Hsp90α forms a stable complex at the cilium neck for the interaction of signalling molecules in IGF-1 receptor signalling

Primary cilia, antenna-like microtubule-based cellular protrusions found in almost every type of mammalian cell, emerge from the distal end of the mother centriole that lies in close proximity to the plasma membrane in growth-arrested cells. They are sensory organelles that coordinate and regulate signalling pathways involved in cell cycle control, cellular differentiation and other cellular activities such as tissue homeostasis or embryonic development (Irigoín and Badano, 2011; Christensen et al., 2012; Badano et al., 2006). They can act as both mechanosensors and chemosensors that detect a range of extracellular stimuli and modulate and integrate these stimuli with the intracellular signalling cascade. Such stimuli include the mechanical force of fluid flow (Schwartz et al., 1997; Praetorius and Spring, 2001; Praetorius and Spring, 2003), hedgehog and Wnt signals (Corbit et al., 2008; Rohatgi et al., 2007), G-protein-coupled olfactory receptor signals (Boekhoff et al., 1990) and receptor tyrosine kinase signals (Schneider et al., 2005; Ma et al., 2005; Teilmann and Christensen, 2005; Zhu et al., 2009).

Receptor tyrosine kinases belong to a superfamily of high-affinity single-pass transmembrane receptors for growth factors and hormones. The binding of ligand to the extracellular domain activates the intracellular kinase domain that catalyses autophosphorylation and tyrosine phosphorylation of adaptors and effectors. Many receptors rely on the primary cilium to transmit their signals to intracellular downstream molecules (Schneider et al., 2005; Ma et al., 2005; Teilmann and Christensen, 2005; Zhu et al., 2009). In the case of IGF-1 receptor signalling, the activation of the IGF-1 receptor in the axoneme membrane leads to the phosphorylation of its intracellular effectors, IRS-1 and Akt-1, at the base of the cilium (Zhu et al., 2009). This receptor and effector coordination along the axoneme-basal body axis forms a signal transduction pathway in which intraflagellar transport can facilitate the interaction of the receptor with its effectors.

Between the distal end of mother centriole in the basal body and the base of axoneme, the nine triplet microtubules in the mother centriole become the nine axonemal doublet microtubules in the cilium (Irigoín and Badano, 2011; Christensen et al., 2012). This transition neck region contains the ciliary necklace and transition fibres (Christensen et al., 2012). It is a unique region that is required for strengthening the cilium and for gating protein transport. Cilia excision, which occurs in response to environmental stresses such as heat shock, usually occurs at this transition neck region (Quarmby, 2004). The rapid resorption of primary cilia that is induced by heat shock is regulated, at least in part, by the 90-kDa heat shock protein, Hsp90 (Prodromou et al., 2012).

Heat shock proteins are molecular chaperones that help nascent peptides to fold correctly and facilitate the productive assembly of multimeric proteins. Hsp90 is a conserved molecular chaperone that targets specific client proteins involved in signal transduction, protein refolding, protein degradation, etc. (Pratt and Toft, 2003; Flandrin et al., 2008; Zhao et al., 2005; Jackson, 2013). For example, it has been reported that Hsp90 can bind to Akt to protect it from phosphatase 2A (PP2A)-catalysed dephosphorylation and to regulate the kinase activity (Sato et al., 2000; Basso et al., 2002). If the binding of the two proteins is blocked, Akt is inactivated (Sato et al., 2000).

Microtubules are highly organised protein complexes. The interaction of Hsp90 with microtubules occurs at several levels. Hsp90 interacts with tubulin dimers to protect tubulin against thermal denaturation and to maintain it in a state compatible with further assembly (Weis et al., 2010). Hsp90 is also a core component of the centrosome that acts to recruit other centrosome-associated proteins and ensure their proper functioning (Lange et al., 2000; Basto et al., 2007; de Cárcer et al., 2001). In this study, we report that Hsp90α at the ciliary base is an important chaperone protein for stabilisation of the cilium and for cilium-mediated IGF-1 receptor signalling. Hsp90α protein localises at the periciliary base and the phosphorylated Hsp90α forms a ring structure around the ciliary neck region. To mediate IGF-1 receptor signalling, IRS-1 interacts with Akt-1 at the transition neck region marked by phosphorylated Hsp90α.

