The Transmembrane Segment S6 Determines Cation versus Anion Selectivity of TRPM2 and TRPM8*

TRPM2 and TRPM8, closely related members of the transient receptor potential (TRP) family, are cation channels activated by quite different mechanisms. Their transmembrane segments S5 and S6 are highly conserved. To identify common structures in S5 and S6 that govern interaction with the pore, we created a chimera in which the S5-pore-S6 region of TRPM8 was inserted into TRPM2, along with a lysine at each transition site. Currents through this chimera were induced by ADP-ribose (ADPR) in cooperation with Ca2+. In contrast to wild-type TRPM2 channels, currents through the chimera were carried by Cl-, as demonstrated in ion substitution experiments using the cation N-methyl-d-glucamine (NMDG) and the anion glutamate. Extracellular NMDG had no effects. The substitution of either intracellular or extracellular Cl- with glutamate shifted the reversal potential, decreased the current amplitude and induced a voltage-dependent block relieved by depolarization. The lysine in S6 was responsible for the anion selectivity; insertion of a lysine into corresponding sites within S6 of either TRPM2 or TRPM8 created anion channels that were activated by ADPR (TRPM2 I1045K) or by cold temperatures (TRPM8 V976K). The positive charge of the lysine was decisive for the glutamate block because the mutant TRPM2 I1045H displayed cation currents that were blocked at acidic but not alkaline intracellular pH values. We conclude that the distal part of S6 is crucial for the discrimination of charge. Because of the high homology of S6 in the whole TRP family, this new role of S6 may apply to further TRP channels.

TRPM2 and TRPM8, closely related members of the transient receptor potential (TRP) family, are cation channels activated by quite different mechanisms. Their transmembrane segments S5 and S6 are highly conserved. To identify common structures in S5 and S6 that govern interaction with the pore, we created a chimera in which the S5-pore-S6 region of TRPM8 was inserted into TRPM2, along with a lysine at each transition site. Currents through this chimera were induced by ADP-ribose (ADPR) in cooperation with Ca 2؉ . In contrast to wild-type TRPM2 channels, currents through the chimera were carried by Cl ؊ , as demonstrated in ion substitution experiments using the cation N-methyl-D-glucamine (NMDG) and the anion glutamate. Extracellular NMDG had no effects. The substitution of either intracellular or extracellular Cl ؊ with glutamate shifted the reversal potential, decreased the current amplitude and induced a voltage-dependent block relieved by depolarization. The lysine in S6 was responsible for the anion selectivity; insertion of a lysine into corresponding sites within S6 of either TRPM2 or TRPM8 created anion channels that were activated by ADPR (TRPM2 I1045K) or by cold temperatures (TRPM8 V976K). The positive charge of the lysine was decisive for the glutamate block because the mutant TRPM2 I1045H displayed cation currents that were blocked at acidic but not alkaline intracellular pH values. We conclude that the distal part of S6 is crucial for the discrimination of charge. Because of the high homology of S6 in the whole TRP family, this new role of S6 may apply to further TRP channels.
The members of the transient receptor potential (TRP) 3 family of non-selective cation channels display an extraordinarily broad spectrum of activation mechanisms reflecting their involvement in manifold biological processes. The basic architecture of TRP channels is shared with the well characterized voltage-gated K ϩ channels. Here, the transmembrane segment S6 contains the activation gate that opens the pore in response to the movement of the activated S4 voltage sensor (1,2). Although TRP channels do not contain a classical voltage sensor, one may presume an interaction of S6 with the pore that governs functional relevance for the properties of TRP channels.
TRPM2 and TRPM8 of the melastatin-related subfamily of TRP channels are the closest relatives within the TRP family (42% identical residues, Ref. 3) but their activation mechanisms are completely different. The principal activators of TRPM2 are intracellular ADP-ribose (ADPR) and reactive oxygen species such as hydrogen peroxide (4,5). Accordingly, TRPM2 is involved in the cellular responses to oxidative and nitrative stress (6 -14).
The TRPM8 channel mediates the cold sensation of the somatosensory system (3,15,16) but was initially discovered as a protein with up-regulated expression in prostate cancer and some other malignant tissues; these roles of TRPM8 are apparently independent of any significant temperature variations (17). TRPM8 is gated by cold temperatures, cooling compounds such as menthol and icilin, and positive membrane potentials in a cooperative manner (16, 18 -20). Recent data have identified several lipid messengers like phosphatidylinositol 4,5-bisphosphate (PIP 2 ) or lysophospholipids (LPLs) as mediators of TRPM8 activation (21,22).
