Primary cilia are specialized calcium signalling organelles


Primary cilia are solitary, non-motile extensions of the centriole found on nearly all nucleated eukaryotic cells between cell divisions. Only 200–300 nm in diameter and a few micrometres long, they are separated from the cytoplasm by the ciliary neck and basal body. Often called sensory cilia, they are thought to receive chemical and mechanical stimuli and initiate specific cellular signal transduction pathways. When activated by a ligand, hedgehog pathway proteins, such as GLI2 and smoothened (SMO), translocate from the cell into the cilium1,2. Mutations in primary ciliary proteins are associated with severe developmental defects3. The ionic conditions, permeability of the primary cilia membrane, and effectiveness of the diffusion barriers between the cilia and cell body are unknown. Here we show that cilia are a unique calcium compartment regulated by a heteromeric TRP channel, PKD1L1–PKD2L1, in mice and humans. In contrast to the hypothesis that polycystin (PKD) channels initiate changes in ciliary calcium that are conducted into the cytoplasm4, we show that changes in ciliary calcium concentration occur without substantially altering global cytoplasmic calcium. PKD1L1–PKD2L1 acts as a ciliary calcium channel controlling ciliary calcium concentration and thereby modifying SMO-activated GLI2 translocation and GLI1 expression.


We generated a transgenic ARL13B–EGFP mouse (Arl13b-EGFPtg) in which primary and motile cilia show spectacular fluorescence labelling throughout the animal, developed a ratiometric Ca2+ sensor directed specifically to cilia, and patch-clamped individual cilia and measured the resting membrane potential and a calcium-permeant conductance (Icilia; see ref. 5). The Arl13b-EGFPtg mice displayed ubiquitous green fluorescence only in primary and motile cilia but not in microvilli (Fig. 1, Extended Data Fig. 1 and Supplementary Video 1). After incubation of E14.5 embryos in clarifying ScaleA2 solution6 (Fig. 1a–c and Supplementary Video 2), three-dimensional imaging of ARL13B–GFP reveals its exclusive localization to cilia. Staining of wild-type and Arl13b-EGFPtg retinal pigmented cell epithelia (mRPE) cells and mouse embryonic fibroblasts (MEFs) with ciliary markers confirmed that ARL13B–EGFP is localized exclusively to cilia without noticeable alteration of cilia morphology (Extended Data Fig. 2).

Figure 1: ARL–GFP localizes to primary cilia and motile cilia in vivo.

a, b, Fixed paw of E14.5 embryo (a), and after incubation in ScaleA2 (b). The inset indicates the area where z-stacks from 0 to 1.2-mm depth were acquired. Scale bar: 500 μm. c, ARL13B–EGFP expression in E14.5 paw. 3 × 4 z-stacks were stitched together, depth colour-coded and projected onto a plane. Scale bar: 200 μm. df, EGFP-positive cilia are present in fibroblasts within the digit (arrows) that overlap with adenylyl cyclase III (ACIII) labelling in e and merged in f. Scale bar: 10 μm; inset, 3 μm; asterisks indicate autofluorescence. Red and green channels are offset to visualize co-localization. gi, Motile cilia in the Fallopian tube express ARL13B–EGFP (g) that overlaps with acetylated tubulin (h). Overlay (i). Inset in i is DIC image. Scale bar: 5 μm.

PowerPoint slide

Primary cilia are not reliably loaded with Ca2+-sensitive dyes, forcing experimentalists to rely on changes in cytoplasmic calcium concentration ([Ca2+]cyto) as an indirect indicator of ciliary calcium concentration ([Ca2+]cilia). To ameliorate this issue, we generated a genetically encoded calcium sensor that is targeted to the cilium by fusing GCaMP3 (ref. 7) to the carboxy terminus of smoothened (SMO–GCaMP3), enabling monitoring of [Ca2+]cilia and [Ca2+]cyto simultaneously. As shown in Extended Data Fig. 3 and Supplementary Video 3, addition of the Ca2+ ionophore, ionomycin, increased fluorescence in both the cytoplasm and cilium of hRPE1 cells stably expressing SMO–GCaMP3, although variation in ionomycin incorporation precludes precise [Ca2+] comparison between the two compartments.

To quantify [Ca2+] we developed a ratiometric SMO–mCherry–GCaMP3 calcium sensor8 (Fig. 2a, b). In hRPE1 cells stably expressing SMO–mCherry–GCaMP3, the cilia-targeted GCaMP3 and mCherry fluorescence co-localized with the cilia-specific marker, acetylated tubulin (Extended Data Fig. 3e–h). To determine whether [Ca2+]cilia could be increased without affecting [Ca2+]cyto, we ruptured the cilia membrane at the tip of the cilia (circle; Extended Data Fig. 3i, j) with an intense 1–2 s laser pulse (405 nm), leading to a rapid increase in [Ca2+]cilia from the tip that travelled to the ciliary base (Supplementary Video 4). Peak [Ca2+] propagated down the cilia at a rate of 4.6 ± 0.6 µm s−1, yielding an apparent diffusion constant (DCa) of 5.3 µm2 s−1 (similar to 5–10 µm2 s−1 for DCa in stellate cell dendrites9).

Figure 2: SMO–mCherry–GCaMP3 localizes to the primary cilium and reports ciliary [Ca2+].

[Ca2+]cilia is poorly coupled to [Ca2+]cyto. a, hRPE1 cells expressing SMO–mCherry–GCaMP3 were treated with 5 μM ionomycin and changes in fluorescence were measured in the cilium and cytoplasm. b, Schematic of ratiometric calcium sensor and quantified fluorescence (F). c, d, After rupture of the ciliary tip with a laser pulse (t = 10 s), [Ca2+]cilia rapidly increases, whereas [Ca2+]cyto near the base of the cilium (circle position 2) increases only slightly 40 s later (t = 61 s). Ionomycin was added for normalization. Circles in c indicate areas where fluorescence was measured. ΔF/F, changes in fluorescence. e, Distribution of lag times between ciliary and cytoplasmic [Ca2+] increases. f, Calcium diffusion from the cytoplasm to the cilium is not restricted. Red dot indicates the position of calcium uncaging whereas the white line indicates length and position of line scan. Arrows indicate cytoplasm to cilium transition. Scale bars: 5 μm. See also Supplementary Video 6.

