Estrogen destabilizes microtubules through an ion-conductivity-independent TRPV1 pathway

Authors


Address correspondence and reprint requests to Dr Tim Hucho, Max Planck Institute for Molecular Genetics, Department for Molecular Human Genetics, Signal Transduction in Pain and Mental Retardation, Ihnestrasse 73, 14195 Berlin, Germany. E-mail: hucho@molgen.mpg.de

Abstract

J. Neurochem. (2011) 117, 995–1008.

Abstract

Recently, we described estrogen and agonists of the G-protein coupled estrogen receptor GPR30 to induce protein kinase C (PKC)ε-dependent pain sensitization. PKCε phosphorylates the ion channel transient receptor potential, vanilloid subclass I (TRPV1) close to a novel microtubule-TRPV1 binding site. We now modeled the binding of tubulin to the TRPV1 C-terminus. The model suggests PKCε phosphorylation of TRPV1-S800 to abolish the tubulin-TRPV1 interaction. Indeed, in vitro PKCε phosphorylation of TRPV1 hindered tubulin-binding to TRPV1. In vivo, treatment of sensory neurons and F-11 cells with estrogen and the GPR30 agonist, G-1, resulted in microtubule destabilization and retraction of microtubules from filopodial structures. We found estrogen and G-1 to regulate the stability of the microtubular network via PKC phosphorylation of the PKCε-phosphorylation site TRPV1-S800. Microtubule disassembly was not, however, dependent on TRPV1 ion conductivity. TRPV1 knock-down in rats inverted the effect of the microtubule-modulating drugs, Taxol and Nocodazole, on estrogen-induced and PKCε-dependent mechanical pain sensitization. Thus, we suggest the C-terminus of TRPV1 to be a signaling intermediate downstream of estrogen and PKCε, regulating microtubule-stability and microtubule-dependent pain sensitization.

Abbreviations used
BIM

bisindolylmaleimide I hydrochloride

DRG

dorsal root ganglion

DTT

dithiothreitol

GFP

green fluorescent protein

MBP

maltose-binding protein

PBS

phosphate-buffered saline

PKCε

protein kinase C ε

+TIP

microtubule plus-end tracking protein

TIRF

total internal reflection

TRPV1

transient receptor potential, vanilloid subclass I

Protein kinase C ε (PKCε) is a central component in various models of pain sensitization (Khasar et al. 1999, 2008; Dina et al. 2000; Joseph and Levine 2003a,b; Sweitzer et al. 2004; Parada et al. 2005; Summer et al. 2006). Recently, we identified estrogen and agonists of the G-protein coupled estrogen receptor GPR30, such as G-1 and ICI 182,780 (fulvestrant) to induce PKCε-dependent mechanical pain sensitivity (Hucho et al. 2006; Kuhn et al. 2008). Nevertheless, the molecular and cellular mechanisms of PKCε-dependent sensitization are mostly elusive.

One target of PKCε phosphorylation is serine 800 (S800) at the C-terminus of transient receptor potential, vanilloid subclass I (TRPV1-Ct) (Bhave et al. 2003; Mandadi et al. 2006). Recently, we identified tubulin and microtubules to interact with TRPV1 close to S800 (Goswami et al. 2004, 2007a). In rats, the microtubule disruptor, Nocodazole, abolished PKCε-dependent mechanical hyperalgesia (Dina et al. 2003). However, if PKCε-activity influences the microtubule cytoskeleton and microtubule-dependent sensitization is unknown.

We now investigated, if PKCε-dependent phosphorylation of TRPV1 alters the interaction of tubulin with TRPV1, if thereby microtubule stability is changed, if TRPV1 ion channel conductivity is essential for this signaling process, if this PKCε-signaling can be induced also by receptor ligands such as the GPR30-agonists estrogen and G-1, and if therefore TRPV1 plays a role in estrogen-induced PKCε-dependent mechanical pain sensitization. We thereby tested the concept that the C-terminus of TRPV1 acts as a signaling intermediate controlling PKCε- and microtubule-dependent signaling towards pain sensitization.

Material and methods

Animals

Male Sprague–Dawley rats (200–300 g; Harlan Winkelmann, Borchen, Germany). Care and use of animals were in accordance with the European-Communities-Council-Directive of 24 November 1986 (86/609/EEC) and were approved by the LaGeSo, Berlin. All efforts were made to minimize the number of animals and their suffering.

Structural modeling

The tubulin structure (1TUB.PDB) and the modeled TRPV1 C-terminus were published earlier (Nogales et al. 1998; Fernandez-Ballester and Ferrer-Montiel 2008). Structures were edited with Swiss-PDB-viewer v3.7 (Guex and Peitsch 1997) and WHATIF (Vriend 1990). Protein–protein interaction prediction was accomplished with GRAMM-X v.1.2.0 (Tovchigrechko and Vakser 2006) using default conditions. For refinements in orientation and optimization of side chains in protein complexes, first, residues with Van-der-Waals-clashes were selected and fitted with ‘Quick-and-Dirty’ algorithms; second, models were energy-minimized (5 × 1000 steps of steepest descent and 5 × 1000 conjugate gradient, cutoff of 10 Å for non-bonded interactions) with Insight II (Biosym/MSI, Accelrys Software Inc., San Diego, CA, USA). The energy status of the protein complexes was evaluated with FoldX (Guerois et al. 2002; Schymkowitz et al. 2005) at the CRG site (http://foldx.crg.es). The molecular graphic representations were created and rendered with PyMOL v0.99rc2.

MBP-pull-down

Maltose-binding protein (MBP)-LacZ and MBP-TRPV1-Ct (aa681–aa838) were expressed in E. coli and purified as described (Goswami et al. 2004). Escherichia coli cells were lysed by three freezing/thawing cycles (lysis buffer: 20 mM Tris–HCl, pH 7.4, 150 mM NaCl, 0.2 g/mL sucrose, lysozyme, benzonase, protease inhibitor cocktail). Lysates were cleared by centrifugation, applied to amylose-resin (NEB), and incubated [1 h, (21–25°C)]. After washing, the amylose-resin with bound proteins was resuspended in PEM-S buffer (50 mm PIPES, pH 6.8, 100 mm NaCl, 1 mm EGTA and 0.2 mm MgCl2). Approximately 50 μL of amylose-resin per tube was used for pull-down experiments. The resin with coupled fusion-protein was incubated with 500 μL soluble F-11 extract [1 mg/mL, 1 h, (21–25°C)] with/without Ca2+ (2 mM), followed by three washes [200 mL PEM-S buffer with/without Ca2+ (2 mM)]. Proteins were eluted in 100 μL 10 mM maltose/PEM, separated by 10% sodium dodecyl sulfate (SDS)–polyacrylamide gel electrophoresis and analyzed by immunoblot.

