Human mutations highlight an intersubunit cation–π bond that stabilizes the closed but not open or inactivated states of TRPV channels

Significance

Transient receptor potential (TRP) channels are key cation channels that respond to thermal, chemical, electric, or mechanical stimuli. There are >60 mutations known in TrpV4 that cause defects in bone or neural development. Two mutations guided us to examine a cation–π bond, which we found to tie the four subunits together to help close the channel gate. This or a similar bond is seen in many structures of TRP channels. In experiments, breaking this bond favors opening and inactivation of TRPV4.

Abstract

An adequate response of a living cell to the ever-changing environment requires integration of numerous sensory inputs. In many cases, it can be achieved even at the level of a single receptor molecule. Polymodal transient receptor potential (TRP) channels have been shown to integrate mechanical, chemical, electric, and thermal stimuli. Inappropriate gating can lead to pathologies. Among the >60 known TRP vanilloid subfamily (V) 4 mutations that interfere with bone development are Y602C or R616Q at the S4–S5 linker. A cation–π bond between the conservative residues Y602 and R616 of neighboring subunits appears likely in many homologous channel structures in a closed state. Our experiments with TRPV4 mutants indicate that the resting-closed state remains stable while the bond is substituted by a salt bridge or disulfide bond, whereas disruption of the contact by mutations like Y602C or R616Q produces gain-of-function phenotypes when TRPV4 is heterologously expressed in the Xenopus oocyte or yeast. Our data indicate that the Y602–R616 cation–π interactions link the four S4–S5 linker helices together, forming a girdle backing the closed gate. Analogous cation–π bonds and the girdle are seen in many closed TRP channel structures. This girdle is not observed in the cryo-EM structure of amphibian TRPV4 (Protein Data Bank ID code 6BBJ), which appears to be in a different impermeable state—we hypothesize this is the inactivated state.

Transient receptor potential (TRP) channels are polymodal cation channels permeable to Ca²⁺. They are found in almost all eukaryotic species, including the systems popular for genetic manipulation—yeast, Caenorhabditis elegans, and Drosophila (1). In mammals, there are seven subfamilies, including the vanilloid subfamily, TRPV, which has many members that are involved with temperature and chemo- and mechanosensations. One member, TRPV4, can be activated by mechanical forces, including swelling, and lipid-intercalating endogenous chemicals, including anandamide and arachidonic acid.

The importance of the TRPV4 channel is exemplified by the fact that mutations in this channel can cause both peripheral neuropathies and a wide phenotypic array of skeletal dyplasias, ranging from mild to fatal (2, 3). Two point mutations, R616Q or V620I, in TRPV4 results in mild brachyolmia (4), and the Y602C mutation is involved in spondyloepiphyseal dysplasia Kozlowski type and spondyloepiphyseal dysplasia Maroteaux type (3, 5). More than 60 mutations causing human heritable diseases are spread throughout the TRPV4 gene (2), and therefore it seems there is no apparent link between mutation sites and phenotypes of diseases. Understanding the underlying mechanisms of these channel malfunctions is essential for treatment of these diseases.

Our understanding of TRP channel structure has been greatly advanced by cryoelectron microscopy and X-ray crystallography. To date, more than five dozen TRP channel structures have been resolved, including the TRPA1 (6), TRPC3 (7), TRPC4 (8), TRPM2 (9), TRPM4 (10), TRPML1 (11), TRPML3 (12), TRPM8 (13), TRPV1 (14), TRPV2 (15), TRPV4 (16), TRPV5 (17), TRPV6 (18), TRPP2 (19), TRPP3 (20), and TRPN (21) channels. The general homotetrameric fold of the transmembrane part of TRP channels resembles one from voltage-gated channels—transmembrane helices S1 to S4 of the same subunit make a peripheral domain, which is connected by an S4–S5 linker to the central pore domain formed by the S5 helix, pore helix, filter loop, and S6 helix contributed by each of the four subunits. All of the structures from the TRPV family and most of the channels from other TRP families (except for TRPP2, TRPML1, and TRPML3) feature a characteristic TRP helix following the gate-bearing S6 helix. The helix, located between the S1-to-S4 bundle and the cytoplasmic domain, contacts essentially all of the major structural modules of a TRP channel, and was suggested to function as a “force hub” integrating the conformational changes induced by stimuli of different modalities (22). An alpha-helical linker between the S4 and S5 helices typically lies along the cytoplasmic membrane and is located right above the TRP helix and immediately contacts the S6 helix near the gate level, suggesting that the linker domain might be involved in control of gating. Helices running along the membrane surface on the cytoplasmic side are a common theme among channels, and often have functional implications (see refs. 23 and 24 for discussions). Indeed, previously, we have shown that a bond between the L596 at the loop preceding the S4–S5 linker and residue W733 on the TRP helix, which falls into this “surface-level” category, functions as a “latch” stabilizing the closed conformation (22).