Structure-function relations of channel activation are well established for many ion channel families but in the relatively young TRP field, only limited data are available. In a recent study, the S4-S5 region of TRPM8 has been identified as a voltage-sensitive domain that additionally affects the sensitivity to temperature and menthol (20). For TRPM2, the analysis of a C-terminal splice variant detected in neutrophil granulocytes has revealed the importance of single amino acid residues of the NUDT9 domain for ADPR-dependent channel gating (5,23). However, no thorough analysis has been undertaken to gain information of how the S5 and S6 regions interact and communicate with the intervening pore region. Hitherto, studies on functional subdomains within the S5-pore-S6 region of TRP channels have concentrated on the cation selectivity determined by charged residues in the pore loop (24 -28).
The aim of our study was the identification of distinct domains within S5 and S6, common for TRPM2 and TRPM8, which govern interaction with the pore loop and thereby bear functional relevance for both channels. We report that a particular site in the C-terminal end of S6 is essential for the cation versus anion selectivity. By insertion of a lysine into corresponding sites of TRPM2 and TRPM8, channels were created that passed currents in response to the same stimuli as the respective wild-type channels but these currents were no longer carried by cations but by anions.

EXPERIMENTAL PROCEDURES
Molecular Cloning-For expression in eukaryotic cells, the plasmid constructs pcDNA3-EGFP-TRPM2 and pcDNA3-EGFP-TRPM8 each containing the full-length open reading frame of the corresponding human TRP channel were used. For the generation of the TRPM2/TRPM8 pore chimeras, we introduced recognition sites for AflII (New England Biolabs, Beverly, MA) at corresponding positions of the open reading frame of both channels; one in transmembrane segment S5 and one in S6. Site-directed mutagenesis was performed using the QuikChange mutagenesis system (Stratagene, La Jolla, CA). Defined oligonucleotides were obtained from MWG-Biotech AG (Ebersberg, Germany). The preparation and ligation of the DNA fragments, which have to be exchanged between the two channels was performed as described elsewhere (23). Every point mutation or deletion as well as the correct orientation of the exchanged S5-pore-S6 segments were verified by DNA sequencing with the Big-DYE-Terminator Kit (Perkin Elmer, Roche Applied Sciences, Branchburry, NJ). To exclude the presence of inadvertent mutations in other regions of the channel, two clones with identical results were tested for each chimera or point mutant. All procedures were performed in accordance to the respective manufacturers' instructions, if not indicated otherwise.
Cell Culture and Transfection-Chinese ovary hamster cells (CHO-K1) were obtained from the German Collection of Microorganisms and Cell Cultures (Braunschweig, Germany) and cultured in Ham's F12 medium (Biochrome, Berlin, Germany) supplemented with 4 mM L-glutamine and 10% (v/v) fetal calf serum (Biochrome). Cells were seeded on glass coverslips at a density of Ͻ10 3 cells/mm 2 and grown for 24 h. Subsequently, the pcDNA3-EGFP-TRPM2 or pcDNA3-EGFP-TRPM8 expression constructs were transiently transfected into the CHO cells, using the Trans-Fast transfection reagent (Promega, Mann-FIGURE 1. Characteristic currents of wild-type TRPM2 and TRPM8 channels. A, whole cell recording from cells expressing wild-type TRPM2. Currents were elicited by infusion of 0.6 mM ADPR through the patch pipette. The standard bath solution (140 mM NaCl) was exchanged with a solution with NMDG as main cation (150 mM) during the time periods indicated with horizontal bars. B, current-voltage relation of currents in the experiment shown in A, obtained during voltage ramps from Ϫ90 to ϩ60 mV. The ramp in the absence of NMDG was taken at the time point indicated in A with an asterisk. C, whole cell recording from a cell expressing wild-type TRPM8. The experiment was started in an extracellular solution with room temperature before the temperature was changed to either 37 or 4°C, as indicated. Note that a current was apparent at room temperature from the beginning of the experiment, in contrast to experiments on TRPM2. D, corresponding current-voltage relations of currents in the experiment shown in C obtained at different bath temperatures. In standard experiments, Cs ϩ and glutamate were the main intracellular ions to minimize potential contaminating currents through K ϩ and Cl Ϫ channels. Dashed lines indicate zero current. The holding potential was Ϫ60 mV. The whole cell configuration was reached at time points indicated with w.c. heim, Germany). As controls, cells were transfected with pcDNA3-EGFP vector alone. The transfection procedure was performed as specified by the manufacturer. Patch-clamp measurements were carried out 24 h after transfection in cells visibly positive for EGFP.
