Two-stage electro–mechanical coupling of a KV channel in voltage-dependent activation

In voltage-gated potassium (KV) channels, the voltage-sensing domain (VSD) undergoes sequential activation from the resting state to the intermediate state and activated state to trigger pore opening via electro–mechanical (E–M) coupling. However, the spatial and temporal details underlying E–M coupling remain elusive. Here, utilizing KV7.1’s unique two open states, we report a two-stage E–M coupling mechanism in voltage-dependent gating of KV7.1 as triggered by VSD activations to the intermediate and then activated state. When the S4 segment transitions to the intermediate state, the hand-like C-terminus of the VSD-pore linker (S4-S5L) interacts with the pore in the same subunit. When S4 then proceeds to the fully-activated state, the elbow-like hinge between S4 and S4-S5L engages with the pore of the neighboring subunit to activate conductance. This two-stage hand-and-elbow gating mechanism elucidates distinct tissue-specific modulations, pharmacology, and disease pathogenesis of KV7.1, and likely applies to numerous domain-swapped KV channels.


Fig. 5e
In the scheme, the hand interaction between S4-S5L and the pore occurs not in the RC state but in the IO state. Also, it is described in the text as follows.
(lines 271-273) "The S4 (upper arm) movement into the intermediate state triggers the Cterminus of S4-S5L(hand) to grip the S6c of the same subunit to promote channel opening at the first conductive IO state." (lines 308-309) "the classic E-M coupling interactions at S4-S5L/S6c engage upon VSD movement into the intermediate state". However, the results of MD simulation in Fig. 5a clearly show that the interactions (e.g. between V254 and L347) occur equally well both in the RC state and IO state. I would like to request full explanation and revisions to clarify this apparent discrepancy.

