Structural biology of cation channels important for lysosomal calcium release

Abstract

Calcium is one of the most important second messengers in cells. The uptake and release of calcium ions are conducted by channels and transporters. Inside a eukaryotic cell, calcium is stored in intracellular organelles including the endoplasmic reticulum (ER), mitochondrion, and lysosome. Lysosomes are acid membrane-bounded organelles serving as the crucial degradation and recycling center of the cell. Lysosomes involve in multiple important signaling events, including nutrient sensing, lipid metabolism, and trafficking. Hitherto, two lysosomal cation channel families have been suggested to function as calcium release channels, namely the Two-pore Channel (TPC) family, and the Transient Receptor Potential Channel Mucolipin (TRPML) family. Additionally, a few plasma membrane calcium channels have also been found in the lysosomal membrane under certain circumstances. In this review, we will discuss the structural mechanism of the cation channels that may be important for lysosomal calcium release, primarily focusing on the TPCs and TRPMLs.

1. Introduction

Ca²⁺ is an important second messenger in cells that regulates many central cellular activities, including cell mobility, enzyme activity, ion channel regulation, and gene expression^1-4. The cytosolic Ca²⁺ concentration is around 100 nM under resting conditions, which is thousands of fold lower than that of the extracellular environment¹. The intracellular concentration of Ca²⁺ is tightly regulated by various mechanisms, including its storage inside and release from intracellular organelles such as the endoplasmic reticulum (ER), mitochondrion, and lysosome⁵. Nicotinic acid adenine dinucleotide phosphate (NAADP), inositol 1,4,5-trisphosphate (IP3), and cyclic adenosine diphosphate ribose (cADPR) are the three major intracellular Ca²⁺ mobilizing messengers². Among them, NAADP is an endogenous metabolite of β-NADP and is regarded as the most potent Ca²⁺-releasing second messenger known to date^6,7. While IP3 and cADPR trigger the Ca²⁺ release from ER, NAADP specifically induces Ca²⁺ release from lysosomes. In animals, lysosomes are acidic organelles containing over 60 different types of hydrolases serving as the center for recycling and degradation of biological materials such as proteins and lipids. Furthermore, lysosomes participate in multiple key cellular activities including autophagy, endocytosis, exocytosis, nutrient sensing^8-12. Dysfunction of lysosomes could cause numerous diseases, such as type IV mucolipidosis (ML-IV), Parkinson's, Alzheimer's, and Huntington's diseases^12-14. In fungal and plant cells, vacuoles are the counterparts of lysosomes and serve as the main Ca²⁺ stores¹⁵.

Ca²⁺ homeostasis plays an important role in maintaining the normal function of lysosomes. The luminal Ca²⁺ concentration of a lysosome is approximately 0.5 mM, about five thousand-fold higher than that in the cytosol^16,17. The lysosomal Ca²⁺ uptake and release are mediated by multiple lysosomal membrane-localized channels and transporters. Two-pore channels (TPCs) and transient receptor potential mucolipin channels (TRPMLs) are the two major lysosome-specific cation channel families that are suggested to be important for lysosomal Ca²⁺ release^18,19. In addition, multiple plasma membrane channels have also been observed in lysosomal membranes and may conduct Ca²⁺ under certain conditions³. These include the purinergic receptor X4 (P2X4)^20,21, the transient receptor potential cation channel subfamily M member 2 (TRPM2)²², the transient receptor potential ankyrin 1 (TRPA1)²³, and the P/Q-type voltage-gated Ca²⁺ channel (Cav2.1)²⁴. In this review, we will primarily focus on the structures and functions of TPCs and TRPMLs, including the plant TPC1 from Arabidopsis thaliana (AtTPC1)^25,26, mammalian TPCs from Mus musculus and Homo sapiens (MmTPC1 and HsTPC2)^27,28, and mammalian TRPML1^29,30.

2. Two-pore channels (TPCs)

TPCs belong to the voltage-gated ion channel superfamily³¹ and are ubiquitously expressed in organelles of animals and plants^32,33. The plant TPC channel (TPC1) is localized to the vacuolar membrane^33,34. The Mammalian TPC family consists of three isoforms, TPC1, TPC2, and TPC3. TPC1 and 2 are localized to the endolysosomal membrane. They are widely expressed and have been extensively studied^18,35-42. TPC3 is only present in some mammals and is not expressed in humans. TPC3 has been localized to the lysosomal membrane in chicken and rabbits but the plasma membrane in zebrafish and its functions are not well defined^18,39,43. TPCs function as a homodimer with each subunit containing 12 transmembrane (TM) segments that can be divided into two homologous copies of the Shaker-like 6-TM domain. Thus, TPCs are believed to be evolutionary intermediates between tetrameric voltage-gated K⁺ channels and the four-domain single subunit voltage-gated Na⁺/Ca²⁺ channels⁴⁴. Plant TPC1 is responsible for generating the slow vacuolar (SV) current observed long before its molecular identification, therefore, plant TPC1 is also called SV channel^33,34,45. Plant TPC1 is involved in various cellular processes, such as germination and stomatal opening⁴⁶, jasmonate biosynthesis^47,48, and long-distance Ca²⁺ wave propagation induced by high salt concentrations⁴⁹. Mammalian TPC1 and TPC2 play critical roles in regulating the physiological functions of lysosomes^12,50,51. The functions of mammalian TPCs are implicated in various physiological processes including nutrient sensing³⁸, autophagy^52,53, lipid metabolism⁵⁴, pigmentation^55-57, blood vessel formation⁵⁸, and acrosome reaction in sperm⁵⁹. A recent study also indicated that TPCs play an important role in controlling Ebola virus trafficking in host cell⁶⁰. Surprisingly, no TPC mutation has yet been identified to be directly associated with severe human disease¹².

3. Structure and function of the plant TPC1 from Arabidopsis thaliana

TPC1 from Arabidopsis thaliana (AtTPC1) has been used as the model system for plant TPC and its structures in the closed state were first determined by protein crystallography^25,26. AtTPC1 is a non-selective cation channel, permeable to various monovalent and divalent cations with a slight preference for Ca^{2+ 61}. The channel is voltage-gated and its activation also requires cytosolic Ca²⁺. The luminal Ca²⁺, on the other hand, can inhibit the channel by modulating its voltage gating^62,63. Each AtTPC1 subunit contains two 6-TM domains (6-TM I and 6-TM II) that are linked by a cytosolic EF-hand domain (Fig. 1A). Two subunits assemble into a rectangle-shaped functional channel equivalent to a tetrameric voltage-gated channel (Fig. 1B). The S1-S4 segment of each 6-TM domain forms the voltage-sensing domain (VSD) at the peripheral of the channel whereas the S5-S6 segment forms the pore domain at the center with S6 lining the ion conduction pathway. Like most classical voltage-gated tetrameric cation channels, the TM region of AtTPC1 is domain swapped with the S1-S4 VSD of one 6-TM interacting with the S5-S6 pore domain of the other 6-TM. The cytosolic EF-hand domain contains two tandem EF-hand motifs and is positioned below the VSD1 from 6-TM I ^25,26.

Fig. 1. Structures of AtTPC1.

(A) Topology diagram of one AtTPC1 subunit. [Note: The N-terminal helix that directly interacts with the EF-hand domain was initially labeled as S0 in AtTPC1 but as an H1 helix in mammalian TPCs. To be consistent, it is labeled as H1 in this review.] (B) Cartoon representations of the AtTPC1 crystal structure (PDB code: 5E1J)²³. Domains are individually colored in one subunit as shown in (A) and the other subunit is colored in grey. Ca²⁺ ions are drawn as green spheres. (C) Ion conduction pore consisting of IS5–IS6 (PD 1 in cyan with front subunit removed for clarity) and IIS5–IIS6 (PD 2, magenta). Right inset: zoomed-in views of ion pathway at filter 1 and filter 2. Lower inset: zoomed-in views of the bundle crossing formed by IS6 and IIS6 pairs. Numbers are diagonal distances (in Å) at the narrowest constriction points. (D) The structure of AtTPC1 VSD2 in the resting state (upper panel with IIS1 removed for clarity) and the partial sequence alignment (lower panel) of IIS4 among AtTPC1, MmTPC1, and HsTPC2. The gating charge region is boxed. IIS4 is stabilized in a resting state by the luminal Ca²⁺ (green sphere). (E) Zoomed-in view of the EF-hand domain with EF-1 in a Ca²⁺-bound state and EF-2 in an apo state. The two EF-hand motifs are shown in orange; the H1 helix and IS6 are shown in cyan. (F) Zoomed-in view of the luminal Ca²⁺ binding site with the Ca²⁺ ion shown as a green sphere. Key residues are colored in yellow. (G) The structure of a partially open AtTPC1 (PDB code: 7FHO)⁷² with Ca²⁺-bound EF-hand domain and activated VSD2. (H) Zoomed-in view of the Ca²⁺-activated EF-hand domain with three bound Ca²⁺ ions (green spheres). The F2 helix breaks into two short helices in the Ca²⁺-bound EF-2 motif. (I) zoomed-in view of the activated VSD2 in AtTPC1 with IIS1 removed for clarity.

4. Ion conduction pore

Each pore domain (PD 1 and PD 2) contains the outer (S5) and inner (S6) helices and the two pore helices (P1, P2) between them, which is distinct from K⁺ channels but similar to Na⁺ and Ca²⁺ channels^64-66. In the closed pore of AtTPC1, the four pore-lining S6s (two IS6s and two IIS6s) form a bundle crossing at the cytosolic side with multiple narrow constriction points, preventing the passage of hydrated cations (Fig. 1C). Those constriction-forming residues include L301s and Y305s from IS6 with the narrowest diagonal distance (atom-to-atom) of about 5.3 Å, and V668s, L672s, and F676s from IIS6 with the narrowest diagonal distance of about 4.2 Å. Different from the K⁺ channel whose filter forms a long narrow ion pathway with four well-defined ion binding sites for dehydrated K⁺, AtTPC1 has a much shorter and wider selectivity filter formed by residues ₂₆₃TTS₂₆₅ in filter 1 and ₆₂₉MGN₆₃₁ in filter 2, allowing ions to cross the filter in a hydrated or partially hydrated state^25,26.

5. Voltage-sensing domain

The two VSDs (VSD1 and VSD2) in each AtTPC1 subunit adopt different structures. A few key features important for voltage-dependent gating in canonical voltage-gated channels are present in VSD2 but absent in VSD1 of AtTPC1. These include the presence of multiple gating charge arginines in S4 at every third position separated by two hydrophobic residues between them (₅₃₇RMLRLIR₅₄₃ in the VSD2 of AtTPC1)^25,26,67, the formation of a 310-helix in part of S4^65,68,69, and the charge transfer center formed of conserved acidic and aromatic residues in S2 and a conserved acidic residue in S3 (Y475 and E478 in S2 and D500 in S3 of AtTPC1) (Fig. 1D) ^60,70. Therefore, only VSD2 contributes to the voltage-dependent gating of AtTPC1^26,28. Because luminal Ca²⁺ or Ba²⁺ can inhibit the channel activity by stabilizing the voltage sensor in the resting state and Ba²⁺ was present in the crystallization conditions, the VSD2 in the AtTPC1 structure was captured in a resting state, providing crucial insight into the voltage-gating mechanism. In an activated VSD as observed in most structures of voltage-gated channels at the time, the S4 helix tends to be positioned with the side chain of its last gating charge arginine housed in the charge transfer center. In the resting VSD2 of AtTPC1, on contrary, it is the first gating charge arginine (R537) positioned in the gating charge transfer center and this resting state S4 is stabilized directly by a bound luminal divalent ion as described below^25,26 (Fig. 1D). The structural comparison between the activated and resting VSDs suggests that the S4 helix undergoes up and down translational movement in voltage gating and this movement is likely coupled to the pore-lining S6 helix through the linker helix between S4 and S5 (S4-S5 linker)^25,28.

6. Ca²⁺ activation and inhibition

Cytosolic Ca²⁺ potentiates voltage activation of AtTPC1 by binding to the EF-hand domain which contains two EF-hand motifs (EF-1 and 2). In the AtTPC1 structure, the EF-1 motif adopts a canonical Ca²⁺-bound EF-hand structure but EF-2 is in a Ca²⁺-free, apo state²⁵ (Fig. 1E). Interestingly, only the Ca²⁺-binding at EF-2 plays the determinant role in Ca²⁺ activation whereas Ca²⁺-binding at EF-1 is dispensable^25,71. Thus, the structure of the EF-hand domain likely represents a deactivated state, despite the presence of Ca²⁺ in the EF-1 motif. Another interesting feature of the EF-hand domain is that its E1 helix comes from the C-terminal part of the exceptionally long S6 helix (IS6) of 6-TM I, allowing the Ca²⁺-induced conformational change in the EF-hand domain to be directly coupled to the pore-lining IS6 upon Ca²⁺ activation.

Luminal Ca²⁺ inhibits AtTPC1 by stabilizing the VSD2 in the resting state, thereby shifting the voltage dependence towards positive membrane potentials. Chelated by D454 from IIS1, E528 from IIS4, and D240 from IS5 of the neighboring subunit, the luminal Ca²⁺ tethers the mobile voltage sensing IIS4 to the static IIS1 and the pore-forming IS5 of the neighboring subunit, and thereby hinders IIS4 movement in response to voltage change²⁵ (Fig. 1F). Consistent with structural observation, one of the key Ca²⁺ coordinating residues, D454, was previously identified to be important for luminal Ca²⁺ binding from a gain-of-function mutant fou2^47,48,63.

7. Recent update on AtTPC1 structures

Cryo-EM structures of AtTPC1 in a closed conformation with deactivated EF-hand domain and resting VSD2 and in a partially open conformation with Ca²⁺-bound EF-hand domain and activated VSD2 were determined recently, revealing the structural basis of voltage gating and cytosolic Ca²⁺ activation of AtTPC1⁷² (Fig. 1G-I). In the Ca²⁺-activated EF-hand domain, its conformational change upon Ca²⁺ binding is directly coupled to the pair of pore-lining IS6 helices, resulting in a wider distance between IS6s at the bundle crossing. In addition to EF-1 and EF-2 sites, a third Ca²⁺ site important for the Ca²⁺ activation of the channel was also identified in the structure of the Ca²⁺-bound EF-hand domain. Ca²⁺ at this site is chelated by the side chains of D39 and D43 and several backbone carbonyls at the C-terminal end of the H1 helix (Fig. 1H).

In the activated VSD2 structure, it is the third gating charge arginine (R543) positioned in the gating charge transfer center, confirming the previous suggestion that the IIS4 helix undergoes simple upward translational movement upon voltage activation (Fig. 1I). Intriguingly, in addition to the local IIS4 translation, the entire VSD2 also undergoes a lateral rotation movement around the central pore upon voltage activation. This lateral rotation movement of VSD2 also induces a global rotation of the EF-hand domain which in turn changes the accessibility of the two Ca²⁺ activation sites (EF-2 and the third newly identified Ca²⁺ site). In other words, the voltage activation of VSD2 appears to be a prerequisite for the Ca²⁺ activation at the EF-hand domain.

8. Structures and functions of Mammalian TPCs

Animal TPCs (TPC1 and TPC2) were initially suggested to mediate NAADP-dependent Ca²⁺ release from lysosomes^18,35,36. However, several recent electrophysiological recordings demonstrate that animal TPCs are Na⁺-selective channels that can be activated by lysosome-specific phosphatidylinositol 3,5-bisphosphate (PI(3,5)P₂) rather than NAADP^37,38. Debate still lingers on whether TPC1 and TPC2 are directly responsible for NAADP-induced lysosomal Ca²⁺ release. Several recent studies have identified small soluble NAADP-binding proteins that may directly mediate the NAADP-evoked, TPC-dependent Ca²⁺ signaling^73-75. The activation of TPC1 but not TPC2 is voltage-dependent. TPC3 is also voltage-gated and its voltage gating can be modulated by phosphoinositides such as PI(3,4)P₂ and PI(3,5)P₂, but not PI(4,5)P₂⁷⁶. Different from TPC1, TPC3 is only activated at extreme membrane depolarization and is probably responsible for generating or maintaining ultra-long action potentials^39,77,78. The structure of TPC3 from zebrafish has also been determined recently⁷⁸. As TPC3 is not present in humans and has been less studied, it will not be further discussed in this review. The following discussion about mammalian TPCs will be focused on the structures of mouse TPC1 (MmTPC1) and human TPC2 (HsTPC2).

Despite low sequence identity between them, mammalian TPCs share a similar overall structure to the plant TPC1^25-28 (Fig. 2A&B). Similar to AtTPC1, the two homologous 6-TM domains of mammalian TPCs are also connected by an EF-hand domain (Fig. 2A-B). However, the EF-hand domain lacks the essential Ca²⁺ chelating acidic residues and the mammalian TPC channels are not Ca²⁺ activated. One unique feature of mammalian TPC1 is that its C-terminal region adopts a horseshoe-shaped structure with four α-helices and two β-strands and wraps tightly around the EF-hand domain (Fig. 2C). It is unclear whether this unique structure feature at the C-terminus plays any functional role²⁸.

Fig. 2. Structures of mammalian TPCs.

(A) Topology diagram of one MmTPC1 subunit. The C-terminal domain is present in mammalian TPC1 but not TPC2. (B) Side and top views of MmTPC1 dimer structure in PI(3,5)P₂-bound open state. Domains in one subunit are individually colored as shown in (A) and the other subunit is colored in grey. PI(3,5)P2 is shown as yellow sticks. (C) Zoomed-in view of MmTPC1 cytosolic soluble region containing the EF-hand domain (green) and the C-terminal domain (salmon). (D) MmTPC1 ion conduction pore consisting of IS5–IS6 (PD 1 in red with front subunit removed for clarity) and IIS5–IIS6 (PD 2, blue). Right inset: zoomed-in views of MmTPC1 filter 2 with a partial sequence alignment of filter 2 region among AtTPC1, MmTPC1 and HsTPC2. (E) Zoomed-in views of the closed and open cytosolic gates formed by IS6 and IIS6 pairs. Numbers are diagonal distances (in Å) at the narrow constriction points. (F) Structure of MmTPC1 VSD 2 with S1 helix removed for clarity. (G) Structure of HsTPC2 VSD 2. (H) Zoomed-in view of the PI(3,5)P₂ binding site in MmTPC1. (I) Zoomed-in view of the PI(3,5)P₂ binding site in HsTPC2. PI(3,5)P₂ is shown in yellow, key residues are shown in white. (PDB codes for the structures used for making the figures are: 6C96 for the apo state MmTPC1; 6C9A for the PI(3,5)P₂ -bound MmTPC1; 6NQ0 for the PI(3,5)P₂ -bound HsTPC2)^25,26

9. Ion conduction pore

Similar to plant TPC1, each pore domain of the mammalian TPCs consists of S5, S6, and two pore helices. In a functional channel dimer, the ion conduction pore consists of two pairs of the pore domains, the IS5-IS6 pair from 6-TM I and the IIS5-IIS6 pair from 6-TM II. The structures of mammalian TPCs were determined in both open and closed conformations, allowing us to visualize the pore opening mechanics^27,28 (Fig. 2D). In the closed structure, the four pore-lining S6 helices form a bundle-crossing at the cytosolic side with narrow hydrophobic constrictions (L317 and F321 in IS6 and V684 and L688 in IIS6) that form the intracellular gate and prevent the passage of hydrated cations. In the open structure, the S6 helices undergo outward movement along with rotational motion, resulting in a much wider opening at the intracellular gate (Fig. 2E). This pore opening is driven by PI(3,5)P₂-binding at 6-TM I and the ensued conformational change at the pair of IS6 helices, as will be discussed later^27,28.

Mammalian TPCs have virtually identical sequences and similar structures at the selectivity filter^27,28. One key difference between mammalian TPCs and plant TPC1 that differentiates their selectivity property lies in the filter (filter 2) from the 2^nd 6-TM domain. As in the example of MmTPC1, its filter 2 has a sequence of ₆₄₇VNN₆₄₉, whereas the equivalent filter 2 in AtTPC1 is ₆₂₉MGN₆₃₁. The side chain of N648 generates a much narrower constriction at the center of the MmTPC1 filter²⁸ (Fig. 2D). The equivalent residue in plant TPC1 is a glycine (G630 in AtTPC1), yielding a wider ion pathway in the filter (Fig. 1C)²⁸. The formation of the narrow point along the filter pathway by a conserved asparagine renders the MmTPC1, as well as other mammalian TPCs, highly Na⁺ selective, and its mutation, can result in a complete loss of Na⁺ selectivity^27,28. Conversely, replacing G630 with Asn in AtTPC1 can convert the channel to be more Na⁺ selective⁶¹.

10. Voltage-sensing domain

Like AtTPC1, MmTPC1 is voltage-gated with its VSD2 contributing to the voltage activation of the channel²⁸. Interestingly, the VSD2 of MmTPC1 adopts an activated conformation in both the apo closed and PI(3,5)P₂-bound open structures (Fig. 2F), suggesting that the voltage sensor of MmTPC1 can be activated without opening the channel and PI(3,5)P₂ binding is the driving force for channel activation²⁸. Although not directly coupled to the pore opening, VSD2 activation appears to be required before PI(3,5)P₂-driven channel opening can occur²⁸. The VSD2 of MmTPC1 contains only two gating charge arginines with a sequence of ₅₄₀RPLQLLR₅₄₆ at its gating charge region where the middle Arg present in AtTPC1 becomes a Gln in MmTPC1^25,28 (Fig. 1D and Fig. 2F). In the structure of the activated MmTPC1 VSD2, it is the last Arg (R546) positioned in the gating charge transfer center. Interestingly, replacing the first arginine (R540) with Gln or Ile yields a mutant channel analogous to TPC2 which is no longer voltage-gated and can be activated by PI(3,5)P₂ alone³⁹. This R540 mutation likely stabilizes the VSD2 in an activated state. Replacing the last Arg (R546) with Gln, on the other hand, yields a channel that can barely be activated by voltage even with the presence of high concentration PI(3,5)P₂, as if the voltage sensor of the mutant is trapped in the resting state²⁸. Structural comparison between the activated VSD2 of MmTPC1 and the resting VSD2 of AtTPC1 reveals about 2-helical turn upward slide of the IIS4 upon voltage activation in TPC1^25,28.

Interestingly, the structure of VSD2 from HsTPC2 resembles that of MmTPC1 but the channel activation is not voltage dependent^27,28. The gating charge region of HsTPC2 VSD2 has a sequence of ₅₅₁IVFRFLR₅₅₇ with its first arginine placed in the middle position instead of the first position observed in MmTPC1 (Fig. 1D and Fig. 2G). The S4 of HsTPC2 VSD2 is positioned with its middle arginine (R554) occupying the gating charge transfer center²⁷. In most mammalian TPC2, there is an Ile instead of Arg at the beginning of the VSD2 gating charge region. When the first Ile (I551) is replaced with Arg in HsTPC2, the mutant channel becomes voltage-gated and requires both PI(3,5)P₂ and voltage for its activation²⁷. Thus, the presence or absence of the first Arg at the gating charge region appears to determine the voltage dependence of mammalian TPC channels.

11. PI(3,5)P₂ binding

In mammalian TPCs, PI(3,5)P₂ binds within the first 6-TM domain at the junction formed by IS3, IS4, and the IS4-S5 linker, and its inositol 1,3,5-trisphosphate (Ins(1,3,5)P3) head group involves in most of the ligand-protein interactions^27,28. As in the example of PI(3,5)P₂-bound MmTPC1 structure, the protein residues that participate in ligand interactions are predominantly basic (Fig. 2H). Some basic residues, including R220, R221, and R224 in the IS4-S5 linker and K331 on IS6, directly interact with the phosphate group on the C3 position of the lipid head group and play critical roles in PI(3,5)P₂ binding and channel activation²⁸. The PI(3,5)P₂ binding site in HsTPC2 is similar to that in MmTPC1²⁷. One notable difference between them is that most of the basic residues involved in ligand interactions, particularly those in the IS4-S5 linker, are predominantly lysines (K203, K204, and K207) in HsTPC2 rather than arginines (Fig. 2I). Mammalian TPCs have high ligand specificity and PI(4,5)P₂ cannot activate the channel. Based on the PI(3,5)P2-bound MmTPC1 structure, this can be explained by missing the C3-phosphate in PI(4,5)P2 lipid that is central for ligand-protein interactions, and the proximity of protein residues (Asn85 and Lys87 in MmTPC1) to the C4 hydroxyl group of the ligand, which occludes the space for the C4-phosphate and thereby sterically prevents PI(4,5)P₂ binding ^27,28. PI(3,5)P₂ binding activates the MmTPC1 channel by pulling IS6 towards the ligand-binding pocket through the interactions between K331 in IS6 and the phosphate groups of the ligand. This movement propagates to the other part of the IS6 helix, resulting in the outward dilation and rotation movements of those constriction-forming residues. To accommodate the conformational change at IS6 and open the pore, the other constriction-forming residues on IIS6 must undergo similar dilation and rotation movements²⁸. Based on the structural comparison between the open and closed MmTPC1, it was proposed that this concurrent movement of IIS6 in response to the PI(3,5)P₂-induced conformational change at IS6 can only occur when IIS4 of VSD2 is in the activated, up state²⁸.

12. Transient Receptor Potential Mucolipin (TRPML)

TRPMLs are Ca²⁺-permeable non-selective cation channels that belong to the large family of TRP ion channels⁷⁹. TRPML family consists of three isoforms, TRPML1, TRPML2, and TRPML3, which share high homology in amino acid sequence but are expressed in different tissues^80,81. TRPML1 is ubiquitously expressed in mammalian cells and is primarily localized to the lysosomal membrane ⁸². In contrast, the tissue expression of TRPML2 and TRPML3 is more restricted. TRPML2 is expressed in the thymus, spleen, and kidney ^83,84, and TRPML3 is expressed in the kidney, lung, and organs of the endocrine system⁸⁵. TRPMLs involve in multiple important cellular activities including lysosome trafficking, lipid accumulation, signaling transduction, and autophagy^8-11,86. Mutations in TRPML1 can directly cause the lysosomal storage disorder mucolipidosis type IV (MLIV), a neurodegenerative disease characterized by abnormal neurodevelopment, retinal degeneration, and iron deficiency anaemia^13,14,87,88. Gain-of-function mutations in mouse TRPML3 attribute to the varitint-waddler (Va) phenotype with symptoms including deafness, pigmentation defects, and circling behavior^19,89.

13. The overall structure of TRPML

Multiple TRPML structures have been determined in various states^29,30,90-95. TRPMLs are 6-TM tetrameric cation channels and are comprised of two structural components: the membrane-embedded S1-S6 transmembrane domain (TMD) and the canopy-like luminal linker domain atop the TMD^29,30,90 (Fig. 3A&B). Like most canonical 6-TM tetrameric channels, the TMD of TRPML is divided into two parts, the S1-S4 VSD and the S5-S6 pore domain, connected by the S4-S5 linker helix³⁰. The VSD resembles other TRP channels by having four straight helices tightly packed against each other but differs from the canonical VSD of voltage-gated channels by lacking the canonical voltage sensing arginines. The ion conduction pore is formed by S5, S6, and two pore helices (P1 and P2). Different from TRPV1 whose filter can undergo a remarkable conformational change upon ligand activation^77,78, the selectivity filter of TRPML is tightly packed with P1 and P2 helices and is unlikely to undergo major conformation change during gating^29,30,90. The luminal linker domain contains a sequence of about 200 amino acids between S1 and S2 (Fig. 3A&B). Structurally, each luminal linker domain is comprised of two long helices and seven strands; four linker domains form a square-shaped canopy with a central hole above the channel pore in a channel tetramer, which is a characteristic feature of group II TRP channels including TRPML and TRPP^79,96 (Fig. 3A-3C). In the polycystic kidney disease channel PKD2, a member of the TRPP family, this linker domain is referred to as the polycystin or top domain⁹⁶. TRPML possesses two unique cytosolic extensions: an N-terminal pre-S1 cytosolic domain containing two short helixes (H1 and H2) enriched in basic residues and a helix–turn–helix (H3–turn–H4) between S2 and S3. Another distinct feature of TRPMLs is the lack of TRP domain, a short peptide sequence around 25 amino acids immediately after S6 that is commonly seen in other TRP family channels, including the TRPC, V, M subfamilies^97,98.

Fig. 3. Structure of mammalian TRPML1.

(A) Topology diagram (left) and the structure (right) of a single HsTRPML1 subunit in the PI(3,5)P₂/ML-SA1-bound open state (PDB code: 6E7Z)⁸⁶. Domains are individually colored. (B) The structure of HsTRPML1 homotetramer in the PI(3,5)P₂/ML-SA1-bound open state, one subunit is colored as that shown in (A) and the other three are in grey. (C) Top view of the luminal linker domain. The three aspartates important for pH-dependent luminal Ca²⁺-inhibition are shown as yellow sticks. (D) The PI(3,5)P₂ binding site in HsTPC1. PI(3,5)P₂ is shown as yellow and red sticks. (E) ML-SA1 (magenta) binding site in HsTPC1. (F) ML-SI3 (blue) binding site in HsTPC1 (PDB code: 7MGL) ⁸⁸.

