Ligand stoichiometry of the cold- and menthol-activated channel TRPM8

Non-technical summary

The ion channel TRPM8, which is present in nerves that innervate the skin and mouth, mediates the sensation of cold temperatures and of chemical substances that evoke a cold sensation, such as menthol or eucalyptol. The molecular structures and biophysical mechanisms that allow this and related channels to integrate thermal and chemical information are poorly understood. In this work, we demonstrate that TRPM8 is able to bind up to four menthol molecules, and that every bound menthol molecule causes an equivalent stabilization of the open state of the channel. These results provide important fundamental insight into the working mechanism of temperature-sensitive channels, and may assist in the development of drugs to treat oversensitivity to cold temperatures and pain.

Abstract

Abstract

Temperature-sensitive transient receptor potential (TRP) channels play a key role in somatosensation, not only as primary thermosensors but also as chemosensors for various compounds that evoke a thermal sensation. The fundamental mechanisms whereby TRP channels translate ligand binding into opening of the ion conducting pore are, however, poorly understood, and the actual number of ligands that bind to these channels is fully unknown. Here, we investigated the ligand stoichiometry of the cold sensor and menthol receptor TRPM8. Based on a detailed biophysical analysis of tandem constructs of wild-type and menthol-insensitive TRPM8 subunits, we now provide direct evidence that each channel has four independent and energetically equivalent menthol interaction sites. These results suggest a fundamentally different ligand stoichiometry in TRP channels, in comparison with other major families of ligand-gated ion channels.

Introduction

Thermal cues from skin and mouth stimuli are conveyed by primary afferent sensory neurons that have their cell bodies in the trigeminal and dorsal root ganglia. Temperature-sensitive cation channels of the TRP superfamily have been identified as the main thermosensors in the sensory system. Intriguingly, several temperature-sensitive TRP channels can also act as ligand-gated channels, and this property underlies the chemesthetic sensation of various natural and synthetic compounds that evoke a thermal sensation. Well-known examples are the cooling agent menthol, which directly activates the cold sensor TRPM8, and the ‘hot pepper’ compound capsaicin, which has a similar effect on the heat sensor TRPV1 (Caterina et al. 1997; McKemy et al. 2002; Peier et al. 2002).

Despite their high physiological and pharmacological importance as ligand-gated channels, very little is known about the coupling between ligand binding and channel opening in TRP channels. Functional TRP channels are tetramers of four subunits with six transmembrane segments (S1–S6) (Hoenderop et al. 2003), and several studies have provided evidence that lipophilic agonists including menthol (TRPM8), capsaicin (TRPV1), and 4α-phorbols (TRPV4) bind to residues in transmembrane domains S2–S4 (Jordt & Julius, 2002; Bandell et al. 2006; Voets et al. 2007; Vriens et al. 2007). This suggested that a single TRP channel can potentially interact with four ligand molecules. However, the actual number of ligands that bind to activate a TRP channel or the energetic contribution of each individual ligand-binding event have remained fully unknown.

In this work, we addressed the ligand stoichiometry of TRPM8, by analysing the menthol sensitivity of channels with a defined composition of wild type and menthol-insensitive subunits. Our results indicate that up to four menthol molecules can independently bind to a single TRPM8 channel, and that each bound menthol causes a similar energetic stabilization of the open channel.

Methods

Construction of tandem tetrameric TRPM8 constructs

The different tandem tetrameric constructs were obtained by linking the coding sequences of wild type (wt) and mutant (mut) human TRPM8 in a head-to-tail fashion. For this purpose, two types of modified constructs were made in the wt and mut background: (1) a 3′-modified construct, in which the last amino acid-coding codon (AAA; Lys) and stop codon (TAC) were mutated to contain a unique PmeI restriction site (GTTTAAAC) followed by a unique AgeI restriction site (ACCGGT), and a new stop codon (TAA); and (2) a 5′–3′-modified construct containing the same 3′ modifications and in addition a unique EcoRV restriction site (GATATC) introduced in front of the 5′-ATG start codon. Dimeric constructs were then obtained by digestion of the 5′–3′-modified construct with EcoRV and AgeI, and ligation of the digestion product in the 3′-modified construct digested with PmeI and AgeI. This procedure was repeated to obtain the different tandem tetrameric constructs. As a consequence of the cloning strategy, two amino acids (Phe–Ile) were introduced between subunits, and four amino acids (Phe–Lys–His–Arg) were attached at the final C terminus (Supplemental Material, Fig. 1). Tandem constructs were verified by restriction digest analysis using NheI, by sequencing over the site of the mut point mutation, and by Western blot analysis (Supplementary Fig. 1C) using a previously described anti-TRPM8 antibody (Mahieu et al. 2007).

