Structural basis for agonist and heat activation of nociceptor TRPM3

Abstract

Detecting noxious heat is vital for survival, triggering protective pain responses. The TRPM3 channel is a key nociceptor and a promising therapeutic target for pain and neurological disorders. Here we show that the rabbit TRPM3 is intrinsically dynamic, with its intracellular domain (ICD) sampling both resting and activated states, but favoring the resting state in the absence of stimulation. We reveal that heat and the synthetic agonist CIM0216 shift the equilibrium toward activation by inducing a similar ICD rearrangement. Mutations that facilitate ICD movement enhance sensitivity to both thermal and chemical stimuli, underscoring the central role of the ICD in channel gating. We also show that the antagonist primidone binds the same site as CIM0216 in the S1–S4 domain but inhibits channel activation. This study provides a structural framework for a mechanistic understanding of thermal and chemical gating of TRPM3 and for guiding the rational design of TRPM3-targeted analgesics and neurotherapeutics.

TRPM3 is an ion channel that helps the body sense heat and contributes to pain. The authors show that both heat and small chemical molecules switch it on through similar changes inside the protein.

Main

Temperature profoundly influences the activity of numerous proteins in the human body, shaping our ability to respond to a wide range of thermal stimuli, including noxious cold, cool, warm and noxious heat^1–17. Detecting extreme temperature, such as noxious heat, is a critical survival mechanism that is closely linked to pain signaling, as heat-induced responses trigger protective mechanisms to prevent tissue damage^3,18. Pain perception begins at the sensory neuron level, where specialized ion channels detect harmful stimuli and transduce signals to the nervous system^4,18–20.

Among these channels, TRPM3, a member of the transient receptor potential (TRP) family, functions as a thermosensor activated by noxious heat^3,9,21–24. It is a voltage-dependent, Ca²⁺-permeable, non-selective cation channel expressed in sensory neurons, including the dorsal root and trigeminal ganglia. TRPM3 shows substantial activation at approximately 35 °C, with markedly increased activity between 40 °C and 45 °C, within the range of noxious heat⁹. Beyond temperature sensing, TRPM3 also acts as a chemosensor and its activity is modulated by various ligands, including neuroactive steroid pregnenolone sulfate (PregS)^25,26, synthetic superagonist CIM0216²⁷ and antagonist primidone^28,29, which enhance or diminish TRPM3 activity, consequently altering the pain effect.

Deletion of TRPM3 in mice causes clear deficits in their avoidance responses to noxious heat and impairs the development of inflammatory heat hyperalgesia, highlighting the channel’s relevance to pain signaling⁹. Indeed, TRPM3 has been implicated in various pain-related conditions, including inflammatory thermal hyperalgesia and female migraine^30–33, making it a promising but underexplored target for pain management. Moreover, gain-of-function mutations in TRPM3 are associated with epilepsy, and its inhibitor primidone—a first-generation barbiturate-type antiepileptic drug—mitigates seizures^{28,29,34–39}. However, beyond primidone, selective drugs targeting TRPM3 for therapeutic applications are lacking. Despite its essential roles in pain signaling and neurological disorders, the mechanisms underlying the temperature-induced and ligand-induced activation and inhibition of TRPM3 remain poorly understood. Addressing these knowledge gaps could pave the way for the development of targeted therapies for TRPM3-related conditions.

To explore how TRPM3 responds to diverse stimuli, we determined its structures under apo conditions and in response to heat, as well as ligands, including CIM0216, and primidone. Structural dynamics analysis provided insights into how these stimuli activate TRPM3, and revealed the molecular determinants of temperature sensing and ligand binding. Furthermore, the findings highlight a converged mechanism through which heat and chemical ligands modulate the channel via shared structural transitions.

Results

Functional characterization of rabbit TRPM3

To investigate the polymodal gating of TRPM3, particularly its activation by heat, we identified rabbit TRPM3 (rTRPM3) as an ideal candidate due to its robust biochemical stability following thermal exposure—a common destabilizing factor for purified proteins. The rTRPM3 shares 98.4% and 99.3% sequence identity with the human TRPM3 splice variant (Uniprot Q9HCF6-10) and the mouse TRPM3 splice variant (Uniprot J9SQF3-13), respectively, which lacks the C-terminal domain (residues 1351–1732 in human TRPM3, and residues 1338–1732 in mouse TRPM3) of unknown function located immediately after the pole helix. Electrophysiological experiments confirmed the thermosensitivity and chemosensitivity of rTRPM3. Heating or exposure to the endogenous agonist PregS induced reversible activation of an outwardly rectifying current, while exposure to the superagonist CIM0216 elicited reversible activation of a nearly linear current (Extended Data Fig. 1a–c). The half-maximal effective concentrations (EC₅₀) of CIM0216 and PregS were determined to be 0.7 µM and 22 µM at 80-mV membrane potential, respectively (Extended Data Figs. 1d and 2a), aligning with previous findings^25,27. Furthermore, the antagonist primidone suppressed rTRPM3 currents induced by PregS. The estimated half-maximal inhibitory concentration (IC₅₀) for primidone was approximately 5 µM at 80 mV membrane potential (Extended Data Fig. 2j), consistent with earlier studies on mouse TRPM3²⁸.

Extended Data Fig. 1. Electrophysiological characterization of TRPM3 in response to chemical and thermal stimulation.

a, Representative whole-cell heat-activated currents measured in tsA cells overexpressing wild-type TRPM3. A voltage protocol was applied every 5 s to monitor current changes: −100 mV for 50 ms, ramped to +100 mV over 200 ms, and held at +100 mV for 50 ms. Currents were measured as the temperature increased from 22 °C to 37 °C (n = 5). Current amplitudes at +100 mV were plotted. The lower panel shows I/V curves corresponding to the time points indicated in the upper panel. b, c, Representative whole-cell currents measured in tsA cells overexpressing wild-type TRPM3 in response to CIM0216 (b, n = 5) and PregS (c, n = 5) using the same protocol as in panel (a). d–l, CIM0216 EC50 measurements for wild-type TRPM3. Upper panels show representative raw whole-cell currents, while lower panels present the fitted data. A two-step voltage protocol was applied every 5 s to monitor current changes: 0 mV for 50 ms, switch to +80 mV for 50 ms, followed −80 mV for 50 ms, and a return to 0 mV for 50 ms. Current amplitudes at +80 mV and −80 mV were plotted. Data are presented as mean ± s.e.m. The number of independent cell measurement (n) is indicated in each panel.

Source data

Extended Data Fig. 2. Electrophysiological characterization of TRPM3 in response to PregS and primidone.

a–i, PregS EC50 measurements for wild-type TRPM3. Upper panels show representative raw whole-cell currents, while lower panels present the fitted data. A protocol was applied every 5 s to monitor current changes: 0 mV for 50 ms, switch to +80 mV for 50 ms, followed by −80 mV for 50 ms, and a return to 0 mV for 50 ms. Current amplitudes at +80 mV and −80 mV were plotted. Data are presented as mean ± s.e.m. The number of independent cell measurement (n) is indicated in each panel. j, Primidone IC50 measurements for wild-type TRPM3 induced by PregS using the same protocol as in panels (a–j). Upper panels show representative raw whole-cell currents, while lower panels present the fitted data. Current amplitudes at +80 mV and −80 mV were plotted. Data are presented as mean ± s.e.m. The number of independent cell measurement (n) is indicated in each panel. k–o, Primidone inhibition on PregS induced currents in wild-type TRPM3 and primidone binding site mutants, using the same protocol as in panels (a–j). Representative current amplitudes at +80 mV and −80 mV are shown in the upper panel. The mean current values at each designated time point (a, b, c) are plotted in the lower panel. Time points a, b, and c correspond to baseline (before PregS application), during PregS application, and after addition of Primidone, respectively. Data are presented as mean ± s.e.m. Each point (n) represents a single independent cell measurement (n = 5 (WT), 3 (F852A), 3 (L892A), 3 (K896M), 4 (R1120A), with bars indicating the mean.

Source data

TRPM3 activation by superagonist CIM0216

Unlike chemical ligands, in which their sites of binding can be directly visualized, where the protein senses temperature might be challenging to detect, in particular, if it is distributed throughout the protein¹. Therefore, we first used CIM0216—a superagonist of TRPM3 that elicits robust, double-rectifying currents²⁷ (Extended Data Fig. 1b)—to study the structural and dynamic changes during channel activation, aiming to establish a foundation for addressing the challenges of studying temperature-dependent activation.

We determined the cryo-EM structure of rTRPM3 in the presence of 1 mM CIM0216 at an overall resolution of 3.0 Å (Extended Data Figs. 3 and 4, Supplementary Fig. 1d,f and Supplementary Table 1). The overall assembly closely resembles other TRPM channels⁴⁰ and is very similar to the mouse TRPM3 apo structure determined earlier⁴¹ (Fig. 1a, left panel). Each protomer consists of a transmembrane domain (TMD) and an intracellular domain (ICD), which includes the N-terminal melastatin homology regions (MHR1–4) and the C-terminal rib and pole helices. The TMD includes the S1–S4 domain and the pore domain (S5–S6), followed by the amphiphilic TRP helix. Notably, a small turret between the pore loop and S6 protrudes into the extracellular space.

Extended Data Fig. 3. Cryo-EM data processing workflow of TRPM3 with CIM0216 at 18 °C.

A representative micrograph, 2D class averages, and key maps are shown, along with the percentages of different subunit and tetramer conformations.

Extended Data Fig. 4. Cryo-EM analysis of TRPM3 bound to CIM0216 at 18 °C and TRPM3 in the apo state at 18°C.

The labels 4R0A, 3R1A, 2R2A_ortho (orthogonal), 2R2A_para (parallel), 1R3A, and 0R4A denote tetrameric TRPM3 configurations composed of varying ratios of resting (R) and activated (A) subunits. For each tetrameric configuration, the local resolution estimation, FSC curves (map vs. map and map vs. model), and angular distribution of particles contributing to the final cryo-EM map reconstruction are displayed from left to right.

Fig. 1. TRPM3 activation by superagonist CIM0216.

a, Structure of TRPM3 in complex with CIM0216 (CIM0216_0R4A), shown in cartoon representation and viewed parallel to the membrane. CIM0216 (green) is fitted into the cryo-EM density map (right). One subunit is highlighted in red, and the rib helix from the adjacent subunit is highlighted in cyan. Chemical structure of CIM0216 is shown in the middle. b, Close-up view of the CIM0216 binding site, detailing key interactions. CIM0216 is shown as stick representation with a transparent surface. c, Whole-cell currents measured in tsA cells overexpressing wild-type (WT) TRPM3 or CIM0216 binding-site mutants at 22 °C. A voltage protocol was applied every 5 s to monitor current changes: −140 mV for 50 ms, ramped to +140 mV over 200 ms and held at +140 mV for 50 ms. CIM0216-induced current amplitudes at +140 mV were calculated by subtracting the steady-state current recorded before CIM0216 application from the steady-state current recorded after CIM0216 application (I_+CIM0216 − I_−CIM0216). Data are plotted as mean ± s.e.m. Each point (n) represents a single independent cell measurement (n = 6 (WT), 4 (Y855A), 5 (Y859A), 4 (Y888A), 4 (L892A), 4 (E895A), 4 (W959R), 4 (R962A), 4 (I966A), 5 (R1120A)), with bars indicating the mean. Statistical analysis was performed using one-way analysis of variance (ANOVA) with Bonferroni’s post hoc test, comparing each mutant with wild type as control group (P = 0.0002 (Y855A), <0.0001 (Y859A), 0.0002 (Y888A), 0.0009 (L892A), 0.0002 (E895A), 0.0002 (W959R), 0.0003 (R962A), 0.0007 (I966A) and 0.9798 (R1120A)). Cell surface expression of WT TRPM3 and mutant proteins is shown, confirming successful trafficking to the membrane. d, Structural comparison of CIM0216-bound activated state (CIM0216_0R4A, red) and apo resting state of mouse TRPM3, superimposed using the TMD (residues 848–1119). r.m.s.d. values are shown for individual domains. Black arrows indicate the approximate direction of domain movements. ECD, extracellular domain.

Source data

The density of the entire TMD, including the pore helix and pore loop, is well resolved, except for part of the turret (residues 1047–1058), which is not visible (Supplementary Fig. 1d). This high-quality map allowed us to resolve a prominent T-shaped density, consistent with the shape of a CIM0216 molecule, within the S1–S4 domain (Fig. 1a, right). The ligand engages residues on the transmembrane helices S1, S2 and S3, as well as the TRP helix, through both hydrophobic and hydrophilic contacts (Fig. 1b). Alanine substitutions at these key residues diminished or abolished CIM0216-evoked currents (Fig. 1c), confirming the functional relevance of these interactions. The horizontal part of the T-shaped CIM0216 fits into the binding pocket of the S1–S4 domain, while the tip of the vertical part penetrates deeper into its core, filling the inner cavity and forcing a 90-degree flip of the side chain of Y855 away from the domain center (Fig. 1b).

The pore radius at the gate of the central conducting pore, defined by the side chains of I1088 and N1092, increases relative to the apo resting state of mouse TRPM3⁴¹; however, it remains only slightly larger than a fully dehydrated sodium ion (Extended Data Fig. 5a,b). This suggests that additional factors, such as membrane potential, are required to fully open the channel, consistent with the rectifying current induced by CIM0216^25,27 (Extended Data Fig. 1b).

Extended Data Fig. 5. Structural comparison of the TMD.

a, Surface representations of the ion-conducting pore in TRPM3 structures determined under various conditions, viewed parallel to the membrane. The pore region in two subunits is shown as a cartoon, with gate and selectivity filter residues depicted as sticks. b, Pore radius profiles of the structures shown in (a), plotted along the pore axis. c–g, Structural superpositions of the TMD between the resting and activated states of TRPM3, and the closed and open states of TRPM4 and TRPM5. Resting and closed states are shown as ribbons; activated and open states as cylinders. h–j, Structural comparisons between the resting and activated states of TRPM3, superimposed using the TMD. Root mean square deviation (r.m.s.d.) values are shown for TMD backbone atoms.

Conformational dynamics of TRPM3 with CIM0216

Although the cryo-EM map of the TMD was of high quality, the ICD—particularly the distal MHR1/2—exhibited conformational heterogeneity. This observation aligns with previous findings on other TRPM channels, where the ICDs adopt distinct conformations that play a critical role in regulating properties such as voltage dependence and sensitivity to temperature and ligands^42,43. Motivated by these insights, we analyzed the structure at the level of individual subunits, identifying two distinct subunit conformations: a major conformation representing 80% of subunits, and a minor conformation accounting for 20% (Fig. 1d and Extended Data Fig. 3, text highlighted in red). Both conformations were bound to CIM0216. The minor conformation closely resembled the apo resting state of mouse TRPM3, with differences largely localized to the S1–S4 domain due to CIM0216 binding. In contrast, the major conformation displayed substantial changes, with a backbone root mean square deviation (r.m.s.d.) of ~7 Å compared to the apo resting state of mouse TRPM3. The difference is even more pronounced when the TMD of a subunit is superimposed, as the MHR1/2 domain between the two states shows a remarkable backbone r.m.s.d. of 12 Å (Fig. 1d). These extensive structural rearrangements, particularly within the ICD, resemble the activation transition observed in TRPM4 upon heat and Ca²⁺ binding⁴³, suggesting that the major conformation represents a novel activated state induced by CIM0216.

In this activated state, CIM0216 binding elevates the TRP helix toward the extracellular side and causes its rotational movement along the pore axis, along with a shift in the S4 helix and the S4–S5 linker. These regions, which are key gating elements in the TRP superfamily, mediate communication between the TMD and ICD, and within the TMD (linking the S1–S4 domain to the S5–S6 pore domain). Consequently, their movements propagate structural changes to both the ICD and the pore domain, driving extensive rearrangements throughout the protein (Fig. 1d). The CIM0216-induced structural changes in the TRP and S6 helices resemble those observed during the transition from closed to open states in TRPM4 and TRPM5 (Extended Data Fig. 5f,g), although the change in the S6 of TRPM3 is less pronounced (Extended Data Fig. 5c–e).

To determine whether these conformational states exist exclusively in separate tetramers or coexist within the same tetramer, we analyzed the composition of all tetrameric particles. Our analysis revealed six distinct tetrameric conformations with varying combinations of resting and activated subunits, including four resting and zero activated (4R0A), two resting and two activated in a para arrangement (2R2A_para), two resting and two activated in an ortho arrangement (2R2A_ortho), one resting and three activated (1R3A) and zero resting and four activated (0R4A) (Fig. 2d and Extended Data Fig. 3). Tetramers with either four or three activated subunits were predominant, while those with two activated subunits were less frequent. Tetramers containing one or no activated subunit were rare. These findings suggest that, even bound to a superagonist and without a membrane potential, TRPM3 channels exist in a dynamic equilibrium between resting and activated conformations.

Fig. 2. TRPM3 conformational dynamics.

a, Cryo-EM density map of the apo resting state (Apo_18 °C_4R0A, left) and apo activated state (Apo_37 °C_0R4A, right). The black double arrow (left) indicates the separation between the MHR1/2 domain and the C-terminal pole helix in the resting state, which come closer in the activated state. b, Structural comparison of apo resting state (Apo_18 °C_4R0A, blue) and apo activated state (Apo_37 °C_0R4A, orange), superimposed using the TMD. r.m.s.d. values are shown for individual domains. Black arrows indicate the approximate direction of domain movement. c, Structural comparison of apo activated state (Apo_37 °C_0R4A, orange) and CIM0216-bound activated state (CIM0216_0R4A, red), superimposed using the ICD (residues 1–747). r.m.s.d. values are shown for individual domains. d, Structures of homotetrameric TRPM3 composed of four resting-state subunits (left, blue) and four activated-state subunits (right, orange). These structures are shown in surface representation, with one subunit displayed in cartoon and viewed parallel to the membrane. The six possible tetrameric compositions (middle) illustrate the coexistence of resting and activated subunits under different conditions (ligand and temperature), along with their respective occupancies. The percentages of the two compositions with two resting and two activated subunits (orthogonal and parallel positions) are combined. These compositions are displayed in surface representation viewed from the extracellular side, with resting subunits in blue and activated subunits in orange.

Heat shifts TRPM3 conformational dynamics

Building on these findings with CIM0216-induced activation, we next sought to investigate how heat activates TRPM3 in the absence of an agonist. To this end, we determined the cryo-EM structures of rTRPM3 at 18 °C and 37 °C, respectively (Extended Data Figs. 4 and 6–8 and Supplementary Tables 2 and 3). We chose 37 °C for the heat-induced study for two key reasons: functional data indicate that TRPM3 shows substantial, but not maximal, activity at 37 °C, as currents activated by 35 °C can be greatly enhanced by PregS, suggesting a mixture of resting and activated conformations ideal for capturing temperature-dependent dynamics⁹. Additionally, temperatures exceeding 37 °C compromised the quality of the cryo-EM data.

Extended Data Fig. 6. Cryo-EM analysis of TRPM3 in the apo state at 37 °C and TRPM3 bound to primidone at 18 °C.

The labels 4R0A, 3R1A, 2R2A_ortho (orthogonal), 2R2A_para (parallel), 1R3A, and 0R4A denote tetrameric TRPM3 configurations composed of varying ratios of resting (R) and activated (A) subunits. For each tetrameric configuration, the local resolution estimation, FSC curves (map vs. map and map vs. model), and angular distribution of particles contributing to the final cryo-EM map reconstruction are displayed from left to right.

Extended Data Fig. 8. Cryo-EM data processing workflow of TRPM3 under apo condition at 37°C.

A representative micrograph, 2D class averages, and key maps are shown, along with the percentages of different subunit and tetramer conformations.

Source data

The cryo-EM maps are of high overall quality, reaching resolutions up to 2.6 Å. The TMD is well resolved, including the pore loop, pore helix and extracellular turret (Supplementary Fig. 1a,b); only the tip of the turret (residues 1048–1056) is less well defined, although still visible at lower map contour levels. In contrast, we observed profound conformational heterogeneity in the ICDs, particularly the distal MHR1/2 domain, as indicated by the less well-defined cryo-EM map in these regions. Structural analysis at the subunit level also revealed two distinct conformations (Fig. 2a and Extended Data Figs. 7 and 8): a resting conformation, and a second conformation similar to the activated state observed with CIM0216, although differing in its TMD due to the absence of ligand binding (Fig. 2b,c and Extended Data Fig. 5h–j). This second conformation likely represents an activated state in the absence of ligand binding. These findings align with the conformational selection model, which posits that proteins intrinsically sample a range of conformations—including both resting and activated states—and that external stimuli, such as temperature or ligands, shift the equilibrium toward specific conformations. Consistent with this model, our data at 37 °C demonstrated that heat markedly altered the conformational distribution: at 18 °C, the resting conformation dominated (84%), whereas at 37 °C the proportion of the activated conformation increased to 40% (Extended Data Fig. 8, text highlighted in red).

