S4–S5 linker movement during activation andinactivation in voltage-gated K+ channelsTanja Kalstrupa and Rikard Bluncka,b,1

aDepartment of Pharmacology and Physiology, Université de Montréal, Montréal, QC H3C 3J7, Canada; and bDepartment of Physics, Université deMontréal, Montréal, QC H3C 3J7, Canada

Edited by Francisco Bezanilla, The University of Chicago, Chicago, IL, and approved June 6, 2018 (received for review November 1, 2017)

The S4–S5 linker physically links voltage sensor and pore domainin voltage-gated ion channels and is essential for electromechan-ical coupling between both domains. Little dynamic information isavailable on the movement of the cytosolic S4–S5 linker due tolack of a direct electrical or optical readout. To understand themovements of the gating machinery during activation and inacti-vation, we incorporated fluorescent unnatural amino acids at fourpositions along the linker of the Shaker KV channel. Using two-color voltage-clamp fluorometry, we compared S4–S5 linker move-ments with charge displacement, S4 movement, and pore opening.We found that the proximal S4–S5 linker moves with the S4 helixthroughout the gating process, whereas the distal portion un-dergoes a separate motion related to late gating transitions. Bothpore and S4–S5 linker undergo rearrangements during C-type in-activation. In presence of accelerated C-type inactivation, the en-ergetic coupling between movement of the distal S4–S5 linker andpore opening disappears.

voltage-gated potassium channels | unnatural amino acids |voltage-clamp fluorometry | inactivation | Anap

Voltage-gated potassium (KV) channels are responsible forrepolarization of the membrane potential during neuronal

and cardiac action potentials and regulate neuronal excitabilityamong many other physiological processes. Mutations in KVchannel genes thus result in severe neuronal and cardiac diseasesand channelopathies, such as episodic ataxia, epilepsy, and car-diac arrhythmia. KV channels thus constitute therapeutic drugtargets for treatments of diverse diseases (1). KV channels’central role in heart and brain underscores the importance ofunderstanding their mechanisms in detail.All mammalian Shaker-like KV channels (KV1 channel sub-

family) share the characteristic assembly of four subunits whicheach consists of six transmembrane helices containing a voltagesensor (S1–S4) and a pore region (S5–S6; Fig. 1A) (2, 3). Thevoltage sensor (VS) contains positively charged arginine residuesin the S4 helix, which are responsible for the generation of gatingcurrents as they rearrange according to changes in the electricfield they sense across the membrane (4). This conformationalchange results in pore opening through a mechanism calledelectromechanical coupling (5–12).Although dynamic information is lacking and the underlying

molecular driving forces are not fully understood, a consensusmodel of the coupling process has emerged (13, 14). First, initialS4 charge movements (Q1) occur independently in each VSupon membrane depolarization, applying a force onto the S4–S5 linker as they move upward within the membrane. Then,during a second S4 charge movement (Q2), energy is released tothe pore domain in a cooperative conformational change, whichfinally results in widening of the bundle crossing at the in-tracellular S6 gates. The opening step has been shown to occurcooperatively among the subunits (15, 16) and comprises at leasttwo transitions (17). In addition to being covalently bound to thepore domain via S5, the S4–S5 linker interacts directly with theS6 bundle crossing via noncovalent interactions of both the same(6, 7, 11, 18) and the neighboring subunit (19). Because dynamics

of the S4 helices and the inner S6 gate are associated with gatingcurrents and ionic currents, respectively, the roles of each regionin the activation process have been well established. The S4–S5 linker movement, on the contrary, is intrinsically not associ-ated to any charge movement. Moreover, the intracellular posi-tion of the S4–S5 linker has made it inaccessible for fluorophore-labeling. Consequently, only limited dynamical information isavailable for the S4–S5 linker movement (20).To investigate S4–S5 linker dynamics, we have made use of the

stop codon suppression technique for the genetic incorporationof a fluorescent unnatural amino acid (Anap) (17, 21, 22), en-abling specific labeling of intracellular sites for the use of voltageclamp fluorometry (VCF) in proteins expressed in Xenopus oo-cytes (17, 23). We aimed to address questions as to where thetransition between the independent voltage sensor movementand the cooperative pore opening occurred and whether the S4–S5 linker rearranges as a rigid body throughout the gating pro-cess or with a degree of flexibility, allowing conservation of theenergy provided by the voltage sensor. To this end, Anap wasinserted into the S4–S5 linker of the Shaker KV channel. Bylabeling the external part of the S4 helix with a thiol-reactive dye,we were able to compare movements of the voltage sensor and ofthe S4–S5 linker simultaneously.

ResultsAssessment of Functional Expression with Anap. All 12 positionsalong the Shaker S4–S5 linker (residues 381–392) were scannedfor Anap insertion (Fig. 1B). Proper protein maturation andfolding were determined from the presence of both ionic and

Significance

Excitability in heart and nervous system is based on sensingand propagation of the membrane potential by voltage-gatedion channels. Despite the increasing availability of high-reso-lution structures of voltage-gated ion channels, key questionsabout their dynamics remain elusive. Here we followed themovements of the gating machinery on the cytosolic surface ofthe Shaker KV channel by introducing a fluorescent non-canonical amino acid at different positions along the linkerbetween voltage sensor and pore. We can thus map themovement of each position and reconstruct the dynamics ofthe gating machinery during activation and inactivation. Con-sidering that the Shaker channel serves as a model system, themechanisms are likely conserved among those KV channelscontaining a sizeable S4–S5 linker.

Author contributions: R.B. designed research; T.K. performed research; T.K. and R.B. an-alyzed data; and T.K. and R.B. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Published under the PNAS license.1To whom correspondence should be addressed. Email: rikard.blunck@umontreal.ca.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1719105115/-/DCSupplemental.