Electrophysiology and intracellular Ca²⁺ measurements

HEK293 cells were grown in Dulbecco's modified Eagle's medium (DMEM) containing 10% (v/v) fetal calf serum, 4 mm l-alanyl-l-glutamine, 100 U ml⁻¹ penicillin and 100 μg ml⁻¹ streptomycin at 37°C in a humidity controlled incubator with 10% CO₂. Cells were transiently transfected with different TRPM8 constructs cloned in the bicistronic pCAGGS-IRES-GFP vector using TransIT-293 transfection reagent (Mirus). Between 36 and 48 h after transfection, currents were recorded in the whole-cell configuration of the patch-clamp technique using an EPC-9 amplifier and PULSE software (HEKA Elektronik, Lambrecht/Pfalz, Germany). Data were sampled at 5–20 kHz and digitally filtered off-line at 1–5 kHz. Between 70 and 90% of the series resistance was compensated, reducing voltage errors to less than 10 mV. All patch-clamp experiments were performed using an intracellular solution containing (in mm): 150 NaCl, 5 MgCl₂, 5 EGTA and 10 Hepes, pH 7.4. The extracellular solution contained (in mm): 150 NaCl, 1 MgCl₂ and 10 Hepes, pH 7.4. When indicated, this solution was supplemented with (−)-menthol at concentrations between 10 and 1000 μm, from a 1 m stock in ethanol. Under these conditions, open TRPM8 channels display a linear current–voltage relation that reverses close to 0 mV. The temperature of the perfusate was controlled using a SC-20 dual in-line heater/cooler (Warner Instruments, Hamden, CT, USA). Reported temperature was measured using a TA-29 thermistor (Thermometrics Corp., Northridge, CA, USA) placed within 500 μm of the patch-clamped cell.

For intracellular Ca²⁺ measurements, cells were incubated with 2 μm Fura-2 acetoxymethyl ester for 30 min at 37°C. The fluorescence signal was measured during alternating illumination at 350 and 380 nm on an Olympus Cell-M fluorescence microscope. Calculation of [Ca²⁺]_i, including the correction for the temperature-dependent changes in Fura-2 affinity, were done as previously described (Vriens et al. 2004). The extracellular solution used in ratiometric [Ca²⁺]_i measurements contained (in mm): 150 NaCl, 6 KCl, 2 CaCl₂, 1.5 MgCl₂ and 10 Hepes, pH 7.4.

Data analysis and modelling

Data analysis was performed using Origin 7.0 (OriginLab Corp., Northampton, MA, USA). Group data are expressed as means ± SEM from at least five independent patch-clamp recordings. It should be noted that the tetramers generally yielded a much lower current density than the respective monomers. To ensure a robust analysis with minimal interference of background current, whole-cell recordings were only included when at 25°C (1) the outward current at +160 mV exceeded 500 pA, and (2) the rectification score (current at +100 mV divided by current at –100 mV) exceeded 10. With these criteria, there was no obvious interference between current amplitude and analysis of the parameters describing the channel's voltage dependence.

To determine the potential for half-maximal activation (V_1/2) and the apparent gating charge (z), steady-state current–voltage relations measured at the end of 100 ms voltage steps were fitted with the following equation:

where G_max is the maximal conductance, F the Faraday constant, R the gas constant and T the absolute temperature. In the case of TRPM8, this type of analysis is preferred to analyses based on tail current measurements, because very rapid deactivation often prevents an accurate determination of the peak tail current amplitude (for a discussion, see Voets et al. 2007). As described earlier, we found that menthol causes a change in V_1/2 without significant alterations of z or G_max (Voets et al. 2007). We therefore quantified the effects of menthol on steady-state TRPM8 currents as ΔV_1/2, the menthol-induced change in V_1/2. To determine ΔV_1/2 values, steady-state current–voltage relations obtained before, during and after menthol application were fitted assuming constant z and G_max, and ΔV_1/2 was then calculated as:

Others have described that menthol not only shifts the activation curve, but can also cause a small but significant increase in G_max (Matta & Ahern, 2007). These apparent differences may be related to the inherent problems associated with the determination of G_max in voltage-gated TRP channels, requiring strong depolarizations to voltages (>+250 mV) that are generally difficult to achieve. Alternatively, there may be functional differences between the human TRPM8 (used in this study) and the rat isoform used by Matta & Ahern (2007).

