Ion channels and calcium signaling in motile cilia

The beating of motile cilia generates fluid flow over epithelia in brain ventricles, airways, and Fallopian tubes. Here, we patch clamp single motile cilia of mammalian ependymal cells and examine their potential function as a calcium signaling compartment. Resting motile cilia calcium concentration ([Ca2+] ~170 nM) is only slightly elevated over cytoplasmic [Ca2+] (~100 nM) at steady state. Ca2+ changes that arise in the cytoplasm rapidly equilibrate in motile cilia. We measured CaV1 voltage-gated calcium channels in ependymal cells, but these channels are not specifically enriched in motile cilia. Membrane depolarization increases ciliary [Ca2+], but only marginally alters cilia beating and cilia-driven fluid velocity within short (~1 min) time frames. We conclude that beating of ependymal motile cilia is not tightly regulated by voltage-gated calcium channels, unlike that of well-studied motile cilia and flagella in protists, such as Paramecia and Chlamydomonas. DOI: http://dx.doi.org/10.7554/eLife.11066.001


Introduction
Cilia are ancient microtubule-based cellular appendages found in eukaryotic organisms (Satir et al., 2008). Motile cilia, like flagella, drive fluid flow via dynein-ATPase action on their 9 + 2 microtubular structure, while most primary cilia are solitary, nonmotile, and lack central microtubular pairs (9 + 0) (Lindemann and Lesich, 2010;Satir and Christensen, 2007). In mammals, almost all cells possess a single primary cilium that houses the Sonic Hedgehog pathway (Drummond, 2012) and mediates aspects of cell-cell signaling. Other nonmotile cilia are found in specialized sensory cells, such as rods and cones of the eye (Satir and Christensen, 2007). In contrast, multiple copies of motile cilia sprout from ependymal cells lining the brain ventricles, and epithelial cells in the airways and Fallopian tubes (Brooks and Wallingford, 2014;Satir and Christensen, 2007). A distinct type of motile cilium (9 + 0) additionally protrudes from cells in the embryonic node during development (Babu and Roy, 2013).
Successful whole-cilia patch clamping of fluorescently-labeled immotile primary cilia revealed nonselective cation currents (PKD2-L1 + PKD1-L1 heteromeric complexes) in primary cilia membranes Delling et al., 2013). Here, we examined ion currents in fluorescently-labeled, voltage-clamped motile cilia of brain ependymal cells and demonstrate that motile cilia are well coupled electrically and by diffusion to the cellular compartment. We show that few Ca V channels are present in the cilia membrane, that resting [Ca 2+ ] is only slightly elevated in motile cilia, and that motile cilia [Ca 2+ ] is driven primarily by changes in cytoplasmic [Ca 2+ ]. Excitation of the ependymal cell by membrane depolarization increases ciliary [Ca 2+ ] with only minor changes in motility and fluid movement, suggesting that beating of ependymal motile cilia is not significantly regulated by the activity of ciliary or cytoplasmic Ca V channels.

