location of some of the top 20 overlapped ciliary genes positively correlated with miR449a. Lower right panel: miR449a expression is positively correlated with DNAI1. Significance was determined by FDR < 0.05.

miR449 Is Induced during Airway Epithelial Differentiation and Regeneration upon Bronchial Challenges
The positive correlation between cilia-associated genes and miR449 in COPD patients is in line with previously published data showing miR449 expression in multiciliated epithelia [8][9][10]. To verify miR449 expression during multiciliated epithelia formation, we analyzed its expression by in situ hybridization (ISH) on mouse embryos (embryonic (E) day 18.5) and found that miR449a is enriched in multiciliated tissues, perceptibly in the developing airway epithelium (Figure 2a). Previous studies have already revealed that the miR449a,b,c cluster is induced far more strongly than the miR34 members during mucociliary differentiation [8,9,11,12]. We confirmed these data in our current study by culturing human airway epithelial cells collected from healthy donors at the air-liquid interface (ALI) for 23 days, a time point when human ALI cultures are fully differentiated ( Figure 2b). Moreover, RNA-sequencing data from ALI cultures of mouse tracheal epithelial cells (MTECs) supported these results [26] (Figure 2c). The onset of miR449 expression occurs along with the expression of other known players of multiciliogenesis including Geminin coiled-coil domain containing (Gemc1), Mcidas, and forkhead box J1 (Foxj1) [6,17]. Notably, the total read counts of miR449 were most strongly upregulated compared to the other pro-ciliogenic factors including miR34b,c and were at their highest between day 4 and 7 ( Figure 2c). These results indicate that miR449 has a much stronger contribution to mucociliary differentiation compared with its homologues. Therefore, we focused our following analyses on investigating the role of miR449 in ciliated airway epithelia regeneration. One commonly used airway regeneration mouse model is the exposure to naphthalene, which leads to extensive club cells' exfoliation [28] and the squamation of ciliated cells accompanied by motile cilia disassembly [2]. Thus, we treated wild type (WT) mice with naphthalene and assessed the lungs at different stages during the airway epithelia regeneration process, which is normally completed after 14 days (Figure 2d). miR449a expression levels immediately increased on day 1 (d1) after naphthalene injection and continued to accumulate until d7, where the expression of miR449a peaked (Figure 2e).
Air pollutants such as diesel exhaust particles (DEPs), which are a major constituent of traffic-related particulate matter, represent another environmental challenge for the respiratory epithelium. DEPs are known to result in acute airway inflammation and the reduction in ciliated and non-ciliated cells [29], as well as a decrease in cilia beat frequency [30]. Due to the influence of DEPs on ciliated bronchial epithelial regeneration, we examined miR449 expression in WT mice after DEP treatment ( Figure 2f). Indeed, transcript levels of miR449a and its host gene cdc20b were strongly increased in DEPexposed WT mice (Figure 2g). In summary, the data reveal an upregulation of miR449 during mucociliary differentiation and upon bronchial challenges, implying that miR449 is involved in the regulation of these two processes. or saline (Ctrl) on day 1, 4, and 7, and analyzed at d9. (g) Mature miR449a and Cdc20b levels were quantified in DEP-and saline-treated lungs (Ctrl), and normalized to U6 or 36b4, respectively. Ctrl levels are used as reference. n = 6 (Ctrl), n = 8 (DEP) mice per group. All data are presented as the mean ± SEM with * p < 0.05 and ** p < 0.01.

miR449 Deficiency Results in Impaired Ciliation in ALI Cultures
To further investigate the role of miR449 in ciliogenesis, we utilized the miR449 −/− mouse model, in which only miR449a, b, and c (miR449 from here on) are selectively depleted, whereas miR34a and miR34b,c [10] are not (Figure 3a,b). Previous studies showed that the expression of miR34b and c is increased upon miR449 depletion and this can compensate for the loss of miR449 in the testis [11,31]. Therefore, miR449 −/− mice are fertile, while double knockout miR34b,c −/− ; miR449 −/− mice are infertile [10][11][12]. Interestingly, the lungs of miR449 −/− mice, in contrast to the testis, did not show increased expression of miR34 members (Figure 3b), as previously shown [11]. Hence, miR449 −/− mice provide the possibility to study miR449-mediated cilia regeneration and pulmonary function without effects resultant from miR34 changes. We measured cilia quantity and cilia length in ALI cultures generated from cells of WT and miR449 −/− mice at d6, a timepoint of the highest miR449 expression in WT cells (Figure 2c). Immunofluorescence (IF) staining for acetylatedalpha-tubulin (Ac-α-TUB) and DNAI1 (both axonemal cilia markers) revealed a significant decrease in the number and length of motile cilia in miR449 −/− compared with WT ALI cultures (Figure 3c-f and Supplementary Figure S1a). Thus, miR449 deficiency impairs the process of multiciliated epithelium generation in vitro. that the expression of miR34b and c is increased upon miR449 depletion and this can compensate for the loss of miR449 in the testis [11,31]. Therefore, miR449 −/− mice are fertile, while double knockout miR34b,c −/− ;miR449 −/− mice are infertile [10][11][12]. Interestingly, the lungs of miR449 −/− mice, in contrast to the testis, did not show increased expression of miR34 members (Figure 3b), as previously shown [11]. Hence, miR449 −/− mice provide the possibility to study miR449-mediated cilia regeneration and pulmonary function without effects resultant from miR34 changes. We measured cilia quantity and cilia length in ALI cultures generated from cells of WT and miR449 −/− mice at d6, a timepoint of the highest miR449 expression in WT cells ( Figure 2c). Immunofluorescence (IF) staining for acetylated-alpha-tubulin (Ac-α-TUB) and DNAI1 (both axonemal cilia markers) revealed a significant decrease in the number and length of motile cilia in miR449 −/− compared with WT ALI cultures (Figure 3c-f and Supplementary Figure S1a). Thus, miR449 deficiency impairs the process of multiciliated epithelium generation in vitro.

miR449 Is Essential for Bronchial Epithelial Regeneration
The formation and proper regeneration of the multiciliated airway epithelium is extremely important for normal respiratory function since the airways are continuously exposed to environmental insults. Since miR449 is upregulated upon environmental stimuli and mucociliary differentiation is impaired in miR449 −/− mouse derived ALIs, we next examined the potential of ciliated bronchial epithelial regeneration in miR449 −/− mice after naphthalene treatment (Figure 2d). Airway cilia coverage was assessed by IF staining and immunoblot for Ac-α-TUB and DNAI1, respectively, at different regeneration stages. Both axonemal cilia markers were significantly decreased after naphthalene injury (d1-d7), and even at the stage of full bronchial epithelium regeneration in WT mice (d14), ciliation in miR449 −/− mice remained drastically reduced (Figure 4a-c). Hematoxylin and eosin staining did not reveal any obvious histological differences between WT and miR449 −/− lungs after naphthalene injury (Supplementary Figure S2a), although miR449 −/− mice experienced higher weight loss in comparison to WT mice (Supplementary Figure S2b). These findings demonstrate that miR449 is necessary for adequate bronchial epithelial regeneration.
In addition to toxic and pollution particles, the airway epithelium is constantly exposed to pathogenic bacteria. Therefore, in our next experiment we studied whether miR449 −/− mice are more vulnerable to nontypeable Haemophilus influenzae (NTHi) (Figure 4d), an opportunistic respiratory bacterium altering cilia beating and associated with chronic airway diseases such as COPD [32]. Motile cilia coverage and expression of cilia markers were indeed significantly reduced in lungs of NTHi-exposed miR449 −/− mice compared to lungs of NTHi-exposed WT mice, as shown by IF for Ac-α-TUB, immunoblot for DNAI1, and gene expression analysis of the outer dynein arm component dynein axonemal heavy

