Distinct properties of Ca²⁺–calmodulin binding to N- and C-terminal regulatory regions of the TRPV1 channel

Transient receptor potential (TRP) vanilloid 1 (TRPV1) is well recognized as a polymodal molecular pain receptor that responds to a wide range of stimuli including noxious heat >43°C, chemical agonists, and protons (Caterina et al., 1997; Tominaga et al., 1998). TRPV1 responses to these stimuli signal the exposure to extreme temperature or tissue damage to elicit appropriate protective mechanisms. TRPV1 is well characterized for its activation by capsaicin, a naturally occurring compound found in chili peppers (Caterina et al., 1997). Prolonged activation of TRPV1 with capsaicin results in acute desensitization, where TRPV1 activity decreases during the course of stimulation. In contrast, repeated stimulation of TRPV1 results in tachyphylaxis, where TRPV1 activation decreases with successive stimulation followed by loss of response to any subsequent stimuli.

TRPV1 belongs to the TRP superfamily of nonselective cation channels with high relative permeability for Ca²⁺ (Caterina et al., 1997). All TRP channels share a similar topology comprising six predicted helical transmembrane segments (S1–S6) and large N- and C-terminal (CT) cytosolic regions for integrating cellular signaling with channel activity (Gaudet, 2008). The N-terminal region of TRPV1 contains an ankyrin repeat domain (ARD), of which a structure is available (Lishko et al., 2007). The cytoplasmic C terminus, of unknown structure, contains multiple sites for interactions with modulatory factors. TRPV1 functions as a tetramer, with S5–S6 forming the pore through which ion conduction occurs. Opening of TRPV1 after channel activation allows the influx of cations, including Ca²⁺, into the cell.

The intracellular Ca²⁺ concentration is kept as low as 10⁻⁷ M compared with a high external Ca²⁺ concentration up to 10⁻³ M (Yamniuk and Vogel, 2004). The steep Ca²⁺ concentration gradient across the membrane in combination with the high relative permeability of TRPV1 for Ca²⁺ ensures rapid Ca²⁺ influx into the cells upon TRPV1 activation. However, intracellular calcium homeostasis is crucial for normal cell function (Han et al., 2007), and continuous capsaicin activation of cells expressing TRPV1 or stimulation of TRPV1 with the potent agonist resiniferatoxin results in severe cytotoxicity and cell death (Caterina et al., 1997; Karai et al., 2004). Ca²⁺ ions are involved in various cellular processes including several Ca²⁺-dependent desensitization mechanisms identified for TRPV1. One is the Ca²⁺-dependent recruitment of the calcineurin phosphatase to TRPV1 via a scaffolding protein, promoting the dephosphorylation and dampening of TRPV1 activity (Docherty et al., 1996; Mohapatra and Nau, 2005; Zhang et al., 2008). Similarly, Ca²⁺-dependent stimulation of PLC may promote cleavage of the sensitizing agent phosphatidylinositol-4,5-bisphosphate (PIP₂) (Liu et al., 2005; Stein et al., 2006; Lishko et al., 2007; Lukacs et al., 2007; Yao and Qin, 2009; Mercado et al., 2010). Although depletion of PIP₂ is thought to play a major role in desensitization (Lukacs et al., 2007; Mercado et al., 2010), artificial depletion of PIP₂ does not completely reproduce the near-full desensitization observed in the presence of calcium (Lukacs et al., 2007).

A third proposed mechanism is the inactivation of TRPV1 via interactions with Ca²⁺–calmodulin (CaM) (Numazaki et al., 2003; Rosenbaum et al., 2004; Lishko et al., 2007; Grycova et al., 2008). CaM is a ubiquitous and highly conserved 17-kD protein. It comprises two globular domains, the N- and C-lobes, connected by a central flexible linker that allows CaM to assume different conformations (Barbato et al., 1992; Tjandra et al., 1995). Each lobe contains two helix-loop-helix Ca²⁺-binding motifs known as EF hands and can thus bind four Ca²⁺ ions (Vetter and Leclerc, 2003). Apo–CaM and Ca²⁺–CaM interact differently with many proteins including ion channels (Gordon-Shaag et al., 2008), making them versatile modulators of various proteins. CaM exhibits diverse mechanisms, sometimes acting through multiple sites, as shown for several TRP channels (Zhu, 2005); its target recognition mechanisms are nearly as promiscuous as its functions (Hoeflich and Ikura, 2002; Yamniuk and Vogel, 2004).

Although TRPV1 does not have a recognizable CaM-binding motif (Rhoads and Friedberg, 1997), at least two CaM-binding sites have been identified in TRPV1 cytosolic domains. In vitro studies demonstrated that Ca²⁺–CaM binds isolated TRPV1 peptides from the TRPV1-ARD (Rosenbaum et al., 2004). In addition, Ca²⁺–CaM competes for a shared binding site on TRPV1-ARD with ATP, which sensitizes the channel to capsaicin and prevents tachyphylaxis (Lishko et al., 2007). A second CaM-binding site was localized to a short 35-residue segment in the C terminus of TRPV1 (TRPV1-CT) through in vitro binding assays (Numazaki et al., 2003). Deletion of this segment disrupts desensitization of TRPV1 to repeated stimulation by capsaicin. Furthermore, several mutations to positively charged residues within this region decreased the Ca²⁺–CaM-binding affinity of TRPV1-CT (Grycova et al., 2008). It is unclear how Ca²⁺–CaM participates in desensitization of TRPV1 through its putative CaM-binding sites. It has been postulated that Ca²⁺–CaM may bridge the two sites, leading to a closed channel (Lishko et al., 2007).