Ependymal motile cilia identification and patch clamp
We initially examined ependymal cell GFP-labeled motile cilia from immunolabeled brain sections of transgenic Arl13b-EGFP tg mice . Ciliary localization of Arl13b-EGFP was confirmed by co-staining with the ciliary marker, acetylated tubulin ( Figure 1A). We also observed GFPlabeled motile cilia in primary cultures. At day 10 in vitro (DIV10),~88% of acetylated tubulin stained multiciliated ependymal cells (n = 400 cells) had GFP-labeled cilia (n = 352 cells). Transgene expression varied, with~one-third of the cells exhibiting weak GFP fluorescence in cilia ( Figure 1B, arrow). Some cells had only a few motile cilia (either mono-, bi-or sparsely ciliated) that often displayed moderate motility (~two-three fold slower beat frequency) as compared to neighboring multiciliated cells ( Figure 1B, arrow head). Sparsely ciliated cells may be the result of ependymal cell maturation, in vitro culturing, or represent a population of previously reported biciliated ependymal cells (Mirzadeh et al., 2008). Sparse cilia likely experience different hydrodynamic forces than large eLife digest Certain specialized cells in the brain, airways and Fallopian tubes have large numbers of hair-like structures called motile cilia on their surface. By beating in a synchronized manner, these cilia help to move fluids across the surface of the cells: for example, cilia on lung cells beat to clear mucus away, while those in the brain help the cerebrospinal fluid to circulate.
Motile cilia in mammals are structurally similar to the flagella that propel sperm cells and certain single-celled organisms around their environments. These flagella have specialized pore-forming proteins called ion channels in their membrane through which calcium ions can move. This flow of calcium ions controls the beating of the flagella. However, it is unclear whether a similar movement of calcium ions across the cilia membrane regulates motile cilia beating in mammals.
Doerner et al. have now used a method called patch clamping to study the movement of calcium ions across the membrane of the motile cilia found on a particular type of mouse brain cell. This revealed that unlike flagella, these motile cilia have very few voltage-gated calcium channels; instead, the vast majority of these ion channels reside in the main body of the cell. Furthermore, the level of calcium ions in the motile cilia follows changes in calcium ion levels that originate in the cell body.
Overall, Doerner et al. demonstrate that the activity of voltage-gated calcium channels does not control the beating rhythm of the motile cilia in the mouse brain or how quickly the fluid above the cell surface moves. Future work should investigate whether this is also the case for the cells that line the trachea and Fallopian tubes.
Motility and multiciliation hinder patch clamp recordings from individual cilia and thus recordings were from sparsely ciliated cells. We verified the 9 + 2 structural arrangement of the cilia by imaging the cells on coded, gridded glass dishes and subsequent staining (Video 1). Spag6, a protein associated with the central pair of microtubules (Sapiro et al., 2002), was detected in motile cilia of multiand sparsely-ciliated cells ( Figure 1C,D). Eighty-nine percent of recorded sparsely ciliated cells exhibited Spag6 immunoreactivity in their motile cilia ( Figure 1D; n = 40/45 cells), consistent with the presence of a central pair of microtubules. In contrast, only a small number of primary cilia from Immunolabeling of motile cilia in a section of the lateral ventricle from an Arl13B-EGFP tg mouse. Ciliary localization of Arl13B-EGFP was confirmed using anti-GFP (green) and anti-acetylated tubulin antibodies (red). (B) Anti-GFP (green) and anti-acetylated tubulin staining (red) of cultured ependymal cells at DIV10. GFP labeling of motile cilia varied, with some cells displaying only weak or barely detectable GFP fluorescence in cilia (arrows). Some cells were only sparsely ciliated (arrowhead). (C) Staining of cultured multiciliated cells with anti-GFP (green) and anti-Spag6 (red). (D) Representative staining of a previously recorded ependymal cell grown on a gridded glass bottom dish (grid size, 50 mm, arrow marks motile cilium). Motile cilia of sparsely ciliated cells were GFP (green) and Spag6 (red) positive (n = 40/45). Nuclei were labeled with Hoechst dye (blue, A-D). Panels B-D display average intensity z-projections of image stacks. Scale bars, 10 mm (A-C) and 5 mm (D). (E) Image showing dye diffusion into a motile cilium after successful break-in (50 mM Alexa 594 hydrazide, n = 9). Scale bar, 3 mm. (F) Example current (bottom) recorded in the whole-motile-cilium configuration in response to increasing voltage steps (top). Holding potential, -80 mV. (G) Mean steady state current after break-in plotted as a function of command voltage (n = 4). External solution (aCSF) with 3 mM KCl (black filled squares), 70 mM KCl (blue filled circles), and 140 mM KCl (red filled diamonds). Arrows in the graph indicate calculated E K values. Error bars; ± SEM. DOI: 10.7554/eLife.11066.003 mIMCD cells were (weakly) immunofluorescent in control experiments (8%, n = 9/114, data not shown).
Typically, we patched the bulging tip of a motile cilium. High resistance seals were obtained by positioning the patch electrode in the focal plane at a position near the end of the cilia's stroke, applying suction at the moment the cilium approached the pipette tip (Video 2). Access to a motile cilium was typically achieved by applying a series of brief voltage pulses. Upon successful 'break-in', we were able to monitor dye diffusion from the electrode into the cilium ( Figure 1E). Recording via the whole-motile-cilium yielded a linear current that reversed at hyperpolarized potentials ( Figure 1F,G). These currents were primarily carried by K + (Figure 1G), reminiscent of the ohmic currents reported for ependymal cell bodies (Genzen et al., 2009a;Liu et al., 2006;Nguyen et al., 2001).

