Molecular Determinants of Mg2+ and Ca2+ Permeability and pH Sensitivity in TRPM6 and TRPM7*

The channel kinases TRPM6 and TRPM7 have recently been discovered to play important roles in Mg2+ and Ca2+ homeostasis, which is critical to both human health and cell viability. However, the molecular basis underlying these channels' unique Mg2+ and Ca2+ permeability and pH sensitivity remains unknown. Here we have created a series of amino acid substitutions in the putative pore of TRPM7 to evaluate the origin of the permeability of the channel and its regulation by pH. Two mutants of TRPM7, E1047Q and E1052Q, produced dramatic changes in channel properties. The I-V relations of E1052Q and E1047Q were significantly different from WT TRPM7, with the inward currents of 8- and 12-fold larger than TRPM7, respectively. The binding affinity of Ca2+ and Mg2+ was decreased by 50- to 140-fold in E1052Q and E1047Q, respectively. Ca2+ and Mg2+ currents in E1052Q were 70% smaller than those of TRPM7. Strikingly, E1047Q largely abolished Ca2+ and Mg2+ permeation, rendering TRPM7 a monovalent selective channel. In addition, the ability of protons to potentiate inward currents was lost in E1047Q, indicating that E1047 is critical to Ca2+ and Mg2+ permeability of TRPM7, and its pH sensitivity. Mutation of the corresponding residues in the pore of TRPM6, E1024Q and E1029Q, produced nearly identical changes to the channel properties of TRPM6. Our results indicate that these two glutamates are key determinants of both channels' divalent selectivity and pH sensitivity. These findings reveal the molecular mechanisms underpinning physiological/pathological functions of TRPM6 and TRPM7, and will extend our understanding of the pore structures of TRPM channels.

TRPM6 and TRPM7 belong to the TRP channel superfamily (1)(2)(3)(4)(5) and are distinguished from other known ion channels by virtue of having both ion channel and protein kinase activities (6 -11). In addition, TRPM6 and TRPM7 uniquely exhibit strong outward rectification, permeation to Ca 2ϩ , Mg 2ϩ , monovalent cations, and a wide array of trace metals (6 -8, 11, 12). The channel activity of TRPM7 is regulated by intracellular Mg 2ϩ (7) and other divalent cations (13)(14)(15), Mg 2ϩ -ATP (7,12,16), phosphatidylinositol 4,5-bisphosphate (14,17), cAMP (18), and internal and external pH conditions (14,19). Similarly, TRPM6 channel activities have been shown to be inhibited by intracellular Mg 2ϩ and potentiated by external protons (8,11). Recent studies have demonstrated that TRPM6 and TRPM7 are key regulators of Mg 2ϩ homeostasis: mutations of TRPM6 cause familial hypomagnesemia and secondary hypocalcemia (20,21); whereas targeted gene deletion of TRPM7 in the DT40 B cell line produced intracellular Mg 2ϩ deficiency and growth arrest (7,22). Consistent with its role in Mg 2ϩ and Ca 2ϩ homeostasis, TRPM6 is abundantly expressed in the intestine and the kidney (8,20,21,23), whereas TRPM7 is ubiquitously expressed, with highest expression in the kidney and heart (5,6). In addition to these channels' regulation of Mg 2ϩ homeostasis, several studies have suggested multiple cellular and physiology functions for TRPM7, including anoxic neuronal death (24), cell adhesion and actomyosin contractility (25,26), and skeletogenesis (27). Although the mechanisms by which TRPM6 and TRPM7 exert their physiological and/or pathological functions are not yet completely understood, it is clear that permeation of Ca 2ϩ and Mg 2ϩ contributes substantially to the known functions of these channels (7, 20 -22, 24, 25, 27). Moreover, a recent study demonstrated that the sensitivity of TRPM7 to external pH may contribute to controlling neurotransmitter release (28). Therefore, it is essential to understand the molecular mechanisms underlying the Ca 2ϩ and Mg 2ϩ permeability of TRPM6 and TRPM7, as well as their sensitivities to changes in pH.
The aim of the present study was to identify the amino acid residues that determine Mg 2ϩ and Ca 2ϩ permeation of TRPM6 and TRPM7. We previously demonstrated that external protons significantly enhance TRPM6 and TRPM7 inward currents (11,19) by decreasing the divalent affinity to the channels. Our results suggested that protons compete with divalents for binding site(s) within the channels' pore. In the present study, we systematically mutated negatively charged amino acid residues within the putative pore-forming region of TRPM7; and identified Glu 1047 and Glu 1052 of TRPM7 as the key residues that confer divalent selectivity and the sensitivity of the channel to pH. Moreover, we demonstrated that mutations of the equivalent positions (Glu 1024 and Glu 1029 ) in TRPM6 produced identical changes, indicating that these two glutamate residues con-stitute the molecular basis of these channels' Mg 2ϩ and Ca 2ϩ permeability as well as their pH sensitivity. The above findings are critical to understanding the physiological/pathological functions of TRPM6 and TRPM7, and provide molecular insight of the pore architecture of these channels.

EXPERIMENTAL PROCEDURES
Molecular Biology-TRPM6 construct was kindly provided by Dr. Joost G. J. Hoenderop. TRPM7 was previously cloned from mouse (6). Amino acid substitutions to the pores of TRPM6 and TRPM7 were made using the QuikChange Sitedirected Mutagenesis Kit (Stratagene) following the manufacturer's instructions. The primers are shown in supplemental materials Table S1.
Functional Expression of TRPM6, TRPM7, and the Mutants-CHOK1 cells were grown in Dulbecco's modified Eagle's medium/Ham's F-12 medium supplemented with 10% fetal bovine serum, 100 units/ml penicillin, and 100 mg/ml streptomycin at 37°C in a humidity-controlled incubator with 5% CO 2 . Cells were transiently transfected with wild-type (WT) 5 TRPM6, TRPM7, and the mutants of TRPM6 and TRPM7 as previously described (6). TRPM7 and its mutants were co-transfected with a green fluorescent protein-containing pTracerCMV2 vector. Electrophysiological recordings were conducted between 36 and 48 h after transfection. Successfully transfected cells were identified by their green fluorescence when illuminated at 480 nm. All patch-clamp experiments were performed at room temperature (20 -25°C).
