Reversible adsorption and nonreversible insertion of Escherichia coli alpha-hemolysin into lipid bilayers

BiophysicalJournal Volume 71 October 1996 1869-1876

Reversible

Adsorption

and Nonreversible Insertion of Escherichia

coli

a-Hemolysin

into

Lipid

Bilayers

Laura Bakas,* Helena Ostolaza,* Winchil L. C. Vaz,# and Felix M.

Gonfi*

*Grupo Biomembranas (UnidadAsociada alC.S.I.C.), DepartamentodeBioquimica, UniversidaddelPais Vasco,48080Bilbao,Spain, and#Unidade de Cieancias ExactaseHumanas, Universidade doAlgarve, P-8000Faro, Portugal

ABSTRACT a-Hemolysin isanextracellularproteintoxin(107 kDa) produced bysomepathogenicstrains of Escherichiacoli.

Although stable in aqueous medium, it can bind to lipid bilayers and produce membrane disruption in model and cell membranes.Previous studies had shown that toxin bindingto thebilayerdid notalwayslead to membranelysis.In thispaper,

wefind that a-hemolysin maybind the membranesinatleasttwoways,areversibleadsorptionandanirreversibleinsertion. Reversibilityisdetected by the ability of liposome-boundtoxin to inducehemolysisof added horseerythrocytes; insertionis accompaniedbyanincrease intheprotein intrinsicfluorescence. Toxin insertion does notnecessarilylead tomembranelysis. Studies ofa-hemolysin insertion into bilayers formed fromavarietyofsinglephospholipids,orbinarymixturesof phospho-lipids,orofphospholipidandcholesterol, reveal that irreversible insertionis favoredbyfluidovergelstates, bylowoverhigh cholesterol concentrations, by disordered liquid phasesover gel orordered liquid phases, and by gel over ordered liquid phases. These resultsarerelevant to the mechanism ofactionofa-hemolysin andprovidenewinsightsinto themembrane

insertion oflarge proteins.

INTRODUCTION

Escherichia coli a-hemolysin (HlyA) is a 107-kDaprotein

secretedbypathogenic strains of thisbacteriumandbelongs

to the so-called RTX protein family, characterized by a glycine-rich nonapeptide repeat region near the C-terminal

end (see Coote, 1992, for a review). In addition to this repeat domain, the protein has a hydrophobic region near the N-end (Menestrina et al., 1994, 1995), andaC-terminal

signal peptide (Zhang et al., 1995; Chervaux and Holland,

1996). The mature protein contains two fatty acyl residues linked to internal lysines (Stanley et al., 1994). HlyA dis-rupts eukaryotic cell membranes andforms cation-selective

channels in planar lipid membranes (Menestrina et al., 1995). The toxin is also capable of inducing leakage of

phospholipid large unilamellar vesicles (Ostolaza et al.,

1993).Ingeneral,there is ample

experimental

evidence that

HlyA, although existing in soluble form after its

secretion,

may become membrane-associated to produce membrane

disruption.

The mechanism ofHlyAinsertion in lipidbilayers is not known in detail. Besides the fact that some cells might contain HlyA receptors, pure lipid bilayers and vesicles constitute a good model for this study. We have recently

described methods for separately

measuring

toxin binding

tomembranes and toxin-inducedlysis, and found that

bind-ing is not

necessarily

followed by membrane damage. In

Receivedfor publication 2May1996and infinalform3 July1996.

Address reprint requests to Dr. Felix M. Gonii, Departamento de Bioqufmica, Universidad del Pais Vasco, Aptdo. 644, 48080 Bilbao, Spain. Tel.: 34-4-464-7700, ext. 2407; Fax: +34-4-464-8500; E-mail: gbzoseth@lg.ehu.es.

Dr.Bak6s'spennanentaddress isCktedradeBiologia,Facultad de Cien-ciasExactas, Universidad Nacional de La Plata, La Plata, Argentina. i 1996bytheBiophysicalSociety

0006-3495/96/10/1869/08 $2.00

particular, HlyA binds lipid bilayers with about the same

affinity in the presence or absence of

Ca2+,

butonly when theprotein has beenpreincubatedwith this cation does the lytic effect follow toxinbinding (OstolazaandGofii, 1995). In this paper, we explore in more detail the influence of some lipid properties onHlyAbinding tomembranes. The interest of thisstudygoes beyondthe mechanismofaction of thetoxin, because it is one case of the moregeneral,and

important, biologicalproblem ofprotein insertionin mem-branes (see reviewsby Jain and Zakim, 1987; Hannavy et al., 1993; Isenman et al., 1995; see also, to mention but a few examples, Dibble et al., 1993; Sankaram et al., 1994;

Pottand Dufourc, 1995; Rytomaa and Kinnunen, 1995). The presentstudy isspecificallydirected toexploring the influence oflipidphasesonthebindingof HlyA. Particular

attentionispaidtothecaseofcoexistingfluidphases, such as phospholipid-cholesterol fluid ordered and fluid disor-deredphases (Sankaram andThompson, 1991; Almeida et

al., 1992; Monette et al., 1993; Mateo et al., 1995; Mc-Mullen andMcElhaney, 1995; Pott and Dufourc, 1995), and to the case ofcoexisting gel and fluid phases (Vaz et al.,

1989; Jorgensen et al., 1993; Sankarametal., 1992, 1994;

Piknova'etal., 1996). Tomita et al. (1992) have studied the

influence of membrane fluidityonthe assembly of

Staphy-lococcus aureus a-toxin, a channel-forming protein not

belonging to the RTX family, in liposomal membranes.

Followingasimilar experimental reasoning, we have estab-lishedbilayer conditions under which HlyA may bind lipid

membranes in either a reversible or an irreversibleway.