Extended Data Fig. 7. Cryo-EM data processing workflow of TRPM3 under apo condition at 18 °C.

A representative micrograph, 2D class averages, and key maps are shown, along with the percentages of different subunit and tetramer conformations. The best tetramer map is highlighted with an orange background.

At the tetrameric level, we also observed that resting and activated subunits coexist within individual tetramers, forming six distinct combinations of resting and activated subunits (Fig. 2d). At 18 °C, tetramers with predominantly resting subunits (three or four) were most common, while tetramers with three or four activated subunits were rare. This distribution, combined with the lack of basal activity in TRPM3 at low temperatures in the absence of an agonist, suggests that channel activation likely requires at least three, if not all four, subunits to adopt the activated conformation. At 37 °C, the proportion of tetramers with three or four activated subunits increased 14-fold relative to 18 °C, rising from 2% to 27%. However, this is still markedly lower than the 83% observed with CIM0216-bound TRPM3 (Fig. 2d). This difference aligns with functional data showing that maximal heat-induced TRPM3 activity occurs at temperatures higher than 37 °C (ref. ⁹) and that heat-induced TRPM3 currents at 37 °C are substantially smaller than those evoked by CIM0216 (Extended Data Fig. 1a,b). Additionally, it should be noted that the temperature at which the cryo-EM sample was ‘frozen’ was probably slightly below the intended 37 °C due to instrument limitations, which could contribute to the lower observed distribution of activated states. Notably, the ion-conducting pore in tetramers with predominantly activated subunits remained closed, consistent with the fact that heat-induced TRPM3 opening is voltage dependent and produces outwardly rectifying currents.

Converged activation mechanism by heat and agonist

Structural comparisons between the resting and activated conformations—derived from all datasets mentioned above—revealed a consistent trend: the ICD undergoes a large counterclockwise inward rotation during the transition from resting to activated states when viewed from the intracellular side (Figs. 1d and 2b and Extended Data Fig. 9a). A key and conserved change was observed at the interface between the MHR1/2 domain and the rib helix: these regions interact in the resting conformation but dissociate in the activated state (Fig. 3a and Extended Data Fig. 9b–d), regardless of whether activation is driven by an agonist or heat.

Extended Data Fig. 9. Conserved conformational changes in the ICD induced by heat or agonist binding.

a, Both heat and agonist binding disrupt the interface between the MHR1/2 domain (depicted as blue and orange surfaces) and the adjacent rib helix (shown as a red cartoon). This disruption leads to a similar counterclockwise inward rotation of the MHR1/2 domain when viewed from the intracellular side. b–d, Close-up views of the interface highlighting the interaction between the positively charged loop in the MHR1/2 domain (blue) and the rib helix (red), shown with surrounding cryo-EM density in the apo resting state (Apo_18°C_4R0A, b), apo activated state (Apo_37 °C_0R4A, c), and CIM0216-bound activated state (CIM0216_0R4A, d). The dashed line indicates missing residues. Mean whole-cell currents measured in tsA cells overexpressing e, wild-type TRPM3 or mutants at the interface between the MHR1/2 domain and the rib helix f, K208A g, R210A h, R210A+R1183A i, K212A j, E1180A k, E1184A l, D1187A and m, E1180A+E1184A+D1187A. Unnormalized currents are plotted on the left, and currents normalized to the value at +140 mV are shown on the right. For clarity, only the positive values of the s.e.m. envelopes are displayed. A voltage protocol was applied every 5 s to monitor current changes: −140 mV for 50 ms, ramped to +140 mV over 200 ms, and held at +140 mV for 50 ms. Currents were measured as the temperature increased from 22 °C to 37 °C. Thermo currents at +140 mV were calculated by subtracting the steady-state current at 22 °C from that at 37 °C in the same cell. The number of independent cell mesurements (n) are e, n=9 f, n=5 g, n=7 h, n=7 i, n=8 j, n=8 k, n=6 l, n=5 and m, n=6.

Fig. 3. The MHR1/2 domain and rib helix interaction as a shared determinant of agonist-induced and heat-induced activation.

a, Left: structure of TRPM3 in white surface representation, with the MHR1/2 and MHR3/4 domains highlighted in blue and green, respectively, and the rib helix of the adjacent subunit highlighted in red. Right: close-up comparison of the interaction between the positively charged loop in the MHR1/2 domain and the rib helix in the resting and activated states. The structures of the two states are superimposed using the rib helix (residues 1163–1192). The rib helix is shown as a surface representation colored according to the electrostatic surface potential (from −5 to +5 kT e⁻¹, red to blue). The loop and its preceding helix in the MHR1/2 domain are shown in cartoon representation for the resting state (blue) and activated state (orange). The three positively charged residues are displayed in stick representation with transparent spheres. The angle between the helices in the resting and activated subunits is indicated. b, Cell surface expression of WT TRPM3 and mutant proteins is shown, confirming successful trafficking to the membrane. The experiment was carried out once. c, Whole-cell currents measured in tsA cells overexpressing wild-type TRPM3 or mutants at the interface between the MHR1/2 domain and the rib helix. A voltage protocol was applied every 5 s to monitor current changes: −140 mV for 50 ms, ramped to +140 mV over 200 ms and held at +140 mV for 50 ms. Currents were measured as the temperature increased from 22 °C to 37 °C. Thermo currents at 140 mV were calculated by subtracting the steady-state current at 22 °C from the steady-state current at 37 °C. The subtraction was performed using paired measurements from a single cell. Mean raw data are presented in Extended Data Fig. 9e–m. Data are presented as mean ± s.e.m and each point (n) represents a single independent cell measurement (n = 9 (WT), 5 (K208A), 7 (R210A), 7 (R210A + R1183A), 8 (K212A), 8 (E1180A), 6 (E1184A), 5 (D1187A) and 6 (E1180A + E1184A + D1187A)), with bars indicating the mean. Statistical analysis was performed using one-way ANOVA with Bonferroni’s post hoc test, comparing each mutant with wild type as control group (P = >0.9999 (K208A), 0.0169 (R210A), 0.0053 (R210A + R1183A), 0.1306 (K212A), >0.9999 (E1180A), >0.9999 (E1184A), >0.9999 (D1187A), 0.0012 (E1180A + E1184A + D1187A)). d, EC₅₀ values for PregS in wild-type TRPM3 and variants from c. Data are presented as mean ± s.e.m and each point (n) represents a single independent cell measurements (n = 6 (WT), 5 (K208A), 5 (R210A), 5 (R210A + R1183A), 5 (K212A), 5 (E1180A), 5 (E1184A), 5 (D1187A), and 5 (E1180A + E1184A + D1187A), with bars indicating the mean. e, EC₅₀ values for CIM0216 in wild-type TRPM3 and variants from c. Data are presented as mean ± s.e.m. Each point (n) represents a single independent cell measurement (n = 8 (WT), 4 (K208A), 4 (R210A), 5 (R210A + R1183A), 5 (K212A), 4 (E1180A), 4 (E1184A), 5 (D1187A), 4 (E1180A + E1184A + D1187A)), with bars indicating the mean. f, Current–voltage (I–V) curves corresponding to c for wild-type TRPM3 and the three mutants, measured after the currents reached a steady state. For clarity, only the positive s.e.m. envelopes are shown for each curve, except for R210A, for which the negative envelope is shown. g, Normalized I–V curves corresponding to f, with normalization based on the current at +140 mV.