14. The modulation of TRPMLs by natural and synthetic ligands

TRPMLs can be activated or inhibited by various ligands, including natural and synthetic molecules. Endogenously, TRPMLs can be activated by lysosome-specific PI(3,5)P₂⁹⁹, but inhibited by PI(4,5)P₂ enriched in plasma membrane ¹⁰⁰. In addition, sphingomyelins have been shown to inhibit the activity of TRPML1, and such inhibition is reversed by treating with sphingomyelinase⁹. Multiple synthetic agonists and antagonists have been identified through drug screenings and development^9,101-104. ML-SA1 is the most well-characterized agonist that can robustly activate TRPML by itself as well as synergistically with PI(3,5)P₂^9,102. In addition to agonists, several antagonists with different chemical structures were identified, such as ML-SI1, 2, and 3. Some of these antagonists are capable of inhibiting both ML-SA1 and PI(3,5)P₂-elicited TRPML currents¹⁰³. Furthermore, two recent studies suggested that rapamycin and reactive oxygen species can both directly and specifically activate TRPML in an mTOR-independent manner^105,106.

The structures of TRPML1 in complex with PI(3,5)P₂ or PI(4,5)P₂ reveal that both lipids bind to the same N-terminal poly-basic region formed by cytosol-extruding H1 and H2, which is distal from the pore region⁹⁴ (Fig. 3D). As the structure of PI(3,5)P₂-bound TRPML1 (in the absence of other agonists) remains in the closed conformation likely because of the low open probability of channel upon PI(3,5)P₂ activation, it is unclear how PI(3,5)P₂ induces the conformational changes that open the channel^30,94. It was proposed that the binding of PI(3,5)P₂ and the ensued interaction between its C3 phosphate and Y355 in S3 promote the formation of a π-cation interaction between Y355 and R403 in S4, which in turn facilitates the pore opening⁹⁴. However, as those PIP₂-bound structures are of low resolution, the protein-ligand and protein-protein interactions at the lipid-binding site could not be accurately defined.

As a potent agonist, the ML-SA1-bound structure captured the TRPML1 channel in an open conformation²⁹. ML-SA1 resides in a hydrophobic pocket formed by several aromatic and hydrophobic residues from P1, S5, S6, and S6 of the neighboring subunit (Fig. 3E). The binding of ML-SA1 causes the pore-forming S5 and S6 to shift slightly away from the pore center, resulting in the opening of the lower gate. To accommodate the ML-SA1 binding and its induced lower gate-opening S6 movement, the filter region including the two pore helices and the selectivity filter also moves slightly away from the center, resulting in a subtle enlargement of the filter pathway ^29,94.

Recently, the antagonist ML-SI3-bound human TRPML1 structure was reported⁹² (Fig. 3F). The structure is in a closed conformation and is virtually identical to the apo structure. ML-SI3 binds at the same hydrophobic cavity as ML-SA1 does, explaining ML-SI3’s competitive inhibition of ML-SA1 activation of TRPML1. However, the two chemicals engage in a different set of protein-ligand interactions, rendering one (ML-SA1) to be agonist whereas the other (ML-SI3) to be antagonist.

15. The modulation of TRPMLs by pH and Ca²⁺

The TRPML activity is also regulated by pH and Ca²⁺ concentration in both lysosomal lumen and cytoplasm sides^80,107-109. Increasing luminal Ca²⁺ concentration inhibits the channel conductance in a dose-dependent manner. The luminal Ca²⁺-inhibition is pH-dependent as lowering the luminal pH can rescue the inhibition¹⁰⁷. Several aspartate residues, including D111, D114, and D115 in both human and mouse TRPML1, are positioned in proximity to the central ion pathway of the luminal linker domain and have been shown to contribute to the pH-dependent luminal Ca²⁺-inhibition of TRPML1⁹⁵. With an acidic luminal environment in the lysosome, the protonation of these acidic residues likely mitigates the Ca²⁺ inhibition of TRPML1 under physiological conditions. Contrary to the promotion of ion conduction by reducing Ca²⁺ inhibition in TRPML1, it has been shown that lower extracytosolic (luminal or extracellular) pH inhibits Na⁺ or Ca²⁺ conduction in TRPML3 and this pH-dependent regulation is mediated by a string of histidines in the luminal linker domain¹¹⁰. In addition, the linker domain has also been shown to be responsible for the inhibition of TRPML2 and TRPML3 by extracytosolic Na^{+ 111,112}. It was suggested that in normal cells, the lysosomal Na⁺ concentration is high enough to keep TRPML3 inactive. However, the channel can become active when the lysosome is damaged or the lysosomal Na⁺ and H⁺ gradient break down ^91,110,111. A study suggests that pathogen-induced pH neutralization of lysosome triggers Ca²⁺ release by activating TRPML3, which in turn facilitates lysosome exocytosis and expulsion of the phagocyted bacteria pathogens¹¹³.