Almost invariantly, the TRPV structures feature a pronounced kink in this location, which allows the S4–S5 linker to lie almost along the cytoplasmic membrane plane, so that the C-terminal end of the linker from one subunit contacts the N-terminal part of the linker from the neighboring subunit. All of the members of the TRPV family (as well as some other TRP families) feature a very conservative pair of residues that come in close proximity from the opposite subunits—aromatic (tyrosine or phenylalanine) and basic (typically arginine) residues. These residues are Y602 and R616 in the rat TRPV4 sequence, which, as discussed above, when mutated in humans to Y602C and R616Q can cause severe skeletal dysplasia. A conspicuous feature of this contact is that the sharpness of the kink and the proximity of these residues vary depending on the putative conformational state of the channel. For example, in the closed conformation of TRPV1 [Protein Data Bank (PDB) ID code 3J5P], the homologous residues Y565 and R579 are in tight contact (25), which suggests a cation–π interaction between the side chains, an interaction that might stabilize the resting conformation. Indeed, our previously published homology model of rat TRPV4 in the resting state based on the TRPV1 (3J5P) structure (22) retains the close proximity of the conserved pair. On the other hand, in the activated state of TRPV1 in complex with DkTx and RTX (3J5Q), the hydrophobic gate is expanded and Y565 and R579 are positioned farther away from each other (26), precluding strong interaction and suggesting that disruption of the bond would facilitate channel opening.

Interestingly, a recently solved TRPV4 structure (6BBJ) revealed a surprising arrangement of this aromatic/basic residue pair different from all of the known TRPV structures (16): While the gate is tightly closed, the kink between the S4–S5 linker and S5 helix is essentially straightened, which relocates the arginine deep into the membrane core and moves the residues far away from each other, making a cation–π bond or any direct interaction extremely unlikely in such a conformational state.

In the current work, we analyze the interaction at the S4–S5 linker of the TRPV4 channel, near the beginning of the S5 helix. We demonstrate that Y602C and R616Q do indeed effect gain-of-function (GOF) channel and in vivo phenotypes, in agreement with their effecting developmental disorders. Given the potential involvement of the Y602–R616 interactions in the gating transitions of the TRPV4 channel, the potential of mutations in this region to effect severe skeletal dysplasia, and the uncertainty in predictions of the positions of these residues in the resting-closed conformation, we decided to test experimentally if this interaction existed in unstimulated cells in vivo, and whether strengthening or disruption of this potential contact will have any effects on the conductive properties and functional cycle of the TRPV4 channel when heterologously expressed in oocytes. These data have proven crucial for understanding the transformations occurring during gating as well as for interpretation of the functional conformational states captured in the current resolved structures of TRPV channels.

Results

TRPV4 Y602C and R616Q Mutations Have Gain-of-Function Channel Phenotypes When Expressed in Oocytes.

The Y602C and R616Q mutations in TRPV4 cause skeletal dysplasia in humans (2, 3). Therefore, it is of interest to determine if TRPV4 activity is affected by these mutations. We examined the channel activities of Y602C and R616Q mutants using a Xenopus oocyte expression system and two-electrode voltage clamp 3 to 4 d after cRNA injection (Fig. 1).

Fig. 1.