Here, we present the first structural view of Ca²⁺–CaM in complex with the 35-residue CaM-binding peptide from the TRPV1-CT (TRPV1-CT35: residues 767–801). Despite the absence of a classical CaM recognition motif, the complex is supported by hydrophobic anchors and electrostatic interactions similar to well-characterized Ca²⁺–CaM peptide complexes. Our in vitro binding studies revealed that the C-lobe of Ca²⁺–CaM is the major determinant of binding to either TRPV1-CT or TRPV1-ARD. We examined whether Ca²⁺–CaM can bridge contacts between TRPV1-ARD and TRPV1-CT35, but our results do not support such a ternary complex. We used mutations that disrupt Ca²⁺–CaM binding at either TRPV1-ARD (K155A) or TRPV1-CT (W787A) to show that the W787A TRPV1 mutant is sufficient to disrupt the interaction of Ca²⁺–CaM with full-length TRPV1, consistent with our in vitro observation that the TRPV1-CT is a high affinity binding site, with K_d = 5.4 ± 0.6 × 10⁻⁸ M. Finally, although the K155A mutation abolished tachyphylaxis, W787A caused slowed and reduced TRPV1 desensitization. In summary, the TRPV1-ARD is crucial for desensitization, although it is unclear whether this is through a direct interaction with CaM, whereas TRPV1-CT is the major CaM interaction site but plays only a minor role in the capsaicin-induced desensitization we measured.

The TRPV1-CT regions (759–802, TRPV1-CT44; and 767–801, TRPV1-CT35) were cloned into SacI and EcoRI sites of pMALC2 (New England Biolabs, Inc.) as N-terminal maltose-binding protein (MBP) fusions with a PreScission (GE Healthcare) protease site (LEVLFQGP). CaM C-lobe (residues 76–148) was cloned into NdeI and BamHI sites of pET21a (EMD Millipore). Full-length TRPV1 was expressed from pTracer-CMV2 (Lishko et al., 2007) for electrophysiology or pCDNA3-FLAG for pulldown assays. TRPV1 containing the glycosylation site mutation N604S (Rosenbaum et al., 2002) was cloned into NdeI and NotI sites of pCDNA3-FLAG. pCDNA3-FLAG was generated by ligating a short double-stranded oligonucleotide (5′-GATCACATATGGGATCCGAATTCGTCGACACTAGTGACGTCGCGGCCGCTGATTACAAGGATGACGACGATAAGTGACTCGAG-3′ and 5′-GGCCCTCGAGTCACTTATCGTCGTCATCCTTGTAATCAGCGGCCGCGACGTCACTAGTGTCG ACGAATTCGGATCCCATATGT-3′) into the BamHI and NotI sites of pCDNA3 and mutating the NdeI site within the CMV promoter to TATATG. Cysteine mutants of CaM and TRPV1-CT were introduced using QuikChange site-directed mutagenesis (Agilent Technologies) and confirmed by DNA sequencing.

Purification of CaM and TRPV1-ARD was performed as described previously (Drum et al., 2001; Lishko et al., 2007). CaM C-lobe, CaM_A15C, and CaM_E127C were purified as CaM. MBP-TRPV1-CT44 or -CT35 was expressed in BL21(DE3) induced at OD₆₀₀ = 0.6 with 100 µM IPTG at 37°C for 2 h. Cells were lysed by sonication in buffer A (20 mM Tris-HCl, pH 7.5, 200 mM NaCl, 0.5 mM PMSF, and 1 mM benzamidine) with 0.1% Triton X-100, 40 µg/ml DNase, 80 µg/ml RNase, and 200 µg/ml lysozyme. The cleared lysate was diluted at 1:1 with buffer A supplemented with 2 mM CaCl₂, loaded onto amylose resin (New England Biolabs, Inc.), washed with buffer B (20 mM Tris, pH 7.5, 200 mM NaCl, 1 mM DTT, and 1 mM CaCl₂) plus 0.5 mM PMSF, and eluted with buffer B plus 10 mM maltose. To obtain TRPV1-CT44/35, MBP-TRPV1-CT44/35 was concentrated and cleaved with PreScission protease overnight at 4°C. TRPV1-CT44/35 was purified on a Superdex 75 (GE Healthcare) in 20 mM Tris, pH 7.5, 50 mM NaCl, and 1 mM DTT, concentrated, and stored at −80°C. Preparation of MBP-TRPV1-CT35 for binding assays was as above with slight modifications: (a) CaCl₂ was excluded from all buffers; (b) 1 mM EDTA was added to lysis buffer; and (c) additional wash with buffer C (20 mM Tris, pH 7.5, 25 mM NaCl, and 1 mM DTT) and 1 mM PMSF was done before elution with buffer C and 10 mM maltose.

To obtain Ca²⁺-CaM–TRPV1-CT44/35 complex, CaM and excess TRPV1-CT44/35 (1:1.25–2.0 molar ratio) were incubated in 20 mM Tris, pH 7.5, 200 mM NaCl, and 0.15 mM CaCl₂. Ca²⁺-CaM–TRPV1-CT44 complex was isolated on Superdex 75 (GE Healthcare).

Crystals of Ca²⁺-CaM–TRPV1-CT35 (mixed at a 1.3:1 ratio at ∼8 mg/ml in 20 mM Tris-HCl, pH 7.5, 100 mM NaCl, 1 mM CaCl₂, and 1 mM DTT) were grown by vapor diffusion at room temperature against reservoir solution (0.1 M MES, pH 6.0, and 2.2–2.8 M (NH₄)₂SO₄), transferred into reservoir plus 25% glycerol, and flash-cooled in liquid nitrogen. Crystals of the two disulfide cross-linked complexes were grown in very similar conditions. Diffraction data, collected at beamlines 24-ID-C and -E (Advanced Photon Source, Argonne National Laboratory) at 0.97949 Å, were processed using HKL2000 (Otwinowski and Minor, 1997). The structure was determined by molecular replacement using CaM coordinates of Protein Data Bank accession number 3DVE (Kim et al., 2008) as a search model. Model building was performed using COOT (Emsley and Cowtan, 2004), and restrained TLS refinement was performed using REFMAC5 (Murshudov et al., 1997). The final wild-type model includes CaM residues 3–148 and TRPV1-CT35 residues 784–798 (deposited in the Protein Data Bank under accession no. 3SUI), with 93% of residues in most favored, 7% in allowed, and none in disallowed regions of the Ramachandran plot. Data and refinement statistics are shown in Table 1 and Table S1.

