Excess microtubule and F-actin formation mediates shortening and loss of primary cilia in response to a hyperosmotic milieu Free

Primary cilia, which are present in most vertebrate cell types, are tiny protruding organelles known to play critical roles as sensory antennae for extracellular signals (Singla and Reiter, 2006; Phua et al., 2015). Various receptors and channels concentrate on primary cilia and are responsible for reception of mechanical and chemical cues (Nauli et al., 2003; Rohatgi et al., 2007; Schou et al., 2015). Primary cilia function has been reported to be critical for cell proliferation and differentiation (Ezratty et al., 2011; Christensen et al., 2012). The structure of primary cilia is supported by well-organized bundles of microtubules, called axonemes. Axonemes of primary cilia are built on the mother centriole. The mother centriole has distal and subdistal appendages radially around its distal end, which create transition fibers to separate the ciliary space from the cytoplasm (Sorokin, 1962; Loncarek and Bettencourt-Dias, 2018; Ma et al., 2023). The proximal ends of centrioles are surrounded by the pericentriolar material (PCM) (Woodruff et al., 2014), which consists of various proteins for organelle trafficking (Lee et al., 2021).

The morphology of primary cilia depends on the balance between ciliogenesis, ciliary elongation and ciliary resorption. Ciliogenesis starts with the assembly of a ciliary vesicle at the distal end of the mother centriole (Lu et al., 2015; Blanco-Ameijeiras et al., 2022). Both distal and subdistal appendages are required for the cilia-directing vesicular transport and ciliary vesicle formation to initiate ciliogenesis (Čajánek and Nigg, 2014). Tubulin dimers subsequently polymerize to form axonemal microtubules at the distal end of the mother centriole. This process requires pericentriolar proteins that constitute the PCM, including ODF2 (Tateishi et al., 2013). The centrioles are docked to the cell membrane, and the polymerizing axoneme protrudes from the cell surface (Garcia-Gonzalo and Reiter, 2012; Reiter et al., 2012). Ciliary elongation depends on anterograde intraflagellar transport (IFT) of ciliary materials. The ejection of ciliary components from primary cilia by retrograde IFT and destabilization of axonemes by deacetylation of tubulin in primary cilia causes resorption of primary cilia (Breslow et al., 2013; Lin et al., 2013; Phua et al., 2017).

The balance between ciliogenesis, ciliary elongation and ciliary resorption is changed by cellular states. Ciliogenesis occurs during the non-dividing phase of the cell cycle, whereas ciliary resorption occurs during the mitotic phase (Plotnikova et al., 2009). Cell-intrinsic circadian rhythms also regulate ciliary elongation and resorption (Tu et al., 2023; Nakazato et al., 2023a). Resorption of primary cilia occurs in response to some stimuli as well. Serum stimulation induces extracellular vesicle release from primary ciliary tips, which triggers rapid loss of the IFT-B complex and resorption of primary cilia (Phua et al., 2017). Heat shock from elevated extracellular temperatures induces resorption of primary cilia mediated by the deacetylase HDAC6 (Prodromou et al., 2012).

Cytoskeletal microtubule and actin filament dynamics affect the morphology of primary cilia. The dynamics of microtubules in the cytoplasm and in the axonemes of primary cilia are also known to contribute significantly to the formation of primary cilia. It has been reported that the amount of free tubulin dimers in the cytoplasm regulates the length of primary cilia; microtubule stabilization shortens the length of primary cilia, whereas accelerated microtubule depolymerization elongates them (Sharma et al., 2011). Docking of the mother centrioles with ciliary vesicles to the cell membrane, which is observed in the early stages of ciliogenesis, requires the removal of cortical actin filaments (Jewett et al., 2021; Tanos et al., 2013). Actin remodeling factors, including LIMK2 and TESK1, have been reported to control ciliogenesis by regulating the transcriptional coactivator Yes-associated protein 1 (YAP), transcriptional coactivator with PDZ-binding motif (TAZ, also known as WWTR1) and directional ciliary vesicle trafficking (Kim et al., 2015). Migration of centrosomes to the apical surface during ciliogenesis in polarized epithelial cells depends on densification of the microtubule network and asymmetrical contraction of the actin cytoskeleton (Pitaval et al., 2017). Actin polymerization within primary cilia also contributes to ciliary resorption, following scission of ciliary tips (Phua et al., 2017).

Unlike normal body fluid osmolarity (≈300 mOsm/l), urine osmolarity fluctuates, ranging from 50 mOsm/l to 1200 mOsm/l due to renal reabsorption (Koeppen and Stanton, 2012; Kitiwan et al., 2021). Primary cilia are present on the epithelial cells of renal collecting ducts. Primary cilia protruding toward the lumen of the collecting ducts are therefore exposed to a wide range of osmolarity caused by urine (Webber and Lee, 1975). Nonetheless, knowledge about the effects of osmotic alteration on primary cilia is limited. Recently, it has been reported that water deprivation and increased urine concentration in mice shortens the length of primary cilia in the kidneys (Kong et al., 2023). Caenorhabditis elegans chemosensory cilia have been reported to accumulate subsets of IFT components at the ciliary tip due to the inhibition of retrograde transport by a hyperosmotic stimulus (Bruggeman et al., 2022). Here, we investigated the impact of hyperosmotic shock on the morphology of primary cilia and their supporting structures in epithelial cells of renal collecting ducts, and probed the underlying mechanisms by using cytoskeleton-targeting reagents.

To determine whether extracellular stress impacts the shape of primary cilia, we investigated the effect of hyperosmotic shock on a cell line derived from murine intramedullary collecting duct epithelial cells (mIMCD-3 cells). The labeled osmolarity of the culture medium used was ≈310 mOsm/l (NaCl, 120 mM; NaHCO₃, 29 mM; osmolarity, 295–325 mOsm/l), which we considered to be iso-osmotic. By adding different amounts of 5 M NaCl to this iso-osmotic medium, we exposed mIMCD-3 cells to hyperosmotic shock ranging from ≈410 mOsm/l to ≈810 mOsm/l for 3 h. We visualized primary cilia by staining ARL13B, a primary ciliary marker, and the ciliary base by staining γ-tubulin, a pericentriolar protein, using immunocytochemistry (ICC) (Fig. 1A).

