Heat- and PIP2-dependent TRPM4 activity underlies mutually exclusive human diseases

Significance

Our study uncovered a signaling network that connects four universal and powerful cellular signaling systems—intracellular Ca²⁺, membrane potential, phosphatidylinositol 4,5-bisphosphate (PIP2), and temperature—linked together by TRPM4, a broadly expressed, Ca²⁺-activated, temperature-sensitive ion channel. We identified PIP2 as a key factor that locks or unlocks TRPM4 activity by modulating the channel’s Ca²⁺ sensitivity. We further proposed the high-affinity PIP2-binding site and key channel residues involved in PIP2 binding and gating transduction. TRPM4 is an important physiological regulator in numerous organs. Gain-of-function mutations in TRPM4 cause severe human hereditary diseases affecting the heart and the skin in a mutually exclusive manner. Our study provides an explanation for this unique phenomenon, which arises from organ-specific regulation within the multidimensional signaling network we have uncovered.

Abstract

Temperature-sensitive transient receptor potential melastatin subfamily 4 (TRPM4) ion channels convert intracellular calcium increases into membrane depolarization, thereby linking these two powerful cellular signaling pathways in diverse physiological processes. TRPM4 mutations cause severe human hereditary cardiovascular and skin diseases; mysteriously, while these mutations are gain-of-function in nature, they affect the heart and the skin in a mutually exclusive manner. Here, we show that phosphatidylinositol 4,5-bisphosphate (PIP2) lipid is a required cofactor for TRPM4 activity by tightly regulating its calcium sensitivity. We detected two PIP2 binding sites and located the high-affinity site adjacent to the S4-S5 linker. We demonstrated that skin disease–associated TRPM4 mutations relieve the tight control of PIP2, resulting in elevated channel activity but only at the body surface temperature. In contrast, heart diseases are associated with mutations known to boost the number of channels, an effect we found to be annihilated by channel desensitization outside the body core. Indeed, dendritic cells from transgenic mice carrying a skin disease mutant exhibited elevated migration at 25-to-30°C range compared to those from normal mice, but no difference was observed at 37°C. These findings shed light on a molecular mechanism for dynamic regulation of cellular signaling in physiology and diseases.

TRPM4 and its homolog TRPM5 are calcium-sensing ion channels, with intracellular calcium ions being the only known physiological activator (1–3). They serve a unique function in physiology, that is, converting increases in the intracellular calcium concentration into membrane depolarization (4). This is because TRPM4 and TRPM5 are selectively permeable to monovalent cations Na⁺ and K⁺ (a distinctive feature among all TRP channels) (5). While other calcium-sensing ion channels do exist (e.g., calcium-activated big and small conductance potassium channels, BK and SK, and calcium-activated chloride channels, CaCC), activities of these channels lead to membrane hyperpolarization due to their selective permeation to K⁺ or Cl⁻ (6). By directly linking calcium signaling to electrical signaling, TRPM4 and TRPM5 are poised to have a powerful impact on cellular physiology. Activities of these channels are also influenced by temperature (7), which markedly shifts the equilibrium between the closed and open conformations (8).

TRPM4 is broadly expressed in mammalian tissues. It is involved in diverse physiological functions, such as cardiovascular activities, immune responses, pancreas release, skin health (4). The critical role that TRPM4 plays in physiology is highlighted by findings that numerous human hereditary heart diseases, including Brugada syndrome and long-QT syndrome, are linked to gain-of-function (GOF) mutations that increase channel density on the cell surface (9, 10). Severe human skin diseases such as progressive symmetric erythrokeratoderma (PSEK) and inflammatory linear verrucous epidermal nevus were also found to be caused by TRPM4 GOF mutations that enhance the channel’s activation gating (11, 12). Strikingly, none of the PSEK patients exhibited detectable heart defects, whereas the heart disease patients were spared of skin problems. Pathogenicity of these GOF mutations is mostly unexplained. Our data reveal tight links among three key regulators of TRPM4 activity—calcium, PIP2, and temperature—which form an intricate network for dynamic regulation of cellular signaling and offer a plausible explanation for how genetic mutations lead to mutually exclusive cardiac and skin phenotypes.

Results

Controlling TRPM4 Calcium Sensitivity by PIP2.

The calcium-sensing function is central for TRPM4 to serve its physiological roles. However, reports since its cloning are dominated by a puzzling observation—the observed calcium concentrations needed to induce channel activation were often higher than 100 μM, which would be generally too high to reach under physiological conditions (4). There was evidence that lower calcium concentrations would suffice in intact cells, and several labs have observed rapid and prominent channel desensitization during patch-clamp recordings (13–15). It suggests that, upon cellular perturbation, TRPM4 undergoes certain changes, such as a gradual loss of ancillary components, for which PIP2 is a strong candidate (16).

TRPM4 desensitization observed in patch-clamp recordings is a rapid decline in current amplitude upon forming the inside-out configuration (13, 14) (Fig. 1A). For human TRPM4 channels expressed in HEK cells and recorded with 3 mM Ca²⁺, we measured a time constant of 5.9 ± 1.3 s (n = 10) for this decline (Fig. 1 A and F), which stabilized at 68.0 ± 3.2% (n = 10) of the peak current amplitude (Fig. 1 A and E). The decline was due to a decrease in calcium sensitivity, as could be seen from a substantial lower and right-shifted calcium concentration–dependence relationship, with an increase in the EC₅₀ value for calcium from 31.5 ± 8.8 µM (n = 6) to 122.3 ± 7.8 µM (n = 7) (Fig. 1B). In agreement with previous observations (13, 14), supplying PI(4, 5)P2 (the predominant phosphoinositide lipid species in plasma membrane (17), referred to simply as PIP2 unless specification is needed) through the perfusate could prevent current decline, or recover current after the decline had occurred (SI Appendix, Fig. S1), confirming that this phase of desensitization (Phase I) is due to a gradual loss of channel-associated PIP2.