For CaM-agarose binding of MBP-tagged TRPV1-CT peptide, 60 µg MBP-TRPV1-CT35 in 500 µl of binding buffer (20 mM Tris-HCl, pH 7.5, 100 mM NaCl, 0.1% TWEEN20, and 1 mM DTT) supplemented with either 1 mM EGTA or 1 mM CaCl₂ was incubated with preequilibrated CaM-agarose (100 µl of 50% slurry; Sigma-Aldrich) at 4°C for 1 h. Beads were washed three times with 1 ml of the respective binding buffers and resuspended in sample dye plus 5 mM EGTA, and bound protein was analyzed by SDS-PAGE.

For CaM-agarose binding of FLAG-tagged TRPV1, HEK293 cells were transiently transfected with the pCDNA3-FLAG–based vectors using Lipofectamine 2000 (Invitrogen). Cells were harvested in TNE buffer (10 mM Tris-HCl, pH 7.5, 150 mM NaCl, and 1 mM EDTA) 30–40 h after transfection, collected by centrifugation at 2,000 g, and lysed in 1 ml TNE buffer supplemented with 1× Complete protease inhibitor (Roche) and 1% IGEPAL (Sigma-Aldrich). The detergent-soluble fraction was divided into two fractions, adding EDTA to 4 mM to one and CaCl₂ to 2 mM to the second. CaM-agarose (100 µl of 50% slurry in TNE buffer supplemented with 1× protease inhibitor and either 4 mM EDTA or 2 mM CaCl₂) was added. After a 2-h incubation, beads were washed five times with 1 ml of the respective buffers and eluted in 80 µl of 2× SDS sample buffer supplemented with 10 mM EGTA.

Fluorescence experiments were performed using a fluorescence spectrophotometer (Cary Eclipse; Agilent Technologies) at room temperature, in 10 mM Tris-HCl, pH 7.5, and 100 mM KCl, with 1 mM CaCl₂ or 5 mM EDTA. Tryptophan excitation was set to 295 nm to minimize contributions of CaM tyrosines to the emission spectra. Emission spectra were recorded from 305 to 400 nm. Excitation and emission bandwidths were set at 5 nm. For intrinsic Trp fluorescence measurements of TRPV1-CT44, 10-mm pathlength cuvettes and a 1-ml volume were used to measure 10 µM TRPV1-CT44 alone or with 10 µM CaM in the presence of CaCl₂ or EDTA. For K_d determination, a 5-mm pathlength quartz cuvette (1 ml volume) was used. TRPV1-CT44 at 0.1 µM in 1 ml was titrated with CaM from stocks of 0.04–1 mM, such that the final added volume was <2% of total volume. Fluorescence intensity at 330 nm was monitored, and results from three replicates were analyzed using one site-specific binding nonlinear regression analysis (Prism 5; GraphPad). Nonspecific binding from titrations of TRPV1-CT44 with CaM in EDTA was subtracted from total binding from the same titrations in CaCl₂.

SEC runs were performed as described previously (Lishko et al., 2007). In brief, 60 µM of TRPV1-ARD and 60 µM of CaM or CaM mutants were preincubated in running buffer (20 mM HEPES, pH 7.4, 140 mM NaCl, 1 mM DTT, and 0.16 mM CaCl₂) for 30 min on ice, followed by separation on a Superdex 75 column (GE Healthcare) at 4°C.

Possible disulfide bonds between Ca²⁺–CaM and TRPV1-CT35 were predicted (CaM_A15/V1-CT35_N789 and CaM_E127/V1-CT35_R785) based on Cβ-Cβ distances using the SSBOND program (Hazes and Dijkstra, 1988). Both pairs were generated by mutagenesis, and all four resulting constructs were purified as above. To test cross-linking, 20 µg CaM_E127C or CaM_A15C alone or with TRPV1-CT35_R785C or TRPV1-CT35_N789C, respectively, was mixed in 20 mM Tris-HCl, pH 7.5, 100 mM NaCl, and 1 mM CaCl₂, and incubated on ice for 20 min. DTT (10 mM final) or CuSO₄-1,10 phenanthroline (1–4 mM final) was added and incubated for an additional 20 min. Samples were resolved on 16% T 3% C Tricine SDS-PAGE with 4% T 3% C stacking gel as described previously (Schägger, 2006), in either reducing or nonreducing sample buffers, using running buffer supplemented with 1 mM CaCl₂.

For purification of CaM_E127C/V1-CT35_R785C, CaM_E127C and TRPV1-CT35_R785C were mixed at a 1:1.5 ratio in 20 mM Tris-HCl, pH 7.5, 100 mM NaCl, and 1 mM CaCl₂, and 1–4 mM CuSO₄-1,10 phenanthroline was added after 20 min, with an additional 20-min incubation on ice. For purification of CaM_A15C/V1-CT35_N789C, CaM_A15C and TRPV1-CT35_N789C were mixed at a 1:2 ratio in 20 mM Tris-HCl, pH 7.5, 100 mM NaCl, 2 mM CaCl₂, 3 mM GSH, and 0.3 mM GSSG and incubated at room temperature overnight. Cross-linked samples were dialyzed in 20 mM Tris-HCl, pH 7.5, 100 mM NaCl, and 1 mM CaCl₂ and purified on a Superdex 75 (GE healthcare) column.

Whole cell patch-clamping recordings were performed in HEK293 cells transiently transfected using Lipofectamine and PLUS reagent (Invitrogen). Currents were recorded 24–36 h after transfection in voltage-ramp experiments as described previously (Lishko et al., 2007) or in voltage-clamp experiments at a holding potential of −60 mV. Standard bath solution contained 10 mM HEPES, pH 7.4, 150 mM NaCl, 5 mM KCl, 2 mM CaCl₂, 1 mM MgCl₂, and 10 mM glucose. Intracellular solution contained 10 mM HEPES-CsOH, pH 7.2, 140 mM Cs-methanesulfonate, 10 mM EGTA, and 2.5 mM NaCl.

