Primary cilia are not calcium-responsive mechanosensors

Abstract

Primary cilia are solitary, generally non-motile, hair-like protrusions that extend from the surface of cells between cell divisions. Their antenna-like structure leads naturally to the assumption that they sense the surrounding environment, the most common hypothesis being sensation of mechanical force through calcium-permeable ion channels within the cilium¹. This Ca²⁺-responsive mechanosensor hypothesis for primary cilia has been invoked to explain a large range of biological responses, from control of left–right axis determination in embryonic development to adult progression of polycystic kidney disease and some cancers^2,3. Here we report the complete lack of mechanically induced calcium increases in primary cilia, in tissues upon which this hypothesis has been based. We developed a transgenic mouse, Arl13b–mCherry–GECO1.2, expressing a ratiometric genetically encoded calcium indicator in all primary cilia. We then measured responses to flow in primary cilia of cultured kidney epithelial cells, kidney thick ascending tubules, crown cells of the embryonic node, kinocilia of inner ear hair cells, and several cell lines. Cilia-specific Ca²⁺ influxes were not observed in physiological or even highly supraphysiological levels of fluid flow. We conclude that mechanosensation, if it originates in primary cilia, is not via calcium signalling.

Main

To examine ciliary Ca²⁺ signalling in vivo, we generated transgenic Arl13b–mCherry–GECO1.2^tg mice in which fluorescent indicators are confined to the primary cilia (Fig. 1a–c and Extended Data Figs 1 and 2). To accurately measure changes in [Ca²⁺] in small moving subcellular structures and to control for motion artefacts and differential bleaching of ratioing dyes, we first examined hair cells in the cochlea’s organ of Corti, the most sensitive of known Ca²⁺-responsive mechanosensors (CaRMS)^4,5,6. The apical surface of each hair cell carries a hair bundle with staggered rows of specialized actin-based microvillar projections called ‘stereocilia’ (Extended Data Fig. 3). Adjacent to the tallest stereocilia is a single non-motile, microtubule-based true cilium (9 + 2; kinocilium). Fine filamentous ‘tip links’ connect adjacent stereocilia and transmit force to mechano-electrical transduction channels. Mechano-electrical transduction channels open in microseconds, and Ca²⁺ entry rapidly raises [Ca²⁺] in stereocilia (10–20 ms time constant)⁷. Using hair cells from the Arl13b–mCherry–GECO1.2^tg mouse, we can compare [Ca²⁺] changes in two structures (stereocilia and kinocilia) of similar geometry (4–10 μm length and 300–600 nm diameter), and determine whether the kinocilium is also a CaRMS, as suggested⁸.

Figure 1: Genetically encoded Ca²⁺ indicator localizes to primary cilia and cochlear hair cell bundles.

a, P14 kidney section expressing Arl13b–mCherry–GECO1.2 in primary cilia. Cilia (white arrows) point into the lumen of an aquaporin-2-positive tubule (*). Scale bar, 5 μm. b, Aqp2 in primary epithelial cells isolated and cultured (3 days in vitro) from kidney papilla of P14–P21 Arl13b–mCherry–GECO1.2^tg/tg mice (Extended Data Fig. 1). Scale bar, 10 μm. c, Embryonic node from an Arl13b–mCherry–GECO1.2^tg mouse. mCherry and GECO1.2 in cilia of the embryonic node overlap with the cilia marker, acetylated tubulin. Scale bar, 10 μm. d, Arl13b–mCherry–GECO1.2 expression in IHCs. P4 organ of Corti explant showing two hair bundles; mCherry fluorescence (red), phalloidin-labelled actin (blue), and antibody to acetylated tubulin (green). Arl13b–mCherry–GECO1.2 protein localizes to all kinocilia as demonstrated by overlapping acetylated tubulin staining (arrows). Later in development, Arl13b–mCherry–GECO1.2 also localizes to stereocilia bundles of some IHCs and OHCs (mCherry-positive bundle, arrowhead; Extended Data Fig. 1). Scale bar, 5 μm. e–g, Deflection of IHC stereocilia bundle. Region of interest (ROI): white outline. e, F_GECO1.2 and F_mCherry in stereocilia before, during, and after deflection. A 1 s flow stimulus (Supplementary Video 1) increased F_GECO1.2 as Ca²⁺ entered the shorter rows of stereocilia. Scale bar, 5 μm. f, Average F_GECO1.2 and F_mCherry intensities (in arbitrary units) within the ROI; bundle displacement during deflection (black). Deflection increased F_GECO1.2 threefold; mCherry bleached only slightly. g, Change of stereocilia F_GECO1.2/F_mCherry during the deflection (blue symbols); bundle displacement, black. After stimulus, the bundle rapidly returned to its resting position while Ca²⁺ fluorescence, F_GECO1.2, decayed slowly (τ = 0.6 s; Extended Data Fig. 4a, b). Images acquired sequentially, 30 ms per paired frame. All images are representative of more than ten images taken of biological triplicates.

PowerPoint slide

Arl13b–mCherry–GECO1.2 is expressed in all primary cilia of the inner ear sensory epithelium, including kinocilia of both inner hair cells (IHCs) and outer hair cells (OHCs) of the organ of Corti. Scanning electron microscopy at ages embryonic day (E)18 to postnatal (P)3 confirmed normal development of stereocilia bundles in transgenic animals (Extended Data Fig. 3). Fortuitously, some hair cell stereocilia also contained the Ca²⁺ indicator (Fig. 1d and Extended Data Fig. 1d). To test for CaRMS, we applied laminar flow via a micropipette to IHC bundles (Supplementary Video 1) or kinocilia and imaged via swept field confocal microscopy (up to 1,000 frames per second) to pinpoint the origin of Ca²⁺ influx with high spatial and temporal resolution. Ratioing of GECO1.2 and mCherry fluorescence (F_GECO1.2/F_mCherry) reduced movement-related artefacts (see Methods and Supplementary Information). As expected, deflection of the stereocilia bundle resulted in a rapid and robust increase in F_GECO1.2 within the bundle, while F_mCherry remained largely unchanged (Fig. 1e–g and Supplementary Video 1). IHC bundle deflection increased the ratio approximately threefold (Fig. 2a), slowly recovering to the resting value within 2.5 s after flow application. The slow dissociation rate of Ca²⁺ from GECO1.2 (τ ≈ 600 ms; Extended Data Fig. 4) provides a valuable means of distinguishing true changes in [Ca²⁺] from movement-related artefacts. In contrast, kinocilia deflected by pressure steps over a range of developmental stages (E14 to P3) exhibited no detectable increase in [Ca²⁺]_cilium (Fig. 2b–f and Supplementary Information). We conclude that kinocilia of mouse hair cells, unlike actin-based stereocilia, are not CaRMS.

Figure 2: No change in [Ca²⁺]_cilium in kinocilia of developing hair cells.

a, Average ratio changes in stereocilia for IHC bundle deflections in P5 + 3 days in vitro organ of Corti explants (n in brackets). b, Average ratio changes for IHC kinocilia (red, n = 36), OHC kinocilia (green, n = 38), and supporting-cell (SC) primary cilia (blue; n = 85). The small positive slope results from differential dye bleaching (mCherry > GECO1.2). c, Normalized average fluorescent ratio changes at times boxed in a and b. Digitonin applied locally to permeabilize the membrane evoked an approximately fivefold ratio increase in hair cell kinocilia and SC primary cilia (brown bars). Black bar, normalized ratio before application; blue bar, during application; red bar, after application. IHC bundle deflection increased stereocilia ratios approximately threefold, persisting well after the bundle returned to its initial position. E14 and E15 explants showed no ratio change in any organ of Corti (OC) primary cilia. d, Kinocilia of IHC, OHC, and SC, from E17 to P3, show no ratio change upon deflection. The slight variability of the ratios from the SC cilia (before, during, after) is similar to the variability of the ratios in IHC and OHC kinocilia. e, f, Individual IHC (e) and OHC (f) kinocilia at P0. F_GECO1.2 (green) and F_mCherry (red) intensities (in arbitrary units) and their ratio (blue) show no Ca²⁺ influx upon deflection. Kinocilium displacement in grey (bottom panel). Student’s t-test, *P < 0.05, **P < 0.01, #P < 0.001. All error bars ± s.e.m.

