bioRxiv preprint doi: https://doi.org/10.1101/2020.01.04.894980; this version posted January 6, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
1 An unorthodox mechanism underlying voltage sensitivity of TRPV1 ion channel
2 Fan Yang1¶,2, Lizhen Xu1, Bo Hyun Lee2, Xian Xiao2,3, Vladimir Yarov-Yarovoy2 and 3 Jie Zheng2¶
4 1Department of Biophysics and Kidney Disease Center, First Affiliated Hospital, 5 Institute of Neuroscience, NHC and CAMS Key Laboratory of Medical Neurobiology, 6 Zhejiang University School of Medicine, Hangzhou 310058, China. 2Department of 7 Physiology and Membrane Biology, University of California, Davis, California 95616, 8 USA. 3Institute for Basic Medical Sciences, Westlake Institute for Advanced Study, 9 Westlake University, Shilongshan Road No. 18, Xihu District, Hangzhou 310064, 10 Zhejiang Province, China.
11 ¶ Correspondence and requests for materials should be addressed to F.Y. (email: 12 [email protected]) or J.Z. (email: [email protected])
13
14 Abstract
15 While the capsaicin receptor TRPV1 channel is a polymodal nociceptor for heat, 16 capsaicin, and proton, the channel’s responses to each of these stimuli are 17 profoundly regulated by membrane potential, damping or even prohibiting its 18 response at negative voltages and amplifying its response at positive voltages. 19 Though voltage sensitivity plays an important role is shaping pain responses, 20 how voltage regulates TRPV1 activation remains unknown. Here we showed that 21 the voltage sensitivity of TRPV1 does not originate from the S4 segment like 22 classic voltage-gated ion channels; instead, outer pore acidic residues directly 23 partake in voltage-sensitive activation, with their negative charges collectively 24 constituting the observed gating charges. Voltage-sensitive outer pore movement 25 is titratable by extracellular pH and is allosterically coupled to channel 26 activation, likely by influencing the upper gate in the ion selectivity 27 filter. Elucidating this unorthodox voltage-gating process provides a mechanistic 28 foundation for understanding polymodal gating and opens the door to novel 29 approaches regulating channel activity for pain managements. 30
31 When the lipid bilayer emerged to enclose living cells more than three billion years 32 ago1, this diffusion barrier between cytoplasmic milieu and external environment 33 allowed the establishment of transmembrane ion concentration gradients, which in 34 turn yielded a transmembrane electric potential. Such a membrane potential (Vm) has 35 been widely utilized in cellular signaling: voltage-gated ion channels alter Vm to elicit 36 electric signals for rapid communications2,3; voltage-sensitive enzymes, e.g., Ci-VSP4, 37 couple changes in Vm to the regulation of enzymatic activities and intracellular 38 signaling. For these “classic” voltage-sensing proteins, high sensitivity to voltage (5- 39 fold/mV) has been attributed to a conserved, densely charged “voltage-sensor” 40 domain in the transmembrane region5. Other membrane proteins—including G 41 protein-coupled receptors6, ion channels and transporters7—have also evolved to take 42 cues from Vm to perform their biological functions. Understanding the origin and 43 operation of voltage sensitivity in these membrane proteins is of fundamental 44 importance.
1 bioRxiv preprint doi: https://doi.org/10.1101/2020.01.04.894980; this version posted January 6, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
45 Transient receptor potential vanilloid 1 (TRPV1) channel, a polymodal nociceptor in 46 higher species8, is a good representative of the non-classic voltage-sensitive 47 membrane proteins. TRPV1 is activated by noxious heat above 40°C9, however, its 48 response to heat at depolarized Vm is markedly larger than that at the resting Vm even 49 when the driving force for its non-selective cation current is equal in magnitude (Fig. 50 1a). Similarly, TRPV1’s responses to capsaicin and proton—which elicits spiciness 51 sensation9 and the sustained phase of pain sensation10, respectively—are also 52 substantially asymmetrical (Fig. 1b). Tuning of TRPV1’s nociceptive sensitivity by 53 Vm is a highly dynamic process occurring at human body temperature, with the range 54 of voltage dependence (reflected by the conductance-voltage, or G-V, relationship) 55 shifting over a more than 200 mV range in the presence of an activation stimulus11 56 (Fig. 1c). Dynamic voltage sensitivity makes TRPV1 more sensitive to noxious 57 stimuli when its host sensory neuron has been previously excited (primed) or 58 damaged, a general feature for nociception known as hyperalgesia12. On the flip side, 59 membrane hyperpolarization is expected to inhibit pain responses, offering 60 opportunities for novel clinical intervention. Despite its physiological significance, 61 how TRPV1 senses Vm remains largely elusive.
62 TRPV1 belongs to the voltage-gated-like (VGL) ion channel superfamily (Fig. 1d)13, 63 for which three voltage-sensing mechanisms have been established. For voltage-gated 64 potassium (Kv), sodium (Nav) and calcium (Cav) channels (blue sections in Fig. 1d), 65 multiple regularly spaced basic residues on the transmembrane S4 helix enable it to 14 66 move in response to Vm changes . For two-pore domain potassium channels (K2P) 67 without a voltage-sensor domain (green section in Fig. 1d), permeant cations have 68 been proposed to bestow voltage sensitivity15. For inward-rectifier potassium channels 69 (Kir) (purple section in Fig. 1d), voltage-dependent pore block by endogenous 70 polyamines and Mg2+ ions introduces voltage sensitivity16. These three mechanisms 71 are applicable to about three-fourth of the VGL superfamily; voltage-sensing is less 72 understood for a significant portion of the family (red section) that is composed of 73 cellular sensor TRP channels including TRPV1. Our study began by determining 74 whether TRPV1 employs one of these three known mechanisms or a new mechanism 75 to detect Vm.
76 Characterizing TRPV1’s voltage sensitivity
77 A fundamental parameter for voltage sensitivity is the total charge movement, q, 78 which determines the steepness of voltage response5. TRPV1 is known to exhibit 79 shallow voltage dependence11. By fitting a Boltzmann function to the G-V curve of 80 mouse TRPV1 recorded in HEK293 cells (Fig. 1c), the apparent q was estimated to be 81 0.72 ± 0.05 e0 (n = 4). Applying this approach to Kv channels yielded an apparent q of 17 82 5.3 e0 . However, it is well known that fitting the shape of G-V curves could yield an 83 under-estimate of q, for example, when there are multiple voltage-dependent closed 84 states that the channel may traverse before reaching the open state17-19. Indeed, the 85 total gating charge of Kv, K2P and Kir channels are about 13 e0, 2.2 e0 and 2.2 e0, 86 respectively15,16,18.
87 The relatively weak voltage sensitivity of TRPV1 precluded direct measurement of q 88 from gating current. Alternatively, the limiting-slope method20,21 is a classic approach
2 bioRxiv preprint doi: https://doi.org/10.1101/2020.01.04.894980; this version posted January 6, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
89 to estimate q, for which voltage dependence was measured at low open probabilities. 90 However, the voltage-driven transition of TRPV1 is allosterically coupled to 91 activation gating22, with the channel open probability approaching a stable level at 92 deep hyperpolarization (Fig. 1e and Extended Data Fig. 1) that reflected the 93 equilibrium of activation gate between the closed and open state23. To better estimate 94 q, we first calculated the derivative of the mean log open probability with respect to 95 voltage and fitted the Boltzmann smoothing function of voltage dependence of mean 96 log open probability (see Methods for details) (Fig. 1f), as was previously performed 97 on the allosteric large conductance calcium activated potassium (BK) channels23. This 98 approach yielded an estimated q of 0.93 e0, which is 30% larger than that measured 99 from fitting a Boltzmann function to the G-V curve (0.72 e0), yet much smaller than 23,24 100 that of BK channels (2.62 e0) . To estimate the potential deviation of the gating 101 charge estimate from its true value, we simulated the voltage sensitivity of TRPV1 102 open probability based on a simple allosteric scheme (Extended Data Figure 2a). With 103 the hypothetical gating charge value set to span a wide range from 1.0 e0 to 13.0 e0, 104 our estimations (Extended Data Figure 2b to 2e, dotted red curves) were close to the 105 true value, with deviations ranging from 3.3% to 5.3% (Extended Data Figure 2f). 106 When the method was applied to BK channels, for which total gating charge could be 107 directly measured from the gating current, an underestimation of 7.7% was seen23,24.
108 TRPV1 S4 does not serve as a voltage sensor
109 The S1-to-S4 domain of TRPV1 channels, like its counterpart in Kv channels25, forms 110 a compact domain surrounding the channel pore26,27. In Kv channels, this domain 111 serves as the voltage sensor28. To test whether TRPV1 also relies on the S1-to-S4 112 domain to sense Vm, we first compared the S4 sequences (Fig. 2a). The Kv channel 113 S4 segments contain seven regularly spaced charged residues. The corresponding 114 residues in TRPV1 are all uncharged except for R558, which matches to R6 of Kv at 115 the end of S4 thought to be insignificant for voltage-sensing28. To test for a potential 116 contribution of R558 to voltage sensitivity, we mutated this residue to leucine. 117 Charge-neutralization did not abolish voltage activation (Fig. 2b and c). While the G- 118 V curve of R558L was left-shifted (Fig. 2d), the estimated q value, at 0.88 e0, was not 119 significantly reduced compared to that of the wild-type channels (Fig. 2e). Similar 120 observations were previously reported when the equivalent residue in rTRPV1 was 121 mutated to A, L, and K29. Therefore, R558 in TRPV1 S4 cannot serve as the main 122 gating charge carrier. These results showed that the S4 voltage-sensing mechanism of 123 Kv channels is not applicable to TRPV1.