An antibody raised against the phosphorylated peptide CVLLA(pS)KEN(pS)APVK of the CASK protein was found to stain a phosphorylated protein at the neck region of primary cilia in 3T3-L1 cells as well as motile cilia in murine tracheal epithelia (supplementary material Fig. S1). This phosphorylated protein was purified from mouse trachea and identified by mass spectrometry as Hsp90. The peptides detected by mass spectrometry matched both Hsp90α and Hsp90β with 18.65% and 15.55% protein coverage, respectively (supplementary material Fig. S1).

Hsp90α and Hsp90β share 86% amino acid sequence identity (Sreedhar et al., 2004). However, their cellular expression patterns are very different. Hsp90β is constitutively and ubiquitously expressed, whereas Hsp90α is less abundant and is heat inducible (Sreedhar et al., 2004). There are several phosphorylation sites that are unique to each protein, e.g. the phosphorylated Thr5 and Thr7 in Hsp90α and the phosphorylated Ser226 in Hsp90β. To determine whether the identified protein was Hsp90α or β, antibodies against Hsp90α phosphorylated at Thr5 and Thr7 [pHsp90α(T5/7)] or against Hsp90β phosphorylated at Ser226 [pHsp90β(S226)] were used to perform immunofluorescent staining. As shown in Fig. 1 and supplementary material Fig. S1, anti-pHsp90α(T5/7) antibody detected the same protein at the ciliary neck region as the antibody against the CASK phospho-peptide. By contrast, anti-pHsp90β(S226) antibody stained a different protein from the original antibody (supplementary material Fig. S2). Furthermore, antibody specific for the Hsp90α isoform stained the region around the periciliary base, whereas antibody specific for the Hsp90β isoform showed no connection to the cilium (Fig. 1F; supplementary material Fig. S2).

Using super-resolution microscopy and three-dimensional (3D) reconstruction, we observed that the phosphorylated Hsp90α formed a ring around the axoneme at the neck region, just above the basal body (Fig. 2A,B,G). This result was confirmed by immunoelectron microscopy (Fig. 2C,D). Cep164 is a protein found previously to form a ring around the ciliary neck (Graser et al., 2007). Double immunofluorescent staining of pHsp90α and Cep164 showed that both proteins localised at the same area of the transition neck region (Fig. 2E,F).

The physical localisation of Hsp90α at the base of the cilium implied its involvement in cilia function or formation. By inhibiting the chaperone activity of Hsp90, we investigated the possible functions of Hsp90α in cilia formation. Hsp90 cycles through several conformations for its chaperone activity (Trepel et al., 2010). Geldanamycin, a natural inhibitor of Hsp90 ATPase activity, blocks the Hsp90 chaperone cycle and inhibits its chaperone activity (Trepel et al., 2010; Wang et al., 2009). A high concentration of geldanamycin (7.2 µM) induced rapid resorption of cilia, but also exhibited a high level of cellular toxicity (supplementary material Fig. S3). It was not used in subsequent experiments. At a low geldanamycin concentration (90 nM), cells could survive for a longer period of time. However, no cilia resorption was observed; instead, the cilia were elongated (Fig. 3A,B). Furthermore, prolonged exposure to a low concentration of geldanamycin (48 hours at 90 nM) dramatically altered cellular microtubules and tubulin acetylation (supplementary material Fig. S3). Although geldanamycin is an inhibitor of both Hsp90α and Hsp90β and cannot specifically target the α isoform, it is only Hsp90α that is related to cilia. It is possible that these effects on cilia were the results of the inhibition of Hsp90α chaperone activity.

Because the primary cilia were already formed in confluent cells, the treatment of geldanamycin affected the existing primary cilia but not the process of cilium formation. To analyse the effect of geldanamycin on cilium formation, cells were treated with both geldanamycin and serum starvation. Under these conditions, the cilia that were formed were, on average, 40% longer and had more misshapes (twists or kinks) than the cilia of cells cultured under control conditions (Fig. 3D–G). Without the chaperone activity to facilitate microtubule assembly during cilium formation, it is possible that some of the microtubule doublets were affected. Although the chaperone activity was inhibited by geldanamycin, the localisation of Hsp90α at the periciliary base or of pHsp90α at the ciliary neck region was not affected by the inhibitor (Fig. 3D,H).

