Structural dynamics at cytosolic interprotomer interfaces control gating of a mammalian TRPM5 channel

Significance

The cytoplasmic domains of transient receptor potential melastatin (TRPM) channels are involved in binding agonists, modulators, and accessory proteins and are therefore important for their physiological function. Despite recent advances in TRPM structural biology, it is not understood how cytoplasmic domains contribute to channel gating. Here, we use electron cryo-microscopy (cryo-EM) and electrophysiology to study the molecular mechanism of Ca²⁺-dependent activation and desensitization of TRPM5 channels. Our findings uncover the highly dynamic nature of the cytoplasmic domains of mammalian TRPM5 channels and provide insight into how these dynamics are associated with the Ca²⁺-dependent modulation of the channel. Our findings offer a mechanistic framework to understand how TRPM5 channel activity underlies processes like taste transduction and insulin secretion.

Abstract

The transient receptor potential melastatin (TRPM) tetrameric cation channels are involved in a wide range of biological functions, from temperature sensing and taste transduction to regulation of cardiac function, inflammatory pain, and insulin secretion. The structurally conserved TRPM cytoplasmic domains make up >70 % of the total protein. To investigate the mechanism by which the TRPM cytoplasmic domains contribute to gating, we employed electrophysiology and cryo-EM to study TRPM5—a channel that primarily relies on activation via intracellular Ca²⁺. Here, we show that activation of mammalian TRPM5 channels is strongly altered by Ca²⁺-dependent desensitization. Structures of rat TRPM5 identify a series of conformational transitions triggered by Ca²⁺ binding, whereby formation and dissolution of cytoplasmic interprotomer interfaces appear to control activation and desensitization of the channel. This study shows the importance of the cytoplasmic assembly in TRPM5 channel function and sets the stage for future investigations of other members of the TRPM family.

The transient receptor potential melastatin (TRPM) family of tetrameric cation channels encompasses eight members (TRPM1 to TRPM8) that are involved in a wide range of biological functions, from temperature sensing and taste transduction (1–4) to regulation of cardiac function (5), inflammatory pain, and insulin secretion (6–8). In recent years, cryo-EM has contributed greatly to our understanding of these ion channels, and we now have representative structures from almost every TRPM family member (9–11). One of their distinguishing features is a very large cytoplasmic domain, which makes up >70 % of the total protein and is fairly structurally conserved throughout the subfamily. It consists of three amino (N)-terminal Melastatin Homology Repeat domains (MHR1/2, MHR3, and MHR4) that are arranged around a C-terminal umbrella-like structure made up of four rib helices that run nearly parallel with the membrane plane, extending from a central coiled-coil domain with an axis perpendicular to the membrane. Mutations in the cytoplasmic assembly of TRPMs can lead to severe disease, including cardiac conductance block (12–14) and complete congenital stationary night blindness (15, 16), suggesting a critical role of these domains in channel function.

To investigate the mechanism by which the TRPM cytoplasmic domains might contribute to gating in TRPM channels, we decided to focus on TRPM5—a monovalent cation-selective channel (17) that primarily relies on activation by cytoplasmic Ca²⁺ ions (17, 18). The TRPM5 channel is important for signaling in type II taste bud cells (19, 20) and for insulin secretion in pancreatic β-cells (6, 7, 21–24). Because of this, TRPM5 is viewed as an attractive pharmacological target for the treatment of type II diabetes (25). Structures of zebrafish TRPM5 were published recently (11), and the study identified two Ca²⁺-binding sites within the channel: one in the voltage sensing-like domain (VSLD) within the transmembrane domain and the other in the cytoplasmic domain at the interface between MHR1/2 and MHR3. The binding site in the VSLD is conserved among many TRPM channels, including TRPM4 and TRPM8 (26–30), but the cytoplasmic binding site is unique to TRPM5. Both Ca²⁺-binding sites in TRPM5 are highly conserved across species (11). However, there is otherwise limited sequence identity between the zebrafish TRPM5 channel and its mammalian orthologs (~57% sequence identity between the zebrafish channel and the rat or human TRPM5 orthologs). We therefore set out to determine the molecular underpinnings of mammalian TRPM5 channel activation by Ca²⁺.

Using a combined approach of cryo-EM and electrophysiology, here, we show that gating of the rat, human, and mouse TRPM5 orthologs is functionally distinct from zebrafish and that the cytoplasmic domains play an important role in mediating Ca²⁺-dependent activation and desensitization. We observe structural dynamics associated with mammalian TRPM5 cytoplasmic domains and interfaces and provide evidence for their contribution to channel function that might also be relevant for other TRPM channels.

Results

Mammalian TRPM5 Channels Have a Distinct Functional Profile.

We began by assessing current responses to intracellular Ca²⁺ in inside-out patches excised from HEK293 cells expressing human, rat, mouse, or zebrafish TRPM5 channels (Fig. 1). TRPM5-specific outward-rectifying currents elicited by 25 μM free Ca²⁺ were similar for all four channel orthologs tested. However, upon exposure to 5.4 mM Ca²⁺, currents reached a peak of similar amplitude at positive and negative membrane voltages, followed by a rapid decay in current that was much more pronounced in the human, rat, and mouse channels than in the zebrafish ortholog (Fig. 1 B–F). This robust difference in the extent of desensitization underscores a key difference in gating between mammalian and zebrafish TRPM5 channels.

Fig. 1.

Mammalian and zebrafish TRPM5 channels possess a distinct activation profile. (A–E) Representative current time courses elicited by pulses from 0 to ±140 mV on inside-out patches obtained from transiently transfected HEK293 cells exposed to intracellular solutions containing free Ca²⁺ concentrations denoted by the horizontal bars. Dotted lines denote the zero-current level. (F) Mean ± SEM currents obtained from experiments as in (A–E), normalized to the peak current at +140 mV in the presence of 5.4 mM free Ca²⁺. Data from individual experiments are shown as circles (n = 5 to 12). Statistical significance was determined with a Tukey HSD test. Statistically significant differences are denoted by the number of asterisks according to the P-values provided in Dataset S1. Nonstatistically significant differences are denoted as n.s. (G) Current magnitudes at +140 mV from individual patches obtained from cells transfected with GFP only, or GFP together with rat (black) or zebrafish (blue) TRPM5 channel cDNA [as in (A), (C) and (E)]. For each patch, data are shown in control solution and in the presence of 25 μM (steady state) or 5.4 mM (peak) free Ca²⁺. Source data included in Dataset S1.

In desensitized channels activation by Ca²⁺ was retained but diminished in amplitude even in the presence of large concentrations of the cation (Fig. 1F), as previously observed (18). To investigate this further, we measured the current response of rat TRPM5 channels to increasing concentrations of free Ca²⁺ before and after complete channel desensitization (SI Appendix, Fig. S1 A–F). We found that before desensitization, depolarization appeared to be sufficient to drive channel opening at partial Ca²⁺-occupancy, evidenced by the 10-fold difference in EC₅₀ for Ca²⁺ activation between currents measured in the same experiment at +140 and −140 mV (SI Appendix, Fig. S1 C and E). The large reduction in Ca²⁺-activated current amplitude caused by desensitization was more pronounced at negative voltages, rendering inward currents indistinguishable from the background (SI Appendix, Fig. S1 B and D). We therefore focused solely on the outward currents after desensitization. In addition to decreased amplitude, desensitized currents at +140 mV had a markedly reduced sensitivity to Ca²⁺, evidenced by a ~30-fold increase in the EC₅₀ for Ca²⁺-activation (SI Appendix, Fig. S1 B, D, and E). These findings suggest that desensitized channels could have decreased binding affinity for Ca²⁺. Alternatively, desensitized channels could have identical binding affinity as non-desensitized channels but require a higher Ca²⁺-binding occupancy to open.

To investigate whether the binding of Ca²⁺ influences the onset of desensitization, we measured how fast the rTRPM5-mediated currents decayed in patches that were continuously exposed to different concentrations of free Ca²⁺ (SI Appendix, Fig. S1 G and H). We found that the time constant of desensitization had a sigmoidal dependence on the concentration of free Ca²⁺ that was well described by the same Hill function that we used to fit the activation curve by Ca²⁺ at −140 mV in non-desensitized channels (SI Appendix, Fig. S1H). This finding suggests that channel desensitization and channel opening at negative membrane potentials are driven by the same Ca²⁺-binding events.

In TRPM4 desensitization is caused by the depletion of phosphoinositide-4,5-bisphosphate (PIP₂) from the membrane (31). Exposure to diC8 PIP₂, a soluble analog of PIP₂, can rescue TRPM4 channels from desensitization with an EC₅₀ of 5 μM and with full recovery observed at 10 μM (31). We therefore explored the effect of diC8 PIP₂ on TRPM5 desensitization. Exposure to 20 μM diC8 PIP₂ had no statistically significant effect on desensitized TRPM5 channel currents (SI Appendix, Fig. S2 A and C). Exposure to a higher diC8 PIP₂ concentration (200 μM) recovered currents elicited by 5.4 mM Ca²⁺ to their pre-desensitization levels, but currents activated by 25 μM were still significantly smaller than those measured before desensitization (SI Appendix, Fig. S2 B and C). This might indicate that diC8 PIP₂ cannot fully substitute for the natural long-chain PIP₂. Alternatively, desensitization in TRPM5 channels could be caused by the binding of Ca²⁺, which is consistent with our observation that TRPM5 channel activation and desensitization have a very similar dependence on the concentration of Ca²⁺ (SI Appendix, Fig. S1H).

Together, our results indicate that mammalian TRPM5 channels undergo multiple conformational transitions following Ca²⁺-binding, involving energetically distinct activation pathways before and after desensitization, and that desensitization likely arises from Ca²⁺-dependent conformational changes in the channel.

Structure of Rat TRPM5 in the Absence of Ca²⁺.

Prompted by our functional data, we set out to determine the molecular mechanisms that underlie Ca²⁺-dependent activation and desensitization of a mammalian TRPM5 channel. First, to investigate the conformation that the channel occupies in the absence of stimulus, we expressed and purified the rat TRPM5 channel and determined its cryo-EM structure in the absence of Ca²⁺ (rTRPM5_EGTA, SI Appendix, Fig. S3). Consistently, an inspection of the previously reported locations for Ca²⁺ binding (11) indicated that no Ca²⁺ ions were present in any of the eight binding sites within the rTRPM5_EGTA tetramer (SI Appendix, Fig. S4 A–H).

We observe a nonprotein density in the rTRPM5_EGTA map which may correspond to PIP₂. This density occupies a cleft created by the pre-S1 domain, helices S1 and S4, the S4–S5 linker, the TRP domain, and the coupling domain and overlaps significantly with the PIP₂ binding site described in mouse TRPM8 (SI Appendix, Fig. S5) (30, 32). This density is observed in all four protomers (SI Appendix, Fig. S5 A–D). As no PIP₂ was added to our sample prior to freezing, we assume that the lipid copurified with the channel. Interestingly, PIP₂—a positive allosteric regulator for both TRPM8 (33) and TRPM5 (18) channels—was also found to copurify with the mouse TRPM8 (32).

The rTRPM5_EGTA structure possesses all the canonical elements of a TRPM channel (SI Appendix, Fig. S6A). Namely, it assembles as a tetramer and each protomer is composed of an N-terminal domain consisting of MHR1/2, MHR3, and MHR4, and a coupling domain (CD) between the cytoplasmic and transmembrane domains (TM). The TM domain is made up of a VSLD bundle (S1–S4 helices) and a pore domain (S5–S6 helices) which are linked by a short helical S4–S5 linker. The C-terminal TRP domain is nestled by the VSLD on top and the helix–loop–helix (HLH) motif of the CD on the bottom. From the TRP domain, a short helix extends down into the cytoplasm and links up to the rib helix. The rib helix then connects to a coiled-coil structure at a ~120° angle. (SI Appendix, Fig. S6A). The MHR1/2 domain is made up of nine helices sandwiching a β sheet (SI Appendix, Fig. S6B) and the MHR3 domain is formed by a stack of helices arranged in a helix-turn-helix (HTH) manner to produce a spring-like structure (SI Appendix, Fig. S6C).

The rTRPM5_EGTA tetrameric channel adopts a unique conformation as it assembles with near-C4-fold symmetry in the TM domain but shows no symmetry (C1) in the cytoplasmic domain. The lack of symmetry is most pronounced in the parts of the channel located deepest in the cytoplasm: the MHR1/2 and the distal parts of the coiled coil (Fig. 2A). This leads to an arrangement at the cytoplasmic side of the channel where each interprotomer interface is unique (Fig. 2 B and C). Namely, the distance between the MHR domain and the rib helix of neighboring protomers is different at each interface—ranging from 23 Å to 15 Å when measured from Cα of K94 in the MHR1/2 to Cα of N1090 in the rib helix (Fig. 2C).

Fig. 2.

Structure of the rTRPM5 in the presence of EGTA. (A) Structure of the rat TRPM5 channel in the presence of EGTA is asymmetric. Individual protomers are overlaid to show that the asymmetry is most pronounced in the cytoplasmic domains, and especially in the MHR1/2 domain (blue cartoon) and the coiled coil (CC, red cartoon). (B) Each cytoplasmic interprotomer interface is unique in rTRPM5_EGTA. The MHR domains of each protomer are shaded differently (A: red; B: marine; C: purple; D: gold). The MHR domains of protomer A (red) extend deep into the cytosol so that the dimensions of the channel from the top of the transmembrane domain (S764) to the bottom of the MHR domain (W296) are ~135 Å. The MHR domains in protomers B (purple) and D (gold) appear to rotate upward, which shortens the length of the channel to ~125 Å as well as the distance to the neighboring protomers. (C) Close-up of the boxed region in (B). The conformation of the MHR domains also dictates the distance between protomers. In protomer A (red), where the MHR domains are extended into the cytosol, the distance to the neighboring protomer D is large (23 Å between the Cα of K94 of MHR A and N1090 of MHR D). This distance shortens upon rotation of the MHR domains, as observed in MHR B, C, and D.

By contrast, these neighboring MHR domain interfaces in the zebrafish TRPM5 channel obtained in the presence of EDTA (PDB ID 7MBP) (11) are tightly coupled (SI Appendix, Fig. S7A), as evidenced by a tight network of interactions between the MHR1/2 and the rib helix and MHR3 of the neighboring protomer (SI Appendix, Fig. S7B). Furthermore, the channel has C4 symmetry. However, the differences between the cytoplasmic domains of rTRPM5_EGTA and the zebrafish TRPM5 in EDTA appear to have no bearing for the conformation of the pore, as the two channels align well in the TM regions (SI Appendix, Fig. S7C).

To determine the source of the asymmetry in rTRPM5_EGTA, we examined its cytoplasmic regions. The bottom view of the channel showed that an interface exists between the coiled-coil (CC) helix and the MHR1/2 domain of the neighboring protomer and that each one of these four CC-MHR1/2 interfaces is different (Fig. 3A) in the tetramer. A closer look at the CC structure showed that the CC contains a double glycine motif (GG hinge) that introduces a flexible point into this rigid assembly of helices and divides the CC into two parts—a channel proximal segment and a distal one (Fig. 3B). An overlay of the CC helices indicates that they begin to diverge below the GG hinge (Fig. 3C). This divergence appears to result in different interfaces between MHR1/2 and the CC.

Fig. 3.

The coiled-coil–MHR1/2 dynamics do not significantly affect activation and desensitization of rTRPM5 by Ca²⁺. (A) Bottom–up view of the rat TRPM5 channel determined in EGTA. Four distinct interfaces are observed between the MHR1/2 domain and the CC of neighboring protomers in this structure. They are colored in red (interface I, MHR A and CC B), blue (interface II, MHR B and CC C), purple (interface III, MHR C and CC D), and gold (interface IV, MHR D and CC A). (B) The distinct interfaces are caused by the arrangement of the CC. The proximal part adopts a tetrameric coiled-coil conformation, but the distal part is highly asymmetric. The density for the CC assembly is rendered as mesh and shown at contour level 0.15. (C) An overlay of the individual CC helices shows that the asymmetry begins at the flexible GG hinge. (D) Representative time courses at ±140 mV obtained from inside-out patches from cells expressing rTRPM5 channel mutants G1092A, G1093A, G1092A G1093A, and the deletion construct Δdistal CC. The dotted line denotes the zero-current level. (E) Group data for experiments in (D), shown as mean ± SEM (n = 5) or as individual experiments (open circles). Data were normalized to the peak current at +140 mV and 5.4 mM free Ca²⁺. Statistical significance was determined with a Tukey HSD test. Statistically significant differences are denoted by the number of asterisks according to the P-values provided in Dataset S4. Nonstatistically significant differences are denoted as n.s. Source data for (C and D) included in Dataset S4.

To test the functional significance of the apparent mobility of the distal part of the CC, we introduced single and double alanine substitutions into the GG hinge to reduce its flexibility (Fig. 3 D and E). We observed no statistically significant differences in the response to Ca²⁺ between these mutants and WT channels. To test for the opposite effect, we deleted the distal CC portion of the C-terminal domain and obtained a similar result: channels with the deletion responded to Ca²⁺ similarly to WT (Fig. 3 D and E). Together, these results indicate that the cytoplasmic assembly in mammalian TRPM5 channels is highly flexible in the absence of Ca²⁺ and that the distal coiled coil has a minimal contribution to the energetics of Ca²⁺-dependent gating. The contrast between the highly similar transmembrane domains of rTRPM5_EGTA and zebrafish TRPM5 in EDTA, and the highly divergent cytoplasmic domains, suggests that the mammalian channel has a more flexible cytoplasmic assembly and that independent conformational dynamics at each of the four cytoplasmic domains observed in the absence of Ca²⁺ in the rTRPM5_EGTA structure are not strongly coupled to the conformation of the transmembrane domain.

Ca²⁺ Binding Leads to Large Conformational Changes in rTRPM5.

Our electrophysiological studies revealed that rTRPM5 channel activation is rapidly followed by desensitization even at concentrations of Ca²⁺ that open only a small fraction of total channels in the patch (Fig. 1A and SI Appendix, Fig. S1 G and H). To attempt to capture rTRPM5 in a non-desensitized Ca²⁺-bound open state, we determined its structure in trace Ca²⁺ (~0.7 μM, see SI Appendix and Dataset S5), a concentration that is expected to weakly activate channels with slow desensitization (SI Appendix, Fig. S1H). Under these conditions, the cryo-EM data revealed a diverse conformational ensemble. We were able to identify three distinct C4-symmetric classes (rTRPM5_trace-1, rTRPM5_trace-2, and rTRPM5_trace-3), each with a unique conformation of the cytoplasmic domains (SI Appendix, Fig. S8). Due to weak local resolution, we did not model Ca²⁺ ions in any of the rTRPM5_trace maps, but the structural dynamics and C4 symmetry observed in the trace Ca²⁺ condition that were absent in the EGTA sample suggest that some of the sites in the trace structures are occupied by Ca²⁺. We also observed a nonprotein density within the putative PIP₂ binding site in all three structures (SI Appendix, Fig. S9 A–C).

Importantly, the three structures differ from each other by the distance between neighboring cytoplasmic protomers. We organized the structures according to the distance between the MHR1/2 of one protomer (Cα K94) and the rib helix of the neighboring protomer (Cα N1090), starting from the longest (rTRPM5_trace-1, ~23 Å) to the shortest (rTRPM5_trace-3, ~12 Å) (Fig. 4A and SI Appendix, Fig. S10A). By comparison, in the tightly coupled zebrafish TRPM5 structures this distance measures ~11 Å (11) (SI Appendix, Fig. S7E). Because the distance between protomers at this interface in rTRPM5_trace-3 is similar to that of the zebrafish TRPM5, we will refer to rTRPM5_trace-3 as fully coupled.

Fig. 4.

Trace amounts of calcium induce large conformational changes in rTRPM5. (A) Comparison of rTRPM5_EGTA with the conformations of rTRPM5 determined in the presence of trace amounts (~0.7 μM) of Ca²⁺ (rTRPM5_trace-1, lilac; rTRPM5_trace-2, purple; rTRPM5_trace-3, blue). The Top panel shows a side view. The cytoplasmic domains are highlighted. The dashed line represents the distance between MHR1/2 helix Cα K94 and the rib helix Cα N1090. The length of the channel (the distance between the Cα carbons of the S2 residue S764 and the MHR1/2 residue W296) condenses from 135 Å in rTRPM5_trace-1 to 125 Å in rTRPM5_trace-3. The Bottom panel shows a top view. The shortening of the channel coincides with a widening of the TM domains, as measured from the Cα of S2 residues V768 of opposing protomers. (B) Alignment of rTRPM5_trace-1, rTRPM5_trace-2, and rTRPM5_trace-3 tetramers_. For clarity, only a single protomer is shown for each structure. The alignment reveals a rotation around the rib helix (indicated by the box and enlarged Inset). The rib helix of rTRPM5_trace-3 is rotated by 12° and displaced laterally by ~4 Å compared to rTRPM5_trace-1. The red dot signifies the position of P1045 in the rib helix. (C) An overlay of the individual protomers shown in (B). (D) A schematic describing the conformational changes observed in rTRPM5_trace.

As the interface between MHR domains becomes tighter, the length of the channel contracts: the rTRPM5_EGTA protomer A and rTRPM5_trace-1 measure ~135 Å from the top of the S2 (S764) to the bottom of MHR1/2 (W296). By contrast, the rTRPM5_trace-2 is shortened to 130 Å and the rTRPM5_trace-3 to 125 Å (Fig. 4A), similar to protomers C and D in rTRPM5_EGTA (Fig. 2B). However, in contrast with the C1-symmetric rTRPM5_EGTA structure, the movements in the cytoplasmic domains are accompanied by significant changes in the dimensions of the TM domain in the C4-symmetric rTRPM5_trace structures. Namely, as the channel contracts vertically, its TM domains expand laterally. When measured across the TM (S2, V768), the channel widens from 86 Å in rTRPM5_trace-1, to 94 Å in rTRPM5_trace-3 (Fig. 4A). The widening in the TM domain also coincides with a widening of the S6 gate of the channel (SI Appendix, Fig. S10 B and C). Therefore, rTRPM5_trace-3—the only structure in the set with a coupled cytoplasmic interface—also has the widest pore (SI Appendix, Fig. S10 B–D).

To determine which conformational changes may be required to transition between rTRPM5_trace-1 and rTRPM5_trace-3, we aligned the three trace-Ca²⁺ tetramers (Fig. 4B). The overlay shows that in rTRPM5_trace-3 the MHR domains swing upward while the TM domains tilt downward when compared to rTRPM5_trace-1 and rTRPM5_trace-2. Interestingly, we observe a 12° rotation and a 4 Å downward movement of the rib helix from rTRPM5_trace-1 to rTRPM5_trace-3 (Fig. 4B). To determine whether this rotation is part of a rigid body movement of the entire protomer, we superposed the three individual protomers. This resulted in a near-perfect alignment (RMSD 1.5 Å), leaving only the proximal CC unaligned (Fig. 4C). This suggests that the key difference between the protomers lies in the relationship between the proximal CC and the rest of the protomer. In other words, the entire protomer undergoes a rotation and pivot around the proximal CC, which acts as a stable, immobile stalk. Such a rotation and pivot can explain the widening of the TM domains and the contraction of the channel length (Fig. 4D).

Based on the concentration of Ca²⁺ in the rTRPM5_trace sample (~ 0.7 μM and SI Appendix, Fig. S1H), it is possible that the ensemble contains channels captured in the pre-desensitized states. At this concentration, we would in principle expect partial Ca²⁺ occupancy and a lower probability of channel opening. This might explain why we did not detect a robust Ca²⁺ signal in neither of the two binding sites per subunit and why only one of the structures (rTRPM5_trace-3) was captured in a potentially open state. However, given the extensive amount of time that the rTRPM5_trace samples were exposed to Ca²⁺, we cannot exclude the possibility that this structural ensemble might represent desensitized states. Desensitized channels still activate upon Ca²⁺ binding, although Ca²⁺-bound closed states are strongly favored over the open states after desensitization has occurred (18, 34). Mechanistically, our trace Ca²⁺ structures suggest that binding of Ca²⁺ stabilizes the MHR1/2 and induces a concerted rotation and pivot of protomers around the CC. These movements result in increased coupling between the protomers at the cytoplasmic interfaces, shortening of the channel along the y axis, and widening of the pore.

Rat TRPM5 Adopts a Range of Desensitized, Closed Conformations in High Ca²⁺.

Our electrophysiological data showed that prominent Ca²⁺-dependent desensitization is a hallmark of mammalian TRPM5 channels. To observe channels in desensitized conformations and to ensure that Ca²⁺ ions can be unambiguously identified, we collected cryo-EM data of rTRPM5 in the presence of high (2 mM) Ca²⁺.

This condition also yielded a diverse conformational ensemble. Using a similar approach to the one applied to the rTRPM5_trace dataset, we were able to identify four classes (rTRPM5_high-1, rTRPM5_high-2, rTRPM5_high-3, and rTRPM5_high-4) (SI Appendix, Fig. S11). Nonprotein density was observed in the putative PIP₂ site in all four classes (SI Appendix, Fig. S9 D–G). In addition, density likely corresponding to Ca²⁺ was observed in both Ca²⁺ binding sites in rTRPM5_high-1, rTRPM5_high-2, and rTRPM5_high-4 (SI Appendix, Fig. S12). In rTRPM5_high-3 we did not assign any Ca²⁺ densities. However, we speculate that—given that the Ca²⁺ concentration is much larger than the EC₅₀ for calcium-dependent channel activation—the ion may be bound but is not fully visible due to the limited local resolution.

The four structures obtained in 2 mM Ca²⁺ can be arranged according to the distance between protomers at the cytoplasmic interfaces. rTRPM5_high-1 has the largest gap between MHR1/2 (Cα K94) of one protomer and the rib helix (Cα N1090) of the neighboring protomer (18 Å), and rTRPM5_high-4 the smallest (~12 Å) (Fig. 5A and SI Appendix, Fig. S13). Indeed, like in the rTRPM5_trace1-3 structures, the protomers pivot and rotate around the CC to result in more compact length and wider TM domains (Fig. 5A and SI Appendix, Fig. S13 A and B). However, the pore in rTRPM5_high remains closed (SI Appendix, Figs. S13D and S14), and we therefore assume that all four conformations represent desensitized closed channels.

Fig. 5.

Structures of rTRPM5 in the presence of high Ca²⁺. (A) A side-by-side comparison of rTRPM5_high1-4 structures: rTRPM5_high-1, light orange; rTRPM5_high-2, gold; rTRPM5_high-3, deep orange; rTRPM5_high-4, maroon. The Top panel shows a side view, and the cytoplasmic domains are highlighted. The dashed line indicates the distance between MHR1/2 α2 helix Cα K94 and the rib helix Cα N1090 of neighboring protomers. The length of the channel is measured between the Cα of the S2 residue S764 and the MHR1/2 residue W296. The Bottom panel shows a top view. (B) An overlay of individual protomers of rTRPM5_trace-3 (blue) and rTRPM5_high-4 (maroon)—structures which exhibit the highest degree of coupling in the two datasets. Notable differences include a small offset in the position of the CC (shown in the red box) and the conformation of the pore (dashed line box). (C) A close-up of the interface between S4, S4–S5 linker, S5, and S6 in rTRPM5_trace-3 (blue) and rTRPM5_high-4 (maroon).

To interrogate the mechanisms of activation and desensitization in rTRPM5, we compared the fully coupled rTRPM5_trace-3, to the closed, fully coupled rTRPM5_high-4 (Fig. 5B). When single protomers of the two channels are superposed, we observe only a small change in the CC (Fig. 5B). This indicates that the extent of rotation and pivot around the rib helix is similar in the two channels. Consequently, the extent of coupling between the MHR domains is also similar, with a ~12 Å distance between MHR1/2 and the rib helix of neighboring protomers (Figs. 4A and 5A and SI Appendix, Fig. S13C). However, there are significant differences between the VSLD, the S4–S5 linker, and the pore domain in the two structures (Fig. 5 B and C). In rTRPM5_trace-3, the S4–S5 linker interacts with the S6 helix, which is positioned closer to the VSLD and away from the center of the pore, apparently pulling on the S6 gate (Fig. 5C).

Our data suggest that four fully coupled cytoplasmic interprotomer interfaces are required to achieve opening of the rTRPM5 channel pore. Full coupling appears to require the presence of Ca²⁺, as observed in rTRPM5_trace-3 and rTRPM5_high-4. However, given that three of the four rTRPM5_high structures possess uncoupled cytoplasmic interprotomer interfaces, we speculate that desensitization in rTRPM5 might be associated with this uncoupling, which disengages critical networks that lead to gating.

Coupling of Cytoplasmic Interfaces Is Important for rTRPM5 Activation.

To interrogate the functional role of the cytoplasmic interprotomer interfaces and their significance in the activation and desensitization of rTRPM5, we took a closer look at the molecular interactions under uncoupled and coupled conditions.

None of the interprotomer interfaces in rTRPM5_EGTA are fully coupled (Fig. 2C). At the interface between protomers A and D the distance is largest—23 Å (Figs. 2C and 6A). Here, the α2 helix of MHR1/2 has no interactions with the neighboring protomer (Fig. 6 A and B). But positively charged residues, R89 and R85, in the α2 helix, extend toward the α3 helix within its own MHR1/2 and are positioned to make cation-pi contacts with H115 (Fig. 6 A and B). These internal MHR1/2 interactions could contribute to keeping the α2 and α3 helices together. When the interface is coupled, e.g., as in rTRPM5_trace-3 and rTRPM5_high-4, the α2 and α3 helices come within interaction distance to the rib helix and the MHR3 α4 helix of the neighboring protomer (Fig. 6 C and D). Here, the internal interactions between helices α2 and α3 of MHR1/2 are preserved and appear critical for stabilizing the interface between α3 and the rib helix (Fig. 6D). At the interface between α2 and the neighboring MHR3 α4 helix, D404 in the MHR3 α4 is within interaction distance to several residues that include K81, S82, and R85 in MHR1/2 α2 (Fig. 6 C and D). Based on our structural data, we hypothesized that the interfaces between MHR1/2 and rib helix and MHR3 of the neighboring protomer influence the functional state of the channel and that a fully coupled interface is necessary for opening of the channel.

Fig. 6.

The interprotomer contacts are functionally important. Interfaces between MHR1/2 and MHR3 and the rib helix of neighboring protomers in rTRPM5_EGTA (A and B) and rTRPM5_high-4 (C and D). (A) Side view of the interface between MHR1/2 and MHR3 and the rib helix from the neighboring protomer. Only the MHR1/2 α2 helix is shown for clarity. The distance between the Cα atoms of K94 and N1060 measures ~23 Å. (B) Top view of the interface in rTRPM5_EGTA including the α3 helix of the MHR1/2 domain. (C) Side view of the MHR1/2 and MHR3 and the rib helix interface in rTRPM5_high-4. The distance between the Cα atoms of K94 and N1060 measures ~12 Å. (D) The top view of the interface shows how α2 and α3 from MHR1/2 and HTH α4 of the neighboring MHR3 condense around the rib helix. The blue residue labels in (C) and (D) reflect results described in (E). (E) Mean ± SEM inside-out patch currents from cells expressing WT or mutant rat TRPM5 channels as shown on SI Appendix, Fig. S15. Data were measured at ±140 mV, normalized to peak at +140 mV and 5.4 mM free Ca²⁺, and quantified at each of the indicated intracellular free Ca²⁺ conditions. Data from individual experiments are shown as circles (n = 4 to 8). Statistical significance was determined with a Tukey HSD test. Statistically significant differences are denoted by the number of asterisks according to the P-values provided in Dataset S6. Nonstatistically significant differences are denoted as n.s. Bars are colored black for WT and WT-like mutants and blue for mutants with significantly reduced responses to 25 μM free Ca²⁺. (F) Mean ± SEM inside-out patch currents from cells expressing WT- or mutant zebrafish TRPM5 channels as in SI Appendix, Fig. S15 and depicted as in (E) (n = 5 to 6). Source data for (E and F) included in Dataset S6.

To test this hypothesis, we introduced mutations into three distinct regions of this interface: 1) the MHR1/2 α2–rib helix interprotomer interaction (mutations K90A, K94A, E1059A, and N1060A), 2) the MHR1/2 α2–α3 intraprotomer interaction (mutations R85A, D86A, and R89A), and 3) the MHR1/2 α2–MHR3 α4 interprotomer interaction (mutations K81A, S82A, and D404A). We recorded currents from inside-out patches under activating micromolar concentrations of Ca²⁺ and under strongly desensitizing Ca²⁺ conditions (Fig. 6E and SI Appendix, Fig. S15). Our data showed that disrupting the MHR1/2 α2–MHR3 α4 interface by introducing alanine substitutions at K81 and D404 resulted in a significant reduction in the response to micromolar concentrations of Ca²⁺ prior to desensitization (Fig. 6E and SI Appendix, Fig. S15). Two of the mutations (R85A and R89A) that disrupt the interactions between the two MHR1/2 helices, α2 and α3, resulted in similar statistically significant alterations of function. Mutations K94A and E1059A that disrupt the MHR1/2 α2–rib helix interface caused a noticeable yet nonstatistically significant reduction in the response to 25 μM Ca²⁺ relative to WT, whereas the two other mutations in this interface, K90A and N1060A, were indistinguishable from WT (Fig. 6E and SI Appendix, Fig. S15A).

Together, these data are consistent with our hypothesis that the coupled conformation is necessary for proper rTRPM5 channel function. The effects of K81A and D404A suggest that the interface between MHR1/2 and MHR 3 of the neighboring protomer plays a role in rTRPM5 activation by Ca²⁺. The intraprotomer interactions between the two MHR1/2 helices, α2 and α3, also appear to play a consequential role in channel activation, presumably because they help stabilize the interaction between MHR1/2 α2 and MHR3 α4.

In contrast to rTRPM5, zebrafish TRPM5 channels containing mutation D387A, equivalent to D404A in the rat ortholog, showed no significant reduction in the response to 25 μM Ca²⁺ relative to WT when measured before desensitization (Fig. 6F and SI Appendix, Fig. S15B). However, after desensitization the current amplitude was significantly smaller than in WT zebrafish TRPM5, and more resembling of the behavior of the mammalian channels. This suggests that destabilization of this interface in zebrafish TRPM5 channels has a negative impact on channel activation. Further, mutation K82A in the zebrafish MHR1/2–rib helix interface caused a small, yet significant reduction in the current elicited by 25 μM Ca²⁺, in contrast to the equivalent mutation K90A in rat TRPM5 that was indistinguishable from WT (Fig. 6F and SI Appendix, Fig. S15B).

In summary, our structural and functional data suggest that full coupling at interfaces formed by MHR1/2–MHR3 and MHR1/2–rib helix is important for activation by intracellular Ca²⁺of both rat and zebrafish TRPM5 channels.

Both Ca²⁺-Binding Sites Contribute to rTRPM5 Activation.

To determine whether channel activation or desensitization might be driven preferentially by one of the two Ca²⁺ binding sites, we mutated both the VSLD site (SI Appendix, Fig. S16A) and the ICD site (SI Appendix, Fig. S16B) and tested their response to activating micromolar Ca²⁺ concentrations and strongly desensitizing millimolar Ca²⁺ concentrations. Mutating residue Q784 in the VSLD binding site, which in the zebrafish channel had proven critical for channel activation (11), resulted in channels that lacked a response to micromolar Ca²⁺ concentrations but that still responded robustly to millimolar Ca²⁺ (SI Appendix, Fig. S16 C and D). We found that individual alanine mutations at each of the five residues forming the ICD site had no significant effect on channel function, with only the mutation E354A causing a slight reduction in the response to 25 μM Ca²⁺ (SI Appendix, Fig. S16 C and D). However, the introduction of two cumulative mutations at the ICD site, D353A+E354A, completely impaired channel functionality, with currents that largely resembled those recorded from cells transfected only with GFP (SI Appendix, Fig. S16 C and D). Our data indicate that both Ca²⁺ binding sites contribute to Ca²⁺-dependent activation.

Discussion

Here, we have described the conformational changes that occur in the absence and presence of Ca²⁺ in a mammalian TRPM5 channel (Fig. 7A). In the absence of Ca²⁺, the cytoplasmic assembly is highly flexible, and its dynamics appear to be uncoupled from the transmembrane domain and pore. The cytoplasmic interfaces between protomers adopt a range of uncoupled conformations, apparently independent of each other. The binding of Ca²⁺ appears to set in motion a series of concerted transitions that couple the movements at the cytoplasmic assembly with the conformation of the transmembrane domain and pore, as observed in our rTRPM5_trace ensemble. Our data suggest that once the cytoplasmic and transmembrane Ca²⁺ sites have been occupied in a defined number of channel subunits, each protomer undergoes rotation and pivot around the CC, and the MHR1/2 is repositioned to engage with the neighboring protomer. Once the channel becomes fully coupled the pore gate can open. Notably, channel opening is likely possible at subsaturating Ca²⁺ occupancies if the membrane potential is sufficiently depolarized, suggesting that pore opening may not require that all subunits are bound to Ca²⁺. We speculate that full Ca²⁺ occupancy enables channels to fully activate at negative membrane potentials, but that it also maximally accelerates the onset of desensitization. Under desensitizing conditions, the interactions at the cytoplasmic interfaces are likely weakened, contributing to the range of distances between protomers at the cytoplasmic interface that we observe in all structures determined in the presence of high Ca²⁺. We therefore posit that the tight coupling at the cytoplasmic interfaces in mammalian TRPM5 channels is a prerequisite for channel opening. During desensitization, the coupling is destabilized, and the cytoplasmic interface becomes dynamic.

Fig. 7.

The role of cytoplasmic domains in mammalian TRPM function. (A) A cartoon representation of the proposed rTRPM5 mechanism. In the absence of Ca²⁺, the cytoplasmic domains possess a high degree of flexibility and assume a range of conformations. However, the cytoplasmic interprotomer interfaces remain uncoupled. The channels can open when Ca²⁺ is bound and the cytoplasmic interprotomer interfaces are fully coupled. Finally, during desensitization, a relaxation occurs—the coupling between protomers decreases and the open state becomes destabilized. Conformational changes remain concerted between all four subunits in the presence of Ca²⁺. (B) Cytoplasmic interfaces in some existing TRPM structures. Cytoplasmic domains have been shaded for clarity.

A fascinating aspect of the mechanism of mammalian TRPM5 channel modulation by Ca²⁺ is that the presence of the divalent cation causes both activation and desensitization. We found that disruption of either of the two identified Ca²⁺ sites impedes normal activation, indicating that occupancy of both sites contributes to opening the channel. We hypothesize that desensitization requires occupancy of the same Ca²⁺ sites as channel activation because desensitization and activation have a very similar dependence on Ca²⁺ concentration.

Interestingly, desensitization of TRPM4 (31) and TRPM8 (33, 35) channels is caused by PIP₂ depletion from the membrane. By contrast, we have obtained several nonconducting conformations in the presence of 2 mM Ca²⁺ that likely represent desensitized states but appear to be bound to PIP₂ at a site that is conserved in TRPM8 (32) and TRPM3 channels (10, 36). This finding indicates that lipids interact strongly with that site and, in conjunction with our electrophysiological experiments, that PIP₂ may not play a central role in TRPM5 desensitization. However, we cannot exclude the possibility that the lipid we observe is not PIP₂, or that there might be additional lower-affinity PIP₂ binding sites in the channel that impact TRPM5 desensitization. Regardless, we found that diC8 PIP₂ has a strong potentiating effect on the Ca²⁺-dependent activation of TRPM5 channels (18). It is unclear whether this potentiation by diC8 PIP₂ involves a distinct, lower affinity binding site that remains to be identified. Finally, it is important to point out that the PIP₂-independent mechanism of desensitization proposed here does not exclude the possibility that PIP₂ depletion could impact TRPM5 channel activity in vivo.

This study confirms other functional similarities between the TRPM4 and TRPM5 channels, such a shared functional importance of the cytosolic interprotomer interface. Namely, in TRPM4, this interface contains the binding site for ATP, an inhibitor of TRPM4 function, as well as decavanadate, a channel modulator (28). ATP apparently inhibits the channel by uncoupling of the interprotomer interface (29). Moreover, several of the TRPM4 mutations that have been identified in patients with inherited cardiac disease occur at the interface between MHR1/2 and MHR3 of neighboring protomers (37, 38).

Interestingly, coupling and uncoupling at the cytoplasmic interprotomer interfaces also occur during gating of the more distantly related TRPM2 (39, 40). This suggests that cytoplasmic interfaces could be important for gating in several types of TRPM channels. Indeed, a cursory look at the cytoplasmic interfaces in the available TRPM channel structures (a small selection is shown in Fig. 7B) reveals many different arrangements of the cytoplasmic domains. This presents us with the question: what do these distinct arrangements tell us about their modes of gating? Additional studies of TRPM conformational ensembles might enable us to answer this question and understand where TRPM gating mechanisms intersect and where they diverge.

Materials and Methods

Protein Expression and Purification.

A gene for rat TRPM5 was synthesized and cloned into a pEG BacMam vector in frame with an N-terminal FLAG tag. Rat TRPM5 baculovirus was obtained as previously described (41). The protein was expressed HEK293 GnTI^- cells grown in Freestyle 293 Media supplemented with 2% (v/v) FBS and 1% (v/v) penicillin–streptomycin. Cultures were grown at 37 °C, 120 rpm, and 8% CO₂ for ~18 h after which 10 mM sodium butyrate was added and the temperature was lowered to 30 °C. After an additional ~48 h, cells were pelleted and resuspended in a buffer A (150 mM NaCl, 50 mM Tris pH 8.0, and 2 mM DTT) supplemented with a protease inhibitor cocktail (1 μg μL⁻¹ leupeptin, 1.5 μg μL⁻¹ pepstatin, 0.84 μg μL⁻¹ aprotinin, and 0.3 mM PMSF), 14.3 mM BME, DNAse I, and 2 mM DTT. For the calcium-free EGTA preparation, 1 mM EGTA was also included.

Cells were solubilized for 1 h at 4 °C in 1% (w/v) glyco-diosgenin (GDN) (Anatrace). Insoluble material was removed by centrifugation at 16,000 rpm for 30 min at 4 °C. The supernatant was then incubated with anti-FLAG M2 resin (Millipore-Sigma) for 1 h at 4 °C and eluted in buffer C (150 mM NaCl, 50 mM Tris pH 8.0, 2 mM DTT, 0.04% GDN, 0.1 mg mL⁻¹ FLAG peptide, and 1 mM EGTA where noted).

The eluate was run on a Superose 6 column (Cytiva) equilibrated with buffer B. Peak fractions were concentrated to ~3 to 4 mg mL⁻¹ for vitrification.

Cryo-EM Sample Preparation.

For the rTRPM5_EGTA preparation, buffer B was supplemented with 1 mM EGTA. The rTRPM5_trace and the rTRPM5_high samples were prepared in buffer B. After concentration, the rTRPM5_high sample was supplemented with 2 mM CaCl₂ and incubated on ice for 30 min. Before vitrification, all samples were centrifuged for 30 min at 4 °C and 16,000 rpm.

The protein was vitrified on freshly glow discharged UltrAuFoil R1.2/1.3 300-mesh (Electron Microscopy Services, rTRPM5_trace and rTRPM5_high) or Quantifoil R1.2/1.3 300-mesh grids (rTRPM5_EGTA), using Leica EM GP2 at 1 to 3 s blot time, 23 °C and >85% humidity.

Cryo-EM Data Collection.

Data were collected using the 300 keV Titan Krios with either Gatan K3 or Falcon III Direct Electron Detectors operating in dose counting mode in conjunction with a Bioquantum K3 Imaging Filter operating at 20 eV. Nominal magnifications used for the acquisitions were 81,000× corresponding to physical pixel sizes of 1.08 Å/pixel (0.54 Å/pixel super-resolution), 1.08 Å/pixel, 1.08 Å/pixel, and 1.12 Å/pixel for calcium-free, trace-calcium, and high calcium, respectively. For the calcium-free TRPM5 sample, 4,778 movies (40 frames/movie) were collected with a total nominal exposure of 40 e⁻/Å. The target defocus range for each dataset was from −1.0 µm to −2.25 µm. For the trace calcium TRPM5 sample, 6,565 movies were collected (40 frames/movie) with total exposure of ~45 e⁻/Å and a nominal defocus range from −1.0 µm to −2.5 µm. For the high calcium TRPM5 sample, 10,055 movies (40 frames/movie) were collected with total nominal exposure of ~43.3 e⁻/Å and a nominal defocus range from −1.0 µm to −2.25 µm.

Cryo-EM Data Processing.

rTRPM5_EGTA.

All data processing was conducted using cryoSPARC V3.3.2 (42). Patch motion correction (43) was used on 4,778 dose-weighted movies. The motion-corrected micrographs were subjected to patch CTF estimation (42) after which they were curated by removal of those where total full-frame motion was >40 pixels, estimated ice thickness was >1.1, and resolution estimations were >5 Å. This yielded a stack of 3,318 movies. From these, we picked a total of 1,462,359 particles using the template picker (160 Å particle size; 6 Å lowpass filter). The particles were extracted, binned at 4 × 4 (2.16 Å/pixel, 180 pixel box size), and subjected to 2D classification, which yielded 180,560 good particles. These particles were then re-extracted to 2× bin (1.08 Å/pixel, 360 pixel box size) and subjected to five-class ab initio reconstruction in C1. From there, particles that contributed toward the best TRPM5 reconstructions (159,140 particles, C1) were subjected to nonuniform refinement (44) that resulted in ~4.2 Å volumes. The C1 map was used as input into 3D variability analysis (45) (simple, three-mode, 5 Å filter). Using 3D variability display, the derived volumes that showed the greatest articulations in the MHR domains were used to bin the original particle set using cryoSPARC heterogeneous refinement. The highest-resolution heterogeneous volumes and particles were then subjected to another round of nonuniform refinement (86,152 particles).

rTRPM5_trace.

Patch motion correction utility and patch CTF estimation were used on 6,565 dose-weighted movies. Exposures with total full-frame motion >40 pixels, estimated ice thickness >1.1, and CTF estimations of >5 Å were removed. Using a combination of manual and template-based picking in conjunction with reference-free 2D classification, a set of 149,532 particles was selected to build a 1-Class ab initio reconstruction. These particles were subjected to nonuniform refinement and the volume used to build a template for picking. Using the template, 2,601,822 particles were picked and extracted (binned at 2 × 2, 2.16 Å/pixel, 180 pixel box size) and then subjected to multiple rounds of reference-free 2D classification to identify 546,821 good particles. These particles were then re-extracted to full resolution (1.08 Å/pixel, 360 pixel box size) and subjected to five-class ab initio reconstruction. From there, particles that contributed toward good TRPM5 reconstructions (510,640 particles) were subjected to a second round of five-class ab initio (475,743 particles, C4). These particles were input into a nonuniform refinement (3.74 Å). These particles used as input into 3D variability analysis (simple, three-mode, 3.5 Å filter). The derived volumes that showed the greatest articulations in the MHR domains were used as reference volumes in heterogeneous refinement. The highest-resolution heterogeneous volumes and particles were then subjected another round of nonuniform refinements with the best particle and volume combinations produced 6.02 Å (49,620 particles), 3.81 Å (139,580 particles), and 4.17 Å (53,230 particles) maps. These particles were subjected to local refinements with a mask covering either the MHR and the coiled-coil domain of a single protomer or the full-length (FL) protein.

r TRPM5_high.

The 10,055 movies were preprocessed as described above. Using the optimal template derived from trace Ca²⁺, the remaining 7,003 movies were picked resulting in 5,766,941 particles. These particles were then extracted, 2 × 2 binned (2.16 Å/pixel, 180 pixel box size), and subjected to multiple rounds of cryoSPARC reference-free 2D classification to identify 512,251 good particles. These particles were then re-extracted to full resolution (1.08 Å/pixel, 360 pixel box size) and subjected to two rounds of two-class ab initio reconstruction. The second round of five-class ab initio resulted in 331,548 particles which were then input into nonuniform refinement that resulted in a 3.65 Å map. These particles were used as input into 3D variability analysis. The derived volumes that showed the greatest articulations in the MHR domains were used as reference volumes in the cryoSPARC heterogeneous refinement. The highest-resolution heterogeneous volumes and particles were then subjected another round of nonuniform refinement. These particles were subjected to local refinements with a mask of either the MHR and the coiled-coil domain of a single protomer or of the FL protein.

All resolution estimates were based on the gold-standard FSC 0.143 criterion as calculated by cryoSPARC (46, 47).

Model Building.

The rTRPM5_EGTA model was built directly into the cryo-EM density using the previously determined structure of the zebrafish TRPM5 (PDB ID 7MBP) as a starting reference. The subsequent models were built using the rTRPM5_EGTA as a starting reference. All models were built in Coot 0.98 (48). The Molprobity (49) server was used to guide the refinement. At the final stage, models were subjected to a round of real space refinement using the PHENIX 1.20 real space refine (50) utility. Analysis and illustrations were performed using Pymol (51) and UCSF Chimera (52).

Cell Culture for Patch-Clamp Electrophysiology.

HEK293 cells were kept at 37 °C with 5% CO₂ and grown in Dulbecco’s modified Eagle’s medium (DMEM) with high glucose, pyruvate, L-glutamine, and phenol red, supplemented with 10% fetal bovine serum (vol/vol) and 10 mg mL⁻¹ gentamicin. For transfection, cells were detached with trypsin, resuspended in DMEM, and seeded onto 3 mL dishes with No. 1 glass coverslips that were pretreated for 30 min with Fibronectin (0.1 mg mL⁻¹ in PBS, Millipore Sigma). Transfections were performed on the same day using FuGENE6 HD Transfection Reagent (Roche Applied Science). Channel constructs were cotransfected with pGreen-Lantern (referred to as GFP in the text, Invitrogen) at a ratio of 1:1. Patch-clamp recordings were done 18 to 36 h after transfection.

Patch-Clamp Electrophysiology.

Patch-clamp recordings were performed on transiently transfected HEK293 cells at room temperature (21 to 23 °C) using the inside-out configuration. Data were acquired and digitized with a dPatch amplifier system and SutterPatch software (Sutter Instrument, Novato, CA) and analyzed using Igor Pro 8.04 (Wavemetrics, Portland, OR). Pipettes were pulled from borosilicate glass (1.5 mm O.D. × 0.86 mm I.D. × 75 mm L; Harvard Apparatus) using a Sutter P-97 puller and heat-polished to final resistances between one and five MΩ using a MF-200 microforge (World Precision Instruments). An agar bridge (1 M KCl; 4% weight/vol agar; Teflon tubing) was used to connect the ground electrode chamber and the main recording chamber. The intracellular and pipette solution consisted of (mM): 150 NaCl, 10 HEPES, and 5 EGTA, pH 7.4 (NaOH). The diC8 PIP₂ (Echelon Biosciences, Salt Lake City, UT) stock solutions were made in water at 1 mM (for 20 μM final concentration) or at 10 mM (for 200 μM final concentration). The free Ca²⁺ concentration in our recording solutions was measured using Fluo4FF as described in the next section.

A gravity-fed motorized perfusion system (RSC-200, BioLogic, France) was used to switch between recording solutions. Currents were elicited by ±140 mV pulses of 50 ms duration applied at 10 Hz from a holding potential of 0 mV. Data were acquired at 5 kHz and low-pass filtered at 1 kHz. The current time courses shown in figures display the current average for the last 5 ms of each pulse and were normalized to the peak current value at 5.4 mM free Ca²⁺ and +140 mV. For dose–response relations, the mean current in zero-Ca²⁺ control solution was subtracted from the current values at the various free Ca²⁺ concentrations, which were then normalized to the current at +140 mV in 200 μM (before desensitization) or 1.56 mM free Ca²⁺ (after desensitization). Curves were fit to the Hill equation.

Statistical Analysis.

Group data are shown as mean ± SEM throughout the manuscript. We utilized the Tukey Honest Significant Difference (HSD) test implemented on Igor Pro to establish statistical significance between different groups of data. All data used for statistical comparisons were normalized to the peak current elicited by 5.4 mM Ca²⁺ at +140 mV. We excluded this condition from our comparisons. We applied the Tukey HSD test to compare between two distinct groups of normalized data: various experimental conditions evaluated in the same patch, and a specific experimental condition (25 μM) evaluated in patches expressing different TRPM5 channel variants. We also utilized a paired t test implemented on Igor Pro to compare the parameters obtained from fits to dose–response relations for Ca²⁺ measured at +140 and −140 mV.

Calcium Concentration Measurements.

The concentration of free Ca²⁺ in the buffers used in electrophysiological recordings and in preparation of rTRPM5 protein was determined using Fluo4FF and Fura 2 dyes (Invitrogen), respectively, according to the manufacturer’s instructions. Please see SI Appendix for more detail.

Cell Lines.

HEK293GnTI^- cells were purc*.369hased from ATCC with authentication records. The cells were not additionally authenticated nor tested for mycoplasma contamination prior to this study. HEK293 cells from ATCC (CRL-1573) used for electrophysiology tested negative for mycoplasma.

cDNAs.

We used the codon-optimized gene for rat TRPM5 cloned into a pEG BacMam vector in frame with an N-terminal FLAG affinity tag. Mouse TRPM5 in pcDNA3.1 was purchased from Addgene (plasmid #85189), human TRPM5 in pcDNA3.1+ (Accession No: NM_014555.3) was purchased from GenScript, and zebrafish TRPM5 in pEG BacMam was kindly provided by Seok-Yong Lee (Duke University).

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Dataset S05 (XLSX)

Dataset S06 (XLSX)

Dataset S07 (XLSX)

Data, Materials, and Software Availability

The atomic coordinates and cryo-EM density maps have been deposited to Protein Data Bank (PDB) and Electron Microscopy Data Bank (EMDB) under the following entries: rTRPM5_EGTA (8SL6 (53) EMD-40574 (54)), rTRPM5_trace-1 (8SL8 (55), EMD-40575 (56), EMD-40600 (57)), rTRPM5_trace-2 (8SLA (58), EMD-40576 (59), EMD-40599 (60)), rTRPM5_trace-3 (8SLE (61), EMD-40577 (62), EMD-40597 (63)), rTRPM5_high-1 (8SLI (64), EMD-40578 (65), EMD-40596 (66)), rTRPM5_high-2 (8SLP (67), EMD-40579 (68), EMD-40595 (69)), rTRPM5_high-3 (8SLQ (70), EMD-40580 (71), EMD-40593 (72)), and rTRPM5_high-4 (8SLW (73), EMD-40581 (74), EMD-40592 (75)).