Electrophysiology-Transfected cells were analyzed with the patch-clamp technique in the conventional whole cell mode, using an EPC 9 equipped with a personal computer with Pulse and X chart software (HEKA, Lamprecht, Germany). The standard bath solution contained (in mM) 140 NaCl, 1.2 MgCl 2 , 1.2 CaCl 2 , 5 KCl, 10 HEPES, pH 7.4 (NaOH). For nominally cation-free extracellular conditions, the solution contained 150 mM N-methyl-D-glucamine (NMDG), 10 mM EGTA, and 10 mM HEPES, pH 7.4 (HCl). In some experiments, the standard extracellular bath solution was modified by replacing 140 mM NaCl with either 140 mM sodium glutamate, 140 mM NaBr, 140 mM NaI, or 140 mM NaF, and the reference electrode in the bath solution was a 140 mM NaCl/2% agar salt bridge. The pipette solution contained (in mM) 145 CsCl or 145 Cs-glutamate (as indicated), 8 NaCl, 2 MgCl 2 , 10 Cs-EGTA, 10 HEPES, pH 7.2 (CsOH). HEPES was replaced with 10 mM MES in experiments in which the pH was 6.0, and 10 mM N-(1,1-dimethyl-2-hydroxy-ethyl)-3-amino-2-hydroxypropanesulfonic acid (AMPSO) at pH 8.8. When a high (1 M) intracellular Ca 2ϩ concentration was desired, the EGTA concentration in the pipette fluid was reduced to 1 mM, and 0.89 mM CaCl 2 was added. For the stimulation of TRPM2 currents, ADP-ribose was added to the pipette solution from a stock, yielding a final concentration of 0.1-1.0 mM. Alternatively, TRPM2 currents were evoked by superfusion of the cells with standard bath solution containing hydrogen peroxide (10 mM; diluted from a 30% stock solution). TRPM8 activation was induced by superfusion with standard bath solution containing 200 M menthol (Sigma-Aldrich) or 10 -30 M icilin (Cayman Chemical, Ann Arbor, MI). Alternatively, 1.5 l menthol (from a 200 mM stock solution in Me 2 SO) or 1.5 l of Icilin (from a 10 -30 mM stock solution in Me 2 SO) was added directly to the bath chamber (1-2 ml volume). In some experiments, TRPM8 currents were stimulated or suppressed by superfusion with standard bath solution which was cooled immediately before application to ϳ4°C or warmed to 37°C in a conventional water bath. If not otherwise stated, cells were held at a potential of Ϫ60 mV at room temperature. In some experiments, the holding potential was periodically varied (indicated in the figures). The current-voltage relations were obtained during voltage ramps from Ϫ90 to ϩ60 mV and back to Ϫ90 mV applied over 400 ms. Changes in the liquid junction potential were not quantified in experiments on shifts in the reversal potential evoked by extracellular anion substitution; instead, a comparison was performed between cation selective TRPM channels and S6-Lys mutants during identical bath exchanges.
4,4Ј-Diisothiocyanostilbene-2,2Ј-disulfonic acid (DIDS, Sigma-Aldrich), used to test inhibition of anion currents, was added from a stock solution (10 mM in water or 50 mM in Me 2 SO) to the pipette solution or the superfusate (final concentration up to 0.5 mM).
N-(p-Amylcinnamoyl) anthranilic acid (ACA; Merck, Germany), used to inhibit currents of wild-type TRPM2, TRPM8, or the M2M8P chimera, was added to the superfusate (final concentration 20 M, diluted from a 50 mM stock solution in Me 2 SO).
The current density represents the maximal current amplitude (pA) divided by the cell capacitance (pF), a measure of the cell surface. In the case of wild-type TRPM8 and M8-S6-Lys, the values were determined shortly after breaking into the cell and in the case of M2-S6-Lys during plateau current levels in the presence of intracellular ADPR (0.6 mM). Values are given as mean Ϯ S.D. with the number (n) of cells studied.

RESULTS
Typical characteristics of whole cell currents through the wild-type channels TRPM2 and TRPM8 are shown in Fig. 1 with respect to their activation and their cation selectivity, which proved to be consistent with previous reports (3-5, 16, 29). CHO cells transiently transfected with TRPM2 developed currents in response to intracellular ADPR (0.1-1.0 mM) that reached the cytosol by diffusion through the patch pipette (Fig.  1A). In standard experiments, Cs ϩ and glutamate were the main intracellular ions to minimize potential contaminating currents through K ϩ and Cl Ϫ channels (see "Experimental Procedures"). The current-voltage relation of ADPR-induced currents obtained during voltage ramps from Ϫ90 to ϩ60 mV showed a reversal potential close to 0 mV, as expected for a nonspecific cation channel (Fig. 1B). Currents in the inward direction were minimized when extracellular Na ϩ was exchanged by the large impermeable cation NMDG (Fig. 1, A FIGURE 2. Design of TRPM2/TRPM8 channel chimeras. The presumed general structure of TRP channels including the exchanged S5-pore-S6 region (gray shading) is illustrated. The corresponding amino acid sequences of the S5-pore-S6 segments of TRPM2 and TRPM8 are shown in single letter code. Residues conserved between both channels are given in bold. The cut paste limits for chimera construction are marked by asterisks. and B), indicating that TRPM2 is a cation channel. Furthermore, TRPM2 currents were sensitive to Ca 2ϩ as previously demonstrated (30). Removal of extracellular Ca 2ϩ in combination with intracellular Ca 2ϩ buffering with EGTA prevented current development in response to ADPR (data not shown).