PowerPoint slide

In order to closely monitor changes in [Ca2+]cyto at the cilia–cytoplasm junction, we loaded the calcium indicator Fluo-4 into hRPE1 cells stably expressing SMO–mCherry–GCaMP3. Ciliary membrane rupture increased [Ca2+]cilia and was detectable at the ciliary base after a 40 s delay (Fig. 2c–e and Supplementary Video 5). More distant parts of the cytoplasm remain unaffected by fluctuations at the base of the cilium. Elevation of [Ca2+]cyto after ciliary rupture occurred infrequently and with varying delay; 40% of the cells did not show any elevation in the cytoplasmic side of the cilia–cell junction even 60 s after [Ca2+]cilia had saturated (Fig. 2e). Apparently, the Ca2+ sensor does not hinder the diffusion of Ca2+ from cilium to cytoplasm after rupture: in this case the cilium fills with 2 mM [Ca2+], which saturates the Ca2+ sensor. Next, we asked whether the protein-dense structure at the base of the cilium might act as a localized Ca2+ buffer and/or impermeant physical barrier to Ca2+ entry into the cytoplasm. We loaded SMO–mCherry–GCaMP3-expressing cells with the caged calcium chelator NP-EGTA-AM and uncaged Ca2+ in the cytoplasm in close proximity to the ciliary base. The velocity of this Ca2+ wave was 16 ± 2 μm s−1 in the cytoplasm and continued at 15 ± 2 μm s−1 in the cilium, indicating no significant delay of Ca2+ entering the cilium from the cytoplasm (Fig. 2f and Supplementary Video 6). In summary, because the ratio of cilium/cytoplasmic volume is exceedingly small (1:30,000), the tiny ciliary Ca2+ rivulet is rapidly diluted in the large cytoplasmic volume. Because the number of free calcium ions at 1 µM concentration within the 0.5 fl cilioplasm is 200 ions at an instant in time, ciliary [Ca2+] does not perturb cytoplasmic [Ca2+] substantially, nor does it initiate a measurable Ca2+-induced Ca2+ release wave in the cytoplasm. These considerations led us to ask whether resting [Ca2+]cilia differs from resting [Ca2+]cyto.

The dissociation constant (Kd) for the ratiometric ciliary Ca2+ sensor in situ was comparable to that in solution (625 nM versus 660 nM (ref. 7); Extended Data Fig. 4). In 2 mM extracellular Ca2+ concentration ([Ca2+]e), resting [Ca2+]cilia was 580 nM in hRPE cells (Fig. 3b). We also calibrated the ratiometric Ca2+ sensor in the cytoplasm using cells that had not yet formed a cilium (SMO in the plasma membrane; Fig. 3a, bottom), and obtained a similar calibration curve and Kd (550 nM), suggesting that the sensor reported [Ca2+] similarly in cilia and cytoplasm. The average ratio for the sensor in the cytoplasm of 0.28 ± 0.02 corresponding to a resting [Ca2+]cyto of <110 nM, a normal resting cytoplasmic [Ca2+]10 (Fig. 3c). To test the surprisingly high resting [Ca2+]cilia using an independent measurement, we patch-clamped SMO–EGFP-labelled cilia in the whole-cell configuration and varied free [Ca2+]cilia. High [Ca2+]cilia (400 nM; IC50 = 445 nM (see Methods), Fig. 3d) inactivates Icilia, enabling an independent calibration of [Ca2+]cilia. We then estimated the undisturbed resting [Ca2+]cilia by comparing the current amplitude from perforated-patch measurements with the [Ca2+] calibration curve. The magnitude of Icilia confirmed that [Ca2+]cilia was 7-fold higher than resting [Ca2+]cyto (Fig. 3c).

Figure 3: Resting cilium [Ca2+] is substantially higher than resting cytoplasmic [Ca2+].

a, Live hRPE1 cell F GCaMP3/F mCherry ratios in 2 mM [Ca2+]e in the cilium (top) and cytoplasm (bottom). Scale bar: 5 μm. b, Plot of measured ciliary fluorescence ratio (average = 0.8 ± 0.12, n = 16 cilia) versus estimated [Ca2+]cilia. Resting [Ca2+]cilia = 580 nM. c, Plot of measured cytoplasmic fluorescence ratio (average = 0.28 ± 0.02, n = 20 cells) versus estimated [Ca2+]cyto. Resting [Ca2+]cyto is 107 nM. d, By measuring current amplitudes in perforated patches (Methods), we estimated resting free [Ca2+]cilia as 742 nM. Black circles indicate current at +100 mV; grey circles indicate current at −100 mV. e, Changes in cell (Vm) and cilia (Vcilia) potentials in response to external [K+]. f, Average potential of the cell body and cilia plotted as a function of external [K+]. Vm differs from Vcilia at all [K+] other than [K+]e = 145 mM; P < 0.05. The measured resting membrane potential is −18 mV for the cilia and −54 mV for the cell (±s.e.m., n = 5 cells and 4 cilia). The grey dashed line is the K+ Nernst potential.

PowerPoint slide

We next determined the cilia’s resting membrane potential (Ecilia) by measuring changes in response to depolarizing concentrations of extracellular potassium ([K+]e) from hRPE1 SMO–EGFP cytoplasm and from its primary cilia. Consistently, Ecilia was >30 mV more positive than the cytoplasm (Ecilia = −18 ± 1 mV; Ecyto = −54 ± 2 mV, respectively; Fig. 3e) and significantly less [K+]e (70 mM versus 129 mM) was required to depolarize the ciliary membrane potential to 0 mV (Fig. 3f). In summary, the cilium is a functionally distinct cell compartment with respect to ions. This calcium compartment is maintained by a favourable influx/efflux ratio: the large number of calcium-permeant channels5 or other ion channels/transporters can easily maintain high [Ca2+] in the small volume of the cilia, despite steady diffusion of Ca2+ into the cytoplasm at its base. An analogy is a water tower (Ca2+ within the cilia) connected to a large lake (cytoplasm) by a small pipe (basal body). Because ciliary [Ca2+] is high (600 nM) compared to the cytoplasm (100 nM), calcium flows from cilium to cytoplasm, but not cytoplasm to cilia. In addition, the approximately 30 mV gradient from cilia to cytoplasm further ensures asymmetry of Ca2+ flux. We speculate that these factors insulate cilia from the <1 µM Ca2+ fluctuations that characterize cytoplasmic Ca2+ signalling.