PowerPoint slide

Next, we examined kidney epithelia primary cilia, widely believed to be CaRMS^1,3,9. Primary inner medullary collecting duct (mIMCD) epithelial cells were isolated from kidneys of P14–P21 Arl13b–mCherry–GECO1.2^tg/tg mice, which express the kidney collecting duct epithelial cell markers, Aquaporin2 (Aqp2; Fig. 1b), and PC2 (Extended Data Fig. 1). In vivo, proximal tubule flow velocities are ~300 μm s⁻¹ (ref. 10). To quantify these forces, we used a flow chamber to measure cilia bending at defined plasma membrane shear stress values in cultured mIMCD cells isolated from Arl13b–mCherry–GECO1.2^tg/tg mice (Methods and Extended Data Fig. 5). Flow velocities ranging from 3 to 400 μm s⁻¹ (shear stress ≈ 0.002–1 dyn cm⁻²) bent cilia (Fig. 3a, b) with half maximal bending at 70 μm s⁻¹ (~0.11 dyn cm⁻²). A similar cilium-bending profile is observed in modelling the cilium as a uniform cylindrical cantilevered beam^11,12.

Figure 3: No change in [Ca²⁺]_cilium during mechanical stimulation of kidney primary cilia.

a, mIMCD primary cilium at eight flow velocities (Extended Data Fig. 5). Scale bar, 5 μm; representative of 14 cilia (8 z-stacks, 12 focal planes each). b, Cilium deflection as a function of flow velocity and shear stress (n = 14). c, Cultured primary mIMCD cells from P14–P21 Arl13b–mCherry–GECO1.2^tg/tg mice imaged from above (top panels, 87 videos, 150 frames each) or the side (bottom, 32 videos × 150 frames); pipette ~4–6 μm from cilium (Supplementary Video 2). Images: before, during, after 2 s, ~250 μm s⁻¹ flow velocity stimulus. A MATLAB tracking algorithm identified the ROI frame-by-frame and quantified F_GECO1.2 and F_mCherry. Scale bars, 2 μm; d, Channel intensities (in arbitrary units) from top panel in c. When flow flattened cilia, average F_GECO1.2 and F_mCherry signals decreased owing to changes in the light path. e, Ratioing F_GECO1.2 to F_mCherry reduced artefact (Extended Data Fig. 6), revealing no change in [Ca²⁺] during deflection. f, Side view fluorescence of bottom panel in c; F_GECO1.2 and F_mCherry were relatively constant. g, Side-imaged F_GECO1.2/F_mCherry ratio unchanged by deflection. h, Normalized F_GECO1.2/F_mCherry for positive controls (IHC stereocilia bundle deflection; data from Fig. 2c, and digitonin application); and for mIMCD primary cilia deflections in high and low [Ca²⁺], n in brackets. Small ratio differences are due to motion artefact (Supplementary Information and Extended Data Fig. 6). i, Perfusion of acutely dissected kidney thick ascending limb tubules (representative of four preparations) (Extended Data Fig. 5). Green, F_GECO1.2 and cytoplasmic autofluorescence; red, F_mCherry. A train of increasing 1 s pressure steps deflected intratubular cilia (arrowheads; Supplementary Video 5). Right, F_mCherry overlay from images before (cyan) and during (magenta) cilia deflection. Scale bars, 5 μm; 16 videos × 200 frames (middle). j, No change in [Ca²⁺]_cilium for the smallest displacement in i. k, No change in [Ca²⁺]_cilium before, during, and after deflection, data from j (1.0 ± 0.06 before versus 0.99 ± 0.06 during, n = 16 cilia). Student’s t-test, #P < 0.001, NS, not significant. Error bars ± s.e.m.

PowerPoint slide

Primary cilia viewed from above were fully deflected by ~250 μm s⁻¹ flow velocity (Supplementary Video 2 and Fig. 3). Fluorescence intensities immediately dropped as flow flattened the cilium (Fig. 3c, d). Ratioing F_GECO1.2/F_mCherry reduced this motion artefact (Fig. 3e and Extended Data Fig. 6). Imaging cilia from the side was most effective in reducing position/motion/path length artefacts; in this configuration, all parts of the primary cilium are in the same focal plane and never overlie the cell’s variable autofluorescence (Fig. 3c, bottom, Supplementary Video 2 and Fig. 3f). Again, no change in [Ca²⁺]_cilium was detected during deflection (Fig. 3g, h; digitonin control, Extended Data Fig. 6). Extending our study to other cell types, we found no CaRMS in osteocyte-like cells, mouse embryonic fibroblasts, or, indeed, any primary cilia examined (Extended Data Fig. 7). Instead, we found that in all cases where [Ca²⁺]_cilia increased, the Ca²⁺ rise was initiated at other sites in the cell and diffused from the cytoplasm into the cilium (Extended Data Figs 8 and 9 and Supplementary Video 3). At low image acquisition rates (>200 ms per frame), this could easily be misinterpreted as cilia CaRMS^3,9,13,14,15. At very high, non-physiological flow velocities and shear stresses (peak, 10 dyn cm⁻¹ (refs 2, 16)), we observed that some cilia tips were ripped from the axoneme (Extended Data Fig. 9 and Supplementary Video 4) and external 2 mM Ca²⁺ filled the breached cilium¹⁷.

In intact kidney tubules, primary cilia are deflected in a pulsatile pattern¹⁸. We microdissected kidney tubules from P21 mice and deflected primary cilia inside isolated tubules (Extended Data Fig. 5). A train of 1 s stepped-amplitude flow stimuli in the lumen facilitated determination of the minimal flow deflecting cilia. Even flow sufficient to fully deflect cilia, however, did not change [Ca²⁺]_cilium. Digitonin control increased Ca²⁺ ratios approximately sixfold (Fig. 3i–k and Supplementary Video 5).

During embryonic development, structures that are initially symmetric along the body axis develop a left–right orientation. This orientation depends on cilia in the embryonic node, a depression at the distal tip of the embryo¹⁹. Specialized motile cilia (Supplementary Video 6) in the central node direct fluid leftwards (viewed dorsally) at ~2–4 μm s⁻¹, which is believed to initiate asymmetric gene expression across the node²⁰. Disruption of flow or mutations in primary cilia proteins result in left–right patterning defects²¹ such as situs inversus, while imposed fluid flow to the right side reverts left–right patterning^22,23. A prominent hypothesis is that left-side crown cell primary cilia are deflected by this flow, which mechanically activates ion channels in primary cilia and raises intracellular [Ca²⁺], leading to altered signalling and gene transcription in those cells and adjacent lateral plate mesoderm^2,24. An alternative hypothesis is that the directed beating of nodal cilia creates a gradient of vesicles²⁵ or a secreted morphogen²⁶.