In patch-clamp experiments performed at room temperature on CHO cells transfected with TRPM8, a basal inward current was apparent that could be suppressed by heating of the cells to 37°C or could be potentiated by cooling to 4°C (Fig.  1C). The current-voltage relation shows a pronounced outward rectification and a reversal potential close to 0 mV (Fig. 1D). Again, substitution of extracellular Na ϩ by NMDG minimized currents in the inward direction without significantly affecting currents in the outward direction (not shown). These data confirm that TRPM8 as well is a non-selective cation channel. TRPM8 currents were not only induced by cold ( Fig. 1C) but also by menthol (0.2 mM) or icillin (0.02 mM, not shown). The currents through TRPM8 desensitized rapidly and irreversibly, typically within 1-2 min (Fig. 1C).
The S5 and S6 segments of TRPM2 and TRPM8 are quite similar, whereas the central pore sequences display little homology (Fig. 2). For the functional characterization of the extended pore domain, we initially exchanged the almost complete S5-pore-S6 segment of TRPM2 and TRPM8. The sequences at the cut-paste limits of the exchanged S5-pore-S6 segments contained single amino acid substitutions at each end where one hydrophobic amino acid residue was changed to a positively charged lysine (see Fig. 2); this was done to enable the use of the chosen restriction enzyme (see "Experimental Procedures"). The corresponding chimeras M2M8P (TRPM2 containing the S5-pore-S6 segment of TRPM8) and M8M2P (TRPM8 containing the S5-pore-S6 segment of TRPM2) were transiently expressed in CHO cells and characterized with whole cell patch clamp analysis. M8M2P was functionally inactive after stimulation with cold temperatures, menthol, icilin, or ADPR (data not shown). However, cells transfected with the reciprocal chimera M2M8P displayed ADPR-dependent currents (Fig. 3). Infusion of ADPR (0.1-1.0 mM) through the patch pipette into cells kept at a holding potential of Ϫ60 mV led to a time-dependent induction of inward currents similar as observed in cells transfected with wild-type TRPM2 (Fig. 3A). However, in contrast to currents of wild-type TRPM2, the inward currents through M2M8P spontaneously and rapidly declined to (mean Ϯ S.D.) 8.2 Ϯ 3.0% of the initial maximum (n ϭ 8; Fig. 3A). To quantify the kinetics, we measured the time over which 90% of the decline had occurred which amounted to 13.4 Ϯ 2.9 s. These inward currents could be restored by depolarization to positive holding potentials. The amplitude of inward currents measured at Ϫ60 mV immediately after a depolarizing pulse was positively correlated with the strength of the depolarization (Fig. 3A) and also with its duration (Fig. 3B). The current-voltage relation of ADPR-evoked M2M8P currents (Fig. 3C) was obtained during voltage ramps from Ϫ90 to ϩ60 mV, starting from a holding potential of Ϫ60 mV. The reversal potential was about Ϫ50 mV, considerably more negative if compared with wild-type TRPM2 and TRPM8 (see Fig.  1, B and D). The corresponding reversal potential obtained when the ramp was started from a holding potential of ϩ60 mV was about Ϫ30 mV (Fig. 3C). As controls, we tested M2M8Ptransfected cells in the absence of ADPR (Fig. 3, C, inset and D) as well as cells transfected with vector construct only (data not shown). Both controls were negative.
As previously demonstrated (30,31), intracellular Ca 2ϩ represents a cofactor of ADPR-dependent activation of TRPM2. Therefore, we compared the ADPR-evoked current development of TRPM2 and M2M8P in the absence (10 mM EGTA) and presence of 1 M Ca 2ϩ in the pipette solution. Similarly to wildtype TRPM2, currents of M2M8P were more strongly stimu-lated when ADPR was infused in non-saturating concentrations (0.2 mM) along with 1 M Ca 2ϩ (Fig. 4A) than when ADPR was infused at a low Ca 2ϩ concentration (Ͻ10 nM). Moreover, similar M2M8P currents as evoked intracellular ADPR were induced by extracellular H 2 O 2 (10 mM; Fig. 4B). Again, these currents declined at a holding potential of Ϫ60 mV and were readily restored by short lasting depolarizations to ϩ60 mV.