The increased Hsp90α phosphorylation observed on western blots of protein from geldanamycin-treated cells was most likely caused by a subset of cells with high levels of Hsp90α phosphorylation (Fig. 3C; supplementary material Fig. S3). It could be the case that geldanamycin induced an apoptotic response, because rapid Hsp90α phosphorylation at Thr5 and Thr7 is induced in apoptotic cells (Solier et al., 2012). In our immunofluorescence analysis, this subset of cells was excluded.

Another way to alter heat shock protein activity is by applying heat shock. When confluent cells were shifted to 42°C, most Hsp90α around the periciliary base was dispersed within 5 minutes, and cilia were absorbed within 15 minutes (Fig. 3I). It appeared that primary cilia were destabilised after Hsp90α from around the base of the cilium was dispersed. In theory, heat shock activates chaperone activity – the opposite effect to that induced by geldanamycin treatment. Its effect on Hsp90α phosphorylation and tubulin acetylation was also the opposite of that observed with geldanamycin (Fig. 3J). Taken together, both the inhibition and activation of Hsp90 chaperone activity affected formation and stabilisation of cilia (Fig. 3C,D,H–J). Although the functional mechanism requires further study, Hsp90α chaperone activity might facilitate rapid protein assembly and disassembly during the formation of cilia.

Even after the cilium was completely absorbed, phosphorylated Hsp90α was still present at the ciliary remnant during heat shock treatment (Fig. 3I). As Cep164 localised to the same place (Fig. 2E,F), the two proteins represented a stable complex at the ciliary neck region. In cilium-mediated IGF-1 receptor signalling, the phosphorylated receptor is present in the axoneme and its phosphorylated signalling molecules, IRS-1 and Akt-1, are detected at the ciliary base (Zhu et al., 2009). The identification of the protein complex at the neck region raised the question of whether IRS-1 and Akt-1 associated with basal body or the neck region protein complex. As shown in Fig. 4A,E, it was clear that Akt-1 was activated by the receptor signal at the ciliary neck region and not the basal body per se.

Hsp90 has been reported to interact with Akt-1 (Sato et al., 2000; Basso et al., 2002). Their colocalisation at the ciliary neck region indicated a chaperone platform at the ciliary neck for cilium-mediated signalling. It implied that the colocalised chaperone could facilitate the interaction between signalling molecules, as the phosphorylation of Akt-1 at the neck complex was sensitive to geldanamycin (Fig. 4C,D). This is in line with the observation that the inhibition of Hsp90 chaperone activity by geldanamycin or other inhibitors decreases Akt-1 protein levels by 80% (Basso et al., 2002). Our current study suggested that the primary cilium is an important platform for the interaction of these two proteins. In addition, Erk1/2, which is not detected in primary cilia during IGF-1 receptor signalling, was less affected by geldanamycin than was Akt-1 (Fig. 4B).

The activation of the Akt-1 signalling pathway by the IGF-1 receptor is mediated by IRS-1, a docking protein capable of recruiting many intracellular signalling molecules (Baserga, 2009; Boller et al., 2012). IRS-1 is found to be recruited to the basal body (Zhu et al., 2009). In 60% of cells, IRS-1 specifically or preferentially associated with the basal body instead of the centrosome (Fig. 5A,B). The distinction was reduced in geldanamycin-treated cells, and 70% of geldanamycin-treated cells had the same level of IRS-1 on both the basal body and the centrosome (Fig. 5A,B). Although both of these structures could recruit IRS-1, only the IRS-1 on the basal body could be phosphorylated by the IGF-1 receptor (Fig. 5C,D). In primary cilia, IRS-1 associated with the basal body, whereas the phosphorylated IRS-1 associated with the neck region (Fig. 5F). The receptor, which was localised in the axoneme, could only interact with proteins transported from the basal body. Because there was no Hsp90α chaperone in the axoneme, the activation of IRS-1 by the IGF-1 receptor was not sensitive to geldanamycin (Fig. 5C,D).