Transfected HEK293 cells were rinsed twice with ice-cold PBS, pH 8.0, and incubated with 0.5 mg/ml sulfo-NHS-Biotin (Thermo Fisher Scientific) in PBS, pH 8.0, for 30 min at 4°C. Cells were rinsed twice with cold 100 mM glycine in PBS and three times with PBS, lysed in 1 ml of RIPA buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, 1 mM EDTA, and 1× protease inhibitor cocktail from Roche), for 1 h at 4°C. Cleared lysates normalized for total protein concentration were incubated with 50 µl of 50% streptavidin-agarose slurry (Sigma-Aldrich) for 1 h at 4°C. The beads were washed three times with RIPA buffer and eluted with 50 µl of 2× SDS sample buffer.

Table S1 contains data and refinement statistics for the two cross-linked Ca²⁺-CaM–TRPV1-CT35 crystal structures. Fig. S1 shows formation of the Ca²⁺-CaM–TRPV1-CT44 complex by SEC. Fig. S2 shows CaM-agarose pulldown assay results for several TRPV1-CT35 mutants. Fig. S3 shows SEC traces of experiments with rat TRPV1-ARD and CaM mutants. Fig. S4 shows SEC traces probing a possible ternary interaction between TRPV1-ARD, CaM, and TRPV1-CT. Fig. S5 illustrates the crystal structures of the two cross-linked Ca²⁺-CaM–TRPV1-CT35 complexes. Fig. S6 shows voltage-clamp experiments and cell surface biotinylation of TRPV1 mutants expressed in HEK293 cells.

A short 35-residue segment (residues 767–801) was identified as a CaM-binding site in TRPV1-CT (Numazaki et al., 2003). In the same study, a longer 44-residue segment (residues 759–802) spanning this site was observed to bind CaM in the absence or presence of Ca²⁺ in a GST pulldown assay. We used SEC to evaluate the interaction of CaM to the 44-residue segment (TRPV1-CT44) and the 35-residue segment (TRPV1-CT35; Fig. 1 A). TRPV1-CT44 and TRPV1-CT35 undergo a large shift in elution volume when mixed with CaM in the presence of Ca²⁺, confirming that a complex is formed (Figs. 1 B and S1; data shown for TRPV1-CT44).

To obtain a quantitative measurement of Ca²⁺–CaM binding to TRPV1-CT, we took advantage of the intrinsic tryptophan present in TRPV1-CT peptides to monitor its binding to Ca²⁺–CaM. We used TRPV1-CT44, which is more readily produced in large quantities and high purity than TRPV1-CT35. The fluorescence emission peak of tryptophan in TRPV1-CT44 is 354 nm (Fig. 1 C). In the presence of CaM in CaCl₂ but not EDTA, the fluorescence emission peak blue-shifted to 330 nm and fluorescence intensity was enhanced, consistent with a tryptophan residue in a hydrophobic environment (Gomes et al., 2000; Weljie and Vogel, 2000). Our results further confirm that TRPV1-CT interacts with CaM in a Ca²⁺-dependent manner. In a titration experiment, we measured that TRPV1-CT44 interacts with Ca²⁺–CaM in a 1:1 ratio with high affinity, K_d = 5.4 ± 0.6 × 10⁻⁸ M (Fig. 1 D), comparable to other Ca²⁺–CaM-binding targets (Crivici and Ikura, 1995). Our results establish the distal TRPV1-CT region as a high affinity binding site for Ca²⁺–CaM.

TRPV1 does not have a classical CaM recognition motif (Rhoads and Friedberg, 1997). To better understand the Ca²⁺–CaM interaction with TRPV1-CT, we determined the crystal structure of Ca²⁺–CaM in complex with TRPV1-CT35 (residues 767–801) to 1.95-Å resolution (Fig. 2 A and Table 1). The asymmetric unit contains a 1:1 Ca²⁺-CaM–TRPV1-CT35 complex. Residues 784–798 of TRPV1-CT were observed in the structure (Fig. 2 A), with the electron density well defined for residues 785–797 (Fig. 2, B and C). As expected, four Ca²⁺ ions are observed in the structure, with two in each CaM lobe. TRPV1-CT35 adopts an elongated structure, with a helix (residues 787–796) flanked by two extended peptide regions (Fig. 2 A). The helical peptide is clasped by both Ca²⁺–CaM lobes in an antiparallel orientation, with its N terminus interacting with the Ca²⁺–CaM C-lobe and its C terminus bound to the N-lobe. This overall Ca²⁺-CaM–TRPV1-CT35 structure is similar to several canonical Ca²⁺–CaM peptide complexes (Yamniuk and Vogel, 2004), including those of smooth muscle light chain kinase (Meador et al., 1992), Ca²⁺–CaM-dependent kinase kinase (CaMKK) (Kurokawa et al., 2001), and endothelial nitric oxide synthase (Aoyagi et al., 2003).

In canonical Ca²⁺–CaM peptide complexes, the peptide is anchored through interactions of two large hydrophobic residues to hydrophobic pockets, one on each Ca²⁺–CaM lobe (Fig. 3 A). The relative sequence position of the anchor residues defines the binding mode of Ca²⁺-CaM–peptide complexes. TRPV1-CT35 is indeed anchored to hydrophobic pockets on the C- and N-lobes of Ca²⁺–CaM through hydrophobic anchor residues W787 and L796 at the C- and N-lobe, respectively (Fig. 3 B). TRPV1-CT35 therefore has a 1–10 motif like that of CaM-dependent kinase II (Meador et al., 1993).

The N-terminal anchor, W787, forms extensive hydrophobic contacts with surrounding residues F92, L105, M124, and M144, lining a deep hydrophobic pocket in the Ca²⁺–CaM C-lobe (Fig. 3 B). These four residues form the FLMM_C tetrad, which is conserved in its interaction with its target anchor, whereas the FLMM_N tetrad in the N-lobe (F19, L32, M51, and M71) is more variable (Ataman et al., 2007). The TRPV1-CT anchor, L796, contacts F19, M51, and M71 from the FLMM_N tetrad (Fig. 3 B). In addition to the anchor residues, other hydrophobic residues along the TRPV1-CT35 α helix, namely F790, A791, L792, V793, P794, and L795, also form extensive hydrophobic contacts with both the N- and C-lobes of Ca²⁺–CaM (Fig. 3 D).

Although hydrophobic residues are major determinants of CaM interactions, basic residues mediate electrostatic contacts with the highly acidic surface of CaM and can drive the orientation of the bound CaM target (Osawa et al., 1999; Yamniuk and Vogel, 2004). Antiparallel Ca²⁺–CaM targets have positive residues near the N-terminal anchor, whereas in the case of parallel Ca²⁺–CaM targets, basic residues flank the CT anchor (Fig. 3 A). This preferential distribution of basic residues around one anchor or the other in an antiparallel or parallel target can be understood from examining the complementary electrostatic surfaces on Ca²⁺–CaM, which are important in determining the orientation of the bound peptide (Osawa et al., 1999). In our structure, basic residues R785 and K788 flank the N-terminal hydrophobic anchor W787 near the highly electronegative opening in the Ca²⁺–CaM structure, whereas the single CT basic residue, R797, is oriented toward the less electronegative opening (Fig. 4 A). In summary, although no canonical CaM-binding sequence is found in the TRPV1-CT, residues 785–797 of TRPV1 interact with Ca²⁺–CaM using hydrophobic anchors and electrostatic interactions similar to those found in well-characterized Ca²⁺–CaM-target complexes.

The Ca²⁺-CaM–TRPV1-CT35 complex exhibits several unique features. The first is the structure of the TRPV1-CT peptide, which consists of a short 10-residue helix flanked by two extended peptide regions (Fig. 2 A), unlike classical Ca²⁺–CaM targets that assume longer α-helical structures (Rhoads and Friedberg, 1997; Yamniuk and Vogel, 2004). Sets of available structures comprising eight parallel and 31 antiparallel Ca²⁺-CaM–peptide structures had Ca²⁺–CaM-interacting helices averaging 18.4 and 18.8 residues in length, respectively. Within these sets, only two other Ca²⁺–CaM targets have short helices similar to TRPV1: the Ca²⁺–CaM-binding domain of myristoylated alanine-rich C kinase substrate (MARCKS; Protein Data Bank accession no. 1IWQ; eight-residue helix) (Yamauchi et al., 2003) and the Ca²⁺-CaMKK (Protein Data Bank accession no. 1CKK; 11-residue helix) (Osawa et al., 1999). The MARCKS and TRPV1-CT peptides share the same antiparallel orientation with respect to Ca²⁺–CaM, whereas the CaMKK peptide has a parallel orientation. The conformation of Ca²⁺–CaM in the TRPV1-CT complex is very similar to that of the MARCKS peptide-bound Ca²⁺–CaM (root-mean-square deviation [RMSD] = 0.646 Å). However, in contrast to the 1–10 binding mode of TRPV1-CT35, MARCKS has no hydrophobic anchor bound to the Ca²⁺–CaM N-lobe hydrophobic pocket. This highlights not only the conformational flexibility of Ca²⁺–CaM to recognize its different targets but also the variability of CaM targets that can fit into the same Ca²⁺–CaM scaffold.

A second feature unique to TRPV1-CT is that the two hydrophobic anchors are at the boundaries of the α-helical region (Fig. 4 B), whereas those of classical Ca²⁺–CaM targets are present within the α helix. Out of the nearly 40 available Ca²⁺-CaM–peptide complex structures that we analyzed, the only Ca²⁺–CaM target that also carries a nonhelical hydrophobic anchor is the CaMKK peptide. However, the binding modes of CaMKK and TRPV1-CT are different in all other aspects. The CaMKK peptide exhibits a parallel 1–16 binding mode, with an N-terminal tryptophan hydrophobic anchor on the α helix and its CT anchor, a phenylalanine residue, on the extended CT region (Osawa et al., 1999). In summary, the structure of the Ca²⁺-CaM–TRPV1-CT complex revealed yet another variation of Ca²⁺–CaM target recognition motifs, with a 1–10 antiparallel binding mode in which both hydrophobic anchors bound the peptide helix.

Interestingly, the Ca²⁺–CaM-binding site observed in our crystal structure overlaps a previously proposed PIP₂-binding site required for PIP₂-mediated inhibition of TRPV1 activity (Prescott and Julius, 2003), although this study’s conclusions are controversial (see Discussion). In the study, substitution of several basic residues by neutral polar residues resulted in increased sensitivity of TRPV1 to both chemical and heat stimuli. This phenotype was interpreted as a relief of inhibitory PIP₂ interactions with TRPV1. In particular, the R785Q/K788Q substitution pair increased TRPV1 sensitivity to heat by shifting the temperature threshold of activation to lower temperatures. Incidentally, R785 and K788 are the two basic residues that flank the hydrophobic tryptophan anchor in our Ca²⁺–CaM-binding site (Fig. 3 A). To examine whether the region observed in our structure represents bona fide determinants for Ca²⁺–CaM binding in solution, we mutated several key residues in an MBP-fused TRPV1-CT35 and examined interactions with Ca²⁺–CaM using a CaM-agarose pulldown assay. Tryptophan is often found in protein interface hot spots responsible for the bulk of binding energy (Bogan and Thorn, 1998), and as anticipated, substitution of the hydrophobic anchor W787 with alanine caused a severe disruption of Ca²⁺–CaM binding, underscoring the importance of the tryptophan anchor in mediating hydrophobic interactions with Ca²⁺–CaM (Fig. 3 C). Alanine substitutions at the R785/K788 pair also impaired Ca²⁺–CaM binding, indicating that these basic residues are likewise crucial for interaction with Ca²⁺–CaM (Fig. 3 C), consistent with a previously published study (Grycova et al., 2008). Charge-neutralizing R785Q/K788Q substitutions, which were previously shown to potentiate TRPV1 activity (Prescott and Julius, 2003), also impair binding to Ca²⁺–CaM (Fig. 3 C).

A separate study aimed at identifying active or hypersensitive TRPV1 mutants using a toxicity-based screen (Myers et al., 2008) identified several mutations that fell within our Ca²⁺–CaM-binding site: W787R, L792P, and L796P. To determine whether these mutations alter Ca²⁺–CaM binding, we studied the interaction of these mutants using our CaM-agarose pulldown assay. As with W787A, W787R severely disrupted Ca²⁺–CaM binding (Fig. 3 C). Although mutation of the L796 CT anchor to proline did not significantly disrupt binding of TRPV1-CT35 to Ca²⁺–CaM in our pulldown assay (Fig. S2), the L792P substitution severely disrupted Ca²⁺–CaM binding (Fig. 3 C). A proline substitution at L792, which is situated in the middle of the TRPV1-CT35 helix in the Ca²⁺–CaM-bound structure, would break the helical conformation of the peptide and thus likely displace other necessary Ca²⁺–CaM interacting groups. In contrast, the L796P substitution is located at the CT end of the helix, providing a likely explanation as to why this mutation is tolerated. Our binding assay agrees well with the observed Ca²⁺-CaM–TRPV1-CT structural features.

To examine whether phosphorylation of S800, which potentiates TRPV1 (Bhave et al., 2003), interferes with Ca²⁺–CaM binding, we mutated S800 to either an aspartic acid or glutamic acid to mimic a phosphorylation state. Neither S800D nor S800E interfered with Ca²⁺–CaM binding (Fig. S3). This result is consistent with our structural data, as S800 lies outside the region observed in our structure (Fig. 3 D) and suggests that disruption of Ca²⁺–CaM binding is not the cause of the observed phosphorylation-induced potentiation in TRPV1.

Collectively, our mutagenesis results confirmed that TRPV1-CT35 interacts extensively with Ca²⁺–CaM through both hydrophobic and electrostatic interactions (Fig. 3 D). Notably, mutation of the N-terminal tryptophan anchor alone is sufficient to severely diminish Ca²⁺–CaM interaction, suggesting that the C-lobe of Ca²⁺–CaM dominates the interactions with TRPV1-CT35.

The interactions of Ca²⁺–CaM with both TRPV1-ARD and TRPV1-CT are Ca²⁺ dependent and both regions have been implicated in TRPV1 desensitization, suggesting that both sites may work in concert. One hypothesis is that Ca²⁺–CaM bridges an interaction between the TRPV1-ARD and TRPV1-CT, leading to a closed channel (Lishko et al., 2007). CaM forms an analogous ternary complex in SK channels, for example, where CaM mediates dimerization of two channel subunits through binding of its N- and C-lobe to two different subunits (Schumacher et al., 2001). Having determined that the C-lobe dominates the TRPV1-CT35 interaction, we asked whether the TRPV1-ARD is similarly dominated by a specific CaM lobe.

We have been unable to measure the affinity of TRPV1-ARD for Ca²⁺–CaM because of technical difficulties with protein stability under various assay conditions, or to cocrystallize a Ca²⁺-CaM–TRPV1-ARD complex to determine its structure. Instead, we tested which CaM lobe principally drives the Ca²⁺-dependent interaction with TRPV1-ARD by using SEC to compare the impact of mutations at the CaM Ca²⁺-binding sites on the interaction. CaM1, CaM2, CaM3, and CaM4 were mutated at the first (D21A), second (D57A), third (D94A), or fourth (D130A) Ca²⁺-binding site, respectively. Rat TRPV1-ARD no longer interacted with any of the single-site CaM mutants (Fig. S3). However, rat TRPV1-ARD exhibits only a small shift by SEC when mixed with Ca²⁺–CaM, perhaps because of a weak interaction that is highly susceptible to perturbations. Slight differences in affinities for the distinct CaM mutants would be missed if any perturbation caused a loss of detectable binding. Therefore, we also tested interactions between the CaM mutants and chicken TRPV1-ARD, which in SEC experiments has a robust shift with wild-type Ca²⁺–CaM (Fig. 5 A). Chicken TRPV1-ARD showed no interaction with CaM1234, which has all four Ca²⁺-binding sites mutated (Fig. 5 B), demonstrating that the interaction with CaM is Ca²⁺ dependent. Chicken TRPV1-ARD retained the ability to interact with CaM12 but not CaM34 (Fig. 5, C and D), pointing to the importance of Ca²⁺–CaM C-lobe in interacting with TRPV1-ARD. Using a truncated CaM (residues 76–148), we further showed that the isolated Ca²⁺–CaM C-lobe can bind TRPV1-ARD (Fig. 5 E). Although chicken and rat TRPV1-ARD may interact differently with Ca²⁺–CaM, their close homology (69.5% identity and 82.0% similarity) suggests that they likely share a common binding mode (Phelps et al., 2007). Thus, our results suggest that the Ca²⁺–CaM C-lobe is crucial for binding to TRPV1-ARD.

Although the Ca²⁺–CaM C-lobe appears to dominate the interactions of both TRPV1-CT and TRPV1-ARD, it remained to be determined whether we could observe a ternary complex between TRPV1-ARD, a TRPV1-CT peptide, and Ca²⁺–CaM. Using SEC, a shift of the rat TRPV1-ARD peak was observed in the presence of isolated Ca²⁺-CaM–TRPV1-CT44 complex (Fig. S4). However, we could not unambiguously ascertain whether the observed TRPV1-ARD shift was caused by its interaction with a Ca²⁺-CaM–TRPV1-CT peptide complex or with Ca²⁺–CaM alone because of overlapping elution profiles of Ca²⁺–CaM and the Ca²⁺-CaM–TRPV1-CT44 complex. To resolve this problem, we introduced cysteine pairs to form disulfide cross-links between Ca²⁺–CaM and TRPV1-CT35. Any SEC shift of TRPV1-ARD in the presence of cross-linked Ca²⁺-CaM–TRPV1-CT35 could thereby be attributed to the binding of TRPV1-ARD to a Ca²⁺-CaM–TRPV1-CT peptide complex.

Two cysteine pairs were engineered based on predicted optimal disulfide bond sites using SSBOND (Hazes and Dijkstra, 1988): TRPV1-CT35 N789C cross-links to A15C in the N-lobe of Ca²⁺–CaM, and TRPV1-CT35 R785C cross-links to E127C in the Ca²⁺–CaM C-lobe (Fig. 6 A). Both sets of mutants display a Ca²⁺-dependent interaction similar to wild-type Ca²⁺–CaM and TRPV1-CT35, as shown in native gel (Fig. 6 B). Although the E127C–R785C pair cross-linked more efficiently under our original conditions (Fig. 6 C), we were able to isolate both cross-linked complexes with good purity (see Materials and methods). Both cross-linked complexes crystallized under very similar conditions and in the same space group as the wild-type complex, and we determined the Ca²⁺-CaM_E127C–TRPV1-CT35_R785C structure to 2.4-Å resolution, and the Ca²⁺-CaM_A15C–TRPV1-CT35_N789C to 2.1-Å resolution. Both structures exhibited only minimal local conformational variations (Fig. S5). Thus, the introduced cross-links in the Ca²⁺-CaM–TRPV1-CT peptide complex did not perturb the interactions observed in our original structure.

We then used SEC to investigate whether the cross-linked complexes interact with rat TRPV1-ARD. Neither TRPV1-ARD nor the cross-linked complexes showed a shift in elution volume when incubated together (Fig. 6 D); therefore, the SEC profiles showed no evidence of interaction between TRPV1-ARD and either cross-linked complex. Although we cannot rule out the possibility that cross-linking may interfere with possible conformational requirements for interaction with the TRPV1-ARD, the fact that neither Ca²⁺-CaM_E127C–TRPV1-CT35_R785C nor Ca²⁺-CaM_A15C–TRPV1-CT35_N789C interacted with TRPV1-ARD makes it unlikely that the cross-link itself prevented the interaction. Our results therefore indicate that TRPV1-ARD and TRPV1-CT do not form a ternary complex with Ca²⁺–CaM under the conditions we tested.

Although both the isolated TRPV1-ARD and TRPV1-CT were shown to interact with Ca²⁺–CaM in solution, it is unclear how these two sites contribute to Ca²⁺–CaM binding in full-length TRPV1. To test this, we introduced mutations K155A or W787A, which were shown to abolish Ca²⁺–CaM binding to TRPV1-ARD (Lishko et al., 2007) and TRPV1-CT (this study), respectively, and determined whether the resulting TRPV1 mutants can still bind Ca²⁺–CaM using a CaM-agarose pulldown assay. Wild-type TRPV1 expressed from HEK293 cells binds CaM in a Ca²⁺-dependent manner (Fig. 7 A). K155A did not affect binding of TRPV1 to Ca²⁺–CaM, whereas W787A severely disrupted binding to Ca²⁺–CaM, suggesting that the TRPV1-CT alone is sufficient for binding to Ca²⁺–CaM. Our results support TRPV1-CT as a high affinity Ca²⁺–CaM-binding site in full-length TRPV1.

Disruption of Ca²⁺–CaM binding to TRPV1-ARD was shown to diminish the Ca²⁺-dependent desensitization of TRPV1 to repeated capsaicin stimulations in whole cell patch-clamp electrophysiology experiments (Lishko et al., 2007). To examine the functional role of the high affinity TRPV1-CT Ca²⁺–CaM-binding site, we introduced mutations that abolish Ca²⁺–CaM binding to either TRPV1-ARD (K155A), TRPV1-CT (W787A), or both TRPV1-ARD and TRPV1-CT (K155A/W787A) and compared the capsaicin-evoked currents in HEK293 cells. We measured the maximal current amplitude evoked by each successive pulse of capsaicin using the whole cell patch-clamp technique in voltage ramps (Fig. 8) or at a holding potential of −60 mV (Fig. S6). As expected, repeated applications of capsaicin resulted in strong tachyphylaxis of wild-type TRPV1 currents, observed as decreased inward currents (Fig. 8 A, black trace). K155A showed little tachyphylaxis and slower activation kinetics as reported previously (Lishko et al., 2007). The slower activation kinetics of K155A is not caused by altered surface expression (Fig. S6) and is therefore an inherent quality of the mutant. Although the slower activation kinetics could suggest a constitutively desensitized phenotype, the steady-state inward currents of K155A are larger than those of wild type (Fig. 8 C). Furthermore, it has been reported that desensitization reduces the sensitivity of TRPV1 to capsaicin (Yao and Qin, 2009), whereas we previously determined that the K155A mutant displays normal capsaicin sensitivity (Lishko et al., 2007).

In contrast, the W787A mutant showed desensitization albeit with slower kinetics and significantly higher residual currents in subsequent capsaicin application compared with wild-type TRPV1. As most tachyphylaxis occurs between the first and second application (Mohapatra et al., 2003), we compared the mean maximal current amplitude at −80 mV of the second application to that of the first application. Wild-type TRPV1-expressing cells only retained 2.3 ± 0.4% of their maximal current at the second application, whereas W787A-expressing cells retained 14.4 ± 3.2% (n = 8 and P = 0.006). Our results with the W787A mutation are similar to those obtained when the TRPV1-CT35 region is deleted in full-length TRPV1 (Numazaki et al., 2003), showing that TRPV1-CT is indeed involved in the fast component of TRPV1 desensitization. Disruption of both Ca²⁺–CaM-binding sites (K155A/W787A) led to capsaicin-induced currents similar to the K155A mutation. In summary, the stronger phenotype elicited by K155A suggests a major role for TRPV1-ARD in TRPV1 desensitization, whereas the more subtle W787A phenotype suggests that TRPV1-CT plays a minor role in capsaicin-induced desensitization but may be involved in other responses not discernible under our experimental conditions.

Here, we present the structure of Ca²⁺–CaM bound to a short 35-residue TRPV1-CT region (TRPV1-CT35), one of two identified Ca²⁺–CaM-binding sites in TRPV1. The second Ca²⁺–CaM-binding site was previously identified in TRPV1-ARD (Rosenbaum et al., 2004), where Ca²⁺–CaM competes with a TRPV1 sensitizer, ATP, for binding (Lishko et al., 2007). Our affinity measurements indicate that TRPV1-CT represents a high affinity binding site for Ca²⁺–CaM (K_d = 5.4 ± 0.6 × 10⁻⁸ M). Although we could not measure the affinity of TRPV1-ARD for Ca²⁺–CaM because of technical difficulties, our SEC experiments suggest that the affinity is likely significantly lower than that of the TRPV1-CT. Consistently, a mutation that abolished Ca²⁺–CaM binding to TRPV1-CT (W787A), but not a corresponding one in TRPV1-ARD (K155A), severely disrupted binding to Ca²⁺–CaM in full-length TRPV1. The W787A mutation was sufficient to slow and reduce TRPV1 desensitization, consistent with a 35–amino acid deletion mutant that also exhibited reduced Ca²⁺-dependent desensitization of TRPV1 (Numazaki et al., 2003).