Under the normal medium osmolarity, ARL13B-positive primary cilia were observed in 81.5±13.0% (mean±s.d.) of total cells (Fig. 1B). Under hyperosmotic shock conditions, the number of ARL13B-positive primary cilia was decreased compared with the number seen in the iso-osmotic condition, and the magnitude of the decrease was dependent on the increase in the extracellular osmolarity (Fig. 1A,B). Cells under hyperosmotic shock at higher than ≈610 mOsm/l exhibited statistically significant loss of primary cilia: 28.9±12.6% of cells in ≈610 mOsm/l culture medium and only 21.9±8.9% of cells in ≈710 mOsm/l medium possessed ARL13B-positive primary cilia (Fig. 1A,B). Under milder hyperosmotic shock at ≈410 mOsm/l or ≈510 mOsm/l, the primary cilia-positive rate decreased less (Fig. 1B; ciliated cells: 70.2±18.5% in ≈410 mOsm/l, 65.1±23.4% in ≈510 mOsm/l).

We quantified the effects of hyperosmotic shock on the length of the remaining primary cilia. The mean length of primary cilia in control cells was 4.5±1.2 µm (mean±s.d.; Fig. 1C). Under hyperosmotic shock, the mean length of primary cilia became shorter (Fig. 1A,C). In contrast to the ARL13B-positive primary cilia rate, the length of remaining primary cilia in ≈510 mOsm/l medium was significantly shorter than that observed for primary cilia in the iso-osmotic condition (Fig. 1C). Furthermore, in higher osmolarity medium, the length of the remaining primary cilia was further reduced (Fig. 1C). These data indicate that in epithelial cells of collecting ducts, milder hyperosmotic shock shortens the length of primary cilia and severe hyperosmotic shock causes the loss of primary cilia.

To track the disassembly of the primary cilia in a time course, we performed the ICC of primary cilia by staining ARL13B in mIMCD-3 cells at different exposure times of hyperosmotic shock: 0.5, 1.0, 1.5, 2.0, 2.5 and 3.0 h (Fig. 1D). The number of ARL13B-positive primary cilia in cells exposed to hyperosmotic shock decreased gradually in a time-dependent manner, with a statistically significant decrease after 1 h in ≈810 mOsm/l culture medium (25.5±7.6%; mean±s.d.) and after 2 h in ≈710 mOsm/l (20.1±5.9%) and ≈610 mOsm/l culture medium (25.7±10.4%) (Fig. 1E).

We also examined the effect of hyperosmotic shock with chemicals other than NaCl on the primary cilia of mIMCD-3 cells. We incubated mIMCD-3 cells for 3 h in hyperosmotic culture medium (≈610 mOsm/l) prepared by adding sucrose (final concentration 300 mM) or mannitol (final concentration 300 mM) and visualized the primary cilia by immunostaining (Fig. S1A). Under hyperosmotic conditions by adding sucrose or mannitol, as with NaCl addition, the number of ARL13B-positive primary cilia decreased (Fig. S1B; ciliated cells: 78.0±2.26% in ≈310 mOsm/l, 34.9±2.52% in ≈610 mOsm/l by sucrose, 40.8±2.20% in ≈610 mOsm/l by mannitol; mean±s.d.) and the length of primary cilia was reduced (Fig. S1C).

We further examined the effect of hyperosmotic shock on primary cilia in cells other than mIMCD-3 cells. Madin–Darby canine kidney II (MDCK-II) cells, an epithelial-like cell line isolated from the distal renal tubule of a dog, and NIH/3T3 cells, a fibroblast cell line isolated from an NIH/Swiss mouse embryo, were chosen. We cultured these cells under iso-osmolarity (≈310 mOsm/l) or hyperosmolarity (≈610 mOsm/l) for 3 h and visualized primary cilia by immunostaining (Fig. S2A,D). In both MDCK-II cells and NIH/3T3 cells, the number of ARL13B-positive primary cilia decreased under hyperosmolarity compared to the number seen in the iso-osmotic condition, as observed for mIMCD-3 cells (Fig. S2B,E; ciliated MDCK-II cells: 50.3±3.55% in ≈310 mOsm/l, 7.20±1.10% in ≈610 mOsm/l; ciliated NIH/3T3 cells: 66.6±3.02% in ≈310 mOsm/l, 19.0±2.52% in ≈610 mOsm/l; mean±s.d.). Interestingly, the hyperosmotic shock did not lead to a reduction in length of remaining primary cilia in both MDCK-II cells and NIH/3T3 cells (Fig. S2C,F), which was a great contrast to mIMCD-3 cells.

We next carried out live-cell imaging to observe how primary cilia shortened and disassembled during the 3 h of hyperosmotic shock. Since overexpressing ciliary marker proteins causes abnormal shapes and elongation in primary cilia (Nachury et al., 2007; Ijaz and Ikegami, 2021), we chose a cell line that stably and endogenously expresses ARL13B tagged with the Venus fluorescent protein (ARL13B–Venus). To eliminate unknown effects of antibiotic-resistance gene expression as part of a gene knock-in approach, we instead directly inserted the Venus sequence into sequence encoding the C terminus of endogenous ARL13B using the HITI method (Suzuki et al., 2016). With the HITI method, we generated a clone where an allele of Arl13b was modified to encode ARL13B C-terminally tagged with Venus. We performed the three-dimensional time-lapse imaging of the ARL13B–Venus-expressing mIMCD-3 cells under iso-osmolarity (≈310 mOsm/l) and hyperosmolarity (≈610 mOsm/l) for 3 h respectively (Movie 1; Fig. 1F). In the acquired movie, almost half of the primary cilia became shorter over time, and most of them eventually disappeared after the 3 h of hyperosmotic shock at ≈610 mOsm/l (Movie 1, right; Fig. 1F, hyperosmotic). In contrast, the majority of primary cilia retained their length throughout the 3 h of recording time under the iso-osmotic conditions of ≈310 mOsm/l (Movie 1, left; Fig. 1F, iso-osmotic), indicating that the laser emission from the confocal microscope had little effect on the length of primary cilia or the fluorescence intensity of Venus. We also measured the length of primary cilia at each time point in the time-lapse data (Fig. 1G; blue curve represents cilium length at ≈310 mOsm/l, red and green curves represent lengths of independent cilia at ≈610 mOsm/l). Primary cilia kept their length in the first hour and began to shorten after 1 h of exposure to the hyperosmotic shock (Fig. 1G; 0–60 min). The reduction in length accelerated in the second hour (Movie 1; Fig. 1G; 60–120 min), and the shortened primary cilia finally vanished in several minutes during the third hour (Movie 1; Fig. 1G; 120–180 min). This time-lapse imaging demonstrates that hyperosmotic shock induces the gradual shrinking and consequent loss of primary cilia.