Recently, it has been demonstrated that N-(p-amylcinnamoyl) anthranilic acid (ACA) is a highly potent and efficient channel blocker of TRPM2, TRPM8, and TRPC6 (32). Because ACA blocked these channels only from the extracellular side, it has been proposed that it reduces the channel open probability. In our hands, the extracellular application of ACA (20 M) inhibited ADPR-induced whole cell currents of M2M8P (Fig.  4C) to the same extent as observed for wild-type TRPM2 (Fig.  4D), suggesting that the putative interaction of ACA and the external pore domain is not disturbed in the M2M8P channel chimera. The experiments with ACA were performed with a pipette solution containing Cl Ϫ as main anion to avoid block by intracellular glutamate (see next two paragraphs). Thus, with respect to channel activation by ADPR or hydrogen peroxide, co-stimulation by Ca 2ϩ and inhibition by ACA, the gating behavior of M2M8P was not noticeably changed in comparison to wild-type TRPM2. The decline of M2M8P current at negative holding potentials and the restitution of inward currents by depolarizing pulses were consistently found when either glutamate or aspartate (not shown) was present in the pipette solution, at concentrations of 90 mM or more, but not at 40 or 70 mM. To analyze the phenomenon which resembles a voltage-dependent block with a slow time course, we first tested conditions with NMDG as main extracellular cation. These experiments revealed a further and more striking difference of M2M8P in comparison to both, TRPM2 and TRPM8. The characteristic current development of M2M8P (see Figs. 3A and 4B) was not changed by NMDG (Fig. 5A); in fact, NMDG did not elicit any obvious effects on inward or outward currents (compare Figs. 3C, 5B). We examined whether these unexpected results could be explained by permeability to NMDG despite the large size of the NMDG cation, because some channels of the TRP family display partial permeability to NMDG (33). However, there was no notable shift of the reversal potential of currents through M2M8P when all extracellular cations were replaced by NMDG (compare, Figs. 3C and 5B). Therefore, we reasoned that the currents through M2M8P might be anion currents and went on to experiments where the intracellular glutamate was substituted by Cl Ϫ (Fig. 5, C and D).
In experiments with Cl Ϫ as the only intracellular anion (Fig. 5C), the time course by which M2M8P currents devel-oped in response to ADPR was profoundly changed. In contrast to previous experiments in which glutamate was present in sizeable intracellular concentrations, no decline of the current occurred and depolarizing pulses did not affect inward currents (Fig. 5C). However, the usual inward currents were larger in amplitude, to an extent that the cells did not tolerate them for more than a few minutes during which they visibly deteriorated (not shown). The corresponding current-voltage relation (Fig. 5D) obtained during voltage ramps from Ϫ90 to ϩ60 mV showed a reversal potential close to Ϫ10 mV, independently of the preceding holding potential (either Ϫ60 or ϩ60 mV). Furthermore, M2M8P currents were again insensitive to NMDG as main extracellular cation (not shown).
Thus, negative currents at hyperpolarizing potentials through M2M8P are not cation currents but are carried by Cl Ϫ that moves out of the cell. The current decline at hyperpolarizing voltages can be attributed to block by glutamate (mimicked by aspartate). The block is released by depolarizing voltages in a time-and voltage-dependent manner.
The switch from a cation to anion conductance when the pore of a cation channel was substituted by the pore of another cation channel was an unexpected result and prompted us to a more refined molecular analysis. Because negatively charged amino acid residues of the outer pore region are critical for the cation selectivity of TRP channels (reviewed in Ref. 34), it is conceivable that positively charged amino acid residues at the pore internal face may influence the anion permeability of M2M8P. In comparison to wild-type TRPM2 and TRPM8, there are two positively charges lysines introduced into M2M8P localized at the cytosolic ends of the transmembrane segments S5 and S6 (Fig. 6: L941KϩI1045K for TRPM2 and F874KϩV976K for TRPM8, respectively). To test whether one or both of these two lysines govern the anion selectivity of M2M8P, we generated the single point mutants TRPM2 (L941K), TRPM2 (I1045K), TRPM8 (F874K), and TRPM8 (V976K) which we refer to as M2-S5-Lys, M2-S6-Lys, M8-S5-Lys, and M8-S6-Lys.