Electrophysiology-Whole cell currents were recorded using an Axopatch 200B amplifier. Data were digitized at 10 or 20 kHz, and digitally filtered off-line at 1 kHz. Patch electrodes were pulled from borosilicate glass and fire-polished to a resistance of ϳ3 M⍀ when filled with internal solutions. Series resistance (R s ) was compensated up to 90% to reduce series resistance errors to Ͻ5 mV. Cells in which R s was Ͼ10 M⍀ were discarded (29).
For whole cell current recordings, voltage stimuli lasting 250 ms were delivered at 1-5-s intervals, with either voltage ramps or voltage steps ranging from Ϫ120 to ϩ100 mV. Unless otherwise stated, following break-in, 3-5 min were allowed to pass to let currents develop to reach a steady-state. A fast perfusion system was used to exchange extracellular solutions, with complete solution exchange achieved in ϳ1-3 s (19).
The standard extracellular Tyrode's solution contained (mM): 140 NaCl, 5 KCl, 2 CaCl 2 , 20 HEPES, and 10 glucose, pH adjusted to 7.4 (NaOH). External solutions at various acidic pH were prepared as previously reported (19, 30 -32). In brief, HEPES (20 mM) was used in the solutions at pH 7.0 and 7.4, and was replaced by 10 mM HEPES and 10 mM MES for the solutions at pH Յ 6. Divalent-free solution (DVF) contained (mM) 145 NaCl, 20 HEPES, 5 EGTA, 2 EDTA, and 10 glucose, with estimated free [Ca 2ϩ ] Ͻ1 nM and free [Mg 2ϩ ] Ϸ10 nM at pH 7.4 (calculated with the MaxChelator software). Appropriate Ca 2ϩ or Mg 2ϩ was added to the DVF at pH 7.4 to prepare solutions containing Յ10 M Mg 2ϩ or Ca 2ϩ (Fig. 5). Solutions containing 0.1, 0.2, 0.5, 1, 2, and 10 mM Mg 2ϩ or Ca 2ϩ were prepared by omitting EDTA and EGTA in the DVF solution, and by adding the appropriate concentrations of Mg 2ϩ or Ca 2ϩ , with reduc-FIGURE 1. Alignment of TRPM7 pore region with other TRPM channels. A, schematic structure of the TRPM7 and positions of substituted amino acid residues in the TRPM7 channel. B, alignment of pore region of TRPM channels. Amino acids in boxes are the ones that were selected for analysis. The GenBank TM accession numbers of the mouse TRPM7, human TRPM1-6, and TRPM8 are AF376052, AAC8000, AAI12343; NP_060106, NP_055370, Q9BX84; and NP_076985, respectively. tions in Na ϩ concentration when necessary to maintain constant osmolarity. Isotonic Ca 2ϩ or Mg 2ϩ solution contained 120 mM Ca 2ϩ or Mg 2ϩ , 10 mM HEPES, 10 mM glucose, with pH adjusted to pH 7.4. Different cation solutions (Fig. 7) at 30 mM contained (in mM): 30 divalents or monovalents, 20 HEPES, 100 NMDG-Cl, 10 glucose (pH 7.4). Zn 2ϩ solution was prepared at 10 mM due to the low solubility (11). All chemical reagents were purchased from Sigma.
Ion permeation ratios are typically calculated from reversal potentials (E rev ) using Nernst and Goldman-Hodgkin-Katz equations based on the assumption that ions permeate independently and that the electric field in the membrane is constant (33). For TRPM7 channels, however, because the monovalent and divalent cations do not permeate the channels independently (7), it is not adequate to use the Goldman-Hodgkin-Katz equation to estimate the relative permeability (7,8,12). Thus, the relative permeability was estimated from the inward current amplitude as previously reported (8,12).
Data Analysis-Pooled data are presented as mean Ϯ S.E. Dose-response curves were fitted by an equation of the form: where E is the effect at concentration C, E max is the maximal effect, EC 50 is the concentration for half-maximal effect, and n is the Hill coefficient (34). EC 50 is replaced by IC 50 if the effect is an inhibitory effect. Voltage-dependent effects of Ca 2ϩ and Mg 2ϩ on TRPM7 and the mutants were analyzed by fitting the I/I 0 ratio curves to the Boltzmann functions: I/I 0 ϭ 1/(1 ϩ exp[V 0.5 Ϫ V]/k) is for the voltagedependent block, and I/I 0 ϭ 1/(1 ϩ exp[V Ϫ V 0.5 ]/k) is for the voltage-dependent relief of block; where I 0 is the current before and I is the current after application of Mg 2ϩ or Ca 2ϩ , V is the membrane potential, V 0.5 is the membrane potential at which the current is blocked by 50%, and k is a slope factor representing the voltage dependence of block (35). The slope factor k is k ϭ RT/z␦F, where z is the valence of blocker and ␦ is the fraction of the membrane electrical field. Statistical comparisons were made using two-way analysis of variance and two-tailed t test with Bonferroni correction; p Ͻ 0.05 indicated statistical significance.