Fig. 5b
I understand this is the IO state model. I would like to see interactions depicted on the structure in the AO state, as well as in the RC state. These interactions is expected to be seen also in the AO and RC states, judging from the data in Fig. 5a. 3. Fig.5 The results of MD analysis is thought to be strongly influenced by the structure model in each state. Thus, the validity of the structure model is very critical. By the detail of the structure model in the Methods section (lines 533-540), the AO and AC models look to be convincing, but I wonder if the IO and RC models are trustable enough. 4. Fig. 2a The kinetics of current activation of W248R is very slow, although it is thought to be in the IO state. Is there any reasonable explanation? 1. Pg. 6. Line 129. The authors identified some mutations featuring strong fluorescence signals and no detectable currents, which could suggest that those residues are important for the classic E-M coupling. But the data presented can't directly suggest any interactions between them. Thus, "these results suggest that in Kv7.1 the E-M coupling interactions are also necessary for pore opening" seems too strong. Could some of these mutations just be important for gate opening, not coupling per se (e.g. A341V, P343A)? 2. Pg. 4. Line 68. "Kv7.1 is the only…" this statement seems too strong. Some studies have shown that KCNQ1 shows one open state (Kass et al 2010, Kubo et al 2014). Also, the GV of Kv7.1 has two components or not? Could it be fitted by double Boltzmann function as FV does? Maybe show the two components in Fig 1? 3. Fig. 2a Show fluorescence from W248R+KCNE1 to show that the channels are expressing and that mutations did not prevent S4 movement, which are alternative explanations for no currents. Please also add the fluorescence in presence of KCNE1 for as many as possible of the mutations in Supplementary Fig 2 to show the same. Understandably, it might not work for all of them, but at least for a couple of them. 4. I have some trouble with the 5-state model. If a mutation prevents AO state, like W248R, then should not current go to 0 at very depolarized voltages (i.e. channel will be pushed into AC by the positive voltage)? Similarly, for a mutation like L251A, which also is supposed to prevent AO state, why are there still two components in the current time course? Previously, the two components were interpreted as IO and AO activation. 5. Fig. 5. The simulation and the experimental data are not always using the same interacting pairs (e.g. Fig. 5c). Please explain and make it clear to the reader when data and simulations match and when they don't match. 6. Fig 5c bottom Supplementary Fig.4 line 52. Analyses should be "analysis" 14. Supplementary Fig 5b and e. I would remove the column marked Assigned State. To me that is just confusing. 15. Supplementary Fig 5e. Shouldn't F232 move relative to F279 when S4 moves up? One would not expect these two residues to always be together in RC, IO, and AO? 16. Supplementary Fig. 6b. Please explain y axis scale? E.g., 3.5 means 10% conserved or 100% conserved? In addition, shouldn't GYG be the most conserved in a K channel alignment? 17. Supplementary Fig. 6c. Show the data with arrows better (e.g. Y axis log scale, or inset with finer scale). Reviewer #3: Remarks to the Author: In this paper by Hou et al., the authors proposes a two-stage ''hand-end-elbow" gating mechanism of Kv7.1 voltage-gated potassium channel. The authors used several experimental and simulation techniques to reach this conclusion. Overall, the manuscript is well written and worthy of publication pending minor to moderate revisions. I have several comments to further improve the manuscript: 1. The SCA analysis is carried out only on domain-swapped Kv sequences (for good reasons) and is clearly mentioned in the results and the methods sections. However, the abstract reads differently and sounds like this mechanism is applicable to any Kv channel (Page 2, Line 39 and 40). Would not it be better to modify "numerous Kv channels" to "numerous Kv channels with domain-swapped architecture"? We have plenty of evidence now that non-swapped topology channels are gating very differently and soluble/regulatory domains playing very significant role in the process, see for example: I do feel that emphasizing apparent differences between two topologies present in Kv family may help a lot with application domain of findings reported and will place it into a broader context.
The authors approached the EM coupling of KCNQ1, using various effective techniques such as electro-physiology and VCF with systematic mutagenesis, MD simulation and also MSA analysis. They observed the classic intra-subunit interaction between S4-S5L (hand) and S6c in both IO and AO state. They also newly identified another inter-subunit interaction of S4/S4-S5L joint (elbow) with S5/S6 on the neighboring subunit, only in the AO state. I judge the elucidation of dynamic molecular mechanisms of two-stage EM coupling is important, as it cannot be achieved solely by structure analysis. I also judge the statistical coupling analysis of other numerous domain-swapped Kv channels using MSA analysis elevated the scientific merit of the work. This paper is written well and will attract attention of wide-ranged readers.
We thank the reviewer for the positive and encouraging evaluation.
I have some specific comments which require attention.

Fig. 5e
In the scheme, the hand interaction between S4-S5L and the pore occurs not in the RC state but in the IO state. Also, it is described in the text as follows.
(lines 271-273) "The S4 (upper arm) movement into the intermediate state triggers the Cterminus of S4-S5L(hand) to grip the S6c of the same subunit to promote channel opening at the first conductive IO state." (lines 308-309) "the classic E-M coupling interactions at S4-S5L/S6c engage upon VSD movement into the intermediate state". However, the results of MD simulation in Fig. 5a clearly show that the interactions (e.g. between V254 and L347) occur equally well both in the RC state and IO state. I would like to request full explanation and revisions to clarify this apparent discrepancy.
We thank the reviewer for raising this good point. The results of MD simulation in Fig. 5a show that interactions involved in the classic E-M coupling "occur equally well" at different states during the channel gating including RC, AC, IO, and AO. These results suggest that even when the VSD is at the resting state: 1) these pairs of residues stay close to each other and the interactions may be already engaged; 2) these interactions may promote channel opening upon the VSD activation to intermediate state and remain effective when the VSD moves to activated state. We have now revised this important point to "the motion of the S4 promotes channel opening through the S4-S5L (hand) grip of the S6c of the same subunit at the first conductive IO state" (pg. 12) and "the classic E-M coupling interactions at S4-S5L/S6c promotes channel opening upon VSD movement into the intermediate state" (pg. 13), we have also revised the cartoon scheme model in Fig. 5e to show the "hand grip" at RC state.