As IRS-1 is the docking site for many signalling molecules (Baserga, 2009; Boller et al., 2012), its localisation at the basal body could recruit signalling molecules into proximity with the receptor in the axonemal membrane. The association of IRS-1 with the basal body was mediated by its C-terminal region (amino acids 816–1231) (Fig. 6A,B,D). IRS-1 associated with the basal body, not the ciliary neck region (Fig. 6C). The association of IRS-1 with the basal body was also independent of IRS-1 phosphorylation, as a Tyr(1171/1218)Phe mutation (CT-2F) had no effect on its basal body association (Fig. 6A,B). Most importantly, deletion of the C-terminal region (Δ816-1231) dissociated the mutant from the basal body and enabled its transport into the axoneme (Fig. 6B). Apparently, IRS-1 contained multiple motifs to regulate its association with the basal body as well as its transport into the axoneme. In fact, at an early stage of insulin stimulation, phosphorylated IRS-1 could be detected in the axoneme (Fig. 5E). With prolonged insulin stimulation, more and more IRS-1 was phosphorylated and accumulated at the neck region, causing IRS-1 signal at the neck region to overshadow the weak signal in the axoneme (Fig. 5B). Given these findings, we postulated that IRS-1 that was initially localised at the basal body was transported by intraflagellar transport to the axoneme, where it was activated by the receptor tyrosine kinase. The phosphorylated IRS-1 was then transported back to the neck region protein complex marked by Hsp90α and Cep164, and then activated downstream signalling molecules, such as the Akt-1 signalling pathway.

The region separating the basal body and axoneme is called the transition neck, in which the transition fibres link the basal body to the plasma membrane and the ciliary necklace links the axoneme to the ciliary membrane (Christensen et al., 2012). The transition fibres and ciliary necklace act as roots to reinforce the ciliary base as well as a ‘doorkeeper’ for proteins and molecules being transported into or out of the cilium. Given the geometric restriction of the neck region, it is important to keep all these proteins in order. Molecular chaperones are proteins that keep other proteins correctly folded and allow multimeric complexes to be assembled productively, while minimising aggregation in a protein-rich environment (Jackson, 2013). The localisation of Hsp90α around the periciliary base and its phosphorylated form at the neck region indicates the involvement of a molecular chaperone in cilia physiology (Figs 1, 2). All the damaging shearing or twisting forces that act upon the axoneme are eventually received at the base of the cilium. The clustering of a molecular chaperone at its base could stabilise the basal protein assembly and facilitate reassembly of these protein complexes. In fact, we found that the association of Hsp90α with the basal body and centrosome was clearly related to protein assembly activity (supplementary material Fig. S4). The high microtubule assembly activity during prophase and metaphase spindle formation corresponded to the greatest clustering of Hsp90α at the spindle poles. This clustering was dissipated as the spindle was disassembled in anaphase and telophase. In quiescent cells, no Hsp90α localised at the centrosome, it all surrounded the ciliary basal body (supplementary material Fig. S4).

The phosphorylated Hsp90α was also found to be associated with the spindle poles and then with the mother centrosome in the daughter cells (supplementary material Fig. S4). It was an appendix protein specific for the mother centrosome and was absent in the newly formed daughter centrosome (supplementary material Fig. S4). In view of its geometric localisation in cilia, the phosphorylated Hsp90α is likely to be a component of the distal appendages in the mother centriole from which cilium emerges (Fig. 2).