MATERIALS AND METHODS

Materials

Egg phosphatidylcholine (PC) was grade I from Lipid Products (South Nutfield, England). Dioleoyl, dimyristoyl, dipalmitoyl, and distearoylphos-1869

Volume 71 October 1996

phatidylcholine (respectively, DOPC, DMPC, DPPC, and DSPC) were

supplied by Avanti Polar Lipids (Alabaster, AL). Cholesterol (Ch) was from Sigma. I-Amino-naphthalene-1,3,6-trisulfonate (ANTS) and p-xylenebispyridiniumbromide (DPX)wereobtained fromMolecular Probes

(Eugene, OR). Plasmid-encoded a-hemolysin (HlyA) waspurified from

the culture filtrates ofanoverproducing strain ofE. coli, accordingtothe

method of Ostolazaetal. (1991); before itsuse,the proteinwasdialyzed

against150 mM NaCl,etMurea,20 mM Tris/HCl,pH7.0 (TCU buffer), towhich1mMEGTAwasadded. Horse red blood cellsweresupplied by

Microlab(Madrid, Spain).

Largeunilamellarvesicles (LUVs) of different compositionswere

pre-pared by extrusion andsized using 0.1-gmporesize Nuclepore membranes asdescribed byMayeretal.(1986);thebufferwas20 mM Tris-HCl, 150 mMNaCl,pH7.0 (TC buffer), ± EGTA, Ca24,orZn2+ as required.

100

75

50 25

0

Measurements ofintrinsic fluorescence ofHlyA Bilayer-toxin interactionsweremonitoredthroughchanges in the intrinsic fluorescenceemissionspectraof HlyA. Small aliquots ofaconcentrated

LUVsuspension wereadded toaprotein solution (0.15-0.30 JIM)ina

cuvettewith continuous stirring. Afterequilibrating for 5min,emission

spectra were recorded with anexcitation light of 295 nm (slit 5 mm).

Fluorescenceintensity measurements were corrected for light scattering

(Surewiczand Epand, 1984).

Toxin binding to liposomes and toxin transfer

to erythrocytes

The method isessentiallytheonedescribed by Tomitaetal.(1992).Horse red blood cellswereusedasindicators inassessingthehemolytic activity

ofHlyAboundtomultilamellar vesicles(MLVs).Toxin(20-30jig)and

liposomes(multilamellar vesicles,1mM)wereincubatedinTCbufferwith

10 mM CaCl2 for 30 min, at the required temperature. HlyA-liposome complexeswererecoveredby centrifugation (16,000Xg,15min, 40C)and washed threetimes toremove anyunbound toxin.The complexes were

resuspendedin 200 ,ul of cold buffer and used forassaysofhemolytic

activity,aswellasinproteinandlipiddetermination. Thiswascalled the

"centrifugation method" formeasuringtotalproteinboundtoliposomes.

For hemolytic activityassays, analiquotof theHlyA-liposome

com-plexeswasseriallydiluted withcold buffer ina96-well microtiterplate.

One hundred microliters of the dilutesuspensionsweremixedwith100,ul

ofastandardizedsuspensionofred blood cells and lefttoincubate for 30 min at 370C. The absorbance ofsupematantswasread at412nm. One

hundredpercentlysiswasestablished afterlysingthered blood cells with Triton X-100(1%finalconcentration).

100

cf)

Cf) Cll 75 50 25 0 100 75 50 25 0

0. 01

0.1

1 10 100

BOUND

PROTEIN

(

,ag/ml)

Time-resolved experimentsand

equilibrium conditions

Preliminarytime-resolved experiments using the varioustechniques

de-scribed above showed that5min after toxin addition, the variouscelland

model systemsused in thisstudy had reachedequilibrium. This is what

would beexpectedfromadiffusion-limitedprocess,asisprobablythecase with toxinbindingand insertion. Thus allmeasurementshavebeencarried

out5mmnafterHlyA addition,exceptin thosecasesinwhich,for the sake

ofconvenience, equilibration was allowed to take placefor 30 min,as stated in each case.

RESULTS

Thecapacityofliposome-bound HlyAtoproduceredblood celllysiswastestedafterincubatingtheproteinwithDOPC, DMPC,orDPPC multilamellarliposomesat0°Cand37°C.

The results are presented in Fig. 1. The dotted line

corre-FIGURE 1 Accessibility ofliposome-bound a-hemolysin toredblood

cells. HlyA was incubated for 30 min at 0°C (0) or 37°C (0) with

multilamellar liposomesofDOPC, DMPC, orDPPC, as indicated. The

resulting liposome-protein complexes werefurther incubated with horse

erythrocytes,and thehemolytic activitywasrecorded. See Materials and

Methods for details. Dotted line: hemolyticactivityof nativetoxin, i.e., preincubatedin the absenceofliposomesand then addedtotheerythrocyte suspension.

sponds to the dose-response curve of free hemolysin, i.e., protein that has been preincubated in the absence of

lipo-somes andthen tested againstred blood cells. When HlyA

has beenpreincubated with DMPC liposomes at0°C (i.e.,

well below

Tc,

the main gel-to-fluidtransitiontemperature of the phospholipid; open circles), incubation with horse

erythrocytes leads to hemolysis, although not all of the

DOPC

v v

.70

-V

V

V~~~

I

-DMPC

v>S

V/L

-DPPC

,,

I a 1870 BiophysicalJournal

a-Hemolysin Insertion intoLipid Bilayers proteinseemstobecapableofproducingcelllysis,because