Electrical coupling of motile cilia to the cellular compartment
To determine whether ependymal motile cilia are electrically coupled to the cellular compartment, we measured their electrical properties and compared them to whole-cell recordings (Figure 2A,B). Input resistances immediately upon break-in were on average~five-fold higher in whole-cilium recordings (~180 MW, n = 8) than in whole-cell recordings (~36 MW, from the cell body and connected cells, n = 12), reflecting the resistance introduced by the cilium (see below). Uncoupling the cells with flufenamic acid (FFA), an anthranilic acid derivative known to inhibit gap junctions (Harks et al., 2001), decreased the measured conductance in both whole-cell and whole-cilium recordings. Bath substitution with TEA-Cl/BaCl 2 blocked the remaining K + conductances (Figure 2A, B).
To directly assess coupling between motile cilia and cell compartments, we analyzed the capacitive current recorded under voltage clamp in the whole-cell and whole-motile-cilium configuration. In response to a hyperpolarizing voltage step, capacitive current relaxation was 8-10 times faster in whole-cell recordings (0.3 ± 0.1 ms in aCSF/FFA and 0.4 ± 0.1 ms in TEA-Cl/BaCl 2 ) than in wholemotile-cilium recordings (3.1 ± 0.3 ms in aCSF/FFA and 3.3 ± 0.4 ms in TEA-Cl/BaCl 2 ; Figure 2C,D). By integrating the current transient to yield net capacitance, we determined that a similar surface area was charged in whole-cell (20.4 ± 3.2 pF in aCSF/FFA, 18.5 ± 2.5 pF in TEA-Cl/BaCl 2 ) and whole-motile-cilium recordings (18.3 ± 2.4 pF in aCSF/FFA, 19.0 ± 2.8 pF in TEA-Cl/BaCl 2 ; Figure 2E). Calculating the series resistance from the time constant and membrane capacitance, we determined a nine-ten fold higher resistance in whole-motile-cilia recordings (180.3 ± 20.6 MW in aCSF/FFA and 188.2 ± 25.0 MW in TEA-Cl/ BaCl 2 ) as compared to whole-cell recordings (18.0 ± 3.4 MW in aCSF/FFA and 20.2 ± 3.7 MW in TEA-Cl/BaCl 2 ; Figure 2F). Together these findings suggest that motile cilia are indeed electrically coupled to the cellular compartment Video 1. Time lapse of a ciliated cell. Ependymal cells were grown on gridded culture dishes (grid size, 50 mm). Sparsely ciliated cells were recorded, fixed and stained with a central pair marker (Spag6). The example movie corresponds to the staining shown in Figure 1D (same grid via a high resistance cable determined by the length and cross-sectional area of a motile cilium (estimated as~350 MW for an ideally insulated cable; Figure 2H, see Materials and methods). In other words, currents recorded in the whole-motile-cilium configuration can be attributed to channel openings in the cell and/or cilia membrane. Finally, we measured an only slightly depolarized resting membrane potential for motile cilia as compared to the ependymal cell body (-77 ± 3 mV, motile cilia; -88 ± 2 mV, cell body; Figure 2G).

Ca V -mediated currents in ependymal cells
When K + currents were blocked in the TEA-Cl/BaCl 2 bath, we frequently observed relatively small voltage-dependent inward currents ( Figure 2B). Consistent with the finding that cilia and cellular compartments are electrically coupled, we recorded Ca V currents in both whole-cell and whole-cilium configurations ( Figure 3A,B; high series resistance in whole-motile-cilium recordings affected the time course and peak of the calcium current; see Materials and methods). These currents were potentiated by BayK8644 and reduced by nimodipine and CdCl 2 ( Figure 3A,B and Figure 3-figure supplement 1A), consistent with the pharmacology of L-type calcium channels (Ca V 1 subfamily) (Catterall et al., 2005). Indeed, Ca V 1.2 and Ca V 1.3 a subunit transcripts were detected in the cDNA of cultured ependymal cells (DIV10; Figure 3F).
To address the question of whether Ca V currents are in motile cilia membranes, cell body membranes, or both, it would be ideal to detach the cilium from the cell. However, we were unable to detach and record from isolated motile cilia, as we were able to do in primary cilia . Thus, we recorded from cell and motile cilia membranes in the membrane-attached configuration. Prolonged single channel openings in the presence of BayK8644 averaged 1.4 pA at 0 mV (from the record shown in Figure 3C) and were frequently observed in cell-attached recordings, but were typically not detected in cilium-attached recordings ( Figure 3C-E). Despite many attempts, brief single channel openings reminiscent of L-type channel openings in the absence of BayK8644 (Hess et al., 1984) were recorded from only a single motile cilium patch (Figure 3-figure supplement 1C,D). Full-length ciliary recordings (Kleene and Kleene, 2012), were also not feasible given the presence of Ca V channels in the cell membrane. Nevertheless, the low percentage of motile cilium patches in which we observed Ca V channel openings suggests that Ca V channels are not enriched in ependymal motile cilia. If channel densities are identical in cell and cilia membranes, we would only expect~3 channels in a motile cilium (see Materials and methods). These Ca 2+ -permeant ion channels and currents in motile cilia are clearly distinct from the nonselective currents recorded from primary cilia .