What are the consequences to cell function of ciliary ion independence from the cytoplasm? As we show in an accompanying paper5, Icilia is mediated predominantly by the heteromeric channel PKD1L1–PKD2L1. Pkd1l1−/− mice have a ciliary phenotype11,12, but in this study we focused on the pore-forming subunit of Icilia, PKD2L1. Although PKD2L1 was initially proposed to be a sour taste receptor, Pkd2l1−/− mice display no sour taste deficit13. Thus, we tested Pkd2l1−/− mice for a potential ciliary defect. Although we did not observe any major organ laterality defects, we observed intestinal malrotation in about 50% of the Pkd2l1−/− knockout animals that was not observed in wild-type mice (Fig. 4a), indicating a mild penetrance of the phenotype. Intestinal malformations are associated with sonic hedgehog (SHH) pathway defects during early development14,15.

Figure 4: Pkd2l1−/− mice exhibit defects in SMO-mediated GLI1 activation.

a, Intestinal malrotation in Pkd2l1−/− mice, and comparison with wild type. Arrow indicates correct rotation of intestine. Schematic orientation of intestines is on the right; red indicates mesentery. b, RT–PCR of MEF and mRPE-derived cDNAs shows that PKD2L1 is expressed in both cell types. c, Western blot of GLI1, SMO, GLI2 and γ-tubulin expression in wild-type and Pkd2l1−/− MEFs after stimulation for 24 h with 400 nM SAG. d, Top panel: GLI1 in wild-type (1.9 ± 0.3 versus 9.3 ± 0.7 after stimulation) and Pkd2l1−/− MEFs (0.9 ± 0.04 versus 3.9 ± 0.6 after stimulation). P < 0.01. Middle panel: GLI2 in wild-type (0.9 ± 0.1 versus 1.1 ± 0.2 after stimulation) versus Pkd2l1−/− MEFs (1.1 ± 0.3 versus 1.2 ± 0.1 after stimulation). Bottom panel: SMO in wild-type (1.3 ± 0.3 versus 1.2 ± 0.09 after stimulation) versus Pkd2l1−/− MEFs (1.3 ± 0.04 versus 1.3 ± 0.1 after stimulation; n = 3 wild-type embryos, 5 Pkd2l1−/− embryos). Error bars show ±s.e.m. e, Localization of GLI2 at the distal tip of Arl13b-EGFPtg and Pkd2l1−/− × Arl13b-EGFPtg MEF cilia. Scale bar: 5 μm. Arrows point to GLI2 at cilia tips. Cilia are magnified in insets. f, Quantification of GLI2 protein accumulation at the ciliary tip after SAG stimulation; wild type = 3.0 ± 0.2 (n = 40 cilia each from 3 embryos) versus Pkd2l1−/− MEFs = 1.8 ± 0.1 fold increase (n = 40 cilia each from 5 embryos). P < 0.05. g, h, Quantification of ciliary length (4.5 ± 1.5 μm for wild-type versus 4.6 ± 1.6 μm for Pkd2l1−/− MEFs) and percentage of ciliated cells (67 ± 8% for wild-type versus 60 ± 6% for Pkd2l1−/− MEFs; n = 40 cilia each from 3 embryos).

PowerPoint slide

Treatment of MEFs with the SMO agonist SAG directly activates the SMO receptor, leading to an upregulation of Gli1 and Ptch1 expression16,17. As Pkd2l1 is transcribed and localizes to cilia in wild-type MEFs (Fig. 4b and Extended Data Fig. 5), we next asked whether the SHH pathway might be affected in Pkd2l1−/− mice by measuring the upregulation of GLI1 in response to stimulation with SAG. As shown in Fig. 4c, d, GLI1 protein increased from 1.9 ± 0.3 to 9.3 ± 0.7 AU (arbitrary units) in wild-type MEFs stimulated for 24 h with 400 nM SAG. In contrast, SAG upregulated GLI1 from 0.9 ± 0.04 to only 3.9 ± 0.6 in Pkd2l1−/− embryonic MEFs. As shown in Fig. 4d, key members of the SHH pathway were not altered. However, Arl13b-EGFPtg MEFs accumulated significant amounts of GLI2 at the distal tip of the cilium (Fig. 4e, f), as is required for activation of GLI2 and full activation of downstream transcription events18. In Pkd2l1−/− × Arl13b-EGFPtg mutant cells, GLI2 accumulation at the ciliary tip was reduced by 50% compared to wild-type cells. PKD2L1 is not required for cilia formation, as ciliary length and the percentage of ciliated cells did not differ significantly between genotypes (Fig. 4g, h). Finally, we asked whether SAG stimulation itself regulates Icilia and thus ciliary [Ca2+]. Although there was no immediate effect on ciliary current stimulation with SAG, after 24 h Icilia amplitude and resting [Ca2+]cilia increased in the cilium of wild-type MEFs (Extended Data Fig. 6). These data suggest that SAG initiates recruitment of PKD2L1 channels into the cilium rather than activating the channel.

Our results indicate that primary cilia are functionally distinct from the cytoplasm. Icilia is encoded by a heteromeric PKD1L1–PKD2L1 channel and resting [Ca2+]cilia is at least 0.4 μM higher than resting [Ca2+]cyto, which regulates trafficking of hedgehog-mediated transcription factors in the cilium. PKD2L1 is increased by SHH pathway stimulation, probably by channel recruitment to the cilium. Interestingly, the IFT25–IFT27 complex seems to be specific for transporting SHH components and an IFT25 mutant MEF showed an impaired GLI2 ciliary trafficking phenotype17 similar to Pkd2l1 mutant MEFs. Because IFT25 has a unique Ca2+ binding site19, the observed SHH defect of Pkd2l1 null mice is indicative that PKD2L1 may ‘tune’ [Ca2+]cilia to optimize IFT25 function. [Ca2+]cilia is probably also adjusted by other PKD members or even other as-yet-unidentified ion pumps or transporters in the cilium during development, which may explain the mild phenotype of the PKD2L1 mutant mouse compared to other SHH-deficient mutant mice (Ift25−/− and Shh−/− mice15,17). An alternative, but not mutually exclusive, hypothesis is that acute Icilia regulation by G-protein-coupled receptors and growth factors dynamically regulates ciliary trafficking. Further studies are necessary to determine whether other intermediates in the SHH pathway, such as SUFU-GLI dissociation or GLI proteolytic processing, are [Ca2+]-dependent. A second conclusion from this work is that the cilium funnels a small but steady Ca2+ load into the peri-ciliary cytoplasm.