miR449 Is Essential for Bronchial Epithelial Regeneration
The formation and proper regeneration of the multiciliated airway epithelium is extremely important for normal respiratory function since the airways are continuously exposed to environmental insults. Since miR449 is upregulated upon environmental stimuli and mucociliary differentiation is impaired in miR449 −/− mouse derived ALIs, we next examined the potential of ciliated bronchial epithelial regeneration in miR449 −/− mice after naphthalene treatment ( Figure 2d). Airway cilia coverage was assessed by IF staining and immunoblot for Ac-α-TUB and DNAI1, respectively, at different regeneration stages. Both axonemal cilia markers were significantly decreased after naphthalene injury (d1-d7), and even at the stage of full bronchial epithelium regeneration in WT mice (d14), ciliation in miR449 −/− mice remained drastically reduced (Figure 4a-c). Hematoxylin and eosin staining did not reveal any obvious histological differences between WT and miR449 −/− lungs after naphthalene injury (Supplementary Figure S2a), although miR449 −/− mice experienced higher weight loss in comparison to WT mice (Supplementary Figure S2b). These findings demonstrate that miR449 is necessary for adequate bronchial epithelial regeneration.
In addition to toxic and pollution particles, the airway epithelium is constantly exposed to pathogenic bacteria. Therefore, in our next experiment we studied whether miR449 −/− mice are more vulnerable to nontypeable Haemophilus influenzae (NTHi) ( Figure  4d), an opportunistic respiratory bacterium altering cilia beating and associated with chronic airway diseases such as COPD [32]. Motile cilia coverage and expression of cilia markers were indeed significantly reduced in lungs of NTHi-exposed miR449 −/− mice compared to lungs of NTHi-exposed WT mice, as shown by IF for Ac-α-TUB, immunoblot for DNAI1, and gene expression analysis of the outer dynein arm component dynein axonemal heavy chain 5 (Dnah5) (Figure 4e-g).  (j) Immunoblot for DNAI1 in lung lysates from 6-month-old WT and miR449 −/− mice; constitutive protein HSC70 was used as a loading control. All data are presented as the mean ± SEM and relative to the WT group with * p < 0.05 and *** p < 0.001.
Although airway epithelial turnover is relatively slow, the epithelium needs to be regularly replaced due to constant challenges during the breathing process, and epithelial regeneration is also important during aging. We thus hypothesized that miR449 depletion could impair airway ciliation in aged mice. We analyzed IF staining for the Ac-α-TUB and DNAI1 of tracheal sections from 6-month-old WT and miR449 −/− mice. We observed a significant reduction in Ac-α-TUB and an almost complete loss of DNAI1 in 6-month-old miR449 −/− mice (Figure 4h,i), despite being barrier maintained under strict pathogen-free conditions including cage content autoclaving and air filtration. Immunoblot analysis for DNAI1 in lung lysates from WT and miR449 −/− mice confirmed the significant reduction in ciliation in 6-month-old mice upon miR449 depletion (Figure 4j). In contrast, ciliation appeared normal in 10-week-old miR449 −/− mice (Supplementary Figure S3a,b). Overall, our data suggest a supportive effect of miR449 in bronchial epithelial repair upon different challenges including aging. in lung lysates from 6-month-old WT and miR449 −/− mice; constitutive protein HSC70 was used as a loading control. All data are presented as the mean ± SEM and relative to the WT group with * p < 0.05 and *** p < 0.001.
Although airway epithelial turnover is relatively slow, the epithelium needs to be regularly replaced due to constant challenges during the breathing process, and epithelial regeneration is also important during aging. We thus hypothesized that miR449 depletion could impair airway ciliation in aged mice. We analyzed IF staining for the Ac-α-TUB and DNAI1 of tracheal sections from 6-month-old WT and miR449 −/− mice. We observed a significant reduction in Ac-α-TUB and an almost complete loss of DNAI1 in 6-month-old miR449 −/− mice (Figure 4h,i), despite being barrier maintained under strict pathogen-free conditions including cage content autoclaving and air filtration. Immunoblot analysis for DNAI1 in lung lysates from WT and miR449 −/− mice confirmed the significant reduction in ciliation in 6-month-old mice upon miR449 depletion (Figure 4j). In contrast, ciliation appeared normal in 10-week-old miR449 −/− mice (Supplementary Figure S3a,b). Overall, our data suggest a supportive effect of miR449 in bronchial epithelial repair upon different challenges including aging.

Spontaneous COPD Development, Associated with an Increased Inflammatory Reaction upon
Challenge, in miR449 −/− Mice A reduced lung function as a long-term consequence of COPD is caused by an impaired pulmonary tissue and airway regeneration. In this context, we show the positive effect of miR449 on bronchial regeneration. We therefore tested whether the loss of miR449 has any effect on lung function. Aged WT and miR449 −/− mice were subjected to invasive lung function while inhaling ambient air (AA). WT mice exposed to CS for 6 months were used as a positive control due to the loss of lung function and resulting tissue destruction [33] ( Figure 5a). Pulmonary function was impaired to a similar extent in miR449 −/− mice that were exposed to AA as well as in CS-exposed WT mice, which was indicated by an increase in total lung capacity and reduction in tissue elastance (Figure 5b,c), whereas no such phenotype observation could be made in young miR449 −/− mice (Supplementary Figure S4a,b). By stereology, we assessed serial histological lung sections by means of the average chord length, thereby confirming the development of spontaneous emphysematous changes in aged miR449 −/− mice, characterized by airspace enlargement (Figure 5d), which represents one hallmark of COPD. Usually, this condition is associated with an increased infiltration of immune cells, i.a., macrophages and neutrophils, which potentially contribute to the imbalance between proteases and antiproteases within the lungs [34]. We therefore counted the inflammatory cells that were present within the bronchoalveolar lavage fluid (BALF). Under specific pathogen-free conditions, it mainly consisted of macrophages. While the number of macrophages was significantly increased in miR449 −/− compared with WT mice, obviously, the number was not increased to a similar extent as in CS-exposed WT mice ( Figure 5e). The pulmonary inflammatory response in mice upon miR449 deficiency was further analyzed in mice treated with NTHi, an administration used to boost inflammation (as described in Figure 4j). NTHi treatment resulted in elevated protein expression of M2A macrophage marker CD206 in the lungs of miR449 −/− compared with WT mice (Figure 5f,g). Accordingly, the transcript expression of the matrix metalloproteinase (Mmp) 12 (Mmp12; macrophage elastase) and Mmp9 (gelatinase), which are associated with COPD and expressed by macrophages [34], was significantly increased in NTHi-treated miR449 −/− mice ( Figure 5h). Contrariwise, we observed unchanged levels of antiprotease upon miR449 depletion, which were measured by the tissue inhibitor of metalloproteinases 1 (TIMP1) protein expression (Figure 5i,j), suggesting no opposing reaction to the observed increase in the protease level. Together, our data strongly suggest that, with an increasing age, miR449 −/− mice exhibit spontaneous manifestations of COPD. These changes are associated with increased pulmonary MMPs and an increased inflammatory response.

Aurora Kinase A Contributes to miR449-Driven Epithelial Regeneration Processes and Ciliary Homeostasis
Next, we aimed to investigate the underlying mechanism of defective mucociliary regeneration in miR449 −/− mice. In order to avoid pathologies, a balance between proliferation and differentiation is indispensable for airway regeneration, which is, in turn, related to ciliary disassembly and re-assembly. The role of miR449 in these key processes was studied in a simplified in vitro model given the lack of convenient models available to quantify and analyze the assembly and disassembly of motile cilia. For this purpose, mouse embryonic fibroblasts (MEFs) can be used to monitor and subsequently measure the assembly and disassembly of primary cilia. This is achieved by rounds of serum withdrawal and serum addition, causing cell cycle exit and cell cycle re-entry, respectively [35] (Figure 6a). In the following, MEFs of miR449 −/− and WT mice were analyzed with regard to their cilia assembly and disassembly at the respective points in time, achieved by IF staining of cilia components belonging to the basal body (pericentrin) as well as the axoneme (Ac-a-TUB). Surprisingly, no change in the assembly rate could be observed between both genotypes, whereas ciliary disassembly in miR449 −/− MEFs was significantly increased, in turn resulting in a reduced ciliation (Figure 6b,c). Structural similarities between primary and motile cilia imply mutual mechanisms regarding cilia assembly and disassembly [1]; thus, an increase in the primary cilia disassembly rate could serve as an explanation for the reduction in motile cilia observed in miR449 −/− mice.
cantly increased in NTHi-treated miR449 −/− mice (Figure 5h). Contrariwise, we observed unchanged levels of antiprotease upon miR449 depletion, which were measured by the tissue inhibitor of metalloproteinases 1 (TIMP1) protein expression (Figure 5i,j), suggesting no opposing reaction to the observed increase in the protease level. Together, our data strongly suggest that, with an increasing age, miR449 −/− mice exhibit spontaneous manifestations of COPD. These changes are associated with increased pulmonary MMPs and an increased inflammatory response. The treatment of WT mice at the age of 4 months took place over a period of 6 months, for 5 h/d and 5 d/w. WT mice were either exposed to CS or maintained at AA, whereas miR449 −/− were exposed, during this period, to AA. A positive control for the development of the COPD phenotype were WT mice exposed to CS. (b,c) Tissue elastance (b) as well as total lung capacity (c) were measured to assess pulmonary function using miR449 −/− , CS-exposed WT and WT mice at 10 months of age. n = 4-5 mice/genotype. Analysis was conducted using the FlexiVent system. (d) By measurements of the mean chord length, emphysema was analyzed in miR449 −/− , CS-exposed WT and WT mice at the age of 10 months. n = 4 mice/genotype. (e) Counting of macrophages in BALF of miR449 −/− , CS-exposed WT and WT mice at 10 months of age. n = 5 mice/genotype. (f) Immunoblot analysis of M2 macrophage marker CD206 in lysates of WT and miR449 −/− mice following NTHi treatment. HSC70 was used as a loading control and quantification thereof (g), respectively; WT serves as reference. (h) Relative mRNA expression of Mmp9 and Mmp12 in WT and miR449 −/− mice treated with NTHi, normalized to 36b4. WT levels serve as a reference. n = 4 mice/genotype. (i) Immunoblot analysis of TIMP1 protein expression and relative quantification thereof (j). HSC70 was used as loading control. All values are presented as the mean ± SEM and relative to the WT group with * p < 0.05, ** p < 0.01, and *** p < 0.001.

