The trajectory of discrete gating charges in a voltage-gated potassium channel

Positively-charged amino acids respond to membrane potential changes to drive voltage sensor movement in voltage-gated ion channels, but determining the trajectory of voltage sensor gating charges has proven difficult. We optically tracked the movement of the two most extracellular charged residues (R1, R2) in the Shaker potassium channel voltage sensor using a fluorescent positively-charged bimane derivative (qBBr) that is strongly quenched by tryptophan. By individually mutating residues to tryptophan within the putative trajectory of gating charges, we observed that the charge pathway during activation is a rotation and a tilted translation that differs between R1 and R2 and is distinct from their deactivation pathway. Tryptophan-induced quenching of qBBr also indicates that a crucial residue of the hydrophobic plug is linked to the Cole-Moore shift through its interaction with R1. Finally, we show that this approach extends to additional voltage-sensing membrane proteins using the Ciona intestinalis voltage sensitive phosphatase (CiVSP).


Introduction 26
The nature of the motion of the voltage sensor in voltage-gated ion channels has been a 27 subject of intensive research. This motion is driven by voltage changes sensed by 28 positively-charged amino acids (typically arginines) found on the fourth transmembrane 29 segment (S4) of each monomer of the tetrameric channel. The number of charged 30 amino acids driving this motion varies from channel to channel, but has been shown in 31 the canonical voltage-gated Shaker potassium channel (Kv) to consist of the four most 32 extracellular arginines (Aggarwal & MacKinnon, 1996;Seoh et al., 1996). However,  Lainé et al., 2003), and accessibility studies have the same limitation (Yang & 41 Horn, 1995). Similarly, site directed fluorimetric approaches typically replace a residue 42 with a cysteine and then attach a fluorescent dye, providing the additional advantage of 43 being able to monitor conformational changes in real-time, but do not directly follow the 44 movement of the gating charges (Cha & Bezanilla, 1997;Mannuzzu et al., 1996;Priest 3 used to replace the positive charge and characterize discrete gating charges (Ahern & 48 Horn, 2004Baker et al., 1998;Larsson et al., 1996). However, these 49 replacement charges cannot be rapidly monitored, limiting their use for observing 50 conformational changes. 51 Ideally, one would like to follow the movement of individual gating charges in real 52 time as they respond to changes in the electric field. This requires a fluorophore that is 53 comparable to the gating arginines. We used monobromo(trimethylammonio)bimane 54 (qBBr), a small molecule fluorescent dye with a permanent positive charge ( Figure 1A). Tryptophan has been shown to strongly quench qBBr fluorescence, with weak 64 quenching by tyrosine, and no quenching from various other amino acids including 65 histidine, phenylalanine, methionine, aspartate, or arginine (Mansoor et al., 2010). This 66 phenomenon of 'tryptophan-induced quenching' in bimane dyes generally (Mansoor et 67 al., 2002), together with their remarkable environmental insensitivity (Mansoor et al., 68 1999), has been taken advantage of to measure conformational rearrangements of 69 various membrane proteins, including the β2-adrenergic G protein-coupled receptor 70 plug and provides the basis of the Cole-Moore shift. This technique should also be 94 transferrable to other voltage-sensing membrane proteins; as a proof of principle, we 95 demonstrate its use in the voltage-sensitive phosphatase CiVSP. 96 97

Results 98
The basic principles of qBBr gating charge tracking in time 99 The idea of the present approach is to study the translocation of the gating 100 charge (now a fluorophore, qBBr) as the membrane potential is changed using the 101 specific quenching of qBBr by a tryptophan (W) that is positioned nearby or in the path 102 of qBBr. 103 If we were tracking fluorescence at the single molecule level, the fluorescence 104 signal we would observe would depend on the position of the W with respect to the 105 moving qBBr. Let us assume that we are applying a positive voltage to activate the 106 voltage sensor that moves between two discrete positions. We can distinguish three 107 extreme cases schematically ( Figure 1C, left panel). If the W is near the resting position 108 of the qBBr, we would see a sudden increase in fluorescence when the qBBr moves 109 away from it and the time lag before that increase corresponds to the waiting time of the 110 sensor before it jumps across the energy barrier ( Figure 1C