Methods Summary

A ratiometric GCaMP3-based calcium sensor was expressed in cilia to measure ciliary calcium in hRPE1 cells. A transgenic mouse model expressing GFP in cilia was generated to measure ciliary calcium and membrane potential by patch-clamp recordings of cilia. Defects in SHH signalling of MEFs isolated from Pkd2l1−/− mutant mice were quantified by western blotting and ciliary localization of GLI1 and GLI2 proteins.

Online Methods

Molecular biology

The hArl13b clone was obtained from Open Biosystems and cloned into the pEGFP-N1 vector. Smo-EGFP, GECO1.2 and GCaMP3 were obtained from Addgene. GECO1.2 is an improved version of GCaMP3 (ref. 20). GCaMP3 was spliced in frame to the 3′ end of Smo. The ratiometric sensor Smo-mCherry-GCaMP3 was obtained by adding short glycine-serine linkers to the 5′ and 3′ ends of mCherry by PCR amplification. mCherry was cloned non-directionally into mSmo-GCaMP3 after linearization with AgeI. Correct orientation of mCherry was confirmed by sequencing. The same approach was used to add mCherry-GECO1.2 to hArl13b.

Transgenic and Pkd2l1 knockout animals

We injected both a human Arl13b-EGFP cDNA construct and a hArl13b-mCherry-GECO1.2 cDNA construct under the control of a chicken actin promoter (CAG) into the pronucleus of mouse C57BL/6 oocytes and obtained two independent founder lines for ARL13B–EGFP and four independent founder lines for ARL13B–mCherry–GECO1.2. Transgenic males and females were viable and fertile. The ScaI/HindIII linearized plasmid was gel-purified and injected into the pronucleus of C57BL/6 oocytes at the transgenic animal core facility at Boston Children’s Hospital. Oviducts of Arl13b-EGFPtg mice were isolated as described previously21. All animal protocols in this work were approved by the IUACUC of Boston Children’s Hospital. Pkd2l1 knockout animals were obtained from Jackson laboratories and have been described previously13. P21 Pkd2l1−/− and age-matched C57BL/6 wild-type mice were used for morphological analysis.

Isolation of primary MEFs and primary mRPE cells

MEFs were isolated from E15 wild-type, Arl13b-EGFPtg, Arl13b-mCherry-GECO1.2tg or Pkd2l1−/−, or Pkd2l1−/− × Arl13b-EGFPtg embryos. Freshly collected embryos were placed in a 10-cm cell culture dish in sterile PBS. The head and inner organs were removed and used for genotyping. The body was minced in 1 ml of 0.5% trypsin (Life Technologies) with a razor blade and incubated at 37 °C for 30 min. Tissue was triturated by pipetting 10–20× and the trypsin reaction was quenched by addition of 10 ml complete medium with serum (DME w/ 10% FCS). The suspension was passed through a cell strainer and plated. MEFs were seeded onto coverslips at 50% cell density and serum-starved the next day for 24 h in DMEM plus 0.2% FCS for electrophysiological recordings, immunostaining, biochemistry and GLI2 translocation assays. MEFs were not used beyond passage 4. mRPE cells were isolated as described previously22.


We used the following antibodies: rabbit anti-adenylyl cyclase III (Santa Cruz); mouse anti-acetylated tubulin (Sigma-Aldrich); rabbit monoclonal anti-villin antibody (Abcam); mouse HMB45 and rabbit PKD2L1 (Thermo Scientific). Goat anti-smoothened and anti-Gli2 antibody was a gift from A. Salic; mouse monoclonal anti-γ-tubulin clone GTU 88 (Sigma-Aldrich); rabbit anti-GLI2 (Sigma Aldrich); rabbit anti-GLI1 V812 (Cell Signaling).

Immunocytochemistry and confocal microscopy

Cells were fixed with 4% formaldehyde, permeabilized with 0.2% Triton X-100, and blocked by 10% goat serum in PBS. Cells were labelled with the indicated antibody and secondary goat anti-rabbit or anti-mouse fluorescently labelled IgG (Life Technologies) and Hoechst 33342 (Life Technologies). Confocal images were obtained using an Olympus FV1000 with a 60× water immersion, 1.2 N.A. objective. Images were further processed using ImageJ (NIH). Differential interference contrast images (DIC) are shown to outline cells.

Three-dimensional reconstruction of cilia

E14.5 embryos were fixed in 4% PFA for 48 h and dissected. Tissue was treated with Scale solutions as described6. In brief, samples were incubated in ScaleA2 and ScaleB4 until they appeared transparent (3 weeks–3 months). Samples were mounted in 2% agarose in ScaleA2 on a Petri dish. Samples were imaged in ScaleA2 with an XLUMPLFL 25× water 1.05 N.A. MPE objective. Tiles of stacks were acquired with an Olympus FV1000 MPE Exclusive with SpectraPhysics MaiTai DeepSee laser set at 915 nm. Images were acquired with single excitation in GFP and RFP channels. Subtraction eliminated autofluorescence. Tiles were stitched together using an ImageJ plug-in23. Three-dimensional reconstruction was done using Imaris software (Bitplane AG).

Cell culture and generation of stable cell lines

hRPE1 cells were transfected using TransIT-LT1 (Mirus Bio). SMO–EGFP-expressing hRPE1 cells have been described previously24. Stable lines were selected in DMEM/F12 (Life Technologies) containing G418 (400 μg ml−1) added 24 h after transfection. Stable cell lines expressing the protein were enriched by fluorescence-activated cell sorting (FACS) 2–4 weeks after initial selection. For SAG (Calbiochem) stimulation experiments, MEFs were serum-starved at 80% confluency in 3.5-cm dishes or on 12-mm coverslips for 24 h in DMEM/0.2% FCS. MEF cells were stimulated and protein levels were analysed as described17. In brief, MEFs were stimulated with 400 nM SAG for 24 h at 37 °C. Cells were washed ×1 with PBS and lysed directly in 2× sample buffer (Life Technologies). Western blots were developed using chemiluminescence (Super Signal West Dura, Pierce Thermo) and a LAS-3000 imaging system (Fujifilm). Detection of RNA levels was performed as described previously13.


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


Unless otherwise stated, all experimental conditions and methods are described in ref. 5. For the experiments in Fig. 3d, intracellular free [Ca2+] was calculated using the Maxchelator website ( and formulated by titrating CaCl2 in 10 mM BAPTA-Cs buffering conditions. The resting membrane potential measurements from the cell and cilia were made in current-clamp mode. Electrical access was established in the perforated-patch configuration using amphotericin B. Electrodes contained (in mM): 95 K-aspartate, 30 KCl, 10 HEPES, 10 NaCl, 2 MgCl2, 5 EGTA, 100 nM free Ca2+. Extracellular solutions in Fig. 3e contained one of following ratios of NaCl/KCl (mM): 140/5, 135/10, 125/20, 75/70, 5/145; and 10 HEPES, 1 MgCl2 and 1.8 CaCl2. Intraciliary [Ca2+] was equilibrated with the known pipette [Ca2+] in whole-cilia recordings. Note that increasing [Ca2+]cilia inhibits Icilia. The dose–response curve was fit to the Hill equation. The IC50 ([Ca2+] concentration at which current is inhibited 50% = 445 ± 18 nM) and used as a calibration curve. We estimated resting free [Ca2+]cilia by comparing this curve to the current amplitude measured in perforated patches (where internal cilia calcium levels are unperturbed).

Ca2+ imaging in cilia

SMO–GCaMP3 and SMO–mCherry–GCaMP3 were imaged on an Olympus FV1000 at a frame interval of 250 or 64 ms. Ionomycin incorporates into both the plasma membrane and cilia; the ratios of F GCaMP3/F mCherry, reflecting changes in [Ca2+], rise simultaneously in the cytoplasm and cilium. Imaging of [Ca2+] transients from cilium to cytoplasm: SMO–mCherry–GCaMP3 hRPE1 cells were serum-starved for 4–5 days and loaded with Fluo4-AM at 25 °C for 60 min in imaging buffer (150 mM NaCl, 2 mM CaCl2, 1 mM MgCl2, 10 mM glucose, 2 mM HEPES, pH 7.4) and then washed for 10–30 min. Cells with cilia extending beyond the cell body were chosen for experiments to avoid possible induction of cytoplasmic calcium signals by membrane rupture. Images were acquired as sequential scans at a frame interval of 2.20 s with spectral detectors set for GFP and mCherry emission on an Olympus FV1000 equipped with a SIM scanner. PMT and offset were adjusted for each channel such that the 16-bit pixel intensity was in the 1,000 range in the GFP channel and 2,500–3,000 for the mCherry channel before each experiment. Under these settings, GCaMP3 in the GFP channel did not reach pixel saturation after ionomycin stimulation. The ciliary membrane was ruptured using the SIM scanner coupled to a 405-nm laser. The region of interest (ROI) for photobleaching was chosen to cover the tip of the cilium; 405 nm laser intensity was set to 25% and pulse duration set to 2 s. Changes of ΔF/F > 10% for three consecutive scans in the cytoplasm were considered significant. Cytoplasmic [Ca2+] was raised by loading hRPE1 SMO–mCherry–GCaMP3 cells with o-nitrophenyl EGTA AM (NP-EGTA AM) for 30 min at room temperature. Calcium was uncaged in the cytoplasm near the base of the cilium with a SIM scanner by a short (30 ms) 405 laser pulse. Diffusion of cytoplasmic Ca2+ to the cilium was measured by imaging changes in GCaMP3 fluorescence of SMO–mCherry–GCaMP3 in the line-scanning mode—the line scanned the cytoplasm and the proximal side of the cilium. Apparent [Ca2+] diffusion speed was 15.8 ± 1.7 µm s−1 for the cytoplasm versus 14.9 ± 2.1 µm s−1 in the cilium, indicating no significant delay of Ca2+ entering the cilium from the cytoplasm. Representative traces are shown in figures. All images were analysed using ImageJ (NIH) and Origin 8 (OriginLab).

Diffusion coefficient (DCa)

The 405-nm SIM laser was used to rupture the tip of a cilium and allow Ca2+ from the external solution (2 mM) to diffuse along the shaft towards the basal barrier. The cilium is modelled with one-dimensional diffusion assuming a constant supply at the boundary (ruptured tip), with diffusion length given by L = 2(Dt)½. The measured fluorescence increase (GCaMP3) along the ciliary shaft immediately following tip rupture progresses at 4.6 ± 0.6 μm s−1, thus providing an estimate for DCa of 5.3 μm2 s−1. The volume of a cilium is roughly 0.5 fl, containing <200 free Ca2+ ions at 1 μM concentration. We are aware that the number of Ca2+ sensor proteins is thus a significant fraction of Ca2+ ions and will tend to ‘clamp’ reported levels to the Kd of the indicator. For this reason we estimated ciliary [Ca2+] by an independent method that does not depend on calcium-binding proteins as calcium sensors. However, DCa is probably a low estimate, as the Ca2+-sensor protein may be a significant immobile buffer.

Calibration of the ratiometric SMO–mCherry–GCaMP3 sensor

Standard solutions of various [Ca2+] concentrations were prepared ranging from 10 nM to 50 μM by adjusting the ratio of EGTA and CaCl2 in each preparation to clamp-free [Ca2+] at the desired value. The solutions contained 137 mM NaCl, 5.4 mM KCl, 10 mM HEPES, 5 mM EGTA and CaCl2 ranging from 0 to 5.04 mM (corresponding to 50 μM free [Ca2+]).