Published online June 29, 2018.

www.pnas.org/cgi/doi/10.1073/pnas.1719105115 PNAS | vol. 115 | no. 29 | E6751–E6759

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 25

, 202

0

gating currents. Four positions resulted in robust expression andvoltage-dependent fluorescence changes with Anap inserted atL382, R387, K390, and A391, whereas four did not express(G381, Y385, T388, and G392). The four remaining residueseither showed limited expression (Q383 and L389) or defectivechannel opening (I384 and G386). This agrees with the essentialrole that I384 has been shown previously to play for electrome-chanical coupling (8). When mapping the other residues onto theavailable KV channel crystal structures (2, 3), the nonexpressingresidues face inward toward the center of the channel (black inFig. 1B), whereas expressing residues point outward, away fromthe channel (red in Fig. 1B). This finding agrees with the fact that alarge amino acid like Anap is more likely to sterically disrupt proteinfolding at positions that interact with the protein versus positionsthat point outward into the cytosol or into the lipid bilayer.It was recently shown that Shaker channels lacking the pore

domain express as functional isolated voltage sensor domains(iVSD), giving rise to gating pore currents (24). These currentswere also observed in mutants with a stop-codon insertion in theS4–S5 linker in the absence of Anap or pAnap (SI Appendix, Fig.S1 A and B). However, although iVSD currents also occurredwhen Anap and pAnap were present, their amplitude dependedon the success of Anap incorporation, such that iVSD currentamplitudes were considerably smaller in oocytes expressing full-length channels (4.4 μA ± 6.7 at −170 mV) compared with oocytesin which full-length channels were absent (20.4 μA ± 7.9 at−170 mV; SI Appendix, Fig. S1C). Having ensured that iVSD cur-rents were not altering gating kinetics and did not form hetero-tetramers (SI Appendix), we were able to proceed to the analysis ofthe fluorescence signals.

The C Terminus of the S4–S5 Linker Rearranges After S4 Activation.The four S4–S5 linker mutants L382Anap, R387Anap, K390Anap,and A391Anap of full-length Shaker exhibited robust expressionlevels, with Anap changes showing relative fluorescence intensitychanges (ΔF/F) in the range of 0.2–4% in response to depolarization(Fig. 2A). Because fluorescent labeling either by insertion of Anapor the thiol-reactive fluorophore tetramethylrhodamine-maleimide(TMR) might influence kinetics and energy of voltage sensormovement and pore opening, we can only directly compare voltagesensor movement, movement of the S4–S5 linker, and pore openingin the same mutant, i.e., tracking all parameters simultaneously. Weshowed previously that fluorescence originating from TMR at-tached to the extracellular end of S4 reliably follows gating chargemovement (8). Therefore, the stop codons were inserted in theShaker-A359C background, to compare S4–S5 linker movement(Anap signals) with voltage sensor movement (A359C-TMR sig-nals), simultaneously (Fig. 2A). In addition, the ionic current—reporting pore opening—was measured at the same time (Fig. 2A).

As expected, the voltage dependence of voltage sensor move-ment (FV-TMR) for all four mutants developed at less depolarizedpotentials compared with pore opening [conductance voltagerelation, GV (Fig. 2 B–E); ΔV1/2 L382Anap −67 mV; R387Anap−33.6 mV; K390Anap −27.3 mV; A391Anap −43.2 mV]. In con-trast, comparison of the voltage dependencies of voltage sensormovement (FV-TMR) and movement at the S4–S5 linker (FV-Anap) varied with the monitored position in the S4–S5 linker.Although the signals overlapped in L382Anap (Fig. 2B), themovements at R387, K390, and A391, in the S4–S5 linker, wereshifted to more depolarized potentials compared with voltagesensor movement by 29, 21, and 21 mV, respectively (Fig. 2 C–Eand Table 1). However, movement at all three positions was stilloccurring at less depolarized potentials than pore opening (GV;ΔV1/2 R387Anap −14.9 mV; K390Anap −6.6 mV; A391Anap−21.4 mV).To ensure that the signal was caused by conformational changes

at the labeling position, we excluded that the fluorescence wasinfluenced by potassium flux through the nearby pore entrance.Conductance might alter local ion concentrations and, thereby,the local electric field, which could influence fluorescence signals.To obtain good fluorescence signals, we need to express theconducting channels at high levels. The accumulation of externalK+ at high expression levels also led to slowing of channel de-activation as evidenced by the slower tail currents (Fig. 2A) (25,26). We used Ba2+ to block the ionic current. Ba2+ has a sizesimilar to K+ and can enter the pore but blocks conductance onceit reaches the selectivity filter with minimal effects on voltagesensor activation (27). Despite significant block by external Ba2+ inR387Anap and K390Anap channels, the total Anap fluorescencechange was not diminished, and the fluorescence time course andvoltage dependency remained unaffected, indicating that the fluo-rescence signals were indeed caused by conformational changesat the labeling site (Fig. 2 F and G). The results suggest that onlythe N-terminal part of the linker (L382) moves together with theS4 helix during activation, whereas the more distal part of thelinker rearranges later in the activation process.

The S4–S5 Linker Experience C-Type Inactivation-Related Rearrangements.A second fluorescence component of opposite direction appearedin A391Anap at depolarizations more positive than 60 mV (Fig.2E). An overlap of the current and fluorescence time courseduring a prolonged depolarization pulse revealed that the secondfluorescence component correlated kinetically with C-type in-activation (Fig. 2H). Moreover, the second fluorescence compo-nent was sensitive to application of 4-aminopyridine (4-AP; Fig.2H). 4-AP is a KV channel blocker, which prevents KV channelsfrom entering the final concerted pore opening transition and thusalso C-type inactivation (28).This was confirmed in nonconducting W434F mutants, in which

ionic currents are blocked due to accelerated C-type inactivation(29). In agreement with the accelerated C-type inactivation (29),the second Anap fluorescence component in A391Anap-W434Fchannels appeared faster and larger than in absence of the W434Fmutation (Fig. 3A andE). The FV fromA391Anap-W434F channelscould no longer be fitted to a single Boltzmann relation be-cause the onset for C-type inactivation now occurred at lowerpotentials than channel opening did (Fig. 3E). These findingsindicate that A391Anap probes rearrangements that are as-sociated with C-type inactivation, in addition to those associ-ated to activation.A second smaller Anap component of opposite direction also

appeared in K390Anap-W434F channels (Fig. 3D). For bothK390Anap-W434F and A391Anap-W434F mutants, the secondfluorescence component was 4-AP sensitive as demonstrated inthe fluorescence time course and in the FV obtained before andafter 4-AP application (Fig. 3 F–I). Because the second fluo-rescence component in the conducting K390Anap channels is

Fig. 1. Overview of KV channel structure. (A) Topology of a Shaker subunitwith relevant positions highlighted. A cysteine was inserted at positionA359 for TMR labeling (green), and Anap was inserted into the S4–S5 linker(red). (B) Bottom view of the KV1.2–2.1 crystal structure of the homotetra-meric transmembrane segment (3). At the bottom the S4–S5 linker sequenceis shown, which was scanned for Anap incorporation. Anap was successfullyincorporated into positions marked in red. The asterisks mark positions usedto characterize the S4–S5 linker movement.