To test the mechanosensitive model, we measured [Ca²⁺] during deflection of left- and right-side nodal primary cilia (Fig. 4a, b, k, Extended Data Fig. 10 and Supplementary Videos 7 and 8). Cilia from developmental stages ‘early allantoic bud’ to two-somite were mechanically stimulated by applying flow using physiological levels (ramp from 0 μm s⁻¹ to ~10 μm s⁻¹), or supraphysiological flow velocity of ~200 μm s⁻¹, similar to that previously shown to revert embryonic organ symmetry (Fig. 4b, c)²². Deflection of left- or right-sided nodal cilia did not change [Ca²⁺]_cilium (Fig. 4d–i). These results demonstrate that primary cilia of the embryonic node are not CaRMS from physiological flow velocities to those 50–100 times greater. Notably, physiologically relevant flow velocities (less than ~10 μm s⁻¹) barely deflected nodal crown cilia (458 ± 22 nm; Fig. 4k–p, Extended Data Fig. 10 and Supplementary Video 9) as predicted (Fig. 3a, b). Resting [Ca²⁺] was ~300 nM both in left- and in right-side cilia (Fig. 4j; for cytoplasmic calcium see ref. 23).

Figure 4: No change in [Ca²⁺]_cilium in primary cilia of the embryonic node.

a, Arl13b–mCherry–GECO1.2^tg embryonic node, early headfold stage (EHF); L, left; R, right. Non-motile cilia; left, white arrowheads; right, black arrows. Scale bar, 20 μm; n = 4 embryos. b, Crown-cell cilia deflected by 2 s/~200 μm s⁻¹ flow velocity (2 videos × 500 frames). Scale bar, 5 μm. c, Cilia from the left outer perimeter before (left), during (middle) and after (right) ~200 μm s⁻¹ flow velocity. Top row, merged GECO1.2 and mCherry signals. White outline, ROI (32 videos × 150 frames). d, e, F_GECO1.2 and F_mCherry (in arbitrary units) of a left (d) and right (e) nodal cilium; ~200 μm s⁻¹. f, g, Ratiometric changes on left (f) and right (g) sides; ~200 μm s⁻¹ cilia deflection (black trace; same cilia as d, e). h, Average relative ratio changes for cilia on the left (burgundy, n = 29) and right (green, n = 15) sides of the node. i, Relative ratio changes; grey boxes in h, time blocks quantified. j, Resting [Ca²⁺] in left- and right-side non-motile primary cilia. [Ca²⁺] calculated as in Extended Data Fig. 2 and Methods (left cilia: 1.0, n = 67, 4 nodes; right cilia: 1.1, n = 72, 5 nodes); resting [Ca²⁺] left: 302 ± 6 nM; right: 320 ± 7 nM. k, l, Cilia on the left outer perimeter (dashed white line) with no net flow (k) or ~10 μm s⁻¹ flow velocity (l) measured near the cilium (arrow). Coloured lines, tracks of beads (Supplementary Video 9). Scale bar, 5 μm; 42 videos × 150 frames. m, Traces of cilium in k (arrow). Displacement of cilium during 0–10 μm s⁻¹ ramp in purple. n, Cilium in k (arrow) at ~0 μm s⁻¹ (left, frame average), ~10 μm s⁻¹ (middle, frame average, end of deflection), and superposition (right, note <1 μm deflection). Asterisk, ciliary base. Scale bar, 500 nm; 42 videos × 150 frames. o, p, Ratio changes of left crown-cell cilia from late bud to three-somite stage during flow from ~0 to ~10 μm s⁻¹ velocities. o, Ratio change for non-displaced cilia (n = 26 cilia, 5 embryos), p, Ratio change for displaced cilia (n = 71 cilia, 5 embryos; average centroid displacement, 458 ± 22 nm). All error bars ± s.e.m.

PowerPoint slide

In summary, mechanical forces do not evoke Ca²⁺ signalling within up to 10 s of stimulation in the cilia of kidney tubules, the embryonic node, several cell lines used as models of cilia function, or kinocilia. Since mechanosensitive channels and putative mechanically gated G-protein-coupled receptors respond in the 10 μs to 100 ms range^27,28, we suspect the reported increases in [Ca²⁺]_cilium after ~10–20 s stimulation^3,15 originate from the cell body. Indeed, imaging up to 100 Hz revealed that non-ciliary origins were always the initial site of Δ[Ca²⁺] (Extended Data Figs 8 and 9 and Supplementary Video 3) and the Ca²⁺ wave initiated in the cytoplasm propagated into the primary cilium. To account for previous observations, we speculate that the two primary sources of error were (1) insufficient time resolution—Ca²⁺ originating from the cytoplasm can diffuse into the cilium in <200 ms and be mistaken as originating from within the cilium, and (2) motion- and light-path-dependent artefacts. Finally, we do not imply that there are no mechanosensitive elements in primary cilia in physiologically relevant ranges, since motor, cytoskeletal, and other proteins are affected by changes in geometry and force without changes in ciliary [Ca²⁺]²⁹. An important implication of these findings is that situs inversus and polycystic kidney disease caused by the loss of polycystins^23,30 are not due to loss of mechanically induced cilia-initiated calcium signalling. These data should motivate investigators to focus on other potential mechanisms for regulation of ciliary ion channels, and to determine whether calcium propagation into the cilium from the cytoplasm affects ciliary function.

Methods

The experiments were not randomized. The investigators were not blinded to allocation during experiments and outcome assessment.

Molecular biology, transgenic animals

The ScaI/HindIII linearized hArl13b–mCherry–GECO1.2 pCAG vector (chicken actin promoter) was gel-purified and injected into the pronucleus of C57BL6/6J oocytes at the transgenic animal core facility at Boston Children’s Hospital (Boston, Massachusetts). The integration site for the Arl13b–mCherry–GECO1.2 transgene was determined by genomic walking (Bio S&T). The genotype of transgenic animals was determined by PCR: primers 372-up, ACATGGCCTTTCCTGCTCTC; 372-down, TTCAACATTTCCGTGTCGCC; and 944-down, GACATCTGTGGGAGGAGTGG. PCR product for the wild-type genomic sequence: ~600 bp; transgene PCR product ~400 bp. All animal procedures of this study were approved by the IACUCs of Boston Children’s Hospital and Harvard Medical School (Boston, MA). Animals were maintained according to ARCH standards at Boston Children’s Hospital and euthanized using CO₂.

Isolation of mIMCD

Arl13b–mCherry–GECO1.2^tg: mIMCD cells were isolated as described³¹. Briefly, ten kidneys isolated from P14–P21 Arl13b–mCherry–GECO1.2^tg mice were cut longitudinally with fine scissors and the outer and inner medulla removed. The tissue was cut into small pieces with a razor blade and digested in collagenase (2 mg ml⁻¹) and hyaluronidase (1 mg ml⁻¹) for 1 h at 37 °C in L-15 medium (Life Technologies). After trituration of the homogenate, cells were washed twice in phosphate-buffered saline (PBS) and plated on laminin-coated dishes (Life Technologies). Cells were grown in DMEM (adjusted to 600 mOsm with urea and NaCl), containing 200 μM dibutyryl-cAMP (db-cAMP), unless stated otherwise. After 2 days, cells were split on laminin-coated coverslips (NeuVitro) and imaged after culturing for an additional 1–2 days to allow confluent cell growth. For side-view imaging, cells were grown on 24 mm Transwell inserts (Corning) until they reached confluency. The membrane was excised with a scalpel and folded before imaging. Where indicated, mIMCD cells were serum starved in DMEM containing 0.2% BSA for 24 h or 48 h. Imaging solutions: L-15 medium (1.3 mM Ca²⁺) or HEPES-containing solution buffered to 50 nM [Ca²⁺] (see ‘Calibration of the ratiometric Arl13b–mCherry–GECO1.2 sensor’ for buffer composition).

