Protein Translocation through the Anthrax ToxinTransmembrane Pore is Driven by a Proton Gradient

Bryan A. Krantz1, Alan Finkelstein2 and R. John Collier1*

1Department of Microbiologyand Molecular GeneticsHarvard Medical School200 Longwood Ave, BostonMA 02115, USA2Department of Physiology andBiophysics, Albert EinsteinCollege of Medicine,1300 Morris Park Ave,Bronx, NY 10461, USA

Protective antigen (PA) from anthrax toxin assembles into a homoheptameron cell surfaces and forms complexes with the enzymatic components:lethal factor (LF) and edema factor (EF). Endocytic vesicles containing thesecomplexes are acidified, causing the heptamer to transform into atransmembrane pore that chaperones the passage of unfolded LF and EFinto the cytosol. We show in planar lipid bilayers that a physiologicallyrelevant proton gradient (DpH, where the endosome is acidified relative tothe cytosol) is a potent driving force for translocation of LF, EF and the LFamino-terminal domain (LFN) through the PA63 pore. DpH-driventranslocation occurs even under a negligible membrane potential. Wefound that acidic endosomal conditions known to destabilize LFN correlatewith an increased translocation rate. The hydrophobic heptad of lumen-facing Phe427 residues in PA (or f clamp) drives translocationsynergistically under a DpH. We propose that a Brownian ratchetmechanism proposed earlier for the f clamp is cooperatively linked to aprotonation-state, DpH-driven ratchet acting trans to the f-clamp site. In asense, the channel functions as a proton/protein symporter.

q 2005 Elsevier Ltd. All rights reserved.

Keywords: translocation; proton gradient; translocase; protein unfolding;planar bilayers*Corresponding author

Introduction

Transport across membranes of ions, small polarmolecules, and macromolecules, such as proteins, isoften facilitated by specific, integral membraneproteins. Membranes are particularly impermeableto soluble proteins, because of their size and theirglobular, amphipathic nature. To remain folded andsoluble, proteins bury hydrophobic surface whileexposing a charged, polar surface to water, but thispolar surface is incompatible with the apolarmembrane interior. Many bacterial toxins (e.g.anthrax, diphtheria, botulinum, and tetanus toxins)take advantage of the low pH conditions in theendosome of mammalian cells to transform theirenzymatic components into a partially unfolded,molten globule (MG) form that can circumvent thebarrier imposed by the membrane more readily.1–3

These toxins chaperone the delivery of their

unfolded enzymatic cargos using separate translo-case domains or proteins that form transmembranepores in the endosomal membrane.4–7 In general,the role of these pores during translocation isunclear; the driving force and underlying mechan-isms for the concerted unfolding and translocationreactions remain only vaguely defined.

The toxin from Bacillus anthracis is an ideal modelsystem for the study of protein translocation acrossmembranes, as it is composed of three separateproteins, a translocase and two substrates, that maybe produced recombinantly in soluble form andstudied independently.8 Protective antigen (PA;83 kDa), the translocase component, transports thetwo enzymatic components, lethal factor (LF;90 kDa) and edema factor (EF; 89 kDa) to thecytosol. LF is a protease that cleaves mitogen-activated protein kinase kinases; EF is a Ca2C andcalmodulin-dependent adenylate cyclase. Toxinaction is initiated by a self-assembly process. First,PA binds to a cell-surface receptor, where it isactivated by a furin-family protease, and a smallfragment dissociates. The remaining 63 kDa,receptor-bound portion (PA63) self-assembles intoa ring-shaped homoheptamer, called prepore.Prepore may then form complexes with up to

0022-2836/$ - see front matter q 2005 Elsevier Ltd. All rights reserved.

Abbreviations used: MG, molten globule; PA,protective antigen;LF, lethal factor; LFN,LFamino-terminaldomain; EF, edema factor; EFN, EF amino-terminaldomain.E-mail address of the corresponding author:

jcollier@hms.harvard.edu

doi:10.1016/j.jmb.2005.11.030 J. Mol. Biol. (2006) 355, 968–979

three molecules of EF and/or LF. These complexesare endocytosed and delivered to an acidiccompartment, where prepore undergoes an acidicpH-dependent conformational rearrangement toform a cation-selective, ion-conducting channel.9

The PA63 channel, or pore, spans the membrane asan extended, 14-stranded b barrel,10,11 and serves asthe conduit for LF and EF into the cytosol.

A model of the PA63 14-strand b barrel revealsthat its lumen is w15 A wide and is able toaccommodate secondary structure only as large asan a helix.3 Translocation through the lumen thusrequires the substrates to unfold.3,12 Thermo-dynamic analysis of the conformational stabilitiesof the LF and EF homologous w250 residue amino-terminal domains (LFN and EFN) indicates thatacidic conditions encountered in the endosome aresufficient to destabilize the native structures of theseproteins.3 Also, a positive membrane potential(CDj) can drive the acid-destabilized LFN throughthe PA63 pore, where the amino terminus precedesthe carboxy terminus into the pore.4,13 However,in vitro and in vivo studies have shown thattranslocation is more complicated than acid-induced unfolding, as followed by electrophoresisthrough a passive pore. The translocase channelcontains a required, solvent-exposed, lumen-facingheptad of Phe427 residues, or f clamp, that canfacilitate unfolding of the MG by interactingdirectly with the substrate and actively chaperoninghydrophobic sequences.5 The f clamp graspshydrophobic sequences as they unfurl from theMG substrate to direct the translocating chainthrough the channel. But how does PA63, a cation-selective channel, accommodate the LFN anionicside-chains, which outnumber its cationic counter-parts? Clearly, the glutamate and aspartate residuesmust be at least sufficiently protonated in thechannel to impart a net positive charge on LFN,otherwise we would have the thermodynamicallyimpossible situation of a negatively charged LFNbeing driven across the membrane by aCDjwithout a compensatory dissipative process.Beyond this, how does the acid-induced destabili-zation of the substrate couple with translocation?Can the proton gradient (DpH) established acrossthe endosomal membrane provide enough drivingforce to support protein translocation? Here, weaddress these issues.