Channel activity of wild-type and mutant TRPV4 channels. (A and C–F) Representative traces of whole-oocyte currents recorded by two-electrode voltage clamp from oocytes injected with (A) 5 ng of wild-type (WT) TRPV4, (C) 0.5 ng of Y602C, (D) 0.5 ng of R616Q, (E) 5 ng of Y602C/R616C, or (F) 5 ng of R616C cRNA. (B) Uninjected oocytes are shown as the negative control (NC). Oocytes were recorded at 3 or 4 d after cRNA injection. Currents were evoked by voltage steps from a holding potential of −60 mV to a test potential between −100 and +60 mV in 20-mV increments (diagrammed in Inset). (G) Comparison of the mean peak currents of WT (5 ng of injection, n = 24; 0.5 ng of injection, n = 6), Y602C (n = 7), R616Q (n = 11), Y602C/R616C (n = 8), and R616C (n = 6) with the indicated amounts of injected cRNAs (gray bars, 5 ng; white bars, 0.5 ng), and NC (black bar; n = 6). Peak currents were recorded during 400-ms depolarization to 60 mV from the holding potential of −60 mV. Data are presented as mean ± SEM.

The oocytes injected with 5 ng of wild-type TRPV4 cRNA produced a moderate peak current (20 ± 2.6 μA, n = 24) when varying voltage from −60 to +60 mV. However, at the same level of cRNA injection, the expression of the Y602C or R616Q channels led to many morphologically unhealthy oocytes, the remaining generating very large currents that were beyond the capacity of the recording system. Thus, we reduced the amount of cRNA injected to 0.5 ng. While only very small or no currents (1 ± 0.4 μA, n = 6) were observed in the oocytes injected with 0.5 ng of wild-type TRPV4 cRNA, both mutant channels demonstrated reasonable currents at this low injection level (19 ± 4 μA for Y602C, n = 7; 16 ± 2 μA for R616Q, n = 11). These currents are comparable to oocytes injected with an order of magnitude more cRNA (5 ng) of WT TRPV4. These results indicate that Y602C and R616Q are both GOF mutations; such channel phenotypes may explain human genetic data.

TRPV4 Y602C and R616Q Mutations Effect GOF When Expressed in Yeast.

The function of TRPV4 GOF mutants also can be examined using a yeast growth assay, as previously described (27, 28). Briefly, the expression of wild-type TRPV4 in yeast does not affect yeast growth; however, the expression of TRPV4 mutated channels with GOF phenotypes, such as W733R (22), will arrest yeast growth, presumably because ion leakage through GOF TRPV4 is toxic to the cell. In uninduced conditions, no effects on yeast growth were observed (Fig. 2). On the inducible medium, however, the expression of Y602C or R616Q significantly inhibited yeast growth in contrast to WT (Fig. 2). These results are consistent with those from oocyte electrophysiology, thus confirming that Y602C and R616Q are sufficient to yield a GOF phenotype.

Fig. 2.

Y602C, R616Q, and R616K mutant channels inhibit yeast cell growth. Starting with ∼5,000 colony-forming units, serial 10-fold dilutions of yeast cells housing plasmids containing conditionally expressed WT, W733R, Y602C, Y602F, R616Q, R616K, or empty vector (p416GAL1) were dropped on a repressive plate (glucose; Left) or an expressive plate (galactose; Right). The plates were incubated at 30 °C for 48 h. Data shown are representative of three independent experiments.

Disulfide Trapping Confirms Close Proximity of the TRPV4 Y602 and R616 Residues and the Requirement for Dynamics of This Region in Channel Gating.

We used a disulfide-trapping approach to test if these residues are also in close proximity in the TRPV4 channel. We first generated the R616C mutation and examined it in the oocyte expression system. As seen in Fig. 1 F and G, the TRPV4 R616C mutated channel yielded currents (22 ± 4 μA at peak; n = 6) similar to the wild-type channel when injected with 5 ng cRNA. If indeed there is a cation–π interaction in the closed state, then the cysteine at this location may maintain this conformation, as in the wild-type channel. This is consistent with other studies giving precedence for SH–π interactions (29, 30). In addition, TRPV4 Y602C displayed slower inactivation compared with the WT (SI Appendix, Fig. S1), suggesting that Y602 interacts with other potential partners in other states.

We then combined the Y602C/R616C mutations to generate the double-mutated channel. Here again, the currents in oocytes injected with 5 ng of cRNA (26 ± 5 μA; n = 8) were similar to wild-type TRPV4 channels (Fig. 1 E and G). Thus, the R616C mutation appears to largely suppress the Y602C GOF channel phenotype even though R616C did not yield an independent loss-of-function (LOF) phenotype.

To more directly confirm proximity, we assayed for the presence of a disulfide bond between 602C and 616C. We compared the size of the complexes for R616C and Y602C/R616C mutant TRPV4 channels by Western blotting in the presence and absence of reducing agent (Fig. 3).

Fig. 3.

Y602C/R616C double mutant shows increased multimerization in the absence of reducing agent. The preparations of the oocytes injected with wild-type, R616C, or Y602C/R616C cRNA and from uninjected oocytes were subjected to SDS/PAGE under conditions without (A) and with 5% 2-ME (B), and detected with an anti-TRPV4 antibody. m, membrane proteins; s, supernatant proteins; t, total proteins.

TRPV4 has a predicted monomeric molecular mass of 98 kDa and 13 cysteine residues in its sequence. We could not detect the Y602C complex, because Y602C is a GOF phenotype, and the same amount of cRNA injection (5 ng) caused morphological decay of most oocytes. For the WT-expressing oocytes in nonreducing conditions, a slight band that is sized as a monomer and a dominant band consistent with a dimer were detected in the membrane preparations but not among supernatant proteins (Fig. 3A), indicating a successful expression and incorporation into the membranes. No TRPV4-specific bands were detected in uninjected oocytes. The R616C mutant protein showed a similar mobility pattern to that of wild-type TRPV4. However, the Y602C/R616C mutant protein exhibited increases in the mobility of the channel protein. After treating with a reducing agent (2-mercaptoethanol; 2-ME), only monomeric bands were observed for wild-type, R616C, and Y602C/R616C TRPV4 channels (Fig. 3B). These results suggest that an additional intersubunit disulfide bond is formed between the two introduced cysteine residues in the Y602C/R616C TRPV4 mutant.

Next, we tested whether a reducing agent (2-ME) and oxidizing agent (H₂O₂) affect the channel activity of the wild type and these mutants using a two-electrode voltage clamp. As shown in Fig. 4A, when directly added to the bath solution, 2-ME immediately reduced the current by ∼30% in oocytes expressing wild-type TRPV4, presumably by reducing some disulfide bridges within the complex. Subsequent H₂O₂ tests showed that endogenous disulfide bridging that influences channel activity is not affected by oxidation. Similar results were seen for the single mutants Y602C and R616C (Fig. 4 B and C). In contrast, 2-ME dramatically (by ∼100%) increased the currents of Y602C/R616C-expressed oocytes, and H₂O₂ greatly reduced the currents down to the starting level (Fig. 4D). Uninjected oocytes showed no changes in response to 2-ME and H₂O₂ (Fig. 4E), indicating that the effect on TRPV4 channels was direct rather than mediated by other membrane proteins. These results suggest that 2-ME disrupts a disulfide bond between residues 602 and 616, and thereby allows greater activation of the mutant channel. Hence, trapping an interaction between 602 and 616 favors TRPV4 closure (although likely does not completely prevent opening).

Fig. 4.

2-ME activates the Y602C/R616C channel. Time course of the effects of 2-ME and H₂O₂ on whole-oocyte currents recorded from oocytes expressing WT (A; n = 8), Y602C (B; n = 3), R616C (C; n = 4), or Y602C/R616C (D; n = 9), and noninjected oocytes as negative control (E; n = 4). Representative data are shown. 2-ME (0.5%) and H₂O₂ (0.15%) were added to the bath solution at the time points labeled with a red arrow and green arrow, respectively. Currents were evoked with serial 250-ms depolarization from a holding potential of −20 mV to +40 or 60 mV with 10-s intervals.

Using Additional Mutagenesis to Dissect the Potential Interaction Between Y602 and R616.

Thus far, the data are consistent with Y602 and R616 interacting in the closed state and needing to be separated for channel opening. As predicted for TRPV1 (25), these residues could form a cation–π interaction—if true for TRPV4, this could serve as the latch stabilizing the closed state. To further test this notion, we generated additional lesions at these positions and tested their effects on channel function. We first mutated Y602 to another aromatic amino acid, phenylalanine, naturally occurring at this position in some of the homologs and capable of maintaining cation–π interactions. As seen in Fig. 5A, Y602F generated currents similar to wild-type TRPV4.

Fig. 5.

Channel activity of mutant TRPV4 channels. Representative traces of whole-oocyte currents recorded by two-electrode voltage clamp from oocytes injected with (A) 5 ng of Y602F TRPV4, (B) 5 ng of Y602D, or (C) 1 ng of R616K. Shown are representative currents of n ≥ 6.

We next generated the mutation Y602D. This is not a conserved substitution, and so potentially could disrupt channel function. However, if the cation–π interaction hypothesis is correct, then the predicted salt bridge formed between Y602D and R616 may stabilize the closed state enough to produce a channel that gates relatively normally. Indeed, as seen in Fig. 5B, Y602D functioned similar to wild-type channels. Notably, Y602F and Y602D also changed the inactivation kinetics of the mutant channel (SI Appendix, Fig. S1), suggesting Y602 also has important interactions with some other residues in other states.

Finally, we generated R616K TRPV4. Although the R616K mutation reflects a conserved substitution, arginine appears to be preferred over lysine in cation–π interactions (31). Indeed, a slight GOF phenotype was observed. Oocytes injected with 5 ng of R616K TRPV4 cRNA showed some morphological degradation and were not easily assayed by voltage clamp. In addition, the R616K TRPV4 channel yielded a GOF phenotype in yeast (Fig. 2). However, this GOF phenotype was not as severe as that observed for Y602D and R616Q, and we were successful at measuring current at 1 ng of cRNA, which showed essentially normal activation and inactivation (Fig. 5C).

Discussion

Overall, our mutation data are consistent with a cation–π nature (31) of interaction between Y602 and R616 in the resting-closed state of the channel. The conservative substitution Y602F, which provides another aromatic residue with essentially the same geometry of π-orbitals and capable of maintaining similar cation–π interactions with R616, shows functional properties close to WT (Fig. 5A). Replacement of arginine in position 616 for lysine did influence channel gating but produced only a mild GOF phenotype (Figs. 2 and 5C), which might be interpreted as some weakening of the interaction due to the change in geometry of the cationic counterpart. It is known that arginine is preferred over lysine in cation–π interactions, and the strength of the bond is sensitive to the geometry of the contact (31). An essentially normal phenotype was observed for the R616C mutant (slight LOF; Fig. 1F), which can reflect the ability of the SH group to engage in polar interactions with aromatic rings, similar to cation–π interactions (32–34). In contrast, the mutants that disrupt cation–π interactions, such as Y602C (absent aromatic ring) and R616Q (zero net charge), effected a pronounced GOF phenotype, both in terms of the current through the channel in oocytes (Fig. 1 C and D) and the inhibition of yeast growth (Fig. 2). Overall, our findings suggest that the channels are hypersensitive, gating inappropriately. Whether the hypersensitivity is exaggerated by one stimulus (e.g., mechanical force or selective agonists) over others will require further study (as an expectation, mechanical force and selective agonists would probably facilitate the gating, because our Y602–R616 intersubunit bond forms a “girdle” around the gate region, and an expansion of the gate induced by any means would most likely require rupturing that ring to let the gate widen—an effect similar to what we have documented for the L596–W733 bond earlier). The observation that these mutations effect human disease as well as GOF channel prototypes when expressed heterogeneously suggests that the mutated channels are processed, trafficked to the cell surface, and adopt a functional fold even when expressed in a living organism. Destabilization of the closed state by the bond disruption suggests that it is indeed present in the resting conformation and needs to be broken for the channel to open. These data can explain the ability of these mutations to effect skeletal dysplasia in humans (3–5). In sum, maintaining an interaction between Y602 and R616 allows normal gating, while destabilization of this interaction leads to increased channel activity and a GOF channel and organismal phenotype.

Several lines of evidence are consistent with the interaction between Y602 and R616 serving as a latch in the resting-closed state that needs to be broken for channel gating. First, we have tested a Y602D mutant, designed to replace the cation–π bond by a salt bridge. The mutant showed an activation level comparable to that of the WT channel (Fig. 5B), indicating the mechanistic importance of an attractive interaction in that region, regardless of the chemical nature of the engagement. Second, we generated a double-cysteine mutant, Y602C/R616C, where the putative cation–π interaction could be replaced by a disulfide bond. Some channel activity was observed in Y602C/R616C mutants when the disulfides were disrupted with the addition of reducing agent (2-ME); however, the oxidant H₂O₂ severely decreased activity (Fig. 4D). Together, the data suggest that stabilization of Y602–R616 interactions stabilizes the resting-closed state of the channel.

The functional information from the presented mutagenesis studies can be interpreted in the context of the known structures of the TRPV channel family. We have previously published a refined homology model for the closed state of the rat TRPV4 channel based on the 3J5P cryo-EM structure of TRPV1 (22). This model predicts proximity between Y602 and R616. The conventional molecular dynamics force fields [specifically, CHARMM36 (35, 36), which was used for refinement of the model] cannot properly reflect cation–π interactions because all of the atoms are represented as point charges, and the electronic π-orbitals are not explicitly present; thus, we did not observe geometry dictating cation–π bonding. But, because Y602 and R616 are close to each other in the model, it is feasible that in a similar arrangement in vivo they would have a cation–π interaction, similar to what is observed in experimentally resolved protein structures (25, 37). In the structure of the closed conformation of the TRPV1 channel (3J5P) that was used as a template for the model, the electrostatic energy of the cation–π interaction for the analogous residues Y565 and R579 was −2.85 kcal/mol (estimated using the web server at capture.caltech.edu), comparable to the strength of a medium-strength H bond, and would amount to −11.4 kcal/mol for the tetramer. In contrast, the activated TRPV1 structures with toxin and capsaicin (3J5Q and 3J5R) do not show any appreciable cation–π interaction. In those structures, these side chains are oriented differently, which makes any cation–π interactions highly unlikely. This observation corroborates the idea that this interaction favors a resting-closed state rather than the open state of the channel.

It should be noted that the TRPV4 channel is known to have at least two nonconductive states—the resting-closed and inactivated (38, 39). These states cannot be distinguished by conductivity measurements. Before stimulus application, the channel spends most of its time in the resting-closed state, and slowly enters the inactivated state only from the sustained open conformation (Fig. 1A). The observation that disulfide bridges are observed in ambient conditions suggests a proximity in the resting-closed state; the mutagenic data are also consistent with this notion. Additional evidence is obtained by assessing the propensity of channels with mutations to inactivate. Regardless of the effect on the opening transition, single mutants at the R616 location (R616Q, R616C, and R616K) had displayed an inactivation rate similar to WT (Figs. 1 D and F and 5C). In contrast, the single mutants at Y602, namely Y602C and Y602D, showed a noticeable decrease in inactivation, while Y602F had more rapid kinetics of transition into the inactivated state (SI Appendix, Fig. S1). This suggests that Y602, rather than R616 or the Y602–R616 interaction, is essential for the inactivation process. It is likely that in the inactivated state, Y602 changes the contact environment relative to the resting-closed conformation. While the data do not exclude that R616 changes its environment as well, the absence of an energetic impact of the tested R616 mutations indicates that rearrangements in the R616 region might not define the cost of the inactivation process.

For this study, we have built a homology model of rat TRPV4 based on the cryo-EM structure of Xenopus tropicalis TRPV4 (6BBJ) (16) and compared it with our previously published model based on TRPV1; both TRPV1- and TRPV4-based homology models are available in SI Appendix, along with a comparison of the key features of S4–S5 linker interactions with the TRP helix (SI Appendix, Figs. S2 and S3) and movies (Movies S1, S2, and S3; see SI Appendix for movie legends) of the conformational transition between them. The TRPV4-based structural model has several prominent distinctions from the TRPV1-based one, including the straightened kink between the S4–S5 linker and S5 helix that essentially makes them a single continuous alpha-helical segment. This rearrangement results in a higher tilt of the S4–S5 fragment and positions R616 deep into the hydrophobic region of the membrane. The separation between Y602 and R616 in this conformation makes any direct contact between the residues unfeasible, which makes cation–π interaction in the WT channel, or formation of a salt bridge or a disulfide in the mutants, very unlikely unless these mutations substantially change the structure of the channel. Both homology models clearly have a tightly closed hydrophobic cytoplasmic gate. In sum, our study strongly suggests that the resting-closed state of the TRPV4 channel is similar to the TRPV1 structure but not the recently resolved TRPV4 structure.

The finding that the TRPV4 nonpermeating structure is apparently not the resting-closed state thus becomes a question of its relevance—does it reflect a true state? A perusal of TRP structures demonstrates that, while the vast majority of them have a kink following the S4–S5 linker region as in what appears now to be the resting-closed state, at least two other TRP resolved channel structures do not, looking more like the Deng Z, et al. (16) TRPV4 structure: TRPM8 (6BPQ) and TRPML1 (5WPT). The latter is remarkable, because there is also a structure of that channel in another conformation (5WPQ) where the S4–S5 region is more kinked and there is a cation–π interaction, as in the resting-closed state. Given that nonkinked conformations are now observed in at least three channel structures and thus presumably represent a natural nonpermeating state, the other low-energy state that does not conduct would be an inactivated state, and both TRPM8 and TRPML1 also show inactivation electrophysiologically (40, 41), we propose that the TRPV4-based model reflects an inactivated state where Y602 and R616 are still separated after the opening transition but the gate has already been rendered nonconductive. In further support of this hypothesis, the closed-state model has a direct contact between the L596 backbone and W733 side chain that was shown to be essential for stabilizing the closed channel (22), whereas this contact is absent in the model of the inactivated state. Remarkably, modeling shows a potential flow between the resting-closed, open, and potential inactivated states. Starting with the closed-state model, we suggest the following rearrangements are occurring during the opening transition (Fig. 6): (i) disruption of the Y602–R616 (or K616) interaction; (ii) release of the S4–S5 linker and S5 helix domains; (iii) disruption of the L596–W733 latch; (iv) displacement of the loop between S4 and the S4–S5 linker; (v) outward motion of the TRP domain; and (vi) opening of the gate. However, it is important to note caveats. First, more experimental evidence is necessary to definitively demonstrate that the recent cryo-EM TRPV4 structure is indeed in the inactivated state. In addition, the proposed order of conformational changes is likely for GOF mutants that have a weakened or absent Y602–R616 bond (e.g., R616Q); however, it might differ for other mutants or a WT channel with alternate stimuli. For example, for WT TRPV4 stimulated by lateral tension, disruption of the contact between the TRP helix and the loop preceding the S4–S5 linker may happen before the outward motion of the TRP helix.

Fig. 6.

Conformational changes likely to be involved in the TRPV gating transition for the GOF mutants with disrupted Y602–R616 interactions (not necessarily in this order): (1) disruption of the Y602–K616 interaction; (2) release of the S4–S5 linker and S5 helix domains; (3) disruption of the L596–W733 latch; (4) displacement of the loop between S4 and the S4–5 linker; (5) outward motion of the TRP domain; and (6) opening of the gate.

Conclusions

Two point mutations, R616Q and Y602C, involved in spondyloepiphyseal dysplasia Kozlowski type and spondyloepiphyseal dysplasia Maroteaux type, lead to GOF channel phenotypes, apparently due to the breaking of a cation–π interaction necessary to stabilize the closed state.

The effects of mutations targeting the Y602–R616 interaction on channel activity and the cell-growth phenotype consistently suggest that (i) this bond is present in the resting-closed state and the connection dissociates during the opening transition, and (ii) the Y602–R616 contact is most likely stabilized by cation–π interactions of the side chains, though (iii) mechanistically, stabilization via a bond of any kind (cation–π, salt bridge, or disulfide) is sufficient to stabilize the resting-closed state.

These observations, supported by analysis of the homology models based either on TRPV4 (6BBJ) or the closed TRPV1 structure (3J5P), indicate that the latter, having Y602 and R616 in close proximity, is likely to resemble the resting-closed conformation of the rat TRPV4 channel, whereas the model based on the amphibian TRPV4 structure might be more similar to an inactivated state of the channel where the gate is closed but Y602 and R616 are spatially separated.

In combination with previous studies (22, 38), the findings highlight the role of interactions of the S4–S5 linker, in the vicinity of the cytoplasmic gate region and the TRP helix, in the control of the gating cycle of the TRPV family of channels.

Methods

Mutagenesis and cRNA Synthesis.

All of the mutations were introduced into pGH19-TRPV4 using the QuikChange Lightning Site-Directed Mutagenesis Kit (Agilent Technologies). The entire cDNA of WT and mutant TRPV4 alleles was verified by sequencing. cRNA was synthesized from Xho1-linearized pGH19-TRPV4 or its mutations using an mMessage mMachine T7 Kit (Ambion).

Oocyte Preparation.

Xenopus laevis ovaries were purchased from Nasco (LM00935MX). The ovaries were cut into small pieces (fewer than 40 oocytes), washed, and incubated in Barth’s solution [88 mM NaCl, 1 mM KCl, 2.4 mM NaHCO₃, 0.82 mM MgSO₄, 0.33 mM Ca(NO₃), 0.68 mM CaCl₂, 10 mM Hepes, pH 7.4] supplemented with 25 μg/mL ampicillin, 6 μg/mL tetracycline, and 0.1 mg/mL gentamicin at 18 °C. Healthy-looking oocytes of V to VI stages were selected after collagenase (type A) digestion and defolliculation treatment. The oocytes were injected with the proper amount of cRNA (0.1 to 5 ng), incubated in ND96 solution (96 mM NaCl, 2 mM KCl, 1 mM MgCl, 1.8 mM CaCl₂, 5 mM Hepes, pH 7.4), supplemented with 0.1 mg/mL gentamicin (Fisher Scientific) at 18 °C for 3 to 4 d, and then used for electrophysiological recording.

Electrophysiological Techniques.

The two-electrode voltage clamp technique was used to study the channel activities of the WT and its mutants as described in previous studies (1, 2). Briefly, currents were recorded using an Oocyte Clamp OC-725 amplifier (Warner Instruments), signal was interfaced through a Digidata 1440A digitizer (Molecular Devices, LLC), and data were analyzed using the software Clamp 10.3 (Molecular Devices, LLC). Pipettes were filled with 1 M KCl and showed resistance of 0.8 to 2 MΩ. The bath solution contained 98 mM KCl, 2 mM MgCl₂, 10 mM Hepes (pH 7.2), and 0.02 mM flufenamic acid (all from Sigma). 2-Mercaptoethanol (0.5%; Sigma) and H₂O₂ (0.15%; Fisher Scientific) were added to both solutions.

Yeast Expression of TRPV4 and Its Mutations.

The cDNA fragment of WT or mutant TRPV4 was PCR-amplified using the KOD Hot Start DNA Polymerase (EMD Millipore). The PCR products were digested with BamHI and HindIII, purified, and ligated with the inducible URA-selectable yeast expression plasmid p416GAL1. After sequence confirmation, the resulting plasmids or the empty vector was introduced into the yvc1Δ yeast strain (BY4742) (MATα, his3Δ1, leu2Δ0, lys2Δ0, ura3Δ0, tok1::KanMX4, yvc1::HIS3) according to high-efficiency yeast-transformation protocols (3). The growth phenotype of yeast cells harboring plasmids with TRPV4 or mutant alleles was investigated using a spot assay. Starting at 1,000 cells per μL, 5 μL culture in serial 10-fold dilutions was spotted on a solid expressive medium (2% galactose and 1% raffinose) or on a repressive medium (2% glucose) and incubated at 30 °C for 2 to 3 d.

Western Blotting.

At 4 d after cRNA injection, 100 oocytes expressing the WT or its mutants were washed and homogenized with 2 mL of ice-cold Tris buffer [50 mM Tris⋅HCl, 150 mM NaCl, 1 mM PMSF, 5% protein inhibitor mixture (P8340; Sigma), pH 8.0]. Oocyte lysate was centrifuged at 10,000 rpm (Thermo Scientific ST 16R centrifuge, rotor 75003652) for 10 min at 4 °C to remove cell debris. The supernatant (as total proteins) was transferred to an ultracentrifuge tube and centrifuged at 70,000 rpm (Backman centrifuge model J25, rotor JA25.50) for 1 h at 4 °C. The supernatants were transferred to a new tube as a supernatant-protein component, and the pellets were resuspended in 2 mL of ice-cold Tris buffer with 0.1% Triton X-100 as a membrane-protein component. Total proteins, supernatant proteins, and membrane proteins of each injection were subjected to SDS/PAGE (4 to 15% Criterion Precast Gel; Bio-Rad) under conditions with and without 2-ME (5%) and analyzed by Western blotting with anti-TRPV4 antibody (OST00185W; Osenses)

Statistical Analysis.

Data are expressed as means ± SEM. The significance of differences was evaluated using an unpaired Student’s t test, with a maximum P value of <0.05 being required for significance.

Homology Modeling.

The homology model of the rat TRPV4 channel based on the template of the cryo-EM structure of an amphibian TRPV4 (6BBJ) (16) was created using the Swiss-Model server (42). Comparison of the modeled structure with results based on the same template but provided by other servers [e.g., I-TASSER (43) and Phyre2 (44)] showed a high consistency of the predictions. The coordinates of the model are available in SI Appendix.