Microdissection of Arl13b–mCherry–GECO1.2^tg kidney tubules

Microdissection was performed as described previously³² with modifications. In brief, 1-mm thick transverse slices of P14–P21 kidneys were incubated with collagenase (1 mg ml⁻¹) and hyaluronidase (1 mg ml⁻¹) for 30 min at 37 °C in L-15 medium (Life Technologies), or gently dissected without prior treatment. The cortex was removed with fine forceps and bundles of tubules were isolated at the transition of inner (white) and outer (red) medulla (Extended Data Fig. 5c). Thick-walled individual tubules with luminal fluorescent cilia were microdissected and mounted on glass or plastic coverslips coated with Cell-Tak (Corning). Under ×4 magnification (upright Nikon NiE), a micromanipulator-mounted 20° micro-knife (Minitool) was used to cut individual tubules from the bundles. A second micromanipulator held a long-tapered micropipette (bent ~20° to ensure the tip of the pipette was parallel to the surface of the coverslip, Extended Data Fig. 5). Under higher magnification (×100/1.1 numerical aperture or ×60/1.0 numerical aperture water dipping lenses), the micropipette was gently inserted into the tubule lumen and the pressure stimulus applied. Regions of the tubule with no direct micropipette contact were used for Ca²⁺ imaging. In some experiments a third micromanipulator was used to deliver digitonin (20 μM) to the tubules (direct injection into the tubule lumen or external application). Cilia from kidney tubule perfusion experiments were collected from three independent microdissections.

Isolation of Arl13b–mCherry–GECO1.2^tg and Arl13b–mCherry–GECO1.2^tg:GCaMP6f^tg:E2aCre^tg embryos

GCaMP6f (B6;129S-Gt(ROSA)26Sor^{tm95.1(CAG-GCaMP6f)Hze}/J) and E2a-Cre (Tg(EIIa-cre)C5379Lmgd) transgenic animals were obtained from Jackson Laboratories. Embryo isolation was performed as described previously³³. Timed pregnancies resulting from mating wild-type C57BL6/6J, Arl13b–mCherry–GECO1.2^tg/- or Arl13b–mCherry–GECO1.2^tg/tg females with Arl13b–mCherry–GECO1.2^tg/tg or GCaMP6f^tg/tg:E2a-Cre^tg/tg males yielded embryos that were then selected for the appropriate developmental stages³⁴. Embryos expressing motile cilia in the embryonic node at stages critical for asymmetric gene expression³⁵ (starting at developmental stages ‘early allantoic bud’ up to two-somite stage) were used for experiments. Embryos were mounted with the embryonic node facing up in a custom-designed embryo mounting plate (Extended Data Fig. 10b–d). Laser cut holes (diameters 0.5–1.2 mm) in 0.8 mm Delrin ensured a good fit of the embryo into the holding well (Extended Fig. 10b, c). All embryonic node imaging was performed in DMEM/F12 with 10% fetal calf serum (Invitrogen).

A similar mating strategy was used to obtain Arl13b–mCherry–GECO1.2^tg:GCaMP6f^tg:E2a-Cre^tg E14 embryos. Mouse embryonic fibroblasts were isolated from E14 embryos as described previously¹⁷. Where indicated, MEF cells were serum starved in DMEM containing 0.2% BSA for up to 48 h. To visualize cytoplasmic Ca²⁺ oscillations, Arl13b–mCherry–GECO1.2^tg: GCaMP6f^tg:E2a-Cre^tg embryos from late allantoic bud to late headfold stage were used. In brief, embryos were mounted in the upright position as described above and imaged for 4–6 min at a frame rate of 0.5 Hz on an upright FV1000 confocal system (Olympus, ×60/1.1 numerical aperture water dipping lens) at either 36 °C or 22 °C (room temperature). Cytoplasmic Ca²⁺ oscillations were quantified using ImageJ as described previously³³ with slight modifications: in brief, fluorescence of all frames was averaged and individual frames were divided by average intensity to generate ΔF/F. Images were thresholded to exclude cells with ΔF/F less than 30%. Furthermore, only regions with area greater than 90 square pixels and circularities greater than 0.6 were used to define cells with cytoplasmic Ca²⁺ oscillations. All Ca²⁺ oscillations within 50 µm surrounding the embryonic node were analysed for occurrence on the left versus the right side of the node.

Reagents

The following reagents were used: mouse anti-acetylated tubulin (Sigma-Aldrich; T7451); CF 405M phalloidin to stain filamentous actin (Biotium; 00034); goat anti-PC2 (Santa Cruz; G-20 sc-10376), rat anti-mCherry (Life Technologies; M11240), chicken anti-EGFP (Aves Labs; GFP-1020).

Immunocytochemistry, confocal microscopy

Cells or embryos were fixed with 4% paraformaldehyde, permeabilized with 0.2% Triton X-100, and blocked by 10% donkey serum in PBS. Cells were labelled with the indicated antibody followed by secondary donkey anti-rabbit, anti-goat, or anti-mouse fluorescently labelled immunoglobulin-G (IgG) (Life Technologies) and Hoechst 33342 (Life Technologies). Confocal images were obtained using an inverted Olympus FV1000 (×60 1.2 numerical aperture water immersion objective lens) and images processed using ImageJ (NIH).

Immunohistochemistry

Fixed (4% paraformaldehyde) 15 μm frozen tissue sections were permeabilized with 0.5% TX100/PBS (pH 7.4) for 15 min and blocked with PBS containing 5% goat serum, 1% BSA, 0.1% fish gelatin, 0.1% TX-100, and 0.05% Tween20. For primary antibodies raised in mice, endogenous mouse IgGs were blocked by incubating sections with the unconjugated Fab fragment goat anti-mouse IgG for 1 h at room temperature. For goat primary antibodies, donkey serum and donkey secondary antibodies were used. Sections were washed twice in PBS-T, and incubated with primary antibodies in blocking solution overnight at 4 °C. Slides were washed twice in PBS-T and goat anti-rabbit/anti-mouse fluorescent-labelled secondary antibodies applied at room temperature for 1 h with Hoechst 33342 (nuclear) dye. Sections were washed twice in PBS-T, mounted in Prolong Gold Antifade (Life Technologies), and imaged (inverted Olympus FV1000; ×60, 1.2 numerical aperture water immersion objective). Images were further processed with ImageJ (NIH).

Ca²⁺-imaging in stereocilia and primary cilia

Arl13b–mCherry–GECO1.2-expressing cilia were observed under an upright Nikon NiE microscope (×100, 1.1 numerical aperture, 2.5 working distance) equipped with an Opterra swept-field confocal imaging system (Bruker Nano Technologies) and a Photometrics Evolve 128 liquid-cooled EMCCD camera (128 pixels × 128 pixels, 120 nm effective pixel size). This system enables fast imaging of up to 500–1,000 frames per second (f.p.s.) in low-light conditions. In most cases, tissue was illuminated sequentially by 488 nm (GECO1.2) and 561 nm (mCherry) laser light, and imaged using the full CCD chip (15 ms exposure per channel; 33 f.p.s.). To increase light delivery to the camera and avoid excessive photobleaching, swept-field confocal imaging was performed in slit mode (35 μm).

Measurements of cilia bending and flow velocities

mIMCD and MEF cells isolated from Arl13b–mCherry–GECO1.2^tg mice were seeded in an IBIDI μ-Slide VI 0.4 flow chamber coated with laminin (see IBIDI Application Note 11; for this chamber, apical membrane shear stress is τ = η ×131.6 × Φ, where η is dynamic viscosity (0.01 dyn s⁻¹ cm⁻²) and Φ is flow rate in millilitres per minute. A syringe pump (Harvard Apparatus) delivered steady flow via 10 ml syringes containing L-15 medium. Z-stacks of primary cilia were recorded on an inverted Olympus FV1000 (×60, 1.2 numerical aperture water immersion objective). The bend angle was measured between ciliary base and tip¹⁸. Fluid velocities were measured by imaging the flow of the solution supplemented with 300 nm green fluorescent beads (Sicastar greenF, Micromod) at the focal plane corresponding to ciliary tips at rest. Images were acquired as line scans (2 ms per line) or in continuous scanning mode (64 or 128 ms per frame) and particles tracked using an ImageJ plugin.