RESULTS
E1047Q and E1052Q Substitutions within TRPM7 Pore Alters Its I-V Relationship-When heterologously expressed, TRPM7 constitutes a channel that is characterized by extremely small inward and large outward currents. External divalent cations such as Mg 2ϩ and Ca 2ϩ are permeable to TRPM7 and at the same time block monovalent cations per-FIGURE 2. Current-voltage relationships of WT TRPM7 and its mutants. A-G, representative currents elicited by ramp protocols ranging from Ϫ120 to ϩ100 mV in WT TRPM7 and its mutants. Note the significant changes in the I-V relation of E1047Q (E) and E1052Q (F). H, normalized I-V curves of WT TRPM7 and its mutants. The I-V curves of D1035N, H1039E, H1039M, and D1054A show outward rectification, and are superimposed with that of WT TRPM7; whereas E1047Q exhibits double-rectification with significantly increased inward currents and markedly decreased outward currents. Inward current of E1052Q is substantially larger than that of WT TRPM7. meating through the pore of the channel. We have previously shown that external protons substantially potentiate TRPM7 inward currents, which may occur through competition of protons with divalent cations for binding sites in the pore of the channel (19). To identify potential binding sites for Mg 2ϩ , Ca 2ϩ , as well as for protons, we systematically exchanged negatively charged residues within the TRPM7 putative pore region with uncharged residues found at equivalent positions in other TRPM family members (Fig. 1). As the increase in inward current induced by pH for TRPM6 is smaller than that observed for TRPM7 (11,19), we also investigated the contribution of His 1039 to the pH sensitivity of TRPM7 because His 1039 is replaced by Glu in TRPM6. Therefore, we additionally mutated His 1039 to H1039E and H1039M, as TRPM1 and TRPM3 have a Met residue at the equivalent positions, respectively. Within the S5-S6 linker of TRPM7, eight residues were singly or doubly mutated (Fig. 1). The resulting TRPM7 mutants were transiently transfected into CHO-K1 cells and their currents examined for sensitivity to pH and permeability to Mg 2ϩ and Ca 2ϩ . Because the endogenous TRPM7-like MIC/MagNuM current is extremely small in CHO-K1 cells (6), the elicited currents obtained upon transfection of the TRPM7 mutants predominantly reflect conductances originating from the expressed TRPM7 pore mutants. Fig. 2 shows currents recorded from various TRPM7 mutants. The current-voltage (I-V) relationships of the mutants D1035N, D1054A, H1039E, and H1039E were similar to that of WT TRPM7. It was surprising that D1054A did not produce a significant change, as this aspartic acid residue is conserved among all TRPM channels. By contrast, the I-V relationships of E1047Q and E1052Q were significantly different from that of WT TRPM7. The inward current of E1052Q was substantially larger than that of WT TRPM7, whereas its outward current was similar to WT TRPM7. E1047Q demonstrated a double rectification I-V profile, with increased inward current and decreased outward current compared with WT TRPM7. The normalized I-V curves of D1035N, D1054A, H1039E, and H1039M were superimposable with that of WT TRPM7, whereas the I-V curves of E1047Q and E1052Q were markedly different from that of WT TRPM7 (Fig. 2H). The average inward and outward current amplitudes obtained for the TRPM7 pore mutants are summarized in Fig. 3A. The ratios of inward currents measured at Ϫ120 mV to the outward currents measured at ϩ100 mV of E1047Q and E1052Q were 12-and 8-fold larger than that of WT TRPM7 (Fig. 3B), respectively; indicating that blockade of monovalent inward current by divalent cations was reduced in E1047Q and E1052Q compared with WT TRPM7. The significant changes in current-voltage profiles in E1047Q and E1052Q indicate that E1047 and E1052 are residues critical to TRPM7 channel function.
In the mutants in which the negatively charged Glu was replaced by positively charged Lys at Glu 1047 and Glu 1052 positions (E1047K and E1052K), a majority of cells transfected with E1047K and E1052K did not produce measurable currents (data not shown). We were unable to detect expression of the E1052K mutant, suggesting that the amino acid substitution may have affected the overall stability of the protein. However, expression of E1047K was confirmed by Western blot analysis (data not shown), suggesting that either the E1047K mutant is completely inactive or unable to traffic to the cell membrane, thereby indicating that Glu 1047 is essential for TRPM7 channel function.
Changes in pH Sensitivity in TRPM7 Mutants-For the WT TRPM7 channels, acidification of extracellular bath solution increased the inward current by about 12-fold when the pH was lowered from pH 7.4 to 4.0 (Fig. 4, A 1 -A 2 ). Similar increases in the inward currents were observed in the H1049E and H1039M mutants ( Fig. 4, F 1 -F 2 and G 1 -G 2 ). The magnitude of the increase in inward current by acidic external bath solutions was considerably smaller for mutants E1052Q, D1035N, and D1054A. However, no significant difference in pH 1/2 (Fig. 4, A 3 and C 3 -G 3 ) was obtained for D1035N, H1039E, H1039M, D1054A, and E1052Q mutants compared with WT TRPM7. Surprisingly, unlike WT TRPM7 and the other pore mutants, external protons inhibited E1047Q currents in a concentrationdependent manner (Fig. 4B 1 ). The maximal inhibition was about 30%, with an IC 50 of pH 5.4. This unexpected result indicates that Glu 1047 substantially contributes to proton binding, and is therefore likely to be the major binding site for divalent cations.
Mutations at Glu 1047 and Glu 1052 Change the Affinity of TRPM7 for Divalent Cations-We next examined whether the divalent affinity for TRPM7 was changed when negatively charged residues in the putative pore region of the channels were mutated to uncharged amino acids. The inward current of WT TRPM7 carried by monovalent cations can be blocked by micromolar concentrations of Ca 2ϩ or Mg 2ϩ (Fig. 5, A and D), with the IC 50 values of 4.1 Ϯ 0.2 (Fig. 5G) and 3.6 Ϯ 0.4 M (Fig.  5H), respectively. Monovalent currents produced by D1035N and D1054A were similarly blocked by Ca 2ϩ and Mg 2ϩ , with the IC 50 values nearly identical to those of WT TRPM7 (Fig. 5, G-H). The dose-response curves (Fig. 5, G-H) of D1035N and D1054A were superimposable with those of WT TRPM7. Unlike D1035N and D1054A, higher concentrations of Ca 2ϩ and Mg 2ϩ were required to block monovalent currents produced by E1047Q and E1052Q (Fig. 5, B, E, C, and F). The dose-response curves for E1047Q and E1052Q were markedly shifted to the right, with IC 50 values increased by 50-(E1052Q) to 140-fold (E1047Q) compared with WT TRPM7. These results indicate that the affinities of Ca 2ϩ and Mg 2ϩ for the TRPM7 mutants E1047Q and E1052Q were significantly decreased, indicating that Glu 1047 and Glu 1052 residues are critical sites for Ca 2ϩ and Mg 2ϩ binding.
We also tested the effects of Ca 2ϩ and Mg 2ϩ on the monovalent currents of H1039E and H1039M. The IC 50 values of the Ca 2ϩ block were 2.3 Ϯ 0.4 M (n ϭ 6, n H ϭ 1.0) and 2.6 Ϯ 0.5 M (n ϭ 6, n H ϭ 1.0) for H1039M and H1039E, respectively; whereas the IC 50 values for the Mg 2ϩ block were 3.4 Ϯ 0.6 M (n ϭ 6, n H ϭ 0.7) and 3.5 Ϯ 0.4 M (n ϭ 6, n H ϭ 0.8) for H1039M and H1039E, respectively. No statistical significant difference in IC 50 values of Ca 2ϩ and Mg 2ϩ block of H1039M and H1039E was observed as compared with WT TRPM7, indicating that the His 1039 residue is not essential for Ca 2ϩ or Mg 2ϩ binding to TRPM7.
