The complete catalytic and gating cycles of TRPM2 chanzyme

The key to studying the coupling mechanism between enzymatic activity and channel gating of chanzymes is to capture the structures of all intermediate states in the catalytic and gating cycles. This is challenging if the enzymatic reaction is fast, such as the well-characterized nvTRPM2 chanzyme10, because the substrate will be depleted in the time frame required for preparing cryo-EM samples using state-of-the-art instrumentation, which typically takes at least 5 to 10 seconds. Phylogenetic analysis suggested that choanoflagellate Salpingoeca rosetta (sr) TRPM2 is also one of the earliest TRPM channels10 (Supplementary Fig. 1). Functional studies suggested that it is a chanzyme with slower ADPR hydrolysis kinetics than nvTRPM2 (ref. 10). To accurately measure the enzyme kinetics of srTRPM2, we developed a highly sensitive and accurate fluorescence-based enzyme kinetics assay capable of detecting ADPR at the nanomolar level (Methods and Extended Data Fig. 1a–e). Our data showed that the srTRPM2 is among the slowest known enzymes with the turnover number (kcat) an order of magnitude lower than those of nvTRPM2 (as detailed in the section ʻsrNUDT9-H modulates agonist availability via slow hydrolysisʼ), making it a promising candidate for capturing catalytic intermediates by performing time-resolved cryo-EM studies on time scales of seconds to minutes. Indeed, by precisely controlling the ligand condition and timing of ADPR hydrolysis during the preparation of cryo-EM samples (Methods), we determined a series of structural snapshots of srTRPM2 in the ADPR hydrolysis cycle, as well as various functional states in the channel gating cycle, including a total of 14 structures (Tables 1–5). These structures were determined at high resolutions (up to 1.97 Å) (Fig. 1a and Extended Data Figs. 2–4), which allowed us to unambiguously define the binding of substrates, lipids, and ions, as well as the closed and open conformations of the ion-conducting pore (Fig. 1b and Supplementary Video 1), thus revealing the complete catalytic and gating cycles of the TRPM2 chanzyme.

Table 1 Cryo-EM data collection, refinement and validation statisticsTable 2 Cryo-EM data collection, refinement and validation statistics, continuedTable 3 Cryo-EM data collection, refinement and validation statistics, continuedTable 4 Cryo-EM data collection, refinement and validation statistics, continuedTable 5 Cryo-EM data collection, refinement and validation statistics, continuedFig. 1: The complete gating and catalytic cycles of the srTRPM2 chanzyme.

a, The unsharpened composite cryo-EM map of srTRPM2–WT–Mg2+–AMP–R5P viewed parallel to the membrane. The density of a representative bound cholesterol molecule is shown in the box. b, The ion-conducting pore in the apo (left) and Mg2+–ADPR-bound open (middle) states viewed parallel to the membrane, with plots of the pore radius along the pore axis (right). The pore region (shown as a cartoon), residues (shown with sticks) forming the gate and the selectivity filter in two subunits are shown. Purple, green and red spheres define radii of >2.3, 1.2–2.3 and <1.2 Å, respectively. c, Example traces of inside-out patch recordings of srTRPM2 in the presence of EDTA–ADPR (n = 6, each lasting 60–120 s), Ca2+–ADPR (n = 6, each lasting 50–100 s) and Mg2+–ADPR (n = 6, each lasting 30–100 s), respectively, clamped at +60 mV. The channel open probability (Po) is reported as mean ± s.d. d, Structural snapshots of the gating and catalytic cycles of the srTRPM2 chanzyme. Of note, the intermediate state II of the ADPR hydrolysis was approximated by the srTRPM2 structure bound with ADPR. The only difference between this structure and the true intermediate state II is the lack of Mg2+ binding in the MgTMD and MgMHR sites. However, our analysis indicates that the absence of Mg2+ binding in these sites does not cause major conformational changes in the MHR1/2 and NUDT9-H domains. Therefore, we consider this structure to be a reasonable approximation of the intermediate state II.

The TRPM2 chanzyme is regulated by five ligand binding sites

In the presence of ADPR, srTRPM2 can be activated by Ca2+ or Mg2+, yielding similar voltage-independent currents with high channel open probabilities (Fig. 1c and Extended Data Fig. 5), whereas ADPR hydrolysis in NUDT9-H occurs only in the presence of Mg2+ but not Ca2+ (refs. 10,21). By contrast, vertebrate TRPM2 opens only in the presence of Ca2+ with a lower open probability22, and cannot hydrolyze ADPR17.

