Tuning the ion selectivity of two-pore channels

Significance

Ion channels selectively transfer ions across cell membranes, and their selectivity is controlled by a special region of the channel protein called the selectivity filter. Two-pore channels (TPCs) belong to the superfamily of voltage-gated tetrameric cation channels and possess a unique set of filter residues that define their ion selectivity. Despite extensive studies, debate still lingers about the selectivity properties of mammalian TPCs. Here, we provide structural and functional insights into the selectivity properties of TPC channels. We confirm the Na⁺ selectivity of human TPC2, identify the key residues in the TPC filters that differentiate the selectivity between mammalian TPC2 and plant TPC1, and reveal the structural basis of Na⁺ selectivity in mammalian TPCs.

Abstract

Organellar two-pore channels (TPCs) contain two copies of a Shaker-like six-transmembrane (6-TM) domain in each subunit and are ubiquitously expressed in plants and animals. Interestingly, plant and animal TPCs share high sequence similarity in the filter region, yet exhibit drastically different ion selectivity. Plant TPC1 functions as a nonselective cation channel on the vacuole membrane, whereas mammalian TPC channels have been shown to be endo/lysosomal Na⁺-selective or Ca²⁺-release channels. In this study, we performed systematic characterization of the ion selectivity of TPC1 from Arabidopsis thaliana (AtTPC1) and compared its selectivity with the selectivity of human TPC2 (HsTPC2). We demonstrate that AtTPC1 is selective for Ca²⁺ over Na⁺, but nonselective among monovalent cations (Li⁺, Na⁺, and K⁺). Our results also confirm that HsTPC2 is a Na⁺-selective channel activated by phosphatidylinositol 3,5-bisphosphate. Guided by our recent structure of AtTPC1, we converted AtTPC1 to a Na⁺-selective channel by mimicking the selectivity filter of HsTPC2 and identified key residues in the TPC filters that differentiate the selectivity between AtTPC1 and HsTPC2. Furthermore, the structure of the Na⁺-selective AtTPC1 mutant elucidates the structural basis for Na⁺ selectivity in mammalian TPCs.

Two-pore channels (TPCs) are organellar cation channels ubiquitously expressed in animals and plants (1, 2) and belong to the voltage-gated ion channel superfamily (3). TPC channels contain two homologous Shaker-like six-transmembrane (6-TM) domains in each subunit and function as homodimers. They are believed to be evolutionary intermediates between homotetrameric voltage-gated potassium channels and the four-domain, single-subunit, voltage-gated sodium/calcium channels (4).

In human and animals, TPC channels (TPC1 and TPC2) are localized to the endo/lysosomal membranes and regulate the ionic homeostasis within these acidic organelles. Their functions have been shown to be involved in various physiological processes, such as endocytosis and endosomal trafficking (5, 6), lysosomal morphology and pigmentation (7, 8), autophagy (9, 10), and nutrient metabolism via the mammalian target of rapamycin (mTOR) complex (11). Not surprisingly, given their central physiological role, defects in TPC channels are associated with a variety of disabling human disorders (9, 12–17).

Despite their physiological importance, some biophysical properties of mammalian TPC channels are still under debate (18). Mammalian TPC channels were initially identified as receptors of nicotinic acid adenine dinucleotide phosphate (NAADP), a Ca²⁺-mobilizing second messenger, and responsible for Ca²⁺ release from the acidic organelles (19–21). Although the NAADP-dependent endo/lysosomal Ca²⁺ release is attributed directly to the Ca²⁺ conduction of TPC channels activated by NAADP in some studies (16, 22–26), several recent studies demonstrate that TPCs are Na⁺-selective channels that are activated by phosphatidylinositol 3,5-bisphosphate [PI(3,5)P₂], instead of NAADP (11, 27), however. The Na⁺ permeability of TPCs was also reported in several other recent studies (26, 28, 29). In this study, we aim to provide structural and functional insights into the ion selectivity properties of TPC channels by taking advantage of our recently reported crystal structure of a plant TPC1 from Arabidopsis thaliana, AtTPC1, which is a voltage-gated, Ca²⁺-regulated, nonselective cation channel localized to the vacuolar membrane (30). Despite the difference in ion selectivity between AtTPC1 and its human counterparts, their filter sequences are quite conserved, particularly between AtTPC1 and human TPC2 (HsTPC2), indicating a similar overall structure at the selectivity filter region. Here, we performed electrophysiological characterization and comparison of the ion selectivity properties of both AtTPC1 and HsTPC2. Through structure-guided mutagenesis, we were able to convert the nonselective AtTPC1 to a Na⁺-selective channel similar to HsTPC2, and thereby identified the key filter residues that are central to defining TPC selectivity. We also determined the structure of the Na⁺-selective AtTPC1 mutant, which, in comparison to AtTPC1, provides structural insights into the selectivity mechanism in TPC channels.

Results

Ion Selectivity of AtTPC1.

We overexpressed AtTPC1 in HEK293 cells (Fig. S1) and measured the channel activity on the plasma membrane using patch clamping in the whole-cell configuration (Materials and Methods). Under this setting, the extracellular side (bath side) is equivalent to the luminal side of AtTPC1 in vacuoles. To determine the selectivity of AtTPC1, the membrane potential was stepped to +80 mV for channel activation and then switched to various testing potentials. The tail currents were recorded to generate a current-voltage (I-V) curve for determination of the reversal potential. The standard pipette solutions (cytosolic side) contained 150 mM Na⁺; therefore, all selectivity measurements of AtTPC1 among various monovalent and divalent cations were compared with Na⁺. Under bi-ionic conditions with 150 mM Li⁺, Na⁺, or K⁺ in the bath solutions, the reversal potentials were all close to 0 mV, indicating that AtTPC1 exhibits almost equal permeability to these ions (Fig. 1 A, B, and D). AtTPC1 is less selective for Rb⁺ and Cs⁺, with reversal potentials of −20.4 ± 4.6 mV and −39.6 ± 3.3 mV, respectively, yielding a permeability (P) ratio P_Na/P_Rb of 2.2 and a P_Na/P_Cs of 4.7. Because AtTPC1 is known to conduct Ca²⁺, we also measured the relative selectivity between Ca²⁺ and Na⁺. Because luminal Ca²⁺ can inhibit AtTPC1 activity by shifting voltage activation toward a more positive potential (30, 31), we generated a mutant channel, AtTPC1ΔCa_i, which contains three mutations (Asp240Ala, Asp454Ala, and Glu528Ala) at the inhibition site; these mutations mitigate the Ca²⁺ inhibition but have no effect on ion permeation (Fig. S2), allowing for the measurement of the Ca²⁺ current. As shown in Fig. 1C, with the presence of 15 mM Ca²⁺ in the bath solution and 150 mM Na⁺ in the pipette, AtTPC1ΔCa_i has a reversal potential of about 0 mV, indicating a higher selectivity for Ca²⁺ over Na⁺ with a P_Ca/P_Na of about 5 (Fig. 1 C and D). Therefore, under our recording conditions, AtTPC1 is permeable to various cations and has a relative permeability sequence of Ca²⁺ > Na⁺ ∼ Li⁺ ∼ K⁺ > Rb⁺ > Cs⁺ (Fig. 1D).

Fig. S1.

Subcellular localization of AtTPC1 and HsTPC2 expressed on HEK293 cells. (A) Confocal images of HEK293 cells coexpressing AtTPC1 tagged with EGFP and the plasma membrane marker KRas GTPase tail tagged with mCherry. Cells were incubated with 1 mM thapsigargin (TG) for 5 min before imaging. The incubation with TG aims to raise the cytosolic [Ca²⁺], mimicking the cytosolic condition used in our electrophysiological recordings of AtTPC1. (B) Confocal images of HEK293 cells coexpressing HsTPC2_L11A/L12A tagged with EGFP and the plasma membrane marker KRas tail tagged with mCherry.

Fig. 1.

Ion selectivity of AtTPC1. (A) Sample traces of whole-cell currents of AtTPC1 overexpressed on the plasma membrane of HEK293 cells with 150 mM Na⁺ in the pipette solution and 150 mM X⁺ (X = Li, Na, K, Rb, or Cs) in the bath solution. All recordings shown here were performed on a single patch, and the bath solutions were exchanged through perfusion. (B) I-V curves generated from the tail currents of the traces shown in A. (C) Sample traces of whole-cell currents with 150 mM Na⁺ in the pipette solution and 15 mM Ca²⁺ in the bath solution and the corresponding I-V curve. (D) Summary of reversal potentials of AtTPC1 with various cations in the bath solutions and the calculated relative permeability of these ions in comparison to Na⁺. The reversal potential values are the mean ± SEM of at least five measurements from different patches. Vm, membrane potential.

Fig. S2.

AtTPC1ΔCa_i mutant is used for Ca²⁺ permeability measurement. The mutant contains three mutations (Asp240Ala, Asp454Ala, and Glu528Ala) at the extracellular/luminal Ca²⁺ inhibition site. (A) Sample traces of voltage activation of wild-type AtTPC1 in the presence of 0 and 10 mM extracellular (bath) Ca²⁺. (B) Sample traces of voltage activation of AtTPC1ΔCa_i mutant in the presence of 0 and 15 mM extracellular (bath) Ca²⁺. The pipette and bath solutions both contain 150 mM Na⁺. (C) G/G_max-V curves of wild-type AtTPC1 and AtTPC1ΔCa_i with or without extracellular Ca²⁺. Boltzmann fit yields V_1/2 = −28.7 mV, Z = 3.9 for wild-type AtTPC1; V_1/2 = −19.8 mV, Z = 3.6 for AtTPC1ΔCa_i without extracellular Ca²⁺; and V_1/2 = 35.0 mV, Z = 2.2 for AtTPC1ΔCa_i with 15 mM extracellular Ca²⁺ (n ≥ 5). Although extracellular Ca²⁺ at 15 mM still partially inhibits AtTPC1ΔCa_i by shifting voltage activation toward a more positive potential, the mutant channel can be fully activated at 100 mV, allowing for the measurement of the Ca²⁺ current. Vm, membrane potential.

HsTPC2 Is a Na⁺-Selective Channel.

Contrary to plant TPC1, mammalian TPC channels have been shown to be selective, despite a disagreement on whether they are selective for Na⁺ over Ca²⁺. To compare the selectivity properties between plant and mammalian TPCs, we chose HsTPC2 as the model system because it has a high sequence similarity to AtTPC1 at the filter region (Fig. 2A). By replacing two leucines with alanines (L11A/L12A) on the N-terminal lysosomal targeting sequence, HsTPC2 can be overexpressed and trafficked to the plasma membrane in HEK293 cells (Fig. S1), allowing for direct measurement of channel activity by patching the plasma membrane (22, 28). We applied this strategy and recorded channel activity of HsTPC2 using the inside-out patch configuration (Materials and Methods). In our recordings, HsTPC2 on the plasma membrane was activated by cytosolic PI(3,5)P₂ (10 μM in bath), but no obvious channel activation by NAADP in a wide range of concentrations (10 nM ∼ 10 µM) was observed (Fig. 2B). Under bi-ionic conditions with 150 mM Na⁺ in the pipette and 150 mM K⁺ in the bath solution, HsTPC2 has a reversal potential of about 81 mV, indicating high selectivity for Na⁺ with a P_Na/P_K of 23.8 (Fig. 2B). To measure the selectivity between Ca²⁺ and Na⁺ or K⁺ for HsTPC2, we included 100 mM Ca²⁺ in the pipette (extracellular) and 150 mM Na⁺ or K⁺ in the bath solution. In this setting, the channel shows high selectivity for Na⁺ over Ca²⁺ with a reversal potential of about −50 mV, yielding a P_Na/P_Ca of about 16.8; the channel shows no selectivity between K⁺ and Ca²⁺ with a reversal potential of about 7 mV, yielding a P_Ca/P_K of about 1.2 (Fig. 2C). Therefore, our study demonstrates that HsTPC2 is a Na⁺-selective channel activated by PI(3,5)P₂. Our results are consistent with the observations from whole-lysosome recordings of PI(3,5)P₂-activated mammalian TPC2 in enlarged lysosomes (11, 27), but different from whole-lysosome planar patch-clamp recordings of NAADP-activated currents (24, 26). Albeit small, we did observe inward Ca²⁺ current in our recording. Considering the large [Ca²⁺] gradient (∼10,000-fold) across the lysosomal membrane, it is possible that a small fraction of HsTPC2 current could come from Ca²⁺ in vivo.

Fig. 2.

Ion selectivity of HsTPC2. (A) Partial sequence alignment of AtTPC1, HsTPC1, and HsTPC2 at the filter regions. IP1, filter I, and IP2 mark the pore helices and selectivity filter from the first 6-TM domain; IIP1, filter II, and IIP2 are from the second 6-TM domain. The three central filter residues are shown in red. (B) I-V curves of HsTPC2 in the presence of cytosolic PI(3,5)P₂ or NAADP. A wide range of NAADP concentrations (10, 30, and 100 nM and 1 and 10 μM) was tested, and all gave the similar results. Therefore, only a representative trace with 30 nM NAADP is shown. The pipette solution contains 150 mM Na⁺, and the bath solution contains 150 mM K⁺. (C) I-V curve of HsTPC2 activated by 10 μM PI(3,5)P₂, with 100 mM Ca²⁺ in the pipette solution and 150 mM Na⁺ or K⁺ in the bath solution. All currents were recorded using inside-out patches. The measured reversal potentials and calculated relative permeability values between Na⁺ and K⁺ or Ca²⁺ are listed in Table 1.

Converting AtTPC1 to a Na⁺-Selective Channel.

Based on the structure of AtTPC1 and the filter sequence alignment, the major differences between AtTPC1 and HsTPC2 lie in the three central filter residues that form the ion pathway (Fig. 2A). AtTPC1 has a sequence of ₂₆₄TSN₂₆₆ at filter I (from the first 6-TM domain) and ₆₂₉MGN₆₃₁ at filter II (from the second 6-TM domain), whereas HsTPC2 has ₂₇₁TAN₂₇₃ and ₆₅₂VNN₆₅₄, respectively. To test if we can recapitulate the selectivity property of HsTPC2 in AtTPC1, we replaced the filter residues of AtTPC1 with the filter residues of HsTPC2 (i.e., a triple mutant with S265A in filter I and M629V/G630N in filter II). We named the mutant channel At2HsTPC2 and performed the same selectivity measurement as for the wild-type AtTPC1. These filter mutations do not change the gating of AtTPC1 (Fig. S3) but have a profound effect on channel selectivity. As shown in Fig. 3 A and B, although At2HsTPC2 is still equally permeable to Li⁺ and Na⁺, the channel becomes highly selective for Na⁺ over larger monovalent cations (K⁺, Rb⁺, and Cs⁺). With K⁺ in the bath, the reversal potential is −95.4 ± 5.8 mV, yielding a P_Na/P_K of about 41. To compare the selectivity between Na⁺ and Ca²⁺, we generated the At2HsTPC2 mutant (At2HsTPC2ΔCa_i) on the background of AtTPC1ΔCa_i to mitigate extracellular Ca²⁺ inhibition. With 15 mM Ca²⁺ in the bath solution and 150 mM Na⁺ in the pipette, At2HsTPC2ΔCa_i has a reversal potential of about −50 mV, indicating a higher permeability for Na⁺, with a P_Na/P_Ca of about 2.5 (Fig. 3 C and D). Thus, by mimicking the HsTPC2 filters, we are able to recapitulate the ion selectivity of HsTPC2 and convert AtTPC1 to a Na⁺-selective channel. The gain of Na⁺ selectivity in the AtTPC1 mutant is also accompanied by the loss of its Ca²⁺ selectivity.

Fig. S3.

Filter mutation has no impact on the voltage activation of AtTPC1. (A) Voltage activation of AtTPC1 and At2HsTPC2 mutants, with 150 mM Na⁺ in both the pipette and bath solutions. WT, wild type. (B) G/G_max-V curves of AtTPC1 and At2HsTPC2. Boltzmann fit yields V_1/2 = −28.7 mV, Z = 3.9 for AtTPC1 and V_1/2 = −27.7 mV, Z = 3.7 for At2HsTPC2, respectively (n ≥ 5).

Fig. 3.

Ion selectivity of At2HsTPC2. (A) Sample traces of whole-cell currents of At2HsTPC2 overexpressed in HEK293 cells with 150 mM Na⁺ in the pipette solution and 150 mM X⁺ (X = Li, Na, K, Rb, or Cs) in the bath solution. All recordings shown here were performed on a single patch. (B) I-V curves generated from the tail currents of the traces shown in A. (C) Sample traces of At2HsTPC2 currents with 150 mM Na⁺ in the pipette solution and 15 mM Ca²⁺ in the bath solution and the corresponding I-V curve. (D) Statistics of At2HsTPC2 reversal potentials (E_rev) measured in various bi-ionic conditions and the calculated relative permeability between Na⁺ and other cations. The reversal potential values are the mean ± SEM (n ≥ 5).

Residues on Filter II Define the Selectivity of TPC Channels.

To test whether all three filter mutations in At2HsTPC2 are necessary for achieving high Na⁺ selectivity, we also measured the selectivity of AtTPC1 with single and double mutations in the filter under bi-ionic conditions with 150 mM Na⁺ in the pipette and 150 mM K⁺ in the bath solution (Fig. S4). As summarized in Table 1, none of the single mutations can change the selectivity of AtTPC1; a double mutation in filter II, M629V/G630N, is necessary to convert AtTPC1 to a Na⁺-selective channel; S265A does not play a determinant role but can further enhance the Na⁺ selectivity of the channel. Thus, having Val and Asn together in filter II appears to be essential for Na⁺ selectivity in mammalian TPCs. To cross-validate this finding, we performed reversed mutagenesis analysis on HsTPC2 by swapping the three equivalent filter residues with the filter residues of AtTPC1, either individually or collectively (Fig. S5 and Table 1). Our results show that the A272S mutation in filter I has a subtle effect on the Na⁺ selectivity of HsTPC2, consistent with the Na⁺ selectivity of mouse TPC2, which has a sequence of ₂₅₅TSN₂₅₇ at filter I (27). Any mutations involving Val651 and Asn652, whether single or double, can significantly decrease the selectivity of the channel, and the triple mutant with the filter sequence equivalent to AtTPC1 has the lowest Na⁺ selectivity. Although we were not able to abolish Na⁺ selectivity in HsTPC2 completely, our mutagenesis results are qualitatively consistent with the results observed in the AtTPC1 mutants, confirming the necessity of having the Val/Asn pair in filter II to achieve high Na⁺ selectivity in HsTPC2. Furthermore, the loss of Na⁺ selectivity in the HsTPC2 triple mutant is also accompanied by more than a 10-fold increase in Ca²⁺ permeability with a P_Na/P_Ca of about 1.3, similar to what was observed in our AtTPC1 study (Fig. S5B and Table 1).

Fig. S4.

Selectivity of AtTPC1 with single- and double-filter mutations. (A) Partial sequence alignments of AtTPC1 and HsTPC2 at filter I and filter II regions. Arrows indicate the residues targeted for mutagenesis. (B) Representative I-V curves of AtTPC1 with single and double mutations in the filter under bi-ionic conditions with 150 mM Na⁺ in the pipette and 150 mM K⁺ in the bath solution. Values of reversal potential for each mutant are listed in Table 1.

Table 1.

Ion permeability of Na⁺, K⁺, and Ca²⁺ in AtTPC1 and HsTPC2 and their filter mutants

	AtTPC1					HsTPC2
Filter sequence	Construct	Naint(150)/Kext(150)		Naint(150)/Caext(15)*		Construct	Naext(150)/Kint(150)		Naint(150)/Caext(100)
I II		Erev, mV	PNa/PK	Erev, mV	PNa/PCa		Erev, mV	PNa/PK	Erev, mV	PNa/PCa
TSN MGN	WT	0 ± 2.8	1.0	0 ± 3.7	0.2	A272S/V652M/N653G	27.1 ± 3.2	2.9	0 ± 2.4	1.3
TAN MGN	S265A	−8.1 ± 3.1	1.4			V652M/N653G	42.0 ± 4.1	5.1
TAN MNN	—					V652M	49.2 ± 3.7	6.8
TAN VGN	—					N653G	48.0 ± 3.3	6.5
TSN VGN	M629V	0 ± 2.1	1.0			—
TSN MNN	G630N	0 ± 3.2	1.0			—
TSN VNN	M629V/G630N	−74.4 ± 4.2	18.0			A272S	75.3 ± 5.1	18.8
TAN VNN	S265A/M629V/ G630N(At2HsTPC2)	−95.4 ± 5.8	41.2	−50.3 ± 6.5	2.5	WT	81.4 ± 4.9	23.8	50.6 ± 3.8	16.8

E_rev, reverse potential; ext, extracellular; int, intracellular; WT, wild type.

^*To mitigate extracellular Ca²⁺ inhibition, constructs of AtTPC1ΔCa_i and At2HsTPC2ΔCa_i were used for P_Na/P_Ca measurements.

Fig. S5.

Effects of filter mutations on HsTPC2 selectivity. (A) Representative I-V curves of HsTPC2 filter mutants recorded on inside-out patches, with 150 mM Na⁺ in the pipette and 150 K⁺ in the bath solution. (B) I-V curve of the HsTPC2 triple mutant (A272S/V652M/N653G). The current was recorded with 100 mM Ca²⁺ in the pipette and 150 Na⁺ in the bath solution. All statistics of reversal potentials and calculated permeability ratios are summarized in Table 1.

Structural Comparison Between AtTPC1 and At2HsTPC2.

At2HsTPC2 exhibits similar ion selectivity properties as HsTPC2, providing us with a valid model system to reveal the structural basis that differentiates the selectivity between AtTPC1 and HsTPC2. We therefore crystallized the At2HsTPC2 channel and determined its structure by molecular replacement using the wild-type AtTPC1 as the search model (30, 32). The structure of At2HsTPC2 is virtually identical to the structure of AtTPC1 except in the filter region, which contains the three mutations. No obvious main chain conformational changes were observed between the wild-type and mutant AtTPC1, and all structural differences reflect changes in side-chain size at the mutated residues. The ion permeation pathway of AtTPC1 is asymmetrical at the filter region (Fig. 4 A and B). Its cross-section along the filter I pair is shorter but wider, with a minimum atom-to-atom distance of about 9 Å (Fig. 4C). The loss of the hydroxyl group, resulting from the S265A mutation, has little impact on the diagonal dimension between the filter I pair and, as expected, has little effect on channel selectivity.

Fig. 4.

Selectivity filter structure of At2HsTPC2. (A) Structure of the At2HsTPC2 ion conduction pore in top view. The pore helix and filter regions are colored in green and salmon, respectively. The putative sodium ion in the filter is shown as a purple sphere. (B) Zoom-in view of the ion selectivity region from the top (Left) and side (Right), with the front filter I removed for clarity. The surface-rendered model is calculated without Asn630 and Asn631 side chains to reveal the dimensions of the ion pathway in the absence of the two constriction points. The electron density (light blue mesh, contoured to 4σ) of the bound Na⁺ is calculated from the 2Fo-Fc map. Yellow dashed lines indicate H-bonds between N630 and neighboring backbone carbonyls from filter I and the ion coordination between N630 and Na⁺. (C) Superimposition of the filter structures between AtTPC1 (gray) and At2HsTPC2 (filter I in green and filter II in salmon). The pore radius diagrams for the cross-sections of filter I (Top) and filter II (Bottom) from both channels are calculated with the program HOLE (35). (D) Anomalous difference maps reveal the binding of Rb (blue mesh, contoured to 5σ) or Cs (magenta mesh, contoured to 5σ) in the At2HsTPC2 filter. (E) Local structural changes caused by the M629V mutation. Residues involved in structural changes are shown in blue for AtTPC1 and in green (filter I) and salmon (filter II) for At2HsTPC2. The yellow arrow indicates the ion pathway along the channel pore.

The filter II residues, on the other hand, build a longer but narrower part of the ion conduction pathway with a 4.8-Å-wide constriction point formed by the side chains of Asn631 from each subunit of the dimer. The G630N mutation adds another layer of constriction along the pathway just below Asn631, with a similar diagonal distance of about 4.8 Å. The Asn630 side chain in the At2HsTPC2 mutant is stabilized by hydrogen bonds with the backbone carbonyls of Thr264 and Ala265 (Fig. 4B). Centered between the twofold-related Asn630 residues is an electron density peak that likely comes from a Na⁺ ion or a water molecule. Because the distance between the side-chain carbonyl oxygen of Asn630 is too short for a hydrogen bond to a water molecule, about 2.4 Å, we assigned this central density peak as a Na⁺ ion (Fig. 4B). Due to the resolution limit, we did not observe any water molecules nearby. However, considering the coordination number and ligand chemistry commonly seen for Na⁺, this bound Na⁺ ion has to be partially hydrated. The two neighboring layers of Asn residues in the At2HsTPC2 filter also create a trapping site for a larger monovalent cation in between. As shown in Fig. 4D, structures from crystals soaked with Rb⁺ or Cs⁺ revealed a strong electron density peak with anomalous scattering sandwiched between the Asn630 and Asn631 residues, indicating a bound Rb⁺ or Cs⁺ ion. The bound Rb⁺ or Cs⁺ in the filter likely serves as a permeating blocker and reduces the Na⁺ current in At2HsTPC2. No equivalent Rb⁺ or Cs⁺ binding was observed in the wild-type AtTPC1.

The M629V mutation does not appear to have a direct impact on the size of the ion pathway. However, replacing the larger side chain of Met629 with the side chain of Val vacates a space beside Tyr656 from the S6 helix of the second 6-TM domain (IIS6), allowing the Tyr656 side chain to move closer to the filter. Consequently, Tyr656 in the M629V mutant engages in new packing interactions with two highly conserved filter residues: Its hydroxyl group forms a hydrogen bond with the Thr263 side chain, and its aromatic ring forms a π-stacking interaction with the aromatic ring of Trp632 (Fig. 4E). These newly formed interactions likely increase the stability of the filter.

Discussion

In this study, we were able to convert the nonselective AtTPC1 to a Na⁺-selective HsTPC2-like channel; specifically, we show that residues on filter II contribute to the difference in ion selectivity between plant and mammalian TPC channels. Structural comparison between the wild-type AtTPC1 and its Na⁺-selective mutant allowed us to propose a simple model to explain the structural mechanism underlining the selectivity difference. Because the At2HsTPC2 filter has an asymmetrical ion pathway with a wider dimension between the filter I pair but narrower between the filter II pair, the permeating ions likely pass through the filter in a partially hydrated form, particularly when crossing the two constriction points formed by the two pairs of symmetry-related asparagine residues (Asn630s and Asn631s). These asparagine side chains (likely the carbonyl oxygen atoms) can participate in ion coordination and stabilize the partially hydrated permeating ions. Commonly seen ion–oxygen coordination tends to have an optimal distance of about 2.4 Å for Na⁺ and 2.8 Å for K⁺ (33). With a distance of about 4.8 Å between the two asparagine residues at each constriction point, their side-chain carbonyl oxygen could provide optimal coordination to a central Na⁺ ion with a distance of about 2.4 Å, but too close for K⁺ or larger ions. In other words, by participating in ion coordination, these asparagine residues can function as a size sieve to sift out larger cations.

Interestingly, when Asn631 is replaced by alanine in At2HsTPC2, the channel remains highly selective for Na⁺ over K⁺ (Fig. S6), suggesting that the two layers of asparagine residues in the At2HsTPC2 filter are not equivalent in defining Na⁺ selectivity, with the Asn630 residues playing the determinant role. One explanation could be the environmental difference between Asn630 and Asn631. The Asn631 side chain forms the external entrance of the filter and, without spatial confinement, its side-chain carboxamide has the freedom to rotate away from the central axis, making the Asn631 constriction point more permissive for ions of different sizes. The Asn630 side chain, on the other hand, is located in the middle of the filter with a confined space and is stabilized by hydrogen-bonding interactions with nearby filter I residues. Thus, the carboxamide groups of the two symmetrical Asn630 residues are in a defined position with less mobility, allowing them to exert stringent size selection for the crossing ions.

Fig. S6.

Sample traces of At2HsTPC2_N631A currents with 150 mM Na⁺ in the pipette solution and 150 mM Na⁺ or K⁺ in the bath solution and the corresponding I-V curves. With K⁺ in the bath solution, At2HsTPC2_N631A has a reversal potential of about −85 mV, indicating high selectivity for Na⁺ over K⁺, with a P_Na/P_K of about 27.

Contrary to the G630N mutation, the M629V mutation does not appear to contribute to a size change in the filter, which raises the question about the necessity of having both the M629V and G630N to achieve high Na⁺ selectivity. The major effect of the M629V mutation is the formation of new protein packing interactions between Tyr656 on the IIS6 helix and two filter residues. We suspect that the newly formed protein packing surrounding the filter adds extra stability to the filter conformation in the mutant and constrains the dynamic motion of the filter, making it more restrictive for ion size. The necessity of having a stable filter to achieve high selectivity has also been observed in our previous study on K⁺ channels, in which weakening or disrupting protein packing interactions surrounding the filter can destabilize the filter conformation and compromise the channel’s selectivity (34).

Materials and Methods

The X-ray diffraction data collection and structure refinement statistics are listed in Table S1. The atomic coordinates and structure factors have been deposited in the Protein Data Bank under ID code 5TUA.

Table S1.

Data collection and structure refinement statistics

Dataset	At2HsTPC2	At2HsTPC2-Rb	At2HsTPC2-Cs
Data collection
Space group	C2221	C2221	C2221
Cell dimensions
a, b, c; Å	88.535, 156.371, 217.128	88.501, 159.196, 216.932	88.768, 159.385, 217.596
α, β, γ; °	90, 90, 90	90, 90, 90	90, 90, 90
Wavelength, Å	1.0332	0.8152	1.0332
Resolution, Å	50.00–3.30 (3.36–3.30)	50.00–3.30 (3.36–3.30)	50.00–3.65 (3.71–3.65)
Rmerge	0.084 (1.244)	0.072 (1.511)	0.081 (1.720)
CC1/2	(0.532)	(0.723)	(0.690)
I/σ	17.7 (0.7)	23.5 (0.6)	21.7 (0.7)
Completeness, %	90.3 (53.1)	96.9 (82.3)	98.5 (96.1)
Redundancy	5.7 (4.1)	6.7 (4.7)	6.7 (5.1)
Refinement
Resolution, Å	50–3.3
No. of reflections	14,619
Rwork/Rfree*	0.3103/0.3229
No. atoms	5,126
Protein	5,114
Ion	12
B-factors	65.22
Protein	65.11
Ion	112.76
Ramachandran, %
Favored/allowed	90.4/9.6
Rms deviations
Bond lengths, Å	0.009
Bond angles, °	0.866

Numbers in the parentheses show the values in the highest resolution shells.

^*R_free was calculated with 5% of the reflection data.

Detailed methods of protein purification, crystallization, structure determination, and electrophysiology are provided in SI Materials and Methods.

SI Materials and Methods

Protein Expression and Purification.

The Na⁺-selective mutant At2HsTPC2 was expressed and purified following the same procedures as the wild-type channel (30). Briefly, the mutant gene was cloned into pPICZ vector (Invitrogen) with a C-terminal EGFP-8× Histidine tag, and the plasmid was linearized, transformed into the Pichia pastoris SMD1163 (Invitrogen), and selected on agar plates containing 500 μg/mL Zeocin (Invitrogen). The transformed cells were grown in minimal glycerol medium + histidine (MGYH medium) to an OD₆₀₀ = 3.0 and then induced in minimal methanol medium + histidine (MMH medium) for 2 d at 28 °C.

The cells were harvested and lysed using an M-110P homogenizer (Microfluidics). The membrane fraction was isolated from whole-cell lysate, and the protein was extracted using 1% (wt/vol) N-dodecyl-β-d-maltopyranoside (Anatrace) at 4 °C for 3 h. The protein was then purified using Talon cobalt affinity resin (Clontech), followed by on-column thrombin digestion (Roche Diagnostics) at 4 °C overnight. The eluted protein was further purified by size exclusion chromatography (Superdex 200 column; GE Healthcare) in a buffer consisting of 20 mM Tris (pH 8.0), 150 mM NaCl, 1 mM CaCl₂, and 0.05% lauryl maltose neopentyl glycol (LMNG). The major fraction was collected and concentrated to ∼10 mg/mL for crystallization.

Crystallization and Data Collection.

The AtTPC1 mutant crystals were grown in similar conditions as the wild type by the conventional sitting drop vapor diffusion method. Typical crystals appeared within 1–3 d and grew to a full size (0.1 mm × 0.2 mm × 0.2 mm) within 1–2 wk in the condition: 24–26% PEG400, 100–150 mM BaCl₂, and 100 mM Hepes (pH 7.0) or 100 mM MES (pH 6.0). For cryoprotection, the PEG400 in the reservoir solution was increased to 38% and crystals were allowed to equilibrate for 12 h before freezing in liquid nitrogen. To reveal Rb⁺/Cs⁺ binding in At2HsTPC2, 3 μL of crystallization solution with an additional 100 mM RbCl or CsCl was directly added to the crystal drop, followed by 12-h equilibration with cryoprotection solution in the reservoir. X-ray diffraction data were collected at the Advanced Photon Source (23IDB, 23IDD, and 19ID) and Advanced Light Source (BL8.2.1 and BL8.2.2). To maximize the anomalous signal, data of Rb⁺-soaked crystals were collected near the rubidium absorption edge (λ = 0.8152 Å). All other data were collected at a wavelength around 1.0 Å (Table S1).

Structure Determination.

The diffraction data were processed and scaled with HKL2000 software (36). Due to the anisotropy of the diffraction, the autocorrection function was applied in HKL2000 during scaling. The AtTPC1 mutant structure was determined by molecular replacement using Phaser (37) with the wild-type AtTPC1 structure absent of the selectivity filter as the search model (Protein Data Bank ID code 5E1J), followed by repeated cycles of refinement using PHENIX (38) and manual adjustment in Coot (39). The geometries of the final structural model of At2HsTPC2 were verified using PROCHECK (40). For Rb⁺/Cs⁺-soaked crystals, anomalous difference Fourier maps were calculated with the phases from the native structure to verify the ion binding in the selectivity filters. Detailed data collection and refinement statistics are listed in Table S1. All of the structure figures were prepared with Pymol (41), and the selectivity filter radius was calculated with HOLE using the default parameters (35).

Electrophysiology.

Various AtTPC1 and HsTPC2 mutants were generated using the Quikchange Site-Directed Mutagenesis Method (Agilent) and confirmed by DNA sequencing. AtTPC1 and its mutants were cloned into pEGFP-C1 (Clontech) plasmid and expressed in HEK cells for functional assay as previously described (30). A total of 1–2 µg of the plasmid containing the GFP-tagged AtTPC1 or its mutant was transfected into HEK293 cells grown in a 35-mm tissue culture dish using Lipofectamine 2000 (Life Technology). At 24–48 h after transfection, cells were dissociated by trypsin treatment and kept in complete serum-containing medium until recording. Due to the low expression level of AtTPC1 on HEK plasma membrane, patch clamping in the whole-cell configuration was used to measure the channel activity. The standard bath solution contained (extracellular side) 145 mM sodium methanesulfonate (Na-MS), 5 mM NaCl, and 10 mM Hepes buffered with Tris (pH 7.4). The pipette solution contained (intracellular side) 150 mM Na-MS, 2.5 mM MgCl₂, 0.3 mM CaCl₂, and 10 mM Hepes buffered with Tris (pH 7.4). It is worth noting that it is necessary to have Ca²⁺ in the pipette solution to observe channel activity and it takes about 3–5 min before the whole-cell current reaches maximum and remains stable. Because most AtTPC1 channels expressed in HEK cells appear to be localized in organelles (endoplasmic reticulum and vesicles), we suspect that the introduction of higher Ca²⁺ into the cytosol from pipette solution triggers vesicle fusion and increases the number of channels on the plasma membrane. For voltage activation, the membrane was stepped from the holding potential (−70 mV) to various testing potentials (−100 mV to +100 mV) for 1 s and then returned to the holding potential. The peak tail currents were used to generate G/G_max-V curves (G = I/V, where G is conductance, I is current, and V is voltage). G_max was obtained from the peak tail current at a 100-mV testing potential, and V_1/2 and Z values were obtained from the fits of the data to the Boltzmann equation.

To measure the reversal potentials in bi-ionic conditions, the bath solution was changed to 145 mM X-MS, 5 mM XCl (X = Li, K, Rb, or Cs), and 10 mM Hepes buffered with Tris (pH 7.4). The membrane potential was stepped to +80 mV for 1 s to activate the channels fully and then switched to various testing potentials. The peak tail currents were recorded to generate an I-V curve for the determination of the reversal potential. To measure the Ca²⁺ permeability, three additional mutations at the Ca²⁺ inhibition site, Asp240Ala, Asp454Ala, and Glu528Ala, were introduced into both AtTPC1 and At2HsTPC2. The bath solution contained 15 mM calcium methanesulfonate (Ca-MS), 100 N-methyl-d-glucamine methanesulfonate (NMDG-MS), 5 NMDG chloride (NMDG-Cl), and 10 Hepes buffered with Tris (pH 7.4). The channel activation potential was increased to +100 mV. The sample traces and the I-V curves shown in each figure were obtained from current recordings on the same patch. The values of reversal potential are the mean ± SEM from at least five measurements from different patches.

The ion permeability ratios of AtTPC1 and its mutants were calculated with the following equations:

$${\mathrm{P}}_{\mathrm{X}}/{\mathrm{P}}_{\mathrm{Na}}={\left[\mathrm{Na}\right]}_{\mathrm{int}}\exp \left({\mathrm{E}}_{\mathrm{rev}}\mathrm{F}/\mathrm{RT}\right)/{\left[\mathrm{X}\right]}_{\mathrm{ext}},\left(\mathrm{X}=\mathrm{K},\mathrm{Li},\mathrm{Rb},\mathrm{or}\hspace{0.17em}\mathrm{Cs}\right),$$ [S1]

$${\mathrm{P}}_{\mathrm{Ca}}/{\mathrm{P}}_{\mathrm{Na}}=\left\{{\left[\mathrm{Na}\right]}_{\mathrm{int}}\exp \left({\mathrm{E}}_{\mathrm{rev}}\mathrm{F}/\mathrm{RT}\right)\left[\exp \left({\mathrm{E}}_{\mathrm{rev}}\mathrm{F}/\mathrm{RT}\right)+1\right]\right\}/\left\{4{\left[\mathrm{Ca}\right]}_{\mathrm{ext}}\right\},$$ [S2]

where E_rev is the reverse potential, F is Faraday’s constant, R is the gas constant, and T is the absolute temperature. The liquid junction potentials were corrected in our measurements.

The HsTPC2 channel used in our study contains two mutations, Leu11Ala and Leu12Ala, at the N-terminal targeting sequence, which allows the channel to be expressed and trafficked to the plasma membrane of the HEK293 cell for activity measurement using patch clamping (22, 28). All filter mutations in our experiments were generated on the background of this plasma membrane-targeting HsTPC2, cloned into pCS2 vectors with a C-terminal EGFP tag, and transfected into HEK293 cells. Because the channel is highly expressed on HEK293 plasma membrane, the channel activity was recorded using the inside-out patch configuration, which allows for easy ligand exchange from the intracellular side (bath solution). The standard bath solution contained (intracellular side) 145 mM K-MS, 5 mM KCl, and 10 mM Hepes buffered with Tris (pH 7.4), and the pipette solution contained (extracellular side) 145 mM Na-MS, 5 mM NaCl, and 10 mM Hepes buffered with Tris (pH 7.4). For the initial test of channel activation, the bath solution (intracellular side) was perfused with 10 μM PI(3,5)P₂ (water-soluble diC8 form; Echelon Biosciences) or 10 nM to 10 μM NAADP. For all selectivity measurements, the bath solution contained 10 μM PI(3,5)P₂. All I-V curves of HsTPC2 and its filter mutants were obtained using voltage pulses ramped from −100 to +100 mV over a duration of 800 ms. To determine the calcium permeability, the bath solution was changed to 145 mM Na-MS, 5 mM NaCl, and 10 mM Hepes buffered with Tris (pH 7.4), and the pipette solution contained 98 mM Ca-MS, 2 mM CaCl₂, and 10 mM Hepes buffered with Tris (pH 7.4).

The ion permeability ratios of HsTPC2 and its mutants were calculated with Eq. S2 for P_Ca/P_Na and Eq. S3 for P_K/P_Na:

$${\mathrm{P}}_{\mathrm{K}}/{\mathrm{P}}_{\mathrm{Na}}={\left[\mathrm{Na}\right]}_{\mathrm{ext}}\exp \left(\mathrm{-}{\mathrm{E}}_{\mathrm{rev}}\mathrm{F}/\mathrm{RT}\right)/{\left[\mathrm{K}\right]}_{\mathrm{int}}.$$ [S3]