124 Permeant ions do not affect voltage sensitivity of TRPV1
125 Since an S4-based voltage-sensing mechanism is inapplicable to TRPV1, we 126 examined other known mechanisms. To test whether TRPV1 uses permeant ions to 127 sense voltage like K2P channels, we took advantage of TRPV1’s low ion selectivity8. 128 We reasoned that, if q is originated from permeant ion movement, switching from 129 monovalent cations to divalent cations would likely alter the measured q value, 130 doubling it if these cations bind to the same site(s) in the pore. When the intracellular 131 permeant ions were switched from Na+ to Mg2+, the voltage activation and 132 deactivation kinetics were both much slower (Fig. 2f). The G-V curve was right-
3 bioRxiv preprint doi: https://doi.org/10.1101/2020.01.04.894980; this version posted January 6, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
133 shifted but without any noticeable change in shape, with an estimated total gating 134 charge of 0.99 e0 (Fig. 2g). Therefore, unlike in K2P channels, permeant ions do not 135 appear to serve as the voltage sensor for TRPV1.
136 TRPV1’s voltage sensitivity is not originated from permeation block
137 Like K2P channels, Kir channels also lack the S1-S4 voltage sensor domain. Their 138 currents exhibit voltage-dependence because of voltage-dependent pore block by 139 endogenous polyamines and Mg2+ ions at depolarized voltages16. Such pore block is 140 much faster than the resolution of patch-clamp recordings so that, when it occurs, the 141 single-channel conductance is reduced. As a result, the single-channel current 142 amplitude exhibits similar voltage dependence as the macroscopic current30. To test 143 whether TRPV1 employs this pore-block mechanism for sensing voltage, we 144 compared single-channel current and macroscopic current recorded at voltages where 145 TRPV1 was clearly activated by depolarization (Fig. 2h and i). It was obvious that the 146 single-channel current amplitude was linearly dependent on voltage and did not 147 follow the voltage dependence of macroscopic current (Fig. 2i). Moreover, in inside- 148 out patch recordings the voltage dependence did not decrease upon continuous 149 perfusion of the bath solution for about three minutes (Extended Data Fig. 3). In 150 comparison, Kir channels lost the inward-rectification feature within two minutes of 151 patch excision due to wash-off of endogenous blockers16. Therefore, unlike Kir 152 channels, voltage sensitivity of TRPV1 is unlikely originated from pore block.
153 TRPV1 voltage-dependent gating behavior resembles a concerted transition
154 As none of the known voltage-sensing mechanisms in the VGL channel superfamily 155 is applicable to TRPV1, an unorthodox mechanism exists. We gained the first glimpse 156 of such a mechanism from kinetic measurements. Voltage activation of Kv2.1 channel 157 exhibited a sigmoidal time course31 (Fig. 3a), where there was an initial delay in 158 macroscopic current upon depolarization (marked by red arrow). This delay is due to 159 multiple closed states Kv channels must traverse before reaching the open state20 (Fig. 160 3c, SCHEME I). In contrast, voltage activation of TRPV1 nicely followed a double- 161 exponential time course (Fig. 3a and Fig. 2b, red solid traces), with no detectable 162 delay in current onset. Indeed, time courses predicted by gating schemes for Kv 163 channel in the form of SCHEME I poorly described the voltage activation kinetics of 164 TRPV1 (Fig. 3a, blue dash trace). In addition, it is well established that for Kv 165 channels holding Vm at more hyperpolarized voltages would produce a longer delay in 166 current activation upon depolarization. Such a longer delay, named Cole-Moore 167 shift32, is again due to the existence of multiple closed states preceding the open state. 168 When TRPV1 channels were voltage-clamped at various hyperpolarized voltages, 169 there was no detectable Cole-Moore shift (Fig. 3b). The absence of both initial delay 170 in macroscopic current activation and Cole-Moore shift, as well as the double- 171 exponential current time course, can be explained by an allosteric model as shown in 172 Figure 3c, SCHEME II, in which there is a single concerted voltage-dependent 173 transition that may represent either the C↔O transition or a separate transition that is 174 allosterically coupled to the C↔O transition. (The observation of a double- 175 exponential activation time course rules out a simpler two-state C↔O model for 176 voltage-dependent activation.) Admittedly, more complex models, for example, with
4 bioRxiv preprint doi: https://doi.org/10.1101/2020.01.04.894980; this version posted January 6, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
177 multiple rapid voltage-dependent transitions and a single rate-limiting voltage- 178 independent transition right before the first opening transition, could also produce 179 similar observations in channel activation time course and the absence of a Cole- 180 Moore shift33. However, none of the currently available experimental evidence 181 requires such a complex gating model and, as discussed above, TRPV1 voltage 182 activation involves an allosteric coupling. Furthermore, the simple allosteric model in 183 Figure 3c, SCHEME II, is consistent with our results described below.
184 Identifying the voltage-dependent transition
185 Besides membrane depolarization, TRPV1 can be allosterically activated by distinct 186 physical and chemical stimuli, many of which have been better characterized both 187 structurally and functionally22,26,27,34. A multi-allosteric gating model initially 188 proposed for BK channels35,36 can adequately describe TRPV1 activation37-40. The 189 polymodal nature of TRPV1 activation offered an opportunity to identify the voltage- 190 dependent transition. For a general allosteric system, there exist three possible 191 scenarios as shown by SCHEME II (Fig. 3c) that are experimentally distinguishable 192 by measuring open probability, Po, at deeply hyperpolarized voltages:
193 A. Voltage directly controls the C↔O transition. In this scenario, the channel 194 can be tightly held in the C state by hyperpolarization; no other stimuli 195 should be able to evade voltage’s control over the C↔O transition to cause 196 channel activation. 197 B. Voltage controls a transition, R↔A, that is allosterically coupled to the 198 C↔O transition, and another stimulus, X, also activates the channel 199 through the same transition. In this scenario, hyperpolarization can 200 effectively prevent channel activation by X but not by other stimuli (for 201 example, Y). 202 C. Voltage controls a distinct transition, independent of other stimuli, that is 203 allosterically coupled to the C↔O transition. In this scenario, all other 204 stimuli can efficiently activate the channel at hyperpolarized voltages.
205 To distinguish these scenarios, we measured NPo from single-channel events at deep 206 hyperpolarization down to -210 mV in the absence or presence of another stimulus. 207 We found that, in the absence of another stimulus, NPo appeared to approach a steady 208 level at deeply negative voltages (Extended Data Fig. 1). In agreement with previous 37,38,40,41 209 reports , the observed baseline Po level was very low, with an upper limit 210 estimate near 0.01 at room temperature (Fig. 1e). Nonetheless, the presence of a 211 voltage-independent baseline Po is inconsistent with Scenario A, suggesting that the 212 C↔O transition itself must be associated with little charge movement and have an 213 intrinsic open probability (Po_intrinsic) of less than 0.01.
214 In the presence of a saturating concentration (3 µM) of capsaicin, TRPV1 current was 215 readily detectable even at -210 mV (Fig. 3d, red arrow). The current could be almost 216 completely blocked by ruthenium red (Fig. 3d, black circles), confirming that 217 contamination by leak current must be small. The estimated Po plateaued above 0.1 218 when voltage was hyperpolarized below -160 mV (Fig. 3e, red dash line; see also Fig. 40,41 219 1c). Similar to previous reports , the presence of a stable Po substantially above 220 Po_intrinsic indicated that capsaicin could allosterically promote the C↔O transition
5 bioRxiv preprint doi: https://doi.org/10.1101/2020.01.04.894980; this version posted January 6, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
221 even at deep hyperpolarization. Therefore, voltage activation cannot be carried out 222 through the same transition promoted by capsaicin, i.e., if Scenario B is true, 223 capsaicin cannot be stimulus X. Indeed, as capsaicin induces TRPV1 activation by 224 linking S4 to the S4-S5 linker42-44, these results offered further supports for the 225 conclusion that voltage activation does not involve S4. Furthermore, the observation 226 of a voltage-independent, elevated Po in the presence of capsaicin also validated the 227 conclusion that voltage cannot work through the C↔O transition to open the channel, 228 i.e., Scenario A is invalid.
229 Like capsaicin, we found that many other TRPV1 activation stimuli, including heat45, 230 were also effective at deeply hyperpolarized voltages (see, for example, Fig. 1c). 231 However, the situation was different when extracellular protons were used as the 232 activator. We found that even in the presence of a saturating concentration of protons 233 (at pH 4.0), current decreased towards zero at hyperpolarizing voltages (Fig. 3f). No 234 plateau could be observed in the Po–V plot (Fig. 3g); instead, Po kept decreasing 235 towards Po_intrinsic (Fig. 3h). It is known that protons partially block TRPV1 236 permeation, which also exhibits a shallow voltage dependence46. However, proton 237 blockade had been corrected for before calculating Po (Extended Data Fig 4.). 238 Therefore, these observations fit the prediction of Scenario B in that deep 239 hyperpolarization can shut down proton activation, identifying that voltage works 240 though the proton activation pathway. These results also voided Scenario C by 241 identifying protons as the stimulus X.
242 Location of the gating charges
243 While unanticipated, the finding that voltage sensitivity of TRPV1 resides in the 244 proton activation pathway might not be surprising. It is well established that proton 245 activation involves protonation of charged residues in the outer pore47, a region 246 intimately involved in TRPV1 activation gating and known to undergo substantial 247 conformational rearrangements48. The outer pore is also noticeably rich in charged 248 residues (Extended Data Fig. 5a). If some of these charged residues locate within or 249 near the Vm field, their movements during channel activation will impart voltage 250 sensitivity to the process5. Indeed, when we simultaneously neutralized titratable 251 negative charges by changing the extracellular pH to 4.0, the total gating charge 252 estimate decreased to 0.47 e0 (Fig. 3i). Given that the pKa value for the sidechain of 253 negatively charged Glu or Asp is around 4, at extracellular pH 4.0 only about half of 254 these charged residues would be neutralized. Therefore, we reasoned that though we 255 observed only an around 50% reduction of q at acidic pH, nearly all of the 0.93 e0 256 total gating charge could be collectively contributed by titratable acidic residues. 257 Consistent with a previous report49, no obvious change in the total charge movement 258 could be confidently identified from channels carrying single or double charge 259 neutralization mutations (Extended Data Fig. 5b and 5c), likely because their 260 individual impacts to q were too small to stand out of the noise. These mutagenesis 261 tests confirmed that the total gating charge of about 1 e0 is not carried by any single 262 charged residue (which would require the residue to move nearly 1/4 of the whole Vm 263 field during activation). Instead, it is most likely that multiple charged residues move 264 collectively during voltage gating. As a negative control we observed no change in 265 total gating charge in the presence of the classic TRPV1 agonist capsaicin (Fig. 3j).