Values of ΔV_1/2 obtained for the different channel arrangements at varying menthol concentrations were compared with the predictions of different gating models (see Supplementary Methods for a mathematical description of the tested models). In all cases, models were fitted to the data for wt–wt–wt–wt and mut–mut–mut–mut channels to obtain values for the affinities of wt and mut binding sites (K_d,wt and K_d,mut) and for the energetic stabilization of the open state caused by ligand binding (ΔΔH) (Voets et al. 2007). Predictions for channels consisting of a mixture of wt and mut subunits were then made using these parameters. To compare the accuracy of the different models to describe the experimental data, we determined the root mean square error (RMSE), i.e. the square root of the mean squared difference between the experimental data points of Fig. 3E and the corresponding model predictions. The Bayesian information criterion (BIC), defined as

where n represents the number of data points and k the number of free parameters, was used to compare the accuracy of between models with different number of free parameters (Schwarz, 1978).

Figure 3. Menthol sensitivity of channels with the indicated subunit stoichiometry.

A, representative whole-cell currents (same voltage protocol as in Fig. 1) in the absence and presence of 100 μm menthol for the channels with the indicated arrangement of wt (blue) and mut (red) subunits. Current scaling is different for each tetrameric construct, and can be derived from the steady-state amplitudes in B. B, steady-state current–voltage relations for the recordings in A. C, average ΔV_1/2 in response to different menthol concentrations for channels with 0–4 wt subunits. Lines represent linear fits. D, root mean square error, comparing the correctness of the different models in predicting the data shown in E. E, dose dependence of the effects of menthol for the indicated channels. Continuous lines represent the predictions of the MWC model. Data are shown as means ± SEM from at least 5 independent patch-clamp recordings.

Results

TRPM8 is a voltage-dependent channel activated by depolarization, and stimuli such as cold and menthol increase TRPM8 activity by shifting the voltage-dependent activation curve toward negative voltages (Voets et al. 2004, 2007; Mahieu et al. 2010). In previous work we have shown that mutating arginine at position 842 in S4 of TRPM8 to alanine (mutant R842A) reduces the channel's gating charge, shifts the temperature sensitivity to lower temperatures and decreases the affinity for menthol (Voets et al. 2007). By mutating the same arginine to histidine (mutant R842H) we obtained a mutant channel that exhibits strongly impaired responses to menthol (in this study, we only used the main natural stereoisomer, (−)-menthol), but apparently unaltered steady-state voltage dependence and cold sensitivity (Fig. 1). As illustrated in Fig. 1A and B, menthol (30 μm) causes a robust current increase in wild-type TRPM8, but is without any significant effect in mutant R842H. The effects of menthol were quantified as ΔV_1/2, the menthol-induced change in the potential for half-maximal activation (V_1/2). For the wild-type channel, we obtained a maximal menthol-induced change in V_1/2, ΔV_1/2,max, of 149 ± 6 mV, and a half-maximal shift at a menthol concentration (EC₅₀) of 39 ± 8 μm (Fig. 1C). For R842H, significant changes in V_1/2 were only observed at concentrations ≥300 μm, and the responses did not saturate at a concentration of 10 mm, which is at the limit of menthol solubility in aqueous solution (Fig. 1C). It should be noted that the R842H mutant exhibited significant desensitization at 10 mm menthol, which may have caused a slight underestimation of the actual ΔV_1/2 at this concentration. Assuming that ΔV_1/2,max is the same for the R842H mutant as in wild-type, we could extrapolate an EC₅₀ of 22 ± 2 mm. The gating charge (z) and V_1/2, both at 25 and 15°C, were not significantly different between wild-type and the R842H (Fig. 1D). Taken together, these results indicate that the R842H mutation selectively affects the menthol-induced gating of TRPM8. Moreover, our finding that menthol can still cause a leftward shift of the voltage-dependent activation curve of R842H, albeit at >500-fold higher concentrations, is fully in line with the notion that this mutation primarily affects the affinity of menthol binding rather than gating steps downstream of menthol binding. Based on these properties, R842H mutant is well suited to further investigate the ligand stoichiometry of TRPM8.