Results and Discussion

Symmetric pH

pH-dependence of translocation

Previous planar lipid bilayer studies demon-strated that LFN could be translocated throughPA63 channels under a Dj of C50 mV at symmetricpH 5.5, but not pH 6.6.4 (The membrane potential isdefined as DjZjcis–jtrans, where jtransh0 mV.)This observation is consistent with the acidic

pH-induced conformational destabilizationdescribed for LFN and EFN,

3 implying that partialunfolding to an MG form is kinetically linked tomore rapid translocation. Extending these studies,we examined the rates of translocation at differentsymmetric pH values. In planar bilayer exper-iments, LFN was added to the cis compartment,where it bound to PA63 channels and blocked ionconductance through the channel. Following per-fusion of the cis compartment to remove unboundLFN, the Dj was stepped to a higher positivevoltage, and the rate of translocation was monitoredby the increase in channel conductance as LFNpassed through the channel (Figure 1(a)). Weobserved that when the cis and trans pH valueswere raised symmetrically from acidic pH to moreneutral pH, the rate of LFN translocation decreasedcorrespondingly (Figure 1(b) and Table 1). Forexample, at C50 mV the translocation half-timewas w5 s at symmetric pH 5.5 and 36 s atsymmetric pH 6.0; at symmetric pH 6.6, transloca-tion was too slow to measure, even at C70 mV.

pH-dependence of LFN stability and rate oftranslocation correlate

We recorded the kinetics of LFN translocationthrough PA63 channels at a Dj of C60 mV andpH values from w5 to 6.5 (Figure 1(b)) andmeasured the translocation half time, t1⁄2 , as a simpleindicator of the complex series of rate constantsdescribing the entire S-shaped translocationkinetics. For each pH value tested, an approximatetranslocation activation energy !DG‡

transl" wascalculated from the half-times as DG‡

translZRT ln tK11=2.

This quantity was plotted against the pH-depen-dence of the free energy difference (DGNI) observedfor LFN between its native (N) andMG intermediate(I) conformations, using:

DGNI!pH"ZRTln!KpH8NI #!1C10pKNKpH"=!1C10pKIKpH"$n"

the known equilibrium stability between the Nand I states, K

pH8NI (at a reference pH of 8), and

the previously defined proton dissociation con-stants, pK, for the n titratable groups responsible fordestabilizing LFN upon protonation.3 The pH-dependent changes in DGNI and DG‡

transl were wellcorrelated, with a slope of unity (1.0 (G0.1))(Figure 1(c)). Thus, translocation kinetics correlatedstrongly with the global stability of the translocat-ing substrate, LFN.Because the correlation of the translocation

activation energy (deduced from the translocationt1⁄2) and the pH-dependence of the conformationalstability covers two orders of magnitude(Figure 1(c)), we infer that the pH-induced desta-bilization of the native state and subsequentpopulation of an MG state limits translocation. Forit is unlikely that the pK and the number oftitratable groups responsible for conversion of LFNto the MG state exactly match other distincttitratable groups in PA63 or LFN that solely affect

DpH-driven Anthrax Toxin Translocation 969

translocation. Moreover, when fold-stabilizingco-solvents were added to the cis compartment ofthe bilayer, the translocation rate slowed (Supple-mentary Data, Figure S1).

pH gradient

Proton gradients drive LFN translocation throughPA63 channels

We found that translocation of LFN (Figure 2(a)),as well as full-length LF (Figure 3) and EF(Supplementary Data, Figure S2), through PA63

channels was markedly stimulated by a protongradient (DpH); that is, when the pH of the transside (pHtrans) of the lipid bilayer was greater thanthat of the cis side (pHcis). The proton gradient isdefined as:

DpHZpHtransKpHcis

The following experiments illustrate this:

(1) At a low Dj of C20 mV, no translocationoccurred at pH 5.5, if the DpHZ0, whereasthere was rapid translocation when the transpH was raised to 6.2 (Figure 2(a)).

(2) Translocation was inhibited (or slowed) whenpHcis was greater than pHtrans (i.e. DpH wasnegative; Figure 4).

(3) Other DpH values, at pHcis higher than 5.5, andeven above neutrality, were effective in promot-ing translocation, so long as pHtrans was greaterthan pHcis. Thus, raising pHtrans from 8.5 to 8.8still increased the translocation rate abouttwofold (Figure 5).

Figure 1. Symmetric pH-dependence of LFN trans-location kinetics through PA63 channels. (a) In a typicalDj-step protein translocation experiment, the cis andtrans compartments contained 100 mMKCl, 1 mM EDTA,and either 5 mM potassium succinate or universal bilayerbuffer (UBB: 10 mM phosphate, 10 mM Mes, and 10 mMoxalic acid) at pH 5.5 at room temperature. Preporeheptameric PA63 (final concentration w2 pM) was firstadded to the cis compartment at DjZC20 mV (wherejtransh0) until channel formation reached steady state(indicated by the stabilization of the current). LFN (finalconcentration w3 nM) was then added to the ciscompartment until the current was blocked; the ciscompartment was then perfused of unbound LFN. Attime zero, Djwas stepped to the final voltage ofC50 mV(Djfinal), and LFN translocation kinetics through the PA63

channel were indicated by the increase in conductanceversus time. (b) Kinetic transients as in (a) for LFNtranslocation (DjfinalZC60 mV) when the cis and transcompartments were bathed symmetrically in UBB at theindicated pH values. (c) Correlation of the symmetric

pH-dependence of LFN translocation kinetics from (b) at aDjZC60 mV (given as an activation free energy of RTln tK1

1=2) to the pH-dependence of the conformationalstability of LFN, or DGNI(pH), using previously deter-mined stability values.3 The relation is fitted to a straightline, which has a slope of 1.0(G0.1), and a correlationR-value of 0.95.

Table 1.Half-times of LFN translocation at symmetric pH

Djfinal

(mV)

t1⁄2 (s)

pH 5.2 pH 5.5 pH 5.7 pH 6.0 pH 6.2

25 7530 2235 1640 7 13 3545 6 8 2050 5 5 10 3660 3 4 15 2870 1.5 3 9 1875 12

Under the indicated symmetric pH conditions, Dj was steppedfrom C20 mV to the indicated Djfinal to initiate LFN transloca-tion. The translocation half-times (t1⁄2 ) were obtained fromrecords such as those in Figure 1(b).

970 DpH-driven Anthrax Toxin Translocation

It is possible, but unlikely, that the rise inconductance induced by raising pHtrans (e.g.Figure 2(a)) resulted from LFN dissociating intothe cis compartment, rather than translocating to thetrans compartment. As a control, we added LFN-biotin (LFNwith biotin attached near its C terminus)

and streptavidin to the cis compartment. Whenstreptavidin binds LFN-biotin, the streptavidinmoiety does not prevent LFN from entering andblocking the PA63 channel,13 but it does preventtranslocation.4 As expected, raising pHtrans underthese conditions did not cause the conductanceto rise, but subsequent addition of a reducingagent to the cis solution, which broke the disulfidebond linking biotin (with its attached streptavidin)to LFN, led to the expected rise in conductance,reflecting LFN translocation under the DpHcondition (Figure 2(b)).

Translocation of whole LF and EF is driven by DpH

Under acidic conditions known to be destabiliz-ing for LFN and EFN (symmetric pH 5.5) and amoderate CDj of 50 mV, little translocation ofwhole LF or EF occurred; however, introduction of aDpH resulted in significant translocation(Figure 3(b) and Supplementary Data, Figure S2).Even at a low Dj of 20 mV, DpH promotedtranslocation of LF in a DpH-dependent manner;i.e. larger CDpH caused more rapid translocation(Figure 3(a)). The conditions of low Dj and a DpH,as in Figure 3(a), are particularly relevant to theconditions existing for LF (and EF) bound to a PA63

pore in an acidic endosome, and hence theseexperiments have particular biological relevance.