The data we present in Fig. 1 show the shortening and loss of primary cilia upon hyperosmotic shock. ARL13B, a ciliary membrane protein, was used as a marker for primary cilia. It was unclear whether the ciliary axoneme behaved the same way as the ciliary membrane. To examine whether the ciliary axoneme is affected by hyperosmotic shock, we exposed cells to hyperosmotic shock at ≈510 mOsm/l or ≈710 mOsm/l for 3 h and performed ICC on both the ciliary axoneme (using α-tubulin as a marker) and ARL13B by fixing cells with a popular fixative, 4% paraformaldehyde (Fig. 2A). Surprisingly, with the hyperosmotic shocks, α-tubulin signals were diminished or lost in primary cilia that were detected with anti-ARL13B antibodies (Fig. 2A; ≈510 mOsm/l or ≈710 mOsm/l). We quantified the percentage of ARL13B-positive primary cilia that were also positive for α-tubulin (Fig. 2B). Following hyperosmotic shock at ≈510 mOsm/l, the percentage of primary cilia positive for α-tubulin was lower (27.6±18.6%; mean±s.d.) than that observed in iso-osmotic conditions (81.2±11.2%). Interestingly, the percentage of α-tubulin-positive primary cilia with hyperosmotic shock at ≈710 mOsm/l (57.0±24.1%) was higher than that with hyperosmotic shock at ≈510 mOsm/l (Fig. 2B).

We also performed electron microscopy of primary cilia exposed to hyperosmotic shock. In transmission electron microscopy (TEM) of longitudinal sections, axonemes of cilia appeared blurred or missing even on centriolar microtubules under hyperosmotic conditions, as compared to iso-osmotic conditions (Fig. 2C). In array tomography of serial cross sections, axonemal microtubules were observed close to the tip of the primary cilia, which were shortened by the hyperosmotic shock (Fig. 2D). However, several microtubules were singlets from the root of primary cilia (Fig. 2D), which is a great contrast to the axonemal structure previously reported for primary cilia of mIMCD-3 cells (Gluenz et al., 2010; Sun et al., 2019).

To further investigate the seemingly discrepant results between fluorescence and electron microscopy, microtubule staining was compared using different fixation methods for fluorescence microscopy. Adding a small amount of glutaraldehyde (0.2%) to the fixation solution restored axonemal microtubule staining even under hyperosmotic conditions (Fig. S3). These data suggest that axonemes of primary cilia are destabilized by hyperosmotic shock.

We next examined whether the PCM is disrupted under hyperosmotic shock. We immunostained two pericentriolar proteins, ODF2 and γ-tubulin, as well as DNA after 3 h of hyperosmotic shock at ≈510, ≈610, ≈710 and ≈810 mOsm/l (Fig. 3A). ODF2 and γ-tubulin showed different responses to hyperosmotic shocks. Although the positive rates of both proteins decreased in a hyperosmotic shock-dependent manner, γ-tubulin showed more tolerance to hyperosmotic shock than ODF2 (Fig. 3B,C). Whereas the ODF2-positive rate significantly dropped at ≈610 mOsm/l, with the lowest rate at ≈710 mOsm/l (Fig. 3B), the γ-tubulin-positive rate showed a statistically significant drop at ≈710 mOsm/l and above (Fig. 3C).

We further observed how the γ-tubulin-positive rate decreased over time for different hyperosmolarity conditions. We immunostained γ-tubulin in mIMCD-3 cells after different exposure times of hyperosmotic shock (Fig. 3D). Out of all the hyperosmolarity conditions we tested, the γ-tubulin-positive rate dropped throughout the 3 h of hyperosmotic shock (Fig. 3D). At ≈710 mOsm/l and ≈810 mOsm/l, the highest decrease in the percentage of γ-tubulin-positive cells occurred between the 0 h and 0.5 h time points (Fig. 3E; 21.8±19.1% in ≈710 mOsm/l medium and 11.3±8.4% in ≈810 mOsm/l medium; mean±s.d.). These observations were highly contrasting with the results of our ARL13B-positive cilia resorption assay: ∼50% of ARL13B-positive primary cilia were still present at 0.5 h after exposure to ≈710 mOsm/l or ≈810 mOsm/l hyperosmotic shock (Fig. 1E). These results demonstrate that pericentriolar proteins, including γ-tubulin and ODF2, disappear from the base of primary cilia prior to the shortening and disassembly of primary cilia during hyperosmotic shocks.

Osmolarity differences across cell membranes create sheer stress on cells, especially on structures with high surface area-to-volume ratio like primary cilia (William, 2005). We aimed to exclude the possibility that primary cilia were being torn apart by the physical and osmotic stress we tested. We incubated the cells at three different temperatures – 4°C, 22°C and 37°C – during hyperosmotic shock for 3 h at ≈610 mOsm/l and ≈710 mOsm/l. We immunostained ARL13B to observe primary cilia (Fig. S4A). At iso-osmolarity, the cilia-positive rate did not significantly change at different temperatures (Fig. S4B). After hyperosmotic shock at ≈610 mOsm/l and ≈710 mOsm/l, the cilia-positive rate decreased significantly at 37°C (Fig. S4C,D). With hyperosmotic shock at ≈610 mOsm/l, the cilia-positive rate dropped as low as 10.7±0.9% (mean±s.d.) at 37°C, as compared to 65.5±4.6% at 4°C and 66.1± 2.2% at 22°C (Fig. S4C). These results suggest that the loss of primary cilia induced by hyperosmotic shock is not caused by sheer physical and osmotic stress but by enzymatic and biochemical processes.

The method of hyperosmotic exposure used resulted in steep increases in osmolarity, particularly when the osmolarity was ≈610 mOsm/l or ≈710 mOsm/l (Figs 1, 2). To exclude the possibility that the shortening and disassembly of cilia and disruption of the PCM were artifacts caused by the acute change of osmolarity, we exposed cells to gradual increases in medium osmolarity: ≈410 mOsm/l for 1 h, followed by ≈510 mOsm/l for 1 h and then ≈610 mOsm/l for 2 h (Fig. 4A). The gradual increase of osmolarity induced the same morphological changes as seen following an acute change in osmolarity (Fig. 4B): loss of ARL13B-positive primary cilia (Fig. 4C; from 75.4±11.2% of cells to 7.7±4.8% of cells; mean±s.d.) and loss of γ-tubulin (Fig. 4D; from 92.4±4.9% of cells to 57.4±16.5% of cells; mean±s.d.). These results indicate that the hyperosmotic shock-dependent loss of ARL13B-positive primary cilia and PCM does not only result from an acute increase in the medium osmolarity.

To examine whether degradation of cilia- and PCM-related proteins caused the observed shortening and loss of cilia and disruption of the PCM, we investigated the total amount of ARL13B and γ-tubulin in the cell lysate with western blot analyses after 3 h of hyperosmotic shock at ≈610 mOsm/l (Fig. 4E). ARL13B, a cilia-related protein, showed almost no significant change in its protein level (Fig. 4F), suggesting that the hyperosmotic shock-induced loss of cilia neither necessarily affects the total levels of cilia-related proteins nor results from ciliary protein degradation. Although both γ-tubulin and ODF2 signals disappeared upon hyperosmotic shock in our ICC experiments, γ-tubulin did not show a statistically significant change in protein level (Fig. 4G), suggesting that hyperosmotic shock does not degrade the pericentriolar proteins but could delocalize them.