Fig. 1.

The unique desensitization profile of hTRPM4. (A) Representative current trace elicited by an application of 3 mM Ca²⁺ alone (gray bar), or together with 10 µg/mL PLL (pink bar), 50 µM PI(4,5)P2 (blue bar), or 50 µM PI(3,5)P2 (cyan bar), recorded at +80 mV from an inside-out patch of a cell expressing WT hTRPM4. The current desensitization process exhibited two distinct phases (Phase I and Phase II). PI(3,5)P2 recovered current to a lower amplitude compared to PI(4,5)P2. (B) Ca²⁺ sensitivity of WT hTRPM4 during initial activation (black circles), after Phase I desensitization (orange circles), after 1-s PLL perfusion (light pink circles), and after complete Phase II desensitization (dark pink circles). Data were normalized to the peak current at +80 mV during initial activation and fitted to a Hill equation (dotted curves). n = 5-to-7. (C) Schematic drawing for PI(4,5)P2 depletion by Dr-VSP. The percentage inhibition of TRPM4 current density at −100 mV and +80 mV was measured before and after Dr-VSP activation. To acquire the current of wild type TRPM4 (black, n = 6-to-7) and the mutant K928A (red, n = 6-to-7), a voltage ramp protocol was applied from −120 mV to +80 mV for 150 ms. To assess the PI(4,5)P2 sensitivity of the channel, the percentage of current density inhibition at −100 mV and +80 mV was calculated. All current data were normalized to the current density of the endogenous leak channel in HEK293T cells. The significance between wild type and K928A was assessed using Student’s t test. ***P < 0.001. (D) Positively charged or polar residues in the vicinity of the intracellular membrane surface of hTRPM4 (PDB entry 9B8Y) were subject to the alanine-scanning mutagenesis. A cluster of residues that showed substantial increases in the speed and extent of desensitization when mutated to alanine (as shown in Fig. 1 E and F) are colored in magenta in one subunit. Those insensitive to the effect of mutation are shown in gray. The transparent light blue rectangle indicates the location of the plasma membrane. (E–G) Summary of functional characterizations of alanine mutants for the reduction of current during Phase I (E), the rate of Phase I desensitization (F), and recovery ratio by PIP2 (G), all conducted with 3 mM Ca²⁺. The WT and those exhibiting a significant change are highlighted in blue and magenta, respectively, in these violin plots. K928A (red arrow) yielded no current. Individual data points are shown as circles (n = 3-to-10). Functional differences versus the WT group were tested using one‐way ANOVA followed by Tukey’s post hoc test. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. WT, wild type; PLL, poly-L-lysine. (H) Representative current trace of the K919A mutant elicited by applications of 3 mM Ca²⁺ alone (gray bar), or with PI(4,5)P2 (dark blue bar) at +80 mV. Note that the K919A mutant Phase I desensitization is near complete. (I) Confocal images of HEK293T cells expressing GFP-tagged TRPM4 wild-type (Top) or K928A mutant (Bottom) stained with plasma membrane dye WGA-Alexa-647.

Two Phases of PIP2 Dissociation.

After Phase I desensitization, perfusing poly-L-lysine (PLL) to inside-out patches caused a further rapid decline that eliminated the calcium-activated current (Fig. 1A). We noticed that the calcium sensitivity continued to decrease along the course of this Phase II current decline (Fig. 1B). Consistent with the assumption that PLL served as a chelator for PIP2 in such treatment (13, 14), supplying PIP2 could recover current but normally only to the level of Phase I desensitization (Fig. 1A). PI(3,5)P2, a structural derivative of PI(4,5)P2, was less effective in recovering the current (Fig. 1A). Decavanadate (DVT), a negative charge-rich molecule used as a surrogate for PIP2 (18, 19), could also slightly recover current from completely desensitized TRPM4 channels (SI Appendix, Fig. S2).

To further confirm the role of PIP2 in supporting TRPM4 activation, we coexpressed a voltage-sensing phosphatase from Danio rerio, Dr-VSP, which removes the 5’-phosphate group from PI(4,5)P2 producing phosphatidylinositol 4-phosphate (PI4P) (20, 21). Current density after activation of Dr-VSP measured at both −100 mV and +80 mV showed nearly 90% inhibition (Fig. 1C). Dramatic inhibition of wild-type (WT) hTRPM4 current was also observed during the protocol for Dr-VSP activation, which involved a +120-mV step for 1 s (SI Appendix, Fig. S3C). Phosphatidylinositol(3, 4, 5)-trisphosphate (PI(3, 4, 5)P3) also regulates ion channels (22, 23). To investigate the potential involvement of PI(3, 4, 5)P3 in TRPM4 activation, we conducted a similar experiment using Ci-VSPTEN, a chimeric construct conjugating the voltage-sensing domain of Ciona intestinalis and PTEN (SI Appendix, Fig. S3D). However, no inhibition of WT hTRPM4 was observed after PI(3, 4, 5)P3 depletion, indicating that it is not involved in TRPM4 channel activation (SI Appendix, Fig. S3 E and F).