500pA

10 s

pH = +0.35

pH = +0.7

cis TCEP+30 mV

+40 mV

20 s 20 s

50pA 100

pA

(b)

(a)

Figure 2. The effect of raising pHtrans on LFNtranslocation. (a) In an example of an LFN translocationexperiment driven by a DpH jump, the cis and transcompartments were initially bathed in symmetric pH 5.5(5 mMpotassium succinate, 100 mMKCl, 1 mMEDTA) ata Dj of C20 mV. PA63 channel conductance was 97%blocked by the addition of LFN (final concentration3.3 nM) (as in Figure 1(a)). The record begins afterperfusion of the cis compartment under a constant Djof C20 mV. At the first arrow, pHtrans was raised to 5.85(with 3.3 mM potassium phosphate); a slow rate of rise inconductance was seen, reflecting the translocation of LFN.At the second arrow, pHtrans was raised to 6.2 (with6.6 mM potassium phosphate), causing a much greaterincrease in the rate of LFN translocation. The t1⁄2 ofw6 s atpH 6.2 likely underestimates the rate of LFN translocation,as the kinetics are limited by the buffer mixing dead-time.(b) In symmetric pH 6.7 buffer (10 mM potassiumdimethylglutarate (DMG), 100 mM KCl, 1 mM EDTA),PA63 channels were formed (as in Figure 1(a)). After asteady-state conductance was reached, LFN-biotin (LFNwith biotin attached near its C terminus4) was added tothe cis compartment (w3 nM), blocking channel con-ductance 98%. Streptavidin was then added to the ciscompartment (10 mg/ml). pHtrans was raised from 6.7 to8.0 (with 5 mM Ches), and the cis compartment wasperfused. After stepping the Dj to C40 mV, there was avery slow rise in conductance (w1.5-fold in 80 s). The Djwas then stepped back toC30 mV. The record begins herewith no discernible rise in the conductance, despite a Djof C30 mV and a DpH of 1.3. However, once Tris(2-carboxyethyl)-phosphine (TCEP) (which reduces thedisulfide bond linking biotin to LFN and thereby freesLFN of liganded streptavidin) was added to the ciscompartment (2.5 mM), there was a continuous rise inconductance, reflecting LFN translocation through thechannel. (The break in the record is 27 s; note the changein current scale after the break.)

Figure 3. Whole LF translocation is driven by a DpH.For whole LF translocation experiments, PA63 channelswere formed as in Figures 1(a) and 2(a), and 25 nM LFwas added to the cis compartment to block conductanceto O 95%. (a) LF translocation was recorded at a Dj ofC20 mV and either a one-unit or two-unit DpH, wherepHcis was 5.5 in both cases. (b) LF translocation wasrecorded both under a Dj alone (C50 mV) and under thesame Dj plus a one-unit DpH (pHcisZ5.5, pHtransZ6.5).Note that there was virtually no translocation of whole LFwith a DjZC50 mV, in the absence of a pH gradient. Thecurrent scales are normalized to overlay the records.

DpH-driven Anthrax Toxin Translocation 971

Dj-driven versus DpH-driven translocation

In comparing the Dj dependence of LFNtranslocation in the presence and in the absence ofa one-unit DpH (pHtransZ6.5, pHcisZ5.5), theDpH was better able to promote translocation atlower voltages than at higher voltages (Figure 6(a)).Moreover, extrapolating the translocation t1⁄2 back to0 mV, we found that this DpH would accelerate

translocation w100-fold relative to the rateexpected in the absence of Dj. The DpH condition(pHtransZ6.5, pHcisZ5.5) is offset 60 mV or morefrom the symmetric pH 5.5 condition (Figure 6(a)).The offset is of the order of the value expected for achemical gradient.

Fold-stabilizing gradients do not drive LFN

translocation

At first glance, it would appear that the re-foldingof LFN that is promoted by the higher trans pH, asLFN emerges into that solution, could be acting as adriving force on translocation, trapping the re-folded protein on the trans side and thereby, inaddition to the transmembrane voltage, addinganother vectorial component to the process. Thisappears to be an unlikely explanation of the transpH effect, however, because a refolding reagentsuch as 1 M glucose did not stimulate translocationwhen added to the trans side (Supplementary Data,Figure S1).

Glucose and other “protecting osmolytes” havebeen shown to stabilize proteins, and cause even“random coil” proteins to collapse into morecompact structures.14 Equilibrium denaturanttitrations of a fluorescently labeled form of LFN!LF%N" that is capable of fluorescence resonanceenergy transfer when folded confirmed that 1 Mglucose stabilized LFN (Supplementary Data,Figure S1). At pH 7.5, the N state was stabilizedw1 kcal molK1 relative to the I state (or MGconformation); the I state was stabilized w0.7 kcalmolK1 over the more expanded, partially unfoldedform (referred to as the J state). Although theaddition of 1 M glucose to the cis compartmentslowed translocation (about threefold) as expected,the addition of 1 M glucose to the trans compart-ment failed to stimulate translocation and insteadslowed it about twofold (Supplementary Data,Figure S1). For the secondary and tertiary struc-ture-stabilizing co-solvent, 2,2,2-trifluoroethanol,15

we found that cis-side addition at 5% (w/v) slowedthe rate of translocation about threefold (at sym-metric pH 5.5,C50 mV), whereas trans-sideaddition had no effect on the translocation rate(Supplementary Data, Figure S1).

The slowing of LFN translocation by non-electrolytes in the cis compartment that stabilize itsnative structure is consistent with the correlation ofthe rateofLFN translocationwith thepH-dependenceof its conformational stability (Figure 1(c)). However,since the addition to the trans side of non-electrolytesthat increase the stability of LFN at either thesecondary or tertiary structural level did notstimulate translocation, the stimulatory effect ofraising the trans pH cannot be attributed to itspromoting the re-folding of LFN.

“Unfolding” and “translocation” barriers

Although we expect that a protein must unfoldand shed its residual tertiary structure in order to

(a)

0 10 20

0

1

Time (s)

Frac

tion

Tran

sloca

ted

pH = 0

-0.15

-0.25

+0.15+0.3+0.55+0.7

(b)

-0.5 0.0 0.5 1.0 1.5-2

-1

0

1

2

Log (

t½ /

t½˚)

pH (pHtrans - pHcis )

Figure 4. Translocation is controlled by the magnitudeand sign of the DpH. (a) LFN translocation driven at theindicated DpH values (where DpHZpHtransKpHcis).pHcis was held at a constant value of 5.45 for eachtransient. The cis and trans compartments were initiallybathed in UBB (pH 5.45), and the indicated DpH wasestablished by either raising or lowering pHtrans with 2 MKOH or 1 M HCl. PA63 channels were formed at a Dj ofC20 mV and blocked by LFN (20 nM) at a Dj of C1 mV.The cis compartment was then perfused, and at time zero,Dj was stepped to C45 mV to initiate translocation.(b) Using the t1⁄2 values of these and other translocationtransients (where pHcis was held at a constant value of5.45), the logarithm of the ratio of the t1⁄2 in the presenceand in the absence of a DpH (t1⁄2 and t1⁄28, respectively), areplotted against the final DpH condition. The ratiocomparisons were made at four different constant Djvalues indicated by symbols as C40 (-),C45 (,),C50(C), and C60 mV (B).

972 DpH-driven Anthrax Toxin Translocation

pass through the narrow channel, we do not knowthe timescales of the pH-induced unfolding andrefolding reactions on the surface of PA63. We havecharacterized only the translocation kinetics usingthe t1⁄2 and not the more elemental kinetic rateconstants. To simplify our discussion, however, wewill define energy barriers, acknowledging that the

observed translocation kinetics are S-shaped, multi-exponential, and complex. The barrier diagramspresented in Figure 8 are intended to represent theenergy surface that unfolded polypeptide encoun-ters as it unravels from the MG state andsubsequently travels through the channel. Weexpect, of course, that multiple kinetic steps areimplied by the periodicity of hydrophobicity,hydrophilicity, and charge over the length of theLFN sequence (Figure 9(a)).Figure 8 depicts a two-barrier, one-well scheme

proposed previously,5 in which translocation maybe limited by the equilibrium population of arequired, partially unfolded species (akin to theMG state observed in solution). Unfolding exposessequence dense in hydrophobic and cationic sitesthat threads into and forms the observed inter-mediate complex with the f clamp.5 The f clampdefines an energy well that partitions the “unfold-ing reaction” in the cis-side “cap” of the PA63

channel from a trans-side “translocation barrier” inthe extended b barrel. At low CDj or DpH%0, thetranslocation barrier is the major barrier. But in thissequential reaction, translocation requires unfoldedprotein sequence to be in the f clamp and therebyis modulated thermodynamically by the stabilityof the protein substrate. At higher voltage andlarger DpH, translocation is kinetically limited(Figures 4(b) and 6). Thus, the translocation barrieris lowered such that a separate barrier kineticallycontrols translocation: this barrier is both largely Djand DpH-independent, and is likely related to theprotein unfolding reaction in the cis-side cap. Insummary, two separate steps (or barriers) areneeded to describe the observed Dj andDpH dependence of the kinetics. We have termedthem: a cis-side unfolding barrier that is affected bydenaturing, acidic pH conditions or stabilizingco-solvents; and a trans-side translocation barrier,modulated by either Dj or DpH.

PA63 f-clamp mutants were less stimulated bya pH gradient

In an effort to identify residues lining theheptameric PA63 channel for their impact on

100pA

10 s

pH = +1.8

200pA

10 s

pH = +2.1

Figure 5. LFN translocation is stimulated at pHtransO8. In symmetric pH 6.7 solutions (10 mM DMG), PA63 channelswere formed and conductance blocked (98%) by LFN (6.6 nM) as in Figure 1(a). The cis compartment was then perfusedof unbound LFN at a Dj of C20 mV. Following perfusion, pHtrans was raised to 8.2 with 10 mM Ches buffer, and the Djwas stepped to C25 mV. The plotted record begins at this point with a DpH of C1.5 and a slow rate of rise ofconductance. Increasing pHtrans to 8.5 (first arrow, DpH of C1.8) with 11.7 mM Ches, initiated an increase in the rate ofconductance rise (i.e. LFN translocation rate). Moreover, raising pHtrans to 8.8 (second arrow, DpH ofC2.1) with 14.2 mMChes caused an additional increase in the translocation rate. (The break in the record is 70 s; note the change in currentscale after the break.)

0 40 80 120 1601

10

102

103

(mV)

(b)

4.955.20 5.45

5.60

5.70

5.856.00

6.206.40

6.60

(a)

0 25 50 75 100 1251

10

10

t½ (s

)

(mV)

2

t½ (s

)

Figure 6. The Dj dependence of LFN translocationkinetics. (a) Dj dependence of LFN translocation kineticsin the presence (C) and in the absence (B) of aDpHZC1(pHcis 5.5). (b) Dj dependence of the rate of LFNtranslocation (indicated by the t1⁄2 ) at the indicatedsymmetric pH conditions. The indicated slope of the Djdependence of the logarithm of t1⁄2 is equivalentenergetically to the presence of about C3 charges in therate-limiting species within the PA63 channel. Note theplateauing of t

1⁄2 at large Dj.

DpH-driven Anthrax Toxin Translocation 973

translocation, we showed earlier that the f clamp, asite composed of a heptad of Phe427 residues, isrequired for translocation in cells and catalyzestranslocation in planar lipid bilayers.5 Here, wereport that a DpH (pHcisZ5.5, pHtransZ6.5) was lessable to accelerate translocation through F427A PA63

channels than through wild-type (WT) channels(Figure 7). By contrast, when F427Y PA63 (afunctional f-clamp variant) was tested, theDpH stimulated the rate of translocation w20-fold(Table 2). A DpH was more effective in stimulatingtranslocation when large aromatic and aliphaticresidues (Phe, Trp, Leu) were present at position 427than when smaller aliphatic or hydrophilic substi-tutions (Ile, Ala, Ser, Asp) were present at thisposition (Table 2)†.

How the f clamp can maintain a DpH

Why is aCDpH less effective in driving transloca-tion through PA63 channels with a mutated f clamp(F427A) than through PA63 channels with a WT fclamp, given that phenylalanine side-chains cannotbe titrated? F427A PA63 channels are defective instably binding the LFN amino terminus, whichflickers in and out of the channel, thereby allowingcations to pass through the channel,5 and wepresume that protons can likewise leak past themutated F427A site. Thus, the WT f clampfunctions as a binding site for hydrophobic proteinsequence, and may effectively maintain the protongradient by acting as a hydrophobic seal. The

higher concentration of proton in the cis-side cap,which is required to acid-destabilize LFN, is keptseparate from the lower concentration of protontrans to the f clamp, which facilitates translocation(Figure 8).

What is titrated during DpH-driven translocation

Assuming that the f clamp maintains the protongradient, we expect that the residues titrated by thehigher pHtrans are trans relative to the f-clamp site.The residues that line the extended b barrel of thePA63 channel satisfy this criterion,10 and we hadbegun testing their involvement in DpH-driventranslocation. (Considering only charged residuesand the 7-fold symmetry of the channel, 42 Asp orGlu and 14 His residues line the b barrel.) Whenpairs of His or Glu residues near the trans-mostopening of the b barrel were doubly substitutedwith Thr, these mutant PA63 channels failed todisrupt DpH-driven translocation (unpublishedresults). Although we have not mutated all of thecharged, lumen-facing residues lining the b barrelof the PA63 channel, we tentatively conclude thatthe titrated residues most affecting DpH-driventranslocation reside not in the channel, but rather inthe translocating substrate, LFN‡.