Calibration of the ratiometric Arl13b–mCherry–GECO1.2 sensor

Calibration was performed using an inverted Olympus FV1000 (×60, 1.2 numerical aperture water immersion objective) as described previously¹⁷. In brief, standard solutions of [Ca²⁺] (ranging from 50 nM to 50 μM) were prepared by adjusting the ratio of EGTA and CaCl₂ (MaxChelator) in 137 mM NaCl, 5.4 mM KCl, 10 mM HEPES. After isolation, mIMCD cells were plated onto 12 mm laminin-coated glass coverslips (Neuvitro) and cultured for 3–4 days to allow cilia formation. For controls, digitonin membrane permeabilization (3 min) was followed by image acquisition in multiple fields of view. Ratios were obtained by dividing the average F_GECO1.2 (per ROI, corresponding to a single cilium) by the average F_mCherry. The average ratios were plotted as a function of free [Ca²⁺] fitted to a sigmoid curve: y = A₂ + (A₁ − A₂)/{1 + exp[(x − x₀)/dx]}, where A₁ = 0.15 ± 0.03, A₂ = 1.58 ± 0.06, x₀ = 442 ± 25, dx = 114 ± 20, with x₀ = K_d (the dissociation constant). The K_d of bacterially expressed/purified GECO1.2 was 1.1 μM (ref. 36), about twice the K_d measured for our GECO1.2 fusion construct in mammalian cells in situ. [Ca²⁺] in embryonic node primary cilia was estimated from an R_min- and R_max-adjusted calibration curve, where both values were calculated from images collected on swept-field confocal imaging system. Late bud to late headfold embryos were permeabilized with 20 μM digitonin (5 min) in either 50 nM or 5 μM [Ca²⁺]. Resting values for nodal primary cilia were measured in DMEM/F12 + 10% FCS using the same imaging settings used to calculate R_min and R_max. Ratios were converted to [Ca²⁺] using the adjusted Ca²⁺ calibration curve.

Primary cilia and stereocilia deflections

Primary cilia and stereocilia bundle deflections were performed using a custom-made fluid-jet system. Briefly, the micropipette pressure at the back of the pipette could be rapidly changed to a desired value by supplying vacuum and/or pressurized air via feedback-controlled solenoid valves (5–10 ms rise time for the pressure step stimulus). The micropipette was filled with bath solution and the pressure at the mouth of the pipette carefully adjusted before approaching to the cilium, ensuring that there was no flow applied to the cilium before the onset of the stimulus. Depending on the experimental design, digitonin was applied to the cells either using the fluid-jet pipette, or via an additional pipette positioned near the cilium and connected to an IM-9C microinjector (Narishige).

For kinocilium deflection experiments, organ of Corti explants were acutely dissected and mounted on a coverslip coated with CellTak, or immobilized with tungsten minutien pins (FST). All hair cell imaging experiments were performed at room temperature in L-15 cell culture medium (Invitrogen), containing in mM: NaCl (138), KCl (5.3), CaCl₂ (1.3), MgCl₂ (1.0), Na₂HPO₄ (1.0), KH₂PO₄ (0.44), MgSO₄ (0.81). For stereocilia bundle Ca²⁺ imaging experiments, organ of Corti explants were dissected at P5 in L-15 medium and placed in culture in DMEM/F12 supplemented with 5% FBS and 10 mg l⁻¹ ampicillin at 37 °C (10% CO₂). Explants were cultured for 3 days to increase the number of cells with sufficient sensor in stereocilia bundles.

All images were analysed using a custom-made MATLAB tracking algorithm (described below), ImageJ (NIH), and Origin 8 (OriginLab).

Osteocyte-like cell lines MLO-Y4 and Ocy454 (ref. 37) were tested for authenticity in the laboratories that supplied them (see Acknowledgements) and for mycoplasma contamination by our laboratory. Both cell types were transfected with a plasmid encoding Arl13b–mCherry–GECO1.2 using electroporation (LONZA, solution V, program T-20), as described previously³⁸. Cells were seeded on coverslips after transfection and cultured at 37 °C. Cells were used after a primary cilium was visible.

Embryonic node cilia were deflected with either a fast fluid-jet stimulus (described above) or with a ramp of slow, physiological level flow, delivering up to ~10 μm s⁻¹ velocities: fluid flowed from a gravity-fed open-ended syringe to the micropipette. Flow rates were calibrated using 100 nm fluorescent beads (Sicastar-greenF, Micromod) and adjusted by gently lifting the syringe up 5–10 mm using a coarse micromanipulator. Perfusion fluid contained 100 nm fluorescent beads at 10 μg ml⁻¹ and was applied directly to the node via micropipette (pipette opening 4–6 μm in diameter) that was 4–6 μm away from the imaging area. Flow rates were adjusted manually as described above such that there was no net flow at the beginning of the experiment and ~10–12 μm s⁻¹ velocity at the end of the 15 s imaging experiment. Bead velocities and tracks were quantified and visualized using the ‘manual tracking’ plugin in ImageJ.

Pipette flow calibration

Pipette flow was calibrated using 300 nm fluorescent beads (Sicastar-greenF, Micromod) re-suspended at 0.5 mg ml⁻¹ in DMEM/F12 10% FCS and loaded into micropipettes following sonication. A pressure stimulus was applied to the back of the pipette and steady flow imaged at 1,000 f.p.s. to ensure accurate frame-by-frame reading for each bead position while exiting the pipette and during its travel across the imaging area (~15 μm). The 1-ms time resolution of the calibration experiment was sufficient to resolve and calibrate the range of velocities used.

Scanning electron microscopy

Cultured mIMCD cells were fixed for 1 h at room temperature with 2.5% glutaraldehyde (Electron Microscopy Sciences) in 0.1 M sodium cacodylate buffer (pH 7.4), supplemented with 2 mM CaCl₂, and stored in distilled water. Organ of Corti explants were fixed following Ca²⁺ imaging experiments, or the entire cochlea was fixed, microdissected in distilled water and prepared for scanning electron microscopy as previously described³⁹. Briefly, specimens were dehydrated in ethanol, critical point dried from liquid CO₂, mounted on a carbon tape, sputter-coated with 5 nm platinum, and imaged on a Hitachi S-4800 field emission scanning electron microscope.

Data analysis

All images were analysed using a custom-made MATLAB tracking algorithm (described below), ImageJ (NIH), and Origin 8 (OriginLab).

Sample size predetermination (power analysis)

In stereocilia we observed a 2.9 ± 1.05 (mean ± s.d.) -fold change (effect size d = 2.76) in F_GECO1.2/F_mCherry after activation of a Ca²⁺-conducting mechanosensor. Assuming a one-sided, paired t-test conducted at the 0.05 level of significance, a minimum of 12 cells would be required to detect an effect size of d = 1 in mechanosensitive [Ca²⁺] increase post-stimulation with 95% power.

Image analysis

Customized image analyses were developed using MATLAB to automatically process the large volume of ratiometric time-lapse data to improve quantitation and objectivity. The analysis was divided into three steps: channel alignment, object detection, and object tracking over time, with subsequent ratio calculations.

Channel alignment. Two factors contribute to misalignment of two channels during image acquisition: chromatic aberration from the optics and time delay due to sequential acquisition. For chromatic aberration, the two channels from each frame of a given time-lapse image were aligned using a translational transformation. A global translational transformation, derived from the individual frame transformations, was applied to all frames.