6 bioRxiv preprint doi: https://doi.org/10.1101/2020.01.04.894980; this version posted January 6, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
266 Based on these findings, we tentatively placed the titratable total gating charge at the 267 voltage-driven R↔A transition that is also affected by protons (SCHEME III, Fig. 3k). 268 A gating model incorporating this feature predicted behaviors that qualitatively 269 resemble our experimental observations (Extended Data Fig. 6).
270 Structural mechanism for voltage/proton activation
271 The TRPV1 closed-state structure and vanilloids/toxin-activated structures have been 272 resolved by cryo-EM26,27,34. However, these structures are missing a part of the outer 273 pore important for voltage and proton activation; in addition, no voltage/proton- 274 induced open structure is currently available. To understand the structural basis for 275 voltage/proton-dependent gating, we first modeled the proton-induced open state 276 structure using Rosetta structural modeling suite (see Methods and Extended Data Fig. 277 7). Comparison between the proton-activated state model and the closed state 278 structure suggested two major conformational changes. Firstly, the selectivity filter 279 region moved away from the central axis to yield an open conformation of this upper 280 gate (Fig. 4a, gray to red), similar to DkTx-induced conformational changes observed 281 in the cryo-EM studies27,34. Secondly, the pore turret regions moved substantially 282 closer to each other upon protonation (Fig. 4a, dark gray to cyan), which has been 283 previously observed by fluorescence recordings during heat activation50. Associated 284 with these structural changes are relocation of multiple charged residues.
285 To functionally verify the predicted outer pore movement, we employed FRET-based 286 patch fluorometry to simultaneously record conformational and functional states of 287 TRPV1 channels50,51. Our structural models predicted that the distance between C622 288 residues of neighboring subunits (measured at Cβ atoms; Fig. 4a, yellow) is reduced 289 by about 13 Å in proton-induced activation. This large distance change would be 290 readily detectable using fluorescein maleimide (FM) and tetramethylrhodamine 291 maleimide (TMRM) as a FRET pair50. Extracellular acidification (pH 5.0) elicited a 292 large current from the fluorophore-labeled channels (Fig. 4b). Fluorescence imaging 293 from the same channel population revealed an increase in the TMRM/FM intensity 294 ratio that indicated an increase in FRET efficiency (Fig. 4b). This change in emission 295 spectrum was absent when fluorophores were attached to cysteine introduced at N467 296 on the S1-S2 linker, or at non-specific sites on untransfected cells (Fig. 4c and d). 297 Therefore, results from the FRET experiments supported the prediction that the turrets 298 move toward each other in the proton-activated state.
299 How does the observed outer pore conformational changes produce voltage-sensitive 300 gating? While much of the detail remains to be elucidated, it is conceivable that 301 relocation of some negatively charged residues within the Vm field might be the origin 302 of the gating charge (Fig. 4e). Indeed, from the activated-state model, we could 303 identify multiple negatively charged residues in the outer pore region that exhibited 304 complex and collective movements (Fig. 4f): D602, D647, E652 and D655 moved 305 downward along the z axis, E601 and E649 moved upward (Extended Data Table 2). 306 Based on the simplified assumption of a uniform Vm field that spans a 30 Å vertical 307 depth, these charge movements alone would be more than sufficient to produce 1 e0. It 308 was also noticed that the modelled gating rearrangements would lead to changes in 309 the electrostatic potential surface of the outer pore, in which a strong negative
7 bioRxiv preprint doi: https://doi.org/10.1101/2020.01.04.894980; this version posted January 6, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
310 potential in the resting state dissipates as the channel transitions into the activated 311 state (Fig. 4g). Such conformational changes present a plausible picture of the 312 voltage-sensitive gating process: it is the collective gating rearrangements of 313 negatively charged residues and the local electric field that produce an unorthodox 314 mechanism of voltage-sensing in TRPV1.
315 Discussion
316 Our study suggested that voltage-sensing in TRPV1 originates from conformational 317 changes of the outer pore, where titratable acidic residues collectively contribute to 318 the total charge movement. Because that the pKa value for the sidechain of negatively 319 charged Glu or Asp is around 4, extracellular acidification to a pH level of 4.0 would 320 only neutralize about half of the charges in these charged residues. It was 321 experimentally infeasible to further decrease extracellular pH in patch-clamp 322 recordings to fully neutralize all the negative charges, though it is most likely that the 323 about 1 e0 total gating charge is collectively contributed by titratable acidic residues. 324 Upon depolarization, these residues appear to undergo complex rearrangements that 325 are directly coupled to opening of the activation gate, most likely the nearby upper 326 gate at the selectivity filter26,27,34 (Fig. 4g). The pore turret is another outer pore 327 channel structure seen to move substantially in our model as well as site-directed 328 fluorescence measurements in our previous studies50,52 and the present study (Fig. 4b- 329 d). Details on the nature of pore turret movements remain unclear, as this segment is 330 missing from the available TRPV1 cryo-EM structures; nonetheless, the equivalent 331 part of the orthologous TRPV2 is seen to move substantially between the closed and 332 open states53.
333 The TRPV1 outer pore is a known hot spot for gating modulation of this polymodal 334 receptor. Small cations such as proton, Na+ and Mg2+/Ba2+/Ni2+/Gd3+, as well as large 335 peptide toxins such as DkTx, RhTx and BmP01 all bind to this region to exert their 336 strong gating effects48. Capsaicin activation has been recently found to affect the outer 337 pore44. Heat also induces large conformational changes of the outer pore50. Exploiting 338 charged residues and dipole movements in this gating structure would allow Vm to 339 function essentially as a “master gain setter” to tune the channel’s sensitivity to all 340 major stimuli (see Fig. 1c), a feature that could be of high practical significance for 341 this nociceptor. For example, under pathological conditions such as bone cancer, 342 TRPV1-expressing dorsal root ganglion neurons exhibit higher excitability54. The 343 resting membrane potential of these neurons is depolarized, which leads to higher 344 nociceptive activities of TRPV1, contributing to the severe and intractable pain in the 345 bone cancer. When TRPV1 activity is modulated by hyaluronan, firing frequency of 346 DRG neurons is also changed55. In this regard, understanding the voltage-sensing 347 mechanism may facilitate future pharmaceutical efforts targeting this pain sensor.
348 Comparing to many noxious stimuli for TRPV1 such as capsaicin and heat, voltage is 349 a relatively mild activator. The weak voltage sensitivity ensures that TRPV1 does not 350 act as a dedicated Vm sensor, which is critical for its nociceptive functions. The total 351 charge movement of about 1 e0, though only a fraction of that in classic S4-based 352 voltage-gated channels, can effectively drive man-made transistors in conventional 353 electronic devices4 and many biological functions of membrane proteins. As pointed
8 bioRxiv preprint doi: https://doi.org/10.1101/2020.01.04.894980; this version posted January 6, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
354 out by Hodgkin56, gating charge movement absorbs energy from electrical signaling 355 and acts as a load of the system, therefore evolution tends to optimize its utility. We 356 showed recently that weak voltage dependence is a crucial feature that allows TRPV1 357 to respond to diverse noxious stimuli including heat, whereas the high voltage 358 sensitivity of Kv channels effectively prohibits the channel’s intrinsic heat 359 sensitivity45. It is plausible that similar trade-offs can be found in other voltage- 360 regulated polymodal TRP channels.
361 When the S4 of Kv channels is functionally decoupled from the pore by introducing 362 mutations, a weakly voltage-dependent (q of ~1.8 e0), concerted transition is 363 revealed57. In CNG channels, S4 (containing multiple charged residues) appears to be 364 naturally decoupled from the gate58, and the channel’s intrinsic weak voltage 59,60 365 dependence (q of ~ 0.74 e0) likely comes from outer pore movement . Therefore, it 366 is likely that the outer pore-mediated voltage-sensing mechanism is broadly used in 367 ion channels. Indeed, many TRP channels exhibit similar shallow voltage dependence, 368 sparse charged residues in S4 but richly distributed charged residues in the outer pore, 369 a region seen to under tremendous evolutional selection61. While activation of these 370 channels is also critically regulated by Vm, their voltage-sensing mechanism is 371 unknown (Fig. 1d, section in red). The unorthodox voltage-sensing mechanism we 372 identify in TRPV1 may help elucidate the function of these channels and other 373 membrane proteins.
374
375 AUTHOR CONTRIBUTIONS
376 F.Y., X.L.Z., B.H.L. and X.X. conducted the experiments including patch-clamp 377 recording, mutagenesis, molecular modeling and data analysis; V.Y.-Y. supervised 378 molecular docking and revised the manuscript; J.Z. and F.Y. prepared the manuscript; 379 J.Z., V.Y.-Y. and F.Y. conceived and supervised the project, participated in data 380 analysis and manuscript writing.
381 ACKNOWLEDGEMENTS
382 We are grateful to Fred Sigworth for reading the manuscript and providing advice, 383 and to our lab members for assistance and discussion. This work was supported by 384 funding from National Institutes of Health (R01NS072377 and R01NS103954) to J.Z. 385 and V.Y.Y., American Heart Association (14POST19820027) to F.Y. and American 386 Heart Association (16PRE26960016) to B.H.L. This work was supported by National 387 Science Foundation of China (31741067 and 31800990) to F.Y. This work was also 388 supported by the Core Facilities in Zhejiang University School of Medicine, including 389 the Olympus FV3000 fluorescence imaging microscope and the bioinformatics 390 computation platform.
391 DATA AVAILABILITY
392 All data needed to evaluate the conclusions in the paper are present in the paper 393 and/or the Supplementary Materials. Additional data available from authors upon 394 request.
395
9 bioRxiv preprint doi: https://doi.org/10.1101/2020.01.04.894980; this version posted January 6, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
396 FIGURE LEGENDS
397 Figure 1. Voltage activation of TRPV1. (a and b) Transmembrane voltage strongly 398 modulates TRPV1 activities as shown in representative inside-out patch recordings of 399 TRPV1 activated by a heat ramp (a), high temperature at 42°C, 1 µM capsaicin (CAP) 400 or pH 5.0 (b). Current traces recorded at depolarizing (+80 mV) and hyperpolarizing 401 (-80 mV) voltages are in black and gray, respectively. (c) Conductance-voltage (G-V) 402 curve of TRPV1 (black: V1/2 = 114.9 ± 1.2 mV, apparent gating charge (qapp) = 0.72 ± 403 0.06 e0, n = 4) was shifted to more hyperpolarized voltages in the presence of an 404 activation stimulus (red, heat at 40°C: V1/2 = 54.8 ± 9.8 mV, qapp = 0.94 ± 0.20 e0, n = 405 3; Blue, extracellular pH 5.0: V1/2 = 33.4 ± 7.6 mV, qapp = 0.46 ± 0.03 e0, n = 4; 406 Orange, CAP at 3 µM: V1/2 = -72.3 ± 2.3 mV, qapp = 0.79 ± 0.19 e0, n = 4.). G-V 407 curves are fitted to a single-Boltzmann function (dash curves). Note that with CAP 408 and heat, the G-V curves approach a steady-state level around 0.1 at deeply 409 hyperpolarized voltages. Shaded in green is the physiologically relevant membrane 410 potential range, in which voltage exhibits a strong influence on channel activity. 411 Conductances were normalized to the maximum conductance measured at the highest 412 depolarization voltage. (d) A rootless phylogenetic tree representing channels within 413 the VGL ion channel superfamily adapted from a previous report13. Sections in blue, 414 green and purple cover channels with a known voltage-sensing mechanism. (e) Semi- 415 logarithmic plot of the voltage dependence of normalized mean conductance 416 measured from steady-state current in whole-cell recordings. (f) The voltage 417 dependence of the derivative of mean log open probability. Dashed curve is a 418 Boltzmann smoothing function determined from fits to mean log open probability. 419 Solid curve is a fit of the smoothing function with a Boltzmann function at voltages 420 where the open probability is low. All statistics are given as mean ± s.e.m.
421 Figure 2. None of the three classic voltage-sensing mechanisms appears applicable to 422 TRPV1. (a) Structural alignment of the S4 helixes in TRPV1 (Orange; PDB ID: 5IRZ) 423 and Kv 1.2-2.1 chimera channel (Blue; PDB ID: 2R9R). (b and c) Representative 424 whole-cell recordings of wild-type (WT) TRPV1 and its mutant R558L activated by 425 depolarization with 10-mV increments from -80 mV to +220 mV and from -40 mV to 426 +150 mV, respectively. The voltage activation time courses are well fitted to a 427 double-exponential function (red traces). (d and e) G-V curves of WT (black) and 428 R558L (green, V1/2 = 71.4 ± 12.3 mV, qapp = 0.96 ± 0.06 e0, n = 3.) in linear plot. (e) 429 The voltage dependence of the derivative of mean log open probability of WT (black) 430 and R558L (green). Dashed curve is a Boltzmann smoothing function determined 431 from fits to mean log open probability. Solid curve is a fit of the smoothing function 432 with a Boltzmann function at voltages where the open probability is low. (f) A 433 representative whole-cell recording with Mg2+ as permeating ions. The voltage 434 activation time courses are well fitted to a double-exponential function (red traces). (g) 435 The voltage dependence of the derivative of mean log open probability of WT 436 measured with intracellular sodium ions (black) and magnesium ions (blue). Dashed 437 curve is a Boltzmann smoothing function determined from fits to mean log open 438 probability. Solid curve is a fit of the smoothing function with a Boltzmann function 439 at voltages where the open probability is low. (h) Representative single-channel 440 recordings from an inside-out patch clamped at +40 mV, +80 mV and +160 mV,
10 bioRxiv preprint doi: https://doi.org/10.1101/2020.01.04.894980; this version posted January 6, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
441 respectively. To determine the single-channel current amplitude, their corresponding 442 all-point histograms are fitted to a double-Gaussian function (curves in red). (i) 443 Voltage dependence of single-channel current amplitude (red) and macroscopic 444 current recorded in the whole-cell configuration (gray) (n = 3-5). All statistics are 445 given as mean ± s.e.m.
446 Figure 3. Voltage-sensing of TRPV1 involves the proton activation machinery. (a) 447 Voltage activation kinetics of Kv 2.1 channels exhibit an initial delay between 448 depolarization (marked by an arrow in red) and the onset of macroscopic current. No 449 such a delay was observed in voltage activation of TRPV1. The activation time course 450 of Kv 2.1 is fitted to Equation (2) (based on SCHEME I in (c); also see Methods in 451 Supplementary materials), whereas a double-exponential function is used to fit that of 452 TRPV1 (solid curves in red). SCHEME I predicted the activation time course of 453 TRPV1 shown by the dashed trace in blue. (b) No Cole-Moore shift was observed in 454 voltage activation of TRPV1. No delay in current onset can be observed from 455 macroscopic currents recorded at +160 mV after a 2-s prepulse from 0 mV to -100 456 mV with 20-mV increments. (c) Gating SCHEME I and II for voltage activation of 457 Kv channels and TRPV1, respectively. (d) A representative recording of the voltage 458 dependence of TRPV1 in whole-cell configuration in the presence of 10 µM CAP. 459 Orange circles represent the steady-state current amplitude at the corresponding 460 voltages. Leak current was measured from the same patch in the presence of 461 ruthenium red (20 µM) to block TRPV1 (black circles). Current through TRPV1 at - 462 210 mV is indicated by a double-arrow in red. A dashed line in black indicates the 463 zero-current level. (e) Voltage dependence of the normalized conductance calculated 464 from the recording in (d). A dashed line in red indicates a steady-state level was 465 reached at deeply hyperpolarized voltages beyond -160 mV. (f) A representative 466 recording of the voltage dependence of TRPV1 in whole-cell configuration in the 467 presence of extracellular pH 4.0. (g) Voltage dependence of the normalized 468 conductance calculated from the recording in (f). A dashed line in red indicates the 469 conductance level kept decreasing at deeper hyperpolarization. (h) The semi- 470 logarithmic plot for the voltage dependence of normalized mean conductance 471 measured from steady-state current in whole-cell recordings at extracellular pH 4.0 472 and pH 7.2, respectively. (i) The fits to the derivative of the mean log open probability 473 at extracellular pH 4.0 and pH 7.2, respectively. (j) The fits to the derivative of the 474 mean log open probability with and without 3 µM capsaicin, respectively. (k) A 475 gating scheme where the voltage- and H+-dependent transition couples allosterically 476 to channel opening.
477 Figure 4. Structural mechanism underlying voltage sensitivity in TRPV1. (a) The pore 478 region of TRPV1 modeled under acidic (pH 4.0) and neutral conditions based on the 479 cryo-EM structure in the closed state (PDB ID: 3J5P). For the model under acidic pH, 480 its selectivity filter and the linker to S6 are colored in red. The turrets under acidic and 481 neutral pH are shown in cyan and dim gray, respectively. Atoms of C622 on the turret 482 are shown as spheres in yellow. An arrow in orange indicates the direction of C622 483 movement upon acidification. (b) Proton-induced conformational changes at the 484 turrets measured by whole-cell patch fluorometry from FM and TMRM attached to 485 C622 sites. 1, 2 and 3 indicate the emission spectra and their corresponding current
11 bioRxiv preprint doi: https://doi.org/10.1101/2020.01.04.894980; this version posted January 6, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
486 recordings at neutral pH, perfusion of a pH 5.0 solution and wash off, respectively. 487 The emission spectra were normalized to the emission peak of FM. Inset shows a 488 representative spectral image of fluorescence emission. (c) Similar measurements as 489 (b) at the N467C site in the S1-S2 linker as a negative control. (d) Percentage change 490 in the TMRM/FM intensity ratio measured from fluorophores labeled at C622, N467C 491 and in blank cells expressing no TRPV1 channel, respectively. *, P < 0.05. n = 3-to-8. 492 All statistics are given as mean ± s.e.m. (e) Left, a schematic diagram illustrating 493 depolarization-induced collective movements of charged residues and dipoles in the 494 outer pore region. Right, comparison of TRPV1 structure in the closed state (grey) 495 and the activated state model (blue). Helices are presented as pipes. Conformational 496 changes in charged residues and helixes underlie the unorthodox voltage-sensing 497 mechanism in TRPV1. (f) Locations of outer pore charged residues in the structurally 498 aligned resting and activated state models. The locations of negative charges are 499 approximated by the oxygen atoms in the sidechain (shown as a large sphere). The 500 dashed arrow in blue indicates the central axis through the pore (also defined as the z 501 axis of the models). (g) The surface electrostatic potential of the resting and activated 502 states. Positive and negative potentials (in kT/e) are colored in blue and red, 503 respectively.
12 bioRxiv preprint doi: https://doi.org/10.1101/2020.01.04.894980; this version posted January 6, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
504 REFERENCES
505
506 1 Cooper, G. M. The Cell: A Molecular Approach, 507
13 bioRxiv preprint doi: https://doi.org/10.1101/2020.01.04.894980; this version posted January 6, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
554 23 Horrigan, F. T. & Aldrich, R. W. Coupling between voltage sensor activation, Ca2+ 555 binding and channel opening in large conductance (BK) potassium channels. J Gen 556 Physiol 120, 267‐305 (2002). 557 24 Ma, Z., Lou, X. J. & Horrigan, F. T. Role of charged residues in the S1‐S4 voltage 558 sensor of BK channels. J Gen Physiol 127, 309‐328, doi:10.1085/jgp.200509421 559 (2006). 560 25 Long, S. B., Tao, X., Campbell, E. B. & MacKinnon, R. Atomic structure of a voltage‐ 561 dependent K+ channel in a lipid membrane‐like environment. Nature 450, 376‐382, 562 doi:10.1038/nature06265 (2007). 563 26 Cao, E., Liao, M., Cheng, Y. & Julius, D. TRPV1 structures in distinct conformations 564 reveal activation mechanisms. Nature 504, 113‐118, doi:10.1038/nature12823 565 (2013). 566 27 Liao, M., Cao, E., Julius, D. & Cheng, Y. Structure of the TRPV1 ion channel 567 determined by electron cryo‐microscopy. Nature 504, 107‐112, 568 doi:10.1038/nature12822 (2013). 569 28 Aggarwal, S. K. & MacKinnon, R. Contribution of the S4 segment to gating charge in 570 the Shaker K+ channel. Neuron 16, 1169‐1177 (1996). 571 29 Boukalova, S., Marsakova, L., Teisinger, J. & Vlachova, V. Conserved residues within 572 the putative S4‐S5 region serve distinct functions among thermosensitive vanilloid 573 transient receptor potential (TRPV) channels. J Biol Chem 285, 41455‐41462, 574 doi:10.1074/jbc.M110.145466 (2010). 575 30 Chang, H. K., Iwamoto, M., Oiki, S. & Shieh, R. C. Mechanism for attenuated outward 576 conductance induced by mutations in the cytoplasmic pore of Kir2.1 channels. Sci 577 Rep 5, 18404, doi:10.1038/srep18404 (2015). 578 31 Islas, L. D. & Sigworth, F. J. Voltage sensitivity and gating charge in Shaker and Shab 579 family potassium channels. J Gen Physiol 114, 723‐742 (1999). 580 32 Cole, K. S. & Moore, J. W. Potassium ion current in the squid giant axon: dynamic 581 characteristic. Biophys J 1, 1‐14 (1960). 582 33 Smith‐Maxwell, C. J., Ledwell, J. L. & Aldrich, R. W. Role of the S4 in cooperativity of 583 voltage‐dependent potassium channel activation. J Gen Physiol 111, 399‐420 (1998). 584 34 Gao, Y., Cao, E., Julius, D. & Cheng, Y. TRPV1 structures in nanodiscs reveal 585 mechanisms of ligand and lipid action. Nature 534, 347‐351, 586 doi:10.1038/nature17964 (2016). 587 35 Horrigan, F. T. & Aldrich, R. W. Allosteric voltage gating of potassium channels II. 588 Mslo channel gating charge movement in the absence of Ca(2+). J Gen Physiol 114, 589 305‐336 (1999). 590 36 Horrigan, F. T., Cui, J. & Aldrich, R. W. Allosteric voltage gating of potassium channels 591 I. Mslo ionic currents in the absence of Ca(2+). J Gen Physiol 1147, 27 ‐304 (1999). 592 37 Cao, X., Ma, L., Yang, F., Wang, K. & Zheng, J. Divalent cations potentiate TRPV1 593 channel by lowering the heat activation threshold. J Gen Physiol 143, 75‐90, 594 doi:10.1085/jgp.201311025 (2014). 595 38 Jara‐Oseguera, A. & Islas, L. D. The role of allosteric coupling on thermal activation of 596 thermo‐TRP channels. Biophys J 104, 2160‐2169, doi:10.1016/j.bpj.2013.03.055 597 (2013). 598 39 Latorre, R., Brauchi, S., Orta, G., Zaelzer, C. & Vargas, G. ThermoTRP channels as 599 modular proteins with allosteric gating. Cell calcium 42, 427‐438, 600 doi:10.1016/j.ceca.2007.04.004 (2007). 601 40 Matta, J. A. & Ahern, G. P. Voltage is a partial activator of rat thermosensitive TRP 602 channels. The Journal of physiology 585, 469‐482, doi:10.1113/jphysiol.2007.144287 603 (2007).
14 bioRxiv preprint doi: https://doi.org/10.1101/2020.01.04.894980; this version posted January 6, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
604 41 Yao, J., Liu, B. & Qin, F. Kinetic and energetic analysis of thermally activated TRPV1 605 channels. Biophys J 99, 1743‐1753, doi:10.1016/j.bpj.2010.07.022 (2010). 606 42 Yang, F. et al. Structural mechanism underlying capsaicin binding and activation of 607 the TRPV1 ion channel. Nature chemical biology 11, 518‐524, 608 doi:10.1038/nchembio.1835 (2015). 609 43 Yang, F. & Zheng, J. Understand spiciness: mechanism of TRPV1 channel activation 610 by capsaicin. Protein & cell 8, 169‐177, doi:10.1007/s13238‐016‐0353‐7 (2017). 611 44 Yang, F. et al. The conformational wave in capsaicin activation of transient receptor 612 potential vanilloid 1 ion channel. Nature communications 9, 2879, 613 doi:10.1038/s41467‐018‐05339‐6 (2018). 614 45 Yang, F. & Zheng, J. High temperature sensitivity is intrinsic to voltage‐gated 615 potassium channels. eLife 3, e03255, doi:10.7554/eLife.03255 (2014). 616 46 Lee, B. H. & Zheng, J. Proton block of proton‐activated TRPV1 current. J Gen Physiol 617 146, 147‐159, doi:10.1085/jgp.201511386 (2015). 618 47 Jordt, S. E., Tominaga, M. & Julius, D. Acid potentiation of the capsaicin receptor 619 determined by a key extracellular site. Proc Natl Acad Sci U S A 97, 8134‐8139, 620 doi:10.1073/pnas.100129497 (2000). 621 48 Zheng, J. & Ma, L. Structure and function of the thermoTRP channel pore. Current 622 topics in membranes 74, 233‐257, doi:10.1016/B978‐0‐12‐800181‐3.00009‐9 (2014). 623 49 Baez‐Nieto, D., Alvarez, O., Brauchi, S., Gonzalez, C. & Latorre, R. Voltage‐ 624 Dependence in Thermo‐Voltage Sensitive Channel TRPV1. A Delocalized Voltage 625 Sensor? Biophysical Journal 104, 456a, doi:10.1016/j.bpj.2012.11.2525 (2013). 626 50 Yang, F., Cui, Y., Wang, K. & Zheng, J. Thermosensitive TRP channel pore turret is 627 part of the temperature activation pathway. Proc Natl Acad Sci U S A 107, 7083‐7088, 628 doi:10.1073/pnas.1000357107 (2010). 629 51 Zheng, J. & Zagotta, W. N. Gating rearrangements in cyclic nucleotide‐gated 630 channels revealed by patch‐clamp fluorometry. Neuron 28, 369‐374 (2000). 631 52 Yang, F., Ma, L., Cao, X., Wang, K. & Zheng, J. Divalent cations activate TRPV1 632 through promoting conformational change of the extracellular region. J Gen Physiol 633 143, 91‐103, doi:10.1085/jgp.201311024 (2014). 634 53 Dosey, T. L. et al. Structures of TRPV2 in distinct conformations provide insight into 635 role of the pore turret. Nat Struct Mol Biol 26, 40‐49, doi:10.1038/s41594‐018‐0168‐ 636 8 (2019). 637 54 Zheng, Q. et al. Enhanced excitability of small dorsal root ganglion neurons in rats 638 with bone cancer pain. Molecular pain 8, 24, doi:10.1186/1744‐8069‐8‐24 (2012). 639 55 Caires, R. et al. Hyaluronan modulates TRPV1 channel opening, reducing peripheral 640 nociceptor activity and pain. Nat Commun 6, 8095, doi:10.1038/ncomms9095 (2015). 641 56 Hodgkin, A. The optimum density of sodium channels in an unmyelinated nerve. 642 Philos Trans R Soc Lond B Biol Sci 270, 297‐300 (1975). 643 57 Ledwell, J. L. & Aldrich, R. W. Mutations in the S4 region isolate the final voltage‐ 644 dependent cooperative step in potassium channel activation. J Gen Physiol 113, 389‐ 645 414 (1999). 646 58 Li, M. et al. Structure of a eukaryotic cyclic‐nucleotide‐gated channel. Nature 542, 647 60‐65, doi:10.1038/nature20819 (2017). 648 59 Martinez‐Francois, J. R., Xu, Y. & Lu, Z. Mutations reveal voltage gating of CNGA1 649 channels in saturating cGMP. J Gen Physiol 134, 151‐164, 650 doi:10.1085/jgp.200910240 (2009). 651 60 Marchesi, A., Mazzolini, M. & Torre, V. Gating of cyclic nucleotide‐gated channels is 652 voltage dependent. Nat Commun 3, 973, doi:10.1038/ncomms1972 (2012).
15 bioRxiv preprint doi: https://doi.org/10.1101/2020.01.04.894980; this version posted January 6, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
653 61 Donate‐Macian, P. & Peralvarez‐Marin, A. Dissecting domain‐specific evolutionary 654 pressure profiles of transient receptor potential vanilloid subfamily members 1 to 4. 655 PLoS One 9, e110715, doi:10.1371/journal.pone.0110715 (2014).
656
16
bioRxiv preprint (which wasnotcertifiedbypeerreview)istheauthor/funder.Allrightsreserved.Noreuseallowedwithoutpermission. a doi:
b c https://doi.org/10.1101/2020.01.04.894980 22°C 46°C 1.0 Control Heat CAP Proton Heat 40°C 0.8 +80 mV CAP 3 M 0.6 pH 5.0 1 nA
400 pA 200 pA 250 pA 0.4 20 s 20 ms 50 ms 50 ms 0.2
‐80 mV Normalized conductance 0.0 -200 -100 0 100 200 ; Voltage (mV) this versionpostedJanuary6,2020. d e f 1 1.0
0.8
0.1 0.6
q (e) 0.4 0.01
0.2 The copyrightholderforthispreprint ?
Normalized conductance 0.001 0.0 -100 0 100 200 -200 -100 0 100 200 Voltage (mV) Voltage (mV)
Figure 1 a b c WT bioRxiv preprint (which wasnotcertifiedbypeerreview)istheauthor/funder.Allrightsreserved.Noreuseallowedwithoutpermission. doi: https://doi.org/10.1101/2020.01.04.894980
d 1.0 e 1.0 WT WT R558L R558L f 0.8 0.8 0.6 0.6
0.4 q (e) 0.4 ; 0.2 0.2 this versionpostedJanuary6,2020.
Normalized conductance 0.0 0.0 -100 0 100 200 -200 -100 0 100 200 Voltage (mV) Voltage (mV) g h i O 200 current whole-cell Normalized 1.0 + 150 16 1.0 Na 100 +160 mV 50 14 2+ Count 0 0.8 Mg C 0.8 The copyrightholderforthispreprint 0 10 20 12 10 pA 0.6 Current (pA) 10 0.6
10 ms 60 8 q (e) 0.4 40 0.4 +80 mV 20 6 0 4 0.2 0 10 20 0.2 60 2 0.0 40 +40 mV 20 (pA) current Single-channel 0 0.0 -200 -100 0 100 200 0 Voltage (mV) 0 50 100 150 200 0 10 20 Voltage (mV) Figure 2 a b c SCHEME I bioRxiv preprint Kv 2.1 TRPV1 TRPV1 +40 mV +200 mV C1 C2 Cn O 1 nA (which wasnotcertifiedbypeerreview)istheauthor/funder.Allrightsreserved.Noreuseallowedwithoutpermission.
200 pA 20 ms SCHEME II doi:
10 ms C https://doi.org/10.1101/2020.01.04.894980 +160 mV L +120 mV J J -10 mV X Y 0 mV O R U -100 mV X Y A L d f h i +CAP pHext 4.0 1.0 1 pHext 4.0 pHext 7.2 5 nA 0.8 +RR ; 2 nA 0.1 0.6 this versionpostedJanuary6,2020. pHext 7.2 pHext 4.0
q (e) 0.4 0.01 0.2
0.001 0.0 Normalized conductance -400 -200 0 200 -210 mV 0 +100 -210 mV 0 +100 -200 -100 0 100 200 Voltage (mV) Voltage (mV) e g j k SCHEME III 1 1 1.0 pHext 7.2 pHext 7.2 6 The copyrightholderforthispreprint 3M CAP + 4 4 0.8 H H+ 2 2 C R N 0.6 J JpH 0.1 0.1 L v ΔV D F
6 q (e) 0.4 4 4 O A P 0.2 2 2 Activation Voltage Proton 0.01 0.01 0.0 sensor sensor Normalized conductance -200 -100 0 100 -200 -100 0 100 -400 -200 0 200 Voltage (mV) Voltage (mV) Voltage (mV) Figure 3 e a bioRxiv preprint C622 ‐ ‐ ‐ ‐
+++ ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ (which wasnotcertifiedbypeerreview)istheauthor/funder.Allrightsreserved.Noreuseallowedwithoutpermission. ΔV ‐ ‐
Acidic doi:
Neutral https://doi.org/10.1101/2020.01.04.894980 ‐‐‐ Closed Open
activated b f resting C622 2: pH 5.0 ;
3 this versionpostedJanuary6,2020. 1 5 nA FM TMRM 100 ms +80 mV 0 mV 500 520 540 560 580 600 -80 mV Wavelength (nm) c d 70 * g resting activated N467C 60 The copyrightholderforthispreprint 50 40 30 20 10
Change in Change TMRM/FM ratio (%) 0 500 520 540 560 580 600 C622 N467C Blank Wavelength (nm) Figure 4 bioRxiv preprint (which wasnotcertifiedbypeerreview)istheauthor/funder.Allrightsreserved.Noreuseallowedwithoutpermission. doi: https://doi.org/10.1101/2020.01.04.894980 a b +200 mV -30 mV 3.0 -40 mV 2.5 2.0 -50 mV 1.5 NPo 100 pA 1.0 -60 mV
0.5 ; 50 ms this versionpostedJanuary6,2020. -70 mV 0.0 -80 -60 -40 -20 Voltage (mV) 10 pA 20 ms The copyrightholderforthispreprint
Extended Data Figure 1 bioRxiv preprint (which wasnotcertifiedbypeerreview)istheauthor/funder.Allrightsreserved.Noreuseallowedwithoutpermission. doi: https://doi.org/10.1101/2020.01.04.894980 a b 1.2 Theoretical gating charge: d 12 Theoretical gating charge: C R 1.0 1.0 e0 10 10.0 e0 Jv L ΔV D 0.8 8 O A 0.6 6 0.4 4 0.2 Gating charge (e) charge Gating 2
0.0 0 f -300 -200 -100 0 100 200 10 -100 -50 0 50 100 ; Voltage (mV) this versionpostedJanuary6,2020. 8 c 3.5 Theoretical gating charge: e 14 Theoretical gating charge: 3.0 3.0 e0 12 13.0 e0 6 2.5 10 4 2.0 8 1.5 6 2 1.0 4 0 0.5 2 Deviation in estimation (%) The copyrightholderforthispreprint 0 4 8 12 0.0 0 Theorectical gating charge value (e0) -300 -200 -100 0 100 200 -100 -50 0 50 100
Extended Data Figure 2 bioRxiv preprint (which wasnotcertifiedbypeerreview)istheauthor/funder.Allrightsreserved.Noreuseallowedwithoutpermission. a b c doi:
Time 0 sTime 185 sOverlay https://doi.org/10.1101/2020.01.04.894980
100 pA
50 ms ; this versionpostedJanuary6,2020.
+220 mV The copyrightholderforthispreprint
0 mV ‐140 mV
Extended Data Figure 3 bioRxiv preprint (which wasnotcertifiedbypeerreview)istheauthor/funder.Allrightsreserved.Noreuseallowedwithoutpermission. doi: https://doi.org/10.1101/2020.01.04.894980
100
80
60 ; this versionpostedJanuary6,2020. 40
20
Inhibition of conductance (%) 0 -300 -200 -100 0 100 200 The copyrightholderforthispreprint Voltage (mV)
Extended Data Figure 4 bioRxiv preprint (which wasnotcertifiedbypeerreview)istheauthor/funder.Allrightsreserved.Noreuseallowedwithoutpermission. a c doi: 1.0 WT https://doi.org/10.1101/2020.01.04.894980 E649 E601Q_E649Q 0.8 E601 D E D 0.6 E E E D K 0.4 S5 S6 0.2
Normalized conductance 0.0 -100 0 100 200 Voltage (mV)
b 1.6 ; this versionpostedJanuary6,2020.
1.2
0.8
0.4 Total gating charge 0.0 The copyrightholderforthispreprint T Q C Q N Q Q N Q N E 1 2 7 7 9 2 5 9 7 0 W 0 0 3 4 4 5 5 4 4 4 6 6 6 6 6 6 6 6 6 6 E D E D E E D E D K _ _ _ Q Q K 1 7 7 0 3 3 6 6 6 E E E
Extended Data Figure 5 bioRxiv preprint (which wasnotcertifiedbypeerreview)istheauthor/funder.Allrightsreserved.Noreuseallowedwithoutpermission. a b doi: https://doi.org/10.1101/2020.01.04.894980 C 1 7 6 L 5 4 [CAP] = 10 M J O J 3 v c CAP R U 2 pH = 4.0 Jcv X Y 0.1
7 ; L 6 this versionpostedJanuary6,2020. A 5 pH = 7.2 4 Open probability JpH 3 H+ 2 N F 0.01 -200 -100 0 100 200 P Voltage (mV) The copyrightholderforthispreprint
Extended Data Figure 6 bioRxiv preprint (which wasnotcertifiedbypeerreview)istheauthor/funder.Allrightsreserved.Noreuseallowedwithoutpermission. doi: https://doi.org/10.1101/2020.01.04.894980 a b Neutral c Acidic
-1500
-1600 CCD KIC1 KIC2 -1700 KIC3 KIC4 KIC5 ;
-1800 KIC6 this versionpostedJanuary6,2020. KIC7 Score (R.E.U.) Score KIC8 -1900 KIC9
-2000
0 2 4 6 8 10 RMS (Å) The copyrightholderforthispreprint
Extended Data Figure 7 bioRxiv preprint doi: https://doi.org/10.1101/2020.01.04.894980; this version posted January 6, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
1 An unorthodox mechanism underlying voltage sensitivity of TRPV1 ion channel
2 Fan Yang1¶,2, Lizhen Xu1, Bo Hyun Lee2, Xian Xiao2,3, Vladimir Yarov-Yarovoy2 and 3 Jie Zheng2¶
4 1Department of Biophysics and Kidney Disease Center, First Affiliated Hospital, 5 Institute of Neuroscience, NHC and CAMS Key Laboratory of Medical 6 Neurobiology, Zhejiang University School of Medicine, Hangzhou 310058, China. 7 2Department of Physiology and Membrane Biology, University of California, Davis, 8 California 95616, USA. 3 School of Life Sciences, Westlake University, Shilongshan 9 Road No. 18, Xihu District, Hangzhou 310064, Zhejiang Province, China.
10 ¶ Correspondence and requests for materials should be addressed to F.Y. (email: 11 [email protected]) or J.Z. (email: [email protected])
12
13 MATERIALS AND METHODS
14 Molecular biology and cell transfection. cDNA of Murine TRPV1 (a gift 15 from Dr. Michael X. Zhu, University of Texas Health Science Center at 16 Houston) and Kv2.1 (a gift from Dr. Jon Sack, University of California, Davis) 17 were used in this study. To facilitate identification of channel-expressing cells, 18 eYFP and GFP were fused to the C-terminus of TRPV1 and the N-terminus of 19 Kv2.1, respectively. Tagging of these fluorescence proteins did not alter 20 channel functions, as previously reported1,2. QuickChange II mutagenesis kit 21 (Agilent Technologies) was used to generate point mutations. All mutations 22 were confirmed by sequencing.
23 HEK293T cells, purchased from and authenticated by American Type Culture 24 Collection (ATCC), were cultured at 37°C with 5% CO2 in a Dulbecco's 25 modified eagle medium with 10% fetal bovine serum and 20 mM L-glutamine. 26 Transient transfection with cDNA constructs was done with Lipofectamine 27 2000 (Life technologies) following manufacturer’s protocol. Patch-clamp 28 recordings were performed 1-2 days after transfection.
29 Chemicals. All chemicals such as capsaicin were purchased from Sigma- 30 Aldrich.
31 Electrophysiology. Patch-clamp recordings were performed with a HEKA 32 EPC10 amplifier with PatchMaster software (HEKA) in inside-out or whole-cell 33 configuration. Patch pipettes were prepared from borosilicate glass and fire- 34 polished to resistance of ~4 MΩ. For whole-cell recording, serial resistance 35 was compensated by 60%. A normal solution containing 130 mM NaCl, 10 36 mM glucose, 0.2 mM EDTA and 3 mM HEPES (pH 7.2) was used in both bath 37 and pipette. For solutions with 6.0 < pH < 7.2, HEPES was used as the pH 38 buffer. For solutions with pH ≤ 6.0, HEPES was replaced by 3 mM 2-(N- 39 morpholino)ethanesulfonic acid (MES). When intracellular Na+ ions were 40 replaced by Mg2+ ions, there was no Mg2+ in the extracellular solution. To 41 determine a G-V curve, the membrane potential was first clamped at -80 mV 42 for 10 ms and then switched to another clamping voltage stepping from -210
1 bioRxiv preprint doi: https://doi.org/10.1101/2020.01.04.894980; this version posted January 6, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
43 mV to +210 mV with a 10-mV interval for 500 ms, which was then switched to 44 -80 mV. Current amplitude at the steady state during the last 100 ms of 45 voltage steps was averaged to construct the G-V curve. Current signal was 46 filtered at 2.9 kHz and sampled at 10 kHz.
47 To apply solutions containing capsaicin or other reagents during patch-clamp 48 recording, a rapid solution changer with a gravity-driven perfusion system was 49 used (RSC-200, Bio-Logic). Each solution was delivered through a separate 50 tube so that there was no mixing of solutions. Pipette tip with a membrane 51 patch was placed directly in front of the perfusion outlet during recording. 52 Each membrane patch was recorded for only once.
53 Temperature control and monitoring. The bath solution was heated using 54 an SHM-828 eight-line heater driven by a CL-100 temperature controller 55 (Harvard Apparatus). We placed a TA-29 miniature bead thermistor (Harvard 56 Apparatus) about 1 mm from the pipette tip to monitor local temperature 57 change. Temperature readout from the thermistor was fed into an analog 58 input port of the HEKA patch-clamp amplifier and recorded simultaneously 59 with channel current. The speed of temperature change was set at a 60 moderate rate of about 0.3 °C/s to ensure that heat activation reached 61 equilibrium during the course of temperature change and the current was 62 recorded at steady state. When the experimental temperature was not 63 controlled, recordings were conducted at room temperature at 24°C. 64 Temperature variation was less than 1 °C as monitored by a thermometer.
65 Site-directed fluorescence recordings. Structural changes in the turret and 66 other extracellular regions were monitored with site-directed fluorescence 67 recordings as previously described 3,4. Briefly, fluorescein maleimide (FM) and 68 tetramethylrhodamine maleimide (TMRM) were used to irreversibly label a 69 native cysteine C622 in the turret region (after removing another native 70 cysteine by mutation C617A), or an introduced cysteine N467C in the S1-S2 71 linker region after removing both C617 and C622 with an alanine mutation. 72 Fluorescence resonance energy transfer (FRET) between FM and TMRM was 73 measured from voltage-clamped HEK293 cells imaged with an inverted 74 fluorescence microscope (Nikon TE2000-U) using a 40X oil-immersion 75 objective (NA 1.3). An argon laser (Spectra-Physics) was used to provide the 76 excitation light, with the exposure time controlled by a Uniblitz shutter 77 synchronized with the camera by the PatchMaster software through HEKA 78 amplifier. Spectral measurements were performed with an Acton SpectraPro 79 2150i spectrograph in conjunction with a Roper Cascade 128B CCD or an 80 Evolve 512 EMCCD camera. The rates of photobleaching FM and TMRM by 81 the excitation light were quantified separately from fluorophores attached to 82 the channel and used to correct for photobleaching during FRET experiments. 83 Proton-dependent changes in fluorescence intensity were also measured and 84 corrected. FRET was quantified from the enhancement of acceptor 85 fluorescence emission due to energy transfer1,5. FRET measurements were 86 done with cells that were patch-clamped throughout the recording. Proton- 87 induced current potentiation and FRET change were recorded simultaneously.
2 bioRxiv preprint doi: https://doi.org/10.1101/2020.01.04.894980; this version posted January 6, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
88 Molecular Modeling. To model the pore region of TRPV1 under neutral and 89 acidic extracellular pH, membrane-symmetry-loop modeling was performed 90 using the Rosetta molecular modeling suite6. First, to model the pore region 91 under neutral pH, coordinates of S1-S5, Pore helix, and S6, were taken from 92 TRPV1 channel structures (PDB ID: 3J5P) and kept rigid during loop modeling 93 and finally relaxed. The loop regions between S1 and S2, S2 and S3, S5 and 94 pore helix (turret), the selectivity filter and its linker to S6 were modeled de 95 novo in each subunit of the tetramer. During the first round of modeling, 96 Rosetta’s cyclic coordinate descent (CCD) loop relax protocol7 was used and 97 the top 20 cluster center models were passed to the second round. During the 98 second to third rounds of modeling, kinematic (KIC) loop relax protocol7,8 was 99 used and the top 20 cluster center models were passed to the next round. 100 During the rest rounds of modeling, kinematic loop relax protocol was used 101 and the top 20 models by score were passed to the next round. From 10,000 102 to 20,000 models were generated in each round. Models reported here 103 represent the lowest energy models from the last round of iterative loop relax. 104 The TRPV1 models we generated clearly entered an energy funnel through 105 the ten rounds of loop modeling (Extended Data Figure 6), suggesting that our 106 final model is located at the bottom of energy well. Top 10 models with lowest 107 energy after the last round of loop modeling exhibited good structural 108 convergence (Extended Data Figure 6).
109 To model the pore region under acidic pH, we followed the established 110 protocols in Rosetta9,10. Briefly, we modified the membrane energy function 111 terms11,12 as described in these reports. Then we added the following 112 command during the loop modeling: 113 -pH_mode true 114 We set the pH value to be 4 by adding this command: 115 -value_pH 4 116 Rounds of CCD and KIC loop modeling were performed with these 117 commands to obtain the model of TRPV1 pore region under acidic pH.
118 The quality of models under neutral and acidic pH were rigorously assessed 119 by four independent protein structure analysis and verification methods13-16. 120 The metrics of our models were similar to those derived from cryo-EM and 121 crystallography (Extended Data Table 1).
122 All the molecular graphics of TRPV1 models were rendered by UCSF 123 Chimera17 software version 1.11.
124
125 Data analysis.
126 To characterize the steady-state G-V curves, a single-Boltzmann function was 127 used:
128 = ∙ (1) /
3 bioRxiv preprint doi: https://doi.org/10.1101/2020.01.04.894980; this version posted January 6, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
129 where G/Gmax is the normalized conductance, V1/2 is the half-activation 130 voltage, qapp is the apparent gating charge and F is Faraday’s constant, R is 131 gas constant, and T is temperature in Kelvin.
132 To describe the voltage activation kinetics of Kv 2.1 channels, the following 133 equation was used:
134 = ∙ 1 − ( )∙ (2)
135 where α and β are microscopic forward and backward transition rates, 136 respectively, n is the number of independent transitions the channel must 137 transverse before opening. For Kv 2.1 channels we set n to be 4.
138 To estimate the total gating charge from the voltage dependence of channel 139 activation, we followed the method employed in the study of BK channels18. 140 Briefly, the mean activation charge displacement qa was first calculated by
( ) 141 = ∙ (3) 142 A smoothing function in the form of equation (1) was imposed on the voltage 143 dependence of Ln(Po), and then the total gating charge was estimated by 144 fitting the foot of the qa-V curve with a function:
145 Qo = + ∙ B[V]
146 where B[V] is a Boltzmann function in the form of equation (1), q0 and q1 147 represent the amplitudes of the voltage-independent and voltage-dependent 148 components of Qo (total gating charge), respectively.
149 Open probabilities of TRPV1 channel lower than 1% were often difficult to 150 quantify reliably; for this reason, we analyzed voltage dependence at lower 151 open probabilities using NPo from single-channel recordings.
152 Multi-allosteric gating model. The approach we used for the study of 153 TRPV1 polymodal gating was inspired by the work on Ca2+-sensitive voltage- 154 gated BK channels by Aldrich, Horrigan, and Cui 18-20. Similar to these earlier 155 studies, our approach to a general gating model for TRPV1 assumed that the 156 closed-to-open transition, CO, and the transition induced by a given 157 activation stimulus, RA, are both reversible transitions that are coupled 158 allosterically. The details in our modeling have been described previously21.
159 Statistics. All experiments have been independently repeated for at least 160 three times. All statistical data are given as mean ± s.e.m.. Two-tailed 161 Student’s t-test was applied to examine the statistical significance. N.S. 162 indicates no significance. ***, p < 0.001.
4 bioRxiv preprint doi: https://doi.org/10.1101/2020.01.04.894980; this version posted January 6, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
163 EXTENDED DATA FIGURE LEGENDS
164 Extended Data Figure 1. Patch-clamp recording of TRPV1 at hyperpolarized 165 voltages. (a) A representative inside-out patch recording with voltage stepped 166 from -70 mV to +200 mV. Single-channel activities were clearly discernable at 167 hyperpolarized voltages. The open probability did not further decrease with 168 deeper hyperpolarization beyond -30 mV. (b) Voltage dependence of NPo 169 beyond -30 mV remained at a stable level (n = 4).
170 Extended Data Figure 2. Estimation of the deviation between the true value of 171 total gating charge and the measured values. (a) The simple allosteric voltage 172 gating scheme employed for the estimation. (b to e) The voltage dependence 173 of open probability was simulated with the gating scheme in (a) and the 174 theoretical total gating charge was set to 1.0 e0, 3.0 e0, 10.0 e0 and 13.0 e0, 175 respectively. The simulated voltage dependence of open probability (open 176 circles in black) was further analyzed as described in Method and in Figure 1 177 (f) to derive the voltage dependence of total gating charge (dotted curve in 178 red). (f) The measured total gating charge (the maximum value of the dotted 179 curve in red) was compared to the theoretical true values to calculate the 180 deviation in total gating charge estimation.
181 Extended Data Figure 3. The voltage activation of TRPV1 does not diminish 182 upon washing off intracellular factors. (a) A representative inside-out patch 183 recording of TRPV1 activated by depolarization during a voltage ramp. This 184 patch was under constant perfusion of bath solution on its exposed 185 intracellular side. (b) After constant wash-off for more than three minutes, the 186 patch exhibited a similar current response to the same voltage ramp as in (a). 187 (c) The overlay of voltage-activated current traces at time 0 s and 185 s 188 demonstrates that washing off intracellular factors did not alter voltage 189 sensitivity in TRPV1. All experiments were performed at 22°C. Only 190 transmembrane voltage was used to activate TRPV1.
191 Extended Data Figure 4. Voltage-dependence of proton block of TRPV1 192 current. This figure is adapted from our previous study22. Proton block is 193 calculated from the OFF response recorded in outside-out patches. A 194 Boltzmann function was superimposed (V1/2 = 26.6 mV, qapp = 0.48 e0, n = 5- 195 to-9). All statistics are given as mean ± s.e.m.
196 Extended Data Figure 5. Neutralizing individual charged residues in the outer 197 pore near transmembrane voltage field did not eliminate voltage activation in 198 TRPV1. (a) A topology plot of the pore region of TRPV1 showing the charged 199 residues near transmembrane voltage field. Two residues, E601 and E649, 200 known to be critical for proton activation are highlighted in red. (b) Total gating 201 charge measured from each single and double mutant. A dashed line in black 202 indicate the total gating charge of WT. All data are given as mean ± s.e.m. (n 203 = 3-to-4). (c) The E601Q_E649Q double mutant (green) exhibits a similar G-V 204 curve (V1/2 = 114.1 ± 10.9 mV, qapp = 0.90 ± 0.04 e0, n = 3) as WT (black).
5 bioRxiv preprint doi: https://doi.org/10.1101/2020.01.04.894980; this version posted January 6, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
205 Extended Data Figure 6. A model of multi-allosteric coupling satisfactorily 206 predicts the behavior of TRPV1 gating by voltage (black), capsaicin (blue) and 207 proton (red). In this model (a), the voltage-dependent transition R↔A is 208 assumed to be completely separated from the proton-driven transition N↔P, 209 as shown in Scheme III (Figure 3i), with the total gating charge movement 210 split evenly between these transitions. Detailed description of the multi- 211 allosteric coupling and determination of the open probability were given in our 212 publication Cao et al. (JGP, 2014). Values for the parameters in the model 213 that yielded the voltage-dependent Po shown in (b) were mostly identical to 214 those used in Cao et al. (JGP, 2014): L = 0.01, q = 0.95 e0, T = 293.15 K, JV = 215 7, V1/2 = 130.4 mV, JC = 150, JpH = 250, JCV = 25, [capsaicin] = 10 μM, E = 5 X 216 104, pH = 4, F = 4 X 103.
217 Extended Data Figure 7. Modeling of the outer pore region of TRPV1 under 218 neutral and acidic pH conditions. (a) Plot of the Rosetta full-atom score versus 219 de novo loop RMS for ten rounds of iterative loop modeling under neutral 220 condition. Low energy models have a more negative value of the full-atom 221 score. (b) Overlap of the top 10 models with the lowest energy, which are well 222 converged after iterative loop modeling under neutral condition. (c) Top 10 223 models with the lowest energy are well converged after iterative loop 224 modeling under acidic (pH 4.0) condition. The turret and the linker between 225 pore helix and S6 are colored in cyan and red, respectively.
226
6 bioRxiv preprint doi: https://doi.org/10.1101/2020.01.04.894980; this version posted January 6, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
227 REFERENCES
228 1 Cheng, W., Yang, F., Takanishi, C. L. & Zheng, J. Thermosensitive TRPV channel 229 subunits coassemble into heteromeric channels with intermediate conductance and 230 gating properties. J Gen Physiol 129, 191-207, doi:10.1085/jgp.200709731 (2007). 231 2 Antonucci, D. E., Lim, S. T., Vassanelli, S. & Trimmer, J. S. Dynamic localization and 232 clustering of dendritic Kv2.1 voltage-dependent potassium channels in developing 233 hippocampal neurons. Neuroscience 108, 69-81 (2001). 234 3 Yang, F., Cui, Y., Wang, K. & Zheng, J. Thermosensitive TRP channel pore turret is 235 part of the temperature activation pathway. Proc Natl Acad Sci U S A 107, 7083- 236 7088, doi:10.1073/pnas.1000357107 (2010). 237 4 Yang, F., Ma, L., Cao, X., Wang, K. & Zheng, J. Divalent cations activate TRPV1 238 through promoting conformational change of the extracellular region. J Gen Physiol 239 143, 91-103, doi:10.1085/jgp.201311024 (2014). 240 5 Zheng, J., Trudeau, M. C. & Zagotta, W. N. Rod cyclic nucleotide-gated channels have 241 a stoichiometry of three CNGA1 subunits and one CNGB1 subunit. Neuron 36, 891- 242 896 (2002). 243 6 Leaver-Fay, A. et al. ROSETTA3: an object-oriented software suite for the simulation 244 and design of macromolecules. Methods Enzymol 487, 545-574, doi:10.1016/B978- 245 0-12-381270-4.00019-6 (2011). 246 7 Wang, C., Bradley, P. & Baker, D. Protein-protein docking with backbone flexibility. J 247 Mol Biol 373, 503-519, doi:10.1016/j.jmb.2007.07.050 (2007). 248 8 Qian, B. et al. High-resolution structure prediction and the crystallographic phase 249 problem. Nature 450, 259-264, doi:10.1038/nature06249 (2007). 250 9 Kilambi, K. P., Reddy, K. & Gray, J. J. Protein-protein docking with dynamic residue 251 protonation states. PLoS Comput Biol 10, e1004018, 252 doi:10.1371/journal.pcbi.1004018 (2014). 253 10 Kilambi, K. P. et al. Extending RosettaDock with water, sugar, and pH for prediction 254 of complex structures and affinities for CAPRI rounds 20-27. Proteins 81, 2201-2209, 255 doi:10.1002/prot.24425 (2013). 256 11 Yarov-Yarovoy, V., Schonbrun, J. & Baker, D. Multipass membrane protein structure 257 prediction using Rosetta. Proteins 62, 1010-1025, doi:10.1002/prot.20817 (2006). 258 12 Barth, P., Wallner, B. & Baker, D. Prediction of membrane protein structures with 259 complex topologies using limited constraints. Proc Natl Acad Sci U S A 106, 1409- 260 1414, doi:10.1073/pnas.0808323106 (2009). 261 13 Bowie, J. U., Luthy, R. & Eisenberg, D. A method to identify protein sequences that 262 fold into a known three-dimensional structure. Science 253, 164-170 (1991). 263 14 Colovos, C. & Yeates, T. O. Verification of protein structures: patterns of nonbonded 264 atomic interactions. Protein Sci 2, 1511-1519, doi:10.1002/pro.5560020916 (1993). 265 15 Hooft, R. W., Vriend, G., Sander, C. & Abola, E. E. Errors in protein structures. Nature 266 381, 272, doi:10.1038/381272a0 (1996). 267 16 Pontius, J., Richelle, J. & Wodak, S. J. Deviations from standard atomic volumes as a 268 quality measure for protein crystal structures. J Mol Biol 264, 121-136, 269 doi:10.1006/jmbi.1996.0628 (1996). 270 17 Pettersen, E. F. et al. UCSF Chimera--a visualization system for exploratory research 271 and analysis. J Comput Chem 25, 1605-1612, doi:10.1002/jcc.20084 (2004). 272 18 Horrigan, F. T. & Aldrich, R. W. Coupling between voltage sensor activation, Ca2+ 273 binding and channel opening in large conductance (BK) potassium channels. J Gen 274 Physiol 120, 267-305 (2002). 275 19 Horrigan, F. T. & Aldrich, R. W. Allosteric voltage gating of potassium channels II. 276 Mslo channel gating charge movement in the absence of Ca(2+). J Gen Physiol 114, 277 305-336 (1999).
7 bioRxiv preprint doi: https://doi.org/10.1101/2020.01.04.894980; this version posted January 6, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
278 20 Horrigan, F. T., Cui, J. & Aldrich, R. W. Allosteric voltage gating of potassium channels 279 I. Mslo ionic currents in the absence of Ca(2+). J Gen Physiol 114, 277-304 (1999). 280 21 Cao, X., Ma, L., Yang, F., Wang, K. & Zheng, J. Divalent cations potentiate TRPV1 281 channel by lowering the heat activation threshold. J Gen Physiol 143, 75-90, 282 doi:10.1085/jgp.201311025 (2014). 283 22 Lee, B. H. & Zheng, J. Proton block of proton-activated TRPV1 current. J Gen Physiol 284 146, 147-159, doi:10.1085/jgp.201511386 (2015).
285
8