Figure 1. Comparison of the gating properties of wild-type TRPM8 and the R842H mutant.

A, representative whole-cell currents in response to the indicated voltage protocol in the absence and presence of 100 μm menthol for wild-type TRPM8 and the R842H mutant. B, steady-state current–voltage relations for the recordings in A. C, average ΔV_1/2 in response to different menthol concentrations for wild-type TRPM8 and the R842H mutant. D and E, average values for V_1/2 and z for wild-type TRPM8 and the R842H mutant. Data are shown as means ± SEM from at least 5 independent patch-clamp recordings.

We constructed vectors encoding tandem tetramers of all possible combinations of wild-type (wt) TRPM8 and R842H mutant (mut) subunits, coupled by a short, two-amino-acid linker (Supplementary Fig. S1). Fura-2-based Ca²⁺ imaging showed that HEK293 cells expressing the wt–wt–wt–wt construct responded to menthol and cold, whereas cells expressing the mut–mut–mut–mut construct responded to cold but not to menthol (Fig. 2A and B). Whole-cell patch-clamp recordings further revealed that the wt–wt–wt–wt and the mut–mut–mut–mut constructs both express as depolarization-activated channels (Fig. 2C). Detailed quantification of the voltage dependence of activation revealed no significant differences between wt–wt–wt–wt and mut–mut–mut–mut in the apparent gating charge (z) or the potential for half-maximal activation (V_1/2) at 25 and 15°C (Fig. 2D and E). Taken together, these data indicate that the tandem constructs produce functional channels, and that tandem tetramers consisting of all wt or all mut subunits retain the basic functional characteristics of channels assembled from wt or mut monomers.

Figure 2. Functional expression of tandem tetramers.

A, Fura 2-based intracellular Ca²⁺ measurement in non-transfected HEK 293 cells, and in cells expressing the indicated wt–wt–wt–wt and mut–mut–mut–mut tandem constructs. B, average responses to cold and menthol. C, whole-cell currents, in response to the same voltage protocols as in Figure 1, measured at 25°C in HEK 293 cells expressing the indicated tandem constructs. D and E, average values for V_1/2 and z for wt–wt–wt–wt and mut–mut–mut–mut tandem constructs. Data are shown as mean ± SEM from at least 5 independent patch-clamp recordings.

Next, we tested the menthol sensitivity of tandem tetrameric constructs that are expected to give rise to all six possible channel compositions (Fig. 3A). Representative whole-cell currents and derived steady-state current–voltage relations in the absence and presence of 100 μm menthol are shown in Fig. 3A and B. Importantly, we obtained similar menthol dependence for wt–wt–wt–wt or mut–mut–mut–mut tandem constructs as for channels formed from wt or mut monomers, which provides further evidence that the tandem construct has a minimal effect on channel gating (Supplementary Fig. 1). Furthermore, we found a quasi-linear relationship between the number of wt subunits in a tetrameric construct and the ΔV_1/2 in response to different concentrations of menthol (Fig. 3C). Given that zFΔV_1/2 (where F is the Faraday constant) is a direct measure of the menthol-induced change in free energy difference between closed and opened states (Yifrach & MacKinnon, 2002; Voets et al. 2007), these results indicate that increasing the number of wt subunits in a tetrameric channel causes a proportional increase in the menthol-induced stabilization of the open state. This implies that up to four menthol molecules can bind to a single channel and that each bound menthol has a similar energetic contribution to channel gating.