Fig. 5b
I understand this is the IO state model. I would like to see interactions depicted on the structure in the AO state, as well as in the RC state. These interactions is expected to be seen also in the AO and RC states, judging from the data in Fig. 5a

rFig. 1 Possible intra-subunit interactions at the classical E-M coupling region in RC and AO states.
Representations of the IO and AO state models of a Kv7.1 subunit. Inset: enlarged structures of the boxed area. Blue, residues identified as important for E-M coupling in experiments (Fig. 1c-g); grey: V255 and T265, which were identified in MD simulations as important for E-M coupling, but for which mutations still show functional currents (Fig.  3c). Residue V255 was found to be important for the AO state E-M coupling (Fig. 3). Each pair of residues is represented by a black dashed line connecting their respective sidechains.
3. Fig.5 The results of MD analysis is thought to be strongly influenced by the structure model in each state. Thus, the validity of the structure model is very critical. By the detail of the structure model in the Methods section (lines 533-540), the AO and AC models look to be convincing, but I wonder if the IO and RC models are trustable enough.
The reviewer makes a good point. We have made different efforts to improve the accuracy of our models at all different states. 1) The VSD salt-bridges of these models ( Supplementary Fig.  5a Fig. 5 (panel g, h), and cited the Figure in the main text (page 33). 4. Fig. 2a The kinetics of current activation of W248R is very slow, although it is thought to be in the IO state. Is there any reasonable explanation?
This is a good observation by the reviewer. W248R demonstrated consistent read-out across all four distinct tests (KCNE1 co-expression; co-mutation with E1R/R4E, E1R/R2E, and F351A) to show that it selectively opens in the IO state. The mutation may change the conformation of the channel in a complex way that also slows down certain transitions during activation to the IO state, which we did not investigate in this study.

Fig. 2d
Current amplitude scale (and also time scale) is missing.

Corrected.
6. I understand this is not the major focus of this work, but I wonder if the mechanistic insight in the present work can explain the reason why IO does not happen in KCNQ1/KCNE1 complex.
This is an important question, but unfortunately, there is not sufficient structural or functional data to address this question at this time.

Reviewer #2 (Remarks to the Author):
The authors here propose a two-stage "hand-and-elbow" gating mechanism for KCNQ1 channel. In voltage-gated K channels, the channel opening has been proposed to be coupled to the voltage-sensing domain (VSD) movement via an electro-mechanical (E-M) coupling using the S4-S5 linker as a coupling domain. In KCNQ1 channels, the VSD activates in two steps: from a resting state to an intermediate-activated state and then to the fully-activated state. The authors first identified mutations that prevent both intermediate opening and fully activated opening. These residues were consistent with the classical E-M coupling. In an additional screen using ML277 to identify mutations that exclusively prevents the ML-277-induced increase in current, the authors proposed that they identified residues important for fully-activated opening. Using four different tests, all of these residues seem important for the fully-activated opening. These residues did not seem to fall into the classical E-M coupling area, but mapped onto the interface between subunits. They further tested some of these pairs of interactions between the S4-S5 linker and the pore in neighboring subunits using Double Mutant Cycle analysis. These new activated-open E-M coupling interactions identified in the study is novel. The authors then performed MD simulation that support their idea that the classic interactions are important in the IO and AO E-M coupling while the newly-identified AO E-M coupling contribute to opening only in the fully-activated state. The authors concludes that when S4 moves to intermediate state, the hand-like C-terminus of S4-S5 linker interact with the pore in the same subunit and thereby promotes opening. This is consistent with the classic E-M coupling in many studies. When S4 then moves to fully-activated state, the elbow-like hinge between S4 and S4-S5 linker also interact with the neighboring pore domain to further activate the conductance. By using Statistical Coupling Analysis, they also suggest that the two-state "hand-and-elbow" model could be applicable to many Kv channels. This newly proposed "hand-and-elbow" mechanism helps to understand how VSD movement is coupled to channel opening and to explain how some mutations change channel function and thus lead to cardiac diseases such as Long QT syndromes. In addition, as different auxiliary subunits (eg. KCNE1 and KCNE3) modulate KCNQ1 gating by changing the E-M coupling in both VSD states, the mechanism provides new insights into the modulation of KCNQ1 by tissue-specific auxiliary subunits. The data are convincing, exhaustive, and robust, and the conclusion are appropriate. However, I would like some additional discussion and more detailed explanations of some figures (see below).
We thank the reviewer for the thorough and positive evaluation.
Major comments. 1. Pg. 6. Line 129. The authors identified some mutations featuring strong fluorescence signals and no detectable currents, which could suggest that those residues are important for the classic E-M coupling. But the data presented can't directly suggest any interactions between them. Thus, "these results suggest that in Kv7.1 the E-M coupling interactions are also necessary for pore opening" seems too strong. Could some of these mutations just be important for gate opening, not coupling per se (e.g. A341V, P343A)?
We appreciate that the reviewer raised a very important issue. We think that the residues in our study that their mutations get rid of the current but show strong fluorescence signal are important for E-M coupling. For residues (V254M and H258W) in the S4-S5L, since they are not in the pore and yet they affect the pore opening, they should be naturally thought to be important for the E-M coupling. On the other hand, residues (A341V, P343A, and G345A) in the S6 segment might be directly affecting pore opening, as the reviewer pointed out. However, the following evidence suggests that they also are important for the E-M coupling. 1) They are close to those residues important for the coupling in the S4-S5L, and MD simulation suggests that these residues interact (Fig. 5a). 2) These residues in Kv7. 1 Science, 2005]. To this point, it makes sense that a residue in the pore might be important for the E-M coupling to sense the signal from the VSD and also important for pore opening. Therefore, it is hard to define clearly whether such residues in the pore directly affect the pore opening or coupling. With these considerations, we modified the sentence to "these results suggest that in Kv7.1 the E-M coupling interactions may also be necessary for pore opening". 3. Fig. 2a Show fluorescence from W248R+KCNE1 to show that the channels are expressing and that mutations did not prevent S4 movement, which are alternative explanations for no currents. Please also add the fluorescence in presence of KCNE1 for as many as possible of the mutations in Supplementary Fig 2 to show the same. Understandably, it might not work for all of them, but at least for a couple of them.
We thank the reviewer for the constructive suggestion. Using   4. I have some trouble with the 5-state model. If a mutation prevents AO state, like W248R, then should not current go to 0 at very depolarized voltages (i.e. channel will be pushed into AC by the positive voltage)? Similarly, for a mutation like L251A, which also is supposed to prevent AO state, why are there still two components in the current time course? Previously, the two components were interpreted as IO and AO activation.  Fig. 5. The simulation and the experimental data are not always using the same interacting pairs (e.g. Fig. 5c). Please explain and make it clear to the reader when data and simulations match and when they don't match.

5.
For experimental data, using double mutant cycle in Fig. 4, we focused on determining intersubunit interactions specifically responsible for the AO state. All residue pairs were chosen from those 17 residues (13 residues from the ML277 effect screening and 4 S4c residues) that have been demonstrated as important for the AO state E-M coupling. Our MD simulation data match well with these experimental data. In addition to those experimentally demonstrated interactions, we also determined possible nearby interacting residue pairs that include at least one of the 17 residues. These interactions fall into three groups of state-dependent interactions as shown in Fig. 5c,d. For residues V241 and F275, which were identified in MD simulations as possibly involved in sidechain interactions but for which mutations did not alter AO states, we were using dark grey color to distinguish them. These similarities and differences were described in the legend of Fig. 5d.   6. Fig 5c bottom. For interacting pairs that are less prevalent in the AO state, shouldn't mutations make it easier to reach AO state (i.e. interactions hold the channel in RC or IO states and mutations should make these state less stable)?
For interacting pairs that are less prevalent in the AO state, these interactions are present in the RC and IO states, and may need to be broken when the VSD activates to the fully activated state to enable AO state E-M coupling. The interaction provides difference in free energy for the transition between AO and other states. When the interaction is altered by the mutations, the experimental results suggest that the free energy differences no longer favor the AO state. It is hard to predict whether a mutation should stabilize or destabilize the AO state simply based on the interaction at a single state. 7. Fig. 6. The SCA analysis cannot distinguish residue-residue interactions important for coupling compared to e.g. gate opening. Maybe point out some residue-residue interactions in Fig 5e that were found here to be important for coupling?
Reviewer 2 makes an excellent point. SCA detects amino acid co-evolution within a sector, which strongly suggests these residues form a functional unit within the protein. However, SCA does not specify the particular function conferred by each sector. We defined sector 2 in our study as an "E-M coupling" sector by its spatial pattern. It is possible that sector 2 also contains residues important in VSD activation and gate opening; however, the large number of residues in the S4-S5 linker strongly suggests that sector 2 includes significant E-M coupling residues. We did not expect to explicitly match SCA sector residues with KCNQ1 AO state E-M coupling residues. This is because SCA outputs protein sectors of the "average" domain-swapped KV channel derived from the sequence alignment. The positions within each sector thus represent important positions "on average". In this light, complete matching between SCA E-M coupling sector and KCNQ1 coupling residues would only be expected if KCNQ1 is an exact "average" KV channel -this is not a claim we make. Our interpretation of the SCA result is that the "average" domain-swapped K V channel features a group of co-evolving residues that are spatially reminiscent and compatible with the two-stage E-M coupling process we elucidated in KCNQ1, suggesting that our findings may broadly translate to other domain-swapped KV channels. To more directly address the reviewer's question, we indeed find that SCA identified interactions we found to be important for AO state EM coupling in KCNQ1. For example, W248-I268 and L251-I268 were found to interact during KCNQ1 activation by mutant cycle (Fig. 4c) and in SCA sector 2 (Fig. 6e). Of the 13 AO state E-M coupling residues identified by our ML277 screen, SCA placed 6 residues (W248, L251, V255, I268, S338, and L342) into sector 2 (E-M coupling sector), while the remaining 7 residues were not categorized into sector 1 or 2. Quantitatively, sector 2 (E-M coupling sector) contains 50 residues out of 200 residues analyzed. If SCA identified E-M coupling residues by random chance, we would expect a 50/200 or 25% positive rate. Reassuringly, SCA picked up 6/13 E-M coupling residues we identified in KCNQ1 for a 46% positive rate, a significant enrichment over random chance. Moreover, we interpreted sector 1 as residues important for independent stabilities/functions of VSD and Pore and not involved in E-M coupling. Sector 1 contains 42 residues, but none of the AO-state E-M coupling residues identified in our screen, which represent a significant negative enrichment over random chance. The combination of the negative/positive enrichment in the SCA sectors 1 and 2 when compared to our ML277 screen lends confidence that our interpretation of the SCA sectors.
We have modified the SCA paragraph in the discussion section to address the important point Reviewer 2 raised in that sector 2 may include residues not exclusive to E-M coupling. 3. Pg. 8. Line 170. Why is exactly the region of the classical E-M coupling? Conclusion that "W248 and S338 are located outside of the region of the classical E-M coupling" seems too strong. And are V255 and F256 not part of classical region, but V254 and H258 are?
This is a really good point. There can be some spatial overlap between the two sets of interactions. We qualified our sentence to "W248 and S338 are located generally outside of the region of the classical E-M coupling". 4. Pg. 2. Line 32. "Here, we leverage" this sentence is too complicated to understand for readers.
We have revised the sentence to make it easier to understand. 5. Pg. 2. Line 36. Please mention that the novel AO state E-M coupling interactions are from neighboring subunits. Otherwise, readers might get confused.
We agree with the reviewer and have changed it to "When S4 then proceeds on to the fullyactivated state, the elbow-like hinge between S4 and S4-S5L engages with the pore of the neighboring subunit to activate conductance." 6. Fig. 2a. Why so few traces in KCNQ1 W248R currents, when GV shows many points with currents?
Thanks for pointing this out. Due to some endogenous currents in the oocytes, W248R currents above 60 mV were not shown, however, these endogenous currents do not seem to alter the G-V relation. Nevertheless, in the revised manuscript, we deleted the data points above 60 mV in the G-V curve to match the data with current traces. The deletion of the points in G-V curve does not affect the results or conclusion. Fig 2e and 2h. The V50 of G-V is missing or too small in both figures.

7.
We assume the reviewer is pointing Fig 3e,g. The black circles are V50s of G-V, which almost superimpose with that of the F1-V (red circles). We have adjusted the figures to make it clearer. 8. Pg. 10. Line 232. How to get the activation energies of voltage-dependent action of channels should be explained.
We thank the reviewer for pointing this out. The activation energy was given by ∆G=-zFV50, where z is the effective charge, F is the Faraday constant, and V50 is the half activation voltage. Both z and V50 were obtained by fitting the G-V relation with Boltzmann equation. We have added more detailed explanation in the Methods (see pg. 27).

Corrected.
14. Supplementary Fig 5b and e. I would remove the column marked Assigned State. To me that is just confusing.

Done.
15. Supplementary Fig 5e. Shouldn't F232 move relative to F279 when S4 moves up? One would not expect these two residues to always be together in RC, IO, and AO?
We agree with the reviewer that F232 (on S4 segment) and F279 (on S5 segment) should have state-dependent interaction, which was nicely demonstrated by Nakajo and Kubo (Nakajo and Kubo, Nat Commun. 2014). In this study, we calculated the distance between Cβ atoms of F232-F279, and used a cut-off of 13 Å to determine if these residues are prone to interact when mutated into Cys. However, this calculation differs from the distance calculation shown in main Fig. 5. The results, 3 4 4 4 over 4 subunits for RC IO AO AC states, suggest the possible statedependent interactions at F232-F279. We realized that we did not specify the method of our calculation in the legend of Supplementary Fig 5e in the original manuscript, and this is a reason to cause confusion. We now added the method of calculation in sFig 5e legend in the revised manuscript.
16. Supplementary Fig. 6b. Please explain y axis scale? E.g., 3.5 means 10% conserved or 100% conserved? In addition, shouldn't GYG be the most conserved in a K channel alignment?
We apologize for the confusion in the y-axis of SFig. 6b, we will update the legend and methods to appropriately explain this axis. The y-axis in supplementary is a metric for first order conservation utilizing a statistical measure termed Kullback-Leibler divergence. The details are described in Rivoire et al. 2016. Briefly, K-L divergence quantifies the observed frequency of a sample against a "background" frequency. In pySCA, each amino acid is assigned a "natural" background frequency. SCA calculates the conservation value Di is then calculated with the equation

= ( )
Where fia is the observed calculated frequency in the input and qa is the background frequency for amino acid a. The value Di is 0 if the observed frequency f is equal to background frequency q, and increases as f deviates from q. Di does not measure how similar the amino acids are within the alignment, but instead measures deviation of the observed amino acid from expected background. High deviation from background is defined as "positional conservation". With respect to the signature sequence GYG, the input alignment shows extremely wellconserved GYG (Supp Fig. 6A). Only 42 out of 1421 total sequences (3%) does not feature "GYG" in the alignment. We included these sequences in the alignment based on their annotations in pfam as voltage-gated K+ channels, but they represent a very small set of the total sequences. We have clarified this in the Methods (page 36).
17. Supplementary Fig. 6c. Show the data with arrows better (e.g. Y axis log scale, or inset with finer scale).
We have revised Fig. 6c to show inset with finer scale. Thanks for the correction, we have revised the legends.

Reviewer #3 (Remarks to the Author):
In this paper by Hou et al., the authors proposes a two-stage ''hand-end-elbow" gating mechanism of Kv7.1 voltage-gated potassium channel. The authors used several experimental and simulation techniques to reach this conclusion. Overall, the manuscript is well written and worthy of publication pending minor to moderate revisions. I have several comments to further improve the manuscript: