Tracking a complete voltage-sensorcycle with metal-ion bridgesUlrike Henriona, Jakob Renhorna, Sara I. Börjessona, Erin M. Nelsona, Christine S. Schwaigerb, Pär Bjelkmarb,c,Björn Wallnerd, Erik Lindahlb,c,1, and Fredrik Elindera,1

aDivision of Cell Biology, Department of Clinical and Experimental Medicine, Linköping University, SE-581 85 Linköping, Sweden; bTheoretical andComputational Biophysics, Department of Theoretical Physics and Swedish e-Science Research Center, Royal Institute of Technology, SE-106 91 Stockholm,Sweden; cCenter for Biomembrane Research, Department of Biochemistry and Biophysics, Stockholm University, SE-106 91 Stockholm, Sweden; anddDepartment of Physics, Chemistry and Biology and Swedish e-Science Research Center, Linköping University, SE-581 83 Linköping, Sweden

Edited by* Michael L. Klein, Temple University, Philadelphia, PA, and approved March 20, 2012 (received for review October 13, 2011)

Voltage-gated ion channels open and close in response to changesin membrane potential, thereby enabling electrical signaling inexcitable cells. The voltage sensitivity is conferred through fourvoltage-sensor domains (VSDs) where positively charged residuesin the fourth transmembrane segment (S4) sense the potential.While an open state is known from the Kv1.2/2.1 X-ray structure,the conformational changes underlying voltage sensing have notbeen resolved. We present 20 additional interactions in one openand four different closed conformations based on metal-ionbridges between all four segments of the VSD in the voltage-gatedShaker K channel. A subset of the experimental constraints wasused to generate Rosetta models of the conformations that weresubjected tomolecular simulation and tested against the remainingconstraints. This achieves a detailed model of intermediate confor-mations during VSD gating. The results provide molecular insightinto the transition, suggesting that S4 slides at least 12 Å along itsaxis to open the channel with a 310 helix region present that movesin sequence in S4 in order to occupy the same position in spaceopposite F290 from open through the three first closed states.

electrophysiology ∣ inactivation ∣ Xenopus oocytes ∣ voltage clamp ∣conformational transition

Voltage-gated ion channels are critical for biological signaling,and they are able to regulate ion flux on a millisecond time

scale. To sense changes in membrane voltage, each ion channelis equipped with four voltage-sensor domains (VSDs) connectedto a central ion-conducting pore domain. The fourth trans-membrane segment (S4) of each VSD carries several positivelycharged amino-acid residues responsible for VSD gating (1). Atleast three elementary charges per VSD must traverse outwardsthrough the membrane electric field to open a channel that cor-responds to a considerable displacement of the S4 helix (Fig. 1A)(2–4). The positive charges in S4 make salt bridges with negativecountercharges on their move through the VSD (4–8). It has evenbeen proposed that the VSD undergoes a conformational altera-tion following the opening, when the channel relaxes to an inac-tivated, that is closed, state (9). In addition to conferring voltagedependence to ion channels, VSDs also regulate enzymes (10), actas voltage-gated proton channels (11, 12), are susceptible to dis-ease-causing mutations (13, 14), and serve as targets for drugs andtoxins (1, 15–18). Therefore, it is of crucial interest to understandthe details underlying voltage sensing by VSDs.

Few, if any, segments of membrane proteins have receivedmore attention than the S4 helix of voltage sensors. In additionto their paramount biological importance, they can help us under-stand fundamental biophysical problems such as why some mem-brane protein segments can be hydrophilic (19), how chargeseffectively move through a membrane, or how a potential triggersstructural changes on a microsecond time scale. These questionsare inherently linked to transient conformations and contacts thatcan be difficult to capture in a single structure. Although active/open-state conformations are known at atomic resolution for K

and Na channels (5, 20), the structure of the resting state has notyet been determined experimentally. Our knowledge about theVSD gating processes, including inactivation, is limited. Thereis strong experimental data suggesting each voltage sensor (inparticular the S4 helix) must adopt at least one intermediate statebetween the conducting and the resting state (3, 4, 21), whichmakes the pathway particularly interesting.

Structural constraints in a working ion channel can be exploredby introducing pairs of cysteines that, if close enough, can belinked by disulfide bonds (Fig. 1B, Upper) to alter the ion channelkinetics (7, 8, 22, 23). Disulfide bonds, however, have the disad-vantage that cysteines up to 15 Å apart can be caught in a bond(24). Moreover, due to their high strength they could capture thechannel in a nonnatural, or at least less common, conformation.In this study, we used weaker Cd2þ bridges that are more likelyto catch the channel in native conformations (25–28) (sulfur dis-tances <6.5 Å (29); Fig. 1B, Middle). In particular, their lowerstrength means the interactions can break and reform, whichmakes it possible to maintain the channel in a working state.Cd2þ bridges are normally formed between cysteines and histi-dines as well as between negatively charged glutamates andaspartates (29). We also explored Cd2þ bridges between one cy-steine and multiple glutamates or aspartates (Fig. 1B, Lower).

To search for possible interactions, we used the voltage-gatedShaker K channel expressed in oocytes from the frog Xenopuslaevis. Residues 325 and 326 in S3 are close to the gating chargesR1ð¼ R362Þ, R2ð¼ R365Þ, and R3ð¼ R368Þ in S4 in the openstate (23). Therefore, to probe for movements within the VSD,we first explored interactions between 325 or 326 and a longstretch of residues in S4 (355–369; green segment in Fig. 1C)using Cd2þ bridges. An interaction in a nonconducting (closed,but not necessarily resting) state is expected to shift the voltagedependence of channel opening in a positive direction along thevoltage axis and to slow down opening (although by itself it doesnot distinguish between different closed states). Similarly, an in-teraction in the open state is expected to shift the voltage depen-dence in a negative direction and to slow down closing (Fig. 1 Dand E). Of the 49 mutations that we investigated, 20 interactionswere found. This investigation follows three steps: (i) We startedby identifying several strong Cys-Cd2þ-Cys bridges, (ii) we usedthese interactions to build molecular models, and (iii) we used

Author contributions: U.H., J.R., S.I.B., E.M.N., C.S.S., P.B., B.W., E.L., and F.E. designedresearch; U.H., J.R., S.I.B., E.M.N., C.S.S., P.B., B.W., E.L., and F.E. performed research;U.H., J.R., S.I.B., E.M.N., C.S.S., P.B., B.W., E.L., and F.E. analyzed data; and U.H., J.R., S.I.B.,E.M.N., C.S.S., P.B., B.W., E.L., and F.E. wrote the paper.

The authors declare no conflict of interest.

*This Direct Submission article had a prearranged editor.

See Commentary on page 8362.1To whom correspondence should be addressed. Email: erik@kth.se or fredrik.elinder@liu.se.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1116938109/-/DCSupplemental.

8552–8557 ∣ PNAS ∣ May 29, 2012 ∣ vol. 109 ∣ no. 22 www.pnas.org/cgi/doi/10.1073/pnas.1116938109

Dow

nloa

ded

by g

uest

on

June

14,

202

0

several weaker Cys-Cd2þ-Glu∕Asp interactions to assess predic-tions generated by the models. These constraints and derivedmodels provided detailed information covering a complete cycleof conformations for a voltage-sensor.

ResultsResidues 325C and 326C in S3 Make Close Contact With a Long Stretchof Residues in S4. Ten μM Cd2þ quickly (within 10–20 s), and al-most completely, reduced the current of 325C∕358C at 100-mslong pulses to þ20 mV (Fig. 2 A and B, SI Appendix, Fig. S1Afor controls). Removal of Cd2þ did not restore the current within10 min (Fig. 2B); however, addition of the Cd2þ chelator EGTA

reversed the effect (SI Appendix, Fig. S1B). This suggests that atsome point during channel gating, 325C and 358C are in closeproximity, forming a metal-ion bridge between 325C and 358Cwith a low Kd value; 10 nM Cd2þ reduced the current by 36�12% (n ¼ 4). Seven other double mutants, though none of the sin-gle mutants, showed similar pronounced current reduction(Fig. 2C, SI Appendix, Table S1). Longer and higher voltage pulsesrestored the current amplitudes and broke the bridges even in thepresence of Cd2þ. Application of a 1-s long pulse to þ50 mV inCd2þ opened the channels, suggesting that these interactions oc-cur in closed states of the channel. As expected for a closed-stateinteraction, the voltage dependence for opening was shifted byþ55� 3 mV (n ¼ 3) (Fig. 3A), and the opening kinetics was slo-wed down by a factor of 490� 50 (n ¼ 3) (Fig. 3B).

326C∕365C and 325C∕364C Stabilize the Open State. In addition tothe clear current reductions shown in Fig. 2C, we found two dou-ble-cysteine mutations with Cd2þ bridges in the open state. Cd2þopened 326C∕365C by shifting the voltage dependence by −21�1 mV (n ¼ 12) (Fig. 3C), and it slowed down the closing bya factor of 31� 6 (n ¼ 4) (Fig. 3D). The interaction is fast(approximately 200 ms) in the open state, suggesting that326C and 365C are close to each other in the open-active state(SI Appendix, Fig. S2). Most likely, the reason for the fast reactionis that Cd2þ is already in place, being attracted by neighboringglutamates in S1 (E247) and S2 (E283) and coordinated by326C (SI Appendix, Fig. S3). In the open-state crystal structure,325 and 364 are very close to each other. The Cd2þ bridge for-mation, however, was slower for 325C∕364C than it was for326C∕365C, probably because of the lack of prebridge Cd2þ co-ordination. The time constant for the bridge formation in theopen state was 34 s, whereas no bridge formation occurred inthe inactivated state (SI Appendix, Fig. S4). This suggests that325 and 364 are close to each other in the open state and moveapart during slow inactivation. This is consistent with the ideathat S4 is relaxing following activation (9).

The Metal–Ion Bridges Suggest a Long Translational MovementBetween S3 and S4.The long stretch of residues in S4 we identifiedthat make close contact with residues 325 and 326 in S3 suggestthat S4 slides a considerable distance along S3 during gating(Fig. 3E). In addition to the open-state interactions, the restof the diagram in Fig. 2C shows three distinct regions with a per-iodicity of three residues, suggesting that there were three non-conducting conformations (C1–C3) that the VSD would pass onthe way from fully open to a down state. The diagram suggeststhat the 325C-361C interaction occurs in the last state beforeopening (C1). The sliding-helix model predicts that this interac-tion should be broken at more negative voltages. The making andthe breaking of the bridge were clearly voltage dependent, thussupporting the idea of C1 as a distinct preopen intermediateconformation (Fig. 3F, SI Appendix, Fig. S5). The interactionsproposed to occur in C2 and C3 were difficult to kinetically se-parate from each other, suggesting more flexible states furtheraway from the open state with C3 as the likely candidate forthe down state (because C1∕C2 would not provide sufficient gat-ing charge). Since some studies (28) have suggested there couldbe a deeper closed state (C4) reachable in some cases, we addi-tionally investigated whether 352C (one helical turn outside 355)interacts with 325C. Cd2þ on 325C∕352C reduced the current ;however, the interaction was not as strong as for the other resi-dues, and it was independent of state (SI Appendix, Fig. S6). Ourinterpretation is that this part is flexible enough in the Shakerchannel to form a bond at any voltage making it difficult to isolatea C4 state from contacts between these residues conclusively.Instead, we searched for interactions deeper in S3. In a potentialC4, 355 should interact with 322, or 358 with 318 or 319. 322C∕355C did not express very well, and was, thus, not possible to