Aurora Kinase A Contributes to miR449-Driven Epithelial Regeneration Processes and Ciliary Homeostasis
Next, we aimed to investigate the underlying mechanism of defective mucociliary regeneration in miR449 −/− mice. In order to avoid pathologies, a balance between proliferation and differentiation is indispensable for airway regeneration, which is, in turn, related to ciliary disassembly and re-assembly. The role of miR449 in these key processes was studied in a simplified in vitro model given the lack of convenient models available to quantify and analyze the assembly and disassembly of motile cilia. For this purpose, mouse embryonic fibroblasts (MEFs) can be used to monitor and subsequently measure the assembly and disassembly of primary cilia. This is achieved by rounds of serum withdrawal and serum addition, causing cell cycle exit and cell cycle re-entry, respectively [35] ( Figure 6a). In the following, MEFs of miR449 −/− and WT mice were analyzed with regard to their cilia assembly and disassembly at the respective points in time, achieved by IF staining of cilia components belonging to the basal body (pericentrin) as well as the axoneme (Ac-a-TUB). Surprisingly, no change in the assembly rate could be observed be- Figure 5. miR449 deficiency results in a spontaneous COPD phenotype with increased inflammatory response. (a) Schematic illustration of the COPD mouse model, smoke-induced. The treatment of WT mice at the age of 4 months took place over a period of 6 months, for 5 h/d and 5 d/w. WT mice were either exposed to CS or maintained at AA, whereas miR449 −/− were exposed, during this period, to AA. A positive control for the development of the COPD phenotype were WT mice exposed to CS. (b,c) Tissue elastance (b) as well as total lung capacity (c) were measured to assess pulmonary function using miR449 −/− , CS-exposed WT and WT mice at 10 months of age. n = 4-5 mice/genotype. Analysis was conducted using the FlexiVent system. (d) By measurements of the mean chord length, emphysema was analyzed in miR449 −/− , CS-exposed WT and WT mice at the age of 10 months. n = 4 mice/genotype. (e) Counting of macrophages in BALF of miR449 −/− , CS-exposed WT and WT mice at 10 months of age. n = 5 mice/genotype. (f) Immunoblot analysis of M2 macrophage marker CD206 in lysates of WT and miR449 −/− mice following NTHi treatment. HSC70 was used as a loading control and quantification thereof (g), respectively; WT serves as reference. (h) Relative mRNA expression of Mmp9 and Mmp12 in WT and miR449 −/− mice treated with NTHi, normalized to 36b4. WT levels serve as a reference. n = 4 mice/genotype. (i) Immunoblot analysis of TIMP1 protein expression and relative quantification thereof (j). HSC70 was used as loading control. All values are presented as the mean ± SEM and relative to the WT group with * p < 0.05, ** p < 0.01, and *** p < 0.001.
Given the increase in the cilia disassembly rate of MEFs from miR449 −/− mice, we aimed to understand whether miR449 regulates key players of the cilia disassembly pathway [35,36] (Figure 6d). Potential targets of miR449 were analyzed with the help of published data [10,12,37,38] as well as in silico prediction tools (TargetScan, PITA) [39,40], subsequently identifying a subset of targets including murine and human histone deacetylase 6 (HDAC6), AURKA, and potassium voltage-gated channel subfamily H member 1 (KCNH1) (Supplementary Figure S5a). Notably, miRNAs that target multiple nodes within one pathway are more likely to provoke major effects on this pathway. The simple reason behind this is that individual miRNA targets are usually only repressed to a small degree [41]. Due to its essential role in cilia disassembly and HDAC6 activation [42][43][44][45], as well as its upregulation in miR34 −/− /449 −/− mice [10,12], we mainly focused on the miR449 target AURKA in the following analysis. miR449 overexpression in human lung carcinoma cells (H1299) was performed to prove AURKA regulation by miR449, and, indeed, the protein as well as the mRNA expression of AURKA was strongly reduced upon overexpression of miR449 (Figure 6e,f). Appropriately, in lung tissue from miR449 −/− mice, aged and naphthalene-treated, AURKA protein levels were elevated when compared with WT controls (Figure 6g,h). Taken together, these findings suggest reduced airway ciliation to be a consequence of increased ciliary disassembly, triggered by miR449 deficiency, and probably due to an upregulation of the AURKA/HDAC6-driven disassembly process. disassembly [1]; thus, an increase in the primary cilia disassembly rate could serve as an explanation for the reduction in motile cilia observed in miR449 −/− mice.
Given the increase in the cilia disassembly rate of MEFs from miR449 −/− mice, we aimed to understand whether miR449 regulates key players of the cilia disassembly pathway [35,36] (Figure 6d). Potential targets of miR449 were analyzed with the help of published data [10,12,37,38] as well as in silico prediction tools (TargetScan, PITA) [39,40], subsequently identifying a subset of targets including murine and human histone deacetylase 6 (HDAC6), AURKA, and potassium voltage-gated channel subfamily H member 1 (KCNH1) (Supplementary Figure S5a). Notably, miRNAs that target multiple nodes within one pathway are more likely to provoke major effects on this pathway. The simple reason behind this is that individual miRNA targets are usually only repressed to a small degree [41]. Due to its essential role in cilia disassembly and HDAC6 activation [42][43][44][45], as well as its upregulation in miR34 −/− /449 −/− mice [10,12], we mainly focused on the miR449 target AURKA in the following analysis. miR449 overexpression in human lung carcinoma cells (H1299) was performed to prove AURKA regulation by miR449, and, indeed, the protein as well as the mRNA expression of AURKA was strongly reduced upon overexpression of miR449 (Figure 6e,f). Appropriately, in lung tissue from miR449 −/− mice, aged and naphthalene-treated, AURKA protein levels were elevated when compared with WT controls (Figure 6g,h). Taken together, these findings suggest reduced airway ciliation to be a consequence of increased ciliary disassembly, triggered by miR449 deficiency, and probably due to an upregulation of the AURKA/HDAC6-driven disassembly process.   [35]). Ca 2+ influx into the cytoplasm is triggered by growth factor stimulation and results in binding of Ca 2+ to calmodulin (CaM). CaM subsequently binds to, and thus activates AURKA by autophosphorylation, which, in turn, phosphorylates HDAC6. In the following, HDAC6 removes acetyl groups from α-tubulin, which triggers ciliary disassembly [35,44]. Kv10.1 (KCNH1) promotes, via an unknown mechanism, ciliary disassembly [36]. (e,f) H1299 cells were transfected with pre-miRNAs from the miR449 family in addition to a natural miRNA (miR302*) without predicted AURKA binding sites and a synthetic control sequence (Ctrl; NC#2), both as references. Forty-eight hours post-transfection, lysates were harvested and AURKA protein and mRNA expression were analyzed in immunoblots (e) and by qPCR ((f), n = 3). (g) Immunoblot analysis of AURKA protein levels in lungs of 6-month-old miR449 −/− and WT mice, and relative quantification thereof. β-ACTIN was used as loading control. n = 4 mice per genotype. (h) AURKA protein expression in lung lysates isolated from naphthalene-treated WT and miR449 −/− mice at d7. β-ACTIN was used as loading control. All data are presented as the mean ± SEM and relative to the WT group with * p < 0.05 and ** p < 0.01.

Mice Lacking miR449 Develop Ultrastructural Cilia Defects and Impaired Mucociliary Clearance
AURKA upregulation causes increased cilia disassembly, resulting in primary cilia instability due to the faulty ultrastructure [45,46]. Electron microscopy of tracheal sections showed aberrant cilia with detached plasma membranes (named aberrant cilia from here on) in mice lacking miR449, while the typical 9 + 2 microtubule arrangement was maintained (Figure 7a,b). Mucociliary clearance disorders can be caused by defects in structural components of motile cilia [3]. The velocity of fluorescently labelled beads was measured over WT and miR449 −/− tracheae to determine mucociliary transport, with only areas with beating cilia investigated. Indeed, in miR449 −/− mice, bead transport was drastically reduced (Figure 7c,d), indicating impaired clearance. Taken together, these findings show that cilia shape and function are harmed when miR449 is depleted.

In miR449 −/− Airway Cultures' Ciliation Can Be Rescued by HDAC6 Inhibition
We investigated whether airway ciliation may be restored by blocking the AURKA/HDAC6mediated cilia disassembly axis. To that aim we used the commercially available selective HDAC6 inhibitor tubastatin [47] to treat ALI cells as HDAC6 is a downstream effector of AURKA ( Figure 6d) and a predicted miR449 target (Supplementary Figure S5a).
In comparison to DMSO-treated cells, miR449 −/− cells treated with an HDAC6 inhibitor have more and longer cilia as shown by IF labeling for the axonemal cilia marker Ac-α-TUB (Figure 8a-c). Thus, inhibition of HDAC6 rescues at least some hallmarks of aberrant ciliation in miR449 −/− airway cultures.
In this study we show that miR449 loss triggers the deficient ciliated regeneration of respiratory epithelia upon environmental insults and aging. This is characterized by reduced airway cilia and faulty particle transport, and accompanied by increased hallmarks of the murine COPD phenotype, i.e., emphysematous and inflammatory changes. Upregulation of AURKA contributes, at least partially, to greater ciliary instability, probably through hyperactivation of its ciliary effector HDAC6, and both the number and the length of respiratory cilia from miR449 −/− cells were restored upon inhibition of HDAC6 activity. Our research uncovers a novel role for miR449 in cilia maintenance and healthy lung aging and emphasizes the significance of balanced ciliated regeneration for pulmonary health.  available selective HDAC6 inhibitor tubastatin [47] to treat ALI cells as HDAC6 is a downstream effector of AURKA ( Figure 6d) and a predicted miR449 target ( Supplementary Figure S5a).
In comparison to DMSO-treated cells, miR449 −/− cells treated with an HDAC6 inhibitor have more and longer cilia as shown by IF labeling for the axonemal cilia marker Acα-TUB (Figure 8a-c). Thus, inhibition of HDAC6 rescues at least some hallmarks of aberrant ciliation in miR449 −/− airway cultures. (c) Representative images of cilia length, (d) Cilia length is measured from (a,c). All data are presented as the mean ± SEM and relative to the miR449 −/− Ctrl group with ** p < 0.01.
In this study we show that miR449 loss triggers the deficient ciliated regeneration of respiratory epithelia upon environmental insults and aging. This is characterized by reduced airway cilia and faulty particle transport, and accompanied by increased hallmarks of the murine COPD phenotype, i.e., emphysematous and inflammatory changes. Upregulation of AURKA contributes, at least partially, to greater ciliary instability, probably through hyperactivation of its ciliary effector HDAC6, and both the number and the