The whole cell patch-clamp analysis of the M2-S6-Lys and M8-S6-Lys revealed that both channel mutants displayed currents with the same time-dependent decline at negative holding potentials as M2M8P, along with relief from glutamate block by depolarization (Fig. 7, A and D). The currents through M2-S6-Lys were ADPR-dependent (Fig. 7A), the currents through M8-S6-Lys developed spontaneously at room temperature (   . Note that no desensitization is detectable. The holding potential in A and D was periodically changed between Ϫ60 mV and ϩ60 mV to enable or disable block by glutamate. extracellular NMDG (shown for M2-S6-Lys in Fig. 7, A and C). The characterization as anion currents was again performed by ion exchange experiments. In particular, the substitution of extracellular Cl Ϫ with glutamate as main anion predominantly decreased the outward currents and shifted the reversal potential to the right (shown for M2-S6-Lys in Fig. 7C; see also Table  1). Moreover, the substitution of intracellular glutamate by Cl Ϫ prevented the voltage-dependent current decline induced by glutamate (current-voltage relation shown for M2-S6-Lys in Fig. 7C; compare with M2M8P in Fig. 5D). When glutamate was the main anion on both sides of the membrane, inward and outward currents were simultaneously minimized. Moreover, extracellular glutamate prevented the restoration of inward currents during depolarizing pulses (Fig. 7A). As control, the currents of wild-type TRPM2 were not visibly affected by substitution of extracellular Cl Ϫ with glutamate (not shown). Thus, the channels M2M8P, M2-S6-Lys, and M8-S6-Lys all behave uniformly as anion channels.
Interestingly, the current density obtained at room temperature in cells expressing M8-S6-Lys was considerably larger (103.2 Ϯ 72 pA/pF; n ϭ 13) than in wild-type TRPM8 (16.2 Ϯ 7.0 pA/pF; n ϭ 14; p Ͻ 0.01 in Wilcoxon rank sum test). Additionally, currents through M8-S6-Lys did not show any desensitization (compare Figs. 1C and 7D). These findings are in agreement with the idea that desensitization of wild-type TRPM8 requires Ca 2ϩ influx through the channel (21) that does not occur in the anion-selective M8-S6-Lys. Furthermore, the pronounced outward rectification typical for wild-type TRPM8 currents was not observed in cells transfected with M8-S6-Lys (data not shown).
In distinct contrast to insertion of one lysine into transmembrane segment S6, insertion into S5 did not produce uniform effects on TRPM2 and TRPM8. M2-S5-Lys was a non-functional channel irresponsive to ADPR. M8-S5-Lys displayed cation currents similar to those of wild-type TRPM8 (data not shown). Therefore, it is the S6 region where insertion of one lysine is sufficient to convert the channel selectivity from cation to anion selective.
Relative permeabilities of these anion-selective currents for various anions were determined in experiments with M2-S6-Lys where the reversal potential of plateau-like ADPR-induced currents were compared in the presence of either Cl Ϫ , Br Ϫ , I Ϫ , F Ϫ , and glutamate ( Table 1). The value for glutamate is probably an overestimation; no true currents carried by glutamate were demonstrated, whereas the reversal potential in glutamate is largely influenced by leak and by remaining Cl Ϫ The anion currents were not inhibited by DIDS (up to 0.5 mM either applicated intracellular or extracellular).
Next we tested whether the conversion of TRPM2 and TRPM8 to anion channels by the S6-Lys mutation is caused by the positive charge of lysine. Initially, we inserted an aspartate instead of the lysine into TRPM2 as well as into TRPM8 at the corresponding position within S6 (see Fig. 6). M2-S6-Asp showed typical ADPR-dependent cation currents with characteristics similar to those of wild-type TRPM2 (data not shown). In contrast, in cells transfected with M8-S6-Asp and stimulated either with cold temperature, menthol, or icilin, no currents were detected. Results on these and other mutations are summarized in Table 2.
As a definite test for the importance of the positive charge of the lysine in S6, we introduced a histidine instead (M2-S6-His). The charge of histidine can be manipulated by variations in the intracellular pH, dependent on the local protein environment (35). The effects of the intracellular pH on the current characteristics of M2-S6-His are shown in Fig. 8. During stimulation with ADPR, currents at pH 8.8 were cation-driven, minimized in the inward direction by NMDG and indistinguishable from currents through wild-type TRPM2 (Fig. 8, A and B). This was profoundly changed at an intracellular (pipette) pH of 6.0. Although the currents were still carried by cations, as indicated by the sensitivity to extracellular NMDG, they were blocked by intracellular glutamate. In the standard conditions when glutamate was the main intracellular anion, the time course of the currents displayed the characteristics of M2-S6-Lys with a strong decline after an initial peak (Fig. 8C). The substitution of intracellular glutamate with Cl Ϫ as main anion prevented the current decline (not shown). Interestingly, the block of M2-S6-

D.) of currents through the M2-S6-Lys mutant in the presence of different intracellular and extracellular anions
All recordings were performed with a 140 mM NaCl/2% agar salt bridge as reference electrode in the bath solution. The holding potential preceding the ramp protocol is indicated. Relative permeabilities were calculated according to Hille (45).

TABLE 2 Summary of experiments on the ion nature of currents through various mutants of TRPM2 and TRPM8
When currents were observed in response to either ADPR, cold temperature, or menthol, the charge selectivity was determined by extracellular NMDG or extracellular and intracellular glutamate (see examples in Figs. 5-8).