A C B

D E

Voltage (mV)Op

en p

rob

abili

ty

0.0

0.5

1.0 Openstatebridge Closed

statebridge

Time (ms)Op

en p

rob

abili

ty

0.0

0.5

1.0

Openstatebridge

Closedstatebridge

Fig. 1. Probing voltage-sensor dynamics with Cd2þ bridges. (A) Multiplemolecular states of the S4 voltage sensor. O, open; I, inactivated; C, closedstates. (B) Three types of interactions. Covalent disulfide bonds (Upper), Cd2þ

bridges coordinating two cysteines (Middle), and Cd2þ bridges coordinatingone cysteine and two or several glutamates/aspartates (Lower). (C) Residues325 (magenta) and 326 (orange) are depicted as space filling. Residues 355–369 (green) are investigated with respect to interactions with 325 and 326.Blue sticks represent the gating charges. (D and E), Possible effects of Cd2þ

bridges (red) on voltage and time dependence of the open probability.

325C/358C

Time (ms)

Cu

rren

t (µ

A)

0 50 100

0.0

0.5

1.0

1.5

+Cd2+

A B 325C/358C

Time (min)

Cu

rren

t (µ

A)

0 5 10 150.0

0.5

1.0

1.5 Cd

C

Max

imu

mco

nd

uct

ance

in C

d(r

elat

ive

con

tro

l)

355C

356C

357C

358C

359C

360C

361C

+362

C36

3C36

4C

+365

C36

6C36

7C

+368

C36

9C0.0

0.5

1.0

1.5

326C

§

325C

#

C3 C2 C1 O

****

********

* *

##

Fig. 2. Cd2þ bridges between 325C or 326C in S3 and residues in S4. (A) Cd2þ

reduces the current amplitude of 325C∕358C at þ20 mV. Holding voltage−80 mV. (B) Bridge formation in 325C∕358C occurs quickly after applicationof Cd2þ. Washout is very slow. (C) Cd2þ effects on maximum conductances (SIAppendix, Table S1; n ¼ 2–12). §: Data for 357 double mutants were dividedby the effect on the single mutant 357C (SI Appendix, Table S1). #, ##: open-state interactions 325C-364C and 326C-365C. C3, C2, C1, and O denote sug-gested states for the interactions. Data shown as mean� SEM. Amplitudessignificantly different from 1.0 are denoted by * if p < 0.05 or ** if p < 0.01.

Henrion et al. PNAS ∣ May 29, 2012 ∣ vol. 109 ∣ no. 22 ∣ 8553

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

June

14,

202

0

investigate. 318C∕358C expressed well; however, the Cd2þ effectswere not different from those of the single mutant 358C. In con-trast, 319C∕358C was readily affected by Cd2þ. The current at theend of a 100-ms pulse to 0 mV was reduced by 58� 7% (n ¼ 5),whereas neither of the single mutants 319C or 358C were signifi-cantly reduced by the same protocol. Thus, we conclude that theShaker K channel goes through at least the C1-C2-C3 sequence ofconformations during closing, and, in some cases, it appears to bepossible to capture the VSD in a C4 conformation.

Atomic-Resolution Models of the Different States. For each state,models were built based solely on the experimental informationwithout any a priori assumption of the position of S4. Modelswere constructed by applying metal-ion bridge constraints in amultistage refinement process (see Methods) followed by ten dif-ferent 100 ns molecular dynamics simulations in a lipid bilayer foreach state (SI Appendix, Fig. S7) without any constraints applied.An additional set of simulations was performed with Cd2þ ionsincluded to assess the influence of the metal-ion bridges on thestructure. The distortion of the VSD from the ion was small (SIAppendix, Fig. S7); however the strong electrostatic interactionalso makes the metal-ion bridges unlikely to break (the S toCd2þ distance is typically 2.23� 0.05 Å for all states). For thisreason, we consider it a more stringent test whether the inter-mediate states simulated without ion bridges still fulfill the dis-tance restraints from our experiments. Because the cysteinesulfurs are free to rotate around the Cα-Cβ bond, the distancesbetween cysteine Cβ atoms were observed, which should fall in-side approximately 9 Å to be compatible with a bridge when theCβ-Sγ bond lengths are taken into account. All models fulfillthese metal-ion distance constraints during the last 50 ns ofunrestrained simulation (SI Appendix, Table S2; SI Appendix,Fig. S7). To predict specific conformations, the simulations re-sults were clustered in two steps, and the most populated clusterswere selected (Fig. 4, SI Appendix, Table S2). In particular theseparation between the C3∕C4 states is important. The mere factthe experimental constraints naturally produce different confor-mations does not, by itself, mean it is necessary to have differentstates. Neither the present C3 state nor the average conformationof the qualitatively similar recent consensus down state (30)appear to fulfill the constraints as well as C4. The shortestNδ∕ε-Nδ∕ε distance for residues A359H-I287H (Lin et al., me-tal-ion bridges (28)) is more than 10 Å in our C3 model and con-sensus model, while it is under 4 Å in C4. Similarly, the Cβ-Cβdistance for A319C-L358C constraint is 10.2 Å in C3 but 5.9 Å in