thedose-response curveis somewhat shifted to higher

pro-tein concentrations. However, when preincubation with

DMPC (Tc = 23°C)has takenplace at37°C (filled circles),

the fraction of HlyA available for hemolysis is much

smaller, andthedose-responsecurveis shiftedbyaboutan

order of magnitude. This can be quantitatively stated by

comparing theamountofprotein requiredin all three cases

toproduce 50% hemolysis. The dataareshown in Table 1,

and the protein concentrations are, respectively, 0.5, 0.95,

and 10.0

jig/ml

for HlyA preincubated without and with

liposomes at0°C andat 37°C.Intermsofpercentage, 53%

of the protein preincubated at 0°C but only 5% ofHlyA preincubated at 37°C with DMPC is "available" for

hemo-lysis. Thephenomenonappearstoberelatedtothephysical

state of the bilayer, because HlyA preincubated with

mul-tilamellar vesicles of DPPC (Tc = 41.5°C) remains

avail-able forhemolysiswhen incubatedat0°or37°C (Fig. 1 and

Table 1). The opposite situationisfound withDOPC,which

remains fluidat bothtemperatures; theproportion ofHlyA available for hemolysis is very small in both cases, and

values producing 50% hemolysis (obtainable only by

ex-trapolation) arevirtuallysimilarandcomparabletothe data forpreincubationwith DMPC in the fluidstate(37°C). Note

that protein binding to the multilamellar liposomes is

es-sentially similar in allcases (Table 1), irrespective of tem-peratureorlipid composition. Control experimentsinwhich

protein-free liposomes are incubated with red blood cells

showed that none ofthe vesicle preparations had any

he-molytic character at all.

Theseresults areinterpreted in termsof the existence of

two populations of membrane-bound HlyA molecules,

re-spectively, reversibly and irreversiblybound. Lipid bilayers

in the fluid state favor irreversible binding. Our previous

studies with LUV in the fluid state indicated that, under

conditionsleadingtomembraneleakage,the toxinremained irreversibly bound tothemembrane (Ostolazaetal., 1993).

In thecase offluidMLV (e.g., DOPC,orDMPC at37°C), virtuallyall of the toxin(>90%)remainsirreversibly bound (Table 1). For bilayers in the gel state (e.g., DPPC, or

DMPCat0WC), theapparentproportion of reversibly bound HlyA is far from 100%; this result is observed either

be-cause some of theprotein is irreversibly boundevenbelow

T,,

or because not all of the reversibly bound hemolysin binds red blood cells under conditions leading to leakage (Ostolaza and Gonii, 1995). A similar situation below Tcwas

found forS. aureus a-toxin (Tomitaet al., 1992).

Some ofthe formerexperimentswith DMPC multilamel-lar vesicles were repeated with preincubated HlyA and

liposomes in TC buffer but with 1 mM ZnCl2, or 1 mM

EGTA, instead of

Ca2+.

Theseexperimentswereperformed

because previous studies had shown thatin the absence of divalent cations, or in the presence of

Zn2+,

HlyA was

inactive (Ostolaza et al., 1995). In agreement with the

previous observations, no hemolysis was detected under

thoseconditions,inspiteof the fact that red blood cellswere

suspendedinacalcium-containingbuffer(datanotshown). Thissuggests that, in the transfer fromMLVtoerythrocyte (whenever this occurs), the calcium-binding domain of

HlyA (Ostolazaetal., 1995) doesnotunfoldtoallowentry ofcalcium ions andsubsequent protein activation.

The hypothesis that HlyA becomes irreversibly boundto

lipid bilayers when they are in the fluid state is also

sup-portedbyanexperimentinwhichtoxinbindingisassessed

as anincreaseinthe intrinsicTrpfluorescence of theprotein

(SurewiczandEpand, 1984;Ostolaza andGoiii, 1995). Fig. 2 shows the relativechange in HlyA intrinsic fluorescence

when incubated continuously with increasingtemperatures inthepresenceofDMPC LUVs (Tc = 23°C). Fluorescence

increases steeply between 20°C and 23°C, i.e., near the

gel-fluid transition. Ifasystemequilibrated well above Tc is thengraduallycooleddown (Fig. 2, triangles), the intrinsic fluorescence doesnotvary.Our interpretation of these data

is that HlyA becomes inserted in the DMPC bilayer in a

nonreversibleway,particularly when the lipid is in the fluid

phase. The tryptophanyl residues being in a less polar

en-vironment than before, the fluorescence emission intensity increases. Once the protein is inserted above Tc, cooling

does not reverse the fluorescence effect, because binding above Tc is irreversible. The peptide becomes kinetically trapped in the gel state. Note that direct measurements of liposome-bound protein do not reveal major differences between the gel and the fluid phase (Table 1), probably

because thedirect centrifugation method gives anestimate

TABLE I Bindingofa-hemolysintophospholipid liposomes and transfer of a-hemolysin from liposome-toxin complexes to horseerythrocytes

%Binding* HlyA concentration(pLg/ml)# % Residualactivity§

Phospholipid 0°C 37°C 0°C 370C 0°C 370C

DOPC 10.9 15.3 15.5" 16.99 3.2q 3.01

DMPC 10.2 14.7 0.95 10.0 53 5.0

DPPC 9.8 13.0 0.78 0.81 64 61

*Totalprotein bound, according to thecentrifugationmethod.

#ConcentrationofHlyA, intheform of toxin-liposome complex, that produces 50% hemolysis in the standard preparation of horse red blood cells. The correspondingvalue for free HlyA is 0.5 ,tg/ml.

§Thepercentage residual hemolytic activity in the toxin-liposome complexes is calculated as (concentration of native toxin required to produce 50% hemolysis/concentration of liposome-bound toxin required to produce50%hemolysis) X 100 (Tomita et al., 1992).

sExtrapolated values.