As the hyperosmotic shock led to disruption of the PCM in our ICC experiments, we performed TEM on centrioles to examine whether the centrioles are disassembled by hyperosmotic shock at ≈610 mOsm/l (Fig. 4H). We observed that the centrioles were not disassembled upon hyperosmotic shock. In contrast, the electron density around the centrioles of cells that had experienced hyperosmotic shock was less than that around the centrioles of cells in iso-osmotic conditions (Fig. 4H). These results further indicate hyperosmotic shock-induced delocalization of the PCM.

To clarify whether hyperosmotic shock-induced cilia loss and PCM disruption result from cell death, we stained the cells with DAPI and propidium iodide (PI) immediately after 3 h of hyperosmotic shock (Fig. 5A). Even under the highest osmolarity we tested, ≈810 mOsm/l, the cell death rate was below 10% (Fig. 5B). The osmolarity condition that was sufficient to induce cilia loss and PCM disruption, ≈610 mOsm/l, led to almost no cell death (Fig. 5B).

To further investigate the lasting effects of hyperosmotic shock and examine whether the morphological changes to the primary cilia and the PCM are reversible, we restored the osmolarity of the medium to iso-osmotic, ≈310 mOsm/l, and incubated the cells for another 3 h after hyperosmotic shock for 3 h at ≈610 mOsm/l, followed by ICC staining of ARL13B and γ-tubulin (Fig. 5C). Both the primary cilia-positive rate and the γ-tubulin-positive rate were restored close to that observed before the hyperosmotic exposure; the primary cilia-positive rate was 58.1% before the hyperosmotic shock and increased back to 57.1% at 3 h after the medium osmolarity was restored to being iso-osmotic (Fig. 5D). The γ-tubulin-positive rate was 72.6% before the hyperosmotic shock and 69.4% after the restoration of iso-osmotic osmolarity (Fig. 5E). Because the hyperosmotic shock we tested did not cause significant cell death, the disassembly of primary cilia and loss of the PCM are temporary phenomena and are reversible without taking longer than the cell cycle.

Next, we applied pharmacological approaches to investigate how cells shorten or disassemble their primary cilia. Our findings showed that hyperosmotic shock caused loss of axonemal microtubules preceding loss of ARL13B-positive ciliary membrane while inducing excessive formation of cytoplasmic microtubule bundles (Fig. 2). To gain mechanistic insight, we thus investigated the contribution of microtubules to the shortening and loss of primary cilia and delocalization of the PCM resulting from hyperosmotic shock. We first tested the effects of nocodazole, which inhibits microtubule polymerization and promotes microtubule depolymerization, on the shortening and loss of primary cilia and the delocalization of the PCM. We exposed cells to hyperosmotic shock (≈610 mOsm/l) for 3 h with or without nocodazole. Nocodazole at 400 nM strongly blocked the hyperosmotic shock-induced excessive formation of cytoplasmic microtubule bundles (Fig. 6A). Under these conditions, we performed ICC to detect ARL13B, γ-tubulin and ODF2 (Fig. 6B). Remarkably, nocodazole treatment prevented the hyperosmotic shock-induced loss of primary cilia (Fig. 6B, top). The percentage of cells with primary cilia in iso-osmotic conditions was ∼70% and decreased to ∼20% in hyperosmotic conditions (≈610 mOsm/l) without nocodazole. The percentage of cells with primary cilia in hyperosmotic conditions recovered to ∼50% in the presence of 400 nM of nocodazole (Fig. 6C). Nocodazole also prevented the hyperosmotic shock-induced shortening of primary cilia (Fig. 6D). Nocodazole prevented delocalization of the PCM more effectively (Fig. 6B, middle and bottom). The percentages of γ-tubulin-positive cells and ODF2-positive cells in hyperosmotic conditions with nocodazole recovered to the same levels as seen in the iso-omolarity control (Fig. 6E,F). These results indicate that the hyperosmotic shock-induced morphological changes of primary cilia and delocalization of the PCM are dependent on excess microtubule formation.

The results of experiments using paclitaxel (PTX), a microtubule polymerization-promoting agent, further supported the idea that the hyperosmotic shock-induced shortening and loss of primary cilia occurred upon excessive microtubule formation. Excess formation of microtubule bundles in cytoplasm following PTX treatment was confirmed by staining microtubules with an anti-α-tubulin antibody (Fig. S5A). PTX treatment failed to prevent the hyperosmotic shock-induced loss of primary cilia at any PTX concentration we tested (Fig. S5B,C). The lengths of remaining primary cilia in the hyperosmotic conditions were also not recovered by any concentrations of PTX tested (Fig. S5D).

We next explored contribution of F-actin to the shortening and loss of primary cilia and delocalization of the PCM by hyperosmotic shock. To this end, we tried latrunculin A (LatA), an inhibitor for F-actin formation. The effect of LatA on F-actin formation was confirmed by labeling cells with fluorophore-conjugated phalloidin. LatA effectively disrupted F-actin in a concentration-dependent manner under normal conditions (Fig. 7A) and suppressed excess formation of F-actin upon hyperosmotic shock (Fig. 7B). The hyperosmotic shock-induced loss of primary cilia was partially prevented by LatA treatment even at 200 nM (Fig. 7C). The percentage of cells with primary cilia was as ∼40% following hyperosmotic shock in the presence of 200 nM LatA, whereas the percentage of cells with primary cilia decreased from ∼80% to ∼21% following hyperosmotic shock without LatA (Fig. 7C,D). Even with the highest concentration of LatA tested (1000 nM) the loss of primary cilia upon hyperosmotic shock was not completely prevented; treatment with 1000 nM LatA held the cilia-positive rate at ∼50% upon hyperosmotic shock (Fig. 7D). Also, LatA partially prevented the hyperosmotic shock-induced shortening of primary cilia at all LatA concentrations tested; the hyperosmotic shock-induced shortening of primary cilia was not completely prevented even with the highest concentration of LatA, 1000 nM (Fig. 7E).

To investigate whether LatA also prevents delocalization of the PCM, we stained the pericentriolar proteins γ-tubulin and ODF2 after 3 h of hyperosmotic shock (≈610 mOsm/l) with 400 nM LatA present (Fig. 7F). The rates of γ-tubulin- or ODF2-positive cells were recovered significantly by LatA treatment under hyperosmotic shock (Fig. 7G,H). Combined with the prevention of hyperosmotic shock-induced shortening and loss of cilia by LatA, these data suggest that both the hyperosmotic shock-induced morphological changes of cilia shortening and loss, and the delocalization of the PCM, are dependent on F-actin.