Motile cilia [Ca 2+ ] can be modified by cytoplasmic [Ca 2+ ]
To further examine the potential functional consequences of Ca V channel activity, we evaluated motile cilia [Ca 2+ ] by targeting a genetically encoded ratiometric Ca 2+ sensor to these cilia . Previous work established Somatostatin Receptor 3 (SSTR3) transgene expression in ependymal motile cilia (O'Connor et al., 2013). We fused GCaMP6s (Chen et al., 2013) with mCherry to the C-terminus of SSTR3 and transduced cultured ependymal cells with a recombinant adenoviral vector (pAd-mSSTR3-mCherry-GCaMP6s), resulting in abundant expression of the sensor in motile cilia ( Figure 4A). Addition of ionomycin evoked a rapid increase in GCaMP6s fluorescence, confirming the ability of the sensor to report changes in ciliary [Ca 2+ ] ( Figure 4B,C). We calibrated the ratiometric sensor (see Materials and methods) and determined the average ratio of F_GCaMP6s and F_mCherry in motile cilia under basal conditions (aCSF, 1.4 mM extracellular Ca 2+ ; Figure 4D, E). To avoid offsets in the position of beating cilia, we scanned non-sequentially and corrected ratios for bleed-through ( Figure 4-figure supplement 1A,B, and Materials and methods). Consistent with a relatively low number of Ca V channels and a hyperpolarized resting membrane potential (at which Ca V channels are inactive), we determined the resting motile cilia [Ca 2+ ] as 165 nM (average    Figure 4D, E). In control experiments in which immobilized cilia were scanned sequentially, a similar resting [Ca 2+ ] was determined (167 nM, Figure 4-figure supplement 1C, and see below). We conclude that motile cilia resting [Ca 2+ ] is only slightly elevated in comparison to cytoplasmic [Ca 2+ ] (~100 nM) (Clapham, 2007) at steady state, in contrast to the~500 nM elevation observed in primary cilia .
To assess Ca 2+ diffusion between the cell body and motile cilia, we loaded transduced ependymal cells with caged Ca 2+ (NP-EGTA-AM). A brief uncaging stimulus in the cytoplasm rapidly increased ciliary GCaMP6s fluorescence ( Figure 4F). However, imaging motile cilia through the z-axis of the cilia (that is, top or bottom views) prevented further analysis of Ca 2+ diffusion. In order to track Ca 2+ waves, we halted ciliary beating and imaged ependymal cilia from the side ( Figure 4G,H, see Materials and methods). As reported previously, sodium metavanadate (SMVD) treatment greatly reduced motility, presumably through inhibition of the dynein ATPase (Gibbons et al., 1978;Nakamura and Sato, 1993) (Figure 4-figure supplement 1D,E). Upon uncaging at the ciliary base (visualized by additional loading of the cells with the Ca 2+ indicator Oregon Green 488 BAPTA-1 AM, OGB-1), Ca 2+ entered the cilium and moved from the base to the tip with an apparent velocity of 22 ± 3 mm/s ( Figure 4G-I, Video 3). Thus, fluctuations in cytoplasmic [Ca 2+ ] can readily diffuse across the cell-cilia boundary to modify motile cilia [Ca 2+ ], as we found for primary cilia .

Depolarization increases ciliary [Ca 2+ ], but not ciliary beat frequency or fluid velocity
We next asked whether activation of Ca V channels increases ciliary [Ca 2+ ] and regulates ciliary motility in ependymal cells. Differences in the rise times of the two Ca 2+ indicators (OGB-1 and GCaMP6s) precluded a meaningful analysis of signal onsets in motile cilia versus the cytoplasm. The homogeneity of voltage in cell body and cilium, as well as the observed rapid diffusion of Ca 2+ from the cytoplasm into motile cilia, suggests that ciliary [Ca 2+ ] will increase quickly, irrespective of the localization of Ca V channels ( Figure 4). By recording ependymal cells in current clamp, we measured the response to increasing external [K + ] and observed a depolarization of membrane potential from -87 ± 3 mV at rest (aCSF 3 K + ) to -48 ± 5 mV, -31 ± 1 mV, -18 ± 0.7 mV, and -1.7 ± 2 mV in aCSF 20 K + , 40 K + , 70 K + , or 140 K + , respectively (     Previous studies suggest that Ca 2+ increases the beating frequency of motile cilia in airways and Fallopian tubes (Di Benedetto et al., 1991;Girard and Kennedy, 1986;Lansley et al., 1992;Schmid and Salathe, 2011;Verdugo, 1980). Nguyen et al, reported a serotonin-mediated Ca 2+dependent increase in ciliary beating frequency of ependymal motile cilia (Nguyen et al., 2001). We thus examined ciliary beating in response to depolarizing external [K + ] by plotting kymographs of individual cilia (Lechtreck et al., 2008). We observed no significant difference in the beating frequency during perfusion of depolarizing external [K + ] solution (aCSF 70 K + or aCSF 140 K + ). For example, cilia beat frequency was initially 9.7 ± 1 Hz (0-1 s) and was unchanged by perfusion with depolarizing high [K + ] solution (9.7 ± 1 Hz at 15-16 s and 9.3 ± 0.8 Hz at 29-30 s; Figure 5E, see Materials and methods). For comparison cilia beat frequency during continuous perfusion of standard external solution (aCSF 3 K + ) was 9.8 ± 1 Hz (0-1 s), 10 ± 1 Hz (15-16 s), and 10 ± 1 Hz (29-30 s). Kymographs depicting the beat cycles of individual cilia additionally revealed that cilia smoothly transitioned from a forward to a backward stroke independent of external [K + ] ( Figure 5E).
While ciliary beating was unaffected in vitro, we next asked whether activation of Ca V channels by a depolarizing stimulus would disturb fluid movement in acute slices of the lateral ventricles. We assessed the velocity of polystyrene beads (500 nm) in proximity to the apical surface (<30 mm; Figure 5F). In agreement with previous studies, we determined an average velocity of 86 ± 5 mm/s in standard bath solution (aCSF 3 K + ) (Lechtreck et al., 2008). ATP application (100 mM) slightly increased average bead velocity (96 ± 8.4 mm/s), but was not statistically significantly different. In increasing external [K + ], we observed a reduction in average velocity, which was reduced by preincubation with 10 mM nimodopine (79 ± 3 mm/s, aCSF 40 K + ; 67 ± 5.5 mm/s, aCSF 70 K + ; 80 ± 4.3 mm/s, aCSF 70 K + + nimodipine; Figure 5G). Note, however, that the velocity of individual beads varied considerably ( Figure 5G). We cannot rule out an effect of Ca V channels in governing overall motile cilia function, but if it exists, is small and variable.

