Regulation of cilia abundance in multiciliated cells

Multiciliated cells (MCC) contain hundreds of motile cilia used to propel fluid over their surface. To template these cilia, each MCC produces between 100-600 centrioles by a process termed centriole amplification. Yet, how MCC regulate the precise number of centrioles and cilia remains unknown. Airway progenitor cells contain two parental centrioles (PC) and form structures called deuterosomes that nucleate centrioles during amplification. Using an ex vivo airway culture model, we show that ablation of PC does not perturb deuterosome formation and centriole amplification. In contrast, loss of PC caused an increase in deuterosome and centriole abundance, highlighting the presence of a compensatory mechanism. Quantification of centriole abundance in vitro and in vivo identified a linear relationship between surface area and centriole number. By manipulating cell size, we discovered that centriole number scales with surface area. Our results demonstrate that a cell-intrinsic surface area-dependent mechanism controls centriole and cilia abundance in multiciliated cells.


Introduction:
In mammals, multiciliated cells (MCC) are terminally differentiated epithelia that line the respiratory tract, brain ventricles and segments of the female and male reproductive organs 1, 2 .
Each MCC contains hundreds of motile cilia, microtubule-based organelles that generate the motive force to move fluid over the surface of the cell. In the airway, motile cilia of MCC beat directionally to propel mucus and inhaled contaminants out of the lungs. Ciliary motility of ependymal MCC lining the brain ventricles is key for movement of cerebrospinal fluid through the central nervous system, while cilia on MCC in the reproductive tracts are important for ovum and sperm transport 1, 2 . Thus, the proper assembly of hundreds of motile cilia is critical for the functions of MCC, and the overall health of the associated tissues. Mutations that result in reduced cilia abundance cause human diseases due to aberrant cilia-based fluid flow [3][4][5][6] , suggesting that establishment of the correct number of cilia per cell is important for function. MCC in different tissues display significant variability in motile cilia number, ranging from 30-600 cilia per cell 1, 2 . However, a major unresolved question is how each cell regulates the precise number of its motile cilia during differentiation.
To template these cilia, each MCC undergoes a process termed centriole amplification to produce hundreds of centrioles, barrel-shaped microtubule structures that form the base upon which cilia are assembled. In the airway, proliferating progenitors (called basal cells) initially contain a single centrosome with two parental centrioles (PC), which they maintain via the canonical centriole duplication mechanism (Fig. 1a, [7][8][9] ). During MCC differentiation, the post-mitotic basal cells undergo massive centriole production that was thought to occur through two parallel pathways: a parental centriole-dependent pathway whereby the original PC template the assembly of multiple procentrioles simultaneously, and an acentriolar or "de novo" pathway mediated by dozens of deuterosomes, spheroidal electron-dense structures that are capable of producing multiple procentrioles ( Fig. 1a, 8,10,11 ). Deuterosomes are transient structures that are absent in proliferating progenitor cells, are formed during early centriole amplification, and mostly disappear after centriologenesis is complete 12,13 . Although both PC-dependent and deuterosome-dependent pathways have been known for decades, the molecular mechanisms that govern the two pathways and their relative contributions to the total complement of centrioles has remained enigmatic.
Most of the key proteins involved in canonical centriole duplication play an essential role in both PC-and deuterosome-dependent centriole amplification pathways in MCC 9,[12][13][14][15] . This 4 indicates that centriole duplication in cycling cells and centriole amplification in MCC share common and conserved molecular mechanisms despite their morphological differences. However, recent studies have characterized proteins that are uniquely associated with deuterosomes. For example, Deup1 and Ccdc78 were recently identified as deuterosome-specific proteins that are essential for the formation of deuterosomes and, subsequently, centriole amplification 12,13 . These data support the theory that the PC-dependent and deuterosome-dependent centriole amplification pathways act in parallel, but use unique molecular components to produce all of the centrioles in a cell. It has been estimated that the deuterosome-dependent pathway contributes the majority (80-90%) of the total centrioles assembled during amplification, while the PC-dependent pathway contributes roughly 10-20% towards the final number 2,8,13,16 . Intriguingly, a recent study found that deuterosomes themselves are nucleated from the original PC, indicating that these two pathways may not be independent of each other after all 16 . If true, this would suggest that all centrioles produced in MCC are ultimately dependent on the PC. Yet, a number of important questions have remained unanswered: are parental centrioles necessary for deuterosome biogenesis? What is the actual contribution of the PC to final centriole number? How do cells that start with exactly two parental centrioles generate such a variable number during differentiation?
In this study, we sought to determine the mechanisms by which airway MCC establish centriole number, using an ex vivo airway culture system [17][18][19] . We began by testing the hypothesis that parental centrioles are essential for deuterosome formation. To ablate PC during proliferation, we used a small molecule inhibitor and shRNA-mediated depletion of Plk4, the major kinase involved in centriole duplication 7,20 . Surprisingly, loss of parental centrioles did not affect deuterosome formation, indicating that PC are dispensable for deuterosome biogenesis. Moreover, the loss of PC and Plk4 function did not disrupt centriole amplification nor decrease the total number of centrioles per cell. Thus, establishment of centriole abundance is regulated by a different cellular property. Airway MCC vary in size and surface area, and display a wide range in centriole number.
Quantification of centriole abundance in vitro and in vivo highlighted a direct relationship between cell-surface area and centriole number. To determine if cell size was a major determinant of centriole abundance, we cultured basal cells on an extracellular matrix of increasing density to manipulate size. We found that centriole number scales with increasing surface area, but not volume, of multiciliated cells. Together, our results point to a cell-intrinsic, surface area-dependent mechanism that establishes centriole-cilia abundance in airway multiciliated cells.