1871 Bakais etal.

Volume 71 October 1996

15 20 25 30 35

TEMPERATURE

(C)

FIGURE 2 The influence of gel-fluid phase transition of DMPC on

HlyA bindingtoLUVs, assessedas an increase in intrinsicTrp

fluores-cenceof the protein. *,Heatingrun;V,coolingrun.

oftotal (reversibly + irreversibly) bound protein, whereas intrinsic fluorescence increases noticeably only when the

protein becomes irreversiblyembedded in alipid bilayer.

Binary phospholipid mixtures offer various possibilities of phase transitions and of coexistence of gel and fluid phases. We have studied the incorporation of HlyA into various mixtures of DMPC and DSPC at 300C. At this temperature,thebinary lipid systemis in theliquid crystal-line phase foraDSPC molar fractionx ' 0.15, and in the

gel phase forx 0.65 (Knolletal., 1981; Vazetal., 1989;

Sankarametal., 1992; Jorgensenetal., 1993; Piknovaetal., 1996); between these boundaries, the gel and liquid crys-tallinephases coexist. Total protein bound, accordingtothe centrifugation method, was 3 ± 0.1/104 (protein/lipid mol

ratio)at30°C, irrespective of thesystemcomposition. HlyA irreversiblebindingtoDMPC/DSPCLUVs is shown inFig. 3 A, as detected through changes in the intrinsic fluores-cence. Again insertion appears to be at a maximum when the system is in the fluid phase, lower when both phases coexist, and virtually zero when the lipids are in the gel phase.A smallbutreproducibleincreaseobserved between

XDSPC = 0.00andxDspc = 0.05 is attributedtoafacilitated

protein insertion when mixed-length phospholipid chains

coexist. Theproportionof insertedproteindoesnotdecrease

precisely in parallel with the theoretical fraction of fluid

phase (Fig. 3A, dotted line).Thismaybe duetofluctuations in the phase boundaries induced by the presence of the

protein. Our differential scanning calorimetry studies of

phospholipid-HlyA systems indeed show a protein-depen-dentwideningof the mainlipidgel-fluid phasetransition(P. Veigaetal., unpublished data). Insertion ofa small (25 aa)

peptideis knowntoleave thephaseboundariesunmodified

(Sankarametal., 1994),but the behavior of the muchlarger

I-C.) 0 IU o 0.0 1.0 cn CO) 0.8

CO

In 0.6 0 z In 0.4 r1 0.2 = CO cn : 0.0 0.2 0.4 0.6 0.8 1.0 XDSPC

FIGURE 3 Binding of a-hemolysin to bilayers composed of binary phospholipidmixturesDMPC/DSPC. (A) Irreversible LUVbinding mea-sured as an increase in intrinsic Trp fluorescence of the protein. Data relative to fluorescence intensity of the pure protein in buffer, in the absence oflipids. (B) Accessibilityof MLV-boundHlyAtoredblood cells

(asinFig. 1). Measurementswerecarriedout at30°C. Thetwo discon-tinuousvertical lines indicate thepurelipid phaseboundaries. The dotted linecorrespondstothe calculatedmassfraction of fluidphaseas afunction

ofsystemcomposition.

and more complex HlyA may be different, particularly in

view of the small size of the fluidphasedomains in DMPC/

DSPC mixturesat30°C (Piknovaetal., 1996).Theputative

effect of the protein on the vesicle curvature and subse-quentlyonthephasebehaviorof thelipidmixture(Brumm

etal., 1996)should also be taken intoaccountinoursystem. When the reversibility ofHlyA binding to DMPC/DSPC

bilayersinthe form of MLVs is testedby incubating

liposome-bound toxin with horse

erythrocytes,

the results are rather

straightforward (Fig.3B).Aslongasthere issomelipidin the 1.55 1.50 1.45 C o= \X 1.40 1.35 0 LL 1.30 h 1.25! 1.20 10

0

Q

X

DSPC

1872 BiophysicalJournal

a-Hemolysin Insertion intoLipid Bilayers

fluid phase,binding is essentially irreversible-no hemolysis is detected. Only when the mixture is in the gel phase, and particularly for

xDspc

= 1.0, does the binding become revers-ible. The pointat

XDSpc

= 0.8 isinterestingbecause it shows virtually the same F/F0 as

xDspc

= 1.0 (Fig. 3 A), yet the recovery ofhemolytic activity is much smaller in the former case. One plausibleexplanation is that

xDspc

= 0.8represents alimiting situation; in fact, the results in Fig. 3 may suggest

that,in thepresenceofHlyA,thelipid phaseboundariesarein fact near

xDspc

= 0.05 and 0.8, respectively. CholesteroV

phosphatidylcholine mixtures are interesting because they of-fer thepossibility ofcoexisting ordered and disorderedliquid

phases (Vistand Davis, 1990; Sankaram andThompson, 1991; Almeida et al., 1992; Mateo et al., 1995; McMullen and McElhaney, 1995). ForDMPC/cholesterol, the phasediagram

publishedbyAlmeidaetal.(1992) shows,at10°C,agelphase

for 7 mol% cholesterol, an orderedliquid phase at 30mol% cholesterol, and coexisting gel and ordered liquid phases be-tween. At 37°C, the phases are disordered liquid, disordered liquid + ordered liquid, and ordered liquid, the two corre-sponding boundaries occurring at lOmol% and -30 mol% cholesterol. The total amount ofliposome-bound HlyA esti-mated by thecentrifugation method for bilayers composed of DMPC/cholesterol decreases steadily and slowly with increas-ing proportions of cholesterol (Fig. 4), apparently with little or no influence of temperature or phase transitions, as was the

casefor pure DMPCbilayers (Table 1).