Discussion
The 9 + 2 structural arrangement of microtubules and inherent dynein ATPase-activity which drives ciliary motion are conserved among motile cilia from protist to humans. However, Ca 2+ modification of cilia function in mammalian ependymal cells is quite distinct from that in protists (Inaba, 2015;Satir and Christensen, 2007). While beat reversal, and changes in waveform critically depend on the activity of ciliary Ca V channels in Chlamydomonas and Paramecium (Fujiu et al., 2009;Kung and Naito, 1973;Matsuda et al., 1998), we found little evidence linking Ca 2+ in ependymal cell bodies or motile cilia to their function. Moreover, unlike primary cilia in which specialized ciliary channels (polycystins) predominate Delling et al., 2013), we found that Ca V channels in the cell body primarily determine changes in motile cilia [Ca 2+ ]. Although we could not accurately quantify the relative Ca V channel densities in motile cilia compared to the cell body or test the possibility of non-uniform channel distribution along the cilium shaft, these channels were not enriched at the ciliary tip. Whether sparsely ciliated cells are less well differentiated or represent a subclass of ciliated ependymal cells is unclear, but the low abundance of Ca V channels revealed by patch clamping of these cilia is consistent with a cytoplasmic control of ciliary [Ca 2+ ] in imaging experiments from multiciliated cells. We thus conclude that ependymal cilia have distinct electrical properties as compared to flagella (cilia) that are used for locomotion in mammalian sperm (Kirichok et al., 2006;Miki and Clapham, 2013;Qi et al., 2007;Ren et al., 2001) and protists (Beck and Uhl, 1994;Dunlap, 1977;Fujiu et al., 2009;Kung and Naito, 1973;Matsuda et al., 1998).
Based on direct measurements of cell body and motile cilia membrane potentials, and voltageactivated Ca V channels in both compartments, we hypothesize that the motile cilium compartment is passively coupled and primarily controlled by channel activity and [Ca 2+ ] in the cell body, rather than being independently regulated. Thus, a cellular response would promote simultaneous signaling to all cilia, perhaps slowly entraining the cilia and enabling signal propagation to neighboring cells (Sanderson et al., 1988;Schmid and Salathe, 2011). Unlike primary cilia, ependymal cilia had only slightly elevated resting [Ca 2+ ], thus enabling smaller changes in cytoplasmic Ca 2+ to equilibrate within cilia. Indeed, membrane depolarization triggered activation of Ca V channels, which resulted in a robust change in ciliary [Ca 2+ ]. While we could not differentiate the onset of the Ca 2+ response in the cilium from that in the cell body, the results suggest that ciliary [Ca 2+ ] follows cytoplasmic [Ca 2+ ].
Previous studies suggest that different sources of Ca 2+ (internal stores, cell membrane flux) can modify the beat frequency of mammalian motile cilia (Di Benedetto et al., 1991;Lansley et al., 1992;Nguyen et al., 2001;Salathe and Bookman, 1995;Schmid and Salathe, 2011). ATP-activated P2X7 receptor-mediated Ca 2+ influx made a minor contribution to the ATP-induced increase in beating frequency of ependymal cilia, while activation of the adenosine A2B receptor made a more significant change, perhaps via slower G protein-dependent pathways (Genzen et al., 2009b). In contrast, we observed that membrane depolarization, or application of ATP, did not significantly alter ependymal motile cilia function. We observed only a slight decrease in bead velocity in the lateral ventricles of acute brain slices that correlated with depolarization and Ca V channel activation. Since we did not detect a change in ciliary beating frequency in in vitro cultures of ependymal cells, we suspect that coupling to other cell types or factors mediates the observed reduction. The variability in velocities of individual beads, however, leads us to question the physiological relevance of the observed changes. Finally, although previous studies suggest that ependymal cell differentiation in vitro reflects in vivo conditions (i.e. postnatal maturation into multiciliated cells [El Zein et al., 2009;Guirao et al., 2010;Spassky et al., 2005]), we cannot exclude the possibility of altered channel expression. Since we did not observe significant changes in ciliary beating in response to membrane depolarization in either acute brain slices or in vitro cultures, we suspect that control mechanisms are similar.
As molecular oars, 9 + 2 motile cilia power fluid flow across the surface of epithelia. Defects in cilia motility are linked to primary ciliary dyskinesia (PCD) and manifest in disease conditions such as chronic sinusitis, male infertility, and hydrocephalus (Ibanez-Tallon et al., 2003;Roy, 2009). Here we have examined the electrophysiological and Ca 2+ signaling properties of motile cilia in ependymal cells. We conclude that they are quite distinct from those of primary cilia Delling et al., 2013) and the flagella that power motile cells. First, membrane potential and [Ca 2+ ] closely follow those in the cytoplasm. This stems naturally from a relative paucity of ion channels in the cilia compared to the nonselective channels in primary cilia . Second, voltage-dependent calcium channels change [Ca 2+ ] in the cytoplasm that propagates readily into the motile cilia, but only marginally alters beat frequency or fluid flow. These findings, although they cannot necessarily be generalized to all motile cilia, suggest that Ca V channels in ependymal cells primarily serve other secretory or differentiation roles in the cell body.