We finally examined whether hyperosmotic shock-induced shortening and loss of primary cilia is associated with excision of primary cilia tips, given that these changes to primary cilia were found to be dependent on F-actin (Fig. 7). We collected extracellular vesicle-like particles from conditioned medium by ultracentrifugation after 3 h of hyperosmotic shock and after culture in iso-osmotic conditions to measure the amount of primary cilia fragments released. A protein anchored to the membrane of primary cilia, ARL13B, and the IFT proteins IFT88 and IFT52 were detected in the conditioned medium pellets by western blot analyses, whereas the ciliary membrane protein PKD2 (also known as polycystin 2, PC2) was not detected in the conditioned medium pellets (Fig. 8A). Interestingly, the three ciliary proteins detected in the conditioned medium pellets showed different release tendencies. The ciliary membrane marker ARL13B showed ∼1.5-fold increase in the conditioned medium pellet from the hyperosmotic shock condition (Fig. 8B). Similarly, IFT52 was ∼1.5-fold increased in the conditioned medium pellet following hyperosmotic shock (Fig. 8C). In contrast, IFT88 showed a decrease in conditioned medium pellet following hyperosmotic shock (Fig. 8D). These results demonstrate that a subset of ciliary proteins are released into the culture medium following hyperosmotic shock.

This work demonstrates that acute hyperosmotic shock causes reversible shortening and disassembly of primary cilia and delocalization of the PCM without structural disassembly of the centrioles in renal collecting duct epithelial cells. These morphological changes occur in a manner dependent on excess formation of microtubule and F-actin. Additionally, these changes are accompanied by the release of a subset of ciliary proteins into the culture medium.

We found that primary cilia were shortened or lost with hyperosmotic shock and that the reduction of primary cilia length was dependent on NaCl concentration (Fig. 1). With milder hyperosmotic shock (≈410 mOsm/l), the primary cilia length was only slightly reduced, and with stronger hyperosmotic shock (over ≈610 mOsm/l), the primary cilia were markedly shortened and disassembled. This phenomenon seems to be evolutionally conserved, as tonicity-induced Chlamydomonas deflagellation has previously been reported (Solter and Gibor, 1978). An argument could be raised as to whether the hyperosmotic shock-induced shortening and loss of primary cilia are artifacts; however, the possibility of artifacts can be excluded. First, a recently published work supports our in vitro findings by showing that water deprivation causes shortening of renal primary cilia in vivo (Kong et al., 2023). Second, the osmolarity we tested is physiological for the epithelial cells of renal collecting ducts, which are subject to a very wide range of osmolarity from 50 to 1200 mOsm/l (Koeppen and Stanton, 2012). Renal epithelial cells are reported to adapt to these hyperosmotic stresses (Dmitrieva and Burg, 2005). The mIMCD-3 cells we used have been generated by passing through hyperosmotic conditions up to 900 mOsm/l (Rauchman et al., 1993). In contrast to a previous study where mIMCD-3 cells were exposed to hyperosmotic pressure for a few days (Pihakaski-Maunsbach et al., 2010), the duration of hyperosmotic shock we tested on these cells did not induce cell death in the majority (Fig. 5). Third, our data demonstrate that the hyperosmotic shock-induced morphological changes are reversible (Fig. 5). This reversibility fits well with oscillation of urine osmolarity, which depends on water availability and/or intake (Koeppen and Stanton, 2012). Lastly, a gradual increase in osmotic pressure also caused the shortening and loss of primary cilia (Fig. 4), which also excludes the possibility of artifacts, as urine osmolarity is expected to change gradually in vivo.

We demonstrated that at least two PCM proteins, γ-tubulin and ODF2, were disrupted with the hyperosmotic shock (Fig. 3); however, neither of these PCM proteins were degraded, and the nine-triplet centrioles were not disassembled (Fig. 4). These results indicate that disruption of the pericentriolar proteins is caused by delocalization. The PCM is expected to appear as an electron-dense region in TEM imaging (Dallai et al., 2016). Our TEM data revealed that the region surrounding the centrioles had reduced electron density following hyperosmotic shock, as compared to that of the corresponding region in cells from iso-osmotic conditions. This is consistent with the PCM delocalization we demonstrated using ICC and western blotting. The delocalization of the two proteins showed different tolerance to hyperosmolarity: γ-tubulin was delocalized at ≈710 mOsm/l or above, whereas ODF2 was delocalized at ≈610 mOsm/l or above (Fig. 3). This difference might result from the different roles of the two proteins. γ-Tubulin is a major component of γ-tubulin ring complexes (γTuRCs), which are responsible for the nucleation of microtubules around the centriolar walls (Tovey and Conduit, 2018; Schweizer and Lüders, 2021). The dissolution of γ-tubulin possibly indicates the inactivation of PCM function as a centrosomal microtubule-organizing center (MTOC) surrounding the centrioles (Magescas et al., 2019). In contrast to γ-tubulin, ODF2 is an initiator for the formation of subdistal appendages. The delocalization of ODF2 upon hyperosmotic shock suggests improper formation of distal and/or subdistal appendages (Ishikawa et al., 2005; Tateishi et al., 2013; Huang et al., 2017). Pericentriolar proteins of the PCM are reported to be disassembled after the onset of anaphase at different time points (Mittasch et al., 2020). The differential tolerance of γ-tubulin and ODF2 to hyperosmotic shock possibly represents the order of deformation of the basal structure of primary cilia upon hyperosmotic shock.

An unexpected finding of this study is the apparent loss of α-tubulin signals in primary cilia that are detected with anti-ARL13B antibodies under hyperosmotic conditions (Fig. 2). Electron microscopy produced seemingly discrepant results, with axonemal microtubules remaining under the hyperosmotic conditions. Since α-tubulin signals were detected in primary cilia even with fluorescence microscopy when glutaraldehyde was used for fixation, the loss of α-tubulin could be an artifact of fixation with only paraformaldehyde before ICC. The immunoreactivity of tubulin or modified tubulin is reported to be drastically affected by fixation methods (Hua and Ferland, 2017), which reflects differences in structural stabilities or antigen exposure states. Thus, the loss or extensive decrease of α-tubulin immunoreactivity to under the detection limit in primary cilia exposed to hyperosmotic shock would also imply changes to axonemal microtubules under the conditions. The appearance of singlet microtubules from the base of primary cilia supports the idea of destabilization of axonemal microtubules. These interesting phenomena should be investigated in future work.

We demonstrated that nocodazole, a microtubule-disrupting agent, partially prevented shortening and loss of cilia and prevented delocalization of the PCM following hyperosmotic shock (Fig. 6). In contrast, PTX, a microtubule polymerization-promoting agent, failed to prevent the hyperosmotic shock-induced shortening and loss of primary cilia (Fig. S5). These data indicate that the hyperosmotic shock-induced changes depend on excess formation of microtubules. Several other works support this idea: excess formation of cytoplasmic microtubules seems to limit the tubulin supply for the axoneme in the ciliary space and suppresses the elongation of primary cilia (Sharma et al., 2011; Nunes et al., 2013). Treatment with PTX would mimic the excess formation of microtubules caused by hyperosmotic shock, and thus fail to prevent the shortening and loss of cilia. Excess formation of bundle-like microtubules could explain the mechanism of a part of the PCM delocalization observed. γTuRCs are reported to also function in non-centrosomal MTOCs (Sanchez and Feldman, 2017). The hyperosmotic shock-induced γ-tubulin delocalization we showed possibly indicates forceful PCM disassembly accompanying inactivation of the centrosomal MTOC. We assume that the delocalization of γ-tubulin possibly results from diffusion of γ-tubulin to nucleate microtubules extensively in the cytoplasm. Delocalization of ODF2 by hyperosmotic shock could result from abnormal interactions between ODF2 and excessively formed bundle-like microtubules, as ODF2 has been proposed to have ability to interact with microtubule lattices (Kunimoto et al., 2012).

The formation of cytoplasmic microtubule bundles upon hyperosmotic shock is also arguable. Several works have reported conflicting results about the relationship between osmolarity and microtubules in the cytoplasm. Hyperosmotic stimulation of HeLa cells with high concentrations of NaCl results in degradation of cytoplasmic microtubules (Skoge and Ziegler, 2016). Chlamydomonas microtubules are disassembled by hyperosmotic stimuli (Ng et al., 2022). Our findings with mIMCD-3 cells seem to oppose these works. A recent work could support our findings: MAP7 has been reported to bind to cytoplasmic microtubules more preferentially under hyperosmotic conditions upon changes to tubulin post-translational modifications (Shen and Ori-McKenney, 2024); however, obvious microtubule bundling is neither demonstrated nor mentioned by the authors. Differences between our work and the others mentioned above include the cell types used and the cell density. Only our work uses fully differentiated epithelial cells that reach full confluence. Alternatively, bundle-like microtubule formation upon hyperosmotic shock might occur specifically in the epithelial cells of the renal collecting duct, which are repeatedly exposed to severalfold increases in osmolarity under physiological conditions. These differences could explain the different responses of primary cilia length to hyperosmotic shock observed for mIMCD-3 cells and for MDCK-II cells and NIH/3T3 cells. Details about the interconnections between microtubules in the cytoplasm and those in primary cilia should be investigated in future works with multiple dimensional analyses that include changes in tubulin post-translational modifications and/or translocation of microtubule-associated proteins between those two compartments.

We found that an actin polymerization inhibitor, LatA, partially suppressed the shortening and loss of cilia and suppressed delocalization of the PCM following hyperosmotic shock (Fig. 7). The results indicate that the hyperosmotic shock-induced morphological changes are dependent of actin polymerization. F-actin formation activates the transcriptional coactivators YAP and TAZ, which leads to upregulation of AURKA and PLK1 expression, resulting in inhibition of ciliogenesis (Kim et al., 2015). Inhibiting actin polymerization has been reported to increase ciliogenesis and elongate cilia (Kim et al., 2010; Avasthi and Marshall, 2012). These observations imply that cells increase disassembly of primary cilia in an actin polymerization-dependent manner with the onset of hyperosmotic shock. We also showed that the hyperosmotic shock-induced γ-tubulin and ODF2 delocalization was prevented by treatment with LatA (Fig. 7). We assume that hyperosmotic shock-induced PCM delocalization occurs in an actin polymerization-dependent manner. The actin cytoskeleton is reported to control the subcellular localization of centrosomes by remodeling of the actin network and increasing network contractility (Pitaval et al., 2017). This report suggests that the cells delocalize the PCM in an actin polymerization-dependent manner with the onset of hyperosmotic shock.

By collecting extracellular vesicle-like particles, we detected a trace of primary cilia components in the culture medium (Fig. 8). Upon hyperosmotic shock, abundance of a ciliary membrane-anchored protein, ARL13B, somewhat increased in the medium pellet. The increased ciliary trace in the medium suggests an increase in ciliary tip scission caused by hyperosmotic shock. Ciliary tip scission and subsequent disassembly of cilia require F-actin formation, as faulty F-actin formation inhibits ciliary tip scission (Nager et al., 2017; Phua et al., 2017; Wang et al., 2019). These reports are consistent with our data showing increased ARL13B in the medium pellet and suppression of hyperosmotic shock-induced cilia disassembly by LatA treatment. The other ciliary membrane protein tested, PKD2, was not detected in the conditioned medium pellet under both iso-osmotic and hyperosmotic conditions. The cargo of ectosomes is dependent on which channels are activated at the time. Activated G-protein-coupled receptors make up the bulk of ectosome cargo as compared to bystander proteins (Nager et al., 2017). This could be the reason for the absence of PKD2 in the conditioned medium pellet. Our data also show interesting results about the release of IFT proteins. The abundance of IFT88, a component of the IFT-B core complex, was decreased in the medium pellet following hyperosmotic shock. In contrast, the abundance of the other IFT-B core complex tested, IFT52, was increased in the conditioned medium pellet under hyperosmotic conditions. We propose that hyperosmotic shock-induced cilia disassembly accompanies a selective retrieval process for ciliary proteins. This suggests that the increased ciliary tip scission and subsequent disassembly of cilia induced by hyperosmotic shock are different from growth stimulation-driven cilia scission, where both ARL13B and the majority of IFT-B proteins are released into the culture medium (Phua et al., 2017). The hypothetical retrieval process we propose would leave ARL13B as ectosome cargo and transport IFT88 back to the cytoplasm upon hyperosmotic shock. The opposite trend of ectosomal release of IFT52 and IFT88 could be explained by some recent advances in understanding of the structure of IFT complexes. Direct interaction between IFT52 and IFT88 seems to be limited to a few domains (Petriman et al., 2022). IFT88 also interacts with IFT144 (also known as WDR19), a protein of the IFT-A complex (Hesketh et al., 2022). Under hyperosmotic conditions, the IFT-B complex could be dissociated and IFT88 could be selectively transported to the soma by the IFT-A complex. Primary cilia in human umbilical vein endothelial cells disassemble under shear stress. In this process, IFT disappears from primary cilia prior to acetyl-α-tubulin degradation (Iomini et al., 2004). In the process of flagellar disappearance in Chlamydomonas due to temperature stress, IFT also disappears prior to flagellar loss (Iomini et al., 2001). The loss of the axoneme prior to ARL13B-positive ciliary membrane retrieval could explain the hypothetical process (Fig. 2). Destabilization of the axoneme could sequester IFT proteins from the tip of primary cilia, while remaining ARL13B-positive ciliary membrane could be pinched off.

An argument can be made for the biological significance of hyperosmotic shock-induced shortening and loss of primary cilia. In human bodies, hypovolemia makes kidneys concentrate urine. High urine osmolarity and decreased kidney function – that is, lower estimated glomerular filtration rate (GFR) – are reported to have a positive correlation (Kitiwan et al., 2021). PKD2 in renal proximal tubules, a Ca²⁺-permeable cation channel localized on the ciliary membrane, is necessary to maintain normal GFR under fluctuating fluid shear stress (Du et al., 2021). Increased fluid shear stress bends primary cilia, increases Ca²⁺ signaling from PKD2 and eventually increases apical endocytosis in renal proximal tubules. The potential loss of primary cilia in high osmolarity conditions could decrease the capability of kidneys to adjust GFR, causing kidneys to form cysts (Raghavan et al., 2014). The morphological changes of primary cilia and PCM shown in this research suggest that these changes are temporary, reversible cellular responses to osmolarity increase. The disassembly of primary cilia accompanied by PCM delocalization is possibly triggered to protect primary cilia from fluid-driven shear stress (Renoux et al., 2019) by increased osmolarity, as primary cilia have a higher surface-to-volume ratio than cell bodies.

In summary, we have demonstrated that the extracellular stress of hyperosmotic shock shortens and disassembles primary cilia and delocalizes the PCM, with the centrioles kept intact, in a manner that is dependent on excess formation of microtubules and F-actin. These morphological changes of primary cilia and the PCM are temporal changes and are thus reversible when the extracellular osmolarity is restored. Although the deeper mechanism behind these changes and how they affect cellular functions are still unclear, our findings suggest that the primary cilium is a structure physiologically responsive to changes in the extracellular milieu.

mIMCD-3 (mouse inner medullary collecting duct-3) cells (ATCC, CRL-2123) were cultured in Dulbecco's modified Eagle's medium (DMEM)/Ham's F-12 (Wako, Osaka, Japan) with 10% fetal bovine serum (FBS; Gibco, CA, USA) and incubated at 37°C with 5% CO₂. MDCK-II cells were generously gifted from Dr Mitsunori Fukuda, Tohoku University, Miyagi, Japan (Homma et al., 2019). MDCK-II cells and NIH/3T3 cells (ATCC, CRL-1658) were cultured in DMEM-high glucose (044-29765, Wako) with 10% FBS. All the cell lines were routinely inspected by microscopy to check cell shape, growth speed, behaviors and the absence of any contaminants. To induce ciliogenesis, the medium was replaced by culture medium without FBS (mIMCD-3 and MDCK-II) or with 1% FBS (NIH/3T3).

To acquire true knock-in cells (Ijaz and Ikegami, 2019), we inserted sequence encoding Venus into the sequence encoding the C terminus of mouse ARL13B using the HITI method (Suzuki et al., 2016). For CRISPR/Cas9, we used a mouse Arl13b 20-bp target sequence and 3-bp PAM sequence (italics): 5′-GCTGTGCGACAGAGACCTAACGG-3′. To construct the Cas9- and gRNA-expression plasmid vector, the 20-bp mouse Arl13b target sequence was sub-cloned into the CMV-Cas9-2A-GFP plasmid backbone (ATUM, CA, USA). To construct the HITI donor plasmid, exon10 of the mouse Arl13b gene had sequence added to encode C-terminal HA (Radeck et al., 2013) and Venus tags (Nagai et al., 2002), and was sandwiched by reversed Cas9/gRNA target sequences. This sequence was cloned into the TOPO vector backbone (pCRII-TOPO; Invitrogen, MA, USA). Cells co-transfected with Cas9/gRNA expression vector and HITI donor vector were allowed to grow for 2–3 days and later expanded into a 10 cm dish. Afterward, cells were collected and single-cell-cloned using the limiting dilution culture method at ∼0.2–0.3 cell/well into 96-well plates. Cells were allowed to grow for 1–2 weeks. For the selection of monoclonal colonies, microscopic observations were made to monitor single-cell colony formation and confluency. Selected colonies were expanded into duplicate multi-well plates, one of which was utilized to make frozen stocks for subsequent use and the other for screening. Western blot analysis was used to screen the positive clones.

Cells were subjected to serum starvation by reducing FBS concentration to 0% for 24 h after the cells reached confluency. We added 5 M NaCl in ultrapure water, 1 M sucrose in DMEM/Ham's F-12 or 0.8 M mannitol in DMEM/Ham's F-12 to fresh serum-free medium to increase its osmolarity from its original osmolarity, rated at ≈310 mOsm/l. The hyperosmotic media ranged from ≈410 mOsm/l to ≈810 mOsm/l, and the existing culture medium was replaced by the relevant hyperosmotic medium. Hyperosmotic medium was prepared with a final concentration of 50 mM (410 mOsm/l) to 250 mM (810 mOsm/l) of added 5 M NaCl and a final concentration of 300 mM (610 mOsm/l) of added 1 M sucrose or 0.8 M mannitol. PI (1 µg/ml; Wako, Osaka, Japan) was added for cell death examination immediately after this hyperosmotic shock. To add cytoskeleton-targeting reagents – nocodazole (Wako, Osaka, Japan), PTX (Taxol; Wako, Osaka, Japan) or latrunculin A (Wako, Osaka, Japan) – we used DMSO as the vehicle (0.1%, v/v). For the PI-DAPI cell death assay, cells were incubated with PI in PBS at 37°C for 20 min, washed with PBS, fixed with 4% PFA and stained with DAPI.

Cells were cultured on nitric acid-washed cover glasses (Matsunami, Japan). They were incubated in each culture medium for 3 h and fixed with 4% paraformaldehyde or with 4% paraformaldehyde and 0.2% glutaraldehyde in phosphate-buffered saline (PBS, pH 7.5) for 30 min at 37°C. Cells were blocked and permeabilized with blocking solution – 5% normal goat serum (Gibco) with 0.1% Triton X-100 in PBS – for 1 h at room temperature. Then, cells were incubated overnight with indicated combinations of these primary antibodies: anti-ARL13B (17711-1-AP; Proteintech, IL, USA; 1:400), anti-γ-tubulin (T6557; Sigma-Aldrich, MO, USA; 1:400), anti-ODF2 (ab43840; Abcam, USA; 1:400) and anti-α-tubulin (T9026; Sigma-Aldrich; 1:400) diluted in blocking solution. Cells were washed with PBS and incubated for 1 h with Alexa Fluor 568- or Alexa Fluor 488-conjugated secondary antibodies (Thermo Fisher Scientific, Rockford, IL, USA; 1:500), Alexa Fluor 488-conjugated phalloidin (Thermo Fisher Scientific; 1:500) and DAPI (DOJINDO, Japan; 1:1000). Fluorescence images were acquired with a laser scanning confocal microscope (Fluoview FV1000, Olympus, Japan; or STELLARIS5, Leica, Germany) equipped with an oil immersion lens (60×, NA 1.35 for FV1000; 63×, NA 1.4 for STELLARIS5).

The cell lysates were made by adding 1× SDS-PAGE sample buffer (Wako, Osaka, Japan) to the cells. Extracellular vesicle-like particles were collected through a series of centrifugations: 1000 g for 20 min, 10,000 g for 30 min and 100,000 g (ultracentrifugation) for 3 h 10 min. Then, 1× SDS-PAGE sample buffer was added to the pellet to make the extracellular vesicle lysate. Lysate samples were heated at 95°C for 5 min and loaded onto an acrylamide gel for SDS-PAGE. Proteins were transferred to PVDF membranes (Millipore, MA, USA), which were blocked with 5% BSA in Tris-buffered saline containing 0.1% Tween-20 (TBST) for 1 h at room temperature. Then, membranes were incubated overnight at 4°C with the following primary antibodies: anti-ARL13B (17711-1-AP; Proteintech, IL, USA; 1:1000), anti-γ-tubulin (GTU88, T6557; Sigma-Aldrich, MO, USA; 1:400), anti-IFT88 (13967-1-AP; Proteintech; 1:5000), anti-IFT52 (17534-1-AP; Proteintech; 1:1000), anti-PKD2 (19126-1-AP; Proteintech; 1:5000) and anti-α-tubulin (F2168; Sigma-Aldrich; 1:1000), which were diluted in TBST containing 1% BSA. After washing in TBST, blots were incubated with HRP-conjugated secondary antibodies for 1 h at room temperature, and chemiluminescence detection was performed using ECL prime (GE, UK). Full blots are shown in Figs S6 and S7.

Time-lapse imaging was performed as described previously (Nakazato et al., 2023b). ARL13B–Venus mIMCD-3 cells were cultured in 35 mm glass-bottom dishes (Iwaki, Japan) with DMEM/Ham's F-12 containing 10% FBS. After 48 h of culture, the medium was replaced with colorless DMEM/F12 medium (Life Technologies, USA) for 24 h. During time-lapse imaging, Z-stack (22 slices of span of 13.86 μm and step size of 0.66 μm) images were captured every 5 min for a total of 3 h using a FV1000 confocal microscope equipped with a 60×, NA 1.35 for FV1000 objective and stage-top incubator (Tokai Hit). Before iso-osmotic or hyperosmotic exposure, five frames of images were acquired, and then the medium for iso-osmotic conditions (≈310 mOsm/l) or hyperosmotic shock (≈610 mOsm/l) was added between the fifth and sixth frame.

Fixed cell samples for TEM were prepared by the following procedures. Cells were first fixed with 2% glutaraldehyde and 4% paraformaldehyde in PBS after the hyperosmotic exposure and stored at 4°C for 2 h, and further fixed with 2.5% glutaraldehyde in 0.1 M phosphate buffer (PB). Fixed cells were washed with 0.1 M PB for 10 min five times at 4°C, and post-fixed with OsO₄ in 0.1 M PB for 30 min at 4°C. The cells were then washed with 0.1 M PB for 5 min twice at 4°C, and subjected to sequential dehydration with 50%, 70%, 90%, 99.5% and anhydrous ethanol for 5 min each. The dehydrated cells were embedded into a resin, TAAB EPON 812, and the resin was polymerized with stepwise increase in temperature (37°C for 24 h, 45°C for 12 h, 60°C for 60 h).

For TEM observation, ultrathin sections of 80 nm thickness were cut at ∼1–1.5 μm above the bottom of the dish to observe centrosomes or from the lateral sides of cells to observe longitudinal sections of primary cilia. Sections were cut using an Ultracut E (Reichert-Jung, Austria) equipped with a diamond knife (DiATOME, Switzerland). The ultrathin sections were subjected to staining with 3% uranyl acetate for 12 min and with lead staining solution (18-0875; Sigma-Aldrich, MO, USA) for 4 min. TEM images were acquired using a JEM-1400 or JEM-1400Plus (JEOL, Japan) TEM with 80 kV of acceleration voltage and a CCD camera.

For array tomography, serial ultrathin sections with 50 nm thickness parallel to the cell bottom were cut using an Artos 3D (Leica, Austria) equipped with a diamond knife (Ultra Jumbo, DiATOME, Switzerland) and were taken up on silicon wafers (Global Top Chemical, Tokyo, Japan). The sections were stained with 1% uranyl acetate for 5 min and with lead staining solution (Sato's Modified Lead staining solution; Hanaichi et al., 1986) for 5 min, then subjected to array tomography using a scanning electron microscope (JSM-IT800, JEOL, Japan) with 5 kV of acceleration voltage and imaged using back-scattered detector. Serial low-magnification images (1000×, 16 nm/pixel) were acquired to locate the primary cilia, which were then re-imaged at higher magnification (100,000× or 50,000×; 1 nm/pixel or 2 nm/pixel, respectively) and serial cross-sectional images of the primary cilia were aligned to determine their spatial architecture.

The ciliary lengths were measured by the Pythagorean theorem by using Z-projected images and Z-slices on which primary cilia propagate. The primary cilia are approximated as straight lines in this method. The statistical significance of each interested pair was evaluated by one-way analysis of variance (ANOVA) on ranks (also known as the Kruskal–Wallis test) with Dunn's multiple comparisons tests using Prism software (GraphPad Software Inc., USA). Single comparisons were analyzed with a Mann–Whitney U test. Paired comparisons were performed with a paired t-test. A P-value of 0.05 was set as the threshold for statistical significance. The results of the immunocytochemical analysis are representative of three independent experiments.

We thank Faryal Ijaz and Madoka Hamada for their technical assistance, and Qushay Umar Malinta for his support with draft writing. A part of this work was conducted using research equipment in the Natural Science Center for Basic Research and Development, Hiroshima University (NBARD-00002) with support from the Ministry of Education, Culture, Sports, Science and Technology (MEXT) Project to promote public utilization of advanced research infrastructure (Program for Supporting Construction of Core Facilities; JPMXS0441300023).

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.261988.reviewer-comments.pdf