When irreversible HlyA insertion into LUV bilayers is considered, assessed by the intrinsic fluorescence method (Figs. 5A and 6 A), thegeneral tendency is, as seen for total binding in Fig. 4, a decrease in irreversible binding as the cholesterol concentrationincreases. At 10°C (Fig. 5 A), the boundary between S and S +

Lo

does not markany

impor-tant change in toxin binding, but beyond the boundary

1.50

0

U-f

U.

1.25

1.00

80

I-->

60

U U

I- 40

-J

0

<

20

I

0

CD 2.00 -J 1.75 0 D 1.50 0 2 1.25 -@ 1.00 a 0.75 z D 0.50 0 < 0.25 I 0.00 FIGURE 4 T4 HlyAtobilayer carriedout at

0

10

20

30

40

50

%

CHOLESTEROL

50

% CHOLESTEROL

FIGURE 5 Binding of a-hemolysin to bilayers composed of DMPC/

cholesterolat10°C.(A)IrreversibleLUVbindingmeasured by theintrinsic

fluorescence method. (B) Accessibility of MLV-bound toxin to erythro-cytes(asinFig. 1).Thediscontinuous verticallines correspond to thelipid

phase boundaries.

l

~~~~~~~between

S +

Lo

and

Lo

binding

is

clearly

decreased,

sug-gesting that the liquid ordered state is not a favorable solvent for

HlyA.

At37°C the

bilayer ability

tobind

hemo-lysin

decreases

steeply

as soon assome

Lo

phase

is

formed,

l l l l l

suggesting

again

the low

affinity

of

Lo

for the toxin. Note

0 10 20 30 40 50 that for cholesterol concentrations ator near50%, when the

DMPC gel-fluid transition has been completely smeared

% CHOLESTEROL

out, HlyA binding at

10°C and 37°C is virtually the same.

tal binding (measured by the centrifugation method) of Tests of

accessibility

to redblood cells of

HlyA

bound to scomposedofDMPC:cholesterolmixtures. Measurements DMPC/cholesterol MLVs (Figs. 5 B and 6 B) confirm in 0°C

(U)

and37°C (-). general the previous observations. At 37°C, accessibility is at

Bakcas et al. 1873

4

Volume 71 October 1996 A

worthy

observation is that, at the temperature of

10°C,

the A

t

370C

liquid

ordered

phase

is somehow

stabilizing

the

membrane-)

~

\

l

l

bound

form

ofthe

toxin-not

only

does

the

recovery

of

he-molytic activity

decrease and remain low as soon as some

cholesterol (andpresumably some

Lo

phase) is present (Fig. 5 B), but the recovery from pure

Lo

(50% cholesterol) is also

L

\ |

L

|

L

] L lower at

1O°C

than at37°C (Fig. 6 B). This is an indication that,

d \ l d 0 0 for a given phospholipid/cholesterol mixture in the same phase

at two different

temperatures,

even when the

proportion

of

irreversibly bound HlyA is similar (Figs. 5 A and 6 A), the fractionof

protein

transferred from

liposomes

to

erythrocytes

is lower at the lower temperature, perhaps indicating a

tem-perature-dependent differential organization of the protein in the

bilayer.

In

turn,

this

suggests

that the mode of

incorporation

) I of HlyA tolipidbilayers may be more complex than the two

0 10 20 30 40 50 possibilities (reversible/irreversible) mentioned above. With

this limitation in mind, we can conclude that thelipidbilayer % CHOLESTEROL properties that facilitate the irreversible insertion of HlyA are thefluid state over the gel state (Figs. 1-3), the low over the highcholesterol concentrations (Figs. 5and6), thedisordered

B

liquid (Ld)over

theordered

liquid

(L.)

orgelstates

(Figs.

5and

VB

370C

_ 6), and the gel over theliquidordered state (Figs. 5 and 6).

+

LoI

Lc

₎

0 10 20 30

%

CHOLESTEROL

FIGURE 6 Binding of a-hemolysin tobilayers comp

cholesterolat37°C. (A) Irreversible LUV bindingmeasure fluorescence method. (B) Accessibility of MLV-bound

cytes(as in Fig. 1). The discontinuous vertical linescorres phase boundaries.

a minimum when the bilayer is in the Ld s

maximumwhenthe

Lo

statepredominates (Fig.( two remarkable features are observed (Fig. 5 1

veryhigh hemolytic activityforHlyAboundtof 10°C (in accordancewith the datainTable 1),

dramatic decreaseas soon as5%cholesterol isac DMPC/cholesterol mixture should be in thegel

asthepureDMPC,but thelargedecrease in hen

(withasimilarextentofbinding accordingtoth measurements; Fig. 5 A) suggests again that changingthephasebehaviorof the mixture.The

DISCUSSION

The mechanism of membrane lysis bya-hemolysin

Our previous studies on this protein have shown that 1) a

specific receptoris notrequired for membrane lysis to occur, because leakage of large molecules is induced by HlyA in vesicles of definedlipid composition, e.g., pure egg PC (Os-tolaza et al., 1993); 2) HlyA may bind lipid bilayers under conditions (e.g., absence of

Ca2+)

that do not lead to lysis

(Ostolaza andGofii, 1995). The data in this paperprovide us with a closer look at the binding step. Thebinding data

ob-40 50 tained by the centrifugation method and by the intrinsic fluo-rescencemethod do not always agree (e.g., Table 1 andFig.2),

particularlybecause the fluorescenceprocedure shows differ-osed of DMPC/ encesbetween

binding

tothe

gel

and fluid

phases,

whereas the

-dbytheintrinsic

centrifugation

method does not. This had not been detected in

toxin toerythro- former studies because only bilayers in the fluid phase were

spondtothelipid tested. When thebinding dataarecomplementedwith assays ofaccessibility ofliposome-bound HlyAto

erythrocytes,

two populationscanbedistinguishedin themembrane-boundtoxin

fraction,onethatistransferredtothe red blood cells(reversibly tate and at a boundHlyA)andonethatisnot(irreversiblybound).Thefact 6B). At10°C, thatirreversible butnotreversible bindingischaracterized

by

3). One is the changes in intrinsic fluorescence supports the

hypothesis

that

pure

DMPCat "binding" actuallyincludestwodistinctphenomena, reversible

followedby a adsorption of theprotein tothebilayerand irreversible inser-lded. The 95:5 tion into thesame.Adsorptionwould bean

early

step

that,

in

state asmuch certaincases(e.g.,fluid

bilayers),

would be followed

by

inser-tolytic

activity tion.Adsorption appearstobe largely

independent

of

bilayer

efluorescence fluidity orcomposition (atleast in the absenceofnet surface the protein is charge), whereas insertion is highlydependenton thebilayer esecondnote- physicalproperties (see Results).

1.50 0

U-f

LL 1.25 1.00 80 70 60 _ 50 _

Ld I

Ld

I-0

0

0,c

40 _ 30 - 1 20 -104 0 1874 BiophysicalJournal

a-HemolysinInsertion intoLipid Bilayers Bycombining data from this andpreviouspapers

(Osto-laza et al., 1995; Ostolaza and Gonii, 1995), it can be concluded thatirreversible insertion doesnotlead

automat-ically to membrane lysis (see the case of binding in the presence of EGTA, accompanied by a similar increase in fluorescence than in the presence of

Ca2+).

Afurther phe-nomenon is required that occurs only in the presence of Ca2 .In a-toxinof S. aureus(Tomitaetal., 1992), aswell as in certainpore-formingcytolysins (Bhakdi and Tranum-Jensen, 1986), insertion is followed by oligomerization.

However, in the case ofHlyA, no oligomer has been iso-lated up to now, although circumstantial evidence in its favor has been produced (Ostolaza etal., 1993; Ludwig et

al., 1993). No data are currently available on the step(s) between HlyAinsertion and membrane disruption.

The insertion of

c!-hemolysin

in lipid bilayers

A fraction of the membrane-associated a-hemolysin is irre-versibly bound to thebilayer, its Trpresidues appeartobe in

a less polar environment than when the protein is free in

solution,and, under theappropriateconditions (e.g.,presence of

Ca2+),

it causes bilayer leakage and/or disruption. These

experimental facts indicate that this particular fraction has become inserted in the bilayer. Apart from the already dis-cussed implications for the mechanism of toxin action, the

experimental data may be of interest in the framework of peptide insertion in membranes, the first stage in membrane

protein folding (Lemmon andEngelman, 1994).

Unlike the casesofproteintranslocation inprokaryotes (de Kruijff, 1994),anionicphospholipidsare notrequiredforHlyA insertion (Ostolaza et al., 1993). Jain and Zakim (1987) have

pointed out the requirement that hydrophobic regions of the

bilayerbecome transiently exposedto the aqueous phase for proteinincorporation to occur. Such transient exposure would befavored by overallchanges in thebilayer properties,suchas fluidity (Chapman, 1975), or by particular properties of a localized microenvironment, or defects (Jain and Zakim,

1987). The requirement of fluidbilayers for insertionisshown inFig. 1, and moreexplicitlyinFig. 3,where, even if gel and fluidphasesmaycoexist, theproportionofirreversiblybound

protein decreases with the fraction of fluid phase. This is in agreement with thepreferential partitioning ofgramicidinA'

influidphospholipidphases coexistingwith gelphases(Dibble et al., 1993). The uneven distribution of a transmembrane

peptide into different coexisting lipiddomains has also been described by Zhang et al. (1995). Moreover, Polozov et al. (1995) described the insertion of two amphipathic peptides in zwitterionicphospholipid bilayers in the fluid, but not in the

gelstate; thus thepreferencefor fluidbilayers by the peptides

to be inserted appears to be a rather general and predictable

phenomenon.

Defectsand intrinsic instabilitiesin thebilayermay

pro-mote protein insertion because defect sites may accommo-date a protein molecule without inducing additional

ener-lateral compression (Jain and Zakim, 1987). Local defects

may arise from lipid mixtures (McIntosh et al., 1983), or

from the presence ofimpurities (e.g., detergents) or other

proteins, etc. Our experimental results show that conditions under which local defects are likely to occur (e.g.,

disor-dered liquid state; Fig. 6) do favor irreversible protein insertion. The overall effect of cholesterol istomakeprotein insertionmoredifficult, probably because its rigidstructure does not help to accommodate the rough protein surface (Figs. 5 A and 6 A). Note also that cholesterol tends to

increase bilayer thickness (Levine and Wilkins, 1971; Nezil andBloom, 1992) andintegral proteins tendtopartition into domains ofagiven thickness (Bretscher andMunro, 1993).

The properties of cholesterol areevidentinthe

cholesterol-rich ordered liquid phases (Figs. 5 A and 6 A), which

support very little insertion. Almeida et al. (1992) have

explainedthisphenomenon intermsof cholesterol

occupy-ing free volume in the bilayer. It is interesting, in this respect,thatS.aureusa-toxin hexamerization is favoredby

these cholesterol-rich phases (Tomita et al., 1992). How-ever, the relationship between the presence of cholesterol and the reversibility of HlyA binding appears to be more

complex than the general anti-insertion effect just

dis-cussed. The effects in the DMPC/cholesterol mixture at

10°C are agoodexample, particularly the surprisingly low

recoveries ofhemolytic activity in the presence of choles-terol (Fig. 5 B) that do notcorrelate with thebinding data (Fig. 5 A). Moreover, the irreversible binding dataarevery similar for the

Lo

phaseat 10°C and 37°C (Figs. 5A and 6 A), whereas the recovery of hemolytic activity is much

higher at the higher temperature (Figs. 5 B and 6 B). Bretscher and Munro(1993) have proposed that changes in cholesterol concentration in cell membranes may induce

segregation ofdomains withhigh and low cholesterol

con-centrations;thisidea has receivedsupportfrom the

biophys-ical studies of Virtanen etal. (1995) in model membranes. If this were the case, segregation of domains would be favoredbylowtemperatures,andHlyA would partition into

domains of a particular composition, irrespective of the

average cholesterol contents in themixture.

The fact thatprotein insertion itself helpsto promotenew structural defects is important because it would support a

cooperativeprocessofinsertion, which inturnis relatedtothe putativeoligomerization of HlyA (Ostolazaetal., 1993; Lud-wigetal., 1993).Pott and Dufourc(1995) have underlined the

creation ofdefectstructuresinthe membraneupontheaddition ofmelittin thatmightmodify the overall elastic properties of the membrane, andMonette et al. (1993) proposed thathigh cholesterol concentrations inducetight lipidpacking, which in

turn prevents penetration of melittin into the bilayer. In our

system,eventhegelstateismorepronetoHlyA insertion than

the

Lo

one,probably becauseofthestructural defects thatare

knowntoexistinliposomes below

T,.

This workwassupported inpartby DGICYTgrantPB91/0441 and bya HumanCapital andMobilityNetworkfrom theEuropeanUnion.

getically unfavorable general disorder, desolvation, or

1876 Biophysical Journal Volume71 October 1996

REFERENCES

Almeida, P. F. F., W. L. C. Vaz, and T. E. Thompson. 1992. Lateral diffusion in the liquid phases of dimyristoylphosphatidylcholine/ cholesterol lipid bilayers: a free volume analysis. Biochemistry. 31: 6739-6747.

Bhakdi, S., and J. Tranum-Jensen. 1986. Membrane damage by channel-forming proteins. Staphyloccal alpha-toxin, streptolysin-O and the C5B-9 complement complex. J. Immunol. 136:2999-3005.

Bretscher, M. S., and S. Munro. 1993. Cholesterol and the Golgi apparatus. Science. 261:1280-1281.

Brumm,T., K. Jorgensen,0. G. Mouritsen, and T. M. Bayerl. 1996. The effect of increasing membrane curvature on the phase transition and mixing behaviour of a DMPC/DSPC lipid mixture as studied by Fourier transform infrared spectroscopy and differential scanning calorimetry. Biophys.J.70:1373-1379.

Chapman, D. 1975. Phase-transitions and fluidity characteristics of lipids and cell membranes. Q. Rev.Biophys.8:185-235.

Chervaux, C., and I. B. Holland. 1996. Random and directed mutagenesis to elucidate the functional importance of helix II and F-984 in the C-terminalsecretion signal of Escherichia coli hemolysin. J. Bacteriol. 178:1232-1236.

Coote, J. G. 1992. Structural and functionalrelationships among the RTX toxin determinants of Gram-negative bacteria. FEMS Microbiol. Rev.

88:137-162.

deKruijff,B. 1994. Anionicphospholipids andprotein translocation. FEBS

Lett.346:78-82.

Dibble, A. R. G., M. D. Yeager, and G. W.Feigenson. 1993. Partitioning ofgramicidinA'betweencoexisting fluid and gel phospholipid phases. Biochim. Biophys. Acta. 1153:155-162.

Hannavy, K., S. Rospert, and G. Schatz. 1993. Protein import into mitochondria: aparadigm for the translocation ofpolypeptides across membranes. Curr. Opin. Cell.Bio.5:694-700.

Isenman, L., C. Liebow, and S. Rothman. 1995. Transport of proteins

across membranes. Aparadigm in transition. Biochim. Biophys.Acta. 1241:341-370.

Jain, M. K., and D. Zakim. 1987. The spontaneous incorporation of proteins into preformed bilayers. Biochim.Biophys. Acta. 906:33-68. Jorgensen, K., M. M. Sperotto, 0.G.Mouritsen, J. H. Ipsen,and M. J.

Zuckermann. 1993. Phaseequilibria and localstructureinbinary lipid bilayers.Biochim. Biophys.Acta. 1152:135-145.

Knoll, W., K. Ibel, and E. Sackmann. 1981. Smallangleneutronscattering study oflipid phase diagrams by the contrastvariation method.

Bio-chemistry. 20:6379-6383.

Lemmon, M. A., and D. M. Engelman. 1994.Specificity and promiscuity in membranehelix interactions. FEBS Lett.346:17-20.

Levine, Y. K., andM. H. F. Wilkins. 1971. Structure of orientedlipid bilayers.NatureNewBio. 230:69-71.

Ludwig, A., R. Benz, andW.Goebel. 1993.Oligomerization of Esche-richiacolihemolysin (HlyA) is involved in pore formation. Mol. Gen. Genet. 241:89-96.

Mateo,C. R.,A. U.Acufia,and J. C. Brochon. 1995.Liquid-crystalline phases ofcholesterol/lipid bilayers asrevealedbythefluorescence of trans-parinaric acid. Biophys. J. 68:978-987.

Mayer,L.D.,M. J.Hope, and P. R. Cullis. 1986. Vesicles of variable sizes produced byarapid extrusionprocedure.Biochim.Biophys.Acta. 858: 161-168.

McIntosh, T. J., R. V.McDaniel, and J. A.Simon. 1983. Induction ofan

interdigitated gel phase infullyhydratedphosphatidylcholine bilayers. Biochim.Biophys.Acta. 731:97-108.

McMullen, T. P. N., and R. N. McElhaney. 1995. New aspects of the interaction of cholesterol withdipalmitoylphosphatidylcholine bilayers

as revealed byhigh-sensitivitydifferential scanning calorimetry. Bio-chim.Biophys.Acta. 1234:90-98.

Menestrina,G.,C.Moser, S.Pellet,andR.Welch. 1994. Poreformationby Escherichia coli hemolysin (HlyA) and other members of the RTX toxinsfamily.Toxicology. 87:249-267.

Menestrina, G., M. Ropele, M. Dalla Serra, C. Pederzolli, F. Hugo, S. Pellet, and R. A. Welch. 1995. Binding of antibodies to functional

epitopes on the poreformed by Escherichia coli hemolysin in cells and modelmembranes. Biochim. Biophys. Acta. 1238:72-80.

Monette, M., M. R.Van Calsteren, and M. Lafleur. 1993. Effect of cholesterol on the polymorphism of dipalmitoylphosphatidylcholine/melittin complexes: an NMRstudy.Biochim. Biophys. Acta. 1149:319-328. Nezil, F. A., and M. Bloom. 1992. Combined influence of cholesterol and

synthetic amphiphilic peptides upon bilayer thickness in model mem-branes. Biophys. J. 61:1176-1182.

Ostolaza, H., B. Bartolome, I. Ortiz de Zarate, F. de la Cruz, and F. M. Gonli. 1993. Release of lipid vesicle contents by the bacterial protein toxina-hemolysin. Biochim. Biophys. Acta. 1147:81-88.

Ostolaza, H., B. Bartolome, J. L. Serra, F. de la Cruz, and F. M. Gonii. 1991. a-Hemolysin from E. coli. Purification and self- aggregation properties. FEBSLett. 280:195-198.

Ostolaza, H., A. Soloaga, and F. M. Gofii. 1995. The binding of divalent cations to Escherichia coli a-hemolysin.Eur.J.Biochem. 228:39-44. Ostolaza, H., and F. M. Gofni. 1995. Interaction of the bacterial protein

toxin a-hemolysin with model membranes: protein binding does not always leadtolyticactivity. FEBS Lett. 371:303-306.

Piknova, B., D.Marsh, and T. E. Thompson. 1996. Fluorescence quench-ing study of percolation and compartmentalization in two-phase lipid bilayers. Biophys.J. Inpress.

Polozov, I. V.,A.I.Polozova, Y. G. Molotkovsky, G. M. Anantharamaiah, J.P. Segrest, and R. M. Epand.1995. Amphipathic peptide effects on the lateraldomainreorganization of lipid bilayers. Biophys. J. 68:455A. Pott, T., and E. J. Dufourc. 1995. Action of melittin on the

DPPC-cholesterolliquid-ordered phase: asolidstate2Hand3'P-NMRstudy. Biophys. J.68:965-977.

Rytomaa,M.,and P. K. J. Kinnunen.1995. Reversibility of the binding of cytochrome c to liposomes. Implications for lipid-protein interaction. J.Bio. Chem.270:3197-3202.

Sankaram, M. B., D. Marsh, L. M.Gierash,and T. E.Thompson. 1994. Reorganization oflipid domainstructurein membranesbya

transmem-branepeptide: anESRspin label studyontheeffect of the Escherichia coli outer membraneprotein A signal peptideonthefluidphasedomain connectivity inbinarymixtures ofdimyristoylphosphatidylcholineand distearoylphosphatidylcholine. Biophys. J. 66:1959-1968.

Sankaram,M.B., D.Marsh,and T.E.Thompson.1992. Determination of fluid andgeldomain sizes in two-component, two-phaselipid bilayers. Anelectronspin resonancespin label study. Biophys. J.63:340-349. Sankaram, M. B., and T. E. Thompson. 1991. Cholesterol-induced

fluid-phase immiscibility in membranes. Proc. Natl. Acad. Sci. USA. 88: 8686-8690.

Stanley, P., L. C. Packman, V. Koronakis, and C.Hughes. 1994. Fatty acylationoftwointernallysineresiduesrequiredfor the toxicactivityof Escherichia colihemolysin.Science. 266:1992-1995.

Surewicz, W. K., and R. M. Epand. 1984. Role ofpeptidestructure in lipid-peptide interactions:afluorescencestudy of thebindingof penta-gastrin-related pentapeptides to phospholipid vesicles. Biochemistry. 23:6072-6077.

Tomita, R., M.Watanabe, and T. Yasuda. 1992. Influence of membrane fluidityontheassembly ofStaphylococcusaureusa-toxin,a channel-forming protein, in liposome membrane. J. Biol. Chem. 267: 13391-13397.

Vaz, W. L.C., E. C. C.Melo, and T. E. Thompson. 1989. Translational diffusion and fluid domainconnectivityinatwo-component,twophase phospholipidbilayer. Biophys. J. 56:869-876.

Virtanen, J. A., M. Ruonala, M. Vauhkonen, andP. Somerharju. 1995. Lateral organizationofliquid-crystallinecholesterol- dimyristoylphos-phatidylcholine bilayers. Evidence for domains with hexagonal and centered rectangular cholesterol superlattices. Biochemistry. 34: 11568-11581.

Vist, M. R., and J.H.Davis. 1990.Phase-equilibriaof cholesterol dipalmi-toylphosphatidylcholine mixture. 2H-Nuclear magnetic resonance and differentialscanning calorimetry. Biochemistry. 29:451-464.

Zhang,Y.P., R. N. A. H. Lewis,R. S.Hodges, and R. N. McElhaney. 1995.Peptidemodels of helicalhydrophobictransmembrane segments of membraneproteins. 2. Differentialscanningcalorimetric andFrIR spectroscopicstudies of theinteraction ofAc-K2-(LA)12-K2-amidewith phosphatidylcholinebilayers. Biochemistry. 34:2362-2371.