Source data

The rib helix, which penetrates the ICD and forms multiple interactions with it, connects to the TRP helix and plays a critical role in transducing signals between the ICD and the TMD. In the resting conformation, the rib helix, containing 11 negatively charged residues, establishes electrostatic interactions with a positively charged loop in the MHR1 domain (Fig. 3a). In the activated conformation, however, this loop flipped away and became disordered, disrupting its interactions with the rib helix and coinciding with a rearrangement of the ICD. We hypothesized that this dissociation is linked to the conformational dynamics underlying heat-dependent and agonist-dependent activation. To test this idea, we neutralized the charged residues involved in the interaction to potentially weaken it, reasoning that a weakened interaction would facilitate the structural transition from the resting to the activated state, thereby enhancing channel sensitivity to both temperature and agonists.

Consistent with this hypothesis, mutants designed to weaken these interactions not only produced markedly larger heat-induced currents, but also altered the voltage sensitivity in electrophysiological experiments (Fig. 3b,c,f,g and Extended Data Fig. 9e–m). Notably, some mutants showed substantial inward currents at negative potentials, whereas the wild-type channel showed no conductance (Fig. 3f,g). Additionally, these mutants demonstrated dramatically enhanced activation by agonists, with up to 13-fold higher sensitivity to the endogenous agonist PregS and up to threefold higher sensitivity to the CIM0216, a highly potent superagonist that strongly favors the activated conformation, in dose–response experiments (Fig. 3d,e and Extended Data Figs. 1d–l and 2a–i).

Together, our structural and functional analysis reveals that the inherent structural dynamics of TRPM enable it to naturally sample both resting and activated states, providing flexibility for diverse activation stimuli. External stimuli, such as heat or agonists, bias this equilibrium toward the activated state, promoting channel activation. Both heat-induced and agonist-induced activation share a common structural transition in the ICD, specifically the dissociation of the MHR1/2 domain from the rib helix. This unified mechanism illustrates how distinct stimuli converge to activate TRPM3, with the ICD playing a central role in mediating channel activation.

Primidone-induced inhibition

The anticonvulsant primidone suppresses pain and treats seizures by inhibiting TRPM3^28,29. To investigate the binding site and mechanism of action of primidone, we determined the structure of rTRPM3 in complex with primidone at a resolution of 2.8 Å (Extended Data Figs. 6 and 10, Supplementary Fig. 1c,e and Supplementary Table 4). A prominent density was observed within the cavity of the S1–S4 domain, which closely matches the shape of the primidone molecule, allowing for an unambiguous fit (Fig. 4a). The binding was further confirmed by mutagenesis studies, where alanine substitutions at key residues markedly reduced or abolished the inhibitory effect (Fig. 4b,c).

Extended Data Fig. 10. Cryo-EM data processing workflow of TRPM3 with primidone at 18 °C.

A representative micrograph, 2D class averages, and key maps are shown, along with the percentages of different subunit and tetramer conformations.

Fig. 4. Primidone inhibition.

a, Structure of TRPM3 in complex with primidone (4R0A), shown in cartoon representation and viewed parallel to the membrane. Primidone (magenta) is fitted into the cryo-EM density map (right). One subunit is highlighted in blue. b, Close-up view of the primidone binding site, detailing key interactions. Primidone is shown as stick representation with a transparent surface. c, Ratio of whole-cell PregS-induced currents measured in the presence and absence of primidone for tsA cells overexpressing wild-type TRPM3 or primidone binding-site mutants at 22 °C. A protocol was applied every 5 s to monitor current changes: 0 mV for 50 ms, switch to +80 mV for 50 ms, followed by −80 mV for 50 ms and a return to 0 mV for 50 ms. The ratio of remaining currents (I/I₀) after the application of primidone was calculated using steady-state currents at +80 mV and −80 mV. Data are presented as mean ± s.e.m. Each point (n) represents a single independent cell measurement (n = 5 (WT), 3 (F852A), 3 (L892A), 3 (K896M) and 4 (R1120A)), with bars indicating the mean. Statistical analysis was performed using one-way ANOVA with Bonferroni’s post hoc test comparing each mutant with wild type as control group (80 mV: P = 0.0030 (F852A), <0.0001 (L892A), <0.0001 (K896M), <0.0001 (R1120A); −80 mV: P = 0.0673 (F852A), <0.0001 (L892A), <0.0001 (K896M), <0.0001 (R1120A)). d, I–V curves showing primidone inhibition of heat-activated whole-cell currents for wild-type TRPM3. The traces represent average currents at 22 °C, 37 °C and 37 °C with primidone (n = 5, paired). For clarity, positive s.e.m. envelopes are shown for the curves at 22 °C and 37 °C, whereas the negative envelope is shown for the 37 °C + primidone curve.

Source data

Interestingly, despite primidone acting as an inhibitor and CIM0216 as an agonist, both compounds occupy the same cavity within the S1–S4 domain (Figs. 1a and 4a). A comparison of the primidone-bound and the CIM0216-bound structures reveals how their opposing effects arise. CIM0216, being a bulky molecule that inserts deeper into the S1–S4 domain, induces substantial movement in the TRP helix, which is sandwiched between the S1–S4 domain and the ICD. This movement propagates to the ICD, shifting the conformational equilibrium toward the activated state (Figs. 1d and 2d). Moreover, CIM0216 displaces the S4 helix and S4–S5 linker, thereby altering the conformation of the S5–S6 pore domain. Together, these changes in the ICD and the pore domain may work allosterically to facilitate channel opening in the presence of membrane potential.

Primidone, by contrast, is smaller and fits snugly into the S1–S4 cavity without causing major conformational changes. Its binding primarily affects the TRP helix and S3 helix, along with localized adjustments in the side chains of residues involved in primidone binding. The small movement in the S3 helix, located at the periphery of the TMD facing the lipid bilayer, is unlikely to impact the overall protein conformation. By contrast, the shift in the TRP helix resembles that observed with CIM0216, albeit smaller, and is therefore expected to propagate to the ICD to a lesser extent. Consistent with this, structural analysis at the subunit level revealed a moderate increase in the population of activated subunits on primidone binding, from 16% to 30% (Extended Data Figs. 7 and 10), accompanied by a proportional rise in tetramers containing three or four activated subunits (Fig. 2d).

Importantly, however, primidone binding does not alter the conformation of the S4 helix or S4–S5 linker, leaving the S5–S6 pore domain in the resting state. This effectively locks the channel in a nonconductive state, regardless of the conformation of the ICD. This inhibition mechanism aligns with electrophysiology data showing that primidone effectively suppresses heat-induced channel activation (Fig. 4d), which primarily drives conformational changes in the ICD. Additionally, primidone probably acts as a competitive antagonist when inhibiting CIM0216-induced currents, as both compounds compete for the same binding site. Our findings demonstrate that two ligands share the same binding site yet have opposing effects because they exert distinct structural effects, driving diverse mechanisms of TRPM3 modulation.

Discussion

Effective pain management involves either reducing neuronal excitation or enhancing inhibition within the nervous system. Opioids, for instance, reduce neurotransmitter release presynaptically and hyperpolarize neurons postsynaptically⁴⁴. However, their high risk of addiction highlights the need for alternative approaches. Targeting nociceptors, specialized sensory neurons that initiate pain signals, offers a promising strategy by modulating ion channels, such as TRPM3, which detect noxious stimuli⁴⁵. This study offers insights into the polymodal gating mechanisms of TRPM3, revealing its ability to respond to diverse stimuli, including heat and chemical ligands, through distinct and shared structural transitions. We identified key molecular determinants for the binding of superagonist CIM0216 and antagonist primidone. Interestingly, CIM0216 and primidone, despite their opposing effects, share the same binding site within the S1–S4 domain. Structural comparisons revealed that their distinct effects arise from differences in size and shape, dictating their ability to allosterically drive conformational changes in the intracellular and pore domains. This contrast underscores the versatility of the S1–S4 domain as a binding site capable of accommodating diverse ligands while producing distinct functional outcomes.

We demonstrated that heat and the agonist CIM0216 promote TRPM3 activation by shifting the conformational equilibrium from the resting state toward the activated state (Fig. 5). Despite their distinct modes of action, both stimuli converge on a shared structural transition where a related conformational change in the ICD is coupled to activation by heat or CIM0216. This transition involves the dissociation of the MHR1/2 domain from the rib helix of the adjacent subunit within the ICD, highlighting a converged mechanism by which distinct stimuli activate the channel.

Fig. 5. Converged activation mechanism by heat and agonist.

Schematic representation of the polymodal activation of TRPM3 by chemical agonists and heat. Both stimuli induce a similar large-scale conformational rearrangement in the ICD, characterized by the disruption of the interface between the MHR1/2 domain and the adjacent rib helix. These changes may facilitate structural transitions in the TMD to promote the opening of the ion-conducting pore in response to membrane depolarization or other cellular cues. The structure of an activated, fully open state of TRPM3 remains to be determined.

Building on our findings, we explored the thermal sensing properties of TRPM3 in comparison to other temperature-sensitive TRP channels, such as TRPM4, TRPM8 and TRPV^43,46,47. In TRPM4, channel activation at high temperature involves substantial intersubunit rearrangements in the ICD⁴³. These large-scale structural reorganizations require substantial energy, making heat alone insufficient to drive activation. Instead, calcium, an additional activation factor, is required to facilitate the structural rearrangement of the ICD, and it works synergistically with heat to activate the channel. In TRPM3, heat-induced conformational changes in the ICD primarily involve the dissociation of the MHR1/2 domain from the rib helix. Notably, the electrostatic interaction between these regions appears relatively limited, relying mainly on a short positively charged loop and the rib helix (Fig. 3a). This limited interaction probably provides sufficient flexibility for the ICD to undergo thermal activation without requiring additional energy input from another factor, such as calcium. This distinction may reflect the role of heat as a direct activation stimulus in TRPM3, compared to its role as a modulator cooperating with calcium in TRPM4. In TRPM8, it has also been observed that replacing the MHR1–3 domain of a cold-sensing ortholog into a cold-insensitive ortholog renders the latter cold-responsive⁴⁸. Together, our studies on TRPM4 and TRPM3, along with published findings on TRPM8, highlight the important role of the ICD in responding to temperature changes across TRPM channels.

In contrast, TRPV channels employ distinct structural mechanisms for thermal activation. In TRPV1, heat induces global yet subtle conformational changes, encompassing the ICD and the TMD⁴⁶. In TRPV3, thermal activation involves structural changes in the S2–S3 linker within the TMD, as well as rearrangements in both the N and C termini⁴⁷. These differences underscore the diversity of thermal sensing mechanisms within and across protein families, probably tailored to their distinct physiological roles. In line with this diversity, thermosensing in TRPM3 may involve multiple regions of the protein rather than a single defined module. Resolving the contributions of these regions will require additional structural insights—particularly from the open state—to capture a more complete picture of temperature-induced conformational changes.

A key unresolved question is the binding site of the endogenous agonist PregS and its mechanism of action. Despite extensive efforts, we could not confidently identify the PregS binding site, probably due to its structural similarity to cholesterol or cholesterol hemisuccinate (CHS), a known TRPM3 modulator²⁶. CHS was indispensable for isolating TRPM3 in our system, complicating the structural observation of PregS binding, as CHS and PregS may compete for the same site(s). Nevertheless, our structures provided clues about potential PregS binding regions. Within the TMD, six CHS-like molecules were identified in the cryo-EM map (Fig. 6a–c), potentially occupying native cholesterol-binding sites. Notably, application of CHS in electrophysiological recordings did not alter channel activation by PregS (Fig. 6d), suggesting that CHS may substitute for cholesterol without introducing functional artifacts. These CHS molecules, along with other lipid-like densities, form a tightly packed ‘lipid belt’ surrounding the TMD (Fig. 6a). We hypothesize that PregS may bind to one or more of these sites to activate TRPM3. Supporting this hypothesis, the gain-of-function mutation P1069Q, located adjacent to a CHS density, increases TRPM3 sensitivity and is associated with developmental and epileptic encephalopathies^34,39,49(Fig. 6b). This suggests that CHS binding sites may overlap with or interact with PregS binding regions, influencing channel gating. Of note, a recent study proposed a putative PregS binding site that corresponds to one of the CHS binding sites identified in our study⁵⁰ (Fig. 6c, CHS site 3). However, given the moderate resolution and quality of the cryo-EM data in this study⁵⁰, further research will be required to disentangle the effects of CHS and PregS and definitively identify PregS-mediated activation sites.

Fig. 6. Binding sites of cholesteryl hemisuccinate.

a, Cryo-EM map of the apo resting state (Apo18 °C_4R0A), highlighting one subunit in green and putative lipid densities in yellow, viewed parallel to the membrane (left) and from the intracellular side (right). Right: the three most well-defined CHS densities are labeled. b, The three CHS densities labeled in a, shown alongside the molecular structure of PregS for comparison. c, Close-up views of the CHS binding sites. The six CHS molecules are shown in yellow, with key interacting residues from the protein depicted in green. The position of the disease-associated mutation P1069Q is highlighted with a rectangle. d, I–V curves showing that CHS application had no effect on PregS-activated whole-cell currents in wild-type TRPM3. Traces show average currents at 100 µM CHS, 100 µM PregS and 100 µM PregS + 100 µM CHS (n = 3, paired). For clarity, the positive s.e.m. envelope is shown for the CHS + PregS curve, and the negative envelope is shown for the PregS curve.

Source data

By elucidating the mechanisms of polymodal gating, this study establishes TRPM3 as a model for understanding ion channel activation by diverse stimuli. The structural plasticity of TRPM3 underscores its ability to accommodate diverse modes of regulation. The identified ligand-binding sites, including the S1–S4 domain cavity and CHS binding sites, represent promising druggable targets. Furthermore, modulating the electrostatic interactions between MHR1/2 and the rib helix with small molecules could provide a new approach to regulating TRPM3 activation. Together, these structural insights pave the way for the development of selective TRPM3 modulators to address critical unmet needs in pain management and neurological disorders.

Methods

Cell lines

Sf9 cells and tsA201 cells were purchased from the American Type Culture Collection and routinely maintained in our laboratory. The cells were not authenticated experimentally in this study. No commonly misidentified cell lines were used.

Constructs and mutagenesis

The full-length rabbit TRPM3 (rTRPM3) gene (Uniprot G1T6W8_RABIT) was subcloned into pEG BacMam vector with an N-terminal 8XHis-Twin-Strep-Alfa followed by a thrombin cleavage site⁵¹. Site-directed mutagenesis was performed by QuickChange site-directed mutagenesis protocol (Agilent) and confirmed by sequencing (Eurofins or Plasmidsaurus).

TRPM3 protein expression and purification

For bacmid generation, pEG BacMam vector was transformed into DH10Bac cells. SF9 cells were transfected with freshly prepared bacmids for generation of P1 and subsequently P2 baculovirus. P2 viruses (8%) were used to infect tSA201 cells grown in Freestyle 293 expression medium (Gibco) in suspension culture. The cells were initially incubated at 37 °C for about 10–12 h. Thereafter, sodium butyrate was added to a final concentration of 10 mM and temperature was reduced to 30 °C to boost expression. The cells were collected after a total growth of about 65 h, washed with buffer containing 20 mM HEPES pH 7.5 and 150 mM NaCl (HBS) and pelleted by centrifugation. The cell pellets were either processed immediately or flash frozen in liquid nitrogen and stored at −80 °C.

The entire protein purification was carried out at 4 °C. The cells were suspended in HBS buffer containing 10 mM β-ME, 1 mM phenylmethylsulphonyl fluoride, 0.8 μM aprotinin, 2 μg ml⁻¹ leupeptin and 2 mM pepstatin A and lysed by sonication. The cell lysate was centrifuged at 8,000g for 20 minutes to remove cell debris. The membrane fractions were collected by centrifugation at 186,000g for 1 h in an ultracentrifuge (Beckman Coulter). The membranes were resuspended in HBS containing 10 mM β-ME, 2 mM Mg²⁺/ATP and protease inhibitors and homogenized using a Dounce homogenizer. After resuspension, membranes were solubilized in 1% LMNG/CHS (5:1 wt/wt) for about 1.5 h at 4 °C followed by ultracentrifugation at 186,000g for 1 h. The supernatant was loaded onto Strep-Tactin Superflow resin (IBA) at a flow rate of 0.1 ml min⁻¹. Thereafter, the resin was washed twice. The first wash was given with HBS buffer containing 0.01% LMNG/CHS (5:1 wt/wt), 10 mM β-ME and 1 mM Mg²⁺/ATP followed by a second wash with HBS buffer containing 0.003% LMNG/CHS (5:1 wt/wt), 10 mM β-ME and 5 mM EDTA. The protein was eluted with HBS buffer containing 0.003% LMNG/CHS (5:1 wt/wt), 10 mM β-ME, 5 mM EDTA and 10 mM desthiobiotin. The elution was concentrated to 0.5 ml and further purified by size exclusion chromatography (SEC) using Superose 6 increase 10/300 GL column (GE Healthcare) in a HBS buffer having 0.001% LMNG/CHS (5:1 wt/wt), 10 mM β-ME and 5 mM EDTA. The peak SEC fractions were pooled and concentrated to about 8 mg ml⁻¹ for freezing cryo-EM grids.

Cryo-EM sample preparation and data acquisition

Grids for the apo condition were prepared at both 18 °C and 37 °C. A 2.5-μl protein sample was applied to a freshly glow-discharged Quantifoil holey carbon grid (Au 1.2/1.3, 300 mesh) and, after a wait time of 15 s at 37 °C, the grids were blotted for 1.5 s in the Vitrobot Mark III set to 100% humidity and a blot force of −15. The grids were plunge-frozen into liquid ethane cooled by liquid nitrogen. For sample preparation at 18 °C, 2.5 μl of protein sample was applied to a freshly glow-discharged Quantifoil holey carbon grid (Au 1.2/1.3, 300 mesh). After waiting for 5 s, the grids were blotted for 1 s in the Vitrobot Mark III set to 18 °C, 100% humidity and a blot force of 0. For CIM0216 and primidone-containing conditions, the purified protein was mixed with 1 mM CIM0216 or primidone and incubated for 1 h. Thereafter, grids were frozen in the same manner as described for the apo condition at 18 °C. The cryo-EM data were collected using an FEI Titan Krios electron microscope operating at 300 kV and a nominal magnification of ×105,000 and an energy filter (20-eV slit width). Videos were recorded in super-resolution mode using a Gatan K3 Summit direct electron detector with a binned pixel size of 0.826 Å. Nominal defocus ranged from −1.4 to −1.1 μm with a total dose of 58 e⁻.

Cryo-EM data processing and analysis

To maintain consistency across datasets, all the data were collected and processed in the same manner. Super-resolution video stacks were motion corrected and 2× binned using MotionCorr2 (v.1.1.0)⁵². The contrast transfer function (CTF) parameters were estimated by CTFFIND (v.4.1.10)⁵³. Particles were picked with RELION template-picking⁵⁴, Gautomatch (v.0.56) (https://github.com/JackZhang-Lab/Gautmatch) and Topaz (v.0.2.4)⁵⁵. Each particle set was independently cleaned by removing only obviously non-TRPM3-like junk particles through rounds of heterogeneous refinement in CryoSPARC⁵⁶. The remaining particles from the three cleanups were merged and duplicates removed. The deduplicated particles were used for homogeneous refinement and nonuniform refinement with C4 symmetry in CryoSPARC. The particles were further subjected to multiple rounds of CTF refinement and Bayesian polishing in RELION, followed by three-dimensional (3D) refinement with C4 symmetry to improve the map resolution.

The resulting particles were symmetry expanded at the single subunit level with C4 symmetry. Using a mask encompassing a single subunit ICD and rib helix from the neighboring subunit, signal subtraction was carried out. The subtracted images were used for 3D classification without image alignment in RELION. The tetrameric particles containing broken/disordered ICD were excluded from further analysis, as they probably represent damaged particles during cryo-EM sample preparation. The remaining intact tetrameric particles were pooled for subsequent analysis. The ratio of subunits in the resting and activated states within these intact tetrameric particles was calculated. Subsequently, subunit particles in the resting and activated states were traced back to their respective tetrameric assemblies, resulting in six distinct compositions: four resting and zero activated (4R0A), three resting and one activated (3R1A), two resting and two activated in orthogonal positions (2R2A_ortho), two resting and two activated in parallel positions (2R2A_para), one resting and three activated (1R3A) and zero resting and four activated (0R4A) particles. The two homotetrameric compositions (4R0A and 0R4A) underwent further local refinement with C4 symmetry. For the heterotetrameric compositions, particles were rotated to align the positions of resting and activated subunits within the tetramer, ensuring proper orientation, and subsequently underwent local refinement with C1 symmetry in RELION.

The detailed image processing steps are summarized and illustrated in Extended Data Figs. 3, 7, 8 and 10. Map resolution estimates were based on the gold standard Fourier shell correlation (FSC) 0.143 criteria for all the datasets.

Model building

For building atomic models of rTRPM3, the published mouse TRPM3 structure (PDB 8DDQ) was docked into the final cryo-EM map by rigid-body fitting. Thereafter, the residues were mutated to rTRPM3 and manually fit into the map using Coot. For obtaining the flipped subunit conformation, the MHR domains, TMD and C-terminal domains of the nonflipped conformations were fitted into the flipped conformation map by rigid-body fitting and then manually adjusted in Coot⁵⁷. Ligands were manually fitted into the density in Coot and refined in real space. All the models were subjected to real-space refinement in Phenix⁵⁸ to improve the model metrics. The final models were validated using MolProbity⁵⁹. Figures 1–6, Extended Data Figs. 3–10 and Supplementary Fig. 1) were prepared using UCSF ChimeraX⁶⁰ and PyMOL (Schrodinger).

Electrophysiology

TsA201 cells expressing plasmids encoding N-terminal GFP-tagged rabbit TRPM3 were used. After one day post-transfection with plasmid DNA (100 ng ml⁻¹) and Lipofectamine 2000 (Invitrogen, 11668019), the cells were trypsinized and replated onto poly-L-lysine-coated (Sigma, P4707) glass coverslips. After cell attachment, the coverslip was transferred to a recording chamber. Whole-cell patch-clamp recordings were performed at room temperature (21–23 °C) or body temperature (36–38 °C). The temperature of perfusion solutions was controlled by thermal control devices (SC-20/CL-100, Warner Instruments).

Signals were amplified using a Multiclamp 700B amplifier and digitized using a Digidata 1550B A/D converter (Molecular Devices). The whole-cell current was measured on the cells with an access resistance of less than 10 MΩ after the whole-cell configuration was obtained. The whole-cell capacitance was compensated by the amplifier circuitry. The ramp pulse from −140 to 140 mV (or from −100 to 100 mV) for 200 ms or the two-step pulse from 80 to −80 mV for 50 ms was continuously applied to the cell membrane every 5 s to monitor the TRPM3 current. Electrical signals were digitized at 10 kHz and filtered at 2 kHz. Recordings were analyzed using Clampfit v.11.3 (Axon Instruments), GraphPad Prism v.10 and OriginPro 2024 (OriginLab). The standard bath solution contained (in mM): 150 NaCl, 5 KCl, 1 MgCl₂, 2 CaCl₂, 12 mannitol, 10 HEPES, pH 7.4 with NaOH. For a whole-cell recording, the extracellular and intracellular solutions contained (in mM): 150 NaCl, 10 HEPES, 1 MgCl₂, 5 EGTA.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41594-025-01692-5.

Author contributions

J.D. and W.L. supervised the project. W.C., S.V. and S.K. initiated the project. S.K. and F.J. carried out protein purification, cryo-EM data collection and processing. R.P.M., S.I.K. and S.K. generated the mutants and carried out construct screening. S.J.P. performed electrophysiological experiments. J.D., W.L., S.K., F.J. and S.P. contributed to the paper preparation.

Peer review

Peer review information

Nature Structural & Molecular Biology thanks Kenton Swartz and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available. Primary Handling Editor: Katarzyna Ciazynska, in collaboration with the Nature Structural & Molecular Biology team.

Data availability

Data and materials can be obtained from the corresponding authors upon request. The cryo-electron microscopy density map and atomic models were deposited in the Electron Microscopy Data Bank (EMDB) and in the Research Collaboratory for Structural Bioinformatics Protein Data Bank (RCSB PDB), respectively. The EMDB accession codes for Apo_18 °C data are EMD-4881 (4R0A), EMD-70209 (3R1A), EMD-70210 (2R2A_ortho), EMD-70211 (2R2A_para), EMD-70212 (1R3A) and EMD-48814 (0R4A). The PDB accession codes for Apo_18 °C data are 9N1G (4R0A), 9O7W (3R1A), 9O7X (2R2A_ortho), 9O7Y (2R2A_para), 9O7Z (1R3A) and 9N1J (0R4A). The EMDB accession codes for Apo_37 °C data are EMD-48812 (4R0A), EMD-70213 (3R1A), EMD-70214 (2R2A_ortho), EMD-70215 (2R2A_para), EMD-70216 (1R3A) and EMD-48815 (0R4A). The PDB accession codes for Apo_37 °C data are 9N1H (4R0A), 9O80 (3R1A), 9O81 (2R2A_ortho), 9O82 (2R2A_para), 9O83 (1R3A) and 9N1K (0R4A). The EMDB accession codes for Primidone data are EMD-48816 (4R0A), EMD-70222 (3R1A), EMD-70223 (2R2A_ortho), EMD-70225 (2R2A_para), EMD-70226 (1R3A) and EMD-48817 (0R4A). The PDB accession codes for Primidone data are 9N1L (4R0A), 9O8C (3R1A), 9O8D (2R2A_ortho), 9O8F (2R2A_para), 9O8G (1R3A) and 9N1M (0R4A). The EMDB accession codes for CIM0216 data are EMD-48813 (4R0A), EMD-70218 (3R1A), EMD-70219 (2R2A_ortho), EMD-70220 (2R2A_para), EMD-70221 (1R3A), EMD-48819 (0R4A) and EMD-48877 (Best_TMD). The PDB accession codes for CIM0216 data are 9N1I (4R0A), 9O86 (3R1A), 9O87 (2R2A_ortho), 9O88 (2R2A_para), 9O89 (1R3A), 9N1N (0R4A) and 9N4J (Best_TMD). Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Extended data

is available for this paper at 10.1038/s41594-025-01692-5.

Supplementary information

The online version contains supplementary material available at 10.1038/s41594-025-01692-5.