Changes in Voltage-dependent Block by Mg 2ϩ and Ca 2ϩ in Mutants E1047Q and E1052Q-It has been shown that divalent cations block monovalent currents of MIC/MagNuM and TRPM7 in a voltage-dependent manner (35,36). We therefore compared the voltage-dependent effects of Ca 2ϩ and Mg 2ϩ on monovalent currents of WT TRPM7, E1047Q, and E1052Q. As shown in Fig. 6, WT TRPM7 monovalent current was the most potently blocked at Ϫ40 mV (Fig. 6, A and D) with an IC 50 of 1.0 M (Fig. 6A), whereas the IC 50 values at Ϫ120, Ϫ80, ϩ40, and ϩ80 mV were 3.6, 1.8, 51.5, and 1573 M, respectively. The smaller inhibition or the relief of Mg 2ϩ inhibition on TRPM7 at hyperpolarized potentials (Fig. 6, A, D, and G) may suggest "punch-through" of Mg 2ϩ to the inside, consistent with the notion that Mg 2ϩ is a permeant blocker for TRPM7 (35). The FIGURE 4. Effects of acidic pH on WT TRPM7 and its mutants. A 1 -G 1 , changes in the inward currents by acidification of external solutions. Current amplitudes were measured at Ϫ120 mV at the indicated pH and normalized to the amplitude obtained at pH 7.4. Note that E1047Q currents were inhibited by acidic pH solutions, with maximal inhibition of ϳ30% at pH 4.0. A 2 -G 2 , representative recordings of WT TRPM7 and its mutants obtained in the external solutions at pH 7.4 and 4.0. A 3 -G 3 , concentration-dependent effects of protons on WT TRPM7 and its mutants. The changes in current amplitude at the indicated pH were normalized to the maximal change in current amplitude. Average data were fitted with the Hill equation with average parameters obtained from best fits to individual cells. The 50% potentiation pH (pH 1/2 ) and Hill coefficient were pH 1/2 ϭ 4.5 Ϯ 0.5 (n H ϭ 1.2, n ϭ 7) for WT TRPM7; pH 1/2 ϭ 4.1 Ϯ 0.2 (n H ϭ 0.8, n ϭ 5) for D1035N; pH 1/2 ϭ 4.5 Ϯ 0.4 (n H ϭ 0.9, n ϭ 5) for E1052; pH 1/2 ϭ 4.7 Ϯ 0.8 (n H ϭ 2.1, n ϭ 6) for D1054A; pH 1/2 ϭ 4.2 Ϯ 0.4 (n H ϭ 1.1, n ϭ 9) for H1039E; and pH 1/2 ϭ 4.6 Ϯ 0.6 (n H ϭ 1.4, n ϭ 8) for H1039M; respectively. E1047Q inward currents were blocked by low pH, with the 50% inhibition pH of pH 1/2 ϭ 5.4 Ϯ 0.7 (n H ϭ 0.9, n ϭ 5).
best fit of Mg 2ϩ block with a Boltzmann equation estimated the equivalent electrical distance across the ␦ membrane from the extracellular side (␦ out ) to be 0.84 for Mg 2ϩ (Fig. 6G and supplementary materials Table S2), indicating that extracellular Mg 2ϩ binds to TRPM7 at a site of 84% of the membrane electrical field. The Boltzmann equation fit to the relief of the Mg 2ϩ block yielded the fractional electrical distance ␦ from the intracellular side (␦ in ) to be 0.25. The fact that our calculated ␦ out and ␦ in values do not add up to exactly 1.0 could be explained in several ways, including: 1) there may be multiple Mg 2ϩ ions binding to the pore (33); 2) the blocking ion Mg 2ϩ may compete with permeating ion Na ϩ ; 3) there may be conformational changes of the channel caused by binding of the blocking ions; and 4) there may be coupled movement of the blocking ion and permeant ion through the ion channels (33,35,37).
Similar to WT TRPM7, E1052Q also exhibited a voltage-dependent block by Mg 2ϩ (Fig. 6, C, F, and I), albeit the voltage dependence was less dramatic compared with that of WT TRPM7, as evidenced by the shallower slope of I/I 0 curves (Fig.  6I). The best fit of the voltage-dependent block with a Boltzmann function yielded the equivalent electrical distance across the membrane ␦ out of 0.52 (Fig. 6I), indicating that a Mg 2ϩ binding site in E1052Q is located close to the outside surface of the membrane. The Boltzmann equation fit to the relief of the voltage-dependent block on E1052Q yielded ␦ in of 0.43 (supplementary materials Table S2). In contrast to WT TRPM7 and E1052Q, the Mg 2ϩ block on E1047Q was barely relieved at hyperpolarizing potentials as evidenced by the flat I/I 0 curves at negative potentials (Fig. 6H) and the virtually identical IC 50 values at Ϫ120, Ϫ80, and Ϫ40 mV (Fig. 6, B and E), indicating that the blocking ion Mg 2ϩ encounters a large energy barrier and cannot penetrate all the way through the pore (38). Thus, E1047Q may not be able to support measurable Mg 2ϩ currents. The best fit of the voltage-dependent block of Mg 2ϩ on E1047Q with the Boltzmann equation estimated the fractional electrical distance across the membrane ␦ out of 0.36, suggesting that Mg 2ϩ binds to a shallow site (Fig. 6) Table S2).
We further examined if E1047Q and E1052Q affected the ability of internal Mg 2ϩ to block TRPM7 currents. We found that mutations at Glu 1047 and Glu 1052 did not change internal Mg 2ϩ inhibition on TRPM7 currents (supplemental materials Fig. 1), indicating that these residues are not the binding sites for internal Mg 2ϩ . This is consistent with the notion that internal Mg 2ϩ is not accessible to the channel pore, because internal Mg 2ϩ blocks TRPM7 in a voltageindependent manner (15). A model proposed in a recent study also suggests that the internal Mg 2ϩ binding sites are located in the C terminus of TRPM7: one site is located in the kinase domain and the second site is located upstream of the kinase domain (39).