JOURNAL OF BIOLOGICAL CHEMISTRY 27605
His by glutamate was not relieved by depolarization to positive holding potentials up to 80 mV (Fig. 8C). Moreover, the effects of NMDG extended to outward currents (Fig. 8D) and prevailed even after intensive washout. Thus, they are best described as an irreversible block. As controls, experiments on M2-S6-His at pH 7.2 revealed the same results as with pH 8.8. Further control experiments on wild-type TRPM2 and M2-S6-Lys at pH 8.8 and 6.0 did not show any differences in comparison to experiments at pH 7.2.

DISCUSSION
The main finding that we have demonstrated is that the substitution of a hydrophobic amino acid residue by a positively charged lysine at the C-terminal end of transmembrane segment S6 converts the ion selectivity of both TRPM2 and TRPM8 from cation selective to anion selective. The activation mechanisms that are quite different between the wild-type channels of TRPM2 and TRPM8 are not affected by the modification, rendering anion channels that are either stimulated by ADPR or are sensitive to temperature. Thus, our study establishes a crucial role of the distal part of the S6 segment for the ion selectivity of TRPM2 and TRPM8; the previous view that the permeation properties of TRP channels are mostly defined by the central pore region located between S5 and S6 segments must be broadened and should include the adjacent S6 segment.
The switch from cation selectivity to anion selectivity and vice versa was intensively studied in a superfamily of ligandgated ion channels, which is remotely related to the TRP family and includes the cation-selective nicotinic acetylcholine receptor as well as the anion-selective ␥-aminobutyric acid, and glycine receptors (for reviews see Refs. 36, 37. Charge selectivity was reversed by mutations to key charged residues located close to the intracellular mouth of the pore (38,39). More closely than this family, cyclic nucleotide-gated (CNG) channels are related to the TRP family. CNG and TRP channels both contain a pore forming P-loop between S5 and S6, which includes a selectivity filter and is typical for various cation channels (40). In TRP channels, mutations in this region may alter the ratio of monovalent over divalent cation permeability (25,26,28). In CNG channels, even the cation versus anion selectivity can be reverted by mutations of this region. When a highly conserved glutamate residue residing at the external end of the selectivity filter was changed to lysine or arginine, anion selective variants The standard holding potential of Ϫ60 mV was intermittently switched to ϩ60 or ϩ80 mV. D, intracellular pH was 6.0 with no intracellular glutamate. The current-voltage relationship during an ADPR-induced current plateau is compared with that obtained during application of NMDG. The effects of NMDG were irreversible even after wash-out. Note that NMDG inhibited both, inward and outward currents. of a CNG channel were created (41), although the conductance was considerably lower than that of the wild-type channel. In our present study, the central pore region did not contribute to the anion selectivity of the S6-lysine variants of TRPM2 and TRPM8. Thus, we have identified a new locus essential for the discrimination of cation versus anion permeability in TRPM channels; this locus is clearly different from that in CNG channels while it resembles that of another ion channel family in terms of general localization with respect to the cytosolic side of the pore.
Our experiments demonstrated the selectivity for anions over cations for the TRPM2 and TRPM8 channels mutated in S6 by two key findings in ion exchange experiments. First, currents through these channels were not affected by substitution of the normal extracellular cations with the large cation NMDG. NMDG is mostly impermeant through wild-type TRPM2 and TRPM8 but does not block the pore, in contrast to the situation in some other TRP channels e.g. TRPC3 where a partial block of outward currents is evoked by extracellular NMDG (42). In the S6-Lys mutations of TRPM2 and TRPM8, NMDG did not reduce the current amplitudes and did not shift the reversal potential to the left. Second, substitution of extracellular or intracellular Cl Ϫ with glutamate or aspartate decreased the currents dramatically and induced large shifts in the reversal potentials. Furthermore, intracellular glutamate and aspartate induced a voltage-dependant channel block that was relieved by depolarizing voltages. The observation that such dramatic changes in the biophysical properties of TRPM channels are caused by a single point mutation located outside of the classical pore region would best fit to the idea that a large structural rearrangement takes place, rather than a local distortion of a hydrophobic region in consequence of the addition of a charged amino acid residue. The same idea has been proposed for the 5-hydroxytryptamine 3 (5-HT 3 ) receptor (38).