Discussion
miR449 has been described in ciliogenesis in earlier studies [8][9][10][11][12][13]48], but this is the first report of its connection with cilia maintenance in response to environmental challenges, and its protective effects on airway biology and function in mouse models of chronic lung diseases and exacerbations. Here we show that the loss of miR449 impairs ciliated epithelial regeneration and mucociliary clearance by increasing AURKA-driven ciliary disassembly (Figure 9). Thus, miR449 −/− mice develop spontaneous COPD accompanied by an exacerbated pulmonary inflammatory response to challenge.
Previous studies on the miR34/449 family using genetic depletion of all miR34/449 family members describe a strong ciliary phenotype of PCD accompanied by defective development and early lethality [10][11][12], thus impairing the long-term analysis of the contribution of miR449 to disease etiology. In this setting, one allele of miR449 or of miR34b,c was sufficient to avoid complete ciliary disruption [10,12], highlighting the strength of one single miRNA allele. Indeed, here we show that the sole loss of miR449 increases susceptibility to chronic lung diseases by disrupting an important innate lung defense mechanism, i.e., mucociliary clearance and epithelial regeneration, thus increasing damage in response to environmental challenges. As miR34b,c are also expressed in the lungs, although to a lesser extent [8][9][10][11][12]49], it would be insightful to challenge miR34b,c knockout (KO) lungs in similar settings, which was beyond the scope of the present study. lung aging and emphasizes the significance of balanced ciliated regeneration for pulmonary health.

Discussion
miR449 has been described in ciliogenesis in earlier studies [8][9][10][11][12][13]48], but this is the first report of its connection with cilia maintenance in response to environmental challenges, and its protective effects on airway biology and function in mouse models of chronic lung diseases and exacerbations. Here we show that the loss of miR449 impairs ciliated epithelial regeneration and mucociliary clearance by increasing AURKA-driven ciliary disassembly (Figure 9). Thus, miR449 −/− mice develop spontaneous COPD accompanied by an exacerbated pulmonary inflammatory response to challenge. Previous studies on the miR34/449 family using genetic depletion of all miR34/449 family members describe a strong ciliary phenotype of PCD accompanied by defective Figure 9. miR449 protects airway regeneration (schematic model). Left panel (WT mice): When environmental toxicants (bacteria, cigarette smoke, diesel exhaust particles, naphthalene) enter the lungs, they are normally trapped in mucus and transported out of the airways by synchronized beating of numerous motile cilia, a phenomenon known as mucociliary clearance. Adequate clearance enables healthy airway regeneration, thereby preventing chronic airway inflammation. miR449 ensures cilia maintenance and healthy clearance by controlling AURKA/HDAC6-mediated ciliary disassembly. Right panel (miR449 −/− mice): by promoting AURKA-mediated ciliary disassembly, loss of miR449 affects ciliated epithelium renewal and mucociliary clearance, resulting in stronger inflammatory response to insult as toxicants may not be cleared efficiently, ultimately resulting in the development of emphysematous manifestations of COPD. Legend: AC = acetylated-alpha-tubulin, P = phosphorylated HDAC6.
It has been proposed that Notch signaling plays an important role in the miR449mediated regulation of ciliogenesis [9]. However, since the deregulation of Notch, which lies upstream of MCIDAS/GEMC1, would influence the ciliated cell's fate [17], it is not likely to be involved in our findings. As determination of the ciliated cell's fate was not affected by miR-34/449 depletion [10,12] and Notch was not upregulated in the miR-34/449 KO nasal respiratory epithelium [12], our recent data suggest the limited contribution of miR449-mediated Notch regulation in mouse cells, although it seems to play a role in the ciliated embryonic frog epidermis [9]. Instead, here we identified that miR449 deficiency decreases airway ciliation by increasing the ciliary disassembly rate, probably through the upregulation of AURKA and potential overactivation of its downstream effector HDAC6, thereby contributing to the development of manifestations of COPD.
Interestingly, significant upregulation of AURKA in response to miR-34/449 depletion has not only been observed in airways [10,12] but also in fallopian tubes [12], an organ covered by MCCs. Notably, AURKA and its direct target HDAC6 are upregulated in COPD lungs, where, at least, HDAC6 contributes to increased cilia destabilization or "ciliophagy" upon environmental challenge, e.g., cigarette smoking [44,50].
Interestingly, increased AURKA levels generated aberrant primary cilia with ultrastructure defects [45,46] similar to those observed in motile cilia in our study. This, together with HDAC6-mediated motile cilia destabilization in response to smoke [44,50], supports the notion that motile and primary cilia share conserved generation/resorption pathways, and both use the AURKA/HDAC6 axis. As primary cilia biology research has generated tremendous data on this topic [35,51], looking into candidates stabilizing primary cilia might be an interesting path to new targets in respiratory disease if expressed in the airways. One such candidate is the predicted miR449 target KCNH1, which encodes a potassium voltage-gated channel (Kv.10.1) and promotes primary cilia disassembly [36]. Notably, a single nucleotide polymorphism (SNP) mutation in KCNH1 was associated with COPD susceptibility [52], although the effect of this SNP on Kv10.1 channel function is currently unknown. One limitation of studying miRNA-based regulation is that we cannot exclude the contribution of further targets to the observed mechanisms, especially as we used a systemic KO mouse model. Nevertheless, miR449 expression was only found in ciliated epithelia [8][9][10]12], suggesting its main functions there, but not excluding communication with other cells, e.g., by exosomes [53]. In this study we showed that the transcript levels of the miR449 host gene Cdc20b and miR449a itself were strongly upregulated in DEP-treated WT mice, in the model of particle-induced airway disease. Interestingly, a recent study demonstrated that mir449a had an antifibrotic effect in silica-induced lung fibrosis via the activation of autophagic activity in vitro [54]. Utilizing single cell data from Adams et al. [55], we could show that CDC20B is expressed preferentially in ciliated cells (Figure 10a). The observed reduction in CDC20B expression in COPD-derived cells suggests the concurrent reduction in miR449 (Figure 10b,c). Interestingly, this dataset shows an increase in AURKA and HDAC6 expression in COPD-derived ciliated cells, mirroring our molecular data. Additionally, new reports have shown that miR449 target HDAC6 is involved in inflammasome assembly in macrophages, suggesting anti-inflammatory effects of HDAC6 reduction, and many groups incl. pharmaceutical companies have started drug discovery programs to inhibit HDAC6 functions. Importantly, it is very rare to see such a strong differential expression of a miRNA target in a miRNA KO mouse, as observed here for AURKA, due to the mainly fine-tuning functions of miRNAs and their putative compensatory mechanisms such as the proteasomal degradation of excess protein [56]. Therefore, we assume that increased AURKA levels and/or HDAC6 activity play a pivotal role in the development of the COPD-like disorders in the miR449 −/− mouse model. The data presented in this study show a supportive effect of miR449 on bronchial epithelial regeneration during the aging process and upon various challenges.
Respiratory motile cilia are of utmost importance for lung health, as underlined by the fact that any single non-synonymous disruptive mutation in any core motile cilia protein generates lung diseases ranging from mucociliary clearance disorder with increased susceptibility to recurrent respiratory infections to severe PCD [3]. However, the link between mucociliary clearance and COPD has not been clearly made. The progression of COPD is driven in most cases by recurrent pathogen-driven exacerbations, which usually target the bronchial epithelium at the first site. Immobilization of the respiratory cilia by the ciliostasis of otherwise functional cilia is sufficient to increase the susceptibility to influenza A infection [57], a well-known trigger of COPD exacerbations. Many pathogens have developed ciliostatic factors to abrogate mucociliary clearance and avoid being evacuated from the lungs, e.g., Pseudomonas [58,59], Haemophilus influenzae [60,61], and fungi [62,63], all typical colonizers of the COPD lung [64]. observed here for AURKA, due to the mainly fine-tuning functions of miRNAs and their putative compensatory mechanisms such as the proteasomal degradation of excess protein [56]. Therefore, we assume that increased AURKA levels and/or HDAC6 activity play a pivotal role in the development of the COPD-like disorders in the miR449 −/− mouse model. The data presented in this study show a supportive effect of miR449 on bronchial epithelial regeneration during the aging process and upon various challenges. Respiratory motile cilia are of utmost importance for lung health, as underlined by the fact that any single non-synonymous disruptive mutation in any core motile cilia protein generates lung diseases ranging from mucociliary clearance disorder with increased susceptibility to recurrent respiratory infections to severe PCD [3]. However, the link between mucociliary clearance and COPD has not been clearly made. The progression of COPD is driven in most cases by recurrent pathogen-driven exacerbations, which usually target the bronchial epithelium at the first site. Immobilization of the respiratory cilia by When pathogens manage to escape mucociliary clearance, they can colonize the lungs. In response to such infections, immune recruitment occurs, and inflammatory signaling is generated including the induction of destructive proteases such as MMP9/12 and neutrophil elastase. Those proteases attack the structural proteins (e.g., elastin) responsible for the recoil properties of the lungs, i.e., elastance and compliance. This suggests that the reduced clearance of pathogens and inflammatory responses associated with airway infections has an impact on the lung function in the long term. Indeed, it is known that exacerbations of COPD, which are often triggered by viral or bacterial infection, are associated with a more rapid decline in lung function. The more often exacerbations happen, the more destruction occurs, and full resolution of the inflammation becomes difficult [21,23,65]. Therefore, hindering pathogen entry by protecting the clearance process might halt COPD progression. Motile cilia are the major factor contributing to mucociliary clearance, and ciliary length [66] and ciliary beat frequency [67] are reduced in COPD. Thus, protecting healthy regeneration of the motile airway cilia, including length and beating ability [47], as well as microbial ciliostasis, might represent new therapeutic approaches to tackle COPD and reduce severe exacerbation rates on top of antiinflammatory drugs. Whether miR449 supplementation may be an option to protect the bronchial barrier is yet unclear, but tool inhibitors addressing aurora kinases and HDAC6 deacetylase activities are already available to study pathway contribution to chronic lung diseases in preclinical models.

Correlation of miR449a and mRNAs in COPD Bronchial Biopsies
Bronchial biopsy specimens were collected from 57 COPD patients who were part of the Groningen Leiden Universities and Corticosteroids in Obstructive Lung Disease (GLUCOLD) study [24]. All patients were stable, either current or ex-smokers, and not on corticosteroid therapy. They had irreversible airflow limitation (post-bronchodilator forced expiratory volume in 1 s (FEV1) and FEV1/inspiratory vital capacity < 90% confidence interval of the predicted value) and chronic respiratory symptoms. The local medical ethics committee approved the study, and all patients gave their written informed consent. The study is filled at clinicaltrial.gov under identification number NCT00158847. mRNA and miRNA profiling were performed at Boston University Microarray Resource Facility using GeneChip ®® Whole Transcript Sense Target Labeling Assay and FlashTagTM Biotin HSR Labeling Kit, respectively. Detailed methods on the profiling and processing of microarray data were described previously [68]. Using the matched mRNA and miRNA profiles of the same patient, Pearson correlation coefficient was used to identify genome-wide mRNAs of which expression was correlated with miR449a expression. Benjamini-Hochberg procedure was applied for multiple test correction. Significant correlation was defined using false discovery rate adjusted p-value (FDR) < 0.05.
To assess the role of miR449a in cilia development and function, the list of mRNAs positively correlated with miR449a was compared to the previously reported list of ciliaassociated genes [26]. Moreover, gene set enrichment analysis (GSEA, v3.0) [25] was performed to identify biological processes (Gene Ontology, v6.1) in which miR449 plays a role. Genome-wide mRNAs were ranked based on their t-statistics representing the strength of their correlation with miR449a expression. A p-value was calculated after 1000 permutations were performed.

Mice
miR449 −/− mutants had a targeted deletion of miR449a, b, and c and were previously described [10]. WT and miR449 −/− mice were maintained in C57Bl/6 background (n8) at the animal facility of the European Neuroscience Institute Goettingen (Germany) in full compliance with institutional guidelines. The Animal Care Committee of the University Medical Centre Goettingen and the authorities of Lower Saxony approved the study under the number G12/963. WT (C57BL/6) mice used for DEP experiments were obtained from Harlan Laboratories (Horst, The Netherlands) and maintained in the animal facility of the Ghent University. The Animal Ethical Committee of the Faculty of Medicine and Health Sciences of Ghent University approved the in vivo manipulations using DEP. Experiments including the exposure to cigarette smoke (CS) and nontypeable Haemophilus influenzae (NTHi) [69,70] were approved by the "Landesamt für Soziales, Gesundheit und Verbraucherschutz" of the State of Saarland following the national guidelines for animal treatment. During the experiment WT and miR449 −/− mice were maintained under specific pathogen-free conditions in the animal facility of the Institute for Experimental Surgery of the Saarland University Homburg.

Transfection of Human Cells
H1299 cells were reverse transfected with 12 nM (RNA analysis) or 20 nM (protein analysis) synthetic pre-miRNAs (Ambion, Thermo Fisher Scientific) of the miR34/449 family using Lipofectamine 2000 (Invitrogen, Thermo Fisher Scientific) for transcript and protein analysis, respectively. Scrambled RNA oligonucleotide (negative control #2 (Ctrl)) and naturally occurring pre-miR-302* were used as controls. Cells were harvested for protein and RNA isolation 48 h (h) after transfection.