E6752 | www.pnas.org/cgi/doi/10.1073/pnas.1719105115 Kalstrup and Blunck

Dow

nloa

ded

by g

uest

on

Nov

embe

r 25

, 202

0

absent or too small to be measured, it was not possible to verifywhether the kinetics of C-type inactivation correlated with thefluorescence signal as was the case with A391Anap (Fig. 2H).The W434F data, nonetheless, suggest that Anap is influencedby C-type inactivation at position K390 in the same way as atposition A391.

C-Type Inactivation Allows the S4–S5 Linker to Assume Final PositionAfter Voltage Sensor Activation. The most remarkable differencein the W434F background was that the shift between voltage

sensor (TMR) and S4–S5 linker (Anap) movement was stronglydiminished (Fig. 3 B–E). The disappearance of the shift indicatesthat in presence of W434F, the entire S4–S5 linker undergoes itsconformational change upon voltage sensor activation. In theconducting channel, in contrast, the C terminus of the S4–S5 linker cannot enter its final position yet after voltage sensoractivation but is energetically coupled to an additional transition.The question remains whether also the sequence of events has

altered, i.e., voltage sensor movement, followed by linker move-ment and finally pore opening. To answer this, we considered

Table 1. V1/2 and z values were obtained from single Boltzmann fits of FV and GV curves of the indicated mutants

Mutant

TMR FV Anap FV GV

V1/2, mV z V1/2, mV z V1/2, mV z

L382Anap −39.2 ± 2.7 1.08 ± 0.09 −43.9 ± 0.7 1.91 ± 0.11 27.8 ± 1.0 1.23 ± 0.05R387Anap −27.4 ± 2.1 1.27 ± 0.10 1.3 ± 1.3 1.46 ± 0.16 16.2 ± 0.5 1.94 ± 0.07K390Anap −42.9 ± 1.9 1.26 ± 0.09 −22.2 ± 0.9 1.24 ± 0.09 −15.6 ± 0.7 1.74 ± 0.06A391Anap −46.4 ± 1.3 1.37 ± 0.08 −24.6 ± 1.0 2.07 ± 0.20 −3.2 ± 1.0 1.47 ± 0.01

Fig. 2. Fluorescence profile of conducting channels. (A) VCF recordings of (Upper) ionic currents, (Middle) Anap, and (Lower) TMR (green) fluorescencechanges obtained from oocytes expressing each of the S4–S5 linker mutants. The arrow next to each fluorescence signal represents 1% ΔF/F. Anap FV, TMRFV, and GV curves of (B) L382Anap, (C) R387Anap, (D) K390Anap, and (E) A391Anap channels. Each dataset is fitted to a Boltzmann distribution (Methods).Error bars indicate mean ± SD. (F and G) Current and fluorescence output upon depolarization from −90 to 20 mV before and 30 min after external ap-plication of 40 mM barium in (F) R387Anap and (G) K390Anap channels. (Inset) Anap FV obtained before and after block. (H) Anap fluorescence (black) andionic currents (red) of A391Anap channels upon prolonged 1-s depolarization from −90 mV to (Left) 50 and (Right) 100 mV. In gray, ionic current and Anapfluorescence after application of 5 mM external 4-AP for the 50-mV pulse is shown.

Kalstrup and Blunck PNAS | vol. 115 | no. 29 | E6753

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 25

, 202

0

the dynamic information. The time constants obtained fromexponential fits of the fluorescence and gating currents wereplotted as a function of the test pulse (Fig. 4 A–D). The mutantsexhibited similar patterns for TMR and gating currents, eachwith two components: one in the 1- to 10-ms range and one inthe 10- to 100-ms range (green and black symbols, Fig. 4 A–D)except for A391Anap-W434F gating currents, which only couldbe fitted to one exponential (Fig. 4D). This could be either be-cause the two components are kinetically similar (yet havedifferent voltage dependencies because the QV was fitted to adouble Boltzmann relation) and therefore appear in-distinguishable or because the slower component is too small ortoo slow to be detected during the test pulse. In fact,A391Anap-W434F generally expressed less than the otherlinker mutants yielding maximally 1 μA of gating currentswhen depolarizing from −90 to 50 mV. It is therefore possible

that the slow component in the gating currents disappearedwithin the experimental noise. Nevertheless, these findings sug-gest that the kinetics of the upper S4 movement and the chargedisplacement were not markedly affected by the relative positionof Anap in the S4–S5 linker in the W434F mutant.However, how did the voltage sensor kinetics compare with

the linker movement monitored by Anap? In L382Anap-W434F,two Anap components were observed: one in the 1- to 10-msrange and another in the 10- to 100-ms range (red symbols, Fig.4A). Although the fast TMR component is 1–5 ms slower thanthe fast Anap component, the kinetics of TMR, Anap, and gatingcurrents are comparable. The result agrees with the overlappingFVs and QV for L382Anap-W434F (Fig. 3B). When overlappingthe fluorescence time courses and charge displacement, it is seenthat TMR relaxes slower than Anap and that Anap is kineticallycloser to the charge displacement in the first miliseconds where

Fig. 3. Fluorescence profile of nonconducting channels. (A) VCF recordings of (Upper) gating currents, (Middle) Anap, and (Lower) TMR (green) fluorescencechanges obtained from oocytes expressing each of the S4–S5 linker mutants with the W434F mutation. Anap FV, TMR FV, and QV curves of (B) L382Anap-W434F, (C) R387Anap-W434F, (D) K390Anap-W434F, and (E) A391Anap-W434F channels. Except for Anap FV in D and E, each dataset is fitted to a Boltzmanndistribution (Methods). Error bars indicate mean ± SD. (F) Anap fluorescence of K390Anap-W434F upon depolarization from −90 mV before (black) and after(gray) external application of 5 mM 4-AP. (G) Same as F but for A391Anap-W434F channels. (H and I) Anap FV before (filled symbols) and after (open symbols)4-AP from (H) K390Anap-W434F and (I) A391Anap-W434F channels. The FV obtained after 4-AP was fitted to a Boltzmann distribution.