C4. The consensus model stops just before the residue corre-sponding to L358; however, if modeled, it would be directed awayfrom the side of the helix facing A319, though it is likely the mod-el would not distort a lot if this distance constraint was added.New omega current data from Lehmann et al. (31) suggestsR362 opposes E293 in the most closed state even though Plesset al. (32) argue E293 does not engage in state-dependent elec-trostatic interactions. For these residues the Cβ distance would be10.4 Å in C4, whereas it is 14 Å in C3 and 15.3 Å in the consensusmodel. For the two latter models the involved residues wouldeven be on opposite sides of F290 (making contact unlikely),while they are naturally adjacent for C4. There is presently noevidence, however, it is functionally necessary for VSDs to alwaysadopt this state in a normal VSD cycle because C3 would producesufficient gating charge, and VSDs without the final argininemight not even be able to reach C4. PDB structures of the modelsare available from the authors upon request and can be down-loaded directly at http://www.tcblab.org/henrion_vsd_models/.

The Shaker O state adopts 310-helix secondary structure forresidues 368–378 that are similar to the chimera template. Theinitial model of C1 only contains three residues of 310-helixaround R3ð¼ R368Þ; however, in the model from the moleculardynamics clustering, this expands to residues 367–377, i.e., down-stream of R3ð¼ R368Þ. For C2, the 310-helix region has movedand covers R2–R4ð¼ R365–R371Þ, and in C3, it moves anotherthree positions to R1–R3ð¼ R362–R368Þ. For all these confor-mations, the region of S4 below the 310-helix converts back intoan α-helix with the arginine above F290 forming salt bridges toE283, and the arginine below F290 forming salt bridges to E293,both in S2 (Fig. 4). The strain induced by these salt bridges ap-pears to be a main stabilizing factor for the 310-helix region.These structures suggest a sequence of events for the C1-C2-C3conformations, which would help explain the rapid transition(though, it should be noted that the experimental constraintsfor C1 and C2 are not very well separated). In contrast, forthe C4 model the final R1 side chain has moved below the hydro-phobic lock formed by F290, and without any salt bridge to E283,the structure relaxes into all α-helix in the model. This would sup-port a gating model where S4 predominantly moves along its axiswith a 310-helix region that slides along the S4 sequence at leastfrom the open state through C3 with one additional argininetranslating across the hydrophobic zone for each subsequent clos-ing step, making C3 the default resting state (supported by thegating charge). The larger structural transition required to movefrom C3 to C4 could indicate this state might be more difficult

326C/365C

500 600 700

-5

0

5

Time (ms)

Cu

rren

t (µµ

A)

A325C/358C

Voltage (mV)

Co

nd

uct

ance

(no

rmal

ized

)

-100 -50 0 50

0.0

0.5

1.0

325C/358C

Time (ms)

Cu

rren

t(n

orm

aliz

ed)

0 500 1000

0.0

0.5

1.0

C

B D F

326C/365C

Voltage (mV)

Co

nd

uct

ance

(µS

)

-50 0 50 100

0

50

100

325C/361C

Voltage (mV)

Rat

e (s

-1)

-110 -90 -700.0001

0.001

0.01

0.1

Bridge making

Bridge breaking

E

Fig. 3. State dependence of Cd2þ bridges. (A–B) Closed-state interaction in 325C∕358C. Cd2þ shifts the G(V) by þ55 mV (A) and slows down channel openingat þ50 mV by a factor of 490 (B). The maximum conductance was reduced by 54� 7% (n ¼ 3). Cd2þ effects in red and control in black. (C and D) Open-stateinteraction in 326C∕365C. Cd2þ shifts the G(V) by −21 mV (C), and slows down channel closure by a factor of 31 in a high Kþ solution at −60 mV following apulse to þ60 mV (D). Cd2þ effects in red and control in black. (E) Summary of the interactions. Residues in S3 interacting with 325C (Left) and 326C (Right)respectively are shown in cyan (closed states) and blue (open state) in the Kv1.2/2.1 chimera structure (6). (F) Voltage-dependence of the kinetics of Cd2þ bridgemaking and breaking for 325C∕361C. Data shown as mean� SEM (n ¼ 2–3).

8554 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1116938109 Henrion et al.

Dow

nloa

ded

by g

uest

on

June

14,

202

0

though sometimes possible, to reach. The structure of the upperpart of S4 in the C4 state is less certain because there are nostrong experimental constraints specifically favoring α-helix,and a helix moving from C3might be able to retain some 310-helixstructure. In Schwaiger et al. (33) the lower stability of 310-helix isreproduced by force fields; however, the sliding 310-helix wouldnot have any net cost between states O through C3.