Analysis of Primary Cilia Assembly and Disassembly
Freshly isolated WT and miR449 −/− MEFs were plated on fibronectin (Merck, Darmstadt, Germany)-coated coverslips and exposed to serum-free medium (DMEM supplemented only with 100 µg/mL streptomycin and 100 U/mL penicillin) for 60 h to induce primary cilia assembly. Afterwards, serum was reintroduced for 6 h to trigger cilia disassembly. To quantify the percentage of ciliated cells at the assembly and disassembly time point, MEFs were fixed with 2% formaldehyde (Merck) at RT for 8 min and blocked with 10% bovine serum albumin (BSA) (Merck) in phosphate-buffered saline containing 0.1% Triton X-100 (PBS-T) for 1 h 30 min at RT. MEFs were first incubated with the primary antibodies diluted in blocking solution overnight at 4 • C (list of antibodies is provided in Supplementary Table S4) and, subsequently, with secondary antibodies for 1 h 30 min at RT (list of antibodies is provided in Supplementary Table S5). Nuclei were counterstained with DAPI. Stainings were visualized using a Leica SP2 laser scanning confocal microscope (Leica Microsystems, Wetzler, Germany). Primary cilia were identified by a double staining of pericentrin and Ac-α-TUB. Only cells positive for both were counted as ciliated cells.

Air-Liquid Interface Cultures
MTECs were isolated and cultivated using ALI conditions as described previously [26]. Note that, minor changes regarding the culture and differentiation medium were applied as highlighted below. Briefly, tracheae from WT and miR449 −/− mice (10-16 weeks old) were excised and after pronase digestion (Merck), epithelial cells were seeded onto collagencoated transwell culture inserts (12 mm diameter, 0.4 µm pores, polyester, Corning, NY, USA). Cells were cultured under submerged conditions in airway epithelial cell growth medium supplemented with growth factors (Promocell, Heidelberg, Germany) 100 U/mL penicillin, and 100 µg/mL streptomycin to allow proliferation. Upon confluence, ALI condition (d0) was generated by removing the apical chamber medium and changing the basolateral medium to the differentiation medium consisting of DMEM/F-12 (Gibco) supplemented with 2% Ultroser G (Pall, Port Washington, NY, USA). For the HDAC6 inhibitor experiment, HDAC6 inhibitor (100 nM, Tubastatin A hydrochloride, Merck, Darmstadt, Germany) or DMSO (control) was added to the basolateral medium of miR449 −/− ALI cultures twice a day from differentiation day (d) 0 to d6. At d6, the entire ALI culture membrane was fixed with 4% paraformaldehyde (PFA) and embedded in paraffin. For one ALI culture, MTECs pooled from three mice were used. Human ALIs were performed in a former study [8], and only RNA was recovered to complete the presented work.

Naphthalene-Induced Injury Model
Naphthalene (Merck) was dissolved in sunflower oil to a concentration of 30 mg/mL in a sterile environment on the day of use. Next, 12-14-week-old male WT and miR449 −/− mice were injected with naphthalene (200 mg/kg) or with oil (vehicle control) intraperitoneally (i.p.) in the morning on day 0.

NTHi-Induced Chronic Inflammation Model
A clinical isolate of NTHi was grown on selective chocolate agar with IsoVitaleX at 300 µL per 10 cm plate (Becton Dickinson, Heidelberg, Germany) for 24 h at 37 • C in 5% CO 2 . After harvesting, bacteria were incubated for 24 h in 700 mL brain-heart infusion broth (Becton Dickinson) containing 3.5 mg/mL β-Nicotinamide adenine dinucleotide (NAD, Merck) and 5% Fildes enrichment (Becton Dickinson). The culture was centrifuged at 2500× g for 15 min (min) at 4 • C, washed, resuspended in 20 mL PBS, heat inactivated at 70 • C on a mechanical shaker for 45 min, and then sonicated three times for 30 s. The protein concentration was adjusted to 2.5 mg/mL in PBS using the Pierce BCA protein assay (Thermo Fisher Scientific). For bacterial challenge, 4-month-old WT and miR449 −/− mice were placed in a plexiglass box connected to a Pari MASTER ®® nebulizer (Pari GmbH, Starnberg, Germany) and exposed to the inactivated bacterial lysate for 40 min 5 times per week for 3 months.

Cigarette Smoke (CS)-Induced Emphysema Model
Four-month-old WT and miR449 −/− mice were exposed to CS (3R4F cigarettes, College of Agriculture, Reference Cigarette Program, University of Kentucky) in a TE-10 smoking machine (Teague Enterprises, Woodland, CA, USA). The concentration of CS was 120 mg/m [3] total suspended particles. Mice were subjected to CS for 5 h per day and 5 days per week over a period of 6 months. The daily smoking protocol consisted of three smoking phases each with 87 min, which were interrupted by room air exposures for 40 min.

Pulmonary Function
Pulmonary function measurement was performed using a FlexiVent system (Scireq Inc., Montreal, QC, Canada). The mice were anesthetized by a mixture of ketamine (Ursotamin, 100 mg/mL, Serumwerk Bernburg, Bernburg, Germany) and xylazine (Rompun, 7 mg/kg, Bayer). The trachea of the anesthetized mouse was exposed and cannulated using the 18 G cannula included with the FlexiVent system. For data acquisition we used the flexiWare 7.1 software (Scireq Inc.).

Bronchoalveolar Lavage
Prior to bronchoalveolar lavage (BAL) the animals were deeply anesthetized with a mixture of ketamine (Ursotamin, 105 mg/mL, Serumwerk Bernburg) and xylazine (Rompun, 7 mg/kg, Bayer). After reaching anesthesia the trachea was exposed and cannulated. The BAL was performed by rinsing the lungs 3 times repeatedly with 1 mL sterile PBS containing protease inhibitors (cOmplete ULTRA Tablets, Mini, Roche Diagnostic). The collected BALF was centrifuged for 5 min at 300× g. The number of viable cells was determined by resuspending the cell pellet in 1 ml of sterile PBS and counting on an improved Neubauer chamber after trypan blue staining. For the cellular composition of the BALF a cytospin was stained with DiffQuik (Medion Diagnostic, Gräfelfing, Germany), and at least 200 cells counted and differentiated based on their morphology.

Stereology
The lung was prepared for uniform random sampling for stereological analysis as described elsewhere [73]. Briefly, the lung was fixed by instillation of freshly prepared PBSbuffered 4% formalin under a constant hydrostatic pressure of 20 cm for 20 min. After that, the trachea was ligated to preserve intrapulmonary pressure. For additional fixation, the lung was placed in PBS-buffered 4% formalin for 24 h. The lung volume used as reference for stereological measurements was determined by fluid displacement. The fixed lung was embedded in agar and cut into regular slices prior to paraffin embedding. Stereological analysis was performed on hematoxylin and eosin (H&E)-stained lung sections. Mean chord length (mean linear intercept, L m ) was calculated using the Visiopharm Integrator System package (Visiopharm, Hoersholm, Denmark) on an Olympus BX51 microscope equipped with an 8-position slide loader (Olympus, Tokyo, Japan).

Histology and Immunostaining
PFA-fixed, paraffin-embedded tissue sections were dehydrated and treated with heat-induced epitope retrieval (sodium citrate, pH 6.0), while cryosections were fixed with 4% PFA for 10 min at room temperature (RT). Following a blocking step with 10% FCS diluted in PBS-T (for cryosections: 1% BSA in PBS-T), primary antibodies diluted in blocking solution were incubated on the samples overnight at 4 • C (list of antibodies is provided in Supplementary Table S4). After washing with PBS-T, sections were stained with fluorescently-labelled secondary antibodies for 1 h at RT (list of antibodies is provided in Supplementary Table S5). Nuclei were counterstained with 4 , 6-Diamidin-2-phenylindol (DAPI) and samples were mounted with Dako fluorescence mounting medium. For histology, PFA-fixed, paraffin-embedded lung sections were dehydrated and stained with H&E. Images were acquired on a ZEISS AxioScope A1 microscope (ZEISS, Oberkochen, Germany), except tracheae images, which were viewed on a ZEISS confocal LSM 510 or 880 microscope.