The structures of srTRPM2 exhibit a characteristic TRPM architecture (Figs. 1d and 2a), from top to bottom, a transmembrane domain (TMD) layer containing the ion-conducting pore, a signal transduction layer consisting of MHR3/4 domain and the C-terminal rib helix, and an ADPR-sensing layer consisting of the N-terminal MHR1/2 domain and the C-terminal NUDT9-H domain. In srTRPM2, the NUDT9-H domain adopts a ‘vertical’ pose, as opposed to the ‘horizontal’ pose observed in human and zebrafish TRPM2 (Extended Data Fig. 6a). This difference accounts for the markedly longer but slimmer shape of srTRPM2 in comparison to its vertebrate orthologs and, more importantly, is closely linked to the molecular evolution of the NUDT9-H domain, which has let to distinct functions of NUDT9-H in early and advanced species TRPM2 (detailed in the last two sections).

Fig. 2: Ligand binding sites.

a, The overall structure of srTRPM2–WT–Mg2+–ADPR in surface representation viewed parallel to the membrane. One subunit is highlighted, with the four domains colored differently and the ligands shown as spheres. b, The ADPRN site. The ADPRN molecule and interacting residues are shown as sticks. Polar interactions are indicated by thick black bars. c,d, The ADPRC site in srTRPM2 (c) and hsTRPM2–WT–Ca2+–ADPR (PDB ID 6PUS) (d). The ADPR molecules and interacting residues are shown as sticks. The Mg2+ cofactors and water molecules are shown as spheres. Note that the orientation of ADPRC molecules in c and d is reversed, despite the two NUDT9-H domains have the same orientation. e,f, The MgTMD site (e) and MgMHR site (f). The Mg2+ cations and water molecules are shown as spheres. Polar interactions are indicated by thick black bars.

Obvious conformational differences were observed between the srTRPM2 structures in ligand-free and ligand-bound states, which is mainly manifested by the rearrangement of the MHR1/2 and NUDT9-H domains (Fig. 1d). We identified five ligand binding sites, including two sets of ADPR-binding sites (one of which also binds the hydrolysis products AMP and R5P) and three sets of cation binding sites (Figs. 1d and 2a). Some of these ligand binding sites are conserved among TRPM2 orthologs, while others are not.

The agonist ADPR was found in both the N-terminal MHR1/2 domain of the channel module (ADPRN) and C-terminal NUDT9-H enzyme module (ADPRC) domains (Fig. 2a). The ADPR binding to the MHR1/2 domain is conserved across all the TRPM2 orthologs (Extended Data Fig. 6b). Two conserved residues, F268 for π–π stacking with the adenine group of ADPR and R275 for hydrogen bonding with the ribose group of ADPR, are essential for ADPR binding in srTRPM2 (Fig. 2b). Replacement of either of these residues with alanine rendered srTRPM2 insensitive to ADPR (Extended Data Fig. 6c,d). These data confirmed that the function of the MHR1/2 domain in activating TRPM2 is preserved from early to advanced species13,14,15,16. By contrast, the ADPR binding to the NUDT9-H domain differs between TRPM2 in early and advanced species. Specifically, ADPR in srNUDT9-H adopts an entirely different conformation from that in human NUDT9-H (Protein Data Bank (PDB) ID 6PUS), and is coordinated by cations (3× Mg2+ or 2× Ca2+) that are absent in human NUDT9-H (Fig. 2c,d).

The cation binding site in the S1-S4 domain of TMD is conserved in all the Ca2+-activated TRPM channels (TRPM2/4/5) and is required for channel activation (Extended Data Fig. 4d)13,14,23,24,25,26,27. However, only srTRPM2 can also be activated by binding of Mg2+ at this site (Figs. 1c and 2e). A cation binding site is found between the MHR3/4 domain and the Rib helix of each subunit, termed MgMHR (Fig. 2f). This site does not distinguish between Ca2+ or Mg2+ either (Extended Data Fig. 7a,b) and is unique to srTRPM2 as the key residue, E1114, is not conserved (Supplementary Fig. 2). Replacing E1114 with alanine eliminated cation binding at MgMHR, resulting in a reduced rotational movement of the intracellular domain around the symmetry axis when ligands were bound, as compared to the wild-type (WT) protein (Extended Data Fig. 7c). The E1114A mutation had no obvious effect on macroscopic currents. However, single-channel analysis showed a small decrease in channel open probability and a small increase in channel close rate (Extended Data Fig. 7d,e and Supplementary Fig. 3a–c), revealing a regulatory role of this cation binding site on channel gating.

Snapshots of ADPR hydrolysis

In the presence of Mg2+, srNUDT9-H is catalytically active, which allowed us to capture five distinct states of the ADPR hydrolysis cycle (Fig. 3a). These structures reveal the detailed molecular mechanisms underlying ADPR hydrolysis. Before the onset of hydrolysis (substrate-bound state), both MHR1/2 and NUDT9-H domains are occupied by ADPR. Within NUDT9-H, the adenine group of ADPR is sandwiched by the indole group of W1264 and the phenyl group of F1372 via π–π stacking (Fig. 2c). The double alanine mutation of the corresponding residues in NUDT5—another ADPR hydrolase from the Nudix hydrolase family like NUDT9 (ref. 28)—reduced its catalytic efficiency by four orders of magnitude29. The terminal ribose group of ADPR forms multiple hydrogen bonds with the side chains of D1330, D1426 and R1428 (Fig. 2c). Three Mg2+ cofactors (Mg1, Mg2 and Mg3) are bound between the two phosphate groups of ADPR and two acidic residues on the Nudix helix, E1386 and E1390 (Fig. 2c). A glutamate residue corresponding to E1386 is conserved in NUDT5, NUDT9 and TRPM2 chanzymes, but is replaced by isoleucine in vertebrate TRPM2 channels (Supplementary Fig. 2); replacing the corresponding residue in NUDT5 or NUDT9 with glutamine or isoleucine, respectively, reduced the catalytic efficiency by four or three orders of magnitude29,30, indicating its key role in metal binding and ADPR hydrolysis. The β-phosphate (Pβ) of ADPR forms salt bridge interactions with the conserved R1360 and R1428 (Fig. 2c).

Fig. 3: Snapshots of ADPR hydrolysis.

a, Cartoon representation of the structural snapshots during ADPR hydrolysis. Note that we were unable to obtain the structure of a bona fide intermediate state II. However, the structure in the presence of EDTA–ADPR provides a reasonable approximation. In this structure, the ADPRN site in the channel module is occupied, while the enzyme module is unoccupied and adopts an apo-like conformation. The only difference between this structure and a bona fide intermediate state II is the lack of MgTMD and MgMHR (Fig. 1d), two cations that cause only minor conformational changes. Note that ADPR binding to the N- and C-terminal sites occurs independently, and each is in a dynamic equilibrium with the environmental ADPR pool. b, Cryo-EM densities of the intact ADPRC, Mg2+ cofactors and putative water molecules in the substrate-bound NUDT9-H from srTRPM2–WT–Mg2+–ADPR–10s. The ADPR molecule is shown as sticks. The Mg2+ cofactors and putative water molecules are shown as spheres. c, Key residues that coordinate the Mg2+ cofactors and nucleophilic water molecule, as well as the two arginine residues that interacts with Pβ of ADPR. The arrow indicates the nucleophilic attack of the water molecule on Pα of ADPR for hydrolysis. d,e, Cryo-EM densities of the hydrolysis products (AMP and R5P) and Mg2+ cofactors in NUDT9-H of intermediate state I from srTRPM2–WT–Mg2+–ADPR–10s (d) and intermediate state III from srTRPM2–WT–Mg2+–AMP–R5P (e). The ADPR molecule is shown with sticks. The Mg2+ cofactors and water molecules are shown as spheres. Note that in intermediate state III, AMP adopts a different conformation from that in intermediate state I, with its phosphate group dissociated from R5P, rendering it flexible, as evidenced by the lack of resolved density of AMP’s phosphate group. This change leads to the loss of Mg1, which interacts with AMP’s phosphate group in intermediate state I, and the movement of Mg2 to fill the position of AMP’s phosphate group in intermediate state I.

The extensive interactions between ADPR and the NUDT9-H domain strategically anchor both termini of the ADPR molecule, exposing only its α-phosphate (Pα) to solvent and making it readily accessible for nucleophilic attack (Fig. 2c). Indeed, we observed a putative water molecule bridging Mg2 and Mg3, as previously found in the crystal structure of NUDT5 (Fig. 3b and Extended Data Fig. 4c)29. On activation by Mg2+ ions, the water molecule is deprotonated by the catalytic base D1460 located in a loop close to the Nudix helix, and is poised for nucleophilic attack on the phosphorus atom of Pα (Fig. 3c). Mutations in the equivalent residues of NUDT5 and NUDT9 have been shown to reduce the catalytic efficiency of the enzymes by two orders of magnitude relative to the WT form29,30. The water molecule breaks down the phosphoester bond, resulting in the production of AMP and R5P, as observed in intermediate state I (Fig. 3a,d and Extended Data Fig. 4c). The release of the hydrolysis products triggers a conformational change of NUDT9-H in preparation for the next catalytic cycle, as seen in intermediate state II (Fig. 3a). It is noteworthy that the MHR1/2 domain remains bound to an intact ADPR in all structures except for one, which was determined using protein samples incubated with ADPR for an extended period. In this instance, the ADPR molecules are presumably depleted, leaving the MHR1/2 domain unoccupied while the NUDT9-H domain bound to the hydrolysis products (intermediate state III; Fig. 3a,e and Extended Data Fig. 4c). Following release of these hydrolysis products, the protein returns to the substrate-free state (Fig. 3a).

Although both Ca2+ and Mg2+ activate the channel, the Mg2+-dependent ADPR hydrolysis of srTRPM2 is inhibited by Ca2+ (Extended Data Fig. 1f). This is because in the Mg2+-bound active site, the three Mg2+ cofactors are closely coordinated around the α-phosphate (Fig. 3c), likely inducing a favorable geometric distortion around the phosphorus atom for hydrolysis31. In contrast, in the Ca2+-bound active site, only two Ca2+ occupy the equivalent positions of Mg2 and Mg3 (Extended Data Fig. 8a,b), which presumably does not induce a sufficient genometric distortion around the α-phosphorus atom necessary for hydrolysis. Therefore, Ca2+ inhibits the enzyme by competing with Mg2+ for the metal binding sites in NUDT9-H.

The srNUDT9-H modulates channel activity via tetramerization

In the apo state, the ADPR-sensing layer, particularly the NUDT9-H enzyme module of srTRPM2, is flexible due to a lack of interface with the channel module (Fig. 4a and Extended Data Fig. 9a). This layer underwent a marked rearrangement on binding of ADPR (or its hydrolysis products) and divalent cations, and is stabilized by the formation of extensive interactions between NUDT9-H and the rest of the protein (Extended Data Fig. 9b,c). We noticed that the stabilization of NUDT9-H was caused, at least in part, by a peptide recruited from the adjacent subunit (Fig. 4b). This intersubunit binding was confirmed by cross-linking experiments (Fig. 4c,d). This peptide, which is situated in the linker region connecting the C-terminal pole helix and NUDT9-H, is composed of 11 residues and forms a short helix. As it buckles the NUDT9-H from the adjacent subunit and tightens all four NUDT9-H domains together, we named it a buckle helix. The sequence of the buckle helix is not conserved (Supplementary Fig. 2), and the corresponding peptide is disordered in the available vertebrate TRPM2 structures13,14,26,27,32.

Fig. 4: The srNUDT9-H modulates channel activity via cation/ADPR-induced tetramerization.

a,b, Organization of the NUDT9-H layer in the apo state (a) and substrate-bound state (b), viewed from cytoplasmic side. The NUDT9-H domains are shown in surface representation, with one subunit highlighted in red. The center-of-mass distances of adjacent NUDT9-H domains, as well as the rotation of the NUDT9-H layer in the substrate-bound state relative to the apo state, are indicated. c, Interactions of the buckle helix (red) with NUDT9-H from adjacent subunit (white). The buckle helix and NUDT9-H are shown in a cartoon representation. The Cα atoms of residues selected for disulfide cross-linking experiments are shown as spheres, with dashed lines marking Cα distances between residue pairs that potentially form disulfide bonds if mutated to cysteine. The IDs of residue pairs match those in d. d, Assessing site-directed disulfide cross-linking assessed through in-gel fluorescence signals. GFP-tagged srTRPM2–WT and single or double cysteine mutants of selected residues shown in d were analyzed by nonreducing SDS–PAGE for three times. The cross-linked tetramer bands were consistently observed. e, Example traces of inside-out patch recordings of srTRPM2–∆NUDT9-H (with both buckle helix and NUDT9-H domain were truncated) (n = 6, each lasting 50–100 s), srTRPM2–BH2A (with all buckle helix residues mutated to alanines) (n = 6, each lasting 90–140 s) and srTRPM2–∆BH (with the buckle helix truncated) (n = 6, each lasting 30–110 s) clamped at +60 mV.

Source data

The tetramerization of the NUDT9-H domain through the buckle helix remodels the interface between the enzymatic module and the channel module (Extended Data Fig. 9b,c). This raises the question of whether the NUDT9-H enzyme module plays a direct role in channel gating of srTRPM2 on ADPR binding. To this end, we generated a truncated construct by removing NUDT9-H, srTRPM2–ΔNUDT9-H. Despite reduced protein expression level at the plasma membrane, the truncation construct still responded to ADPR and Ca2+, generating currents with similar characteristics to the WT (Extended Data Fig. 10a–c). This is distinct to human TRPM2 (hsTRPM2), in which removal of NUDT9-H abolished ADPR-induced current13. Single-channel analysis of srTRPM2–ΔNUDT9-H revealed channel gating kinetics similar to the WT, albeit with a reduced open rate (Extended Data Fig. 10d and Supplementary Fig. 4a–c). Notably, this construct exhibited long-lived closed states between bursts of channel opening, a pattern not observed in the WT protein (Figs. 1c and 4e). In agreement with the functional data, the overall structure of srTRPM2–ΔNUDT9-H closely resembles the channel module of srTRPM2, but with some minor conformational changes that may have contributed to the subtle differences in gating kinetics (Extended Data Fig. 10e–h). Our result indicates that the NUDT9-H domain is dispensable for the channel activation of srTRPM2, which is consistent with the nvTRPM2 chanzyme15, but plays a role in the regulation of the channel activity.

We then set out to understand how NUDT9-H directly modulates channel activity. When srTRPM2 was incubated with ADPR and Mg2+ for a long enough time before cryo-EM sample preparation, ADPR was completely hydrolyzed in the structure. We found that the MHR1/2 domain became ligand-free and adopted the same conformation as in the apo structure. On the other hand, NUDT9-H was occupied by the hydrolysis products AMP, R5P and Mg2+ (Extended Data Fig. 8d,e), but adopted almost the same conformation as when it was complexed with intact ADPR and Mg2+. Therefore, by comparing this structure with the apo structure, we can approximately analyze how the binding of ADPR and Mg2+ to NUDT9-H affects the conformation of the channel module of the chanzyme. In the apo structure, the NUDT9-H enzyme module formed only one interface with the MHR1/2 domain of the same subunit and was thus flexible and unlikely to have a major impact on the conformation of the channel module (Fig. 4a and Extended Data Fig. 9b). By contrast, in the presence of ligands, the NUDT9-H layer underwent tetramerization via the buckle helix, causing a clockwise rotation viewed from the intracellular side and a contraction of the NUDT9-H layer (Fig. 4b). This created multiple additional interfaces of the NUDT9-H enzyme module with cognate and adjacent subunits (Extended Data Fig. 9c), ultimately leading to a rotational movement of the signal transduction layer—the MHR3/4 domain and the rib helix (Extended Data Fig. 8f). Our previous studies on vertebrate TRPM2 have shown that rotational movement of the MHR3/4 domain is a key factor in transducing the signal from agonist binding to channel gating13,14. We therefore propose that binding of ADPR (or its hydrolysis products) to the NUDT9-H enzyme module leads to buckle helix-mediated self-tetramerization, which facilitates the movement of the signal transduction layer, thereby regulating channel function. Consistent with this notion, replacing residues of the buckle helix with alanine (srTRPM2–BH2A) or truncating the buckle helix (srTRPM2–ΔBH), both of which are expected to weaken the tetramerization of NUDT9-H, resulted in the occurrence of long-lived closed states, similar to those observed in srTRPM2–ΔNUDT9-H (Fig. 4e).

srNUDT9-H modulates agonist availability via slow hydrolysis

As an ADPR hydrolase, the NUDT9-H domain of srTRPM2 also couples indirectly to channel gating by regulating the local concentration of agonist. This indirect coupling, however, requires that the kinetics of the hydrolysis must be substantially slower than the kinetics of the channel activation, otherwise the agonist would be depleted before it has a chance to induce channel opening. Indeed, srTRPM2 is among the slowest enzymes with a kcat of 3 (Fig. 5a,b), similar to RuBisCo involved in carbon fixation33.

Fig. 5: The srNUDT9-H is a slow ADPR hydrolase.

a, Plot of the rate of AMP formation as a function of substrate concentration, representing the rate of ADPR hydrolysis by the WT srTRPM2. Each enzymatic reaction was performed three times independently and the converted ε-AMP was measured by fluorescence detection. The solid line indicates the fit to the Michaelis–Menten equation (n = 3) with Michaelis constant (KM) indicated, and the circles and error bars represent mean ± s.d. b, Estimated kcat values for WT srTRPM2 (n = 3), isolated srNUDT9-H (n = 3), srTRPM2–∆BH (n = 3) and srTRPM2–BH2A (n = 3) from the experiments in a. The bars and error bars represent mean ± s.d. and the P value was derived from two-tailed analysis. c, Superposition of NUDT9-H in the apo (predicted by AlphaFold) and ligand-bound states using the cap region of NUDT9-H. The NUDT9-H and buckle helix are shown in a cartoon representation. ADRP, R1360 and R1428 are shown as sticks. The movement of the core region of NUDT9-H on ligand release is manifested by an increased Cα distance between R1360 and R1428, as well as the rotation of the Nudix helix. d, Cytoplasmic view of the 3D classes in the cryo-EM data of ADPR being completely hydrolyzed, showing successive steps of detetramerization of the NUDT9-H domain. The unbuckled NUDT9-H domains disassociate from other subunits, thus becoming flexible and less well defined in the cryo-EM map. By contrast, the buckled NUDT9-H domains as well as the channel module remains well defined, and the channel module maintains C4 symmetry.

Source data

To understand how the srTRPM2 chanzyme has evolved an extremely slow enzyme module to accommodate the function of the channel module, we compared the active site in the ligand-free state and when bound to ADPR (or the hydrolysis products). We expect different conformations, as the free enthalpy of the hydrolysis reaction provides the driving force to overcome the stability of the protein and disassemble the active site, allowing the products to leave. Indeed, an AlphaFold-predicted34 model of the srNUDT9-H domain in the ligand-free form—which is otherwise challenging to obtain in high resolution due to its high flexibility in the full-length protein—has an open clamshell conformation, in contrast to the closed clamshell conformation bound to ADPR (or the hydrolysis products) in the cryo-EM structure, as indicated by the change in distance between the arginine pair R1360–R1428 involved in ADPR binding (Fig. 5c). Structural comparison further revealed that the opening of the clamshell requires the movement of the Nudix helix that forms part of the active site (Fig. 5c, black arrow). While the Nudix helix is free to move in an independent monomeric NUDT9-H domain, it is constrained in the srTRPM2 chanzyme when loaded with ADPR, as the NUDT9-H domains tetramerize via the buckle helix, which is inserted into the cleft between the Nudix helix and a nearby helix (Fig. 5c), preventing the movement of the Nudix helix to release the products. To release hydrolysis products, the buckle helix must disassociate from the adjacent NUDT9-H domain, an additional step that slows down the enzymatic reaction. In support of this idea, successive steps of detetramerization of the NUDT9-H domain were observed in the cryo-EM data of ADPR being completely hydrolyzed (Fig. 5d). Our data suggest that buckle helix-mediated tetramerization exerts an inhibitory effect on the enzyme module. This is consistent with measurements of enzyme kinetics, which showed that the kcat of the full-length srTRPM2 is more than twofold lower than that of the isolated srNUDT9-H domain. Furthermore, as expected, replacing residues of the buckle helix with alanine (srTRPM2–BH2A) or truncating the buckle helix (srTRPM2–ΔBH) resulted in a kcat between those of the full-length srTRPM2 and the isolated srNUDT9-H domain (Fig. 5b).

In summary, the enzymatic NUDT9-H domain modulates the channel gating in two different ways. Directly, it up-regulates the channel activation by facilitating signal transduction from the ADPR-sensing layer to the channel gate. Indirectly, it down-regulates the channel activation by reducing the available ADPR through its slow hydrolase activity.

Loss-of-function of NUDT9-H as hydrolase in vertebrates

To understand the molecular basis of how the NUDT9-H domain loses ADPR hydrolase activity in advanced species, we analyzed the amino acid sequences and performed structural comparisons of srTRPM2 and hsTRPM2, focusing on two sets of important residues responsible for ADPR hydrolysis and binding, respectively.

Residues responsible for ADPR hydrolysis include those that coordinate the three Mg2+ cofactors and the catalytic water molecule in the active site, which vary considerably between early and advanced species10 (Supplementary Fig. 2). For instance, D1460, responsible for deprotonating the catalytic water molecule in srTRPM2, becomes S1469 in hsTRPM2, and E1386, responsible for coordinating the catalytic Mg2, becomes I1405 in hsTRPM2. As a result, human NUDT9-H can no longer bind Mg2+ and activate the nucleophilic water, becoming catalytically inactive