The interaction of Ca²⁺–CaM with TRPV1-CT35 is similar to that observed in canonical Ca²⁺–CaM/target complexes. The bound peptide adopts an α-helical conformation and binds Ca²⁺–CaM in an antiparallel manner through two hydrophobic anchors, W787 and L796, corresponding to a 1–10 binding motif. Sequence alignment of the TRPV1-CT CaM-binding site identified in the rat TRPV1 sequence with TRPV1 from other species shows that this site is conserved (Fig. 7 B), suggesting a conserved function.

The TRPV1-CT Ca²⁺–CaM-binding site identified in our structure overlaps a region previously identified as an inhibitory PIP₂-binding domain, although no direct binding was demonstrated (Prescott and Julius, 2003). A subsequent study found that the inhibitory effects of PIP₂ on TRPV1 were only detectable in whole cells, not excised patches, suggesting that PIP₂-mediated inhibition is indirect (Lukacs et al., 2007). Furthermore, PIP₂ can also potentiate TRPV1 (Liu et al., 2005; Lishko et al., 2007; Lukacs et al., 2007) and does so through a proximal CT region (residues 682–725) (Ufret-Vincenty et al., 2011). Together with these findings on TRPV1 regulation by PIP₂, our data support the distal CT region of TRPV1 as a physiologically important Ca²⁺–CaM-binding site.

The interactions of Ca²⁺–CaM with both TRPV1-ARD and TRPV1-CT are implicated in TRPV1 desensitization (Numazaki et al., 2003; Lishko et al., 2007). This brings up the question of whether Ca²⁺–CaM regulation of TRPV1 at TRPV1-ARD and TRPV1-CT constitutes integrated or separate mechanisms. Several findings suggested that both sites may work together in the desensitization mechanism. First, mutations or deletion of the Ca²⁺–CaM-binding site either in TRPV1-ARD or TRPV1-CT disrupts desensitization (Numazaki et al., 2003; Lishko et al., 2007). Second, a toxicity-based screen for active or hypersensitive TRPV1 phenotype (Myers et al., 2008) identified mutations that impair Ca²⁺–CaM binding in either TRPV1-ARD, such as K155E/A and K160E/A (Lishko et al., 2007), or TRPV1-CT, such as W787R and L792P (Fig. 3 C). One hypothesis is that Ca²⁺–CaM may bridge an interaction between TRPV1-ARD and TRPV1-CT (Lishko et al., 2007). A similar mechanism is postulated for CaM regulation of voltage-gated Ca_v1 channels (Dick et al., 2008), and CaM bridges two subunits of a Ca²⁺-activated K⁺ channel (Schumacher et al., 2001). Here, we generated two different cross-linked Ca²⁺-CaM–TRPV1-CT complexes but did not observe any ternary assemblies of these cross-linked complexes with TRPV1-ARD. This suggests that the Ca²⁺–CaM-mediated desensitization of TRPV1 may instead involve separate pathways. However, we cannot rule out the possibility that the introduced cross-links, which maintain the interactions observed in the un–cross-linked complex (Fig. S5), may restrict conformational changes required for interaction with TRPV1-ARD.

We observed a more pronounced disruption of desensitization in response to capsaicin when mutating the N-terminal K155A than the CT W787A, suggesting that the TRPV1-ARD is most important for tachyphylaxis. Mutation of the CT site, on the other hand, led to a more subtle phenotype of slowed and reduced desensitization (Fig. 8). Notably, the CT site was most important for Ca²⁺–CaM binding in our pulldown assays. Therefore, our data suggest that Ca²⁺–CaM binding may not be the dominant regulatory factor for TRPV1 desensitization. Overall, our data are consistent with a model where PIP₂ depletion is a major contributor to the acute desensitization of TRPV1 after activation by capsaicin (Lukacs et al., 2007; Mercado et al., 2010), with Ca²⁺–CaM binding to the TRPV1-CT likely playing a supporting role.

It is possible that the CT Ca²⁺–CaM-binding site plays a more important role in regulating other TRPV1 responses not tested in the current study. Several studies hint at a role of the TRPV1-CT Ca²⁺–CaM-binding site in tuning channel sensitivity. Increased sensitivity to heat and/or hypersensitivity was observed in several CT mutants (Prescott and Julius, 2003; Myers et al., 2008) and in the vampire bat TRPV1-S isoform in which the CT-binding site is absent (Gracheva et al., 2011).

In summary, we showed that TRPV1-CT represents a high affinity binding site for Ca²⁺–CaM. The TRPV1-ARD and TRPV1-CT likely provide two separate regulatory pathways, a model that is supported by both our biochemical and electrophysiological analyses. Our work provides new biochemical and structural information that enables future studies to fully decipher the molecular mechanisms behind the physiological regulation of TRPV1 responses.

We thank current and former laboratory members for technical help and discussions. Special thanks to Wilhelm Weihofen for advice and assistance on x-ray crystallography; Marcos Sotomayor and Ahmet Vakkasoglu for x-ray data collection; Alejandra Beristain-Barajas for technical assistance; and Hitoshi Inada for assistance with electrophysiology, discussions, and insights to this work. We are grateful to Narayanasami Sukumar and staff at the Northeastern Collaborative Access Team (NE-CAT), beamline 24-ID-C, for assistance with data collection.

This work was supported by National Institutes of Health (NIH; grant R01GM081340) and Klingenstein Awards to R. Gaudet. Use of APS beamlines was supported by NIH award RR-15301 and Department of Energy contract DE-AC02-06CH11357.

Edward N. Pugh Jr. served as editor.