Quantification of Cilia Markers
The expression of cilia markers in immunofluorescence images (converted to 8-bit) was quantified using the ImageJ software [74]. Briefly, after selecting the region of interest, unspecific signals were removed, and a threshold was set to identify the positive signal. To measure the area of the positive cilia staining the Analyze Particle tool was applied. Values of the cilia area were normalized to the length of the respiratory epithelium measured. For length measurements of motile cilia in ALI cultures, the line tool from the ImageJ software was used. Cilia length was defined by the area positive for both cilia markers used.

RNA Extraction and Quantitative PCR
RNA from snap-frozen lung tissue and cells (H1299) was isolated by Extrazol (7Bio-Science, Hartheim, Germany)/Chloroform as recommended by the manufacture. The miRVana kit (Invitrogen, Thermo Fisher Scientific) was used for RNA isolation from human ALI cultures. cDNA from 1 µg RNA was synthesized using the M-MuLV reverse transcriptase (New England Biolabs, Ipswich, MA, USA) and a mixture of random nonameric and Oligo-dT primers (Metabion, Planegg, Germany). For the qPCR reaction, a self-made SYBR Green qPCR Mix (Tris-HCl (75 mM), (NH 4 ) 2 SO 4 (20 mM), Tween-20 (0.01% v/v), MgCl 2 (3 mM), Triton X-100 (0.25% v/v), SYBR Green I (1:40,000), dNTPs (0.2 mM) and Taq-polymerase (20 U/mL)), and 250 nM of each gene-specific primers were used (list of primers is provided in Supplementary Table S6). The relative expression of transcripts from each sample was obtained by averaging the cDNA levels of technical triplicates and normalizing it to the housekeeping gene ribosomal phosphoprotein P0 (Rplp0, or 36b4). For miRNA quantification, TaqMan MicroRNA Assay (Applied Biosystems, Thermo Fisher Scientific) for murine and human members of the miR34/449 family was performed according to the manufacturer's instructions with U6 snRNA as reference gene.

Western Blot
Samples were lysed in RIPA buffer (20 mM Tris-HCl pH 7.5, 150 mM NaCl, 9.5 mM EDTA, 1% Triton X100, 0.1% SDS, 1% sodium deoxycholate) supplemented with urea (2.7 M) and protease inhibitors (Complete Mini EDTA-free, Roche Diagnostic). Note that, for lung samples, lungs were snap-frozen in liquid nitrogen and mechanically homogenized. Equal amounts of protein lysates were resolved by SDS-polyacrylamide gels followed by a transfer to nitrocellulose membranes. Membranes were incubated with primary antibodies overnight at 4 • C (list of antibodies is provided in Supplementary Table S4), washed, and stained with horse radish peroxidase-conjugated secondary antibodies for 1 h at RT prior to chemiluminescence detection (list of antibodies is provided in Supplementary Table S5). The volume intensity of specific protein bands was quantified using the Image Lab software (Biorad, Hercules, CA, USA).

Electron Microscopy
Trachea samples from 7-9-week-old WT and miR449 −/− were fixed by immersion using 2% glutaraldehyde in 0.1 M cacodylate buffer (Science Services, München, Germany) at pH 7.4 overnight at 4 • C. Post-fixation was performed using 1% osmium tetroxide diluted in 0.1 M cacodylate buffer. After pre-embedding staining with 1% uranyl acetate, tissue samples were dehydrated and embedded in Agar 100 (Plano, Wetzlar, Germany). To rule fixation artefacts out, a second embedding technique was performed. For this, samples were placed in aluminum platelets of 150 µm depth containing 1-hexadecen [75]. The platelets were frozen using a Leica EM HPM100 high pressure freezer (Leica Microsystems, Wetzlar, Germany). The vitrified samples were transferred to an Automatic Freeze Substitution Unit (Leica EM AFS2) and substituted at −90 • C in a solution containing anhydrous acetone, and 0.1% tannic acid for 24 h and in anhydrous acetone, 2% OsO 4 , and 0.5% anhydrous glutaraldehyde (EMS Electron Microscopical Science, Hatfield, PA, USA) for additional 8 h. After a further incubation over 20 h at −20 • C, samples were warmed up to +4 • C, and washed with anhydrous acetone. The samples were embedded at RT in Agar 100 at 60 • C for 24 h. Thin trachea sections (100 nm) were examined using a Philips CM 120 BioTwin transmission electron microscope (Philips Inc., Amsterdam, The Netherlands) and images were taken with a TemCam F416 CMOS camera (TVIPS, Gauting, Germany). Longitudinaland cross-sections from motile cilia were evaluated and grouped in normal and aberrant cilia. Aberrant cilia were defined as cilia with a detachment of the outer membrane. Images were taken and quantified from at least two different tissue locations from each sample.

Mucociliary Transport Assay
Mucociliary transport assay was performed as described in [26]. Briefly, freshly isolated vital tracheae were longitudinally cut using a microtome (VT1200S, Leica Microsystems) and rinsed with Ringer's solution containing 98 mM NaCl, 2 mM KCl, 1 mM CaCl 2 , 2 mM MgCl 2 , 5 mM glucose, 5 mM sodium pyruvate, 10 mM Hepes (230 mOsm, pH 7.8). Fluorescent microspheres (FluoSpheres ®® , 0.17 µm, PS-Speck, Invitrogen, Thermo Fisher Scientific) diluted in Ringer's solution were added onto the multiciliated epithelium of the tracheae. In areas of 280 µm × 280 µm (512 × 512 pixels) the movement of the fluorescent microspheres was recorded using a custom-built confocal line illumination microscope at 61 Hz for 2000 frames. Note that only areas with beating cilia were imaged. For each frame, particles were detected and tracked using functions previously described [26]. Absolute full-trajectory velocities along the focal plane were combined within one measurement. Averaged velocities of independent measurements from different tissue areas and different mice were assigned into the following groups: WT, miR449 −/− , and diffusion. Diffusion measurements were obtained from non-living dried tracheae and pooled from WT and miR449 −/− mice since they were identical. Individual trajectory velocities from all measurements within the respective group were included to obtain particle velocity distributions. All data evaluation was carried out using a custom software written in Matlab (MathWorks, Natick, MA, USA). Age-matched WT and miR449 −/− mice (12 weeks) were used.

Statistical Analyses
Two tailed, unpaired or paired t-test with Welch-correction (used only for unequal variances) was used to calculate statistical significance for pairwise comparisons. Except for Figures 5d and 6c, one tailed unpaired t-test and two tailed paired t-test were applied, respectively. Normal distribution was assumed for all analyzed data. Statistical analyses were carried out using the GraphPad Prism software (version 6.01, San Diego, CA, USA). The following indications of significance were used throughout the manuscript: * p < 0.05, ** p < 0.01, *** p < 0.001. Results are shown as the mean ± standard error of the mean (SEM).