Coexpressing PIP5KIγ, a kinase that promotes PIP2 synthesis (24), had the opposite effect to Dr-VSP. The Phase I desensitization was substantially delayed, though its extent was boosted, as if overexpressed TRPM4 channels in HEK cells were not saturatedly associated with PIP2 unless coexpressed with PIP5KIγ (SI Appendix, Fig. S4). Taken together, our results revealed that PIP2 is a required cofactor for TRPM4 activation, and it exerts its functional role through tight regulation of the channel’s calcium sensitivity. Both 4’ and 5’ phosphate of PIP2 appear to contribute to interactions with TRPM4. In addition, there are two types of PIP2 binding sites in TRPM4—low-affinity sites associated with Phase I desensitization and high-affinity sites associated with Phase II desensitization.

A Hotspot for Determining PIP2 Binding.

To search for the PIP2-binding sites, we first conducted alanine-scanning mutagenesis on all positively charged residues in the vicinity of the intracellular membrane surface (Fig. 1D and SI Appendix, Table S1). We found that while TRPM4 activity was insensitive to charge neutralization mutation at most positions, it exhibited high sensitivity at a cluster of residues adjacent to the S4-S5 linker, which serves as a gating transducer in many ion channels of a similar architecture: Mutations at this location caused substantial increases in the speed and extent of desensitization (Fig. 1 E–G). Noticeably, currents from K919A mutant declined to zero in a single phase without the need for the PLL treatment (Fig. 1H). Even more dramatic changes were observed in the K928A mutant, which could not be activated despite being expressed in the plasma membrane, as indicated by the strong fluorescence signal from a GFP tag (Fig. 1I). Current density measurements from cells expressing K928A showed neither any change upon Dr-VSP activation (Fig. 1C) nor difference from nontransfected HEK293T cells (SI Appendix, Fig. S3 A, B, and E). These data suggest that mutations at the hotspot (oval in Fig. 1D) may have affected calcium activation through PIP2, making its binding unstable or impossible.

To further confirm that charge-neutralization mutations at the identified hotspot weakened PIP2 binding instead of exerting their gating effect through nonspecific structural perturbations, two polar residues within the hotspot, N913 and Q915, were picked for alanine mutagenesis. We found that indeed N913A and Q915A exhibited phenotypes consistent with weakened PIP2 binding (Fig. 1 E–G). Our results are consistent with a recent report of the open TRPM4 structure (8), for which the negative charge-rich molecule DVT bound at the same location [but not in the closed TRPM4 structures (18)] was required to stabilize the open-state structure. These structural observations lend further support to the idea that the highly negatively charged PIP2 binds to this location to support channel activation.

Two PIP2 Binding Sites Identified In Silico.

None of the published TRPM4 structures contains PIP2. To identify potential PIP2-binding sites, we used two complementary computational approaches. First, we used Rosetta GALigandDock, a genetic algorithm-based docking protocol (25). Our docking simulations identified a preferred binding pocket for PIP2 adjacent to the S4-S5 segment (Fig. 2 A and B). At this position, PIP2 directly interacts with residues K928, Q915, K914, and N913 (Fig. 2A, zoomed-in view, Right). Interestingly, we also observed PIP2 binding at the position DVT occupies in the open state TRPM4 structure, where PIP2 interacts with residues K925, K928, and R664 (Fig. 2A, zoomed-in view, Left). All these residues are within the PIP2 binding hotspot determined functionally (Fig. 1 E–G). Since in silico simulations identify low energy states, it is likely these two adjacent sites represent candidates for the high-affinity PIP2 binding sites.

Fig. 2.

PIP2 binds to a cavity adjacent to the S4-S5 linker. (A&B) Lateral (A) and top–down (B) view of TRPM4 (PDB entry 9B8Y) showing the PIP2 binding pocket identified by GALigandDock. The cavity is located near residue K928 in a cytoplasmic region adjacent to the S4–S5 linker. The two Bottom panels present close-up views of the potential PIP2 poses identified by GALigandDock. (C) Top–down view from the extracellular side of the membrane showing the five PIP2 binding sites from MD simulation. The two Right panels present the close-up views of Site 1 and Site 4 from the 550 ns simulation time point. (D) Average contacts with PIP2 for different TRPM4 residues. Blue bars represent the average contact time across sites 1, 2, 3, and 4.1, pink bars represent the contact time for the site 4.2. (E–G) Summary of functional characterizations of alanine or glutamine mutants for the Phase I current reduction (E), the rate of Phase I desensitization (F), and the rate of recovery by PIP2 (G). Significance was assessed using one‐way ANOVA followed by Tukey’s post hoc test. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 versus WT. (H) Schematic illustrations of WT and K928A-WT dimer (Left) and comparison of Ca²⁺ sensitivities of WT (black circles) and K928A-WT dimer (pink circles) during initial activation (Right). Data were normalized to the peak current at +80 mV during initial activation and fitted to a Hill equation (dotted curves). n = 5-to-6. (I and J) The K928A-WT dimer exhibits substantial increases in the extent (I) and rate (J) of desensitization compared to WT. Data were normalized to the peak current at +80 mV during initial activation. Significance was assessed using Student’s t-test. *P < 0.05; ****P < 0.0001 versus WT.

Second, we used all-atom molecular dynamics (MD) simulations to confirm the protein contacts with PIP2. Four PIP2 molecules were placed at sites approximately 5 Å away from the DVT site in the cryo-EM structure, here referred to as sites 1, 2, 3, and 4, thereby reflecting the fourfold symmetry of the TRPM4 channel. We observed that each of the PIP2 molecules rapidly approached the DVT binding site during the initial phase of the MD simulation and reached a stable binding pose (Fig. 2C and SI Appendix, Figs. S5 and S6). PIP2 forms the most consistent contacts with four basic residues: R601, R664, K925, and K928 (Fig. 2C, zoomed-in view, Top; Fig. 2D and SI Appendix, Fig. S6). The 4’ phosphate of PIP2 contributed significantly more polar interactions than its 5’ phosphate (SI Appendix, Fig. S7).

Encouraged by the Rosetta GAligandDock results, we explored the possibility of introducing two PIP2 molecules in site 4. While one PIP2 went to the DVT site (referred to here as site 4.1, see Fig. 2C), the second PIP2 molecule remained stably bound throughout the simulation and shared some protein contacts with the site 4.1 PIP2 molecule, namely residues R601 and K928. The site 4.2 PIP2 also established some unique contacts with TRPM4 residues N913, K914, and Q915, all of which were implicated by functional tests (Fig. 1 E–G). Site 4.2, overlaps with the other site identified by GALigandDock. Interestingly, K919, another residue seen in functional assays to be important, did not make significant contact with any of the PIP2 molecules directly; instead, it made significant contact with residue Q915, forming a hydrogen-bond network to stabilize PIP2 binding in site 4.2 (SI Appendix, Fig. S8).

Functional Validations of PIP2 Binding Sites.

To validate that PIP2 binds to the presumably high-affinity sites to support channel activation, we further tested selected residues identified in computational analyses with less invasive K→Q mutations (Fig. 2 E–G). We observed that, while K928A eliminated channel activation, a polar substitution to Q at this position (K928Q) recovered current (SI Appendix, Fig. S9A). This mutant channel, however, desensitized more profoundly than the wild type, supporting the predicted role of K928 in directly interacting with PIP2 (Fig. 2 A, C, and D). Likewise, K919Q mutant exhibited strong desensitization phenotypes, presumably by perturbing the H-bond network (SI Appendix, Fig. S8), whereas K914Q mutant exhibited weakly altered phenotypes, similar to those of K914A. We conducted an additional test at the K928 position by constructing a K928A-WT tandem dimer (Fig. 2H and SI Appendix, Fig. S9B). If the K928A single-point mutation indeed eliminated current by preventing PIP2 binding, channels made of the K928A-WT dimer would have two intact PIP2 binding sites in the two wild-type subunits. We observed that, indeed, calcium-activated currents could be elicited in cells expressing this dimeric construct but with substantially weakened sensitivity similar to partially desensitized channels. The dimeric channels also exhibited severely altered desensitization phenotypes (Fig. 2 I and J). All these observations are consistent with in silico findings of the potential PIP2 binding site(s).

In conclusion, our search identified a candidate PIP2 binding location that can be divided into two subsites, one of which overlaps the DVT site. We assume one of the subsites represents the high-affinity PIP2 binding site. While we assume that an unidentified site is responsible for the low-affinity PIP2 binding, it might be possible that transition of PIP2 between the two subsites seen in simulations constitutes Phase I desensitization, and exiting from these sites yields Phase II desensitization. Along with the identification of potential PIP2 binding sites, the regulatory role of PIP2 for calcium sensitivity of TRPM4 started to emerge.

Skin Disease GOF Mutants Evade PIP2 Control.

Our data showed that activation of TRPM4 requires binding of PIP2 to the high-affinity sites; activation is further stabilized by PIP2 binding to the low-affinity sites. Given the tight coupling between PIP2 binding and calcium sensitivity (Fig. 1B), it is anticipated that physiological or pathological changes affecting this coupling would have a strong impact on TRPM4 activation by calcium. We compared two skin disease PSEK mutants with increased calcium sensitivity, I1033M and I1040T (11) (Fig. 3A), to the wild type and the heart GOF mutant E7K that boosts channel expression without affecting gating (9). We observed that the spontaneous (Phase I) desensitization became less pronounced in the skin disease mutants. This is reflected by both a slower pace and a reduced extent of current decline, as shown in Fig. 3B recorded at 300 μM Ca²⁺. These phenotypes were also observed at 3 mM Ca²⁺ (Fig. 3 C and D), a higher concentration that was more convenient for observing Phase II desensitization. Upon applying PLL, we found that I1033M and I1040T currents showed a less prominent Phase II desensitization. A noticeable consequence of this change was that the current did not decline to zero as the wild-type channel current did, that is, PIP2 is no longer a required cofactor for the mutant channels to activate (Fig. 3E). As expected, E7K activated and desensitized like wild-type channels (Fig. 3 C and D).

Fig. 3.

Skin disease–causing TRPM4 GOF mutants yield persistent currents after removing PIP2. (A) Positions of the skin disease mutants (I1033M and I1040T) mapped onto the TRPM4 cryo-EM structure (PDB entry 9B8Y). (B) Representative current traces (Left) and comparison of residual current amplitudes after spontaneous desensitization (Right) between I1033M (pink), I1040T (green), and WT (black) in response to 300 µM Ca²⁺ (dark gray bar). Significance was assessed using one‐way ANOVA followed by Tukey’s post hoc test. ***P < 0.001; ****P < 0.0001 versus the WT group, n = 4-to-6. (C) Representative current traces of WT, I1033M, and E7K elicited by 3 mM Ca²⁺ alone (dark gray bar), or together with 10 µg/mL PLL (pink bar) or 50 µM PIP2 (blue bar) at +80 mV. Dash lines denote the baseline current level in the absence of Ca²⁺. (D) Two skin disease–causing mutants (magenta for I1033M and green for I1040T) exhibit significantly higher current amplitudes compared to WT (black) and E7K mutant (red) after Phase I desensitization (Left), after Phase II desensitization (Middle), and after PIP2 generated recovery (Right). Data were normalized to the peak current at +80 mV during initial activation. Significance was assessed using one‐way ANOVA followed by Tukey’s post hoc test. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 versus WT group. (E) Schematic illustration that I1033M and I1040T mutants escape PIP2 control. (F) Increased Ca²⁺ sensitivities of the mutants, quantified during initial activation (Left), after Phase I (Middle) and Phase II (Right) desensitization compared to the WT group. Data were normalized to the peak current at +80 mV during initial activation and fitted to a Hill equation (n = 4-to-8).

We examined both skin disease mutants for calcium-dependent activation and found they exhibited a clear left-shift in their calcium dependence compared to the wild-type channels (Fig. 3 F, Left and SI Appendix, Fig. S10). Similar observations were made after Phase I desensitization (Fig. 3 F, Middle). After Phase II desensitization, I1033M exhibited clearly observable calcium-dependent activation (Fig. 3 F, Right); I1040T also appeared activatable after the PLL treatment, though the remaining currents were too small for reliable quantification (Fig. 3 F, Right). In summary, the skin disease mutations partially decouple calcium activation and PIP2 regulation in TRPM4.

Sustained Activity Only at Room Temperature.

Escaping tight control of PIP2 would permit GOF mutants such as I1033M to exert sustained activities at the body surface to disrupt skin homeostasis. Indeed, we observed long-lasting currents from I1033M upon continuous calcium application at 25 °C (Room T), while currents from the wild-type and E7K channels dropped quickly (Fig. 4A and SI Appendix, Fig. S11A). Interestingly, this dramatic difference disappeared when the recording temperature was raised to 37 °C to mimic the body core temperature at the heart (Body core T) (Fig. 4B and SI Appendix, Fig. S11B). Therefore, the elevated activity of I1033M at lower than 37 °C (mimicking the skin temperature) (Fig. 4C) is possibly a direct cause of the PSEK disease, potentially through effects on keratinocyte proliferation or through immune mechanisms related to dendritic cell tracking (see below). This effect disappears nonetheless at the body core temperature.

Fig. 4.

Skin disease TRPM4 GOF mutant evades PIP2 control at body surface temperature but not body core temperature. (A and B) Representative current traces of WT (black), I1033M (magenta), and E7K (green) mutants induced by 300 µM Ca²⁺ at room temperature (A) and body core temperature (B). (C) At body core temperature, the skin disease mutant exhibits a marked increase in the extent of desensitization, like WT and E7K mutant, compared to its behavior at room temperature (n = 5-to-12). Functional differences versus the WT group were tested using one‐way ANOVA followed by Tukey’s post hoc test. ****P < 0.0001 versus WT; ns, no significance. (D) Schematic drawing of the chemotaxis assay (Top), comparisons of the number of migrated cells and chemotaxis ratios of BMDCs from WT (black circles), heterozygous (HET, pink circles), and homozygous (HOM, magenta circles) mice at different temperatures (n = 4). Student t test, versus WT CCL21 group. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. (E) Working model of the temperature-dependent mechanism underlying mutual exclusivity of cardiac and skin diseases caused by two types of TRPM4 GOF mutants.

Active Migration of DC Cells Expressing I1033M.

TRPM4 activity has been reported to contribute to mouse bone marrow–derived dendritic cells (BMDCs) migration but not to their maturation (26). Therefore, we tested our working hypothesis in a mouse model by examining how I1033M affected migration of BMDCs. We previously reported that PSEK exhibits many cellular and phenotypic similarities to psoriasis (11). We made a transgenic mouse model carrying the I1029M (equivalent to human I1033M) mutant TRPM4 and showed that they were predisposed to more severe psoriasiform dermatitis and enhanced dendritic cell migration (27). We found that indeed BMDCs from the TRPM4^I1029M mice exhibited much stronger migration at room temperature compared to BMDCs from the wild-type mice (Fig. 4 D, Left). The difference was more prominent at slightly elevated temperatures that likely encouraged cell migration (Fig. 4 D, Middle). Interestingly, such difference began to diminish as the experimental temperature increased further and disappeared completely at 37 °C (Body core T). In conclusion, the behavior of BMDCs mirrored closely the activity of I1033M at these temperatures, offering a plausible explanation of the skin disease pathogenesis.

Discussion

Our data revealed intricate connections of calcium ions, PIP2, and temperature to TRPM4 activity. Collectively, these findings suggest an explanation for the striking mutual exclusivity of heart and skin diseases caused by TRPM4 GOF mutants (Fig. 4E): E7K and other GOF mutants cause heart diseases by boosting the cardiac excitatory current due to the increased number of TRPM4 channels in cardiomyocytes. Such an effect would be canceled in the skin as the activity of these channels is inhibited at the lower body surface temperature. On the other hand, I1033M and I1040T are GOF due to enhanced activation and escape from tight PIP2 control, hence they are capable of yielding sustained excitatory currents at body surface temperature. Indeed, PSEK affects body extremities (hands, feet, etc.) where the skin temperature tends to be the lowest (11). However, at the higher body core temperature in the heart, these mutant channels exhibit similar activities to the wild-type channels and do not alter the heart’s electrophysiology. The rhythmic cardiac calcium waves activate TRPM4 repetitively, unlike the situation in the skin, which would likely amplify the behavior difference discussed above. While details of pathogenesis await future investigation at cellular and systemic levels, findings from the present study shed light on a molecular mechanism for dynamic regulation of cellular signaling in physiology and diseases, in which TRPM4 serves as the linkage between calcium signaling and electrical signaling.

The phospholipid PIP2 is known to be a universal and potent regulator of membrane proteins, including ion channels (17, 28). Similarly, intracellular calcium and membrane potential are two powerful regulators of cellular physiology in virtually all organs and systems (29, 30). TRPM4, as a calcium-activated membrane depolarizer tightly controlled by PIP2, serves the role of a multidirectional signal converter connecting these physiological regulators. The signal conversion function of TRPM4 in maintaining homeostasis or inducing dynamic changes are emerging in a number of physiological systems at the body core or surface, for which temperature acts as another key regulatory factor. The intricate connection among these factors in TRPM4 revealed in the present study offers a framework to better understand these systems and a potential explanation for numerous apparent conflicts in experimental observations in the literature (4).

Materials and Methods

Constructs and Mutagenesis.

The wild-type human TRPM4 (hTRPM4) complementary DNA (cDNA) in pCMV6-AC-GFP vector was kindly provided by Zhimiao Lin (Peking University First Hospital). All point mutations and dimeric constructs were designed using wild-type human TRPM4 plasmid as a template and generated by GenScript. Dr-VSP (GenBank# AB308476) was obtained from Yasushi Okamura at Osaka University, Osaka, Japan; Ci-VSPTEN was distributed by Christian Halaszovich at the University of Aachen, Aachen, Germany; the PIP5KIγ plasmid was kindly provided by Sharona Gordon at University of Washington, Seattle, USA.

Animals.

Wild-type C57BL/6J mice as well as homozygous and heterozygous mice carrying a TRPM4 mutation I1029M (equivalent to the human PSEK disease–causing I1033M mutation) were used in this study. Transgenic mice were generated using the CRSPR-Cas9 technology, as described previously (27). Mice were bred and housed at specific pathogen free (SPF) conditions with standard 12 h light/dark cycle. All animal experiments were performed following the guidelines and approved by the Institutional Animal Care and Use Committee at the University of California, Davis.

Cell Culture and Transfection.

Human embryonic kidney 293T (HEK293T) cells were purchased from American Type Culture Collection (ATCC) for human TRPM4 plasmid expression and patch-clamp recordings. Cells were kept at 37 °C with 5% CO₂ and cultured in Dulbecco’s modified Eagle’s medium (DMEM, Cytiva) supplemented with 10% (vol/vol) fetal bovine serum (GenClone or Cytiva) and 0.2% penicillin/streptomycin (Cytiva). Transient transfections were performed using Lipofectamine 2000 reagent (Invitrogen) with cells grown to 40 to 60% confluence on 25 mm glass coverslips in 35 mm dishes (Fisher Scientific). To assess effects of PIP2 on the channel, 500 ng of TRPM4 cDNA and 1,000 ng of Voltage-Sensing Phosphatase from Danio rerio (Dr-VSP) cDNA or a chimeric construct conjugated with the voltage-sensing domain of Ciona intestinalis and PTEN (Ci-VSPTEN) cDNA were cotransfected. After 24-to-36 h of transfection, cells were plated onto a coverslip coated with 0.1 mg/ml Poly-L-lysine (PLL) (Sigma-Aldrich). Patch-clamp recordings were conducted 18 to 24 h after transfection.

Electrophysiology and Data Analysis.

Most functional data were collected in the cell-free inside-out configuration. The bath and pipette control solutions contained 156 mM NaCl and 10 mM HEPES (pH 7.2). For the calcium-free working solution, 5 mM EGTA was added. The 1 µM, 3 µM, and 10 µM free Ca²⁺ solutions were made by adding the appropriate amounts of CaCl₂ calculated by the MAXCHELATOR program (http://maxchelator.stanford.edu) to a 5 mM EGTA-containing solution. For Ca²⁺ concentrations higher than 10 µM, CaCl₂ was added to an EGTA-free bath solution. PLL was diluted in the bath solution to make a 10 µg/mL working solution. The diC8-PI(4,5)P2 and diC8-PI(3,5)P2 (Echelon Biosciences) stock solutions at 5 mM were made in the bath solution and diluted into working solutions to a final concentration of 50 µM. Patch-clamp recordings in the inside-out configuration were performed at room temperature or at elevated temperatures. To increase the solution temperature, the bath temperature was controlled by an in-line heater plus controller (Warner Instruments TC324B, Harvard Apparatus, Holliston, MA, USA). Temperature was monitored by a thermometer with voltage-proportional output (BAT-12 Microprobe Thermometer, Physitemp Instruments, LLC, Clifton, NJ, USA). Its flexible microprobe (IT-24P, Physitemp Instruments, LLC, Clifton, NJ, USA) was placed near the pipette. An EPC10 USB patch-clamp amplifier controlled by PatchMaster software (HEKA Elektronik) was used. Pipettes were pulled from borosilicate glass (Sutter instrument) using a Sutter P-97 puller and heat-polished using an MF-830 microforge to a final resistance between 1.5 MΩ and 4 MΩ (Narishige Scientific Instruments Lab). A gravity-driven perfusion system (RSC-200, BioLogic) was used to switch between perfusates. Current signal was sampled at 10 kHz and filtered at 2.25 kHz. Currents were elicited by +80 mV pulses of 50 ms duration from a holding potential of 0 mV. This protocol was delivered every second. Averaged current amplitude was estimated during the last quarter of the +80 mV pulse.

Patch-clamp data were exported from the PatchMaster software and analyzed using Igor Pro (WaveMetrics). For calcium concentration–response relationships in the wild-type, mutant, and dimeric channels, the current in the Ca²⁺-free control solution was subtracted from that at various free Ca²⁺ concentrations with or without PLL or PIP2, which were then normalized to the initial current amplitude at +80 mV. Normalized data were fitted to the Hill function. The time constants during Phase I and Phase II desensitization and PIP2-induced recovery phase were obtained by fitting current traces within each phase to an exponential function. Statistical data are shown as mean ± SEM throughout the manuscript. Statistical analyses were performed using GraphPad Prism 8 (GraphPad Software, Inc). Student’s t test was used to compare between two groups. For multiple groups, one-way ANOVA followed by Tukey’s post hoc test were applied. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 versus WT group were considered statistically significant.

Voltage-Sensitive Phosphatase Experiments.

For these recordings, the whole-cell configuration was used. The data were acquired using an EPC-10 amplifier with PatchMaster software (HEKA Elektronik) at room temperature (22-to-25 °C). The bathing solution for current measurement contained (in mM)156 NaCl₂, 5 CaCl₂, 10 Glucose, and 10 HEPES, adjusted pH 7.2 with NaOH. The pipette solution contained 156 CsCl, 1 MgCl₂, 10 EGTA, and 10 HEPES with CsOH for adjusting pH 7.4. For the 10 μM of free Ca²⁺, CaCl₂ was added to pipette solution calculated by the MAXCHELATOR program (http://maxchelator.stanford.edu). Glass pipettes were pulled from fire-polished borosilicate glass (Sutter Instrument) using P-97 micropipette puller (Sutter Instrument) that had a resistance between 2 to 4 MΩ and series resistance errors were compensated by 60%. All voltage protocols were held at 0 mV as holding potential. For measuring PI(4,5)P2 sensitivity, the two ramp protocols were applied before and after depolarization at 120 mV for 1 s to activate Dr-VSP. Additional hyperpolarization pulse of −150 mV for 400 ms was put right after to avoid voltage-dependent inhibition. The ramp protocol was applied from −120 mV to +80 mV for the duration of 150 ms. For the Ci-VSPTEN test, a ramp protocol was used at −120 ms to +50 mV for the same duration. Current density was analyzed by measuring peak current during a voltage step protocol that was applied from −100 mV to +100 mV for 100 ms with 20 mV increment followed by −100 mV for the same duration. Data were analyzed using Igor Pro 6.37 (WaveMetrics). Statistics were shown as mean ± SEM. Statistical analysis for differences between two groups was carried out with Student’s t test using R software (R Core Team, Vienna, Austria). Differences were considered significant at the *P < 0.05; **P < 0.01; and ***P < 0.001.

Confocal Fluorescence Microscopy.

HEK293T cells transiently transfected with either hTRPM4 WT or K928A mutant were imaged in a buffer solution containing 145 mM NaCl, 5 mM KCl, 1 mM CaCl₂, 10 mM glucose, and 10 mM HEPES, pH 7.4. Following media replacement, Wheat Germ Agglutinin (WGA) conjugated to Alexa Fluor 647 (W32466, Invitrogen; 1:50) was added to the buffer and incubated for 40 min in the dark at room temperature to label the plasma membrane. Excess dye was removed by washing with the buffer before imaging. Imaging was performed using a Zeiss LSM 880 confocal microscope (Zeiss) with Airyscan detector and a Plan-Apochromat 63×/1.40 NA oil DIC M27 objective. For detection of GFP-tagged hTRPM4 (WT or K928A), a 488 nm laser was used for excitation, and emission was collected within the 493 to 598 nm range; for WGA conjugated to Alexa Fluor 647, excitation was at 633 nm, with emission collected between 638 to 756 nm. Images were acquired using Zen software (Zeiss). To assess colocalization, ImageJ/FIJI was used.

Rosetta GALigand Dock.

We used Rosetta GALigand Dock, a docking protocol that employs a genetic algorithm to explore ligand conformations and placements within a binding site using Rosetta’s energy function (25, 31, 32). Its ability to sample a vast conformational space made it well-suited for docking PIP2, a flexible and highly charged lipid. We performed two independent Rosetta GALigand Dock docking simulations to generate one million PIP2 poses in each simulation using the cryo-EM structure of TRPM4 representing the open state at 37 °C, resolved in the presence of decavanadate (PDB: 9B8Y) (8). To account for different possible PIP2 protonation states, we docked two -4 net charge variants of PIP2: one deprotonated at the 4-position phosphate in one simulation and one at the 5-position phosphate in another simulation.

MD Simulations.

The TRPM4 protein was based on the PDB entry 9B8Y with much of the intracellular domains truncated using UCSF ChimeraX (33). Specifically, residues 558 through 1,088 were retained. The four Ca²⁺ ions at the transmembrane binding sites were also retained. Four PIP2 molecules were manually placed using UCSF ChimeraX near the TRPM4 channel with headgroups placed approximately 5 Å away from the DVT-binding sites, here referred to as sites 1 through 4 corresponding to different subunits of the fourfold symmetric homotetrameric ion channel structure. An additional PIP2 molecule was placed in site 4 adjacent to the first PIP2 molecule, referred to as site 4.2. CHARMM-GUI membrane builder (34–36) was used to generate the MD simulation systems using the CHARMM36m protein force field (37), CHARMM36 lipid force field (38), standard CHARMM ion force field parameters (37), and TIP3P water model (39). The 1-Stearoyl-2-Arachidonoyl-sn-glycero-3-PI(4,5)P2 lipid model with -4 net charge, deprotonated at the 1st and 4th and protonated at the 5th position phosphates (residue name SAPI25, see https://charmm-gui.org/?doc=archive&lib=lipid) (38) was used for PIP2. The TRPM4 channel and accompanying PIP2 molecules were embedded in a lipid bilayer composed of 1-Palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) and cholesterol in a 9:1 ratio. The system was solvated with 0.15 M NaCl aqueous solution. Amber20 PMEMD (40) was the MD engine used to perform the simulations. The standard minimization protocol followed by a six-step staged restrained equilibration from the CHARMM-GUI generated Amber input files were used for the MD simulation systems (35). The equilibration was then extended for an additional 350 ns with a 0.1 kcal/mol/Ų harmonic restraint placed on all backbone Cα protein atoms. Subsequently, all restraints were removed, and the production MD simulation was run for an additional 200 ns. All the presented MD simulations were run at 310 K (37 °C) and 1 atm pressure in the NPT ensemble using recommended nonbonded interaction scheme for lipid membrane simulations (38). Similar runs were carried out at 298 K (25 °C); no detectable difference was observed. The trajectory was analyzed using custom Python scripts making use of the MDAnalysis Python library (41, 42).

In Vitro Culture of Mouse Bone Marrow–Derived Dendritic Cells (BMDCs).

Mice were euthanized in a method consistent with the recommendations of the AVMA Guidelines on Euthanasia as outlined in our animal protocol mentioned above. Hind legs were cut off from the hip joint. Femurs were cut free for isolating the bones, followed by immersing in 70% ethanol for a few seconds and rinsing thoroughly with phosphate-buffered saline (PBS). Femur was broken on both sides at the epiphyseal line and bone marrow was flushed out with ice-cold RPMI1640 medium (Gibco™). The flush fluid was collected, centrifuged, and red blood cells (RBCs) were lysed with the RBC lysis buffer. Isolated bone marrow cells were cultured in the RPMI1640 growth medium with recombinant mouse GM-CSF protein (Cat # 415-ML, R&D Systems) at 20 ng/mL and recombinant mouse Interleukin-4 (IL-4, carrier-free, Biolegend) at 10 ng/mL for 7 d to induce dendritic cell differentiation; the medium was intermittently refreshed. Recombinant mouse TNF-alpha protein (Cat # 410-MT, R&D Systems) (50 ng/mL) was added on Day 7 for maturation. Lipopolysaccharide (LPS, eBioscience™, Cat # 00-4976, 500×) was added to stimulate the cells on day 10 for 24 h, followed by functional assays.

Chemotaxis Cell Migration Assay.

Matured BMDCs were collected for performing migration assay with the QCM Chemotaxis Cell Migration Assay (96-well, 5 µm, Cat. No. ECM512, MilliporeSigma). Briefly, the clear plastic 96-well microplate, also called feeder tray, was first prepared and filled with 150 µL culture medium per well in the presence or absence of the chemoattractant agent, Chemokine C-C motif ligand 21 (CCL21). BMDCs were gently resuspended in the culture medium without chemoattractant at a concentration of 5 × 10⁵ per milliliter. A 96-well migration chamber with 5 µm pore size membrane at the bottom of each well was put over the feeder tray before filling 100 µL of the cell suspension per well. The plate was covered and incubated for 3 h at the indicated temperatures (25 °C, 30 °C, and 37 °C), in a CO₂ incubator (5% CO₂). The migration chamber plate was removed from the assembly after incubation. Cells/media from the top side were gently discarded by flipping out the remaining cell suspension and placed onto a new, unused 96-well feeder tray containing 150 µL of prewarmed cell detachment solution in each well. After a 30-min incubation at room temperature, 150 µL of the medium solution from the first feeder tray containing cells that had migrated through the membrane, along with 150 µL of cell detachment solution containing cells released from the membrane, were transferred to a test tube. Cell numbers were counted with a flow cytometer (BD Accuri™ C6 Plus Flow Cytometer), and the cell migration ratio quantified and analyzed.

Note.

After the submission of our manuscript, a new study came out [Teixeira-Duarte et al. (43), NSMB; 10.1038/s41594-025-01705-3] in which a PIP2-binding site in TRPM4 was identified with cryoelectron microscopy. The site overlaps with our proposed high-affinity binding site. A SI Appendix, Fig. S12 is added to illustrate the comparison of our proposed PIP2-binding site and the site identified by cryoelectron microscopy.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.