Object detection. Frame-by-frame superimposed images of both channels were created. When cilia motion was faster than acquisition time, channels were significantly misaligned. To create a combination image, two channels were added and smoothed using a Gaussian filter. Local image background was subtracted from the combination image, and Otsu thresholding used to detect objects. Small objects (less than three pixels) were filtered as noise.

Object tracking with ratio calculations. Different cilia in the same image vary with relative orientation. During flow application, many cilia undergo large deflections and even cross one another, further complicating the tracking of individual cilia. A tracking algorithm based on object overlap was implemented. It included features such as splitting of crossed cilia, linking cilia with no spatial overlap, and closing gaps over a given number of time frames. Once an individual cilium was tracked, the analysis code measured the signal of the cilium from both channels, calculated the ratio, and plotted it against time or spatial displacement.

Code availability

The algorithms were developed in MATLAB in open-source, and are available upon request.

Extended data figures and tables

Extended Data Figure 1 Arl13b–mCherry–GECO1.2 identifies primary cilia.

Arl13b–mCherry–GECO1.2 contains an improved genetically encoded calcium indicator³⁶ (GECI) with an apparent K_d of 450 nM (Extended Data Fig. 2), well suited to work within the reported range of ciliary Ca²⁺ concentrations ([Ca²⁺]_cilium)⁴⁰. The genomic integration site of the transgene is within a non-coding region of chromosome 1 (Extended Data Fig. 2) and transgenic animals maintained as homozygotes (Arl13b–mCherry–GECO1.2^tg) are viable, have average litter sizes, and do not show phenotypes consistent with cilia defects (for example, situs inversus, organ malformation; Extended Data Fig. 2). a, Frozen tissue section of P21 mouse kidney. GECO1.2 and mCherry are preferentially localized to cilia, identified by the cilia-specific marker, anti-acetylated tubulin antibody. b, c, Primary mIMCD cells isolated from kidneys of P14–P21 Arl13b–mCherry–GECO1.2 transgenic mice. Ciliary localization (arrow) of anti-polycystin 2 antibody in c. Scale bars, 10 μm. d, Two OHC hair bundles marked with the actin-binding peptide, phalloidin. One of the stereocilia bundles expresses Arl13b–mCherry–GECO1.2. Kinocilia on both OHCs marked with an antibody to acetylated tubulin, express Arl13b–mCherry–GECO1.2. Scale bar, 5 μm. e, Scanning electron micrograph of a primary mIMCD cell. Left, red dashed line outlines a single mIMCD cell; white circle indicates the primary cilium. Scale bar, 10 μm. Right, magnified image. No defects in cilia formation were evident following 3–4 days in culture. Scale bar, 500 nm. All images are representative of more than ten images taken of biological triplicates.

Extended Data Figure 2 Arl13b–mCherry–GECO1.2 transgenic mouse.

a, Transgene orientation and integration site. The transgene was integrated into the non-coding region of chromosome 1 (position 174,611,500). b, The genotype of transgenic animals was determined by PCR using the following primers: 372-up, ACATGGCCTTTCCTGCTCTC; 372-down, TTCAACATTTCCGTGTCGCC; and 944-down, GACATCTGTGGGAGGAGTGG. The PCR product for wild-type genomic sequence was ~600 bp; the transgene PCR product was ~400 bp. c, mIMCD cells isolated from Arl13b–mCherry–GECO1.2^tg mice were imaged after permeabilization with 15 μM digitonin in varying extracellular [Ca²⁺]. Average ratios (n = 12 cilia per each data point) are plotted versus free [Ca²⁺]. Arl13b–mCherry–GECO1.2 calibration fitted by a Boltzmann curve (R² = 0.98; K_d = 442 nM). d, Phenotype of Arl13b–mCherry–GECO1.2^tg/tg mice. Mouse organ morphology/orientation appeared normal (heterotaxy was not observed) and breeding animals had normal litter sizes (6–8 for C57Bl/6 (ref. 41)). All error bars ± s.e.m.

Extended Data Figure 3 Arl13b–mCherry–GECO1.2^tg/tg mouse organ of Corti hair cells develop normal stereocilia bundles.

Hair bundles from Arl13b–mCherry–GECO1.2^tg/tg mice during development. a–d, Cochlear hair cells acutely dissected at age E18 appear normal. a, IHC, with kinocilium not attached to stereocilia. b, c, OHC stereocilia bundles with some kinocilia attached to the bundles. d, IHC bundle with kinocilium attached at tip. e–i, Cultured organ of Corti explant dissected at P1 + 1 day in vitro. e, OHC with normal shape and stereocilia staircase structure. f, OHC stereocilia bundle with kinocilium. g, IHC with normal shape and stereocilia staircase structure; tip links and other links present. h, Pair of stereocilia (dashed box in f) at higher magnification. i, IHC with kinocilium attached at tip. Scale bars, 1 μm (except h, 100 nm). All images are representative of more than three images.

Extended Data Figure 4 GECO1.2 k_off for Ca²⁺ dissociation from GECO1.2, measured in cells.

a, b, Fig. 1g at higher time resolution. F_GECO1.2/F_mCherry (open circles) relative to bundle motion (black line) at the initial deflection (a) and after return to the resting position (b). Bundle deflection preceded the F_GECO1.2/F_mCherry increase by two frames (~60 ms). At the termination of the flow stimulus and return to the resting position, the ratio remained elevated owing to the slow decay (τ ≈ 0.6 s) of the Ca²⁺ from GECO1.2. c, Rapid Ca²⁺ uncaging was used to measure the Ca²⁺ decay rates for GECO1.2 (not bound to the membrane), Arl13b–mCherry–GECO1.2 (bound to the membrane), and the Fluo-4 control with its established time constant of decay. HEK293 cells were transfected with GECO1.2 or Arl13b–mCherry–GECO1.2 constructs and loaded with caged Ca²⁺ (NP-EGTA), or loaded with a combination of Fluo4-AM and NP-EGTA. Caged Ca²⁺ was released by a local 100–200 ms pulse of ultraviolet laser illumination (white box); images were acquired in line scan mode (2 ms per vertical line). Representative of more than 16 images. d, Representative fitted fluorescence intensity decays of Fluo-4 in HEK293 cells (black), GECO1.2 (red), and Arl13b–mCherry–GECO1.2 (blue), compared with GECO1.2 from Arl13b–mCherry–GECO1.2^tg IHCs following deflection (green). e, Table summarizing molecular properties of genetically encoded calcium indicators (GECIs) used in current and previous reports describing ciliary Ca²⁺ signalling^{14,15,16,17,36,42,43,44,45}. f, Average decay rates, τ, for indicators in c, d. Fluo-4: τ = 154 ± 36 ms (n = 16); GECO1.2 in HEK293 cells: τ = 203 ± 50 ms (n = 19); Arl13b–mCherry–GECO1.2 in HEK293 cells: τ = 358 ± 55 ms (n = 16); Arl13b–mCherry–GECO1.2 in IHC stereocilia: τ = 601 ± 70 ms (n = 10). Averages ± s.d.

Extended Data Figure 5 Kidney tubule dissection and flow velocity calibration.

a, Calibration of the fluid velocity exiting the stimulus pipette versus distance from the mouth of the pipette (3-μm pipette, 6.4 mm Hg pressure step). The pipette was positioned ~4–6 μm from the cilium, delivering a flow velocity of 250–300 μm s⁻¹ (n = 5). b, Velocity measured at the tip of the cilium and calculated shear stress at the plasma membrane (n = 6). c, Microdissection of kidney tubules. c1, Coronal section of P15 kidney; white box indicates the microdissected area. c2, Area from c1 following microdissection. c3, Small bundle of tubules; individual tubules gently separated from the bundle (arrow). c4, Tubules with thick walls and fluorescent cilia (asterisk) used for experiments. Scale bar, 5 μm. Images representative of more than six preparations. d, Maximum intensity z-projection of mIMCD primary cilia deflected in the flow chamber using defined fluid velocities. Scale bar, 3 μm; representative of 14 cilia each with 7 z-stacks containing 12 frames. e, Relative ratio changes in primary mIMCD primary cilia during deflection, three experimental conditions. Black and blue lines represent the averaged normalized ratio changes for top views of cilia deflection in 1.3 mM and 50 nM [Ca²⁺], respectively. The increase in ratio upon cilia deflection is comparable between high and low external [Ca²⁺], suggesting that the ratio change did not result from Ca²⁺ entry (Fig. 3h for P values). In addition, the return to baseline with cilium movement was much faster than the Ca²⁺ indicator response time, providing further evidence that it is a motion artefact. Such fast responses were not observed in the bona fide Ca²⁺ entry into IHC stereocilia (Fig. 2a and Extended Data Fig. 4b, d). Purple line represents the average normalized ratio changes for side-imaged cilia deflections in the presence of 1.3 mM [Ca²⁺]. Ratio changes were negligible in side views, as the motion artefact (light path length change upon motion) is small. The positive slope seen in top views results from differential dye bleach, faster for mCherry than GECO1.2. Bleaching has a much more pronounced effect in top views, probably from the change in geometry, contribution from underlying autofluorescence, and relative light exposure upon bending. f, Micropipette used for cilia deflection inside the kidney tubules. Scale bar, 0.5 mm. g, Insertion of micropipette into the tubule. Scale bar, 100 μm. All error bars ± s.e.m.

Extended Data Figure 6 Deflection of primary cilia in the presence of 50 nM [Ca²⁺] reveals focal-plane-dependent artefact present in ‘top view’ imaging conditions; saturation controls for sensor.

a–c, An mIMCD cell cilium was repeatedly deflected in a low (50 nM) [Ca²⁺] solution. The same flow stimulus (same pipette) was applied to the cilium while imaging in different focal planes. Top panels, experimental arrangement and fluorescence images of the cilium; red dashed lines indicate the focal plane for each set of images. Middle panels (a′–c′), average fluorescence intensity change for F_GECO1.2 (green) and F_mCherry (red) during deflection (black). Lower panels (a′′–c′′), ratio change (blue) during deflection (black). a, Primary cilium near its attachment to the cell. b, At a slightly higher focal plane, deflection enlarges the cross section of the cilium. c, A focal plane ~1 μm above the cell surface. Different segments of the same cilium were imaged upon deflection. Note that the artefact increases with larger cross section changes. Thus, top view imaging of cilia is fraught with two interrelated artefacts: (1) at high z-resolution (0.8 μm), the section of the cilia being imaged changes upon deflection, thus conflating fluorescence changes from different regions of the cilium; (2) at lower z-resolution, the path length, fluorescent indicator volume, and optical properties of the cilium above and below the image plane all change upon deflection and thus contribute to the apparent [Ca²⁺] reporter changes. d, e, Digitonin permeabilization indicates that the Arl13b–mCherry–GECO1.2 sensor is not saturated in measurements of primary cilia and kinocilia. Scale bars, 2 μm. d, Top: mIMCD cell cilium deflected by fluid flow containing 10 μM digitonin. Bottom left: F_mCherry (red) decreased owing to cilia motion (black). F_GECO1.2 (green) rose ~1 s later, as permeabilization initiated Ca²⁺ influx. Right: F_GECO1.2/F_mCherry increased ~4.2-fold upon permeabilization. Representative of 13 videos × 400 frames. e, Kinocilia of E15 cochlear hair cells. Digitonin (10 μM) increased the kinocilium’s normalized F_GECO1.2/F_mCherry ratio by 4.6-fold. Representative of 6 videos × 400 frames. Similar results were obtained in P3 hair cell kinocilia (data not shown).

Extended Data Figure 7 No change in [Ca²⁺]_cilium during mechanical stimulation of primary cilia in MLO-Y4 and Ocy454 osteocyte-like cells, and primary MEF cells.

a–c, Cultured MLO-Y4 (a, representative of 24 videos × 150 frames), Ocy454 (b, representative of 11 videos × 150 frames), and MEF cells isolated from E14 Arl13b–mCherry–GECO1.2^tg:GCaMP6f^tg:E2a-Cre^tg mice (c, representative of 25 videos × 150 frames) were imaged from above; stimulus pipette was placed ~4–6 μm away from the cilium. Images: cilium before, during, and after deflection by a 2 s, ~250 μm s⁻¹ flow stimulus. The ROI was identified frame-by-frame by a MATLAB tracking algorithm and F_GECO1.2 and F_mCherry quantified. Scale bar, 1 μm. d–f, Quantification of the channel intensities from the cilia in a–c. Average fluorescence intensity for both F_GECO1.2 and F_mCherry decreased as the cilia flattened and the light path length via the cilium volume changed. Cilium ROI displacement is superimposed in grey. g–i, Ratioing F_GECO1.2 and F_mCherry compensated for the path-length artefact (see also Extended Data Fig. 6), revealing no change in [Ca²⁺] during deflection. Cilium ROI displacement is superimposed in grey. j, Average F_GECO1.2/F_mCherry for MLO-Y4 (red, n = 24), Ocy454 (purple, n = 11) and MEF (blue, n = 25) primary cilia. The small continuous positive slope during the entire course of the experiment results from differential dye bleaching (mCherry > GECO1.2). All error bars ± s.e.m.

Extended Data Figure 8 Flow-dependent [Ca²⁺] increases originate in the cytoplasm in MEF cells.

a, Representative image sequence of cultured MEF cells responding to 1 dyn cm⁻² fluid flow in a flow chamber. Cells were isolated from E14 Arl13b–mCherry–GECO1.2^tg:GCaMP6f^tg:E2a-Cre^tg mice, expressing Arl13b–mCherry–GECO1.2 in primary cilia, and GCaMP6f in the cytoplasm. F_GCaMP6f is presented as pseudocolour heatmap; arrows point to the cells that respond to the flow stimulus, circles indicate ROI used for analysis in f; imaging rate, 0.6 f.p.s.; scale bar, 10 μm. Representative of 19 videos × 60 frames. b–b′′, Representative image sequence of cultured MEF cell responding to a single cycle of oscillatory fluid flow (OscFF¹⁶). Top row, F_GCaMP6f + F_GECO1.2 presented as pseudocolour heatmap. Middle row, mCherry fluorescence. Bottom row, merged F_GCaMP6f+ F_GECO1.2 (green) and F_mCherry (red) signals. Left panel, b, average of about ten consecutive frames before stimulus application; 1.2 s into the experiment, an alternating pressure stimulus was applied to the cell membrane, away from the cilium (positive, then negative, ~1.5 s each, diagram shown in b′). Following the stimulus (b′′; each image is an average of three to five consecutive frames at the time point reflected on the image), a Ca²⁺ wave originating from the plasma membrane/cytoplasm spreads across the cell body. As seen on the image sequence, a single cycle of strong OscFF application to the cell membrane initiates a Ca²⁺ increase in the cytoplasm (whether from across the plasma membrane or from intracellular stores was not determined). Negligible cilium movement has been detected in this particular case (~200 nm, grey trace in c), as the cilium was located under the cell, between the cell and the coverslip. Scale bar, 5 μm. Representative of 40 videos × 450 frames. c, Quantification of (F_GECO1.2 + F_GCaMP6f) (green) and F_mCherry (red) channel intensities in the cilium in b. The ROI was identified frame-by-frame by a MATLAB tracking algorithm and average fluorescence plotted as a function of time. Cilium ROI displacement is superimposed in grey. d, Average F_GCaMP6f (cytoplasm) in the ROI depicted in b′ (white circle), superimposed with ciliary (F_GECO1.2 + F_GCaMP6f)/ F_mCherry (purple trace) showing an earlier Ca²⁺ onset for the cytoplasmic GCaMP6f indicator. Dashed box outlines the data shown in e. e, Same as in c, but with higher time resolution. Ciliary [Ca²⁺] increase is ~200 ms delayed from the cytoplasmic Ca²⁺ increase, showing the necessity for high imaging rates for Ca²⁺ imaging of cilia. f, Quantification of F_GCaMP6f change over time for four cells from a; ROIs depicted in a. Dashed line: threshold for labelling a cell as responsive (20% change in fluorescent intensity). All selected cells, except that represented by the magenta trace, responded to the flow. g, MEF cell response rate in flow chamber and pipette flow application experiments. Blue and red bars represent 1 dyn cm⁻² application in flow chamber experiments; ~40% of the cells responded to the flow. Apyrase (7 units per millilitre) application did not change the response rate, suggesting that ATP release is not a major contributor to flow-induced Ca²⁺ response in this cell type. Remaining bars represent single-cell imaging experiments with local flow delivery via pipette, using cells in three conditions: low confluency cells (~15%), with 24 h (green bar) and 48 h (purple bar) serum starvation to promote cilia formation, and highly confluent coverslips without serum starvation (light blue bar). Arrest of MEF cells in G0 sensitizes cells to respond to flow stimulus with intracytoplasmic Ca²⁺ changes. All error bars ± s.e.m.