Changes in Relative Permeability by Mutations at Glu 1047 and Glu 1052 -The larger inward currents observed in E1052Q and E1047Q mutants (Fig. 2) and their reduced apparent affinity for Ca 2ϩ and Mg 2ϩ (Fig. 5 and 6) prompted us to determine whether the relative permeability of these mutants to Ca 2ϩ , Mg 2ϩ , and other divalent cations were altered as well (11,12). We assessed the relative permeability by evaluating changes in the current amplitude as previously reported (7,8,12). Divalent cations at 30 mM were used to achieve larger inward currents so that more precise measurements of the relative permeability could be obtained. Zn 2ϩ was prepared at 10 mM due to its low solubility at pH 7.4 (11). Fig. 7, A-E, shows the ratios of current amplitude of the indicated cations normalized to the current amplitude measured in the solution containing 30 mM Ca 2ϩ . The mutants D1035N and D1054A exhibited sub-  Table S2). stantial permeation to different divalents with a relative permeability to different cations similar to that observed for WT TRPM7 (Fig. 7, A, B, and E). In contrast with D1035N and D1054A, E1052Q exhibited a decreased divalent permeability, as was evident from the ratio of I Tyrode /I Ca being substantially larger than that of I Ni /I Ca (Fig. 7D). In addition, Ba 2ϩ permeation through E1052Q appeared smaller than that of WT TRPM7, D1035N, and D1054A (Fig. 7, A-D). Nevertheless, all the tested divalents permeated through E1052Q. Intriguingly, E1047Q exhibited very little permeation to the divalents such that Mg 2ϩ , Ca 2ϩ , and Zn 2ϩ currents were barely detectable (Fig. 7C). By contrast, the ratio of I Tyrode /I Ca for E1047Q was significantly larger than that for WT TRPM7 and other mutants, indicating that currents through E1047Q in Tyrode solutions were mainly carried by monovalent cations. Fig. 7F shows the normalized Mg 2ϩ and Ca 2ϩ currents versus the current amplitude obtained in Tyrode solution. In E1047Q, the current amplitude of Mg 2ϩ and Ca 2ϩ was only 1.1 and 2.3% of that observed in WT TRPM7, respectively. In E1052Q, the current amplitude carried by Mg 2ϩ and Ca 2ϩ was 24.3 and 24.1% of that observed in WT TRPM7, respectively. These results strongly suggest that Glu 1047 is the dominant residue that confers Ca 2ϩ and Mg 2ϩ permeability to TRPM7. In contrast to the changes to divalent permeability, the sequence for monovalent permeability (K ϩ Ͼ Cs ϩ Ͼ Na ϩ ) (Fig. 7, A-E) was not changed in all the mutants tested compared with WT TRPM7.
Mutation of Glu 1047 Diminishes Ca 2ϩ Permeation and Largely Eliminates Mg 2ϩ Permeation-We further studied the Ca 2ϩ and Mg 2ϩ permeation properties of E1047Q and E1052Q using isotonic Ca 2ϩ and Mg 2ϩ solutions (120 mM Ca 2ϩ or Mg 2ϩ ). Currents were recorded using a P2 pipette solution to minimize outward currents. In WT TRPM7, the inward current amplitude in isotonic Ca 2ϩ and Mg 2ϩ solutions was similar to that in Tyrode solution or in 2 mM Ca 2ϩ , 150 mM monovalent solutions (Fig. 8, A, D, and G). Changes in reversal potentials of TRPM7 in isotonic Ca 2ϩ and Mg 2ϩ solutions were also similar to those in 2 mM Ca 2ϩ /monovalent solutions (Fig. 8J). In clear contrast to WT TRPM7, the inward current amplitude of E1047Q in isotonic Ca 2ϩ and Mg 2ϩ solutions was significantly smaller than those in 2 mM Ca 2ϩ Tyrode solution (Fig. 8, B and E). There was almost no Mg 2ϩ conductance in isotonic Mg 2ϩ solution, as shown in Fig. 8, B and E. The average current amplitude shown in Fig. 8H also indicates that the Ca 2ϩ current was significantly reduced, whereas the Mg 2ϩ current was almost undetectable in the E1047Q mutant. The isotonic Mg 2ϩ and Ca 2ϩ current amplitude of E1047Q (Fig. 8H) was 2.1 and 6.0% of the current amplitude of WT TRPM7 (Fig. 8G), respectively. Consistent with the small conductances in the isotonic solutions, the reversal potentials of E1047Q in isotonic Ca 2ϩ and Mg 2ϩ solutions were much more negative than that in Tyrode solution (Fig. 8, B and K). Unlike E1047Q, E1052Q exhibited substantial inward Ca 2ϩ and Mg 2ϩ currents (Fig. 8, C, F, and I), albeit the current amplitude was smaller compared with that of WT TRPM7. Changes in reversal potentials under the indicated conditions in reference to the value obtained in Tyrode solution were smaller than those of WT TRPM7 (Fig. 8J), but much larger than those of E1047Q (Fig. 8L). Based on the above results, we conclude that E1047Q and E1052Q are critical for Mg 2ϩ and Ca 2ϩ permeation through TRPM7.
Mutations of Glu 1024 and Glu 1029 Changes Divalent Permeability and pH Sensitivity of TRPM6-The above results strongly suggest that Glu 1047 and Glu 1052 are key amino acid residues that confer Mg 2ϩ and Ca 2ϩ selectivity to TRPM7. To further confirm the importance of Glu 1047 and Glu 1052 in Ca 2ϩ FIGURE 7. Changes in divalent permeability of TRPM7 mutants (mean ؎ S. E., n ‫؍‬ 6). A-E, currents recorded under the indicated conditions were normalized to the current amplitude value obtained in 30 mM Ca 2ϩ external solutions. The sequence of monovalent permeability was not changed (K ϩ Ͼ Cs ϩ Ͼ Na ϩ ) in the mutants compared with WT TRPM7; however, the monovalent permeabilities in E1047Q were significantly larger than that of WT TRPM7. F, currents obtained in 30 mM Ca 2ϩ or Mg 2ϩ were normalized to the current amplitude obtained in Tyrode solution. Note that the ratios of I Ca /I Tyrode and I Mg /I Tyrode for E1047Q were 0.014 and 0.0096, respectively. The I Ca(E1047Q) /I Ca(TRPM7) was 0.023, and I Mg(E1047Q) /I Mg(TRPM7) was 0.011, respectively. and Mg 2ϩ permeation, we generated mutants E1024Q and E1029Q in TRPM6 at the equivalent positions for Glu 1047 and Glu 1052 in TRPM7, respectively. As shown in Fig. 9, A-C, TRPM6 mutants E1024Q and E1029Q exhibited significant changes in their I-V relationship compared with WT TRPM6. The normalized I-V curves (Fig. 9D) show that the ratios of inward currents at Ϫ120 mV versus outward currents at ϩ100 mV of E1024Q (0.7) and E1029Q (0.3) were 14-and 6-fold larger than that of WT TRPM6 (0.05). The I-V relationships for E1029Q and E1024Q in TRPM6 were similar to the TRPM7 mutants E1052Q and E1047Q (Fig. 2).
