Temperature sensitivity is a defining feature of many macromolecules, affecting their function in physiology12,13,14,15,16. Studies of these macromolecules at the biophysical level are commonly conducted at subphysiological temperatures to preserve protein integrity. However, this practice may not accurately reflect their function in the human body—a consideration that is particularly crucial for thermosensing ion channels, such as members of the transient receptor potential (TRP) channels17,18,19,20,21,22,23,24,25,26,27. Among them, the TRPM4 channel, a member of the melastatin subfamily4,28, has been identified by us as a notable example.

TRPM4 is widely expressed in various tissues and has important roles such as cellular depolarization, cardiac rhythm generation and immune response29,30,31,32. Mutations in TRPM4 are linked to serious cardiac conditions such as Brugada syndrome32. TRPM4 and its closest homologue TRPM5 are monovalent cation-selective channels; they are activated by intracellular Ca2+ with a weak voltage dependence; and their activities are regulated by small molecules and lipids5,7,8,33,34,35,36. Like many TRP channels, they exhibit pronounced temperature sensitivity, with activities strongly intensifying near physiological temperatures6,37. However, the molecular mechanism underlying this temperature impact remains unclear, as existing structural data, from our group and others, have been limited to non-physiological temperature10,11,38,39,40.

Here we investigated the structure, function and pharmacology of human TRPM4 channel at physiological temperature. Our findings revealed a striking phenomenon: TRPM4 can adopt distinct conformations at different temperatures, markedly influencing where and how ligands interact with them. Similar temperature-dependent structural changes and ligand recognition have not been reported in the structures of TRPV1 and TRPV3 activated by heat41,42. Thus, our findings and those of others suggest a complex and diverse thermal response across different protein families.

Structure determination of TRPM4 at 37 °C

At room temperature (around 22 °C), cells overexpressing human TRPM4 exhibited relatively small Ca2+-activated currents (Extended Data Fig. 1a). By contrast, near physiological temperature (about 37 °C), we observed a substantial increase in current magnitude (Extended Data Fig. 1a), consistent with a previous report that TRPM4 is a temperature-sensitive channel6. This observation underscores the distinct properties of TRPM4 at physiological temperatures versus room temperature, hinting that the differences may be even more pronounced compared with at the lower temperatures (4–18 °C) that are commonly used in cryo-electron microscopy (cryo-EM) sample preparation. This prompted us to investigate the molecular mechanism of TRPM4 under physiological temperatures. Purified human TRPM4 demonstrated high stability (Extended Data Fig. 2), providing a basis for its structural determination at physiological temperature. We incubated TRPM4 with saturating Ca2+ at 37 °C before grid preparation in a Vitrobot chamber, which was also maintained at 37 °C.

In the cryo-EM map, the transmembrane domain (TMD) is well resolved; however, the intracellular domain (ICD) containing the MHR1/2 and MHR3/4 domains showed a pronounced deterioration. In particular, the MHR1/2 domain appeared predominantly disordered with spike-like features (Extended Data Fig. 3), suggesting a mixture of different conformations. Indeed, three-dimensional (3D) classification at the single-subunit level revealed two classes with distinct ICD conformations (Extended Data Fig. 3). The first, representing around 25% of the particles, resembled the TRPM4 conformation observed at 4–18 °C (refs. 10,11,38,39,40). The second, representing around 75% of the particles, is a conformation in which the MHR1/2 domain is tilted towards the membrane compared with the first class (Extended Data Fig. 3). We refer to these conformations as ‘cold’ and ‘warm’, reflecting the temperatures at which they were identified and for ease of discussion. As discussed later, we speculate that both conformations may co-exist at physiological temperatures, with their equilibrium controlled by local Ca2+ concentration.

Motivated by the identification of temperature-dependent conformations, we investigated whether ligand recognition and modulation of TRPM4 may also be affected by temperature. We focused on the ligands targeting the ICD given its pronounced conformational changes, including decavanadate (DVT), which positively modulates TRPM4 voltage dependence8, and ATP, a TRPM4 inhibitor9. To this end, we performed cryo-EM studies on the TRPM4 samples prepared with saturating Ca2+ at 37 °C in the presence of DVT and ATP, respectively (Extended Data Figs. 4 and 5). These conditions yielded both the cold and warm conformation, and we identified ligand-binding sites that are unique to the warm conformation. We also determined the TRPM4 structure in the presence of Ca2+ at 18 °C, as well as in the presence of EDTA at 37 °C, both of which yielded exclusively the cold conformation (Extended Data Fig. 6a,b), indicating that neither Ca2+ nor temperature alone shifts the cold-to-warm equilibrium. Given that all of the cold conformations in this study resembled previously published TRPM4 structures10,11,38,39,40, our structural analysis and discussion will mainly address the warm conformation and its ligand-binding sites.

Temperature shapes TRPM4 architecture

We determined human TRPM4 structures at physiological temperature in different functional states, including the apo state (EDTA–TRPM4), an agonist-bound state (Ca2+–TRPM4warm), an agonist- and positive-modulator-bound state (Ca2+/DVT–TRPM4warm) and an agonist- and inhibitor-bound state (Ca2+/ATP–TRPM4warm), at resolutions of 3.1–3.2 Å (Extended Data Tables 1 and 2 and Extended Data Fig. 7). All of the structures had a vertical arrangement of the TMD, MHR3/4 and MHR1/2 domains from top to bottom (Fig. 1). The warm conformation showed pronounced vertical compression and horizontal expansion of the ICD compared with the cold conformation (Fig. 1 (blue versus yellow)).

Fig. 1: The overall structures of TRPM4cold bound to Ca2+ and TRPM4warm bound to Ca2+, Ca2+ and DVT, or Ca2+ and ATP.
figure 1

The structures are shown as a surface representation with one subunit in cartoon, viewed parallel to the membrane (top row) or from the intracellular side (bottom row).

Three new ligand-binding sites were identified exclusive to the warm conformation. A Ca2+-binding site, named Cawarm, is located at the interface between the MHR1/2 and MHR3/4 domains (Fig. 1). A DVT-binding site is situated at the TMD and ICD interface (Fig. 1). This site, termed DVTwarm, is distinct from the two DVT-binding sites known in the cold conformation10. Lastly, an ATP-binding site, ATPwarm, is located between the MHR1/2 domain and the rib helix of the neighbouring subunit (Fig. 1), approximately 20 Å from the previously identified ATPcold site11.

Looking into the TMD, the Ca2+-bound cold and warm conformations have subtle but noticeable differences in the S1–S4 domain, S4–S5 linker and TRP helix due to the movement of the ICD that is in direct contact with these regions (Supplementary Video 1). However, their pore domains are nearly identical, each featuring a closed ion-conducting pore (Fig. 2). This implies that the combined effect of Ca2+ and physiological temperature, although inducing large conformational changes in the ICD, is insufficient to open TRPM4 at 0 mV potential. This aligns with electrophysiological data demonstrating that Ca2+-induced TRPM4 activation at 37 °C is outward rectifying and the currents at negative membrane potentials desensitize rapidly (Extended Data Fig. 1a,c,e). We therefore propose that the Ca2+–TRPM4warm structure represents a pre-open state or desensitized state.

Fig. 2: The ion-conducting pore.
figure 2

a, The profiles of the ion-conducting pore (shown as a surface representation) in different functional states, viewed parallel to the membrane. The pore region (shown as a cartoon) and residues (shown as sticks) forming the gate and the selectivity filter in two subunits are depicted. b, The pore radius along the pore axis.

Notably, the addition of DVT resulted in an open pore in the Ca2+/DVT–TRPM4warm structure with a radius of 2.5 Å (Fig. 2), allowing monovalent cations to pass through4. This open configuration is coherent with electrophysiological evidence that DVT renders TRPM4 largely voltage independent8, but was not observed in the previously reported structure in the DVT-bound cold conformation10. Thus, our finding reveals an open state of TRPM4, highlighting the intricate interplay between temperature, ligand binding and channel gating. By contrast, the ATP-bound warm conformation yielded a closed pore, representing an inhibited state (Fig. 2).

TRPM4 conformational dynamics

To elucidate the molecular basis of the conformational dynamics between cold and warm conformations, we compared the structures of Ca2+–TRPM4warm and Ca2+–TRPM4cold, and observed marked structural changes throughout the protein. This conformational disparity is especially evident in the MHR1/2 domains, which, when superimposing a warm subunit with a cold subunit using the MHR3/4 domains, showed a pronounced rigid-body rotation towards the MHR3/4 domain (Fig. 3a (yellow versus blue) and Supplementary Video 1). This movement converged specific residues from the MHR1/2 and MHR3/4 domains, creating the Cawarm site (Fig. 3b). Furthermore, our analysis revealed that the TRPM4 warm conformation closely resembled the Ca2+-bound zebrafish TRPM5 structure (Fig. 3a (yellow versus grey)), with the Cawarm site in TRPM4 being similar to the Ca ICD site in TRPM543, although the latter does not depend on physiological temperature.

Fig. 3: Temperature and Ca2+ govern the cold-to-warm transition.
figure 3

a, Comparisons between subunits of Ca2+-bound wild-type TRPM4 cold and warm conformations, Ca2+-bound TRPM4(E396A) and Ca2+-bound zebrafish TRPM5 (Protein Data Bank (PDB): 7MBQ), superimposed using MHR3/4 (residues 391–687 in TRPM4 and 332–627 in TRPM5). The root mean squared deviation is shown for domains. b, The Cawarm site; interactions between Ca2+ and coordinating residues are indicated by grey lines. c, Comparison of the Cawarm site in the TRPM4 cold (blue) and warm (yellow) conformations by aligning helix α13 (residues 374–382), which forms half of the site. The angle between α12, which forms the other half of the site, is indicated by the black arrow. d, Whole-cell voltage-clamped currents were measured using patch pipettes containing 1 µM free calcium in tsA cells overexpressing wild-type TRPM4 (left) and the TRPM4(E396A) variant (right) at 37 °C. A protocol was applied every 5 s to monitor current changes, initiating at −100 mV for 50 ms, ramping to +100 mV over 200 ms, then maintaining +100 mV for 50 ms; the holding potential was 0 mV. The black and orange traces represent average peak and steady-state current traces (n = 5 each for WT and E396A). e, After reaching the steady-state current in d, additional measurements were made using a multistep voltage-clamp protocol from −120 mV to 160 mV. Representative traces are shown in Extended Data Fig. 1a,b. Current amplitudes at the end of each pulse are plotted as a function of clamp voltage. n = 16 (22 °C, WT), n = 15 (37 °C, WT), n = 8 (22 °C, E396A) and n = 7 (37 °C, E396A) particles. Data are mean ± s.e.m. Statistical analysis was performed using two-way analysis of variance (ANOVA) with Bonferroni’s post hoc test; *P < 0.05, **P < 0.01, ***P < 0.001. The P values for the 60 to 160 mV steps on the left are as follows: P = 0.0015, P< 0.0001, P < 0.0001, P < 0.0001, P < 0.0001 and P < 0.0001, respectively.

To investigate the role of Cawarm in TRPM4 channel function, we mutated a key residue in the binding site, Glu396, to alanine and performed electrophysiological studies. The E396A mutant displayed two strong phenotypes compared with the wild type: the loss of the temperature-induced potentiation and the inability to activate at negative membrane potentials (Fig. 3d,e and Extended Data Fig. 1b,d,f). The crucial role of Glu396 in coordinating Cawarm led us to propose that these phenotypes stem from disrupted Ca2+ binding at this site. To test this idea, we determined the structure of TRPM4(E396A) at 37 °C in the presence of saturating Ca2+, yielding exclusively the cold conformation (Fig. 3a and Extended Data Fig. 6c). Accordingly, the Cawarm site was absent, despite a pronounced Ca2+ signal at the Ca TMD site, suggesting that the two Ca2+-binding sites are not allosterically coupled. This diverges from zebrafish TRPM5, of which abolishing Ca2+ binding at the Ca ICD site markedly reduced the Ca TMD binding43.

Together, our data demonstrate that the Cawarm site and physiological temperature cooperatively have an indispensable role in cold-to-warm transition of TRPM4, with the warm conformation being critical for TRPM4 activation under physiological conditions (Fig. 3d). This is particularly relevant as TRPM4 is expressed in various non-excitable cells with physiological membrane potentials that are normally negative. Even in excitable cardiomyocytes, TRPM4 contributes to cardiac electrical activity through its activation at negative membrane potentials44.

Temperature determines binding of DVT

At physiological temperature, although the agonist Ca2+ triggered substantial structural rearrangements in the ICD of TRPM4, the ion-conducting pore remained closed. This aligns with the fact that TRPM4 currents became outwardly rectifying shortly after initial exposure to intracellular Ca2+, indicating a low open probability at zero or negative potentials (Fig. 3d,e). Our previous effort to capture an open-state structure at low temperatures using DVT was unsuccessful10, despite the identification of two DVT-binding sites, DVTcold1 and DVTcold2 (Fig. 4a). This earlier discrepancy between structural and functional data now makes sense in light of our new data showing that mutations at these sites did not affect the modulatory effect of DVT (Fig. 4a and Extended Data Fig. 8a–i). This observation suggests that these sites are functionally irrelevant, and that DVT may bind non-specifically to negatively charged cavities.

Fig. 4: Temperature determines DVT binding and modulation.
figure 4

a,b, The rectification index (RI) ratio for TRPM4 wild type and mutants at the two DVTcold sites (a) and mutants at the DVTwarm site (b). The RI is defined as the current ratio of I(−120 mV)/I(+120 mV), and the RI ratio is defined as RI(+DVT)/RI(−DVT). Statistical analysis was performed using one-way ANOVA with Bonferroni’s post hoc test (WT was compared with each mutant). The P values in b from left (R597A) to right are as follows: P = 0.0013, P > 0.9999, P = 0.0042, P = 0.0100, P = 0.0085, P = 0.0122 and P = 0.0042, respectively. Representative traces are provided in Extended Data Fig. 8. Insets: the structure of Ca2+/DVT–TRPM4cold (a) and Ca2+/DVT–TRPM4warm (b) as a surface representation, with one subunit as a cartoon and DVT molecules as yellow spheres. The magnified view in b shows the interactions within the DVTwarm site, which is encompassed by the MHR3/4 domain, pore domain and TRP helix of one subunit, along with the S4–S5 linker of the adjacent subunit. The DVT molecule is shown as sticks with a transparent surface, and the surrounding positively charged residues are shown as sticks. c,d, Comparisons of Ca2+-bound cold (blue) versus warm (yellow) conformations (c), and Ca2+-bound warm versus Ca2+/DVT-bound warm conformations (d), by aligning the tetrameric pore helix and loop (residues 958–989). A single subunit is depicted, with the centre-of-mass movement of the MHR1/2 domain indicated to represent the motion of the ICD. The centre-of-mass movement of the S1–S4 movement is also indicated. e,f, The pore domain in the Ca2+-bound warm (e) and Ca2+/DVT-bound warm (f) conformations viewed from the intracellular side. The movements of the S6 helix, TRP helix and S4–S5 linker caused by DVT binding are indicated. The side chain of Ile1040, which forms the channel gate, is shown as sticks. The DVT molecule is shown as sticks with a transparent surface.

Prompted by the identification of the warm conformation, we revisited this discrepancy by determining TRPM4 with Ca2+ and DVT at 37 °C, revealing both cold and warm conformations (Extended Data Fig. 4). While the cold conformation retained the two DVTcold-binding sites, the warm conformation featured a robust DVT density in a different cavity at the TMD–ICD interface, termed DVTwarm (Fig. 4b). To assess the functional relevance of the DVTwarm site, we neutralized the charge on each of the seven positively charged residues and performed electrophysiological experiments. Notably, apart from K928A, which did not generate Ca2+-activated currents, six out of the remaining seven mutants became insensitive to DVT’s voltage modulation effect (Fig. 4b and Extended Data Fig. 8a,j–p). Our data therefore support that the DVTwarm site is the relevant site for voltage modulation. This DVTwarm cavity, which is present in all TRPM channels, may represent a universal drug site for modulating TRPM family channels.

To understand how temperature allocates DVT into different sites in TRPM4, we analysed the electrostatic surface potential of the cold and warm conformations, considering that DVT is strongly negatively charged (Extended Data Fig. 9). We found that the structural rearrangement of the ICD from cold to warm reversed the electrostatic surface potential of the DVTcold sites from positive to negative, effectively preventing DVT from binding to these sites. On the other hand, the DVTwarm cavity remained positively charged in both conformations. However, it became compatible with DVT binding only when the cavity is narrowed—probably to better accommodate the size of a DVT molecule—as MHR3/4 moved towards the TMD after the cold-to-warm transition.

TRPM4 channel opening at 37 °C

To elucidate the activation mechanism of TRPM4 at physiological temperatures, we compared the structural differences between Ca2+–TRPM4cold and Ca2+–TRPM4warm (Fig. 4c), as well as between Ca2+–TRPM4warm and Ca2+/DVT–TRPM4warm (Fig. 4d). This analysis revealed that physiological temperature and Ca2+ synergistically triggered an upward swing of the ICD, creating the Ca2+warm site and priming the DVTwarm site for binding (Fig. 4c). However, this movement was confined to the ICD, leaving the ICD–TMD interface and the TMD largely unchanged. This suggests the need for other factors, such as membrane potential or allosteric modulators, to facilitate channel opening, as demonstrated by DVTwarm binding.

DVTwarm, densely packed with six negative charges, is situated between the ICD and TMD near the cytoplasmic surface of the membrane, attracting the positively charged membrane-facing side of the MHR3/4 domain towards it. As a result, the entire ICD moved towards the TMD in a rigid-body manner, elevating the S1–S4 domain towards the extracellular side through the TRP helix that bridges the ICD to the TMD (Fig. 4d). Meanwhile, the pull from the four DVTwarm molecules on the positively charged TRP helix and S4–S5 linker caused the pore-lining S6 helix to move outwards, therefore opening the ion-conducting pore (Fig. 4e,f and Supplementary Video 1).

Temperature dictates ATP binding

We next extended our studies to a third type of ligand, ATP, a known endogenous TRPM4 inhibitor that also binds to its ICD9,11,33. At room temperature, ATP effectively inhibited the Ca2+-induced TRPM4 current with a half-maximum inhibitory concentration (IC50) in the micromolar range9,11,33 (Fig. 5a,b). This was puzzling because it implies that TRPM4 would be constantly inhibited under cellular conditions in which the concentration of cytosolic free ATP is an order of magnitude higher than the IC50 (ref. 45). However, when we conducted electrophysiological experiments at 37 °C, we observed a substantial decrease in ATP’s inhibitory effect, with less than 50% of the current inhibited even at millimolar concentrations (Fig. 5a,b). This result indicates that ATP is not a potent inhibitor under physiological conditions, emphasizing the important role of temperature in modulating the inhibitory potency of ATP.

Fig. 5: Temperature dictates inhibitor binding and action.
figure 5

a, Whole-cell currents were measured in tsA cells overexpressing wild-type TRPM4 at 22 °C (left) and 37 °C (right). A protocol was applied every 5 s to monitor current changes, initiating at −100 mV for 50 ms, then ramping to +100 mV over 200 ms, and finally holding at +100 mV for 50 ms. The black and orange traces represent the average steady-state current traces measured with 1 µM free Ca2+ and 1 µM free Ca2+ plus 5 mM ATP, respectively, in the pipette solution. n = 8 (−ATP, 22 °C), n = 8 (+ATP, 22 °C), n = 11 (−ATP, 37 °C) and n = 9 (+ATP, 37 °C). b, Normalized current amplitudes in the presence and absence of ATP (at 50 ms of holding potentials of +100 and −100 mV) of the experiments in a were plotted. Each point represents a single cell and the bars represent the mean. c, The structure of Ca2+/ATP–TRPM4cold (left) and Ca2+/ATP–TRPM4warm (right) as a surface representation, with one subunit as a cartoon and ATP molecules as orange spheres. The dashed white circle marks the position of the ATPcold site hypothetically mapped in the Ca2+/ATP–TRPM4warm structure, with its distance to the ATPwarm site indicated. d, A magnified view of the interactions within the ATPwarm site. One subunit is coloured grey, while the adjacent subunit is coloured magenta. The ATP molecule is shown as sticks with a transparent surface, and the surrounding residues are shown as sticks.

Motivated by the identification of the DVTwarm site, we examined the possibility that ATP may also exhibit a temperature-dependent binding and inhibition mechanism. Cryo-EM analysis of TRPM4 prepared with saturating Ca2+ and ATP at 37 °C yielded both cold and warm conformations (Extended Data Fig. 5). Importantly, we observed that ATP indeed occupied different locations in these two conformations (Fig. 5c). In the cold conformation, ATP bound at the interface between the MHR1/2 domain and the adjacent MHR3/4 domain, consistent with the previous findings11. At this ATPcold site, the adenosine moiety of ATP was tightly encapsulated in a cleft in the MHR1/2 domain, forming extensive hydrophilic and hydrophobic interactions with surrounding residues, while its triphosphate tail loosely engaged with a positively charged cap (Arg421) on the adjacent MHR3/4 domain11. The transition to the warm conformation shifted the MHR1/2 domain toward the TMD, moving the ‘adenosine cleft’ to near the N-terminal tip of the rib helix. Notably, this movement also transferred ATP within the cleft to this new location, ATPwarm (Fig. 5c). Here, while the interactions of the adenosine group within the MHR1/2 domain remained essentially unchanged, the triphosphate tail became surrounded by positively charged residues from the rib helix (Fig. 5d).

Comparing the structures of the non-conducting warm conformation with and without ATP revealed that ATPwarm binding did not induce major conformational changes. Instead, it appears that ATP binding locked the channel in a non-conducting warm conformation by impeding the ICD movement necessary for channel activation (Fig. 4d). To test this hypothesis, we determined the structure of TRPM4 with Ca2+, DVT and ATP at 37 °C (Extended Data Fig. 6d). Indeed, this structure closely resembled the ATP-bound non-conducting warm conformation, but not the DVT-bound warm open conformation, with a weak ‘dust-like’ density at the DVTwarm site and a closed pore (Extended Data Fig. 6e–g). We speculate that ATPcold similarly inhibits the channel at lower temperature by locking it in a non-conducting cold conformation. Furthermore, it is possible that ATPcold binding has a higher affinity compared with ATPwarm, which may explain ATP’s increased inhibitory potency at room temperature. Together, our data suggest that, although the adenosine cleft has a critical role in ATP recognition, the mechanism of ATP inhibition is determined by its spatial location, which is temperature dependent.

Discussion and conclusion

Our study of the temperature-sensitive TRPM4 channel at physiological temperatures offers critical insights into its temperature-dependent structural dynamics, ligand recognition and gating mechanisms. A key finding is the warm conformation, which differs from the cold conformation observed at lower, non-physiological temperatures. Our analysis—although currently limited to the biophysical level—suggests that both cold and warm conformations coexist in physiological conditions, with their equilibrium modulated by the local Ca2+ concentration. The identification of the open-state structure of TRPM4 at physiological temperature underscores the importance of the cold-to-warm transition in channel activation. However, the precise interplay of temperature and Ca2+ during this transition remains to be further explored. A plausible explanation is that elevated temperatures may weaken the intersubunit interface in the cold conformation46,47, potentially facilitating, in conjunction with Cawarm binding, the transition to the warm conformation with new intersubunit interfaces. The cold-to-warm transition of TRPM4 also profoundly affects its ligand recognition, highlighted by the identification of temperature-dependent binding for ligands across three principal categories: Ca2+, the endogenous agonist; DVT, a positive modulator; and ATP, an inhibitor. Importantly, we provide compelling evidence that these sites—in contrast to those observed or absent in the cold conformation—are functionally relevant.

It is important to acknowledge that discrepancies may arise between the intended and actual protein temperatures due to the limitations of the cryo-EM sample-freezing process. Indeed, the conformational dynamics captured in the cryo-EM data are likely to reflect a temperature lower than 37 °C at which the sample was prepared. We therefore propose that the warm conformation may exist even below 37 °C. This notion is supported, in part, by our electrophysiological studies on DVT using an inside-out configuration at room temperature (as conducting these experiments at 37 °C posed practical challenges, possibly due to the fragility of excised membrane at elevated temperatures). In these experiments, we were able to discern the phenotypes of mutations at the DVTwarm site, suggesting that the warm conformation is present in the cellular environment at room temperature.

In conclusion, our study elucidates the temperature-dependent mechanisms of ligand recognition and gating of TRPM4 and, more broadly, underscores the importance of considering temperature as a pivotal factor in both mechanistic studies and drug development. Moreover, while a direct role of TRPM4 in temperature sensing is yet to be established, its marked structural similarity to well-known thermosensors in the same family, including the heat sensor TRPM3 and the cool senor TRPM827,48,49,50, hints at a universal molecular framework for how these proteins sense temperature changes.


Human TRPM4 protein expression and purification

The gene encoding human full-length TRPM4 (UniProtKB: Q8TD43) was subcloned into pEG BacMam vector with a 2×Strep tag, GFP and a thrombin-cleavage site at the N terminus51. Bacmid and baculovirus of TRPM4 in a BacMam vector were generated, and P2 viruses were used to infect a suspension of tsA cells. Cells were incubated at 37 °C for 12 h. Subsequently, 10 mM sodium butyrate was added to the culture and the temperature was changed to 30 °C. The cells were collected 72 h after infection and resuspended in a buffer containing 150 mM NaCl and 20 mM Tris pH 8.0 (TBS buffer) in the presence of 1 mM phenylmethylsulphonyl fluoride, 0.8 μM aprotinin, 2 μg ml−1 leupeptin and 2 mM pepstatin A. The cells were lysed by sonication. The membrane fraction was collected by centrifugation at 186,000g using a 45 Ti rotor (Beckman Coulter) for 1 h at 4 °C. It was then homogenized with a Dounce homogenizer in TBS buffer supplemented with protease inhibitors. The protein was extracted from the membrane with TBS buffer supplemented with 1% GDN and protease inhibitors for 3 h at 4 °C. The solubilized proteins were loaded to Strep-Tactin resin. After washing with TBS buffer supplemented with 0.02% GDN, TRPM4 was eluted with the same buffer, supplemented with 10 mM desthiobiotin. The proteins were concentrated and further purified by size-exclusion chromatography (Superose 6). The peak fractions containing the channel were pooled and concentrated to 6 mg ml−1.

EM sample preparation and data acquisition

For sample preparation at 37 °C, purified TRPM4 was incubated with 5 mM calcium chloride or 5 mM EDTA, as required for each experiment, for 10 s. For samples requiring DVT or ATP, TRPM4 mixed with 5 mM calcium chloride was incubated at 37 °C for 30 s before the addition of DVT or ATP to final concentrations of 2 or 5 mM, respectively, followed by further incubation for 2 min. In experiments including both DVT and ATP, TRPM4 was first incubated with 5 mM calcium chloride at 37 °C for 30 s, then with 1 mM DVT for 2 min, and finally with 2 mM ATP for an additional 2 min. For all conditions, 2.5 μl of the 6 mg ml−1 sample was applied to a glow-discharged Quantifoil holey carbon grid (gold, 2/1 μm size/hole space, 300 mesh). The grids were blotted for 1.5 s in the Vitrobot Mark III set to 100% humidity and 37 °C with a 15 s wait time before being plunge-frozen into liquid ethane cooled by liquid nitrogen.

For the Ca2+-containing sample at 18 °C, purified TRPM4 was incubated with 5 mM calcium chloride at 18 °C for 10 s. After incubation, 2.5 μl of the 6 mg ml−1 sample was applied to a glow-discharged Quantifoil holey carbon grid (gold, 2/1 μm size/hole space, 300 mesh). The grids were then blotted for 1.5 s in the Vitrobot Mark III set to 100% humidity and 18 °C, with a 15 s wait time, before being plunge-frozen into liquid ethane cooled by liquid nitrogen.

Images for all samples, except for those detailed below, were recorded using the FEI Titan Krios electron microscope at 300 kV and a nominal magnification of 105,000×. Data were collected on the Gatan K3 Summit direct electron detector in super-resolution mode, resulting in a binned pixel size of 0.828 Å, using SerialEM for automated acquisition52. Nominal defocus ranged from −1.2 to −1.9 μm.

For the Ca2+-containing sample prepared at 18 °C, images were recorded on an FEI Arctica electron microscope at 200 kV and a nominal magnification of ×45,000. A Gatan K2 Summit direct electron detector in super-resolution mode was used, yielding a binned pixel size of 1.16 Å, with SerialEM for automated acquisition. Nominal defocus ranged from −1.4 to −2.2 μm.

For the Ca2+/DVT/ATP-containing sample prepared at 37 °C, images were recorded on the FEI Arctica electron microscope at 200 kV and a nominal magnification of ×45,000. A Gatan K2 Summit direct electron detector in super-resolution mode was used, yielding a binned pixel size of 0.92 Å, with SerialEM for automated acquisition. Nominal defocus ranged from −1.4 to −2.2 μm.

Cryo-electron microscopy data analysis procedure

The detailed workflow for the data-processing procedure is summarized in Extended Data Figs. 37. In general, the raw super-resolution .tif video files for each dataset were motion-corrected and 2× binned using MotionCor2 (v.1.1.0)53. The per-micrograph defocus values were estimated using ctffind (v.4.1.10)54. Particle picking was performed using gautomatch (v.0.56) ( or topaz (v.0.2.4)55 or RELION’s template picking56. Junk particles were removed by rounds of 3D heterogeneous refinement using CryoSPARC57. Good particles were selected for homogeneous refinement with C4 symmetry in CryoSPARC to generate a 3D map. Multiple rounds of CTF refinement and Bayesian polishing were performed in RELION to further improve the map resolution.

For datasets showing noticeable conformational heterogeneity in the ICD, symmetry expansion at the single-subunit level was done from the map refined with C4 symmetry, followed by subtraction of the ICD. The subtracted images of the ICD underwent a round of local refinement with C1 symmetry followed by 3D classification without image alignment in RELION, yielding two distinct conformations—warm and cold. A similar procedure was used to further improve the local resolution of the MHR1/2 domain. The best parts of all focused maps were combined in phenix58 and used for model building. Homotetrameric particles of the warm conformation were identified and refined to generate the final tetrameric map.

For the Ca2+/DVT/37 °C dataset, symmetry expansion at the single subunit level was done from the map refined with C4 symmetry, followed by monomer subtraction. The subtracted images of the monomer underwent a round of local refinement with C1 symmetry followed by 3D classification without image alignment in RELION. Classes that showed an open-pore conformation with strong DVT density were selected. Homotetrameric particles of this conformation were identified and refined to generate a final tetrameric map.

Map resolution estimates were based on the gold standard Fourier shell correlation 0.143 criterion for all datasets56.

Model building

Atomic models were generated by rigid-body fitting of the TMD, MHR1/2 and MHR3/4 and C-terminal domains from a published human TRPM4 model (PDB: 5WP6) into the final cryo-EM maps. Ligands were fitted into the density through real-space refinement using COOT59. The CIF file of DVT was generated using eLBOW60. DVT molecules, lacking distinctive features, were aligned within the density so their long axes corresponded with those of the densities. The models were then manually adjusted in COOT and subjected to phenix.real_space_refine to improve the model metrics. The final models were validated using phenix.molprobity61. Figures were generated using PyMOL (Schrödinger LLC) and UCSF ChimeraX62. The profiles of the ion-conducting pore were calculated using HOLE63.


TsA201 cells expressing plasmids encoding N-terminal GFP-tagged human TRPM4 were used. Then, 1 day (for whole-cell recording) or 2 days (for inside-out recording) after transfection with plasmid DNA (100 ng μl−1) and Lipofectamine 2000 (Invitrogen, 11668019), the cells were trypsinized and replated onto poly-l-lysine-coated (Sigma-Aldrich, 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). Signals were amplified using the Multiclamp 700B amplifier and digitized using the 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. A typical TRPM4 current shows a transient activation by intracellular Ca2+ and subsequent desensitization to reach a steady state within 1–2 min (ref. 64). The ramp pulse from −100 to 100 mV for 200 ms with a holding potential of 0 mV was continuously applied to the cell membrane every 5 s to monitor the transient of the TRPM4 current. The step pulse from −120 to 160 mV for 200 ms was applied after steady state was reached. The inside-out patch was performed at room temperature (21–23 °C) to study the effect of DVT. A 200 ms step pulse from 160 mV to −120 mV (intracellular side relative to extracellular side) was applied. Electrical signals were digitized at 10 kHz and filtered at 2 kHz. Recordings were analysed using Clampfit v.11.3 (Axon Instruments), GraphPad Prism 10 and OriginPro 2024 (OriginLab). The standard bath solution contains 150 mM NaCl, 5 mM KCl, 1 mM MgCl2, 2 mM CaCl2, 12 mM mannitol, 10 mM HEPES, pH 7.4 with NaOH. For a whole-cell recording, the extracellular solution contains 150 mM NaCl, 10 mM HEPES, 1 mM MgCl2 and 2 mM CaCl2. The intracellular solution contains 150 mM NaCl, 10 mM HEPES, 5 mM EGTA, 4.45 mM CaCl2 (1 μM of free Ca2+). 5 mM di-sodium ATP is included in the intracellular solution described above to test the effect of ATP. In the inside-out patch mode, the extracellular solution (pipette) contains 150 mM NaCl, 10 mM HEPES, 1 mM MgCl2 and 2 mM CaCl2. The intracellular solution (bath) contains 150 mM NaCl, 10 mM HEPES, 5 mM EGTA. 5 mM EGTA was replaced by 5 mM CaCl2 (or 5 mM CaCl2 + 10 μM DVT).

Reporting summary

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