Extended Data Figure 9 Flow-dependent [Ca²⁺] increases originate in the cytoplasm in primary mIMCD cells.

a, Cultured primary mIMCD cells, isolated from Arl13b–mCherry–GECO1.2^tg:GCaMP6f^tg:E2a-Cre^tg mice, respond to 1 dyn cm⁻² shear stress in a flow chamber. F_GCaMP6f fluorescence intensity is presented as pseudocolour heatmap. Arrows point to the cells that respond to the flow stimulus. Imaging rate, 0.8 f.p.s.; scale bar, 10 μm. Representative of 6 videos × 100 frames. Circles indicate ROIs used for analysis in f. b, Cultured mIMCD cell with cytoplasmic Ca²⁺ oscillations following flow application. Pipette outline on the image represents its approximate position; an arrow inside the pipette represents the direction of the flow. An alternating flow stimulus deflects the primary cilium (red) in positive and negative directions. This deflects the cell membrane, resulting in Ca²⁺ increases in the cytoplasm. Although it is well known that mechanical force can initiate increases in cytoplasmic Ca²⁺, there are several potential sources (plasma membrane rupture, mechanosensitive ion channels, ATP release and purinergic receptor activation, intracellular Ca²⁺ stores) that appear to depend on cell type and conditions. No change in ciliary [Ca²⁺] is evident until the cytoplasmic Ca²⁺ reaches the cilium (arrowhead, 2.55 s). Top row, F_GCaMP6f + F_GECO1.2 presented as a pseudocolour heatmap. Middle row, mCherry signal intensity. Bottom row, merged F_GCaMP6f + F_GECO1.2 (green) and F_mCherry (red). Asterisks indicate the base of the cilium. Imaging rate, 33 f.p.s.; scale bar, 5 μm. Representative of 17 videos × 450 frames. c, Fast imaging during supraphysiological flow application to primary cilium of cultured mIMCD cells reveals ciliary damage and subsequent increase of ciliary [Ca²⁺]. F_GCaMP6f + F_GECO1.2, pseudocolour heatmap (left); F_mCherry, red (right). At flow rates greater than ten times those used for ciliary deflection, the ciliary membrane disintegrates and distal parts of the cilium (arrowheads) detach. Following ciliary tip damage, Ca²⁺ enters the cilium from the break point. Ciliary Ca²⁺ influx does not occur before detachment of ciliary tip, presumably when the membrane experiences the highest force and stretch. Arrow, direction of the flow. Imaging rate, 100 f.p.s.; Scale bar, 2 μm. Representative of 13 videos × 3,000 frames. d, e, Quantification of F_GCaMP6f + F_GECO1.2 and F_mCherry from the cilium in c. Arrow, ciliary tip detachment event. The ROI was identified frame-by-frame by a MATLAB tracking algorithm and F_GECO1.2 and F_mCherry (d) and ratio (e) plotted as a function of time. Cilium ROI displacement is superimposed (purple). f, Quantification of F_GECO1.2 and F_mCherry fluorescence intensity change over time for four representative cells in a; ROIs depicted in a (left). Dashed line, 20% threshold defining responsive cells. All selected cells (except that represented by the green trace) responded to flow. g, Cultured primary mIMCD cell response rate in flow chamber and pipette flow application experiments. Blue bar, flow application performed in a flow chamber; ~46% of the cells respond to 1 dyn cm⁻² shear stress. Green and purple bars represent single-cell imaging experiments for highly confluent cells (~80%), without (green bar) or following (purple bar) 48 h serum starvation. Light blue bar, cells following 48 h serum starvation in response to damaging flow stimulus. All error bars ± s.e.m.

Extended Data Figure 10 Cytoplasmic Ca²⁺ signalling in the embryonic node of Arl13b–mCherry–GECO1.2^tg: GCaMP6f^tg:E2a-Cre^tg embryos.

a–a′′, Representative images of embryos within the developmental window used: from early bud (EB, a), early headfold (EHF, a′′) to two- to three-somite stage (2–3S, a′). Scale bars, 200 μm. Each panel is a representative of at least five images. b, Drawing (top modified after ref. 22) and image (bottom) of the embryo mounting plate used for imaging. Scale bar, 10 mm. c, Embryos were mounted with the node facing up. Pipette used for cilia deflection is shown in blue. d, Differential interference contrast image of embryo with the node outlined by the dashed line. Scale bar, 100 μm. Representative of 14 images. e, Representative image of an embryonic node of an Arl13b–mCherry–GECO1.2^tg:GCaMP6f^tg:E2a-Cre^tg EHF embryo, expressing GCaMP6f in the cytoplasm to visualize cytoplasmic Ca²⁺ oscillations. Scale bar, 20 μm. Representative of 3 videos × 300 frames. f, Mapping of cytoplasmic Ca²⁺ oscillations in close proximity to the embryonic node. Only ΔF/F > 30% were included in the analysis. Nodal perimeter is outlined with white dashed line. g, Quantification of cytoplasmic Ca²⁺ signals shown in e, f occurring within 0.01 mm² on either the left or the right side of the embryonic node. Left side: 8.3 ± 2.3 min⁻¹; right side: 7.6 ± 1.2 min⁻¹; n = 5 embryos; this difference was not statistically significant between the late bud and late headfold (LHF) stage (see also ref. 33) (Supplementary Video 10). h, i, Average ratio changes (F_GECO1.2/F_mCherry) of crown cell primary cilia from the left side of the embryonic node of Arl13b–mCherry–GECO1.2^tg/tg:GCaMP6f ^tg:E2a-Cre^tg expressing EHF embryos during the application of physiological levels of flow (slow deflection; note longer imaging time of 15 s). Flow velocity, calibrated in-frame using fluorescent beads, was slowly increased (ramped); see also Fig. 4 k, l. Cilia were divided into two groups: h, the average ratio change for displaced cilia (n = 34 cilia from 5 embryos; average centroid displacement was 409 ± 35 nm); i, the average ratio change for non-displaced cilia (n = 23 cilia from 5 embryos). All error bars ± s.e.m.

PowerPoint slides

PowerPoint slide for Fig. 1

PowerPoint slide for Fig. 2

PowerPoint slide for Fig. 3

PowerPoint slide for Fig. 4