130
sensors that retain a bare cysteine. For R1C-qBBr, gating currents appeared to be 177 faster than those of the wild type channel but still comparable in voltage sensitivity 178 (Figures 2A and 2B). With R2C-qBBr, gating currents were also more rapid, with voltage 179 sensor movement beginning at more hyperpolarized potentials than in the wild-type 180 channel and producing a shallow Q-V curve (Figures 2A and 2B). These findings 181 suggest that the additional bulk of qBBr may destabilize the resting state of the voltage 182 sensor when it is conjugated to the more intracellular R2C, but not when it is conjugated 183 to the more extracellular R1C. This is in good agreement with numerous computational 184 models of the resting state of the Kv channel that suggest that R1 is less sterically 185 inhibited in the resting state than R2 (Vargas et al., 2012). They also suggest that qBBr 186 would act as a faithful mimic of the movement of the gating charge of R1, and should 187 also mimic the movement of R2, albeit with energetic differences.    To map the trajectory of R1 and R2 using qBBr, we created a series of constructs 223 based on the R1C:W454A and the R2C backgrounds. We used these two constructs as 224 backgrounds because they only showed a very small depolarization-induced increase in 225 fluorescence that had a fast time course with no voltage dependent kinetics ( Figure 2E) 226 which we were unable to link to any endogenous tyrosine or tryptophan (      cluster of tryptophans inserted at L294, I287, and I241. As seen with R1, there is a 313 rotation in the movement of R2; however, the rotation of R2 is less evident. 314 Interestingly, the fluorescence responses differ qualitatively between the two 315 gating charges for tryptophans substituted at L294, T326, and W454. Thus, qBBr 316 mapping reveals that R1 and R2 travel distinct paths in relationship to the voltage-317 sensing domain during activation. Notable exceptions to this include Y415, W289, and 318 F244, which produce similar responses from both R1C and R2C-qBBr. As it faces the 319 lipid membrane, it is unsurprising that W289 does not quench either R1C or R2C-qBBr; 320 however, this suggests that the S2 in which it resides is unlikely to undergo any large 321 rotations that would expose this residue to the gating charges. Y415W quenches both 322 both activation and deactivation are slowed, but activation much more strongly (Hong & 355 Miller, 2000). Thus, we have strong evidence through both the differing kinetics and the 356 leftward shifts in the FVs of R1C-qBBr fluorescence that a discrete gating charge takes a different path in activation than during deactivation; this is also seen with R2C-qBBr 369 ( Figure 6-figure supplement 1). 370 In addition to a resting state and an active state, the voltage sensor has a third 371 state, called the relaxed state, which it enters after prolonged depolarization (F. Previous studies have shown that at negative potentials R1C in the closed state can 410 spontaneously link to I287C, which is a full turn above F290 (Campos et al., 2007). Here, 411 we observe that at extreme hyperpolarizing potentials (-160 mV) R1C-qBBr moves near 412 F290W, producing a slow quenching of fluorescence. We explain this as a result of 413 moving the VSD with a strong hyperpolarization to such an extreme intracellular position

Voltage-sensitive membrane protein CiVSP is interrogable by qBBr 435
Our technique presented here allows for the interrogation of other VSD movements.  In this paper we have demonstrated a technique that allows tracking of a gating charge 458 surrogate in the pathway of a voltage sensor using qBBr. We have obtained new 459 information on the trajectories of the first two gating charges in Shaker (Figures 4 and 5). We propose that this is a flexible tool that should prove readily applicable to other 461 voltage-sensitive proteins. As we discuss the details of the trajectories, we will point out 462 some of the strengths and limitations of the technique. 463 464

Spatial resolution 465
In principle, one would expect qBBr mapping to be limited in its spatial resolution. pathway, we obtain much higher spatial resolution than measuring static fluorescence 476 quenching in two positions. This is because we can detect very small changes in 477 relative fluorescence over time as the qBBr approaches or recedes from the quencher. 478 Take as an example qBBr moving from a position adjacent to W to a distance of 479 7 Å away from the quencher during a depolarization. The qBBr-W distance is always 480 shorter than the 10 Å known to be the qBBr-W quenching distance. However, the 481 degree of quenching will vary from extremely high when qBBr is close to W to less as it 482 moves to its final position. The change in fluorescence may be only 1 to 2%, but this change still reveals that the qBBr-quencher distance is increasing. In other words, we 484 infer that it moved away even though we do not know the exact distance. 485 We cannot absolutely calibrate the degree of quenching, mainly because of 486 spurious background fluorescence. Therefore, the technique presented here takes full 487 advantage of the insertion of a quencher in the putative path: when a fluorescence 488 change occurs during a voltage pulse, we can infer the charge is moving with respect to 489 the quencher while if the signal does not change, it means that either the quencher is 490 far from the charge's path or there is no movement with respect to the quencher. 491 As a result, even though the interpretation is qualitative rather than quantitative, fluorescence changes are always in the same direction (either an increase or a 508 decrease), albeit with different kinetics and more than one time constant. The different 509 kinetics observed between different constructs, therefore, may be explained by the brief 510 dwell time in the intermediate state; however, we cannot exclude that the introduction of 511 a W mutation alters the underlying gating current kinetics. Consequently, although it 512 would be ideal to be able to compare kinetics of one construct to another, the possible 513 differences produced by the W mutations preclude us from doing so. 514 However, using the same construct, voltage pulses of different durations or 515 potential can be used to interrogate different transitions of the voltage sensor, and the 516 kinetics of different processes from the same construct can then be compared. For 517 example, the deactivation kinetics of the fluorescence change produced by R1C-518 qBBr:W454A;T326W are more rapid than its activation kinetics, suggesting that the 519 interaction of R1C-qBBr with T326W is different during activation than during 520 deactivation. Additionally, we observe that the QV and FVs of deactivation have a 521 leftward shift to that of activation, thus demonstrating that the path of deactivation is 522 different than that of activation. This in turn informs us that the deactivation pathway of 523 R1C-qBBr must be different from its activation pathway, a conclusion that would be 524 difficult to obtain with any other method. Future studies could use qBBr optical mapping 525 to investigate the movement of discrete gating charges during conductive events to 526 correlate features of gating with ion conductance. kinetics of the main motion of the voltage sensor. This result gives direct evidence that 530 R1 can populate a region of the sensor closer to the intracellular side when the 531 membrane is strongly hyperpolarized and provides a molecular basis for the  Moore shift. 533 534

Trajectories of discrete gating charges 535
Our data indicate that in the normal resting state, it appears that R1 does not 536 come into close contact with F290 or I294 in the S2 segment, while R2 does come into 537 proximity with I294. In the normal active state, R1 comes into close contact with W454 538 and Y415, while R2 comes into close contact with Y415, but does not interact with 539 W454. One observation potentially related to the position of the resting state is that two 540 of our tested constructs were lethal: R1C:W454A;I287W and R2C:F290W. Whether the 541 lethality of these constructs stems from a disruption of the normal resting state 542 interactions of discrete gating charges with the gating pore remains to be investigated.