Mice
Transgenic Arl13B-EGFP tg  and C57BL/6 (wild type) mice were used in this study. Animal research protocols were approved by the IACUC of Boston Children's Hospital.

Cell culture Ependymal cell culture
Primary ependymal cell cultures were established as described previously (Delgehyr et al., 2015;Guirao et al., 2010). Briefly, brains of postnatal day 0-3 transgenic Arl13B-EGFP tg mice  or wild type mice were collected, telencephala dissected, the choroid plexus and meninges removed, the tissue enzymatically digested, cells mechanically dissociated, plated and grown to confluence. Confluent cell layers were trypsinized, and replated on poly-L-lysine coated coverslips at a density of 10 7 cells/ml in Dulbecco's modified Eagle's medium (DMEM) with high glucose, GlutaMAX supplement, and pyruvate (Life Technologies), 10% decomplemented fetal bovine serum (FBS, Atlanta Biologicals), and 100 units/ml penicillin/100 mg/ml streptomycin (Atlanta Biologicals), which was replaced by serum-free media the next day. Ependymal cells were cultured for up to two weeks.

Immunostaining
Paraformaldehyde (PFA, 4%) fixed frozen sections of brain ventricles (from a juvenile male) and cultured ependymal cells were permeabilized with 0.1-0.5% Triton X-100, blocked in 5% NGS, 1% BSA, 0.1% Triton X-100, 0.05% Tween20, and incubated with unconjugated Fab fragment goat antimouse (Jackson Immuno Research) to block endogenous mouse IgGs. Primary antibodies (goat anti-GFP FITC conjugate, Novus Biologicals or rabbit anti-GFP Alexa Fluor 488 conjugate, Life Technologies; rat anti-mCherry, Life Technologies; mouse anti-acetylated tubulin, Sigma-Aldrich; rabbit anti-Spag6, Sigma-Aldrich) were applied overnight at 4˚C. Sections and cells were washed in PBS-T and incubated with secondary antibody (Alexa Fluor 568 or 647 conjugates, Life Technologies and Jackson Immuno Research). Nuclei were stained with Hoechst 33342 (Life Technologies) and sections/ cells mounted in Prolong Diamond Antifade (Molecular Probes). Cells were cultured on glass coverslips (VWR) or gridded glass-bottom dishes (Ibidi) for staining. mIMCD3 cells were fixed and stained accordingly and images acquired with settings in the range of the acquisition settings used for Spag6-stained ependymal cells. All images were acquired with a FluoView1000 confocal microscope (Olympus) and processed using ImageJ (NIH).

Data acquisition
Recordings from cultured Arl13b-EGFP expressing primary ependymal cells (cell bodies and motile cilia) were typically performed at 7-14 days in vitro (DIV) after serum starvation at room temperature. Electrodes were pulled from borosilicate glass (Sutter Instruments) with an inner diameter of 0.75 mm or 0.86 mm and a resistance of~16-30 MW or 4-8 MW for motile cilia and cell body recordings, respectively. All data were acquired using an Axopatch 200B amplifier (Molecular Devices). Wholecell and single-channel currents were digitized at 20 kHz and filtered at 5 kHz or 2 kHz, respectively. For capacitive current analysis, currents were digitized at 50 kHz and filtered at 10 kHz. A voltage protocol was applied for motile cilium break-in. Liquid junction potentials were measured using a salt bridge filled with 3 M KCl solution, as described by Neher (Neher, 1992). Voltages were corrected for the liquid junction potential between the bath and pipette solution before acquisition. A 150 mM KCl agar bridge was used in all other recordings. Membrane capacitance and series resistance were not compensated in whole-motile-cilium/whole-cell recordings. Electrodes were routinely Sylgard-coated in single-channel recordings. The image series illustrating seal formation (see Video 2) and the image showing dye diffusion (50 mM Alexa Fluor 594 hydrazide, Life Technologies) from the recording pipette into a motile cilium were captured with an Evolution QEi monochrome camera (Media Cybernetics).

Data analysis
Data were analyzed using Clampfit (Molecular Devices), Origin (OriginLab) and IgorPro (Wavemetrics). Membrane capacitance and series resistance were determined from capacitive currents obtained from hyperpolarizing voltage steps (-80 mV to -100 mV). Briefly, the steady state current was subtracted and the transient current integrated to determine the transferred charge, Q. The membrane capacitance (C m ) was then calculated as C m = Q/4V, where 4V is the size of the voltage step. The series resistance (R series ) was determined from C m and time constant (t) as R series = t/C m .
Voltage-gated calcium channel recordings were analyzed after removal of passive components resulting from depolarizing voltage steps. Three consecutive current responses to a hyperpolarizing voltage step (20 mV or 10 mV in case of nimodipine) were recorded before each voltage pulse. The current responses to the hyperpolarizing pulses were summed, inverted and scaled, and then subtracted from the current response to a given depolarizing voltage step. High series resistance, as observed in whole-motile-cilia recordings, results in slow or incomplete voltage clamp of membranes. Slower V m changes induced Ca V currents with a lag (t = R series C m , see exemplary trace Figure 3-figure supplement 1B), and V m becomes larger than the voltage command, introducing a voltage error that scales with the current amplitude (V m = V p -ÀÀ I m R series ).
The displayed single channel recordings were filtered offline at 1 kHz and capacitive and leak currents subtracted. Histograms were plotted with either all points included or confined to open points (events with fractional amplitude of 0.9).
The total number of channels per cell was calculated as N = I m /P o (i), where I m is the average peak current (recorded from ependymal cells in the presence of BayK8644 at 0 mV; 297.8 pA), P o is the single channel open probability (calculated from a single channel recording in a cell membrane patch by integrating 400 ms voltage pulses over 9 sweeps; 0.32), and (i) is the single channel amplitude at 0 mV (analyzed from the record shown in Figure 3C; 1.4 pA). The surface area of a motile cilium was estimated using the formula for a cylinder: A = 2pr (r + h), where r is the radius and h is the height. If the average length of an ependymal motile cilium is~11.5 mm (Spassky, 2013) and the diameter 0.25 mm, then the surface area of an ependymal motile cilium is~9.1 mm 2 . Since C m is approximately 1 mF/cm 2 in biological membranes, the motile cilium capacitance is~0.091 pF. The average cell capacitance was 18.5 pF (after gap junction uncoupling in TEA-Cl/BaCl 2 bath, see Figure 2E). The number of Ca V channels per motile cilium was estimated by multiplying the total number of channels per cell with the quotient of cilium to cell surface area.
The resistance of a motile cilium as an insulated cable was estimated using the equation R = rl/A, where r is the resistivity (150 WÁcm), l is the length of the cilium (11.5 mm) and A is the cross-sectional area (pr 2 , with r = 0.125 mm). The actual internal resistivity of a motile cilium is unknown. The value for r is based on the assumption that the internal resistivity of a motile cilium is similar to cytoplasmic resistivity.

Adenoviral vector
Expression clones were created by cloning mSSTR3-mCherry-GCaMP6s into pENTR 3C and subsequent LR recombination with the destination vector pAD/CMV/V5-DEST vector (Life Technologies). Glycine-serine linkers were introduced between mSSTR3-mCherry and mCherry-GCaMP6s as described previously . The expression vector was digested with PacI and transfected into HEK293 AD cells. The crude adenoviral stock was further amplified by infection of HEK293 cells. Cultured ependymal cells (derived from wild type pups) were typically transduced with the recombinant adenoviral vector (pAD-mSSTR3-mCherry-GCaMP6s) at DIV6 or 8 and functional expression of the sensor in motile cilia assessed at DIV10 to 12.

Ca 2+ imaging Calibration of the sensor
The calibration curve for the cilium-targeted ratiometric sensor (mSSTR3-mCherry-GCaMP6s) was obtained by incubating adenovirus-transduced ependymal cells in solutions with free [Ca 2+ ] ranging from 50 nM to 10 mM. Digitonin was added at a final concentration of 16 mM to permeabilize the cells and images were acquired with a FluoView1000 confocal microscope (Olympus) for multiple cells after a 5 min permeabilization. The standard solution contained (in mM): 145 NaCl, 5 KCl, 5 EGTA, 10 HEPES, and CaCl 2 ; pH 7.3 with NaOH. MaxChelator (maxchelator.stanford.edu) was used to calculate the total CaCl 2 needed to obtain the desired free [Ca 2+ ]. The acquisition parameters were initially established for GCaMP6s and mCherry fluorescence in the solution containing the highest free [Ca 2+ ] and then held constant across experiments. Non-sequential scan mode was chosen to avoid offsets in the position of beating cilia (basal conditions). Bleed-through of GCaMP6 fluorescence emission into the Cherry signal was subtracted as explained below. Image analysis was performed using ImageJ (NIH) and data plotted using Origin (OriginLab). Images were backgroundsubtracted and a mask created by thresholding the Cherry channel to exclude pixels with no sensor fluorescence. The Cherry mask was then applied to the images with GCaMP6s fluorescence and the resulting image divided by the background-subtracted image of the Cherry channel to obtain ratioed images. Saturated points were subtracted by thresholding the Cherry channel before background subtraction and multiplying it by the ratio image. The mean ratio of F_GCaMP6s and F_mCherry for various free [Ca 2+ ] was measured by averaging the fluorescence intensity of all motile cilia per cell, of multiple cells. To correct for bleed-through, images of HEK293 cells expressing GCaMP6s (in pcDNA3.1) were acquired upon stimulation with 2 mM ionomycin in varying external CaCl 2 concentrations, using the identical acquisition settings as for the acquisition of the calibration curve. Images were background-subtracted and ROIs selected in the cytoplasmic compartment. The GCaMP6s fluorescence was plotted as a function of mCherry fluorescence and the data points fitted linearly. The resulting slope accounts for the bleed-through and was subtracted using the following formula: A c = F_GCaMP6s/F_mCherry -aGCaMP6s, where A c is the corrected average ratio and aGCaMP6s is the slope of the fit. The corrected ratios were then plotted as a function of free [Ca 2+ ] and the data points fitted with the Hill equation. The basal free [Ca 2+ ] in motile cilia was calculated from the average of the corrected fluorescence ratio in standard bath solution (aCSF) and the parameters from the Hill fit of the calibration curve.
imaged from the side as follows: Cells were grown on polyester membrane inserts (Transwell, Costar), the membrane cut, flipped, pinned down, and covered by a self-adhering cover with dual ports for solution exchange (180 ml volume, Electron Microscopy Sciences). Ciliary beating was reduced by continuous perfusion of 100 mM sodium metavanadate (SMVD) for 5 min. Cells with halted ciliary beating were chosen for the experiments. The uncaging stimulus (405 nm, 200 ms laser pulse) was focused to the ciliary base and diffusion measured by scanning a single line from the cytoplasm to the tip of a motile cilium. A few recordings displayed two apparent velocities, presumably related to the different kinetics of OGB-1 and GCaMP6s. In those cases, we analyzed the velocity for the distal part of the cilium. Non-sequential image series were also acquired at a frame rate of 0.254 s.

High speed imaging of motile cilia
Cilia motility and fluid flow were assessed at 60x magnification on an inverted Olympus IX70 microscope. Images were acquired with an OrcaFlash 4.0 high speed camera and HCImageLive imaging software (Hamamatsu).

Analysis of beating frequency
Cultured ependymal cells (DIV12) were imaged at room temperature at 200 frames/s for 2 s in aCSF (see Electrophysiology) pre-and post-incubation with SMVD (100 mM, 5 min). Acquisition areas (typically several per condition and coverslip) were chosen randomly. The beating frequency was analyzed from kymographs (Lechtreck et al., 2008) created in ImageJ using the re-slice function and counting peaks/s. Typically 1-3 cilia were analyzed per acquisition area. For comparison of cilia beat frequencies from cultured multi-and sparsely-ciliated ependymal cells (DIV10), images were acquired at 100 frames/s for 4 s. In experiments in which cilia motility was assessed in response to aCSF 70 K + or aCSF 140 K + , images were captured at a frequency of 100 Hz for 30 s. Gravity-driven bath perfusion was used to continuously perfuse cells with aCSF (controls) or to exchange the bath solution to aCSF 70 K + or aCSF 140 K + . Solution exchange was initiated immediately before image acquisition to capture the transition (5-10 s perfusion delay). Kymographs were plotted for 2 cilia per recording and beating frequencies analyzed at 3 different time points (0-1 s, 15-16 s and 29-30 s).

Analysis of bead velocity
Brain hemispheres of postnatal day 10-14 wild type or transgenic Arl13B-EGFP tg mice  were sectioned with a vibratome (Leica) to 200 mm slices in an ice-cold oxygenated solution containing in mM: 87 NaCl, 25 NaHCO 3 , 1.25 NaH 2 PO 4 , 2.5 KCl, 75 sucrose, 25 glucose, 7.5 MgCl 2 . Slices were stored on ice in an oxygenated solution containing in mM: 87 NaCl, 25 NaHCO 3 , 1.25 NaH 2 PO 4 , 2.5 KCl, 75 sucrose, 25 glucose, 7 MgCl 2 , 0.5 CaCl 2 until use (typically 1-3 h after slicing). Recordings were performed at room temperature in an enclosed 180 ml perfusion chamber (Electron Microscopy Sciences). All aCSF-based solutions (see Electrophysiology) used for brain slice imaging additionally contained 10 mM glucose. Slices were routinely perfused for 2 min with aCSF prior to the perfusion of polystyrene beads (500 nm in diameter, Sicastar-greenF, Micromod) in the respective solution for 30 s. The perfusion was subsequently stopped to record motile cilia driven bead movements along the surface of the lateral ventricle (4 s at 100 Hz). Some slices were perfused with beads in aCSF control solution and, e.g., aCSF 70 K + , and slices washed for 2 min with aCSF between recordings. In experiments with nimodipine, slices were perfused for 2 min with aCSF plus nimodipine (10 mM) prior to bead perfusion in aCSF 70 K + plus nimodipine. Bead velocities were analyzed using the ImageJ manual tracking plugin. Only beads that could be tracked for at least 4 frames were included in the analysis.
A 2-Log DNA ladder (NEB) and PCR samples were loaded on 1% NuSieve 3:1 agarose gels (Lonza) and images acquired with the Syngene GBox system and processed using ImageJ. The PCR reactions for Ca V 1.1 and its controls were run separately and a larger volume (1.5x) was loaded on the gel (weak amplification of Ca V 1.1 transcripts from control tissue).