E6754 | www.pnas.org/cgi/doi/10.1073/pnas.1719105115 Kalstrup and Blunck

Dow

nloa

ded

by g

uest

on

Nov

embe

r 25

, 202

0

the fast component is dominant (arrow, Fig. 4E). Furthermore,the absence of a delay in the onset of the Anap fluorescencecompared with TMR indicates that their movements areprompted simultaneously (black arrow, Fig. 5A). These data aresimilar to earlier data of V234Anap on the S1 helix, which is only9.6 Å away (17) (Fig. 4F). Based on the kinetic analysis and FV

and QV comparisons, the results show that the L382 movementis similar to that of the voltage sensor.For the C-terminal part of the S4–S5 linker, however, Anap

fluorescence only exhibited a single time constant (red circles,Fig. 4 B–D) when disregarding the 4-AP–sensitive slow compo-nent of opposite direction in K390Anap-W434F and A391Anap-W434F (red open squares, Fig. 4 C and D). In contrast toL382Anap-W434F, superposition of the TMR and Anap fluo-rescence time courses revealed a delayed onset for Anap (Fig.5G). This was best observed in R387Anap-W434F channelswhere it amounted to 1.40 ± 0.38 ms at 0 mV (Fig. 5B). Thedelay in K390Anap-W434F channels was slightly shorter with0.93 ± 0.37 ms at 0 mV (Fig. 5C), which was likely due to theoverall faster kinetics of gating and ionic currents in K390Anap-W434F. The delay was not visible in A391Anap-W434F due todecreased signal-to-noise ratio as a result of limited expression.The delay of L382Anap-W434F was within the error margin.To ensure that the delay was not related to instantaneous C-type

inactivation in the W434F mutant, we analyzed the TMR and Anapsignals in the conducting L382Anap, R397Anap, and K390Anapmutants (Fig. 5 D–G). Although due to smaller expression level, thesignal to noise ratio was less favorable, the signals displayed thesame—even slightly longer—delay for R397Anap and K390Anap,whereas the delay was absent in L382Anap.The presence of a delayed onset in the Anap fluorescence of

the C-terminal linker mutants means that at least one transitionneeds to occur before linker rearrangement also in the W434Fbackground. Therefore, even though the S4–S5 linker is free totransition into its final position upon voltage sensor activation, itis not an independent movement like the voltage sensors.

The C-Terminal S4–S5 Linker Follows both Early and Late Transitions.Except for L382Anap-W434F, the QVs (Fig. 3 B–E) were best fitto a two-transition Boltzmann relation, reflecting the two majorcharge systems Q1 and Q2 (30). There were subtle differences

Fig. 4. Kinetical analysis of fluorescence and gating current time course. (A–D) Time constants obtained from single or double exponential fits of TMRfluorescence (green), Anap fluorescence (red), and gating currents (black)for each of the S4–S5 linker mutants during a test pulse. (E) OverlappingTMR (green), Anap (red), and charge displacement (black) for an oocyteexpressing L382Anap-W434F. Arrow highlights kinetic correlation betweenthe fast component of Anap fluorescence and charge displacement. (F)Structural view from the KV1.2–2.1 crystal structure (3) showing the four S4–S5 linker positions in red and the distance from L382 to V234 in green in S1.

Fig. 5. Comparison of onset of TMR and Anap fluorescence signal. (A–F) Overlap of TMR (green) and Anap (red) fluorescence upon depolarizations from−90 mV. Arrows in A and D highlight the absence of a delayed onset, whereas arrows in B, C, E, and F highlight the presence of a delayed onset for Anap comparedwith TMR. In the boxes a vertical zoom of the curves is shown to better visualize the delay. The depolarizations were chosen with respect to the respective V1/2

values. In the zoomed display, the data were filtered by moving average to remove white noise. (G) Voltage dependence of the delay for all three mutants. Solidsymbols indicate -W434F mutants; open symbols indicate conducting. The fit relates to the W434F data, which were obtained with better signal-to-noise ratio.

Kalstrup and Blunck PNAS | vol. 115 | no. 29 | E6755

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 25

, 202

0

among the FVs of TMR and Anap and the QV. In fact, bycomparing the midpoint values of Q1, Q2, and the FVs of thethree C-terminal S4–S5 linker mutants, it is tempting to concludethat Anap follows Q2, whereas TMR is closer to Q1 (Table 2).However, we would have to separate both charge movements todetermine whether these subtle differences are significant. Tothis end, we inserted the F290A mutation into the R387Anapand K390Anap background.F290 is situated in the S2 helix where it stabilizes the voltage

sensor in the activated state, presumably by controlling the transferof the fourth gating charge (31, 32). When mutated from phe-nylalanine to alanine, the early gating transitions remain nearlyunaffected, whereas the final transitions shift to higher potentials.In agreement with WT-F290A channels (32), the GV curve wasshifted to higher potentials when F290A was inserted into R387Anapand K390Anap (open black circles, Fig. 6 C and D) with GV mid-point values of 134.8 and 95.5 mV, respectively (Table 3).In the W434F background, wild type (F290A-W434F) and both

mutants, R387Anap-F290A-W434F and K390Anap-F290A-W434F,exhibited a biphasic QV in which ∼20% of the charge activatedat higher potentials (Fig. 6 A and B) (31). The second chargemovement developed in the voltage range of channel openingwith Q2 (Table 3). The Anap-FV also split into two components(Fig. 6 C and D). The first component (F1) remained shifted by10–12 mV to potentials more depolarized than Q1, whereas thesecond component (F2) developed in the voltage range of Q2.The slight shift between Q2 and F2 is caused by the shift of theoff-gating currents versus the on-gating currents (8). The Q2 hadto be obtained from off-gating currents because the oocytes ex-hibit endogenous K+ currents at potentials more positive than+50 mV (33, 34) (black arrows, Fig. 6 A, B, and E), making itchallenging to determine the displaced charge during de-polarization in that voltage range. The F290A data show that the C-terminal part of the S4–S5 linker rearranges at least twice duringactivation. The first rearrangement occurs with a voltage de-pendence falling between Q1 and Q2, and the second rearrange-ment coincided with Q2.