In vitro phosphorylation

MBP-TRPV1-Ct was coupled to amylose-resin (1 mL) and eluted with 10 mM maltose in 500 μL phosphate-buffered saline (PBS) buffer. One hundred micrograms of MBP-TRPV1-Ct or MBP-LacZ in 140 μL PBS were used for in vitro phosphorylation. A kinase reaction mixture [MgCl2 (12 μL from 1 M stock), dithiothreitol (DTT) (12 μL from 1 M stock), lipid mix (phosphatidylserine f.c. = 200 μg/mL, diacylglycerol f.c. = 20 μg/mL in 10 mM HEPES (pH 7.4), 0.03% Triton X100; 80 μL from a 10× stock)], cold ATP (12 μL from 10 mM stock), 5 μL of γP32-ATP (10 μCi/μL), and PBS was added to a final volume of 160 μL. 20 μL were complemented with PKCε (1 μL) or buffer. The reaction-mix was incubated over night and stopped with 20 μL 5× Laemmli-buffer followed by 10% SDS–polyacrylamide gel electrophoresis and transfer to a polyvinylidene difluoride (Milipore) membrane. The radioactivity incorporated into the proteins was detected on a PhosphorImager STORM 820.

Tubulin-binding assay

Purified MBP-TRPV1-Ct and MBP-LacZ was coupled to amylose-resin and suspended in PEM-S buffer (50 mM PIPES pH 6.8, 1 mM EGTA, 0.2 mM MgCl2 and 100 mM NaCl). One hundred microliters suspension were incubated with 20 μg tubulin [200 μL of a 100 μg/mL stock, soluble tubulin purified from adult porcine brain, 1 h, (21–25°C)]. As control, 200 μL PEM-S buffer was added instead of tubulin. The amylose-resin was washed 3× with PEM-S buffer (600 μL/each) and PKCε-phosphorylation was performed. Bound proteins were eluted with 10 mM maltose/PEM-S and separated on SDS gels. Tubulin and MBP-TRPV1-Ct were detected by immunoblot using anti-MBP- and anti-β-tubulin-antibodies. Phosphorylation was monitored with PhosphorImager STORM 820.

PKCε phosphorylation was performed using 100 μL of MBP-TRPV1-Ct or MBP-LacZ coupled amylose suspension as template. MgCl2 (12 μL from 1 M stock solution), 1.5 μL DTT (from 0.1 M stock solution), 1.5 μL Lipid Mix, 15 μL cold ATP (0.01 M), 1.5 μL hot ATP (γP32 with 10 μCi/μL) and 4 μL PKCε were incubated with 100 μL of TRPV1-coupled amylose-resin. Reaction mixture without PKCε served as control. Samples were incubated for 1 h at room temperature, washed three times with PEM-S (800 μL each) and incubated with 20 μg tubulin [20 μL from 100 μg/mL solution, 1 h, (21–25°C)]. Amylose-resin was washed three times with PEM-S (600 μL each), was removed using a Hamilton-syringe, bound proteins were eluted with 10 mM maltose/PEM-S, separated on SDS gels, and analyzed by immunoblot using anti-MBP- and anti-β-tubulin-antibodies.

Immunoprecipitation

Approximately, 200 μg of MBP-TRPV1-Ct in 200 μL PEM-S buffer (PEM/100 mM NaCl) was phosphorylated by PKCε. The kinase was activated by MgCl2 (2 μL from 1 M stock), DTT (2 μL from 1 M stock), lipid mix (20 μL from a 10×stock; composition described above), cold ATP (2 μL from 10 mM stock), 1 μL of hot ATP (γP32 with 10 μCi/μL) and 1 μL of PKCε. Phosphorylation [3 h, (21–25°C)] was stopped by addition of bisindolylmaleimide I hydrochloride (BIM). One hundred microliters of the phosphorylated MBP-TRPV1-Ct was mixed with 100 μL non-phosphorylated MBP-TRPV1-Ct and 20 μg of purified tubulin dimer was added (final tubulin:phospho MBP-TRPV1-Ct:non-phospho MBP-TRPV1-Ct ratio is 1:5:5). Tubulin-MBP-TRPV1-Ct complexes were established [1 h, (21–25°C)] and bound to protein-G-resins coupled with 10 μg of tubulin antibody [1 h, (21–25°C)]. The flow-through and wash fractions (each 1 mL PEM-S buffer) were collected for analysis. The proteins bound to protein-G beads were eluted with Laemmli buffer (5×). Equal volume (10 μL) of all input, flow-through, wash and elution fractions were loaded and analyzed by coomassie staining and autoradiogram. The samples were also probed by western blot for tubulin (by DM1A, 1 : 1000) and MBP-TRPV1-Ct.

Preparation of cytoskeleton in situ

F-11 cells were transfected on glass cover slips with TRPV1 coupled to green fluorescent protein (TRPV1-GFP), TRPV1-ΔN or TRPV1-S800A-GFP and cultured for 48 h. Medium was replaced by serum-free media 1 h before stimulation. Cells were stimulated with 17-β-estradiol (1 nM), G-1 (10 nM), ICI 182,780 (100 nM) or phorbol 12-myristate 13-acetate (1 μM). Stimulation occurred at 21–25°C.

To ensure homogeneous dispersion of the stimulants, half of the 1 mL medium was removed, mixed thoroughly with the respective stimulans, and added back to the culture. All stock solutions of reagents were dissolved in 100% dimethylsulfoxide (f.c. on cells = 0.2%). Buffer-controls were treated alike without adding any stimulus. After treatment cells were extracted with an isotonic cell-extraction buffer [1 min, (21–25°C), 50 mM PIPES, pH 6.8, 1 mM EGTA, 0.2 mM MgCl2, 10% v/v glycerol, complete™ protease inhibitor cocktail, digitonin 50 μg/mL]. Buffer was removed and cells fixed with 4% paraformaldehyde [10 min, (21–25°C)].

Fixed cells were analyzed immunocytochemically after permeabilization [0.1% Triton X-100, 10 min, (21–25°C)], wash (3×, PBS), and block of non-specific binding-sites [5% bovine serum albumin or 10% normal goat serum in PBS, 1 h, (21–25°C)]. Cells were probed with primary antibodies in 1% bovine serum albumin/PBS-T [overnight 4°C or 1 h, (21–25°C)]. Antibody dilutions: rabbit polyclonal anti-TRPV1 IgG (1 : 1000; AbCam, Cambridge, UK), rat monoclonal anti-tyrosinated tubulin IgG (Clone YL ½, 1 : 1000). Cells were washed three times with PBS-T and incubated with Alexa 488/Alexa594-coupled antibodies [1 : 1000, 1 h, (21–25°C)]. After three final washes cells were mounted with Fluoromount-G (Southern Biotech/Biozol, Birmingham, AL, USA) containing 4′,6-diamidin-2-phenylindol (5 μg/mL). Confocal images were taken on a Zeiss LSM 510 Meta with 40× or 63× objectives and analyzed with the Zeiss LSM image-examiner-software.

F-11 cell culture and transfection

F-11 cells were cultured in Ham’s F12 medium (Invitrogen, Darmstadt, Germany) supplemented with 20% fetal bovine serum. For transient transfection, Lipofectamine 2000 and Plus reagents (Invitrogen) were used according to the instruction provided. Experiments were done approximately 48 h after transfection.

Life cell imaging of F-11 cells

F-11 cells were transfected with plasmids encoding TRPV1-GFP or TRPV1-S800A-GFP along with m-Cherry-tubulin. Cells were imaged 36–60 h after transfection in a metal live cell chamber covered with Hank's balanced salt solution buffer. Stimulation agents were diluted in 50 μL Hank's balanced salt solution buffer and added onto the culture during imaging.

For analyzing the involvement of protein kinases C (PKCs), TRPV1-GFP expressing cells were selected before application of the PKC inhibitor BIM (f.c. 5 μM, 12 min pre-incubation). As BIM shows high autofluorescence, thereafter only phase contrast images were taken (1 frame/min).

Cells were imaged with an inverted Zeiss LSM 510 Meta confocal microscope with either 40× or 63× objectives.

TIRF imaging of F-11 cells

F-11 cells were seeded on glass cover slips, transfected with TRPV1-GFP and cultured for 24 h after transfection. Medium was replaced by serum-free media 1 h before stimulation. Cells were stimulated with estrogen [1 nM, 1 min, (21–25°C)] or PBS (solvent control) and cells were fixed with methanol (10 min, −20°C). Unspecific binding sites were blocked with 10% normal goat serum/PBS [1 h, (21–25°C)] followed by incubation with primary antibodies [rabbit polyclonal anti-TRPV1 IgG, 1 : 1000, AbCam and rat monoclonal anti-tyrosinated tubulin IgG, Clone YL ½, 1 : 1000, 1 h, (21–25°C)], washes [3×, 5 min, (21–25°C)], and secondary antibodies (Alexa 568 anti-rat, 1 : 1000 and Alexa 488 anti-rabbit, 1 : 1000). After washing [3×, 5 min, (21–25°C)] cells were mounted on glass slides with PBS and fixed with nail polish. Total internal reflection (TIRF) images were taken with a 100×-objective on a microscope (Zeiss) with a TIRF illuminator and fiber-optic-coupled laser-illumination. Images were analyzed with Image J. For quantification, filopodia structures were analyzed from three independent experiments with identical exposure times for tubulin signals in filopodia.

Data analysis

Data are presented as mean ± SEM. All statistical comparisons were made with one-way anova followed by Dunnett’s test for comparisons with one control value, or Tukey–Kramer post hoc test for multiple comparisons. A p-value of < 0.05 was considered statistically significant.

Results

Structural modeling indicates binding of microtubular protofilaments and free tubulin to TRPV1

We reported microtubules to bind to the C-terminus of TRPV1 (Goswami et al. 2004) and identified the binding domain of TRPV1 (Fig. 1a) (Goswami et al. 2007a). But, which part of the microtubule binds TRPV1, is not known. With a three-dimensional structural model of TRPV1 (Fernandez-Ballester and Ferrer-Montiel 2008), we tested if tubulin dimers, microtubule protofilaments or whole microtubules filaments can bind to TRPV1.

Figure 1.

 Modeling indicates interference of tubulin-binding and PKCε phosphorylation at TRPV1-S800. (a) Regulatory regions (arrows) and tubulin-interaction domains (green cylinders) of the TRPV1-C-terminus. (b) Side view of the interaction of soluble α (yellow) and β (magenta) tubulin dimers with the tetrameric-TRPV1-C-terminus (green). PKCε-phosphorylation site, TPRV-S800, is indicated by the black arrow. Most computed tubulin-TRPV1 interactions are centered on Ser800 and involve both the α- and β-tubulin interface.

For interaction, the surface areas of the binding partners have to match. The interaction involves the globular C-terminal region of TRPV1 (Goswami et al. 2007a), which is predicted to form a hook-like structure (Fig. 1b, Figure S1a) (Fernandez-Ballester and Ferrer-Montiel 2008). The surface area of one C-terminal globular region (residue 716–839) is calculated to 81.7 nm2. Thus, microtubules (13 protofilament, 25 nm in diameter) are with 490 nm2 much too large for binding in this narrow hook/groove structure. Microtubules are larger even than the whole tetrameric TRPV1 C-terminus (490 nm2 versus 412.5 nm2, respectively) and are therefore similar unlikely to bind.

In contrast, αβ-tubulin-dimers meassure to 4.6 nm × 8.0 nm × 6.5 nm (Nogales et al. 1998). Thus, individual tubulin-dimers or tubulin at the tip of single microtubule protofilaments (plus end of open, straight or frayed conditions) can fit into the hook-structure formed by the TRPV1-Ct surface. Accordingly, in a random trial of 200 possible tubulin-TRPV1 interactions (Tovchigrechko and Vakser 2006), 155 possibilities showed an interaction of tubulin-dimers with TRPV1-Ct indicating a high binding probability.

Modeling indicates interference of tubulin-binding and TRPV1 phosphorylation

Tubulin binds the TRPV1-C-terminus at amino acids 771–797 (Goswami et al. 2007a) thus binding close to the PKCε-phosphorylation site, Serine 800 (Fig. 1a) (Mandadi et al. 2006). In our model, TRPV1-S800 is solvent exposed and accessible to PKCε-phosphorylation. Tubulin can bind TRPV1 in multiple orientations. Most potential interactions between tubulin and TRPV1-Ct suggested by our modeling approach are centered on the S800 position of TRPV1. Figure 1b and Figure S1a/b show an example where the tubulin-dimer forms with S800 several hydrogen bonds [S800/R802 with D163 (β-tubulin), R797 with H406 (main chain, α-tubulin), R802 with E97 (main chain, α-tubulin)]. The space between S800 and TRPV1-bound tubulin is limited. Thus, a phosphate group at S800 position should clash both sterically and by charge with Asp163 of β-tubulin, as well as repel the negatively charged Glu411 of α-tubulin. Thus, our modeling approach indicates, that phosphorylation of S800 should interfere with the binding of tubulin to TRPV1.

TRPV1-C-terminus forms a complex with tubulin and PKCεin vitro

To evaluate the functional implications of our modeling, we characterized the relation of TRPV1, PKCε and tubulin biochemically in vitro. We performed pulldown experiments with the purified C-terminal region of TRPV1 fused to maltose-binding-protein (MBP-TRPV1-Ct). We found PKCε and tubulin to associate with MBP-TRPV1-Ct in extracts of dorsal root ganglion (DRG) neuron derived F-11 cells (Fig. 2a).

Figure 2.

 Interaction of PKCε and tubulin with the TRPV1 C-terminus. (a) In pull-down experiments with purified MBP-TRPV1-Ct from F-11 extracts, PKCε and α-tubulin were detected in complex with TRPV1-Ct (Western Blots tested with tubulin- and PKCε-specific antibodies). (b) Phospho-S729-PKCε interacts with MBP-TRPV1-Ct. After phosphorylation of purified PKCε in presence of γATP and immobilized MBP-TRPV1-Ct, amylose resin was washed extensively and eluted with maltose. In the eluate, the pulled down complex containing MBP-TRPV1-Ct, pPKCε and PKCε were detected via Autoradiogram (1) and via western blot [anti-pPKCε (2), anti-PKCε (3), anti-MBP (4)]. (c) Detection of phosphorylation of the PKCε-phosphorylation site, TRPV1-S800, using phospho-TRPV1-S800 specific antibodies. Incubation of MBP-TRPV1-Ct with purified PKCε resulted over time in increased signals of the phospho-TRPV1-S800 antibody (1–60 min). Phosphorylation signal was strongly reduced by the PKC inhibitor BIM. The Coomassie stained gel served as loading control. (d) Kinase assays with purified PKCε and MBP-TRPV1-Ct showed PKCε-mediated phosphorylation of the TRPV1-C terminus, which could be blocked by the PKC inhibitor BIM (5 μM). No phosphorylation was detected on the MBP-LacZ control.

Phosphorylation of S729 on PKCε is taken as indicator for activation. We investigated if activated, that is, phospho-S729-PKCε could be pulled down with MBP-TRPV1-Ct out of a solution of purified MBP-TRPV1-Ct, PKCε, and P32γ-ATP. Activation of PKCε was initiated by addition of an activating lipid mix. Autoradiogram (Fig. 2b, lane 1) and western blot (Fig. 2b, lane 2) analysis of the eluate showed the MBP-TRPV1-Ct-complex to contain phospho-S729-PKCε.

To test if active PKCε indeed phosphorylates TRPV1, we used an antibody specific for the phosphorylated PKCε-phosphorylation site, TRPV1-S800 (Mandadi et al. 2006). Incubation of MBP-TRPV1-Ct with PKCε resulted in a time-dependent increase in phosphorylated MBP-TRPV1-S800 (Fig. 2c). Block of PKCε with BIM strongly reduced the antibody detected phospho-MBP-TRPV1-Ct signal (Fig. 2c). To test for the specificity of the antibody signal we next incubated purified MBP-TRPV1-Ct with PKCε and P32γ-ATP. Indeed, MBP-TRPV1-Ct was radioactively labeled and resulted in a clear signal in the autoradiogram (Fig. 2d, lane 4). Supporting the specificity of the phosphorylation reaction, the reaction was nearly completely abolished by BIM (Fig. 2d, last (fifth) lane).

PKCε phosphorylation of S800 on TRPV1 competes with tubulin binding

Our modeling suggested that phosphorylation of S800 by PKCε would reduce the subsequent binding of tubulin to TRPV1. Indeed, pre-incubation of MBP-TRPV1-Ct with active PKCε reduced the amount of bound tubulin by 22% (Fig. 3a and b). Vice versa, incubation of MBP-TRPV1-Ct with tubulin before adding active PKCε reduced the phosphorylation signal by 33% (Figure S2a and b).

Figure 3.

 Regulation of the tubulin-PKCε-TRPV1 complex. (a) Phosphorylation of the TRPV1-Ct reduces tubulin binding. PKCε mediated phosphorylation of MBP-TRPV1-Ct resulted in a decrease of tubulin binding compared to non-phosphorylated controls. (b) Quantification of tubulin band densities of (a), error bars represent SEM. n = 8, ***p < 0.001. (c) Tubulin binds preferentially to non-phosphorylated TRPV1-Ct. In a competitive binding assay with purified tubulin and a mixture of non-phosphorylated and P32-phosphorylated MBP-TRPV1-Ct followed by an IP with anti-α-tubulin antibodies, MBP-TRPV1-Ct was detected in western blots. The following fractions are presented: 1. Input of MBP-TRPV1-Ct (not phosphorylated), 2. Input of MBP-TRPV1-Ct (phosphorylated by PKCε), 3. Input for the IP (mixture of 1 and 2 and tubulin), 4. Flow through of the IP, 5. first wash flow through, 6. second wash flow through, 7. third wash flow through, 8. IP elution sample. In the elution sample (lane 8) no phosphorylation of MBP-TRPV1-Ct could be detected on an autoradiogram even after 7 days of exposure.

As kinase reactions mostly are non-exhaustive, the reduction of tubulin binding by prior PKCε-phosphorylation could be caused by two differential mechanisms: prior phosphorylation (i) reduces or (ii) abolishes the binding affinity of tubulin to phosphorylated TRPV1. To differentiate between these two options, we exposed tubulin in a competitive binding experiment to equal amounts of non-phosphorylated MBP-TRPV1-Ct and P32-phosphorylated MBP-TRPV1-Ct before adding limited amounts of tubulin. Subsequent immunoprecipitation of α-tubulin confirmed the presence of MBP-TRPV1-Ct in the immunoprecipitates. However, testing for P32-phosphorylation with a sensitive autoradiogram we could not detect any phosphorylation of MBP-TRPV1-Ct even after 7 days of exposure (Fig. 3c, lane 8). In contrast, presence of MBP-TRPV1-Ct in the immunoprecipitate is readily visibly by Coomassie staining (Fig. 3c, lane 8). Thus, tubulin binds to non-phosphorylated but not to phosphorylated MBP-TRPV1-Ct.

Estrogen reduces the number of microtubule containing filopodia

Our modeling and biochemical data indicates that PKCε phosphorylation of TRPV1-S800 interferes with tubulin-binding to TRPV1. Previously, we found TRPV1 to stabilize microtubules (Goswami et al. 2004, 2006). Therefore, we next addressed, if disruption of the TRPV1-microtubule interaction results in destabilization of microtubules.

Few microtubules if any reach into filopodial structures. Nevertheless, we recently described microtubules in filopodial structures in various TRPV1-expressing cells including DRG-neuron-derived F-11 cells. Visualizing them with high-resolution TIRF microscopy, we found microtubules in 84.26% of the evaluated TRPV1-expressing F-11 filopodia. Estrogen treatment (1 nM, 1 min) resulted in a drastical drop in numbers of microtubule-containing filopodia to 19.79% (Fig. 4a and b) apparently destabilizing microtubules in these structures.

Figure 4.

 Estrogen induces microtubule destabilization in TRPV1 expressing cells. (a) Estrogen leads to retraction of microtubule from filopodia structures in TRPV1-GFP expressing F-11 cells [shown are TIRF-microscopic images, both, black and white (upper rows) as well as false colour images (lower rows)]. (b) Quantification of percentage of microtubule-containing filopodia structures. Estrogen reduced the fraction of microtubule containing filopodia from 84.3 ± 4.8% to 19.8 ± 3.3% [evaluated: n = 316 filopodia (controls), n = 228 filopodia (estrogen-treated)]. (c) Estrogen induces microtubule destabilization in TRPV1 expressing F-11 cells. After estrogen treatment (10 nM, 1 min) the soluble cytoplasmic components were removed by digitonin extraction. Microtubule structure was analyzed via confocal immunofluorescence microscopy using anti-tubulin and anti-TRPV1 antibodies. Only in TRPV1-expressing cells (upper row) estrogen lead to near complete destruction of filamentous microtubules, whereas in TRPV1-negative cells (lower row) no destruction was detectable. (d) In cells expressing a TRPV1 mutant lacking the PKCε phosphorylation site, TRPV1-S800A-GFP, no microtubule destabilization was observed in response to estrogen. Analysis of TRPV1-S800A-GFP expressing cells after estrogen treatment (10 nM, 1 min) and digitonin-extraction showed complete microtubule filaments after staining with anti-tubulin and anti-TRPV1 antibodies using fluorescent confocal microscopy. (e) Quantification of mean tubulin immunofluorescence signals of (c) and (d), n = 21, 29, 20, respectively. (f) Estrogen-induced microtubule destabilization is independent of the ion conductivity of TRPV1. Pre-treatment with the general cation-channel blocker, ruthenium red (RR, 2 μM, 20 min), does not inhibit estrogen-induced microtubule destabilization observed in TRPV1-GFP expressing F-11 cells. (g) Quantification of tubulin signals of (f), n = 19, 35, respectively. Error bars represent SEM. ***p > 0.001.

Estrogen induces TRPV1-dependent destabilization of microtubules

We next tested if the estrogen-induced microtubule destabilization can be observed also on a whole cell level just monitoring the total amount of microtubule-derived signal independent of its subcellular localisation. To allow differentiation between fully polymerized microtubular structures and soluble tubulin, the cells were extracted with a digitonin-containing isotonic buffer shortly before fixation. Digitonin washes out soluble cytoplasmic tubulin but leaves the polymeric cytoskeleton unaffected (Lieuvin et al. 1994; Goswami et al. 2006). The then fixed cells were immunfluorescently labeled and the total immunfluorescence quantified. Application of estrogen (1 nM, 1 min) to TRPV1-expressing cells resulted in rapid decrease of microtubule immunfluorscence signal down to 52.28% (Fig. 4c and e). Microtubules no longer appeared as smooth filaments and the remaining immunofluorescence signal was restricted to the vicinity of the microtubule-organizing center. In contrast, TRPV1-negative cells were not affected by estrogen and intact microtubules could be visualized all over the cells (Fig. 4c and e).

PKCε phosphorylation at TRPV1-S800 is essential for estrogen-induced microtubule disassembly

Recently, we identified PKCε to be activated by estrogen-induced signaling (Hucho et al. 2006; Kuhn et al. 2008). Thus, we investigated if estrogen produces PKCε-mediated destabilization of microtubules.

Lipids as well as hydrophobic substances can activate TRPV1 at its Capsaicin-binding site. To exclude a direct action of estrogen at this site, we performed the estrogen treatment in presence of the Capsaicin antagonist, 5′I-RTX (1 μM, 30 min). Still, estrogen strongly destabilized the microtubular network (Figure S3a).

Next, we used the PKC inhibitor BIM, which blocks all PKC isoforms including PKCε. In live cell experiments, pre-treatment with BIM (5 μM, 12 min) abolished estrogen-evoked (1 nM, 1 min) morphological alterations in TRPV1-expressing F-11 cells (Fig. 5b). In contrast, the inhibitor Gö6976, which blocks classical PKCs but not PKCε, did not prevent the estrogen-induced morphological changes in TRPV1-GFP expressing F-11 cells (data not shown).

Figure 5.

 Estrogen treatment results in rapid morphological alterations in TRPV1-expressing cells. (a) Estrogen induced no changes of the microtubular cytoskeleton in F-11 cells only expressing tubulin-cherry but no TRPV1 (upper row). In contrast, live cell imaging of tubulin-cherry and TRPV1-GFP expressing F-11 cells showed cell retraction, growth cone retraction and varicosity formation in response to estrogen (10 nM) within 5–10 min. (b) In live cell imaging experiments of TRPV1-expressing F-11 cells incubation with the PKC inhibitor, BIM (5 μM), 12 min prior to estrogen application (1 nM) inhibited estrogen-induced morphological changes. Because of the strong autofluorescence of BIM, one fluorescence image of TRPV1-GFP/tubulin-cherry expressing cells was taken before BIM application and imaging was continued in the bright field mode for at least 20 min (1 frame/min). (c) Live cell imaging of 4–5 day cultures of DRG neurons showed rapid varicosity formation in response to estrogen (10 nM) within 5–15 min only in a subset of neurons. (d) Similar to estrogen, treatment with the GPR30 agonist G-1 (100 nM) induced varicosity formation within 5–15 min in a subset of DRG neurons.

To further narrow down the mediating kinase, we expressed the TRPV1-S800A mutant, which lacks the PKCε-specific phosphorylation site (Mandadi et al. 2006). No destabilization of microtubules was detected after estrogen treatment (1 nM) in fixed cells (Fig. 4d and e).

Microtubule destabilization is independent of TRPV1 ion conductivity

To investigate, if estrogen-induced microtubules destabilization is dependent of TRPV1s ion conductivity we expressed TRPV1 lacking the entire N-terminal domain (TRPV1-ΔN). TRPV1-ΔN was reported to localize to the plasma membrane but not to be activated by TRPV1 agonists (Jung et al. 2002). TRPV1-ΔN expressing cells showed microtubule disassembly in response to estrogen similar to disassembly observed in cells expressing full length TRPV1 (Figure S3b).

To corroborate that estrogen-induced rapid microtubule disassembly is independent of TRPV1 ion-pore opening and calcium influx, we applied the general cation-channel blocker, ruthenium red, before estrogen treatment (20 μM, 20 min pre-incubation). Still, estrogen evoked a reduction of the mean microtubule signal to 58,18% (Fig. 4f and g). To ensure that the ruthenium red concentration and the pre-treatment time were sufficient for suppression of calcium-influx-dependent destabilization, we tested the effect of ruthenium red on capsaicin-induced microtubule destabilization. Ruthenium red abolished capsaicin-induced (1 μM, 1 min) microtubule destabilization completely (Figure S3c).

Estrogen induces TRPV1-dependent morphological changes in F-11 cells

Fixation procedures are prone to alter the cytoskeleton. To assure that the observed destabilization occurs in the living cell, we performed live cell imaging. Changes in cellular morphology are a convenient indicator for cytoskeletal reorganization. Application of 1 nM estrogen had no effect on F-11 cells lacking the expression of TRPV1 (Fig. 5a, upper row). In contrast, estrogen resulted in rapid (1–10 min) morphological changes in TRPV1-GFP and tubulin-cherry expressing cells. Also cell retraction, growth cone shortening and varicosity formation was observed (Fig. 5a, lower rows). Expression of TRPV1 with mutated PKCε-phosphorylation site (TRPV1-S800A-GFP) abolished the morphological changes (data not shown).

GPR30 agonist G-1 induces TRPV1-S800-dependent but 5′I-RTX-insensitive destabilization of microtubules

PKCε activation was induced by estrogen and agonists of the novel estrogen receptor, GPR30, such as G-1 and ICI 182,780 (Kuhn et al. 2008). RT-PCR analysis verified presence of GPR30 mRNA in F-11 cells (data not shown). G-1 treatment like estrogen had no effect on TRPV1-negative F-11 cells (10 nM, 1 min, Figure S4a). In contrast, in TRPV1-GFP expressing cells G-1 reduced the mean tubulin fluorescence intensity to 46.81% (Figure S4a and b). Genetic deletion of the PKCε phosphorylation site TRPV1-S800 abolished this effect (Figure S4a and b).

Also in live cell experiments, G-1 induced similar morphological changes as estrogen. Morphological alterations included cell and growth cone retraction (Figure S4c). F-11 cells lacking TRPV1 expression showed no morphological changes (Figure S4d). G-1 actions were not blocked by the lipophilic agonists antagonist, 5′I-RTX (Figure S4d). In live cell experiments, G-1 treatment resulted in microtubule disassembly and varicosity formation in wild type TRPV1-GFP but not in TRPV1-S800A-GFP expressing cells (Figure S4d). However, the potent TRPV1 agonist, RTX, induced varicosity formation within 3 min even in TRPV1-S800A-GFP expressing cells (Figure S4e). Thus, the GPR30-agonist, G-1, mirrors the described estrogen effect on cell morphology and microtubule stability.

Estrogen and G-1 causes microtubule disassembly in a subset of DRG neurons

To confirm the relevance of the described signaling pathway for nociceptive neurons, we examined if PKCε activation by estrogen and G-1 influences the microtubular cytoskeleton of DRG neurons. Live cell imaging of DRG-neuron neurites showed varicosity formation after stimulation with estrogen (10 nM) or G-1 (100 nM) within 5–15 min, indicating the disassembly of microtubule filaments (Fig. 5c and d). As TRPV1 is not expressed in all neurons, accordingly, not all neurons responded to this treatment.

TRPV1 regulates PKCε-mediated and microtubule-dependent mechanical pain sensitization

Epinephrine-induced and PKCε-dependent mechanical sensitization has been shown to be modulated by microtubule destabilizing drugs such as Nocodazole but to be permissive to Taxol pre-treatment (Dina et al. 2003). We now find G-1-induced sensitization, indeed, like epinephrine-induced sensitization to be completely abolished by Nocodazole pre-treatment but not by Taxol-pretreatment (Fig. 6).

Figure 6.

 Presence/absence of TRPV1 alters the influence of microtubule-drugs on G-1 induced and PKCε-dependent sensitization in male rats. Injection of the estrogen-receptor GPR30 agonist, G-1 (1 μg/2.5 μL), reduces mechanical nociceptive thresholds in naïve but also TRPV1 antisense or mismatch oligodeoxynucleotides (ODN) treated male rats. In naïve and mismatch ODN treated animals, pre-treatment with Nocodazole (1 μg/2.5 μL) inhibits G-1-induced sensitization, whereas pre-treatment with the microtubule stabilizer Taxol (1 μg/2.5 μL) had no effect. In TRPV1 antisense ODN-treated animals, the effects of Nocodazole and Taxol were reversed (n = 6, p < 0.001).

We described in this manuscript PKCε-phosphorylation of TRPV1 to alter the microtubular network. We therefore asked if knock-down of TRPV1 alters the effect of Taxol and/or Nocodazole on estrogen/G-1-induced mechanical sensitization. Indeed, knock-down of TRPV1 (Christoph et al. 2006) altered dramatically their influence. Their effect was turned around so that now Taxol but not Nocodazole abolished G-1-induced sensitization (Fig. 6, Figure S5b).

Discussion

Ion channels have been studied intensively in the context of pain and pain sensitization. In contrast, cellular and molecular processes underlying sensitization beyond changes of electrical properties of ion channels are only starting to emerge (Hucho and Levine 2007). We hypothesized, that the interaction of microtubules with TRPV1 could be modulated by phosphorylation. We tested this hypothesis via structural modeling, biochemical interaction studies, as well as in cell biological model systems and tested for a role in sensitization in behavioral experiments.

Our computational approach indicates that only soluble tubulin or tubulin exposed at the tip region of microtubular protofilaments is small enough to bind to TRPV1-Ct. We found that tubulin binds only to TRPV1, which is not phosphorylated at S800. Activation of PKC resulted in phosphorylation of the PKCε phosphorylation site, TRPV1-S800. Subsequent microtubular destabilization was dependent on PKC activity and the PKCε phosphorylation site TRPV1-S800. Although dependent on TRPV1, microtubule destabilization was independent of the ion conductivity of TRPV1. And finally, the effect of microtubule-altering drugs onto sensitization was strongly dependent on the presence of TRPV1 rats. This not only introduced novel components of estrogen-induced intracellular sensitization-signaling pathways but also proved that TRPV1 can function as signaling intermediate rather than signaling starting receptor or signaling modulated effector.

Based on these data, we discuss the following aspects: (i) conductance-independent TRPV1 signaling, (ii) TRPV1 as dynamically regulated microtubule plus-end tracking protein (+TIP), (iii) TRPV1 as downstream mediator of estrogen signaling and (iv) TRPV1, a role in cytoskeleton-dependent mechanical hyperalgesia.

Microtubule destabilization is independent of the ion conductivity of TRPV1

Microtubules destabilization can be induced by calcium channel opening and the influx of calcium. This was shown for example for TRPC and TRPV-channels such as TRPC5 (Greka et al. 2003) and TRPV1 (Goswami and Hucho 2007), where the calcium channel opening controled growth cone movements and neurite outgrowth. But an ever increasing plethora of proteins like microtubule-associated proteins have been identified to regulate microtubule stability independent of intracellular calcium-concentration-changes (Akhmanova and Steinmetz 2008). Thus, the opening of a calcium channels is not necessarily required for rapid destabilization of microtubules. Although for TRPC5 so far only calcium-dependent mechanisms have been described, indeed, recently some researchers indicated calcium- and ion-conductivity-independent functions for TRPV1. Several cellular functions such as filopodia initiation, smooth muscle cell relaxation, and the enhancement of kidney cell uptake, have been reported to depend on the presence or activation of TRPV1, but remained insensitive to TRPV1 inhibitors or calcium depletion (Creppy et al. 2000; Fujimoto and Mori 2004; Myrdal and Steyger 2005; Fujimoto et al. 2006; Goswami and Hucho 2007). Nevertheless, prove of a calcium independent function of TRPV1 was still missing. With the non-activatable TRPV1-ΔN (Jung et al. 2002) and with the blocker Ruthenium Red, we now prove that signaling toward microtubules involves TRPV1 but that microtubule destabilization is independent of TRPV1′s ion conductivity.

This elucidates an interesting additional point. Disassembly of microtubules is based on phosphorylation of TRPV1-S800. Phosphorylation of this site also has been reported to sensitize TRPV1. But as microtubule destabilization occurs in absence of ion conductivity of through TRPV1, apparently destabilization is independent of sensitization evoked channel opening. Therefore, we now provide evidence, that PKCε-induced sensitization of TRPV1 and PKCε-induced microtubule destabilization are two phenomena dependent on one and the same phosphorylation event but at least in part independent of each other. This raises the question, if only microtubule destabilization is independent of sensitization or if also in turn S800-dependent TRPV1 sensitization can be separated from microtubule destabilization. An appealing third option would be that the loss of microtubule interaction is a pre-requisite for the sensitization of the ion channel as discussed further below.

Although increasingly ion conductivity-independent functions of voltage-gated ion channels have been reported, to our knowledge this is the first report on conductivity-independent TRPV1 signaling. Indeed, not directly ion-conductivity-linked mechanisms are likely to be involved in the sensitization machinery. We recently made the case that cells present a plethora of aspects and mechanisms beyond ion channel opening, which so far have not been investigated in the context of pain (Hucho and Levine 2007). To find a non-ion conductivity-based mechanism in pain sensitization indicates further the need to investigate nociceptive neurons beyond electrophysiological properties of single ion channels (Hucho and Levine 2007).

TRPV1 beyond ion conductivity

Our findings are of interest also in a general cell-biological context. Increasingly, submembranous localization of microtubules is gaining attention (Goswami and Hucho 2008). But knowledge about microtubule stabilizing mechanisms at the membrane is still sparse. Physical interaction of microtubules with membranes (Stephens 1986; Wolff 2009) but also with transmembrane proteins such as TRPC1, TRPC5, TRPC6, Na/K ATPase (van Rossum et al. 1999; Saugstad et al. 2002; Vladimirova et al. 2002; Bollimuntha et al. 2005; Campetelli et al. 2005; Goel et al. 2005; Ma et al. 2007) and TRPV1 (Goswami et al. 2007a) have been reported. Of special interest among microtubule-binding proteins is the class of plus end (+TIP) binding proteins (Morrison 2007; Akhmanova and Steinmetz 2008). We find the C-terminus of TRPV1: (i) to stabilize microtubules (Goswami and Hucho 2007; Goswami et al. 2007a), (ii) to regulate growth cone behavior (Goswami et al. 2007b), (iii) potentially to bind to the very end of microtubular protofilaments, and (iv) to interact with microtubules in a dynamically phosphorylation-regulated manner. All four properties of TRPV1 are shared properties with +TIP binding proteins. Thus, it is tempting to describe TRPV1 as a +TIP binding protein. This opens the doors to investigate the role of TRPV1 in microtubule dynamics, which of the aspects of +TIP proteins plays a role in pain sensitization, and if also other +TIP proteins are involved in the sensitization process.

Regulation of the cytoskeleton plays a crucial role in neuronal functionality. Recently, we described activation of TRPV1 to induce growth cone collapse (Goswami et al. 2006). Interestingly, in cases like diabetes sensory fiber retraction/degeneration is a common phenomenon. Is there a role of TRPV1-mediated cytoskeleton regulation? Indeed, TRPV1 has been described to be involved in neurite retraction, neuronal degeneration, and cell death (Kawakami et al. 1993; Kim et al. 2005; Hong et al. 2008). The loss of interaction could thereby result in a destabilization of microtubules and therefore increased accessibility of TRPV1 for PKCε. PKCε-phosphorylation of TRPV1 in turn could block further microtubule interaction but also sensitize TRPV1, which resulted in increased ion pore openings and thus further increased microtubule destabilization by calcium influx. If the loss of interaction between TRPV1 and the microtubule could indeed result in a degeneration-feed-forward loop remains to be investigated in detail.

TRPV1 as downstream mediator of estrogen signaling

The novel role of TRPV1 in microtubule regulation is interesting. We started the presented series of experiments as we were following up on our results, that estrogen activates PKCε, and attempted to identify novel downstream components of this sensitization signaling. Indeed, TRP channels and steroid hormones are functionally interconnected. Capsaicin increases the expression of steroid receptors (Malagarie-Cazenave et al. 2009). Steroids induce TRP channel expression (Tong et al. 2006). Steroids can activate TRP channels by direct binding (Irnaten et al. 2008; Nilius and Voets 2008; Wagner et al. 2008). And steroids can modulate TRP channel activity via yet to identify mechanisms (Diogenes et al. 2006). We now describe TRPV1 to be post-translationally modified through fast non-genomic estrogen signaling involving PKC at S800, a PKCε-phosphorylation site.

But which estrogen receptor mediates this effect? Recently we suggested the G-protein coupled estrogen receptor GPR30 (Revankar et al. 2005), to mediate estrogen-induced PKCε-dependent mechanical sensitization (Kuhn et al. 2008). We now observed estrogen and the GPR30-ligand, G-1, to cause rapid microtubule-disassembly. A GPR30-mediated effect would be in accordance with studies suggesting TRPV1 to be downstream of G protein-coupled receptors such as Neurokinin-1 (NK-1) receptor (Zhang et al. 2007), P2Y2 (Moriyama et al. 2003; Malin et al. 2008), and G protein coupled receptor mediated actions of the vasoactive endothelins (Plant et al. 2007; Malin et al. 2008) in part mediated by PKC-phosphorylation of TRPV1-S800. Thus, our current results suggest that also the activation of NK-1, P2Y2 receptors, and other G protein coupled receptors results in a TRPV1-dependent destabilization of the microtubular cytoskeleton. In turn it is tempting to speculate that estrogen-induced microtubule destabilization depends on TRPV1 also in cellular systems beyond the nociceptive neuron.

Detailing cellular mechanisms of GPR30 agonists is also therapeutically of interest. The ERα/ERβ-antagonist and GPR30-agonist, fulvestrant, is widely used in anti-breast cancer therapy. One prominent side effect in about 20% patients is therapy-induced long-lasting pain (Vergote and Abram 2006). Finding now GPR30-agonists to result in TRPV1-mediated microtubule destabilization, urges novel therapeutic research; (i) is fulvestrant-induced pain dependent on TRPV1 expression? Could therefore analgesic substances be specifically introduced into these neurons by opening the pore of TRPV1 (Binshtok et al. 2007)? And (ii) is the action of fulvestrant on the microtubule network crucial for the onset of fulvestrant-induced pain? Thus, could the block of the microtubule–TRPV1 interaction through TRPV1 down-regulation or interaction-inhibiting peptides abolish the sensitizing effect?

TRPV1, a role in cytoskeleton-dependent mechanical hyperalgesia

Estrogen and GPR30-agonists injected into the skin of male rats induces strong mechanical hyperalgesia via a PKCε-dependent signaling pathway (Hucho et al. 2006; Kuhn et al. 2008). Controversy exists in the literature about the role of TRPV1 in mechanical sensitization. Evidently mechanical hyperalgesia can be induced in TRPV1 knock out animals by, for example, formalin, complete Freud's adjuvant, carrageenan, and partial nerve lesion (Caterina et al. 2000; Bolcskei et al. 2005; Jin and Gereau 2006; Kanai et al. 2007). Thus, seemingly a central role of TRPV1 in mechanical hyperalgesia can be ruled out. Nevertheless, recently TRPV1-dependent mechanical sensitization was reported (Jones et al. 2005; Tender et al. 2008; Wang et al. 2008; Ro et al. 2009; Vardanyan et al. 2009). A solution for this apparent conundrum has not been suggested. Our data confirm both aspects and might suggest a solution. PKCε-dependent mechanical sensitization can be induced in wildtype but also in TRPV1 knock down animals, confirming that TRPV1 is not an essential component in mechanical hyperalgesia. Nevertheless, the presence/absence defines the effect of microtubule-altering drugs on the induction of PKCε-dependent mechanical hyperalgesia. Thus, we propose that TRPV1 is not part of the mechano-sensitization machinery but is a modulator of some signaling pathways toward this machinery. Therefore, the opposing reports of a role for TRPV1 in mechanical hyperalgesia might reflect different signaling mechanisms inducing the mechanical hyperalgesia, being for example microtubule-dependent versus microtubule-independent sensitization paradigms.

Conclusion

We describe a novel function of TRPV1. We find the C-terminus of TRPV1 to be necessary for PKCε-signaling toward microtubule destabilization in an ion conductivity-independent manner. The TRPV1 signaling complex serves as signaling intermediate but not as start or endpoint of signaling. Demonstrating the influence of estrogen- and GPR30-agonist-initiated intracellular signaling on TRPV1 opens the TRP-channel field to the recent progress in fast hormone signaling as well as the multitude of aspects of cytoskeleton regulation. It will be interesting which aspect of the dynamic interaction with the microtubular network is of importance in pain. In turn, the potential of the modulation of the microtubule-TRPV1 interaction e.g. by down-regulation as done in our rat behavioral experiment or by inhibition of the interaction with small compounds bears potential for novel therapeutic approaches has to be tested.

Acknowledgements

The work was funded by the Max Planck Institute for molecular Genetics, by the Studienstiftung des Deutschen Volkes, by MICINN (SAF2006-02580, CONSOLIDER-INGENIO 2010, CSD2008-00005) and Fundación la Marató de TV3.

Ancillary