Hsp90 interacts with a wide range of proteins, including many signalling molecules, which play an essential role in regulating cell survival, cell cycle and hormone signalling (Zhao et al., 2005; Jackson, 2013; Echeverría et al., 2011). The binding of these proteins to Hsp90 is required for their stability and functional maturation (Pratt and Toft, 2003; Sato et al., 2000; Basso et al., 2002). In cilium-mediated cell signalling, all sorts of extracellular signals are received by the axoneme on the cell surface, and these signals converge on the ciliary base for signal relay or crosstalk between signalling pathways (Christensen et al., 2012). The presence of Hsp90α at the ciliary base forms a hub at which intracellular signalling molecules are gathered and stabilised, creating opportunities for these molecules to interact with one another. In IGF-1 receptor signalling (Zhu et al., 2009), the phosphorylated Hsp90α forms a platform on which IRS-1 mediates the activation of the PI3 kinase–Akt-1 pathway by the IGF-1 receptor (Figs 4, 5). The multiple signalling steps from the receptor to Akt-1 include the phosphorylation of IRS-1 by the receptor kinase and the activation of PI3 kinase to generate phosphatidylinositol phosphates that recruit Akt-1 to be activated by phospholipid-dependent kinase 1 (Chitnis et al., 2008). In this signal transduction pathway, IRS-1 serves as a secondary messenger between the receptor in the axonemal membrane and Akt-1 at the neck region. IRS-1 is transported from the basal body to the axoneme, phosphorylated by the receptor tyrosine kinase in the axonemal membrane and then transported back to the transition neck region where it activates the PI3 kinase–Akt-1 pathway (Figs 4–f05,6). These observations indicate that the signalling molecules are activated at the ciliary neck, not the basal body. Hsp90α is most likely to serve as a molecular chaperone to stabilise signalling molecules and facilitate their interaction.

Because it is rather difficult to biochemically separate Hsp90 isoforms (Hsp90α and Hsp90β), many studies are carried out without strictly differentiating between the two. However, the viable Hsp90α-knockout mice and embryonic-lethal Hsp90β-knockout mice clearly indicate that these two highly similar isoforms have very different physiological functions (Voss et al., 2000; Grad et al., 2010). In mice lacking Hsp90β, the defective allantois–trophoblast interaction impairs the development of the placental labyrinth and causes embryonic lethality (Voss et al., 2000). Hsp90α-knockout mice, in which the Hsp90β levels are unchanged, develop normally (Grad et al., 2010). The lack of Hsp90α only affects spermatogenesis in male mice, causing meiosis I arrest at the pachytene stage. In our current study, Hsp90α was found to localise at the periciliary base as well as at the transition neck region (Figs 1, 2). In fact, the sperm tail is essentially a specialised cilium. It is possible that Hsp90α is also involved in the development of sperm tails. The different cellular localisations and functions of Hsp90α and β suggest that they should be studied separately as two distinct proteins.

Antibodies against Hsp90α (ab74248), Hsp90β (ab51145), pHsp90β (S226) (ab63562) and α-tubulin (ab15246) were from Abcam. Anti-pHsp90α(T5/7) (#3488) antibody was from Cell Signaling Technology. Antibodies against pIRS-1(Y632) or (Y989) (sc-17196 or sc-17200), IRS-1 (sc-7200), Hsp90α/β (sc-7947), Akt-1 (sc-5298), pAkt-1(S473) (sc-7985), Cep164 (sc-240226), acetylated α-tubulin (sc-23950), Erk (sc-94) and pErk(Y204) (sc-101761) were from Santa Cruz Biotechnology. Anti-γ-tubulin (T6557) and anti-GAPDH (G8795) antibodies, horseradish-peroxidase-conjugated secondary antibody, geldanamycin, insulin and DAPI were from Sigma. Alexa-Fluor-488- or Alexa-Fluor-546-conjugated secondary antibodies were from Molecular Probes (Invitrogen). The Leica laser-scanning confocal microsystem, including the Leica TCS SP2 confocal microscope, Leica confocal scanner and Leica confocal acquisition software, was used with the HCX PL APO 1bd. BL 63.0×/1.4 oil objective at 1.4 numerical aperture at a working temperature of 22°C. The fluorescence medium used was DABCO (Sigma).

For cell staining, cells on coverslips were fixed with 4% paraformaldehyde in phosphate buffered saline (PBS) at room temperature for 10 minutes or in 100% methanol at −20°C for 5 minutes. For tissue staining, mouse trachea was fixed with 4% paraformaldehyde in PBS overnight, dehydrated with 30% sucrose in PBS at 4°C overnight, embedded in Jung Tissue Freezing Medium (Leica, Germany) and sectioned at −20°C with a cryostat (Leica CM3050). The staining was carried out as described previously (Huo et al., 2003). Western blotting was carried out as described previously (Qiu et al., 2001).