The S6 as well as the S5 segments of TRPM2 and TRPM8 show 74% identical positions and charged residues are almost completely absent (see Fig. 2). Moreover, these hydrophobic stretches of S6 and of S5 are highly conserved in various TRP channels including members of the TRPC, TRPV, and TRPM subfamilies (43). Therefore, any charged amino acid within this hydrophobic environment is expected to elicit a considerable and functionally relevant distortion of the local structure. However, the manipulations of S6 but not of S5 were decisive to create anion-selective channels. After demonstrating that anion selectivity was achieved by a single lysine, we examined whether the positive charge of this amino acid was the instrumental factor and tested the two mutants M2-S6-Asp and M2-S6-His. M2-S6-Asp showed the phenotype of wild-type TRPM2, ruling out the possibility that a negative charge may have the same effect as a positive one. The M2-S6-His mutant allowed us to study the importance of the positive charge because histidine becomes neutral at alkaline pH values but cationic at acidic pH values. Importantly, M2-S6-His behaved like wild-type TRPM2 at an intracellular pH of 8.8 but considerably changed its properties at an intracellular pH of 6.0. In particular, intracellular glutamate induced a current block. This block could not be relieved by depolarization, in contrast to that in M2-S6-Lys, even though currents through M2-S6-His were carried by cations and sensitive to NMDG. Thus it can be concluded that the positive charge of M2-S6-Lys and M2-S6-His is responsible for an interaction with glutamate. Subtle structural details in addition to charge decide whether anions or cations are allowed to permeate. A comparison of the pore properties of wild-type TRPM2, M2-S6-Lys, and M2-S6-His is illustrated in Fig. 9.
In the literature, there are numerous reports on point mutations that completely abolish channel activity. In the case of TRPM2, examples include the mutation N1326D in the Nudix box or C996S in the pore (23,44). In contrast, the S6-Lys-modification represents a change of the channel structure that profoundly affects certain pore properties, whereas other channel FIGURE 9. Summary of the pore properties of wild-type TRPM2, the mutant M2-S6-Lys, and the mutant M2-S6-His at an intracellular of pH 6.0, with respect to charge selectivity, current inhibition by extracellular NMDG, and block by intracellular glutamate. The S5-pore-S6 segments are depicted with the cation selectivity filter symbolized by a rhombus containing a negative charge. A, in wild-type TRPM2, NMDG inhibits currents but glutamate does not exert a block. B, in M2-S6-Lys, the lysine residue in S6 changes the pore selectivity in favor of anions and additionally renders the channel sensitive to a reversible voltage-dependent block by glutamate. C, M2-S6-His is cationselective as wild-type TRPM2 and sensitive to NMDG. The positive charge of the histidine at acidic intracellular pH values enables glutamate block. The effects of both NMDG and glutamate are irreversible. SEPTEMBER 21, 2007 • VOLUME 282 • NUMBER 38 functions are preserved. Hence, the mutated TRPM8 channels in our study allow to probe previous ideas on some aspects of channel regulation. Wild-type TRPM8 currents display a strong desensitization in their response to activating stimuli like temperature, menthol and icilin. It has been proposed that this is due to Ca 2ϩ influx because activation of a Ca 2ϩ -dependent phospholipase C results in depletion of phosphatidylinositol 4,5-bisphosphate (PIP 2 ), which is essentially required for channel activity (21). Our experiments provide an elegant support of this proposition because desensitization is no longer observed in the anion-selective M8-S6-Lys mutant that does not allow Ca 2ϩ entry. Moreover, ACA inhibited TRPM2 and TRPM8 from the outside; it has been suggested to act on the pore. Strikingly, ACA effects on the anion currents of the S6-lys mutants of TRPM2 and TRPM8 were identical with those on the cation currents of the wild-type channels, in agreement with the proposed mode of action. Finally, it has been suggested that the gating of TRPM8 is associated with a net loss of hydrophobic interactions within a putative temperature sensor domain (18). The same group has recently presented a molecular model of the TRPV1 channel that involves a direct interaction of the S5 and S6 domains with the channel activator PIP 2 (46). Such an idea might be used to explain why the M8-S6-Lys mutant displays larger currents at room temperature than wildtype TRPM8. In any case, the role of the hydrophobic stretch in S6 may not be confined to charge discrimination but may be also involved in the gating process. However, definite conclusions on this second role of S6 are limited at present and any generalizations to other TRP channels are hypothetical. In contrast, the importance for ion selectivity is uniform for TRPM2 and TRPM8. Because of the high homology of S6 between TRPM2, TRPM8 and also many other channels of the TRP superfamily, straightforward experimental approaches are available to test how far the function of S6 for charge selectivity represents a general feature of TRP channels.

S6 Controls Charge Selectivity of TRPM2 and TRPM8
In conclusion, we have identified a locus within the S6 region of TRPM2 and TRPM8 that determines whether these channels are cation channels or anion channels. Our results establish a new role of the S6 segment; this role is common between TRPM2 and TRPM8 and may extend to other members of the TRPM family or even the TRP superfamily because of the high degree of homology in this region.