Single Cysteines Coordinate Cd2þ Together With Endogenous Gluta-mates or Aspartates. Seven of the single-cysteine mutants were af-fected by 10 μM Cd2þ suggesting that they may coordinate Cd2þ

togetherwith endogenous residues.Basedon themolecularmodels(Fig. 4), we evaluatedpossible interactions. For three evenly spacedresidues (356C, 359C, and 362C),Cd2þ kept the channel in a closedstate. The opening kinetics was slowed down (Fig. 5A) and more

positive voltages were needed to open the channel (Fig. 5B). Allthree residues were close to two glutamates in S1 (E247) and S2(E283) in states C3 and C4 (Fig. 6A), C2 and C3, and C1 andC2 respectively. 365C is close to this glutamate pair in the O stateand, interestingly, Cd2þ opens this channel (Fig. 5B). The Cd2þeffects on 356C, 359C, and 362C were rescued by mutating E283or E247 to glutamines (Fig. 6 B–E, SI Appendix, Table S3; SIAppendix, Figs. S8 and S9), confirming that these residues are closeto E283 andE247 in three consecutive closed states. Formutations358C and 361C, Cd2þ opens the channel (Fig. 5B). Residues 358Cand 361C could possibly interact with a cluster of four negativelycharged residues at the top of S3 in the O state. The closeness isshown for 358C and the cluster (Fig. 6F). Neutralizing this clusterdeleted the Cd2þ effect (Fig. 6G andH; SI Appendix, Table S3; SIAppendix, Fig. S10) which supports the finding that 358 and 361 areclose to EEED333-336 in the O state.

DiscussionIn the present investigation we have described 20 additionalinteractions between transmembrane segments of the VSD (SIAppendix, Table S4). With molecular modeling we have builtmolecular models of several different intermediate VSD states.These five models are consistent with the interactions describedin the present investigation (SI Appendix, Table S4) and fulfill pre-viously published experimental data (SI Appendix, Table S5).The open-state model, O, with R1–R4ð¼ R362–R371Þ outsideF290, is close to the previously determined structure of a Kchannel (5). Both the C1 and C2 states are compatible with someprevious experimental data, and C3 is very similar to previousdown-state models (30, 34–36). The C4 state, with all positivecharges inside F290, has S4 even deeper intracellularly, but giventhe efforts required to capture this state and the fact that C3 issufficient to explain the gating charge, C4 might only be presentat significant hyperpolarization (which could remove it from thenormal VSD cycle). Another important finding is that there is alarge relative motion between S3 and S4. The C4 model is con-sistent with recent experimental data showing that not onlyR2–R4ð¼ R365–R361Þ, but also R1ð¼ R362Þ might be able topass inside F290 (28, 31, 37). The cascading motion of thearginines is similar to recent simulations (38); however, in the ex-perimentally derived models, this is achieved through a sliding310-helix present in all states from O through C3 in the VSD.

A

B

G(V

) sh

ift

(mV

)

356C

357C

358C

359C

360C

361C

+362

C36

3C36

4C+3

65C

366C

367C

+368

C36

9C

-20

-10

0

10

20 **

**

***

***

***

*

t 75,

Cd/t

75,c

on

tro

l

356C

357C

358C

359C

360C

361C

+362

C36

3C36

4C+3

65C

366C

367C

+368

C36

9C

0

2

4

6

8 *

**

*

*

Fig. 5. Effects of Cd2þ on single-cysteine mutants. Summary of Cd2þ effectson activation time constants (A) and G(V) shifts (B) of the single mutations.Data shown as mean� SEM (n ¼ 2–6). Amplitudes significantly differentfrom 1.0 (A) or 0 (B) are denoted by * if p < 0.05, ** if p < 0.01, or *** ifp < 0.001.

Fig. 4. Molecular models of VSD states and the gat-ing process. (A) Metal-ion constraints and molecularsimulation relaxation predicts each C1 through C3,and, under some conditions, C4, state to correspondto one additional arginine side chain in S4 (bluesticks) translating across the hydrophobic zone lockformed by F290 (green sticks), forming salt bridgesto negatively charged residues in S1–S3 (red sticks;E247 in S1and E283 in S2 above F290, and E293 in S2and 316D in S3 below F290. The region of S4 close toF290 adopts 310-helix (purple) in all models; how-ever, C4 with the rest of S4 in α-helix (yellow). Thissuggests a gating model where a virtually constant-length 310-helix region slides along the sequence ofS4 without any net free energy cost that avoidsoverall rotation of the entire S4 helix during gat-ing—only the outer ends rotate. (B) Cartoon high-lighting some of the features from A.

Henrion et al. PNAS ∣ May 29, 2012 ∣ vol. 109 ∣ no. 22 ∣ 8555

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

June

14,

202

0

In the C4 state, the salt bridge to E283 above F290 is lost, andconsequently S4 relaxes to α-helix in the model. We have alsoshown that the open-state model is transformed to a new config-uration during inactivation.

To conclude, our set of models covers a complete sequence oftransformations from the deepest closed state to an open state.This set strongly suggests that the VSD activation involves S4moving at least 12 Å relative to the rest of the VSD (transferringthree to four charges across the membrane) from the O to the C3state, and that it could be possible to reach an even more closedC4 state in some cases with larger structural changes (althoughthis would require a 17 Å motion). The motion can primarily bedefined as a sliding helix, and for O through C3 there is a con-stant-length 310-helix region moving along the sequence of S4 tomaintain its location in the hydrophobic region around F290 ofthe VSD. The structural difference as well as vertical translationis larger between C3 and C4 compared to the other states. It is animportant question whether C3 or C4 is more likely to correspondto the resting state. Considering the Cole-Moore effect wheresignificant hyperpolarization makes it harder to subsequentlyactivate the channel (39), this could be explained by C3 being thedefault resting state (supported by recent simulations of ωome-ga;-currents (40)), while C4 might only be reached after long hy-perpolarization. This would agree well with the model of a morenatural motion from O through C3, while C3 to C4 requires alarger change, and could help reconcile many of the different con-straints observed in experiments and models.

MethodsOocytes and Expression. All experiments were performed on the Shaker H4channel (Acc No NM_167595.3) (41) made incapable of fast inactivation bythe Δ(6–46) deletion (42). Xenopus laevis surgery, oocyte dissection, channelmutagenesis, and cRNA preparation and injection were performed aspreviously described (15). Injected oocytes were maintained at 11 °C in mod-ified Barth’s solution (15) containing pyruvate and antibiotics until subjectedto electrophysiological experiments. For double-cysteine mutants, 0.5 mMDTTwas added to the MBS solution to prevent disulfide-bond formation dur-ing incubation.

Electrophysiology. Currents were measured with the two-electrode, voltage-clamp technique (CA-1B amplifier, Dagan Corporation) 3–6 d after injection.The amplifier’s leak and capacitance compensation were used, and thecurrents were low-pass filtered at 5 kHz. All experiments were done in roomtemperature (20–23 °C). The control solution contained (in mM): 88 NaCl, 1KCl, 15 Hepes, 0.4 CaCl2, and 0.8 MgCl2. pH was adjusted to 7.4 with NaOH toyield a final sodium concentration of approximately 100 mM. Solutions of10 μM CdCl2 (unless stated otherwise) were prepared in control solution.To study channel voltage dependence and kinetics, steady-state currentswere achieved by stepping to voltages typically between −80 and þ50 mVin 5 mV increments from a holding voltage of −80 mV. For mutants with sub-stantially altered voltage dependence, the voltage range was shifted accord-ingly. Cd2þ was applied during 0.1 or 1 s pulses by stepping from a holdingvoltage of −80 to þ20 mV. For other pulse protocols, please refer to the de-tails in the corresponding figure legend. WT channels were essentially inertto 10 μMCd2þ (SI Appendix, Table S1). All chemicals were from Sigma-Aldrich,unless stated otherwise.

Analysis of Electrophysiological Data. All data analysis was carried out inClampfit (Version 10.2, MDS Analytical Technologies) or Graphpad Prism(Versions 4.03 and 5.04 for Windows, GraphPad Software, Inc.). The conduc-tance (G) was calculated from the steady-state currents (I) according to G ¼ I∕ðV − VeqÞ, where V is the membrane potential and Veq the reversal potentialfor Kþ (−80 mV). The Cd2þ-induced shifts were determined at 50% of themaximum conductance. The activation speed was assessed by determiningthe time needed to reach 75% of the maximum current amplitude. Forthe analysis of 325C∕358C in Fig. 3, we decomposed the G(V) curve in twoBoltzmann expressions: one following the control G(V); and the other as-sumed to be affected by Cd2þ. Similarly, we decomposed the activation timecourse in two independent time courses and measured the time constant;however, in SI Appendix, Table S1 we used the same analysis as for all othermutations.

Rosetta Modeling. A homology model of the voltage sensing domain ofShaker was built by aligning the sequence of Shaker to the sequence ofthe crystal structure of the Kv1.2-Kv2.1 paddle chimera (PDB: 2R9R) withHHalign (43) and then building the model with Modeller (44). The sequenceidentity is over 40%. This initial model was used as a starting point for mod-eling based on the experiments. A sequence of distinct constraints believedto be nonoverlapping in structure (SI Appendix, Table S4) were selected forthe models and represented as harmonic distance restraints centered at 6 Å

BA

D356C

-80 -40 0 40

0.0

0.5

1.0

Voltage (mV)

Co

nd

uct

ance

(no

rmal

ized

) 356C/283Q

-80 -40 0 40 80

0.0

0.5

1.0

Voltage (mV)

Co

nd

uct

ance

(no

rmal

ized

)

356C

0 100 200

0.0

0.5

1.0

Time (ms)

Cu

rren

t(n

orm

aliz

ed) 356C/247Q/283Q

0 100 200

0.0

0.5

1.0

Time (ms)

Cu

rren

t(n

orm

aliz

ed)

State C4

C

E

G358C

-80 -40 0

0.0

0.5

1.0

Voltage (mV)

Co

nd

uct

ance

(no

rmal

ized

) 358C/333-336QQQQ

-80 -40 0

0.0

0.5

1.0

Voltage (mV)

Co

nd

uct

ance

(no

rmal

ized

)

F

H

Open state

Fig. 6. Cd2þ is coordinated by one cysteine and two or several negatively charged residues. (A) Glutamates 247 and 283 are close to 356C in the C4-model.(B) Cd2þ slows the opening kinetics of 356C. (C) Neutralization of E247 and E283 abolish the Cd2þ effect. (D) Cd2þ shifts the G(V) of 356C with þ19.5 mV (Top).(E) Neutralization of E283 abolishes the Cd2þ effect. (F) A cluster of negatively charged residues in the S3-S4-linker (E/D333-336) is close to 358C in theO-model. (G) Cd2þ shifts the G(V) of 358C with −13.2 mV. (H) Neutralization of E/D333-336 abolishes the Cd2þ effect.

8556 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1116938109 Henrion et al.

Dow

nloa

ded

by g

uest

on

June

14,

202

0

between the cysteine sulfurs involved in Cd binding. The entire S3–S4 loop aswell as the S4 segment (residues 332–379) were rebuilt completely from frag-ments in Rosetta allowing the S4 helix to adopt different positions relative tothe membrane. The transmembrane backbone of S1–S3 was restrained, andonly their side chains were modified. To enhance sampling of realistic modelsduring the Rosetta step, a weak constraint was also used to keep the arginineresidues (R1–R4ð¼ R362–R371Þ) in S4 close to the negatively charged E247,E293, or D319. In each modeling stage 10,000 models were sampled. Theywere subsequently clustered using 2 Å rmsd between the membrane regions.The model from the largest cluster with the lowest energy was selected.

Molecular Dynamics. The simulation systems were generated by superimpos-ing each VSD model onto the structure of an already equilibrated VSD-mem-brane system (33) simulated in a hexagonal unit cell. Overlapping moleculeswithin 1 Å around the protein were deleted, such that, in general, 43 lipidsper monolayer and about 6,900 water molecules remained. The system wasneutralized by randomly replacing two water molecules with Kþ counterions. The final configuration comprises about 28,000 atoms. The VSD was de-scribed with the Amber99SB-ILDN force field (45). POPC interactions weredescribed with the Berger force field parameters (46) using united atomsin the tails to reduce the number of particles and water with the TIP3P model(47). The systems were energy minimized with steepest descent for 1,000steps. Subsequently, a short equilibration simulation of 100 ps was performedrestraining the water molecules along the z–axis and the protein backboneto its initial position. It was followed by yet another 100 ps of equilibrationwithout restraints on the water molecules. No restraints were applied tomaintain the experimental salt bridge distances in the model simulations;however, explicit ions were included in a similar set of control simulations.

All simulations were carried out with the Gromacs version 4.5.3 (48) pack-age using virtual interaction sites, 4 fs time steps, and all bond lengths con-strained. Electrostatic interactions were evaluated using particle mesh Ewaldsummation every step. A 10 Å cutoff was used for electrostatics and van derWaals interactions with neighbor lists updated every 10 steps. Simulationswhere performed at 300 K by using a Bussi thermostat (49). Semiisotropicpressure coupling was used with a Parrinello–Rahman barostat (50) with atime constant of 5 ps and a compressibility of 4.5 10−5 bar−1 in the planeof the membrane and along the membrane normal.

Ten 100 ns production runs with different random seeds for velocitieswere performed for each state without any restraints on the system. Rmsdclustering was performed with the Gromacs g_cluster program with a 0.7 Åcutoff using frames spaced 1 ns apart from the last 50 ns of each simulation,after which centroids from different simulations were clustered with a 2 Åcutoff. The centroid of the most populated cluster was selected as the finalmodel for each state.

ACKNOWLEDGMENTS.We thank IngridWalan andMattias Larsson for some ofthe experiments and Peter Larsson and Gunnar von Heijne for comments onthe manuscript. This work was supported by the Swedish Research Council,the Swedish Heart–Lung Foundation, the Swedish Brain Foundation, theCounty Council of Östergötland, Queen Silvia’s Anniversary Foundation, KingGustaf V and Queen Victoria’s Freemasons Foundation, Stina and BirgerJohansson’s Foundation, the Swedish Society for Medical Research, theSwedish Foundation for Strategic Research, and the European ResearchCouncil.

1. Borjesson SI, Elinder F (2008) Structure, function, and modification of the voltagesensor in voltage-gated ion channels. Cell Biochem Biophys 52:149–74.

2. Schoppa NE, et al. (1992) The size of gating charge in wild-type and mutant Shakerpotassium channels. Science 255:1712–1715.

3. Zagotta WN, Hoshi T, Aldrich RW (1994) Shaker potassium channel gating III:Evaluation of kinetic models for activation. J Gen Physiol 103:321–62.

4. Keynes RD, Elinder F (1998) Modelling the activation, opening, inactivation andreopening of the voltage-gated sodium channel. Proc Biol Sci 265:263–70.

5. Long SB, et al. (2007) Atomic structure of a voltage-dependent Kþ channel in a lipidmembrane-like environment. Nature 450:376–382.

6. Papazian DM, et al. (1995) Electrostatic interactions of S4 voltage sensor in ShakerKþ channel. Neuron 14:1293–301.

7. DeCaen PG, et al. (2008) Disulfide locking a sodium channel voltage sensor reveals ionpair formation during activation. Proc Natl Acad Sci USA 105:15142–15147.

8. DeCaen PG, et al. (2009) Sequential formation of ion pairs during activation of asodium channel voltage sensor. Proc Natl Acad Sci USA 106:22498–22503.

9. Villalba-Galea CA, et al. (2008) S4-based voltage sensors have three major conforma-tions. Proc Natl Acad Sci USA 105:17600–17607.

10. Murata Y, et al. (2005) Phosphoinositide phosphatase activity coupled to an intrinsicvoltage sensor. Nature 435:1239–1243.

11. Sasaki M, Takagi M, Okamura Y (2006) A voltage sensor-domain protein is a voltage-gated proton channel. Science 312:589–592.

12. Ramsey IS, et al. (2006) A voltage-gated proton-selective channel lacking the poredomain. Nature 440:1213–1216.

13. Ashcroft FM (1999) Ion Channels and Disease (Academic, San Diego), 1st Ed.14. Sokolov S, Scheuer T, Catterall WA (2007) Gating pore current in an inherited ion

channelopathy. Nature 446:76–78.15. Borjesson SI, et al. (2010) Electrostatic tuning of cellular excitability. Biophys J

98:396–403.16. Catterall WA, et al. (2007) Voltage-gated ion channels and gating modifier toxins.

Toxicon 49:124–141.17. Catterall WA (2010) Ion channel voltage sensors:structure, function, and pathophysiol-

ogy. Neuron 67:915–928.18. Borjesson SI, Elinder F (2011) An electrostatic potassium channel opener targeting the

final voltage sensor transition. J Gen Physiol 137:563–577.19. Hessa T, White SH, von Heijne G (2005) Membrane insertion of a potassium-channel

voltage sensor. Science 307:1427.20. Payandeh J, et al. (2011) The crystal structure of a voltage-gated sodium channel.

Nature 475:353–358.21. Schoppa NE, Sigworth FJ (1998) Activation of Shaker potassium channels. III. An acti-

vation gatingmodel for wild-type and V2mutant channels. J Gen Physiol 111:313–342.22. Larsson H, Elinder F (2000) A conserved glutamate is important for slow inactivation in

Kþ channels. Neuron 27:573–583.23. Broomand A, Elinder F (2008) Large-scale movement within the voltage-sensor paddle

of a potassium channel-support for a helical-screw motion. Neuron 59:770–777.24. Careaga CL, Falke JJ (1992) Thermal motions of surface alpha-helices in the

D-galactose chemosensory receptor Detection by disulfide trapping. J Mol Biol226:1219–1235.

25. Webster SM, et al. (2004) Intracellular gate opening in Shaker Kþ channels defined byhigh-affinity metal bridges. Nature 428:864–868.

26. Campos FV, et al. (2007) Two atomic constraints unambiguously position the S4 seg-ment relative to S1 and S2 segments in the closed stateof Shaker K channel. Proc NatlAcad Sci USA 104:7904–7909.

27. Li M, et al. (2011) Gating the pore of P2X receptor channels. Nat Neurosci 11:883–887.28. Lin MC, et al. (2011) R1 in the Shaker S4 occupies the gating charge transfer center in

the resting state. J Gen Physiol 138:155–163.29. Rulisek L, Vondrasek J (1998) Coordination geometries of selected transition metal

ions (Co2þ, Ni2þ, Cu2þ, Zn2þ , Cd2þ , and Hg2þ) in metalloproteins. J Inorg Biochem71:115–127.

30. Vargas E, Bezanilla F, Roux B (2011) In search of a consensus model of the resting stateof a voltage-sensing domain. Neuron 72:713–720.

31. Lehmann C, Heldstab H, Greeff NG (2012) Positioning and Guidance of the VoltageSensor S4Within the Omega/Gating-Pore in the Shaker K-Channel. Biophys J 102:532a.

32. Pless SA, et al. (2011) Contributions of counter-charge in a potassium channel voltage-sensor domain. Nat Chem Biol 7:617–623.

33. Schwaiger CS, et al. (2011) 3-helix conformation facilitates the transition of a voltagesensor S4 segment toward the down state. Biophys J 100:1446–1454.

34. Khalili-Araghi F, et al. (2010) Calculation of the gating charge for the Kv1.2 voltage-activated potassium channel. Biophys J 98:2189–2198.

35. Schow EV, et al. (2010) Down-state model of the voltage-sensing domain of a potas-sium channel. Biophys J 98:2857–2866.

36. Pathak MM, et al. (2007) Closing in on the resting state of the Shaker K(+) channel.Neuron 56:124–140.

37. Tao X, et al. (2010) A gating charge transfer center in voltage sensors. Science328:67–73.

38. Delemotte L, et al. (2011) Intermediate states of the Kv1.2 voltage sensor fromatomistic molecular dynamics simulations. Proc Natl Acad Sci USA 108:6109–6114.

39. Cole KS, Moore JW (1960) Potassium ion current in the squid giant axon:dynamiccharacteristic. Biophys J 1:1–14.

40. Khalili-Araghi F, et al. (2012) Molecular Dynamics Investigation of the omega-Currentin the Kv1.2 Voltage Sensor Domains. Biophys J 102:258–267.

41. Kamb A, Tseng-Crank J, Tanouye MA (1988) Multiple products of the DrosophilaShaker gene may contribute to potassium channel diversity. Neuron 1:421–430.

42. Hoshi T, Zagotta WN, Aldrich RW (1990) Biophysical and molecular mechanisms ofShaker potassium channel inactivation. Science 250:533–538.

43. Soding J (2005) Protein homology detection by HMM-HMM comparison. Bioinfor-matics 21:951–960.

44. Sali A, Blundell TL (1993) Comparative protein modelling by satisfaction of spatialrestraints. J Mol Biol 234:779–815.

45. Lindorff-Larsen K, et al. (2010) Improved side-chain torsion potentials for the Amberff99SB protein force field. Proteins 78:1950–1958.

46. Berger O, Edholm O, Jahnig F (1997) Molecular dynamics simulations of a fluid bilayerof dipalmitoylphosphatidylcholine at full hydration, constant pressure, and constanttemperature. Biophys J 72:2002–2013.

47. Jorgensen WL, et al. (1983) Comparison of simple potential functions for simulatingliquid water. J Chem Phys 79:926–935.

48. Hess B, et al. (2008) GROMACS 4.0:Algorithms for highly efficient, load balanced, andscalable molecular simulation. J Chem Theory Comput 4:435–447.

49. Bussi G, Donadio D, Parrinello M (2007) Canonical sampling through velocity rescaling.J Chem Phys 126:014101–014107.

50. Parrinello M, Rahman A (1981) Polymorphic transitions in single crystals:A new mole-cular dynamics method. J Appl Phys 52:7182–7190.

Henrion et al. PNAS ∣ May 29, 2012 ∣ vol. 109 ∣ no. 22 ∣ 8557

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

June

14,

202

0