Results:
Parental centrioles are dispensable for centriole amplification We previously established a mouse tracheal epithelial cell (MTEC) culture system to study centriologenesis and ciliary assembly in MCC 14,18,19,21 . This airway model entails seeding freshly isolated mouse tracheal basal cells onto a porous filter suspended in medium, and allowing the progenitors to proliferate into a confluent, polarized layer. Cells are induced to differentiate into MCC and other cell types found in the airway (for example, secretory cells) by exposure to an air-liquid interface (ALI). During proliferation, basal cells contain two centrosomal centrioles that are maintained via the canonical centriole duplication mechanism (Fig. 1a). During post-mitotic MCC differentiation, the centriologenesis program proceeds in distinct stages from ALI days 0 to 12: centriolar protein aggregates and deuterosomes form near the PC, and procentrioles are nucleated from both PC and deuterosomes (stage I, ALI1-2); procentrioles continue to grow and elongate from PC and deuterosomes (stage II, ALI3-4); centrioles disengage from PC and deuterosomes, start to mature into basal bodies and migrate towards the apical membrane (stage III, ALI4-8); fully mature, apically docked basal bodies template ciliary assembly (stage IV, ALI8-12). By ALI12, most cells have completed differentiation into fully mature MCC (Fig. 1a).
It was recently proposed that deuterosomes are nucleated from the original PC, suggesting that the de novo centriole assembly pathway is dependent on PC 16 . To test this theory, we ablated PC in basal cells using Centrinone 22 , a selective inhibitor of the kinase Plk4 known as the "master regulator" of centriole assembly 20,[23][24][25][26] . Centrinone blocks centriole duplication, resulting in cells that lack centrioles following a few rounds of cell division 22 . Basal cells were cultured with or without 1 µM Centrinone, roughly 3-8 times higher concentrations than needed to suppress centriole formation in most cells 22 , and the compound was maintained in the medium throughout differentiation (described in Methods). Samples were fixed at various stages of differentiation, immunostained for centrioles and analyzed using either confocal or superresolution structured illumination microscopy (3D-SIM). Control cells at ALI0 predominantly contain the original two PC ( Fig.1b-c). Upon exposure to ALI, cells proceeded through the various stages of centriole amplification, resulting in a population of cells comprised of roughly 60% MCC and 40% non-MCC as expected (Fig. 1c). Growth of basal cells in the presence of 1 µM Centrinone resulted in >80% of cells lacking PC at ALI0 (Fig. 1b-c). Remarkably, the formation of centriolar aggregates and subsequent centriole amplification proceeded normally in the absence of PC. Moreover, there was no significant difference in the timing of centriologenesis, or the fraction of MCC in the population 6 ( Fig. 1b-c). To ensure that Centrinone-treated cells were indeed lacking PC prior to differentiation, we analyzed cells using transmission electron microscopy (TEM). Cells at ALI0 were serially sectioned beginning at the apical surface and down to the basal membrane. Whereas PC were evident in control cells, there was >80% decrease in cells containing PC upon Centrinone treatment ( Fig. 1d and Supplementary Video 1), consistent with the immunofluorescence analysis.
The abundance of Plk4 protein is regulated by autophosphorylation-induced degradation 27 .
Centrinone-mediated inhibition of Plk4 kinase activity blocks this degradation, resulting in elevated (yet inactive) Plk4 protein levels 22 . In control cells, Plk4 expression increased during early stages of centriole amplification, peaking at ALI3 and subsequently decreasing upon centriole maturation at ALI8 ( Fig. 2a-b). Centrinone-treated cells showed accumulation of Plk4, which was already evident in a higher proportion of cells at ALI0-1 (Fig. 2a). Unlike control cells, the high Plk4-expressing fraction persisted at ALI8 and beyond ( Fig. 2a-b), indicating that Plk4 kinase activity was indeed blocked. Moreover, the non-multiciliated cells in the population did not reform/regain their PC ( Collectively, these results indicate that parental centrioles and Plk4 kinase activity are not necessary for centriole amplification and ciliogenesis in airway MCC.

Parental centrioles are dispensable for deuterosome formation and centriole abundance
Next, we tested whether loss of PC altered the expression pattern or number of deuterosomes in MCC. Cells at ALI3 were immunostained for Deup1, a core component of deuterosomes essential for their formation 13 . In control cells, deuterosome formation begins ~ALI1 and peaks at ALI3, and they are subsequently lost by ALI8 ( Fig. 3a-b). Surprisingly, Deup1 expression was grossly unchanged in cells lacking PC ( Fig. 3a-b). However, Deup1 foci were evident in a higher proportion of cells at ALI1, suggesting that deuterosome formation may start sooner in PC-less cells. Since MCC are variable in size and centriole number (discussed in more detail in the subsequent section), we measured deuterosome number and cell surface area (using ZO-1 to outline apical cell-cell junctions), then compared deuterosome abundance in cells of similar 7 size (Fig. 3c). Centrinone-treated cells lacking PC displayed a slight but significant increase in average deuterosome number per cell (Fig. 3d). Serial-section TEM analysis of cells at ALI3 confirmed the presence of deuterosome structures that were nucleating procentrioles in PC-less cells (Fig. 3e). These deuterosomes appeared consistent in size and morphology with control cells.
The deuterosome-dependent pathway is thought to contribute the majority (80-90%) of centrioles assembled during amplification, whereas the PC-dependent pathway contributes roughly 10-20% towards the final complement 2,8,13,16 . Therefore, we reasoned that loss of PC might result in a reduction in the final complement of centrioles in mature MCC. Quantification of centriole number in control cells at ALI12 showed a range from 100-500, and correlated with increasing surface area ( Fig. 4a-b). Loss of PC in Centrinone-treated cells did not disrupt this relationship (Fig.   4c). Importantly, the average number of centrioles per mature MCC did not decrease; in contrast we noted a slight but significant increase in centriole abundance and density ( Fig. 4d-e). All together, these results indicate that loss of PC does not inhibit deuterosome biogenesis nor result in decreased centriole abundance in MCC.

Modulating Plk4 expression delays centriole assembly but does not affect number
We were surprised to discover that inhibiting Plk4 activity had no detrimental effect on centriole amplification, as this kinase is known to play a critical role in centriole assembly in the majority of cell types examined in mammals. Plk4 expression becomes elevated during centriole amplification in airway MCC 38 , where it localizes to both PC and deuterosomes 13 , thus it has been generally assumed to play an important role in centriole amplification. However, our Centrinone treatments suggest that although Plk4 kinase activity is necessary for centriole duplication during the proliferation phase of basal cells, it is not essential for post-mitotic centriole amplification ( Fig. 1-3). One possibility is that Plk4 acts in a kinase-independent manner to control centriole amplification, for example as a scaffold to recruit other components of the centriole assembly pathway. To test this, we depleted the protein in basal cells using shRNA-expressing lentivirus. Depletion of Plk4 during the proliferation phase resulted in near complete loss of PC by ALI0 ( Fig. 5a- However, the process of centriole amplification appeared delayed, as roughly 50% of Plk4-depleted 8 cells at ALI12 were in Stages I-II or III-IV instead of being fully mature MCC ( Fig. 5a and f). We interpret these results to suggest that loss of Plk4 delays the process, but does not impact overall centriole assembly. Indeed, culturing these Plk4-depleted cells for an additional 9 days caused an increase in the fraction of mature MCC ( Fig. 5a and g). Importantly, once cells completed maturation we did not observe a significant decrease in centriole number per cell (Fig. 5h). These results indicate that loss of Plk4 does not impact centriole abundance, consistent with observations using Centrinone.
Loss of PC and Plk4 function did not affect centriole abundance in mature MCC, suggesting that establishment of centriole number is not dependent on them. To further test this hypothesis, we induced the formation of excess PC in basal cells by overexpressing Plk4, which drives nucleation of ectopic centrioles in a diversity of cells and organisms 20,26 . To overexpress Plk4 in a temporally controlled manner, we used a recently developed mouse model (Tg::mChPlk4) whereby mCherrytagged Plk4 is regulated by tamoxifen-induced Cre-recombinase expression 39,40 . To conditionally express Plk4 in MTEC, we bred Tg::mChPlk4 mice to a Tg::Rosa26-Cre ERT2 41 strain that globally expresses Cre upon tamoxifen addition. Basal cells were harvested from trachea of

Centriole number scales with surface area
Since establishment of centriole abundance appears to be independent of the PCdeuterosome mediated pathway, we wondered if other cellular properties influenced centriole-cilia abundance. We discovered that airway multiciliated cells grown in vitro display a broad range in centriole number, from 100-500 centrioles per cell (Fig. 4). We wanted to determine whether this broad range in centriole number, and its relationship to surface area, was retained in vivo. Trachea isolated from wild-type mice was cut longitudinally to expose the lumen, immunostained with centrin and ZO-1, then imaged by 3D-SIM (Fig. 6a). Quantification of centriole number showed a similar variation in centriole abundance (Fig. 6b). Consistent with the measurements in vitro, there was a 9 direct relationship between centriole number and surface area, as larger MCC contained more centrioles compared with smaller cells (Fig. 6b-d).
Next, we wondered if cell size (surface area or volume) was a major determinant of centriole abundance. In MTEC cultures, cell size and surface area are established roughly 1-2 days before the ciliogenesis transcriptional program is initiated. This temporal separation provided us with a means to determine whether the centriole amplification machinery responds to the differences in cell size set at the end of the proliferation phase. To test this, we manipulated cell shape and size prior to inducing differentiation by growing basal cells on varying concentrations of extracellular collagen.
Modulating extracellular matrix density is a commonly used approach to manipulate cell volume and/or surface area [42][43][44][45] . Culture and differentiation of basal cells on increasing concentrations of extracellular collagen resulted in a dose-dependent increase in cell surface area ( Fig. 7a-b).
Remarkably, centriole abundance also increased in a dose-dependent manner (Fig. 7c). To determine whether the increase in average centriole number per cell was a function of increased cell volume, we measured cell depth (apical-basal distance). Although the cells contained a larger surface area, they were shallower in depth, resulting in a similar overall cell volume ( Fig. 7a and de). Thus, the size of surface area, but not overall cell volume, appears to dictate centriole abundance during amplification. Finally, we sought to determine whether the higher centriole number in cells with larger surface areas was regulated at the level of the deuterosome. Indeed, cells grown on higher concentrations of collagen formed a higher proportion of deuterosomes per cell on average ( Fig. 7f-g). Collectively, these results indicate that a cell-intrinsic surface areadependent process establishes centriole abundance, potentially by modulating deuterosome number. 10

Discussion:
In this study, we sought to determine how MCC establish the number of centrioles and cilia that each cell contains. We first tested the relative contributions of the PC and Plk4 to deuterosome formation, centriole amplification, and establishment of centriole number. Loss of PC had no impact on deuterosome formation and function indicating that, although deuterosomes can be nucleated from the original PC 16 , their presence is not essential for the de novo centriologenesis pathway. In contrast, the average number of procentriole-forming deuterosomes was slightly but significantly higher upon loss of the PC (Fig. 3 and 5). One intriguing possibility for the increase may be that the deuterosome pathway compensates for the loss of PC-derived centrioles by increasing deuterosome-mediated centriole assembly. This is consistent with our observation that cells lacking the PC formed a similar number of centrioles on average when fully mature ( Fig. 4-5). Similarly, inhibition of deuterosome formation was shown to result in increased procentriole nucleation by the PC 13 , suggesting that these two centriole assembly pathways likely communicate to regulate centriole biogenesis. However, loss of the PC did impact the kinetics of deuterosome formation; they were evident in PC-less cells slightly earlier than control cells, and persisted at stages (e.g. ALI8) when they would normally have disappeared ( Fig. 3 and 5). We interpret these data to suggest that, although the PC are not critical for deuterosome biogenesis, they may help to control the timing of their assembly and disassembly. Nonetheless, the overall process of centriole amplification (initiation, growth and maturation) can proceed normally is the absence of PC.
Although Plk4 kinase function was needed for canonical centriole duplication in proliferating basal cells, it was dispensable for post-mitotic centriole amplification. This was somewhat surprising for two reasons: (a) expression levels of Plk4 increase significantly during centriole amplification in MCC 13, 38 , suggestive of a role during those stages, and (b) the kinase is essential for initiating procentriole formation in the majority of mammalian cells examined 7,26 . Nonetheless, depletion of Plk4 did cause a delay in centriologenesis, indicating that the protein itself might be needed for proper progression through the various stages of centriole assembly. This is reminiscent of what was recently described for other kinases involved in coordinating centriole assembly and cell cycle progression in MCC. For example, it was shown that differentiating, non-dividing MCC repurpose the mitotic regulatory circuitry involving CDK1/Plk1/APC-C to control the timely progression of centriole amplification, maturation, and motile ciliogenesis while avoiding reentry into mitosis 46 .
Another study found that CDK2, the kinase responsible for G1-S phase transition, was also required in MCC to initiate the motile ciliogenesis program independent of cell cycle progression 47 . Thus, 11 one possible reason for the elevated Plk4 protein is to coordinate the timing of centriole assembly and maturation in post-mitotic cells. Indeed, we found that MCC lacking Plk4 initiated centriole assembly to the same extent as control cells, were delayed in passage through the growth and maturation phases, but eventually "caught up" (Fig. 5). Importantly, multiciliated cells lacking Plk4 contained the same number of centrioles on average when fully mature at ALI21, further indicating that it is not critical for regulating number per se. Consistent with this observation, overexpression of Plk4 in MTEC (Fig. 5) or in Xenopus larvae MCC 12 did not result in increased centriole number.
Thus, Plk4 may play a similar role as CDK1/CDK2/Plk1/APC-C, by participating in a temporal regulatory mechanism that mediates passage through the various centriole assembly steps.
Centriole abundance in MCC scales with surface area, a phenomenon we observed in airway tissues in vivo and in MTEC cultures in vitro. However, it is unclear which of those properties influences the other: does having a larger surface area result in the formation of more centrioles, or does a cell that forms a larger number of centrioles expand its surface area to accommodate them?
One advantage of using the MTEC culture system is that the ciliogenesis program initiates roughly 2 days after basal cells have already established their size and surface area at ALI0. Therefore, we could temporally separate these two events. By growing cells on increasing extracellular collagen matrix density during the proliferation phase, we caused the enlargement of cell surface area before the transcriptional ciliogenesis program initiated. Remarkably, we discovered that cells formed more centrioles once fully differentiated, suggesting that the centriole amplification machinery responds to the change in surface area. We attempted the reciprocal experiment, which was to induce the formation of excess centrioles and test whether the size of the surface area changed accordingly.
Although constitutive overexpression of Plk4 did result in the formation of excess PC, it did not alter final centriole number or surface area ( Fig. 5 and data not shown).
How are the variations in cell surface area communicated to the centriole amplification pathway to establish centriole number? There are at least three possible ways we envision this could occur. First, larger cells might increase transcription of genes essential for centriole and cilia assembly. This would be consistent with the "limiting component" model of organelle abundance [48][49][50] , where a fixed quantity of a precursor protein(s) would be expressed then "used up" as centriole assembly occurs. In this scenario, the number of centrioles assembled would stop once the limiting component is no longer available. We did note an increase in deuterosome number in cells with enlarged surface area (Fig. 7), suggesting that transcriptional output of at least some precursors is 12 likely elevated. However, it remains unclear whether there is a single limiting component, or if all centriolar proteins become expressed at higher levels in larger cells. Alternatively, cells with larger surface areas might spend a longer period of time in stages of centriole assembly compared to smaller cells. In this scenario, the rate of centriole assembly would be the same in cells of different size, but ones with larger surface areas would spend longer periods in stages I-III to generate more total centrioles. Since the amount of time spent in these assembly stages is coordinated by proteins such as CDK1/Plk1CDK2/APC-C 46, 47 , one possibility is that these pathways help relay information about cell surface area/size to the centriole assembly machinery, and fine-tune the length of time spent in assembly to achieve the desired final number. Finally, it is possible that the apical cytoskeleton helps instruct the centriole amplification pathway to regulate centriole abundance per cell. It is well established that an apical actin network in MCC plays a critical role in modulating the surface area, regulating centriole migration and docking at the plasma membrane, ensuring the even distribution of centrioles across the cell surface, and orienting (planar polarization) of centrioles [51][52][53][54][55][56][57][58] . What remains unclear is whether this actin network communicates with the centriole amplification machinery to regulate centriole number. One possibility is that the variations in surface area may result in corresponding changes to the size of the apical actin lattice, which could then instruct the centriologenesis program to generate more centrioles. Our future studies will focus on defining the pathway(s) coordinating the relationship between cell surface area and centriole amplification. These will have important clinical implications for ciliopathies caused by reduced cilia abundance in MCC, such Primary Ciliary Dyskinesia (PCD) and hydrocephalus. 13

Acknowledgments:
We thank X. Zhu (Shanghai Institutes for  All animal studies were performed following protocols that are compliant with guidelines of the Institutional Animal Care and Use Committee at Washington University and the National Institutes of Health. Mouse Tracheal Epithelial Cell (MTEC) cultures were established as previously described 14,18,19,21 . Briefly, C57BL/6J mice were euthanized at 2-4 months of age, trachea were excised, opened longitudinally to expose the lumen, placed in 1.5 mg/ml pronase E in DMEM/F-12 supplemented with antibiotics and incubated at 4°C overnight. Tracheal epithelial progenitor cells were dislodged by gentle agitation and collected in DMEM/F12 containing 10% FBS. After centrifugation at 4°C for 10 min at 800 g, cells were resuspended in DMEM/F12 with 10% FBS and plated in a Primaria Tissue Culture dish (Corning) for 3-4 h at 37°C with 5% CO 2 to adhere contaminating fibroblasts. Non-adhered cells were collected, concentrated by centrifugation, resuspended in an appropriate volume of MTEC-complete medium (described in 18,19 ), and seeded at a density of 9X10 4 cells/cm 2 onto Transwell-Clear permeable filter supports (Corning) treated with 0.06mg/ml rat tail Collagen type I. Air liquid interface (ALI) was established after cells reached confluence by feeding cells with MTEC serum-free medium 18,19 only in the lower chamber. To ablate parental centrioles using Centrinone, cells were incubated in medium supplemented with 1 µM Centrinone A 22 beginning at 2 days after seeding. Cells were cultured at 37°C with 5% CO 2 , and media containing Centrinone replaced every 2 days for up to 21 days. Samples were harvested at ALI days 0, 1, 2, 3, 4, 6, 8, 12 and 21, fixed and analyzed as described below. All chemicals were obtained from Sigma-Aldrich unless otherwise indicated. All media were supplemented with 100 U/ml penicillin, 100 mg/ml streptomycin, and 0.25 mg/ml Fungizone (all from Invitrogen).
To induce constitutive overexpression of Plk4 in MTEC, transgenic Tg::mCherry-Plk4 animals containing a Chloramphenicol Acetyl Transferase (CAT) coding sequence that includes a stop codon flanked by two loxP sites 39,59 were mated to Tg::Rosa26-Cre ERT2 mice 41 that conditionally express Cre-recombinase in the majority of tissues. Trachea were isolated from Tg::mChPlk4-Rosa26-Cre ERT2 and basal progenitor cells isolated as described. Cre-recombinase expression was induced by adding 1 µm 4-hydroxytamoxifen (Sigma) during the proliferation stage, beginning at 2 days post-seeding. Cells were incubated with tamoxifen for a total of 4-6 days up to ALI 0 to induce expression of mChPlk4. Air liquid-interface was established once cells reached full confluence, samples harvested and fixed at ALI days 0, 1, 2, 3, 4, 8, 12 as described above. 15  Alexa Fluor dye-conjugated secondary antibodies (Invitrogen) were used at a dilution of 1:500 at room temperature for 1 h. Samples were mounted in Mowiol antifade medium containing N-propyl gallate (Sigma). For whole-mount in situ staining, freshly-excised trachea were fixed using 4% paraformaldehyde in PBS at room temperature for 5 min, then ice-cold methanol at −20°C for 5 min.

Lentivirus production and cell infection:
The tissue was washed in PBS, blocked and incubated with primary antibody as described.
Images were captured using a Nikon Eclipse Ti-E inverted confocal microscope equipped with a 60X (1.4 NA) or 100X (1.45 NA) Plan Fluor oil immersion objective lens (Nikon, Melville, NY). A 16 series of digital optical sections (z-stacks) were captured at 0.2 µm intervals using an Andor Neo-Zyla CMOS camera at room temperature. Three-dimensional superresolution structured illumination microscopy (3D-SIM) images were captured on an inverted Nikon Ti-E microscope using 100X oil objective (1.45 NA) at the Washington University Center for Cellular Imaging. Optical z-sections were separated by 0.216 µm, and 3D-SIM images were processed and reconstructed using the Nikon Elements AR 4 Software.

Electron microscopy:
MTEC grown on Transwell filters were fixed with freshly prepared 2.5% glutaraldehyde in PBS, washed in PBS for 30 min, pre-stained with 2% osmium tetroxide and 1% uranyl acetate, dehydrated in graded ethanol series, and then embedded in Embed 812 resin. Samples were serially sectioned from the cell's apical surface to the basal membrane, above the filter. The thickness of the sections was 80 nm (for ALI3 samples), or 120 (for ALI0 samples). Serial sections were transferred onto formvar coated copper slot grids, stained with 2% uranyl acetate and 0.4% lead citrate, and imaged using a transmission electron microscope (FEI Tecnai T12) operating at 80 kV. The alignment of the serial sections and image analysis was performed in Adobe Photoshop and Fiji (NIH).

Image analysis and measurements:
Quantification of centriole abundance, deuterosome number, surface area and volume were performed using Nikon Elements AR 4 Software. To determine the extent of parental centriole loss, cells at ALI 0 were scored for centriole number (0, 1, 2, 3, 4 centrioles) per cell. To determine the fraction of cells in the population undergoing centriole amplification during differentiation, cells at ALI 1, 2, 3, 4 and 8 were scored for the presence of centriolar aggregates (Agg.), which then become defined as multiciliated cells (MCC). This approach was applied to each centriolar marker, and the percentage of the population undergoing centriole amplification determined. The surface area was defined by staining cells with ZO-1 to mark the apical cell-cell junctions, and measured using the thresholding algorithm (Elements AR 4). Cells were binned into three categories that ranged from what we define as small (<50 µm 2 ), medium (50-100 µm 2 ) and large (>100 µm 2 ). Cell volume was determined by measuring the apical-basal distance (depth), which was then multiplied by the surface area. To quantify deuterosome number, MCC at ALI 3-4 were stained for markers of the deuterosome and procentrioles. The number of procentriole-forming deuterosome rings/rosettes 17 were scored and compared in cells of similar size. To determine the number of centrioles per cell in mature MCC at ALI12 or 21, the spot detection tool (Elements AR 4) with a spot circular ROI of 0.26 µm was used. Centriole abundance was then grouped based on cell size as described. Centrioles density was determined by dividing the number of centrioles per cell by the surface area.

Manipulation of cell surface area:
Freshly isolated basal cells were seeded onto Transwell filters coated with varying concentrations (from 0.06mg/mL -2.4mg/mL) of type 1 rat tail collagen. Cells were cultured for 4-6 days until confluent then ALI established. Samples were fixed on the indicated days as analyzed for centriole number, surface area and volume. We found that the size of the surface area did not increase significantly beyond 1.2mg/mL collagen concentration, thus we set that value as the upper limit for all subsequent analyses.

Statistical analyses:
Statistical analyses were performed using GraphPad PRISM 7.0 or Microsoft Excel. The vertical segments in box plots show the first quartile, median, and third quartile. The whiskers on both ends represent the maximum and minimum values for each dataset analyzed. For bar graphs, data are reported as mean ± SEM. Collected data were examined by two-tail unpaired Student's t-test.
Statistical significance was set at p < 0.05.      Supplementary Video 1 -Centrinone-mediated inhibition of Plk4 results in loss of parental centrioles at ALI0. Samples were serially sectioned from at the cell's apical surface until the basal membrane above the filter. The thickness of each section was 120 nm. Serial sections were imaged using a transmission electron microscope, and images aligned using Adobe Photoshop and Fiji software. Sections within first 2 µm are shown to illustrate the absence of parental centrioles, which are evident in control cells at or near the apical surface. Scale bar = 4µm.