1 10 100

0

1

Frac

tion

Tran

sloca

ted

Time (s)

Figure 7.A PA63 f-clampmutation is defective in DpH-driven LFN translocation. Kinetic records acquired as inFigures 1(a) and 4(a) for LFN translocation through WTPA63 channels driven by Dj alone (DjZC40 mV, DpHZ0 (symmetric pH 5.5); broken black line) or by Dj andDpH (DjZC40 mV, DpHZC1 (pHcisZ5.5); continuousblack line) compared to translocation through F427A PA63

channels driven by Dj alone (DjZC50 mV, DpHZ0(symmetric pH 5.5); red broken line) or by Dj andDpH (DjZC50 mV, DpHZC1 (pHcisZ5.5); red continu-ous line). See Table 2 for additional data on f clampmutants under a DpH.

Table 2. The DpH rate-enhancement for a panel of fclamp PA63 mutants

PA63

F427X Dj (mV) t1⁄2 (Dj) (s)t1⁄2 (Dj, DpH)

(s)

DpH rate-enhancement

(fold)

WT C40 25 1.5 17L C40 74 2.8 26W C40 51 4.7 11Y C50 88 5 18I C50 41 8 5A C50 32 12 3S C50 83 18 5D C50 120 150 0.8

Translocation t1⁄2 values were measured for LFN driven throughthe PA63 F427X mutant channel by either a Dj alone (atsymmetric pH 5.5) or by both a constant Dj (C40 to C50 mV)and a positive, one-unit DpH (pHcisZ5.5, pHtransZ6.5). TheDpH rate-enhancement is defined by the ratio: t1⁄2 (Dj)/t1⁄2 (Dj,DpH). Each tested PA63 mutant was given equal opportunity tobe stimulated by a DpH; i.e. not be limited by the known DpH-independent process (Figure 4(b)), so that the translocation rateunder a Dj alone had more or less consistent t1⁄2 values (of theorder of w30 to 100 s). Despite this consideration, it is known5

that at lower Dj (less than C30 mV) the DpH rate-enhancementfor WT and F427/(W, Y or L) PA63 channels would be evengreater than that shown here (at C40 to 50 mV); but the mutantsF427/(I, A, S, and D) are so defective that the quantitativecomparison at these lower Dj cannot be reported.

† F427I PA is an exception to this generalization, aspreviously discussed.5

‡We do not mean to imply that further mutagenesis inthe b barrel would not eventually affect DpH-driventranslocation. For example, loss of cation selectivity in thechannel would be expected to impair DpH-driventranslocation by the following model.

974 DpH-driven Anthrax Toxin Translocation

A charge-state ratchet translocation model

Here, we propose a novel charge-state Brownianratchet mechanism for DpH-driven translocationbased on the chemical asymmetry created by aDpH. Considering two degrees of conformationalfreedom in the backbone, a polypeptide chain suchas LFN would have w150 kcal molK1 of Brownianthermal energy available. This large pool of

diffusive thermal energy, by its nature, cannot douseful work, such as driving the translocating chainthrough a transmembrane pore in a unidirectionalmanner. However, a transmembrane chemicalgradient can bias the Brownian fluctuations in thepolypeptide chain toward productive, or rectified,unidirectional translocation. Protonation or deproto-nation of acidic side-chains comprising the translo-cating polypeptide can occur on either the cis or thetrans side of the membrane and decrease or increase,respectively, the energy barriers for entry of thetranslocating polypeptide into the pore. This modelis essentially electrostatic by nature. Given that thePA translocase pore is cation-selective,9 we expectthe LFN anionic side-chains to play a significant rolein the proposed ratchet model.First, consider that the titrated residues are Asp

or Glu (of which there are 54 in LFN; Figure 9(a)).The LFN acidic residues are chosen as they areanionic (when deprotonated) and have low pK near4 in solution. The pK can be elevated several units indesolvated environments, however, such as thenarrow b barrel of the channel or the desolvatedinterior of the MG form.3 Since the rate of LFNtranslocation increases with increasing CDj,4 andLFN has more negatively charged residues thanpositively charged residues, we expect that thenegative charges are neutralized in stretches ofpolypeptide as they traverse the channel. In fact, wehave shown here that the translocating chain mostexposed to the Dj gradient has aC3 chargedependence in the translocation rate (Figure 6(b)),requiring neutralization of anionic charges inalmost any part of the LFN primary sequence,excepting the most N-terminal end, which is likelynot limiting to translocation (Figure 9(a)). Finally,anion-neutralized stretches of translocating poly-peptide would be favored in the cation-selectiveportion of the PA63 channel.In this model, a stretch of anionic polypeptide

sequence from LF !LF!K"cis " resides initially in the cis

chamber of the channel and may enter the cation-selective portion of the channel once it is protonatedand enough negative charge is neutralized to makethe stretch neutral or net-cationic !LF!C"

PA63". As theLF!C"

PA63 stretch exits the channel it may thendeprotonate, making LF!K"

trans. The rates at whichthese stretches of anion-neutralized polypeptideboth enter and exit the cation-selective portion ofthe channel from either the cis or the trans side arelinked to the pH-dependent, microscopic rates ofprotonation and deprotonation. Under favorableDpH conditions (pHtransOpHcis), low pHcis drivesformation of the intermediate LF!C"

PA63 species. LF!C"PA63

then partitions productively to LF!K"trans and translo-

cates, because protonation is slower in the higherpH of the trans solution, and a single cycle oftranslocation is completed:

When the DpH conditions are unfavorable(pHtrans!pHcis), retrograde translocation is more

cistrans

Ser/Thr-rich-barrel

-cla

mp

well

Solvated cap

Translocationbarrier

Unfoldingbarrier

Low pHcis High pHtrans

cis trans=0pH

Kinetically controlled

pH

cis transStability controlled

Stabilizinghigher pHcis

GNI

Figure 8. Structural and energetic models of transloca-tion. Translocation energy diagrams with two barriersand one well modeled on a cross-sectional depiction ofthe PA63 channel colored by domain: D1 0 (magenta), D2(green), D3 (gold), and D4 (blue). The Phe427 f clamp(red) defines an energy well for hydrophobic and cationicstretches of translocating polypeptide.5 The Dj gradientis indicated (at the top in gray) such that one-third of thevoltage-drop occurs cis to the f clamp.5,30 The DpH isindicated (at the top in magenta) such that the f clampdefines a boundary, since this hydrophobic site creates aseal blocking the passage of small ions,5 partitioninghigher pH conditions of the cytosolic (trans) side fromlower acidic pH conditions on the endosomal (cis) side. Inthe energy diagrams (at bottom), the barrier on either thecis or trans side of the f-clamp well, termed “unfoldingbarrier” and “translocation barrier,” respectively, areshaded to indicate their location in the channel. AcidicpH conditions on the cis side of the f clamp destabilizethe substrate protein, allowing unfolded polypeptidechain to pass into the f clamp. Either a favorable DpH orCDj can reduce the trans-side b-barrel translocationbarrier. Translocation can be limited kinetically by thetranslocation barrier (continuous line arrow), but whenthis barrier is reduced by large Dj or DpH, the unfoldingbarrier is rate-limiting (dotted line arrow), which islargely independent of Dj and DpH. Alternatively,translocation may be dominated by the equilibriumstability of the substrate protein (affected by pHcis

conditions, as in Figures 1(c) and S1).

DpH-driven Anthrax Toxin Translocation 975

likely, because the probability that the translocatedspecies, LF!K"

trans, is populated is decreased,and LF!C"

PA63 and LF!K"cis accumulate:

Thus, under appropriate DpH driving force, thecharge-state ratchet acts processively upon individ-ual frames of translocating polypeptide sequencerich in acidic residues. In a sense, the channel isfunctioning as a proton/protein symporter. Thistranslocation model is kinetically linked to therelative kinetic rates of protonating acidic residues

and not simply to their thermodynamic pK values.As the kinetic rate of protonating acidic residues isdependent on pH (or specifically [HC] and [OHK]),this model is consistent with two distinguishingcharacteristics of the phenomenon: (i) positiveDpH conditions even at near-neutral cis pH arestill effective in driving translocation (Figure 5); (ii)negative DpH values cause protein translocation toslow abruptly even at low cis pH, which destabilizethe LFN tertiary structure (Figure 4). While the LFNtranslocation t1⁄2 values (1 to 10 s) are incomparablyslower than the protonation timescales (!10K3 s)expected for Asp or Glu side-chains in neutralpH buffer, the translocation rate hinges on thetimescale of the Brownian diffusion constant of theunfolded protein within the confines of the narrowtranslocase channel (or possibly on the rate ofprotein unfolding on the cis side). This frictionimposed by the channel may severely reduce theoverall timescale of translocation.

A previous study16 has elegantly demonstrated abasic ratchet model for chaperone-assisted translo-cation through a translocase pore. The rate oftranslocation (or velocity) is favorably modulatedby either the cis dissociation rate (kK) or the transassociation rate (kC) of chaperone proteins, whichbind sites contained in the translocating chain.These sites would be analogous to proton-bindingsites on acidic side-chains in LFN. Their molecularmechanics model indicated that either increasingkK or kC effectively increased the observed rate oftranslocation. The model described here is novel, inthat the DpH-driven charge-state ratchet results notfrom pH-dependent differences in substrate poly-peptide folding, or “coiling,” as suggested,16 butrather from electrostatic repulsion between thechannel and anionic charges in the translocatingchain, as reflected in the cation-selectivity of thechannel.

Tandem synergistic ratchets

Experimental evidence presented in Figure 7 andTable 2 indicates that the DpH-driven charge-stateratchet works in conjunction with the f clamp,which we proposed earlier to be a hydrophobicratchet (Figure 9(b)). The tandem ratchets providetwo distinct strategies of interacting with a chemi-cally complex polypeptide substrate that hashydrophobic, hydrophilic, anionic and cationicfunctionalities distributed across its sequence(Figure 9(a)). For example, during a cycle oftranslocation, the f clamp located cis to theextended b barrel5,10 can stably bind hydrophobicsequence from the translocating polypeptide.Hydrophilic polypeptide accumulated as a morecompact conformer, such as an a helix, in the trans-side b barrel may be translocated as an extendedconformer via the DpH-driven, charge-state ratchet,which we expect operates trans to the b barrel. Inturn, the charge-state ratchet imposes force uponthe extended translocating chain and may sub-sequently induce upstream hydrophobic sequence

LFtrans

LFcis

LFPA63

(a)

1 Residue 263

(b)

I II III IV

Figure 9. A tandem Brownian ratchet translocationmechanism. (a) Chemical complexity of a proteinsubstrate, LFN (residues 1–263), colored by functionality:hydrophobic (green), greater than K1.75 kcal molK1 insolvation energy for a ten residue average;31 cationic(blue), Arg, His or Lys; and anionic (red), Asp or Glu.(b) Depicted is a model of some possible partiallyunfolded intermediates of LFN populated during translo-cation, illustrating how the hydrophobic, f-clamp ratchetand the protonation-state ratchet may work together on achemically complex protein substrate (as shown in (a)).Stage I: Initially, the amino terminus of LFN is outside thechannel and negatively charged residues are in adeprotonated state, LF!K"

cis . Stage II: A conductance-blockedintermediate forms when hydrophobic, aromatic and/orcationic sequence binds the f clamp site, which is thenarrow constriction on the cis side of the extended bbarrel5,10 that effectively maintains the DpH. Negativelycharged residues in the translocating chain that haveentered the channel are protonated, making the stretchnet-positive, LF!C"

PA63; this chain is then compatible with thecation-selective channel. Stage III: The translocating chainproceeds through the f-clamp site as hydrophilicsequence enters the b barrel and exits to the trans-sidesolution (via either electro-diffusive or Brownian thermalfluctuations.) Translocation is favored over retro-translo-cation to the cis side under aCDpH. Stage IV: Successiveiterations of this DpH-driven, charge-state ratchet lead toprocessive translocation.

976 DpH-driven Anthrax Toxin Translocation

to dissociate from the f clamp, thereby completinga cycle of translocation (Figure 9(b)).

Physiological relevance of DpH-driven translocation

A translocase in the thylakoid membrane of thechloroplast can transport substrates across mem-branes via a DpH-dependent mechanism.17 TheDpH-driven translocation in the thylakoid, which isindependent of the secretase, Sec61p, as well asNTP hydrolysis, occurs through the twin-argininetranslocation system (TAT) and, unlike anthraxtoxin, it is not believed to be coupled with theunfolding of substrate proteins.18 The lack of anunfolding requirement in the TAT system reflectsthe fact that the translocase forms much largerpores, as shown in electron microscopy studies.19

The translocase of the mitochondrial inner mem-brane forms a narrow pore and can translocatesufficiently under either a DpH or a Dj alone(created by a potassium-diffusion potential20).

The interchangeability of the Dj and DpH inmitochondria is somewhat analogous to anthraxtoxin. The route of anthrax toxin entry into the cellrequires endocytosis into vesicles that acidify asthey mature.21 Although the sign of the Dj in anacidified endosome is certainly positive (jendosome-

Ojcytosol), the reported magnitude varies widelyfrom C10 to C300 mV; the Dj is highly dependenton the concentration of chloride ions, which havebeen shown to concentrate in the endosome as itacidifies, lowering the expected Dj to C10 toC30 mV.22–24 The DpH across the endosomalmembrane is more certain, and drugs or mutationsin the cellular machinery that block endosomalacidification affect the toxicity of anthrax toxin andother analogous toxins.21,25

Experimental evidence presented here demon-strates that a DpH promotes anthrax toxin translo-cation through PA63 channels in planar lipidbilayers and suggests that it is likely a principledriving force of translocation in cells. TheDpH driving force would be expected to be greaterin late endosomes (i.e. pHendosomez5.5 andpHcytosolZ7.3). At pH 5.5, LFN can be translocatedrapidly in planar bilayers at a Dj ofC50 mVand noDpH, or alternatively with little or no Dj (as low asC5 mV) and a DpH of one unit. Full-length LF, onthe other hand, is unable to be translocated at anappreciable rate, even at C70 mV, in the absence ofa DpH, and a DpH is required to observe significanttranslocation. Thus, it appears that DpH-driventranslocation is more relevant to the biologicalaction of the toxin.

In the amino acid sequence of the naturalsubstrates of the PA63 translocase, there is a roughlyequal proportion of basic, positively chargedresidues and acidic, negatively charged residues.This argues that, on the one hand, the acidicresidues must be protonated for aCDj to effec-tively drive translocation. On the other hand, it isapparent from these DpH studies that these acidicresidues may play a role in driving translocation, by

the proposed DpH charge-state ratchet model, andtheir inclusion in the sequences suggests they areintegral to the physiological mechanism of translo-cation.

Materials and Methods

Proteins

PA (83 kDa) and its mutants, full-length EF, and LFN(residues 1–263 of LF) and its mutants were expressedrecombinantly and purified as described.3–5,26 The amino-terminal His6 affinity tags were removed from LFN and itsmutants by treatment with bovine a-thrombin (10 units/mg of protein) for 4 h at room temperature in 20 mM Tris–HCl (pH 8), 150 mM NaCl. Full-length, recombinant LFfreed of its His6 expression tag was provided by MerckResearch Laboratories, and was prepared by them asdescribed.27 LFN* (LFN K14C N242C fluorescently labeledwith Cys-reactive Alexa Fluor 488 and 546 C(5) malei-mides (Molecular Probes)) was prepared as described.3

LFN-biotin was prepared by covalently modifying LFNN242C with a Cys-reactive biotinylation reagent (Pierce),which leaves a disulfide bond between the biotinylmoiety and the protein that is cleavable with a reducingagent such as Tris(2-carboxyethyl)-phosphine (TCEP).13

The heptameric prepore form of PA63 was prepared bynicking PA83 with trypsin and purifying the PA63

heptamer from the smaller 20 kDa fragment usinganion-exchange chromatography.28

Planar lipid bilayers

Bilayers were formed by the brush technique29 acrosseither a 500 mm aperture in a Teflon partition or a 200 mMaperture in a Delrin cup (Warner Instruments, Hamden,CT). Membranes separated two compartments of either3 ml or 1 ml containing buffered solutions of 100 mMKCl,which could be stirred by small magnetic bars. Thebuffers used were either 5 mM potassium succinate,10 mM potassium dimethylglutaric acid (DMG), oruniversal bilayer buffer (UBB; containing 10 mM oxalicacid, 10 mM Mes, and 10 mM phosphoric acid). Allsolutions contained in addition 1 mM EDTA. SolutionpH values were changed by additions of appropriateamounts of 0.2 M or 2 M KOH, 0.1 M or 1 M HCl, orconcentrated buffers of Ches or potassium phosphate.Agar salt-bridges (3 M KCl, 3% agar) linked Ag/AgClelectrodes in 3 M or saturated KCl baths to the cis andtrans compartments. The membrane-forming solutionwas 3% diphytanoylphosphatidylcholine (Avanti PolarLipids, Alabaster, AL) in n-decane, and membraneformation was monitored either visually or by capaci-tance. All experiments were done under voltage-clampconditions; voltages are those of the cis solution (to whichprotein was added) with respect to the trans solution,which was held at virtual ground. Current responseswere either filtered at 10 Hz and recorded on a Narcophysiograph chart recorder (Houston, TX), or filtered at100 Hz by a low-pass eight-pole Bessel filter (WarnerInstruments) and recorded by computer via an Instrutechanalog-to-digital converter (Instrutech, Port Washington,NY) using AXOGRAPH 4.0 (Axon Instruments, UnionCity, CA) software.

DpH-driven Anthrax Toxin Translocation 977

PA63 channel formation and LF, EF or LFN

conductance block

Following membrane formation, PA63 prepore hepta-mer was added to the cis compartment, to a finalconcentration of w1 ng/ml (w2 pM), which was held ata Dj of C1 to C20 mV with respect to the transcompartment. LF or LFN was added to the cis compart-ment (final concentration w3 nM) after PA63 channelformation stabilized. The progress of LF, EF or LFNbinding to PA63 channels was monitored by the continu-ous fall of conductance. In most experiments, O95% ofthe conductance was blocked by LF, EF or LFN beforetranslocation experiments were begun.

Voltage-step translocation experiments

After LF, EF or LFN conductance block of PA63 channelswas complete, excess ligand was removed from the ciscompartment by perfusion, using a pair of syringesconfigured in a push/pull arrangement that were driveneither manually or automatically by a Hamilton Microlabtitrator. The exchange of more than six volumes wasaccomplished in several minutes, while Dj was heldconstant at fromC1 toC20 mV. Translocation of LF, EF orLFN was initiated by stepping Dj to RC40 mV.

DpH-jump translocation experiments

After LF, EF or LFN conductance block of PA63 channelswas complete, excess ligand was removed from the ciscompartment with perfusion, and translocation wasinitiated or stimulated by jumping pHtrans by adding tothe trans compartment an appropriate amount of 0.2 M or2 M KOH, 0.1 M or 1 M HCl, or other concentrated buffersuch as Ches or phosphate. Altered pH values wereconfirmed in situ using a micro pH electrode.

Fluorescence equilibrium denaturation

Guanidine hydrochloride denaturation profiles of LFN*in UBB at pH 7.5 in the presence and absence of 1 MD-glucose were measured using a computer-controlledHamilton Microlab titrator interfaced to an ISS K-2fluorimeter for fluorescence resonance energy transfer(ArC laser excitation at 488 nm, emission ratio of520–570(G16) nm) as described.3

Acknowledgements

This work was supported by an NRSA fellow-ship, AI062204 (to B.A.K), and NIH grants,AI022021 (to R.J.C.) and GM29210 (to A.F.) Wethank Roman Melnyk for helpful advice andCharles Peskin for useful discussions of Brownianratchets. R.J.C. holds equity in PharmAthene, Inc.

Supplementary Data

Supplementary data associated with this articlecan be found, in the online version, at doi:10.1016/j.jmb.2005.11.030

References

1. Zhao, J. M. & London, E. (1986). Similarity of theconformation of diphtheria toxin at high temperatureto that in the membrane-penetrating low-pH state.Proc. Natl Acad. Sci. USA, 83, 2002–2006.

2. Ren,J.,Kachel,K.,Kim,H.,Malenbaum,S.E.,Collier,R.J.& London, E. (1999). Interaction of diphtheria toxin Tdomain with molten globule-like proteins and itsimplications for translocation. Science, 284, 955–957.

3. Krantz, B. A., Trivedi, A. D., Cunningham, K.,Christensen, K. A. & Collier, R. J. (2004). Acid-inducedunfolding of the amino-terminal domains of the lethaland edema factors of anthrax toxin. J. Mol. Biol. 344,739–756.

4. Zhang,S.,Udho,E.,Wu,Z.,Collier,R. J.&Finkelstein,A.(2004). Protein translocation through anthrax toxinchannels formed in planar lipid bilayers. Biophys. J. 87,3842–3849.

5. Krantz, B. A., Melnyk, R. A., Zhang, S., Juris, S. J.,Lacy, D. B., Wu, Z. et al. (2005). A phenylalanine clampcatalyzes protein translocation through the anthraxtoxin pore. Science, 309, 777–781.

6. Koriazova, L. K. & Montal, M. (2003). Translocation ofbotulinum neurotoxin light chain protease throughthe heavy chain channel. Nature Struct. Biol. 10, 13–18.

7. Oh, K. J., Senzel, L., Collier, R. J. & Finkelstein, A.(1999). Translocation of the catalytic domain ofdiphtheria toxin across planar phospholipid bilayersby its own T domain. Proc. Natl Acad. Sci. USA, 96,8467–8470.

8. Collier, R. J. & Young, J. A. (2003). Anthrax toxin.Annu. Rev. Cell Dev. Biol. 19, 45–70.

9. Blaustein, R. O., Koehler, T. M., Collier, R. J. &Finkelstein, A. (1989). Anthrax toxin: channel-formingactivity of protective antigen in planar phospholipidbilayers. Proc. Natl Acad. Sci. USA, 86, 2209–2213.

10. Nassi, S., Collier, R. J. & Finkelstein, A. (2002). PA63

channel of anthrax toxin: an extended b-barrel.Biochemistry, 41, 1445–1450.

11. Benson,E.L.,Huynh,P.D., Finkelstein,A.&Collier,R. J.(1998). Identification of residues lining the anthraxprotective antigen channel. Biochemistry, 37, 3941–3948.

12. Wesche, J., Elliott, J. L., Falnes, P. O., Olsnes, S. &Collier, R. J. (1998). Characterization of membranetranslocation by anthrax protective antigen.Biochemistry, 37, 15737–15746.

13. Zhang, S., Finkelstein, A. & Collier, R. J. (2004).Evidence that translocation of anthrax toxin’s lethalfactor is initiated by entry of its N terminus into theprotective antigen channel. Proc. Natl Acad. Sci. USA,101, 16756–16761.

14. Qu, Y., Bolen, C. L. & Bolen, D. W. (1998). Osmolyte-driven contraction of a random coil protein. Proc. NatlAcad. Sci. USA, 95, 9268–9273.

15. Buck, M. (1998). Trifluorethanol and colleagues:cosolvents come of age. Recent studies with peptidesand proteins. Quart. Rev. Biophys. 31, 297–355.

16. Simon, S. M., Peskin, C. S. & Oster, G. F. (1992). Whatdrives the translocation of proteins? Proc. Natl Acad.Sci. USA, 89, 3770–3774.

17. Settles, A. M., Yonetani, A., Baron, A., Bush, D. R.,Cline, K. & Martienssen, R. (1997). Sec-independentprotein translocation by the maize Hcf106 protein.Science, 278, 1467–1470.

18. Hynds, P. J., Robinson, D. & Robinson, C. (1998). Thesec-independent twin-arginine translocation system

978 DpH-driven Anthrax Toxin Translocation

can transport both tightly folded and malfoldedproteins across the thylakoid membrane. J. Biol.Chem. 273, 34868–34874.

19. Gohlke, U., Pullan, L., McDevitt, C. A., Porcelli, I., deLeeuw, E., Palmer, T. et al. (2005). The TatA componentof the twin-arginine protein transport system formschannel complexes of variable diameter. Proc. NatlAcad. Sci. USA, 102, 10482–10486.

20. Pfanner, N. & Neupert, W. (1985). Transport ofproteins into mitochondria: a potassium diffusionpotential is able to drive the import of ADP/ATPcarrier. EMBO. J. 4, 2819–2825.

21. Friedlander, A. M. (1986). Macrophages are sensitiveto anthrax lethal toxin through an acid-dependentprocess. J. Biol. Chem. 261, 7123–7126.

22. Van Dyke, R. W. & Belcher, J. D. (1994). Acidificationof three types of liver endocytic vesicles: similaritiesand differences. Am. J. Physiol. 266, C81–C94.

23. Rybak, S. L., Lanni, F. & Murphy, R. F. (1997).Theoretical considerations on the role of membranepotential in the regulation of endosomal pH. Biophys.J. 73, 674–687.

24. Sonawane, N. D., Thiagarajah, J. R. & Verkman, A. S.(2002). Chloride concentration in endosomesmeasured using a ratioable fluorescent ClK indicator:evidence for chloride accumulation during acidifica-tion. J. Biol. Chem. 277, 5506–5513.

25. Merion, M., Schlesinger, P., Brooks, R. M., Moehring,J. M., Moehring, T. J. & Sly, W. S. (1983). Defective

acidification of endosomes in Chinese hamster ovarycell mutants “cross-resistant” to toxins and viruses.Proc. Natl Acad. Sci. USA, 80, 5315–5319.

26. Pimental, R. A., Christensen, K. A., Krantz, B. A. &Collier, R. J. (2004). Anthrax toxin complexes:heptameric protective antigen can bind lethal factorand edema factor simultaneously. Biochem. Biophys.Res. Commun. 322, 258–262.

27. Shoop, W. L., Xiong, Y., Wiltsie, J., Woods, A., Guo, J.,Pivnichny, J. V. et al. (2005). Anthrax lethal factorinhibition. Proc. Natl Acad. Sci. USA, 102, 7958–7963.

28. Cunningham, K., Lacy, D. B., Mogridge, J. &Collier, R. J. (2002). Mapping the lethal factor andedema factor binding sites on oligomeric anthraxprotective antigen. Proc. Natl Acad. Sci. USA, 99,7049–7053.

29. Mueller, P., Rudin, D. O., Tien, H. T. & Westcott, W. C.(1963). Methods for the formation of single bimole-cular lipid membranes in aqueous solution. J. Phys.Chem. 67, 534–535.

30. Blaustein, R. O. & Finkelstein, A. (1990). Voltage-dependent block of anthrax toxin channels in planarphospholipid bilayer membranes by symmetrictetraalkylammonium ions. Effects on macroscopicconductance. J. Gen. Physiol. 96, 905–919.

31. Wesson, L. & Eisenberg, D. (1992). Atomic solvationparameters applied to molecular dynamics of pro-teins in solution. Protein Sci. 1, 227–235.

Edited by G. von Heijne

(Received 10 October 2005; received in revised form 4 November 2005; accepted 8 November 2005)Available online 1 December 2005

DpH-driven Anthrax Toxin Translocation 979