For imaging, hRPE1 cells were plated onto 12-mm glass coverslips, serum-starved for 4–5 days to allow for cilia formation, and imaged in the various standard solutions following digitonin membrane permeabilization. Briefly, coverslips were washed twice in Ca2+-free solution to remove residual Ca2+ and incubated in the Ca2+ standard solution. The samples were then imaged using an Olympus Fluoview FV1000 laser point-scanning confocal microscope (60× water immersion, 1.2 N.A. objective) with spectral detectors set up for optimal detection of GFP and mCherry fluorescence with sequential excitation with 488 nm and 543 nm lasers, respectively. The settings were adjusted for a GCaMP3 signal in the 1,000 sub-range and mCherry signal in the 2,000 sub-range of the 16-bit intensity range. Identical settings were used for all Ca2+ standard solutions. Subsequently, the cells were permeabilized on the microscope stage by addition of an identical volume (0.5 ml) of 32 µM digitonin dissolved in the same Ca2+ standard, resulting in a final concentration of 16 µM digitonin. Images were acquired for multiple fields of view after allowing permeabilization to occur for 1 min. Cytoplasmic [Ca2+] was measured in RPE cells that had not formed a cilium and thus had significant levels of SMO–mCherry–GCaMP3 protein in the plasma membrane.

To obtain the standard calibration curve, the acquired images were processed with ImageJ in the following way. Briefly, after background subtraction, the images were thresholded in the mCherry channel to only take into account pixels with a minimum expression level of the sensor (a threshold of 20× background signal was generally used). Dividing the GCaMP3 fluorescence image by the mCherry channel intensity generated ratio images. ROIs exclusively located in cilia of multiple cells were selected, and the average ratios measured for multiple cells and coverslips were reported. The average ratios obtained were plotted as a function of free [Ca2+] and the resulting points were fitted with a sigmoid curve. Images in Fig. 3a–d were acquired using the same acquisition settings as in Extended Data Fig. 5.

For quantification of GLI2 at ciliary tips, MEFs isolated from Arl13b EGFPtg and Pkd2l1−/− × Arl13b EGFPtg mice were stained with anti-GLI2 antibody25. Cilia were outlined in ImageJ based on the EGFP signal. Mean background signal in the GLI2 channel was determined using ImageJ, multiplied by 1.2 and subtracted from the image. Integrated fluorescence intensity was measured within the ciliary outline. GLI2 quantification was compiled from measurements of 40 cilia from MEFs each isolated from three wild-type and five Pkd2l1−/− embryos.

Data analysis

Group data are presented as mean ± s.e.m. Statistical comparisons were made using unpaired t-tests (Origin 8). Statistical significance is denoted with asterisk (*P < 0.05; **P < 0.01).


  1. 1

    Corbit, K. C. et al. Vertebrate Smoothened functions at the primary cilium. Nature 437, 1018–1021 (2005)

  2. 2

    Huangfu, D. & Anderson, K. V. Cilia and Hedgehog responsiveness in the mouse. Proc. Natl Acad. Sci. USA 102, 11325–11330 (2005)

  3. 3

    Goetz, S. C. & Anderson, K. V. The primary cilium: a signalling centre during vertebrate development. Nature Rev. Genet. 11, 331–344 (2010)

  4. 4

    Praetorius, H. A. & Spring, K. R. Bending the MDCK cell primary cilium increases intracellular calcium. J. Membr. Biol. 184, 71–79 (2001)

  5. 5

    DeCaen, P. G., Delling, M., Vien, T. N. & Clapham, D. E. Direct recording and molecular identification of the calcium channel of primary cilia. Nature (this issue)

  6. 6

    Hama, H. et al. Scale: a chemical approach for fluorescence imaging and reconstruction of transparent mouse brain. Nature Neurosci. 14, 1481–1488 (2011)

  7. 7

    Tian, L. et al. Imaging neural activity in worms, flies and mice with improved GCaMP calcium indicators. Nature Methods 6, 875–881 (2009)

  8. 8

    Yasuda, R. et al. Imaging calcium concentration dynamics in small neuronal compartments. Sci. STKE 2004, pl5 (2004)

  9. 9

    Soler-Llavina, G. J. & Sabatini, B. L. Synapse-specific plasticity and compartmentalized signaling in cerebellar stellate cells. Nature Neurosci. 9, 798–806 (2006)

  10. 10

    Clapham, D. E. Calcium signaling. Cell 131, 1047–1058 (2007)

  11. 11

    Vogel, P. et al. Situs inversus in Dpcd/Poll−/−, Nme7−/−, and Pkd1l1−/− mice. Vet. Pathol. 47, 120–131 (2010)

  12. 12

    Field, S. et al. Pkd1l1 establishes left-right asymmetry and physically interacts with Pkd2. Development 138, 1131–1142 (2011)

  13. 13

    Horio, N. et al. Sour taste responses in mice lacking PKD channels. PLoS ONE 6, e20007 (2011)

  14. 14

    Martin, V. & Shaw-Smith, C. Review of genetic factors in intestinal malrotation. Pediatr. Surg. Int. 26, 769–781 (2010)

  15. 15

    Ramalho-Santos, M., Melton, D. A. & McMahon, A. P. Hedgehog signals regulate multiple aspects of gastrointestinal development. Development 127, 2763–2772 (2000)

  16. 16

    Humke, E. W., Dorn, K. V., Milenkovic, L., Scott, M. P. & Rohatgi, R. The output of Hedgehog signaling is controlled by the dynamic association between Suppressor of Fused and the Gli proteins. Genes Dev. 24, 670–682 (2010)

  17. 17

    Keady, B. T. et al. IFT25 links the signal-dependent movement of Hedgehog components to intraflagellar transport. Dev. Cell 22, 940–951 (2012)

  18. 18

    Kim, J., Kato, M. & Beachy, P. A. Gli2 trafficking links Hedgehog-dependent activation of Smoothened in the primary cilium to transcriptional activation in the nucleus. Proc. Natl Acad. Sci. USA 106, 21666–21671 (2009)

  19. 19

    Bhogaraju, S., Taschner, M., Morawetz, M., Basquin, C. & Lorentzen, E. Crystal structure of the intraflagellar transport complex 25/27. EMBO J. 30, 1907–1918 (2011)

  20. 20

    Zhao, Y. et al. An expanded palette of genetically encoded Ca2+ indicators. Science 333, 1888–1891 (2011)

  21. 21

    Miki, K. & Clapham, D. E. Rheotaxis guides mammalian sperm. Curr. Biol. 23, 443–452 (2013)

  22. 22

    Gibbs, D. & Williams, D. S. Isolation and culture of primary mouse retinal pigmented epithelial cells. Adv. Exp. Med. Biol. 533, 347–352 (2003)

  23. 23

    Preibisch, S., Saalfeld, S. & Tomancak, P. Globally optimal stitching of tiled 3D microscopic image acquisitions. Bioinformatics 25, 1463–1465 (2009)

  24. 24

    Kim, J. et al. Functional genomic screen for modulators of ciliogenesis and cilium length. Nature 464, 1048–1051 (2010)

  25. 25

    Tukachinsky, H., Lopez, L. V. & Salic, A. A mechanism for vertebrate Hedgehog signaling: recruitment to cilia and dissociation of SuFu-Gli protein complexes. J. Cell Biol. 191, 415–428 (2010)

Download references


We thank the Mouse Gene Manipulation Facility of the Boston Children’s Hospital Intellectual and Developmental Disabilities Research Center (IDDRC; NIHP30-HD 18655). We thank the Image and Data Analysis Core (IDAC) at Harvard Medical School for help with three-dimensional reconstruction. We also thank A. Salic for anti-SMO and anti-GLI2 antibodies. P.G.D. was supported by NIH T32-HL007572. We thank M. Desai for graphical assistance and A. von Gise and H. Tukachinsky (Harvard Medical School) and the members of the Clapham laboratory for advice and discussion.

Author information

All authors designed or conducted experiments and wrote the manuscript.

Correspondence to David E. Clapham.

Ethics declarations

Competing interests

The authors declare no competing financial interests.

Extended data figures and tables

Extended Data Figure 1 ARL13B–EGFP identifies primary and motile cilia in transgenic mouse tissue.

al, Frozen tissue sections were prepared from P0 intestine (ac); E15 nasal cavity (df); 6-week-old hippocampus (gi); P0 retina from Arl13b-EGFPtg mice (jl). First column, Arl13b-EGFPtg (green) fluorescence; second column (red) labels villin in b, ACIII in e, h, or HMB-45 (human melanoma black antibody, retinal pigmented epithelial cells) in k. ac, Intestine: EGFP fluorescence in a does not overlap with the anti-villin staining in b as shown in the merged image in c, indicating that ARL13B–EGFP is absent from microvilli (arrows). Scale bar: 100 μm. df, Nasal cavity: (d) several ARL13B–EGFP-positive cilia (arrows) that face the lumen of the turbinate co-localize with ACIII (e) as shown in the merged image (f). Scale bar: 100 μm. Inset shows magnification of nasal cavity surface. Scale bar of inset in f, 5 μm. gi, CA1 region of hippocampus: (g) prominent cilia (arrows) overlap with ACIII staining (h) as shown in the merged image (i). Scale bar: 50 μm. Inset shows magnification of hippocampal cilia. Inset scale bar: 5 µm. Merged red and green channels were offset for clarity (inset). jl, Retinal pigmented epithelia (RPE): (j) short cilia (arrow) are visible at the intersection between RPE (labelled with HMB-45 antibody) and (k) developing photoreceptor cells, as shown in the merged image (l). Scale bar: 10 μm. Inset shows magnification of RPE–photoreceptor interface. Inset scale bar: 2 μm.

Extended Data Figure 2 ARL13B–EGFP labels a primary cilium in cultured MEF and RPE cells.

al, The first column shows ARL13B–EGFPtg (green) fluorescence; second column labels ACIII (b, e) or acetylated tubulin (h, k). Third column: merged red and green channels were offset for clarity. Primary MEFs of Arl13b-EGFPtg (ac) and wild-type (df) mice isolated from E14.5 embryos. ARL13B–EGFP co-localizes with ciliary ACIII in c. MEFs isolated from wild-type mice show no fluorescence in the cilium (488 nm excitation). gi, Primary RPE cells isolated from P12 Arl13b-EGFPtg mice. Cells were fixed and stained with antibody to acetylated tubulin. ARL13B–EGFP (g) exclusively localized to the primary cilium identified by antibody to acetylated tubulin (h, i). jl, Stable cell line (hRPE1) expressing SMO–EGFP. After 2 days of serum starvation, SMO–EGFP labelled the primary cilium of hRPE1 cells (j) as indicated by acetylated tubulin labelling (k, l). Scale bars: c, f, i, l, 5 μm. m, n, To determine whether ARL13B–EGFP expression adversely affected ciliogenesis, ciliary length and per cent of cells with cilia were quantified from wild-type and ARL13B–EGFP-expressing MEFs stained by anti-acetylated tubulin. m, Ciliary length was similar in wild-type and ARL13B–EGFP-expressing MEFs (2.6 ± 0.5 μm versus 2.9 ± 0.8 μm, respectively, n = 200). n, The number of cells with cilia was also comparable between wild-type and ARL13B–EGFP-expressing MEFs (60.2% ± 5.1% versus 65.5% ± 7.3%; n = 120).

Extended Data Figure 3 Ciliary [Ca2+] changes in stably transfected RPE cell cilia.

ac, Live hRPE1 cells stably expressing SMO–GCaMP3 were treated with 5 μM ionomycin. Fluorescence increases were measured in the cilium and cytoplasm. Image in c is DIC. d, Changes in fluorescence (ΔF/F) of the calcium sensor, GCaMP3, are plotted for both cytoplasm and cilium. eh, hRPE1 cells expressing SMO–mCherry–GCaMP3 were stained with acetylated tubulin. GCaMP3 (e) and mCherry (f) fluorescence overlaps with acetylated tubulin staining (g, h). Scale bar: 5 μm; merged channels were offset for clarity. i, The tip of a cilium was ruptured with an intense 1-s laser pulse (405 nm, hRPE1 cell expressing SMO–mCherry–GCaMP3). Circle indicates area of rupture. Numbered arrowheads indicate positions where changes in fluorescence were measured. Asterisk indicates mCherry fluorescence outside of cilium, indicating that some SMO–mCherry–GCaMP3 is retained in the ER. j, Quantification of changes in fluorescence at the positions marked in i. Rupture of the ciliary membrane rapidly increases [Ca2+]cilia at the tip and travels along the cilium at 4.6 ± 0.6 µm s−1 (n = 16).

Extended Data Figure 4 [Ca2+]cilia calibration.

a, Images of hRPE1 cells stably expressing SMO–mCherry–GCaMP3 were acquired after permeabilization with 15 μM digitonin in varying extracellular [Ca2+]. b, Averages of several ratios (n = 12–16; ±s.d.) per concentration are plotted against free [Ca2+], yielding the calibration curve for SMO–mCherry–GCaMP3: Kd = 625 nM.

Extended Data Figure 5 Overexpressed and endogenous PKD2L1 localizes to the primary cilium.

ac, Rabbit anti-PKD2L1 (Thermo Scientific) recognizes overexpressed PKD2L1. HEK cells were transfected with hPKD2L1-IRES mCherry construct and stained with PKD2L1 antibody. PKD2L1 staining (a) is specific to cells that also express mCherry (b). c, Overlay. df, Overexpressed hPKD2L1 localizes to the primary cilium in mIMCD3 cells. mIMCD3 cells were transfected with HA-tagged hPKD2L1 and stained with an anti-HA antibody (d) and anti-acetylated tubulin antibody (e). HA immunoreactivity is visible both in the cytoplasm and in the cilium. f, Overlay. g, h, PKD2L1 antibody labels the primary cilium of mIMCD3 cells. g, Confluent mIMCD3 cells were stained with anti-PKD2L1 antibody used in a; and h, acetylated tubulin antibody to label cilia. PKD2L1 immunoreactivity is visible in the primary cilium. i, Overlay. jo, Primary MEFs of Arl13b-EGFPtg (jl) and Arl13b-EGFPtg × Pkd2l1−/− mice (mo) isolated from E14.5 embryos were stained with anti-PKD2L1 antibody used in a. In Arl13b-EGFPtg MEFs, PKD2L1 immunoreactivity (j) co-localizes with ARL13B–EGFP signal (k) labelling the primary cilium. l, Overlay. mo, PKD2L1 staining is absent in cilia of Arl13b-EGFPtg × Pkd2L1−/− mice. Scale bars: 10 μm.

Extended Data Figure 6 [Ca2+]cilia increases 24 h after SAG stimulation.

a, MEFs expressing ARL13B–mCherry–GECO1.2 were stained with acetylated tubulin. GECO1.2 and mCherry fluorescence overlaps with acetylated tubulin staining. Scale bar: 5 μm; merged channels were offset for clarity. b, Ratio maps of MEFs isolated from ARL13B–mCherry–GECO1.2 mice stimulated with 0.05% DMSO (left) or 400 nM SAG (right) for 24 h. Scale bar: 5 μm. c, Quantification of ciliary GECO1.2/mCherry ratios obtained for MEFs with and without SAG stimulation. Ratio increases from 0.4 ± 0.05 to 0.8 ± 0.2 after SAG stimulation (*P < 0.05; n = 20–30 cilia). d, Example ciliary current measured from MEFs treated with 500 nM SAG (SMO agonist) or with DMSO vehicle (0.05%) in culture for 24–36 h in control conditions and after activation with 10 μM calmidazolium (CMZ). e, Scatter and whisker (±s.d.) plots from cilia show total outward (+100 mV) and inward (−100 mV) current measured for both treatment groups. Averages are indicated by the thick horizontal lines and individual cilium current magnitudes are represented as circles. P values resulting from Student’s t-test comparing treatment groups are indicated (*< 0.05; n = 11 cilia).

Supplementary information

Oviductal motile cilia from an 8-week-old Arl13B-EGFPtg mouse

The oviduct was isolated after hormone treatment (Methods), opened longitudinally, and mounted on a coverslip. Motile cilia were imaged at a frame rate of 65 ms. Scale bar, 5 µm. Cilia sweep oviductal debris from left to right. (MOV 3935 kb)

3D projection of cilia expressing Arl13B-EGFP

Rotating 3D projection of the middle digit of an E14.5 Arl13B-EGFPtg mouse paw showing all cilia in this developing tissue. Scale bar, 200 µm. (MOV 94206 kb)

Smo-GCaMP3 reports [Ca2+] changes in cilium and cytoplasm

Addition of 5 µM Ionomycin to Smo-GCaMP3 expressing hRPE1 cell increases fluorescence in the cilium and cytoplasm. Scale bar, 5 µm. Recorded at 250 ms per frame. (MOV 932 kb)

: Laser rupture of a cilium expressing Smo-mCherry-GCaMP3

The tip of a cilium was ruptured by a 1-2 s pulse from a 405 nm laser. Calcium influx is visible at the right tip of the cilium. The resulting increase in [Ca2+] diffuses along the cilium. From a Smo-mCherry-GCaMP3-expressing hRPE1 cell recorded at 65 ms per frame. Scale bar, 5 µm. (MOV 1783 kb)

Ciliary [Ca2+] can increase independently of the cytoplasm

The cytoplasm of a Smo-mCherry-GCaMP3-expressing hRPE1 cell was loaded with Fluo-4 Ca2+ indicator (Fluo-4-AM, 10 µM) and fluorescence measured in the cilium and cytoplasm. The tip of a cilium was then ruptured by a 1-2 s pulse from a 405 nm laser. 2 mM [Ca2+] entering the cilium overwhelms the ciliary buffers, saturates the ciliary Smo-mCherry-GCaMP3 reporter, and after an ~ 40 s delay, barely increases cytoplasmic [Ca2+] (t= 59 s). Ionomycin added to the bath at 1 min 20 s results in a very large increase in the cytoplasm. Recorded at 1.1 s per frame. Scale bar, 5 µm. (MOV 413 kb)

Cytoplasmic [Ca2+] rises propagate into the cilium

The cytoplasm of a Smo-mCherry-GCaMP3-expressing hRPE1 cell was loaded with o-Nitrophenyl EGTA AM (NP-EGTA AM, 10 µM) and fluorescence measured in the cilium and cytoplasm. Both GCaMP3 (left) and mCherry (right) fluorescence are shown. Calcium was uncaged in the cytoplasm by a 1 s pulse from a 405 nm laser. A calcium wave travels inside the cytoplasm and enters the cilium. Scale bar, 5 µm. (MOV 5836 kb)

PowerPoint slides

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Delling, M., DeCaen, P., Doerner, J. et al. Primary cilia are specialized calcium signalling organelles. Nature 504, 311–314 (2013) doi:10.1038/nature12833

Download citation

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.