We next determined whether the binding affinity of the TRPM6 pore mutants for divalents was changed. Similar to what we observed for E1047Q and E1052Q, the affinity of Ca 2ϩ and Mg 2ϩ was significantly decreased in TRPM6 mutants E1024Q and E1029Q (Fig. 9, E and F) 2ϩ and Mg 2ϩ , the permeation of Ca 2ϩ and Mg 2ϩ was also largely decreased. As shown in Fig. 9G, the inward current amplitude of the WT TRPM6 obtained in isotonic Ca 2ϩ and Mg 2ϩ solutions were similar to the current amplitude in Tyrode solution. By contrast, Ca 2ϩ currents of E1024Q and E1029Q obtained in the isotonic solutions were significantly smaller than those obtained from WT TRPM6 (Fig. 9, G-I).
Moreover, there was almost no Mg 2ϩ permeation through E1024Q (Fig. 9, H and J). These results indicate that similar to Glu 1047 and Glu 1052 in TRPM7, Glu 1024 and Glu 1029 residues are essential for the Mg 2ϩ and Ca 2ϩ permeability of TRPM6.
As TRPM6 is sensitive to acidic pH (11), we next tested if the pH sensitivity was changed in mutants E1024Q and E1029Q. As shown in Fig. 10, external protons enhanced inward currents of E1029Q (Fig. 10, C, F, and I). The degree of increase in E1029Q inward currents was smaller than the increase in TRPM6 inward current induced by low pH, but similar to the changes in FIGURE 8. E1047Q diminishes Ca 2؉ permeation and eliminates Mg 2؉ permeation. A-C, representative currents of TRPM7, E1047Q, and E1052Q elicited by ramp protocols in 2 mM Ca 2ϩ /Tyrode, NMDG, isotonic Ca 2ϩ , and isotonic Mg 2ϩ solutions, respectively. P2 pipette solution containing reduced Cs ϩ concentration (10 mM) was used. D-F, inward currents measured at Ϫ120 mV under different conditions in the representative cells for TRPM7 (D), E1047Q (E), and E1052Q (F). G-I, average current amplitudes (mean Ϯ S.E., n ϭ 6) measured at Ϫ120 mV under the indicated conditions for TRPM7, E1047Q, and E1052Q. J-L, changes in reversal potentials of TRPM7 (J), E1047Q (K), and E1052Q (L) (mean Ϯ S.E., n ϭ 6). ⌬E rev was obtained by subtracting the reversal potential under indicated conditions by the reversal potential in Tyrode solution. E1052Q (Fig. 4, C 1 -C 3 ). In contrast to E1029Q, E1024Q was blocked by acidic external solutions by 42% at pH 3.0 (Fig. 10, B, E, and H) with an IC 50 of pH 5.0. Thus, similar to what we found for residues Glu 1047 and Glu 1052 in TRPM7, TRPM6 residues Glu 1024 and Glu 1029 are critical determinants of the sensitivity of the channel to external pH.
Double Mutant E1047Q/E1052Q Exhibits Similar Properties to Those of E1047Q-Because mutation of Glu 1047 in TRPM7 produced dramatic changes in TRPM7 channel properties, and mutation of Glu 1052 also generated substantial changes in Ca 2ϩ and Mg 2ϩ permeability as well as pH sensitivity, this prompted us to ask how mutation of both sites (E1047Q/E1052Q) would affect TRPM7 channel properties. Fig.  11A shows that the E1047Q/ E1052Q current elicited by a ramp protocol exhibited a double rectifying I-V relation similar to that of E1047Q (Fig. 2, E and H). Like E1047Q, the inward currents in isotonic Ca 2ϩ or Mg 2ϩ solutions (120 mM) were almost undetectable (Fig.  11, B-D), and the reversal potentials of E1047Q/E1052Q in isotonic Ca 2ϩ and Mg 2ϩ solutions were almost identical to those in NMDG solutions (Fig. 11B), suggesting a largely reduced permeability to Ca 2ϩ and Mg 2ϩ . Consistent with this notion, Mg 2ϩ affinity for E1047Q/E1052Q was significantly smaller than that of WT TRPM7. At Ϫ120 mV, IC 50 of the Mg 2ϩ block on the monovalent currents of the double mutant was 132.7 M (Fig.  11E). The voltage-dependent effect of Mg 2ϩ on the E1047Q/E1052Q monovalent currents (Fig. 11, E and  F) was also similar to that of E1047Q. IC 50 values of Mg 2ϩ block at Ϫ120, Ϫ80, and Ϫ40 mV were almost identical, indicating that there was no relief of Mg 2ϩ block at hyperpolarizing potentials. Consistent with this notion, the current ratio (I/I 0 ) and voltage relation (Fig.  11F) shows a virtually flat line at hyperpolarizing voltages, further suggesting that the punch-through mechanism of Mg 2ϩ permeation had been disabled in E1047Q/E1052Q. The best fit of a Boltzmann equation to I/I 0 curves produced a slope factor k of 29 mV, generating the estimated distance of the electrical field ␦ out ϭ 0.44 from the outside surface of the membrane. Similar to the effects of Mg 2ϩ , the IC 50 values of Ca 2ϩ block on E1047Q/E1052Q were 164.6 Ϯ 20.7 M at Ϫ120 mV, 170.3 Ϯ 24.6 M at Ϫ80 mV, 166.9 Ϯ 26.7 M at Ϫ40 mV, 723.1 Ϯ 89.4 M at ϩ40 mV, and 2.4 Ϯ 0.3 mM at ϩ80 mV, respectively (n ϭ 5). The lack of voltage-dependent relief of Ca 2ϩ block at hyperpolarized potentials further suggests the diminished divalent permeation through E1047Q/E1052Q. The above similar properties between E1047Q/E1052Q (Fig. 11) and E1047Q (Fig. 6) strongly suggest that, like the E1047Q mutation, double mutation of Glu 1047 and Glu 1052 largely eliminated a high-affinity divalent binding site that is present in the deep pore of WT TRPM7.
Unlike E1047Q currents, which were slightly inhibited by acidic external solutions (Fig. 4, B 1 -B 3 ), the inward currents of the double mutant E1047Q/E1052Q were increased to a small degree by high concentrations of external protons (Fig. 11, E and F). This potentiation of inward currents of E1047Q/ E1052Q by low pH was similar to that of E1052Q, although the degree of increase was much smaller than that in E1052Q. These results provide further evidence that residues Glu 1047 and Glu 1052 are indeed important in determining divalent permeability and pH sensitivity of TRPM7.

DISCUSSION
In the present study, we demonstrate that mutation of Glu 1052 decreases its Ca 2ϩ and Mg 2ϩ permeability; whereas mutation of Glu 1047 largely eliminates Ca 2ϩ and Mg 2ϩ permeability and converts TRPM7 into a monovalent selective cation channel. In addition, external protons, which significantly increase TRPM7 inward currents, fail to enhance the inward currents of E1047Q. Furthermore, mutations at the equivalent sites in TRPM6, E1024Q and E1029Q, produce similar phenotypic changes to those observed in E1047Q and E1052Q. These results indicate that Glu 1047 /Glu 1052 in TRPM7 and Glu 1024 /Glu 1029 in TRPM6 are essential for the Mg 2ϩ and Ca 2ϩ permeability and pH sensitivity of these channels.
to divalent permeability. In the present study, by systematically substituting the negatively charged residues in the putative pore region of TRPM7 with uncharged residues, we identified Glu 1047 and Glu 1052 as the key residues for Ca 2ϩ /Mg 2ϩ permeability and pH sensitivity. This conclusion is supported by the fact that mutations at the equivalent positions in TRPM6 (Glu 1024 and Glu 1029 ) generated similar changes in divalent permeability and pH sensitivity. It is remarkable that neutralization of a single amino acid residue, E1047Q (or E1024Q), largely eliminated divalent permeation, converting the divalent selective TRPM7 (or TRPM6) to a virtually monovalent selective channel, and at the same time abolished external pH sensitivity. The feature that the divalent selectivity and pH sensitivity are conferred by the same residue in TRPM6 and TRPM7 is unique among TRP channels. It is known that TRPV5 (or TRPV6) is highly Ca 2ϩ selective (34,40) and also sensitive to external pH (41,42). However, the pH sensitivity mediated by Glu 522 in TRPV5 is independent of its Ca 2ϩ selectivity determined by Asp 542 (42,43). TRPV1 channel activity is regulated by acidic pH through Glu 600 and Glu 648 residues (30), whereas its divalent permeability appears to be mediated by Asp 646 (44). External low pH activates TRPV4 (45), yet it is unknown whether this pH sensitivity of the channel is correlated with its Ca 2ϩ permeation determined by Asp 672 and Asp 682 residues in the pore (46). The monovalent selective channel TRPM5 is also inhibited by protons (47). Nonetheless, the molecular mechanism of pH sensitivity and divalent permeability of TRPM7/TRPM6 resembles to some extent the properties of the voltage-gated Ca 2ϩ channels (VGCCs). It has been demonstrated that protons block VGCCs by binding to the Ca 2ϩ binding sites, Glu residues, in the pore-forming region (48,49). Mutation of Glu by Gln replacement in repeats I or III abolished the high-conductance state of VGCCs, as if the titration site had become permanently protonated (48). Like VGCCs, mutation of Glu 1047 in TRPM7 (or Glu 1024 in TRPM6) increased inward monovalent currents, resembling the effect produced when the external bath solution is acidified on WT TRPM7 (or TRPM6) currents (11). Furthermore, TRPM6 and TRPM7 lose their high affinity binding sites when Glu 1024 /Glu 1047 and Glu 1029 /Glu 1052 are replaced by Gln, similar to the loss of high affinity binding sites for divalent cations in VGCCs when the pore Glu residues are mutated (38).
It is noticeable that both E1047Q and E1052Q produced substantial changes in channel properties compared with WT TRPM7, although the changes such as I-V relation, decreased affinity to Ca 2ϩ and Mg 2ϩ , and diminished divalent permeation in E1052Q are less dramatic than those observed in E1047Q. The double mutant E1047Q/E1052Q exhibited similar properties to those of E1047Q, suggesting that although both Glu 1047 and Glu 1052 are essential, Glu 1047 dominates in determining divalent permeability and pH sensitivity. Using the Eyring rate model (35), Kerschbaum and colleagues predicted a high affinity site for binding Mg 2ϩ within the electric field and two low affinity sites for MIC/TRPM7 channel. In agreement with this notion, our data indicate that Glu 1047 and Glu 1052 compose a high affinity binding site. Mutation of either or both of these residues disrupts the high affinity site (Figs. 5 and 6). As we used single Boltzmann functions for data analysis, our data suggest FIGURE 11. Changes in Mg 2؉ and Ca 2؉ permeability and pH sensitivity in the double mutant E1047Q/E1052Q of TRPM7. A, typical recording of the E1047Q/E1052Q current elicited by a voltage ramp protocol ranging from Ϫ120 to ϩ100 mV. B, currents of E1047Q/E1052Q recorded in Tyrode, NMDG, isotonic Ca 2ϩ and Mg 2ϩ (120 mM) solutions. Note the diminished inward currents in isotonic Ca 2ϩ and Mg 2ϩ solutions. C, inward current amplitude measured at Ϫ120 mV under various conditions. D, average current amplitudes at Ϫ120 mV in isotonic Ca 2ϩ and Mg 2ϩ solutions normalized to the current amplitude in the Tyrode solution. The current amplitude of E1047Q/E1052Q in isotonic Ca 2ϩ and Mg 2ϩ (120 mM) solutions was 0.1 Ϯ 0.06 and 0.08 Ϯ 0.04 (n ϭ 6) of that in Tyrode solution, respectively. E, concentration-dependent effects of Mg 2ϩ on E1047Q/E1052Q at various voltages. The IC 50  that there is one low affinity site after mutation of the high affinity site Glu 1047 /Glu 1052 .