Model for the Cytosolic Gating Machinery. In this study, the fluo-rescent unnatural amino acid Anap was site-specifically in-corporated into various positions in the S4–S5 linker of theShaker channel. Two-color voltage clamp fluorometry was usedfor detection of voltage-gated fluorescence changes, allowing thedynamical study of protein rearrangements from two regions in

the same protein simultaneously. Anap was successfully insertedinto both the N-terminal part (L382) and the C-terminal part(R387, K390, and A391) of the S4–S5 linker. In a previous work,we already studied the movement of the cytosolic end of theVSD (V234Anap) and pore opening (H486Anap) (17). Bycombining these results, we are now in a position to follow theprogression of the voltage-gated rearrangements throughoutthe voltage-gated K+ channel Shaker: the extracellular endof the VSD (A359C-TMRM), the cytosolic end of the VSD(V234Anap), the N and C termini of the S4-S5 linker (L382Anap,R387Anap, K390Anap, and A391Anap), and the cytosolicpore gate (H486Anap; Fig. 7).We found that the VSD move independently coinciding with

their gating charge movement. The cytosolic end of VSD movesfaster than the external surface, but they share their voltagedependence (17). Here we found that the N terminus of the S4–S5 linker also follows the gating charge movement (Fig. 7,transition A). The C terminus of the S4–S5 linker, in contrast,moves at potentials intermediate between gating charge move-ment and pore opening (Fig. 2). The difference in movement ofthe N and C termini indicates that due to flexibility betweenL382 and R387, energy is briefly stored in this region until theS4–S5 linker responds by relaxing into its final state (Fig. 7,transition B). The S4–S5 linker thus constitutes the transitionbetween the independently acting voltage sensors and the co-operative pore opening step.Pore opening, monitored at H486, occurred in two steps (Fig.

7, transitions B and C), one of which also precedes conduction(GV) (17) (Fig. 6F). The sequence of events would thus beginwith VSD activation and S4–S5 linker bending. This is followedby “tension” or energy relief in the S4–S5 linker and pre-activation of the pore and finally pore opening and C-type in-activation (Fig. 7, transition C). Although the essential step of C-type inactivation occurs in the selectivity filter (35), it has beenshown to interact with the cytosolic pore gate at the helicalbundle crossing (36). We found accordingly rearrangementscorrelating with C-type inactivation in both the N and C terminiof the S4–S5 linker (Fig. 2) and the S4 (V234Anap) (17). Weverified the rearrangements at the helical bundle crossing atH486Anap and found, indeed, a slow fluorescence change cor-relating with entry into C-type inactivation upon prolongeddepolarization (Fig. 6G). C-type inactivation thus leads to con-formational changes in the cytosolic pore gate, the entire S4–S5 linker, and the VSD.

Table 2. V1/2 and z values were obtained from two- or three-state Boltzmann fits of FV and QV (W434F) curves of the indicatedmutants

Mutant

TMR FV Anap FV QV

V1/2, mV z V1/2, mV z A1/A2 V1/2 (1), mV z(1) V1/2 (2), mV z(2)

L382Anap-W434F −41.5 ± 0.9 1.13 ± 0.03 −51.0 ± 0.4 1.58 ± 0.05 — −45.0 ± 0.6 1.42 ± 0.03 — —

R387Anap-W434F −38.5 ± 0.8 2.27 ± 0.10 −26.1 ± 0.4 2.40 ± 0.13 0.05 −72.2 ± 4.0 3.9 ± 0.3 −30.3 ± 0.5 2.0 ± 0.1K390Anap-W434F −48.1 ± 0.2 2.58 ± 0.05 −40.5 ± 0.5 2.29 ± 0.08 0.09 −77.9 ± 2.2 3.2 ± 0.3 −37.7 ± 1.5 2.6 ± 0.1A391Anap-W434F −42.9 ± 0.4 2.36 ± 0.04 −39.4 ± 0.2 3.42 ± 0.18 0.25 −61.6 ± 4.0 3.3 ± 0.5 −36.0 ± 2.2 3.8 ± 0.05

Table 3. V1/2 and z values obtained from single or three-state Boltzmann fits of FV, QV, and GV curves of F290A mutants

Mutant

Anap FV QV GV

A1/A2

V1/2 (1),mV z(1)

V1/2 (2),mV z(2) A1/A2

V1/2 (1),mV z(1)

V1/2 (2),mV z(2) V1/2, mV z

F290A-R387Anap-W434F

0.4 −24.3 ± 5.1 1.75 ± 0.29 79.1 ± 7.1 1.23 ± 0.32 0.7 −39.8 ± 0.3 2.16 ± 0.06 110.4 ± 6.4 1.41 ± 0.25 134.8 ± 2.7 1.61 ± 0.10

F290A-K390Anap-W434F

0.7 −49.5 ± 0.7 2.32 ± 0.05 69.6 ± 3.4 1.17 ± 0.16 0.95 −58.0 ± 0.3 2.55 ± 0.04 79.2 ± 1.7 2.34 ± 0.27 95.5 ± 1.2 1.29 ± 0.05

E6756 | www.pnas.org/cgi/doi/10.1073/pnas.1719105115 Kalstrup and Blunck

Dow

nloa

ded

by g

uest

on

Nov

embe

r 25

, 202

0

Most significantly, the behavior of the S4–S5 linker is alteredin the presence of accelerated C-type inactivation (W434F) (29).Here we found that the shift in the voltage dependencies be-tween voltage sensor movement and S4–S5 linker rearrange-ments disappears in W434F (Fig. 3). However, the S4–S5 linkerrearrangements are temporally delayed versus the voltage sensormovement, indicating that they do occur subsequently. In con-trast, the movement of the cytosolic pore gate (H486) is notaffected by W434F and still shows two transitions that occur atpotentials more depolarized than gating charge movement (17).These results suggest that the S4–S5 linker, although hindered toenter its final state in the closed WT channel, is no longer hin-dered to directly enter its final conformation in the presence ofaccelerated C-type inactivation. Because pore gate opening isnot altered in W434F, it suggests that it is C-type inactivation,which evens the path to the final state.In the proposed model in Fig. 7, the voltage sensor and S4–

S5 linker would transition directly into their relaxed state uponopening and inactivation (Fig. 7, Right, transition B), whereas wefind the same two transitions at the C-terminal S6 (Fig. 7,transitions C and D). Although we detected the movements ofthe S6 in the presence of W434F, we cannot measure the con-ducting state of the pore and thus cannot be certain whether thetwo transitions are identical to the conducting mutant.

DiscussionOur findings provide us with a detailed picture about the rear-rangements following membrane depolarization in voltage-gatedK+ channels including the voltage sensing domain, S4–S5 linker,and cytosolic pore gate. Temporally correlating the signals withelectrical signals from gating charge movement and ion conduc-tion allowed us to establish a sequence and kinetics of the gatingmechanism at the cytosolic side of voltage-gated K+ channels.The mechanism underlying the fluorescence changes is

quenching of the fluorophore most likely by the surroundingprotein. As such, we probe movement of the labeled site relativeto the quenching group. We therefore have to take more than

one position into account when assigning movements to thefluorescence changes. Due to the α-helical structure of the S4–S5 linker, for instance, the amino acid residues of K390 andA391 point into different directions but still display verysimilar fluorescence signals, indicating that the changes arecaused by movement of the S4–S5 linker rather than itsenvironment.It was remarkable that the C-terminal S4–S5 linker and the

cytosolic pore gate react differently to the presence of W434F.Although the S4–S5 linker is free to move with the voltage sensorin W434F, the cytosolic pore gate still opens at more depolarizedpotentials. Our data suggest that despite uncoupling of S4–S5linker movement from S6 movement in the presence of W434F,the channel still opens normally. This seems to contradict ourcurrent understanding of electromechanical coupling, accordingto which the annealing of the S4–S5 linker and the comple-mentary C-terminal S6 leads to coinciding movements (6, 8, 9–12). A mechanism similar to what we see here, where uncouplingof the S4–S5 linker from S6 is linked to pore inactivation, hasbeen suggested for closed state inactivation in KV4 channels (37),and it will thus be interesting to investigate S4–S5 linker move-ment in KV4.An important feature in the C-terminal S4–S5 linker rear-

rangements was that we still observed a slow conformationalchange even in the presence of W434F. This component wassensitive to 4-AP but occurred at potentials more depolarizedthan pore opening. It might be an indication that the S4–S5 linker, despite not entering a “strained” state in presence ofW434F, only enters its final state after pore opening.Considering the conservation of the electromechanical cou-

pling for many voltage-gated ion channels as well as the occurrenceof C-type inactivation for many voltage-gated potassium channels, weexpect the mechanism described here to be conserved throughoutfor those voltage-gated ion channels that contain a sizeable S4–S5 linker. In this context, also the role of auxiliary subunits of KVchannels will be of interest. The cytosolic β subunits bind tothe lower T1 domain (2) and could only indirectly influence the

Fig. 6. Separation of the final gating transition using F290A. (A and B) (Upper) Gating currents and (Lower) Anap fluorescence changes. The arrows highlightendogenous leak currents during high depolarizations. Anap FV and QV for (C) F290A-R387Anap-W434F and (D) F290A-K390Anap-W434F plotted togetherwith the GV obtained from the respective conducting mutants. (E) QVon (gray) and QVoff (black) obtained from the gating currents in A and B, respectively.For comparison with QVon, QVoff has been normalized to the amplitude of the saddle point. (F) QV, Anap FV, and GV for Shaker-H486Anap-W434F. (G) C-typeinactivation of H486Anap. Current and fluorescence change elicited upon a 1-s depolarization from −90 to +100 mV to induce C-type inactivation inH486Anap channels.

Kalstrup and Blunck PNAS | vol. 115 | no. 29 | E6757

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 25

, 202

0

S4–S5 linker position. In contrast, KChip in KV4 channelsbind laterally to the T1 domain (38) and could therefore in-fluence directly the S4–S5 linker region. How far cytosolic aux-iliary subunits might influence the mechanisms described herewill have to be confirmed in further studies.Recent structures of the EAG and Herg channels (39, 40)

revealed that these channels have very short S4–S5 linkers.Consequently, their voltage sensing domain directly neighborsthe S5–S6 pore domain of the same subunit, as opposed to theneighboring one as observed in Shaker-like KV channels (“do-

main swapping”). In the Shaker-like KV channels, the linker annealsto the C-terminal S6 but also forms a cuff around the pore and“presses” it inward (20). This is not possible in the Herg and EAGchannels, and it will be interesting to study in how far our findingshere are valid in those channels. It had been suggested that in thesechannels, the direct interface between voltage sensor and S5–S6 plays a more important role (40–42).Because both EAG and Herg channels display a depolarizing

shift between the voltage dependencies of charge movement andpore opening, just as Shaker-like KV channels (43–45), the shorterS4–S5 linker does not seem to influence the independent move-ment of the first voltage sensors. However, although HERG showsvery rapid C-type inactivation, EAG channels do not inactivate.One could thus speculate that the longer linker in Shaker-like KVchannels is required for the delay between activation and C-typeinactivation. This would suggest that these channels rather followthe transitions of the W434F mutant (Fig. 7, Right). Only a de-tailed study will be able to establish this.

MethodsXenopus Oocyte Injection. All mutations were introduced into the Dro-sophila Shaker H4 gene in the pBSTA vector with a deletion of amino acids6–46 to remove N-type inactivation (19). For Anap incorporation, 9.2 nL of0.1 ng/nL cDNA encoding the tRNA/AnapRS pair (pAnap) (17, 22) was in-jected into the nucleus of Xenopus laevis oocytes 6–24 h before coinjectionof 23 nL 1 mM Anap (custom synthesis TCRS-170; Abzena TCRS) and 35 ngin vitro transcribed RNA. All steps during and after Anap injection wereperformed under red light to avoid bleaching. Oocytes were then in-cubated for 1–3 d at 18 °C in Barth’s solution supplemented with 5%horse serum.

Electrophysiology. Voltage clamp was performed with a CA-1B amplifier(Dagan). Currents were recorded in the cut-open oocyte voltage-clampconfiguration as described (8) and analyzed by using GPatch (Departmentof Anesthesiology, University of California, Los Angeles). Linear capacitancecurrents were subtracted online using the P/4 protocol (46) or offline using anegative pulse from −100 to −120 mV (F290A mutants). For ionic recordingsthe external solution contained, in mM, 5 KOH, 110 NMDG, 10 Hepes, and 2Ca(OH)2, and the internal solution contained, in mM, 115 KOH, 10 Hepes,and 2 EDTA. For gating current recordings, KOH was replaced by NMDG. Forblocking experiments, 5 mM NMDG or KOH was replaced by 5 mM 4-aminopyridine (4-AP), and the command potential was held at 0 mV for5 s before recordings. All solutions were adjusted to pH 7.1 with methanesulfonic acid (Mes). Conductance (G) was calculated from the steady statecurrents (I) via G = I/(V − Vrev), where Vrev is the reversal potential. Con-ductance–voltage relationships (GV) were fitted to a Boltzmann relation ofthe form G/Gmax = (1 + exp(−zF(V − V1/2)/RT))

−1. The total gating charge (Q)was calculated from the gating currents by integration, and charge–voltagerelationships (QV) were fitted to a three-state Boltzmann relation of theform Q/Qmax = A1(1 + f−1 + g)−1 + A2(1 + f(1 + g))−1, with f = 1 + exp(−z1F(V –

V1,1/2)/RT) and g = 1 + exp(−z2F(V – V2,1/2)/RT). Because the gating charge wasnormalized to its maximal value, A2 equals 1.

Voltage Clamp Fluorometry. For two-color VCF, oocytes were incubated in 5 μMTMR-maleimide in depolarizing labeling solution [in mM, 115 KOH, 10Hepes, and 2 Ca(OH)2, pH 7.1, with Mes] for 20 min at room temperature,before recordings. The oocytes were washed three times in labeling solutionto remove excess dye. Fluorescence intensities were recorded with a pho-todiode detection system (Photomax 200; Dagan) using ex-545/25, dc-570,em-605/70 (TMR) or ex-377/50, dc-409, and em-470/40 (Anap) filter sets.Fluorescence–voltage (FV) relationships were fitted to a single or three-stateBoltzmann relation of the form ΔF/ΔFmax = (1 + exp(−zF(V − V1/2)/RT))

−1 andΔF/ΔFmax = A1(1 + f−1 + g)−1 + A2(1 + f(1 + g))−1, respectively, with f = 1 +exp(−z1F(V – V1,1/2)/RT) and g = 1 + exp(−z2F(V – V2,1/2)/RT). Bleaching effectsduring step protocols were accounted for by scaling fluorescence tracesusing the first sweep as reference. See Kalstrup and Blunck (23) for visualdemonstration of two-color VCF.

ACKNOWLEDGMENTS. We thank Mireille Marsolais for technical assistancein mutagenesis. pAnap was a kind gift of Peter Schultz (Scripps ResearchInstitute). This work was funded by the Canadian Institutes for HealthResearch Grants MOP-102689 and MOP-136894 and the Natural Sciences andEngineering Research Council Grant RGPIN-2017-06871 (to R.B.). R.B. is aSenior Research Fellow of the Fonds de Recherche de Québec–Santé.

Fig. 7. Stylized model for cytosolic gating machinery. (Upper) The spotsindicate the positions from which we obtained fluorescence data. (Lower)Voltage sensors switch between resting (R), activated (A), and relaxed (X)state, indicated by orange letters; the pore assumes the closed (C), open (O),and inactivated (I) state, indicated by blue letters. The channel undergoestransitions A–D. Here the spots indicate at which position the respectivetransition has been detected. Voltage sensor activation was detected with atemporal delay toward the C terminus of the S4–S5 linker, which is reflectedin the spot color for this transition (blue before orange). The model is shownfor (Left) wild type and (Right) the W434F mutant.

E6758 | www.pnas.org/cgi/doi/10.1073/pnas.1719105115 Kalstrup and Blunck

Dow

nloa

ded

by g

uest

on

Nov

embe

r 25

, 202

0

1. Wulff H, Castle NA, Pardo LA (2009) Voltage-gated potassium channels as therapeutictargets. Nat Rev Drug Discov 8:982–1001.

2. Long SB, Campbell EB, Mackinnon R (2005) Crystal structure of a mammalian voltage-dependent Shaker family K+ channel. Science 309:897–903.

3. Long SB, Tao X, Campbell EB, MacKinnon R (2007) Atomic structure of a voltage-dependent K+ channel in a lipid membrane-like environment. Nature 450:376–382.

4. Bezanilla F (2008) How membrane proteins sense voltage. Nat Rev Mol Cell Biol 9:323–332.5. Blunck R, Batulan Z (2012) Mechanism of electromechanical coupling in voltage-gated

potassium channels. Front Pharmacol 3:166.6. Labro AJ, et al. (2008) Kv channel gating requires a compatible S4-S5 linker and bottom

part of S6, constrained by non-interacting residues. J Gen Physiol 132:667–680.7. Labro AJ, Raes AL, Snyders DJ (2005) Coupling of voltage sensing to channel opening

reflects intrasubunit interactions in kv channels. J Gen Physiol 125:71–80.8. Haddad GA, Blunck R (2011) Mode shift of the voltage sensors in Shaker K+ channels

is caused by energetic coupling to the pore domain. J Gen Physiol 137:455–472.9. Wall-Lacelle S, Hossain MI, Sauvé R, Blunck R, Parent L (2011) Double mutant cycle

analysis identified a critical leucine residue in the IIS4S5 linker for the activation of theCa(V)2.3 calcium channel. J Biol Chem 286:27197–27205.

10. Muroi Y, Arcisio-Miranda M, Chowdhury S, Chanda B (2010) Molecular determinantsof coupling between the domain III voltage sensor and pore of a sodium channel. NatStruct Mol Biol 17:230–237.

11. Lu Z, Klem AM, Ramu Y (2002) Coupling between voltage sensors and activation gatein voltage-gated K+ channels. J Gen Physiol 120:663–676.

12. Lu Z, Klem AM, Ramu Y (2001) Ion conduction pore is conserved among potassiumchannels. Nature 413:809–813.

13. 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.

14. Chowdhury S, Chanda B (2012) Perspectives on: Conformational coupling in ionchannels: Thermodynamics of electromechanical coupling in voltage-gated ionchannels. J Gen Physiol 140:613–623.

15. Pathak M, Kurtz L, Tombola F, Isacoff E (2005) The cooperative voltage sensor motionthat gates a potassium channel. J Gen Physiol 125:57–69.

16. Mannuzzu LM, Isacoff EY (2000) Independence and cooperativity in rearrangementsof a potassium channel voltage sensor revealed by single subunit fluorescence. J GenPhysiol 115:257–268.

17. Kalstrup T, Blunck R (2013) Dynamics of internal pore opening in K(V) channelsprobed by a fluorescent unnatural amino acid. Proc Natl Acad Sci USA 110:8272–8277.

18. Long SB, Campbell EB, Mackinnon R (2005) Voltage sensor of Kv1.2: Structural basis ofelectromechanical coupling. Science 309:903–908.

19. Batulan Z, Haddad GA, Blunck R (2010) An intersubunit interaction between S4-S5 linker and S6 is responsible for the slow off-gating component in Shaker K+channels. J Biol Chem 285:14005–14019.

20. Faure É, Starek G, McGuire H, Bernèche S, Blunck R (2012) A limited 4 Å radial dis-placement of the S4-S5 linker is sufficient for internal gate closing in Kv channels.J Biol Chem 287:40091–40098.

21. Lee HS, Guo J, Lemke EA, Dimla RD, Schultz PG (2009) Genetic incorporation of asmall, environmentally sensitive, fluorescent probe into proteins in Saccharomycescerevisiae. J Am Chem Soc 131:12921–12923.

22. Chatterjee A, Guo J, Lee HS, Schultz PG (2013) A genetically encoded fluorescentprobe in mammalian cells. J Am Chem Soc 135:12540–12543.

23. Kalstrup T, Blunck R (2017) Voltage-clamp fluorometry in Xenopus oocytes usingfluorescent unnatural amino acids. J Vis Exp 123:e55598.

24. Zhao J, Blunck R (2016) The isolated voltage sensing domain of the Shaker potassiumchannel forms a voltage-gated cation channel. eLife 5:e18130.

25. Stefani E, Toro L, Perozo E, Bezanilla F (1994) Gating of Shaker K+ channels: I. Ionicand gating currents. Biophys J 66:996–1010.

26. Swenson RP, Jr, Armstrong CM (1981) K+ channels close more slowly in the presenceof external K+ and Rb+. Nature 291:427–429.

27. Hurst RS, Roux MJ, Toro L, Stefani E (1997) External barium influences the gatingcharge movement of Shaker potassium channels. Biophys J 72:77–84.

28. Claydon TW, Vaid M, Rezazadeh S, Kehl SJ, Fedida D (2007) 4-aminopyridine preventsthe conformational changes associated with p/c-type inactivation in shaker channels.J Pharmacol Exp Ther 320:162–172.

29. Yang Y, Yan Y, Sigworth FJ (1997) How does the W434F mutation block current inShaker potassium channels? J Gen Physiol 109:779–789.

30. Bezanilla F, Perozo E, Stefani E (1994) Gating of Shaker K+ channels: II. The compo-nents of gating currents and a model of channel activation. Biophys J 66:1011–1021.

31. Lacroix JJ, Bezanilla F (2011) Control of a final gating charge transition by a hydro-phobic residue in the S2 segment of a K+ channel voltage sensor. Proc Natl Acad SciUSA 108:6444–6449.

32. Tao X, Lee A, Limapichat W, Dougherty DA, MacKinnon R (2010) A gating chargetransfer center in voltage sensors. Science 328:67–73.

33. Parker I, Ivorra I (1990) A slowly inactivating potassium current in native oocytes ofXenopus laevis. Proc R Soc Lond B Biol Sci 238:369–381.

34. Lu L, Montrose-Rafizadeh C, Hwang TC, Guggino WB (1990) A delayed rectifier po-tassium current in Xenopus oocytes. Biophys J 57:1117–1123.

35. Bernèche S, Roux B (2005) A gate in the selectivity filter of potassium channels.Structure 13:591–600.

36. Cuello LG, et al. (2010) Structural basis for the coupling between activation and in-activation gates in K(+) channels. Nature 466:272–275.

37. Barghaan J, Bähring R (2009) Dynamic coupling of voltage sensor and gate involved inclosed-state inactivation of kv4.2 channels. J Gen Physiol 133:205–224.

38. Pioletti M, Findeisen F, Hura GL, Minor DL, Jr (2006) Three-dimensional structure ofthe KChIP1-Kv4.3 T1 complex reveals a cross-shaped octamer. Nat Struct Mol Biol 13:987–995.

39. Whicher JR, MacKinnon R (2016) Structure of the voltage-gated K+ channelEag1 reveals an alternative voltage sensing mechanism. Science 353:664–669.

40. WangW,MacKinnon R (2017) Cryo-EM structure of the open human ether-à-go-go-relatedK+ channel hERG. Cell 169:422–430.e10.

41. Li-Smerin Y, Hackos DH, Swartz KJ (2000) A localized interaction surface for voltage-sensing domains on the pore domain of a K+ channel. Neuron 25:411–423.

42. Fernández-Mariño AI, Harpole TJ, Oelstrom K, Delemotte L, Chanda B (2018) Gatinginteraction maps reveal a noncanonical electromechanical coupling mode in theShaker K+ channel. Nat Struct Mol Biol 25:320–326.

43. Tang CY, Bezanilla F, Papazian DM (2000) Extracellular Mg(2+) modulates slow gatingtransitions and the opening of Drosophila ether-à-go-go potassium channels. J GenPhysiol 115:319–338.

44. Goodchild SJ, Fedida D (2014) Gating charge movement precedes ionic current acti-vation in hERG channels. Channels (Austin) 8:84–89.

45. Piper DR, Varghese A, Sanguinetti MC, Tristani-Firouzi M (2003) Gating currents as-sociated with intramembrane charge displacement in HERG potassium channels. ProcNatl Acad Sci USA 100:10534–10539.

46. Bezanilla F, Armstrong CM (1977) Inactivation of the sodium channel. I. Sodiumcurrent experiments. J Gen Physiol 70:549–566.

Kalstrup and Blunck PNAS | vol. 115 | no. 29 | E6759

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 25

, 202

0