One caveat with this approach is that the expressed channels may not have the expected composition, for example by cleavage of the tandem protein in mono- or dimers, or by exclusion of subunits from the assembled channel. However, Western blot analysis of total membranes obtained from HEK293 cells expressing tetrameric tandem constructs yielded specific bands of the expected tetrameric size, and no indications of proteolytic breakdown of the tandem protein into monomers (Supplementary Fig. S1). Moreover, comparison of tandem tetramers containing one mut and three wt subunits revealed that the menthol sensitivity was independent of the relative position of the mut subunit (ΔV_1/2 in response to 100 μm menthol: 68 ± 3 mV for mut–wt–wt–wt, 65 ± 3 mV for wt–mut–wt–wt and 69 ± 3 mV for wt–wt–wt–mut; n = 3–6). Taken together, these results provide strong evidence that all four subunits are reliably incorporated in the fully assembled tetrameric channel.

The experimentally derived menthol sensitivity of the different tandem channels allowed us to evaluate various gating models for a channel with four ligand binding sites. Since channel activation can occur in the absence of menthol, and menthol shifts the voltage dependence of channel activation to more negative values by slowing channel deactivation, we explored several allosteric models whereby the binding of one or more ligand molecules stabilizes the channel in the open conformation: (1) a Hodgkin–Huxley (HH)-type model, where all four channel subunits can independently transit from a resting to an activated state, and binding of ligand to a subunit stabilizes its active state; all four subunits must be in the active state for the channel to open; (2) a coupled dimer (CD) model, where two adjacent subunits form a functional dimer, and undergo a concerted transition between the resting and activated state; binding of one or two ligands to a dimer causes a proportional stabilization of its activated state, and both dimers must be in the activated state for the channel to open; (3) a Monod–Wyman–Changeux (MWC)-type model, where the four subunits undergo a concerted transition between the closed and open state; each subunit binds ligand independently, and each binding step leads to an equivalent stabilization of the open conformation; and (4) specific ligand number (SLN) models, where the four subunits also undergo a concerted transition between the closed and open state as in the MWC model, but the binding of a specific number of ligands (ranging from 1 to 4) is required to cause maximal stabilization of the open conformation. A full mathematical description of these models is provided in the Supplemental Material.

The different models were fitted to the data for the menthol sensitivity of wt–wt–wt–wt and mut–mut–mut–mut channels, yielding values for the affinities of wt and mut binding sites (K_d,wt and K_d,mut) and for the energetic stabilization of the open state caused by the binding of ligand (ΔΔH). These fits and the obtained parameters are shown in Supplementary Fig. 2 and Supplementary Table 1. Subsequently, these three parameters were used to calculate the predicted menthol sensitivity of the different constructs with one to three mut subunits (Fig. 3E and Supplementary Fig. 2). To obtain an objective measure of the quality of the fit for the different models, we calculated the root mean square error (RMSE) from a comparison of the model predictions with the experimental data (Fig. 3D). This analysis revealed that the MWC model provides the most accurate prediction of the menthol sensitivity for TRPM8 channels of different subunit composition (Fig. 3E). In this model, the menthol binding sites in an open channel have an affinity of 29 μm for wt and 2620 μm for mut subunits, and each bound menthol stabilizes the open state by 3.1 kJ mol⁻¹. The coupled dimer model yielded a 44% higher RMSE value, and predicts differences in menthol sensitivity between channels with two mut subunits in either cis or trans positions, in contrast to our observations. All other models provided a poor fit to the data, reflected in more than 4-fold higher RMSE values than for the MWC model.

Note that in this analysis we assumed that the R842H mutation specifically affects the affinity of the menthol binding site (K_d) without changing the efficacy in stabilizing the open state (ΔΔH). However, since we were unable to achieve a saturating response to menthol for the R842H mutant, we cannot fully exclude that the mutation also affects the menthol efficacy. We therefore repeated the model analysis for the MWC and coupled dimer models while allowing ΔΔH to differ between wt and mut subunits. Despite the fact that this analysis increases the number of free parameters from 3 to 4, it had only a marginal effect on the model parameters and the RMSE, and the rank order of the models did not change (Supplementary Table 1). We therefore conclude that the MWC model provides the best description of our experimental data, with a minimal number of free parameters.

Discussion

We have presented an analysis of the menthol sensitivity of tetrameric TRPM8 channels with varying numbers of subunits with strongly reduced menthol affinity. Our results show a linear relation between the number of wt subunits and the menthol-induced shift in the voltage-dependent activation curve of TRPM8. This indicates that TRPM8 can simultaneously bind up to four menthol molecules, resulting in a stepwise stabilization of the open state.