Analysis of the voltage-dependent effects of Mg 2ϩ /Ca 2ϩ suggests that the high affinity site is located at about 84% of the electrical distance of membrane, similar to the value estimated by Kerschbaum and colleagues (35) for the MIC/ TRPM7 channel. The fractional electrical distance from the outside surface of the membrane ␦ out estimated based on the Mg 2ϩ block changes from 0.84 in WT TRPM7 (Fig. 6) to 0.36 in E1047Q, and 0.52 in E1052Q. In the double mutant, the ␦ out (0.44) is close but not identical to that of E1047Q or E1052Q, suggesting that a low affinity site is located between ␦ out of 0.36 and 0.52. The differences in ␦ out for the potential low affinity site may be a reflection of experimental variations, or may represent reorganization of the pore structure after mutating of Glu 1047 , Glu 1052 , and Glu 1047 /Glu 1052 . In addition, we used the simplest model for data analysis, which assumes one permeant ion occupies a binding site in the channel pore (50). This may also contribute to the variances in estimated ␦ out for the low affinity site. Nonetheless, our data indicate that a low affinity site is located close to the outer surface of the cell membrane. Taken together, our results suggest that a high affinity binding site comprising Glu 1047 and Glu 1052 is located deep within the pore (82 to 84%), whereas a low affinity binding site is shallower, within 36 -52% of the electrical distance of membrane.
Selectivity Filter of TRPM6 and TRPM7-TRP channels vary from each other by their gating mechanisms and permeability profiles. TRPV5 and TRPV6 have the highest Ca 2ϩ selectivity (51), whereas TRPM4 and TRPM5 are monovalent selective and impermeable to divalent cations (52)(53)(54)(55). Similar to VGCCs, the Ca 2ϩ selectivity of TRPV5/TRPV6 is determined by a single pore residue, Asp 542 (43). The acidic residues in the putative pore of TRPV1 (Asp 646 ) (44) and TRPV4 (Asp 682 ) (46) are also involved in divalent permeability. The pore structure of TRPM channels is less well defined (51). A recent study by Nilius and colleagues (56) revealed that an acidic stretch of six residues ( 981 EDMDVA 986 ) between TM5 and TM6 is the selectivity filter for TRPM4. According to our sequence alignment, the selectivity filter of TRPM4 corresponds to the region between residues 1051 YEIDVC 1056 in TRPM7 and 1028 GEI-DVC 1033 in TRPM6. Different from this prediction, our data indicate the residues critical to ion selectivity in the pore region are 1047 EVYAYE 1052 for TRPM7 and 1024 EVYAGE 1029 for TRPM6. We demonstrate that the E1047Q mutant produced the most dramatic changes in TRPM7 channel properties. Not only does E1047Q eliminate Ca 2ϩ and Mg 2ϩ permeation, but it also abolishes a proton-induced increase in inward currents. Thus, the Glu 1047 residue should be considered a key component of the selectivity filter. In addition to E1047Q, E1052Q also influenced Ca 2ϩ /Mg 2ϩ permeability. Furthermore, the double mutant E1047Q/E1052Q exhibited similar properties to those of E1047Q. Therefore, it is likely the Glu 1047 and Glu 1052 residues form a high affinity binding site for divalent cations. Asp 1054 is a conserved residue in all TRPM channels. However, the exchange of Asp at position 1054 for Ala (D1054A) did not alter the apparent affinity for Ca 2ϩ and Mg 2ϩ . The negatively charged residue preceding Glu 1047 , Asp 1035 , has no influence on Ca 2ϩ and Mg 2ϩ permeability and selectivity. Therefore, our results predict that the selectivity filter of TRPM7 lies between Glu 1047 and Glu 1052 (or Glu 1024 -Glu 1029 for TRPM6).
The negatively charged residue Glu 1047 is conserved in TRPM1, TRPM3, TRPM6, and TRPM7. Interestingly, mutation of Gln 977 by Glu in TRPM4, the analogous residue to Glu 1047 in TRPM7, altered the monovalent cation permeability sequence and rendered the TRPM4 pore permeable to Ca 2ϩ (56), suggesting that the Glu 1047 position is indeed crucial for determining the divalent selectivity of TRPM channels. It will be of interest to investigate whether mutation of this conserved residue influences the divalent permeability of TRPM1 and TRPM3 as well. Nonetheless, it seems that the pore structure of TRPM channels is more complicated than first imagined. For example, Oberwinkler and colleagues (57) recently identified several TRPM3 spliced variants, TRPM3␣1-5. Whereas the TRPM3␣-2 displays high permeability to Ca 2ϩ and Mg 2ϩ , TRPM3␣-1, which includes an additional stretch of 12 amino acids following the invariant Asp, a residue equivalent to Asp 1054 in TRPM7, exhibited a more than 10-fold lower permeability for divalent cations (57). This could be attributed to a possibility that the additional 12 residues result in re-arrangement of the pore structure. Because Asp 1054 is conserved in all the TRPM channels, including both divalent permeable and monovalent selective (TRPM4 -5) ones, it may be less influential than Glu 1047 and Glu 1052 on divalent selectivity. Taken together, our results suggest that 1047 EVYAYE 1052 in TRPM7 and 1024 EVY-AGE 1029 in TRPM6 constitute the selectivity filter, and that Glu 1047 /Glu 1024 and Glu 1052 /Glu 1029 are essential for Ca 2ϩ / Mg 2ϩ selectivity and pH sensitivity.
Conclusions-The importance of Mg 2ϩ and Ca 2ϩ permeation of TRPM6 and TRPM7 has been well demonstrated. Mutations in TRPM6 lead to familial hypomagnesemia and secondary hypocalcemia (20,21), and deletion of TRPM7 causes cell death (7,10,22). Moreover, it has been demonstrated recently that the sensitivity of TRPM7 to external pH may contribute to the ability of the channel to control neurotransmitter release (28). In the present study, we provide strong evidence demonstrating that the amino acid residues Glu 1024 /Glu 1047 and Glu 1029 /Glu 1052 in the putative pore region of TRPM6 and TRPM7 are essential for their Ca 2ϩ and Mg 2ϩ permeability and pH sensitivity. These findings provide the molecular mechanisms by which TRPM6 and TRPM7 exert their physiological functions, and may in fact serve as a platform for future work aimed at discovering pharmacological agents to manipulate TRPM6 and TRPM7 channel functions. In addition, our results provide valuable information on the pore architecture of TRPM6 and TRPM7 channels, knowledge that will help guide future investigations into the complex nature of TRPM channel family members.