Isolation of cortices was performed as described previously (Anderson, 1974). Briefly, the trachea was trimmed of connective tissue in ice-cold PBS and placed in cortex isolation buffer containing 0.25 M sucrose, 20 mM HEPES pH 7.5, 25 mM KCl, 2 mM EDTA, 0.05% Triton X-100 and phosphatase and protease inhibitors for 30 minutes at room temperature with intermittent vortexing. Strong vortexing was carried out as a 1-minute strong vortex at 2-minute intervals, allowing the cortices and cilia to be separated as individual cilium. Mild vortexing was carried out as a 20-second mild vortex at 5-minute intervals, a protocol that preserved large chunks of cortices. For immunofluorescent staining, the extraction buffer was collected and cell debris, large chunks of cortices or cilia aggregates were collected by centrifugation at 600 g for 10 minutes. The pellet was resuspended in a small volume of cortices isolation buffer and was smeared onto a coverslip for immunofluorescent staining.

A Nikon N-SIM microscopy system on an Eclipse Ti inverted microscope run with Nikon Elements software (Nikon Instruments Inc.) was used. The samples were imaged with a 100×/1.49 NA objective and an iXon DU897 EM-CCD camera (Andor Technology PLC, Northern Ireland). Structured illumination microscopy (SIM) images were acquired with laser excitation and emission filters in 3D SIM mode (for each SIM image 15 images with five different phases of three different angular orientations of illumination were collected, as well as z-stacks). The sequential z-sections were reconstructed to create a 3D-SIM image and make the 3D deconvolution with the α-blending function using NIS-E software (Nikon). SIM images were processed with the Nikon Elements software.

Mouse trachea was fixed with 2.5% glutaraldehyde for 2 hours at 4°C and 1% osmium tetroxide for 2 hours at 4°C. After washing three times in distilled water, samples were dehydrated in ethanol and acetone as follows: 30% ethanol, 50% ethanol and 70% ethanol each for 15 minutes, 80% ethanol and 95% ethanol each for 20 minutes, 100% ethanol twice for 20 minutes each, and then 100% acetone three times each for 30 minutes. The dehydrated samples were infiltrated with a 1∶2 mixture of acetone∶Epon 812 ethoxyline for 2 hours and with pure Epon 812 ethoxyline overnight, and embedded in Epon ethoxyline resin. The samples were ultrathin-sectioned (DiATOME knife and Leica ultramicrotome), stained with 2% uranyl acetate and 2% lead citrate and observed in FEI Tecnai G2 Spirit transmission electron microscope.

For immunoelectron microscopy, mouse tissue was fixed with 4% paraformaldehyde instead of 2.5% glutaraldehyde. The rest of the steps were the same as the process for electron microscopy. The thin sections were blocked with 3% BSA in Tween Tris-buffered saline (TTBS). After incubation with primary antibody for 2 hours at room temperature, the sample was incubated with secondary antibody labelled with 10-nm colloidal gold for 1 hour. After washing in TTBS four times, the sample was fixed with 2.5% glutaraldehyde in water and stained with 2% uranyl acetate and 2% lead citrate for observation.

IRS-1 cDNA was cloned from 3T3-L1 cells. The mutants were constructed and N-terminally tagged with eGFP (pEGFP plasmid). The eGFP fusion proteins were expressed in 3T3-L1 cells and visualised by eGFP fluorescence.

For heat shock treatment, pre-warmed 42°C culture medium was added to the cells, and the samples were incubated in a 42°C water bath for the indicated time. For geldanamycin treatment, 90 nM geldanamycin was added to culture medium for up to 24 hours. At this time, cells were harvested for western blotting or fixed for immunofluorescent staining.

The length of cilia was measured with the measuring tool of the Leica confocal acquisition software. Positive staining of pAkt-1, pIRS-1 or IRS-1 at the cilium base or/and centrosome was determined from confocal microscopy images. The exact numbers of cells analysed are indicated in each figure legend. Data are presented as the mean±s.d. Differences were analysed by Student's t-test. P<0.05 was considered to be statistically significant.

We thank Ms Wei Bian at the Analytical Imaging Center of Institute of Biochemistry and Cell Biology for assistance with electron microscopy.