The minimal model that provided the most accurate description of these findings was a MWC model in which the four channel subunits activate in concert, and each occupied menthol binding site causes a stabilization of the open channel conformation by 3.1 kJ mol⁻¹ (Fig. 4). Based on this model, we can predict that increasing menthol in the concentration range of 0–1000 μm causes TRPM8 to transit gradually from a ligand-free to a fourfold liganded channel (Fig. 4). One possible molecular mechanism could be that menthol intercalates between the transmembrane domains where it can interact with residues in both the S2 (e.g. Y745) and S4 segments (e.g. R842), thereby stabilizing the active conformation of the voltage sensors within the plasma membrane. Clearly, further structural studies are required to validate this and to obtain a full picture of the channel residues that interact with menthol. Based on such structural information, more detailed mutagenesis experiments can be designed to refine the mechanistic details of menthol-dependent gating of TRPM8.

Figure 4. Binding of up to four menthol molecules to the tetrameric TRPM8 channel.

A, schematic visualization of the MWC model, showing the concerted voltage- and temperature-dependent activation of the four sensor domains (in blue or orange) to open the central conducting pore. Binding of up to 4 menthol molecules (red diamonds) causes a gradual stabilization of the open state. B, probability of the occurrence of TRPM8 tetramers with 0–4 bound menthol molecules in function of menthol concentration (at –80 mV).

Ligand-induced activation of other voltage-dependent channels such as the hyperpolarization-activated cyclic nucleotide-gated (HCN) channels and large conductance Ca²⁺-activated (BK) K⁺ channels, also primarily involves a ligand-induced shift in the voltage dependence of channel activation. Our present results suggest, however, that the underlying mechanisms are fundamentally different. For HCN channels, available evidence indicates that the cytosolic cyclic nucleotide-binding domains associate and activate as two dimers, and the coupled dimer model provided an excellent description of the channel's cAMP sensitivity (Ulens & Siegelbaum, 2003). In contrast, we found no differences in menthol sensitivity between channels with two mut subunits in either cis or trans positions, arguing against dimerization of adjacent TRPM8 subunits. In the case of BK channels, a voltage-dependent MWC model similar to the one used in this study described many aspects of the channel's Ca²⁺ and voltage dependence (Cox et al. 1997). However, further mutagenesis studies indicate that each of the four subunits contains two distinct high-affinity and one low-affinity cytosolic Ca²⁺ sensors, and combined loss-of-function mutation of all three sites is required to fully eliminate the channel's Ca²⁺ sensitivity (Lee & Cui, 2010). In contrast, single point mutations in the S2–S4 region of TRPM8 are sufficient to fully eliminate high-affinity menthol binding (Bandell et al. 2006; Voets et al. 2007). The ligand stoichiometry of TRPM8 is also fundamentally different from that of pentameric ligand-gated channels, such as the nicotinic acetylcholine and glycine receptors, where in most cases the binding of two to three ligands is sufficient for maximal channel gating (Colquhoun & Sivilotti, 2004).

In conclusion, our data indicate that a functional TRPM8 channel has four independent and energetically equivalent menthol binding sites. Given that binding sites for lipophilic agonists in other TRP channels are confined to the same transmembrane region (S2–S4) (Jordt & Julius, 2002; Chuang et al. 2004; Bandell et al. 2006; Voets et al. 2007; Vriens et al. 2007), this may represent a general paradigm for ligand-induced gating in this large channel superfamily.

Glossary

Abbreviations

BK channel

large conductance Ca²⁺-activated K⁺ channel

CD

coupled dimer

G_max

maximal conductance

HCN channel

hyperpolarization-activated cyclic nucleotide-gated channels

HH

Hodgkin–Huxley

MWC

Monod–Wyman–Changeux

SLN

specific ligand number

TRP

transient receptor potential

Author contributions

All experiments were performed in the Laboratory of Ion Channel Research. Both authors contributed to all aspects of the work, and approved the final version of the manuscript.

Supplementary material

Supplementary Figures 1 and 2

Supplementary Table 1

Supplementary Material and Methods: full mathematical description of gating models

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer-reviewed and may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors