How SNARE molecules mediate membrane fusion: Recent insights from molecular simulations

(1)

How SNARE molecules mediate membrane fusion: Recent insights from molecular simulations

Herre Jelger Risselada and Helmut Grubmu¨ller

SNAREmoleculesarethecoreconstituentsoftheprotein machinerythatfacilitatefusionofsynapticvesicleswiththe presynapticplasmamembrane,resultinginthereleaseof neurotransmitter.Onamolecularlevel,SNAREcomplexes seemtoplayaquiteversatileandinvolvedroleduringallstages offusion.Inadditiontomerelytriggeringfusionbyforcingthe opposingmembranesintocloseproximity,SNAREcomplexes arenowseentoalsoovercomesubsequentfusionbarriersand toactivelyguidethefusionreactionuptotheexpansionofthe fusionpore.Here,wereviewrecentadvancesinthe

understandingofSNARE-mediatedmembranefusionby molecularsimulations.

Address

TheoreticalMolecularBiophysicsGroup,Max-Planck-Institutefor BiophysicalChemistry,Go¨ttingen,Germany

Correspondingauthor:Grubmu¨ller,Helmut(hgrubmu@gwdg.de)

CurrentOpinioninStructuralBiology2012,22:187–196 Thisreviewcomesfromathemedissueon

Theoryandsimulation

EditedbyJuanFernandez-RecioandChandraVerma Availableonline23rdFebruary2012

0959-440X/$–seefrontmatter

#2012ElsevierLtd.Allrightsreserved.

DOI10.1016/j.sbi.2012.01.007

Membrane fusion is a fundamentalprocess in cell bio- physics,beinginvolvedinviralinfection,endocytosisand exocytosis,andfertilization.Overthepast30yearsithas becomewidely acceptedthatfusionproceedsthrougha hemifusionstatewhereaninitialhour-glass-shapedlipid structure, the so-called fusionstalk, is formedbetween theadjacent membrane leaflets[1^!,5,6^!].Thisstructure eventually evolves into a fusion pore [2^!–6^!]. SNARE molecules are the core constituents of the protein machinerythatfacilitatesfusionofsynapticvesicleswith the presynaptic plasma membrane, resulting in the release of neurotransmitter [7]. It remains unclearhow SNARE complexesaltertheenergylandscapeof fusion and steerthe transition from thelipidic stalk to fusion pore on a molecular level [4]. Here, we review recent advances in the understanding of SNARE-mediated membrane fusion bymolecular simulations.Because of computationalcostrequiredtocoverthelengthandtime scaleofSNARE-mediatedmembranefusion,simulations

were typically applied to explore the membrane partitioning and dynamics of individual SNARE fractions [8,9,2^!,3^!],thestabilityofthecoiled-coilcomplex[10^!], orusingasimplifiedrepresentationoftheSNAREcom- plex tofuse twosmallvesicles[11].

Recently,SNARE-mediatedfusioneventsbetweentwo vesicles have also been simulated at near-atomic resol- ution [12].Here, acoarse-grained model[13,14],where severalatomsarerepresentedbyasingleinteractionsite, wasusedtobothcapturetheatomisticcharacteristicsof theneuronalSNAREcomplexsolvedbyX-ray[15^!],such as secondarystructure, andto simultaneouslyovercome the computational expense involved with SNARE- mediated membrane fusion. Combining the available experimental andsimulationresults, wewill attemptto draw a consensus picture of how SNARE molecules might overcome the various lipidic fusion barriers and, especially, toidentifythesebarriers.

Mechanical couplingbetween theSNARE complex and thebilayer

Atomistic simulations have revealed that the partly assembledcoiled-coilcomplexformsaconsiderablystiff platform that allows force transduction between the SNAREcomplexandthemembraneviathetransmem- branedomains(TMDs)oftheSNAREmolecules[10^!].

During SNARE zipping a considerable fraction of the releasedenergyisexpectedtodissipate,andtheremain- ing fractionis storedas molecularbending stressin the individual SNARE molecules. This mechanical stress playsanimportantroleintheself-organizedarrangement at the fusion site and is minimized when the SNARE complexislocatedattheperipheryofthecontactregion andtherebyallowsacloserproximitybetweenthemem- branes (Figure 1a–c) [12]. A central positioning of the SNARE complex, where its excluded volume would rather hindersuchcloseproximity(Figure 1c), requires strongerbendingoftheSNAREmoleculesandisthere- foremostlikelyunfavorable.

Tobeabletoefficientlyexertforceonthemembrane,the semi-flexiblelinkersofboththeSNAREmoleculesboth need to be sufficiently stiff. Despite a non-conserved a-helical structure upon bending, atomistic studies suggested a stiffness for the syntaxin linker of 1.7–

50calmol^"¹deg^"² (11#0.2calmol^"¹deg^"² for the

coarsegrainedSNAREmodel[12])[8].Whereassyntaxin is readily a-helical before SNARE complex formation, synaptobrevinisunstructured[16].Figure1fillustratesa

(2)

scenariowherethesynaptobrevinlinkerremainsunstruc- turedduringSNAREzipping. Insuchacase, thetrans- missionofforcetowardthemembranewouldbeimpaired because the bending stress that is otherwise stored in syntaxinnowalternativelyrelaxesbyadditional‘kinking’

of the more flexible synaptobrevin. Such ‘kinking’ of synaptobrevinwouldimposeanadditionalbarrieragainst subsequent a-helical nucleation [15^!,17] and the pro- gressionofSNAREzipping[15^!].Thus,itseemsessen- tial thata-helical nucleationin synaptobrevin precedes SNAREzipping.

Stalkformation

Whenthemembranesarebroughtintosufficientlyclose proximity,a stalkcan be formed [1^!].Here, continuum descriptionstendtoimplyatransientstalkofinfinitesimal radiusbeforeitexpandsintothewellcharacterizedhour- glass-shapedstalkstructure—incontrasttothemolecular natureofthelipidmembrane.Infact,thereisagrowing bodyofevidencefromrecentmolecularsimulationsthat stalkformationisneithertheinitialnortheratelimiting

step in membrane fusion [3^!]. These simulations show stalk intermediates, such as a single or several splayed lipids connecting the adjacent monolayers (Figure 1d), whichleadto rapidstalkformationwithin severalnano- seconds[12,18^!–26^!].Inaway,onemightrefertosucha splayedlipid(orasimilarperturbation)asessentiallythe smallest‘stalk’thatispossibleinmolecularterms.Impor- tantly,oncesuchsplayedlipidstateisreached,formationof theactualstalkisenergeticallydownhill(i.e.spontaneous).

Crucially,andincontrasttopreviousviews,inthisscenario thestalkstructuredoesnotrepresenttheratedetermining barrier,butrathera(local)freeenergyminimum.Thisview is strongly supported by the observation of stalk-like structuresinrecentX-raystudies[27^!,28^!].Oneimportant consequenceisthatitistheenergybarrierassociatedwith thesplayedlipidstatethatdeterminesthekineticsofstalk formation,andwhichthereforeneedstobeovercomeby thefreeenergyreleasedbySNAREcomplexformation(cf.

Figure2).Further,thefreeenergyofthemetastablestalk can either be larger [20^!,29^!!] or, in the rhombohedral phase or stalk phase [27^!,28^!,30^!!], even slightly lower

Figure1

(a) (c) (e)

(b) (d) (f)

Current Opinion in Structural Biology

TheSNAREcomplexbeforefusion.(a)and(b)SNAREzipping(arrows)bringstwovesiclesincloseproximity.(c)Flexiblelinkers(coloredblack)allowa centralpositioningoftheSNAREcomplexatthefusionsiteandtherebyhinderacloseproximity.(d)Splayedlipidintermediateformedattheonsetof stalkformation(coloredcyan)(e)SNARE-inducedcurvatureinthetargetmembrane.Notethehydrophobicnatureoftheadjacentfusionsites(colored cyan).(f)Examplewheretheforcetransmissiontothetargetmembraneisimpairedbythepresenceofunstructured(i.e.flexible)synaptobrevin (coloredblue)linkers.

CurrentOpinioninStructuralBiology2012,22:187–196 www.sciencedirect.com

(3)

than that of theunfused bilayers. A stalk ismetastable because it simultaneously faces a barrier to dissociate [20^!,29^!!]andtoexpand[30^!!].

The free energy of the stalk intermediate (i.e. stalk barrier)ismostdirectlydeterminedbyboththedistance between thefusing leaflets and thelength/solubility of thelipidtails[31^!,32^!,3^!,23^!,20^!],andonlyindirectlyby lipid shape or, equivalently, spontaneous curvature [31^!,32^!,3^!,23^!,20^!].Thus, one way to promote thefor- mationofasufficientlylowenergystalkintermediateisto bring the adjacent leaflets within a critical distance.

Indeed,weconsideritnotacoincidencethatessentially allfusogenicconditions,suchasanegativespontaneous curvature, tension/osmotic pressure,and positivemem- brane curvature (curvature stress), lower the energy of leaflet approach[33^!–35^!].All thesefactorsincrease the hydrophobicityofthefusionsite,andtherebylowerinter- membranerepulsionbyalsoreducingtheenergeticcost

ofsolventremovalbetweentheapproachingleaflets.The positive membrane curvature in vesicles and ‘dimples’

decreasethisrepulsionevenfurtherbyalsoreducingthe exposedareaof thefusionsite.

In SNARE-mediated simulations, stalk formation requires a minimumdistance between theheadgroups of the opposing bilayers of only about 1nm. Similar distances were found by recent X-ray studies of stalk- phaseformation[36]aswellasbyothersimulationstudies [23^!].Letusassumethatmostoftheenergyavailablefor subsequentfusionisstoredin thestifflinkerregionsof the SNARE complex. Because the angle between the transmembrane domains (TMDs) is about 1208 [12], basedonatomisticsimulation-derivedvaluesofthelinker stiffness[8],mechanicalenergyofupto$10k_BTisstored ineachSNAREcomplexduringstalkformation.Onlya smallfractionofthisenergy,however,isreleasedduring stalk formation, because the angle between the TMD (a)

(b)

∆G DOPE

negative positive

stalk intermediate

Unfused bilayers

unstable stalks

metastable stalks

90º

elongated stalks Rombohedral phase (R)

free energy

free e xpansion inverted hexagonal phase (HII) DOPC

(c)

(d)

(e)

90º

expan^sion

Energeticsofstalkformation.(a)Unfusedbilayers.(b)Ataconstantinter-membranedistance,thefreeenergyofthesplayedlipid(DG)isnotexpected todependonthehead-grouptype(spontaneouscurvature).(b),(c)and(e)Thefreeenergyofthestalkitself,however,dependsonboththe spontaneouscurvature(plottedcone)andseparationdistanceandthesewilldeterminewhetheraformedstalkismetastableornot[31^!].(c)Atsuch aninter-membranedistance,thestalkformedbetweentwoDOPCmembranesisanunstabletransientintermediate.(d)Thestalkismetastablewhen itsfreeenergyislowerthanboththedissociationbarrierandexpansionbarrier.(e)IntheHII-phasetheformedhourglass-shaped(rhombohedral)stalk structureisunstable(noexpansionbarrier)andthestalkwillcontinuouslyelongatetomaximizeitsenergeticallyfavorablenegativelycurvedperimeter.

Thebarrierofstalkformationisdeterminedbythefreeenergyofthestalkintermediate(DG).However,thisbarrierexcludestheadditionalenergythatis requiredtobringthetwomembraneswithinsuchproximity(inter-membranerepulsion).Thisinter-membranerepulsionincludestheknownrelation betweenspontaneouscurvatureandfusogenicity.Hence,anintrinsicnegativecurvaturereflectstheinabilityofthelipidhead-grouptosufficiently shielditshydrophobicmoietywhenarrangedinaplanarconformation.Asaresultthemembranesurfacebecomesmorehydrophobicwhichlowers theenergeticcostofleafletapproach.

(4)

remainsnearly constantduring thatprocess [12].More- over,stalkformationwasalsoobservedinoursimulations whenforcetransmissiontothemembraneswasblocked by artificially ‘freezing’ the bent conformation of the SNAREmolecules.Theseresultssuggestthattheenergy releasedbySNAREzippingaswellasthelinkermainly serves to bring theopposing leaflets withincritical distance—initiallybyanoveralltranslationoftheopposing bilayers(Figure1a,b),andlaterlikelybylocalmembrane surfaceperturbation, for example, byforming adimple (Figure1e).Indeed,experimentshaveindicatedthatthe

TMDs play an important role in stalk formation. For example, SNARE molecules where the TMDs are replacedbylipidsdonotfacilitatefusion[37],whereas isolated TMDs are able to induce fusion [38^!]. Also recent simulations suggest an intrinsic propensity of TMDstobothperturbthelipidpacking(enhancinglipid splaying)andtherebyincreasethehydrophobicity(redu- cingrepulsions)ofthefusionsite[12].Accordingtothis new consensus picture that emerges, stalk formation requires (a) proximity and (b) is driven by reducing themain free energybarrier atthe splayedlipid state.

Figure3

Stalk evolution

inverted micelle

intermediates (IMI) expanded hemi-fusion diaphragm (HD)

hemi-fusion stalk

meta-stable stalk

fusion pore initial stalk

Stalk elongation Protein inhibited stalk elongation metastable leakage pore stalk thinning/

widening POPE

SNAP-25 Sx1a Syb2 TMD

POPC CHOL

(a)

(b)

(e)

(f)

(g) (c)

(d) (h)

(i)

(j)

(k) Stalk instability

90º

DifferentpathwaysobservedintheSNARE-mediatedfusionbetweena20nm-sizedvesicleandlipidbilayerofvaryingcomposition.(a)Initialstalkin thepresenceoffour[74]SNAREcomplexes.(b)Stalkelongation(HII-phaseregime).(c)Proteininhibitedstalkelongation(HII-phaseregime).(d) Metastablestalk(POPCtargetmembrane).Theformedstalkfacesasubstantialbarrieragainstexpansion[30^!!,40^!!]andprogressionoffusiondoes notoccurinthecourseofthe6mssimulations.(e)Invertedmicelleintermediate(IMI).Progressionoffusiondoesnotoccurinthecourseofthe6ms simulations.(f)Non-leakystalkexpansion/wideninginthepresenceof40%cholesterol[75].Sterolsseemtospecificallyenhancenon-leakyfusion [51^!].(g)LeakystalkexpansioninapurePOPEsystem.(h)Spontaneousstalk-poretransition.(i)Expandedhemifusiondiaphragm(HD).(j)Pore formedattherimoftheHD.(k)Toroidalfusionpore.

(5)

purpose,inparticularthoseinducedbyTMDs.

Fusionpathways: thegood, thebad, theugly Afterstalkformation,simulationshavesuggestedmainly threepathwaysthroughwhichSNARE-mediatedfusion canproceed[2^!,3^!].Asanexample,Figure3showsthese different pathways in the SNARE-mediated fusion be- tweenalipidbilayerand20nm-sizedvesicle,represent- ingasynapticvesicle (30–50nmsized[39^!]):

(i) Stalk elongation (Figure 3b,e). Stalk elongation relates to the inverted hexagonal phase transition (HII-phase)[30^!!,40^!!–42^!],whereastackedbilayer system spontaneously rearranges into inverted hexagonally packed cylinders. In the presence of sufficiently (small) rigid vesicles, however, the topology is limited to that of an inverted micelle (IMI)[12,40^!!,2^!,3^!](Figure3e).

(ii) Stalk widening (Figure 3c,f). A radial expansion (widening) can only occur when the stalk is metastableand,bydefinition,facesabarrieragainst expansion[30^!!,40^!!].Duringstalkwideningthecis- leafletswilleventuallymeet(blackbar,3f)toeither form a metastable single bilayer, that is, the hemifusion diaphragm (Figure 3i), or to sub- sequently rupture (Figure 3h). Importantly, such wideningfacesaconsiderablefreeenergybarrierof 15–63kBT [30^!!]. This barrier can be reduced by increasingthenegativespontaneouscurvatureofthe cis-leaflets[43];however,suchincreaseislimitedby theoccurrenceoftheHII-phasetransitionbecauseat thispointthestalkwillspontaneouslyelongate(IMI- pathway)ratherthanexpandradially[30^!!,40^!!,43].

Inthelattercase,stalkelongationmightbeimpaired byacircumferenceof multipleSNAREcomplexes that‘lock’ thestalk(Figure 3cvsb).

(iii) Leakyexpansion(Figure3g).Onthispathway,the formedstalkismetastableandfurtherexpansionfaces a substantial barrier [40^!!], that is, the barrier of competitivestalkwideningistoolarge[44^!,40^!!,2^!].

Althoughinvivosynapticfusionitselfisbelievedtobe anon-leakyprocess,invitroSNARE-mediatedfusion hasoftenbeenobservedtobetransientlyleaky[45–

48].Likestalkwidening,alsotheleakyexpansioncan either stabilize a hemifusion diaphragm or instan- taneouslyprogressintoafusionpore[3^!,2^!].Thereis accumulatingevidence thatthepresence of astalk dramaticallyincreases the probability to nucleatea porein its directvicinity (stalk-pore complex), and vice versa [2^!,40^!!,44^!,49], that is, the favorable partitioning of a stalk atthe rim of a pore renders thefreeenergyof theresultingstalk-pore complex lowerthanthatofisolatedstalksorpores[44^!,40^!!,2^!].

Theprobability to formsuchstalk-pore complexis

lipids [40^!!,47^!]—quite in contrast to a ‘usual’

membraneporethatwouldratherbefavoredbythe presenceofpositivelycurvedlipids[6^!].

If the latter scenario were true, one would expect that transientleakage,thatis,porenucleation,canbereduced orevenbeinhibitedbyatleasttwomodifications.(i)The opposing fusion-sites are made attractive, rather than repulsive (e.g., the presence of counter ions or hydro- phobicpolymers)[26^!,25,50^!].Inthatcase,porenucleation would reduce thefavorable membrane–membrane contactareaatthefusionsite.Molecularsimulationshave suggested that these attractionsfacilitate an alternative adhesion-condensationmechanismwhereasinglemem- brane-thick hemifusion-diaphragm is instantaneously formedbyanon-leakylateralreorganizationofthelipids [26^!,25,50^!].(ii)Thepresenceofhydrophobicmembrane ordering/strengtheningmoleculessuchassterols[51^!]or carboxylic acids [18^!] might specifically oppose pore nucleation. Notably, sterols are ubiquitous in the presynaptic plasmamembrane[39^!].

Formation ofthe fusionpore

Fusionporeformationseemsrelatedtoboththepresence ofmechanicalstressintheSNAREcomplexaswellasthe TMRends. Onone hand,thebinding affinitybetween SNARE molecules and the nature of the linkers have been linked to fusion pore formation [52^!,53]; on the otherhand,deletionsofTMRendresiduesandaddition of polaraminoacidsto theSyb2C-terminus havebeen showntoarrestfusionporeformation[54,55^!].Further,in agreement with currentmodels [4,56^!,57], conductance measurementssuggestedthatthenegativelychargedC- termini of the TMDs reside in or near the appearing fusionpores[58^!].Duringhemifusionbothsynaptobrevin and syntaxin can only release their mechanical stress through widening of the stalk or formation of a fusion pore,duetothelargebarrierthatpreventspenetrationof the chargedC-termini in theTMRs throughthemem- branes (Figure4).Thisobservationsuggeststhat(i)the mainactionofSNAREcomplexesistoactivelypromote fusion pore opening, (ii) this process is driven by the stored mechanical stress in the SNARE complex that resulted from SNARE zipping, and (iii) the C-termini playanessentialroleintheunderlyingmechanism[12].

Asaresult ofthemechanicalstress storedwithinin the SNAREcomplex,theC-terminiexerta‘squeezing’force on the trans-leaflets [12] of the hemifused membranes.

Because stalk widening goes hand-in-hand with simul- taneousstalkthinning(Figure4),thisforceenhancessuch process.Self-consistentfieldcalculations havesuggested thatstalkwideningcanrequireevenmorefreeenergythan stalk formation [30^!!], which is also supported by our simulations (Figure 3d) as well as those of others

(6)

[3^!,2^!,18^!].Thisfindingmightprovideanexplanationwhy muchoftheconservedmechanicalstressintheSNARE complexisreleasedduringsuchprocess(Figure4a).

Thisidea alsoraises thequestion, how muchforcethe SNARESactuallyexerttodrivestalkexpansion?Because the distance between the separated C-termini is about8nm(i.e. the length of the stalk),and assuming that most of the stress arises from bending of the SNARE linkers (1208) with a maximal linker stiffness

of50calmol^"¹deg^"²[8],oneobtainsanaverageforceof

5pN per SNARE complex. In this estimate we have allowed for an additional energy of ca. <10kBT that is releasedwhentheTMDseventuallyassemble[15^!,59].

Withthe a-helicalTMDs beingsufficiently stiff, however,thisfreeenergyislikelytobeonlyavailablewhen thebarrierof stalkwideningisreadily surpassed[12].

Further,thesqueezingactionthattheTMDC-terminiof partlyassembledSNAREcomplexesexertonthetrans- leafletslocallythinstheleaflet(createsawell)andresultsin ahydrophobicmismatch.Thelatterenforces,asidefrom intrinsicTMDattractions[38^!,60,15^!],additionalnearC- termini attractions between the TMDs of multiple SNAREcomplexes[38^!,60,15^!](Figure3a,f).Wewilllater rationalize why these C-terminal attractions might be important.

Afterthebarrieragainststalkexpansionisovercome,fusion pore formationproceeds—either spontaneously(Figure 4) [61–63^!,24^!,11], or by an activated non-instantaneous transition via the formation of a metastable hemifusion diaphragm(HD,Figure5)[3^!,18^!,26^!,25,50^!,62,63^!].The

co-existenceofthesedifferentpathwayswouldresultina heterogeneousfusionkinetics[47^!].MetastableHDshave beenobserved in model membranefusion [64^!,65^!,35^!].

Further,theproximityoftheinitialporetotherimofthe HD[26^!,25,50^!,65^!]suggeststhat,despitethelargesizeof anHD,onlyalimitednumberofthepotentiallyavailable SNAREcomplexescanactively participateinnucleating suchrim-pore(Figure5).Incontrast,inthe spontaneous mechanism, all available SNARE complexes can simul- taneouslyreleasetheirmechanicalenergy,andthuscoop- erativelycontributeto(toroidal)fusionporeopening[12]

(Figure4).

OnceaHDhasbeenformed,anddespitebeingthermo- dynamically less stable than the toroidal fusion pore [25,30^!!,50^!],theHDiskineticallystabilizedbyakinetic barrieragainsttheformationofarimpore.Similartothe leakinessdiscussedabove,suchbarrierislikelyincreased bysterols[66^!–68^!]and reduced by,for example, poly- unsaturatedlipids[69]orlyso-lipids[70].Duringexpan- sionoftheHD,itsinitialtensionrelaxesand,therefore, the probability of rim-pore nucleation and subsequent expansiondecreases(Figure 5f).

IftheexpandedHDisinfactametastablefusioninter- mediatewithaslow escaperate[64^!,66^!,35^!,26^!,25,50^!], thisstateneeds tobecircumventedbythefastsynaptic fusionmachinery.Indeed,simulationssuggestanefficient strategytobothcatchandrupturesuch‘slipperyeel’:Ifthe stalkiscircumventedbymultipleopposingSNAREcom- plexes, the presence of the above mentioned near C- terminiattractions may notonly limitfurther expansion oftheHD,butalsoredirecttheseexpandingforcesintoa

Figure4

(a) (b) (c)

Detailedlookatthenon-leakyspontaneousstalktoporetransition.(a)SchematicillustrationofaSNARE-mediatedstalkexpansion.Thecirclesdepict hydrophilicregionsthatinteractwiththestalk.BluecirclesdepicttheN-terminalandredcirclestheC-terminalregionsofthetransmembrane domains.ThezippingactionoftheSNAREcomplexesallowswidening(bluecircles)andsimultaneouslyenforcesthinning(redcircles)ofthestalk.In thepresenceofnearC-terminaltransmembraneattractionsbetweenopposingSNAREcomplexes(redcircles),themotionbetweentheblueandred circlesbecomecoupled(blackframe-work).Insuchacase,expansion(bluecircles)ofaformedhemifusiondiaphragmwoulddirectlytranslateintoan additionalsqueezingforce(redcircles)thatenhancesprematurerupturingofsuchdiaphragmbeforeitcanexpandtoalessstressedandmorestable size.(b)Ruptureoccurswhenthestalksufficientlywidens/thins.Here,theC-terminislipthroughtheinitialpore.(c)Fusionpore.

(7)

‘squeezing’forcethatenforcessimultaneousthinningand ruptureoftheHD(Figure4).Thus,theTMDattractions between multiple SNARE complexes seem to play an important rolein preventingtheexpansion andstabiliz- ationofanHD,whichwouldotherwiseretardfusionpore formation[38^!,60].Thisideacanalsoexplainwhyfastin vivosynapticfusionrequiresatleastthreeSNAREcom- plexes,whereasfusiononlyrequiresone[71^!,72^!].Further, tobeabletohinderHDexpansion,theTMDN-termini needtoarrangeatoppositesidesofthestalk/HDsuchthat theirTMDspointawayfromeachother(Figure4).The presenceofN-terminalattractions[73^!]wouldopposesuch positioningandratherenhanceaparallelpositioningwhere theTMDsarealignednexttoeachother(Figure5),such thattheHDcouldescapeviaexpansion.Asapossibletest, therefore, the presence of N-terminalattractions should retardfusionporeformationbytrappingthefusionreaction atthestageofanexpandedHD.

Conclusions

Inthelastdecenniamolecularsimulationshaverevisited the original stalk-pore hypothesis from different angles.

Initially, and mainly motivated by continuum elastic models,itwasthestalkthatwasbelievedtobetherelevant fusionbarrier.Accordingly,themainfocuswasonpredict- ingthefreeenergyofthestalkstructure.Thediscoveryof therhombohedralphase(stalk phase)byX-rayexperiments in 2003 proved that thestalkcanbeastable structure,that is,

afreeenergyminimum[27^!].Inthesameyear,molecular simulationssuggestedthatstalkformationisprecededby theformationofsplayedlipidintermediates[19^!].Today, therefore,thefocushasshiftedtofurtheridentify,quantify, and structurally characterizethe intermediatesand molecular free energy barriers involved in stalk formation [3^!,20^!,22^!,21^!,23^!,31^!,32^!]. In addition, the importance of subsequent fusion barriers, that is, the expansion of thestalk[2^!,3^!,30^!!,40^!!]andtheformationofafusionpore [64^!,66^!,35^!,26^!,25,50^!],hasnowbeenrecognized.

This paradigm shift has also changed our view on SNARE-mediatedmembranefusion.Intheoriginalscaf- foldmodels,theroleofSNAREmoleculeswasconfined to bringing the membranes into close apposition by exerting mechanical force to overcome the activation energy barrier [4]. In particular, the SNARE complex was notassumedto beinvolvedinthetransitionstates, whichwereconsideredtobeexclusivelylipidic[4].The emerging view today is that such simple and clear-cut separation between the role of the SNARE complexes and that ofthe purelipid membrane misses theirclose coupling,whichturnsouttobeessentialforaquantitative understanding ofSNARE-inducedmembranefusion.

First, whereas theSNARE complexesindeedbring the twomembranesintocloseapposition,itisthelipidsthat determine the mechanical energy which is required to

(a) (b) (c) (d) (e) (f)

(a)–(e)SNARE-mediatedporeformationneartherimofan8-nmsizedexpanded/equilibratedPOPEhemifusiondiaphragm(HD).Therimporeremains stableovera1ms(flickeringstage)beforeiteventuallyexpandsintothe(toroidal-shaped)fusionpore.Rim-poreexpansionrequirestheabsorptionof excessmaterialfromtheremainingHD.Inturn,poreclosureisopposedbythepresenceofthenegativelychargedTMDC-terminiwithintheporelumen [58^!].(f)Structureandenergeticsofarim-pore(POPC:Cholesterolmixture).Arim-pore(top-viewontheHD)iscomposedofanenergeticallyfavorable toroidal-shapedfusionporeedge(bluecoloredline)andacostlymembraneedge(redcoloredline)andadoptshalfacircularshapetominimizeitsfree energy.Thefreeenergyofarimporeisconsiderablylowerthanthatofa‘usual’circular-shapedmembranepore,whichisonlycomposedofthecostly membraneedge.Thefreeenergyofsuchrim-poreisgivenby,F(Lfp,Le,A)=(lfp"lhd)Lfp+Lele"sA,whereLarethelengthsandlthelinetensions (forces)ofrespectivelythefusionpore(fp),hemifusiondiaphragm(hd)andmembraneedge(e)fractions,Aistheareaoftherim-pore,andsthetension present.Belowthecriticalsize(blackline),expansionisopposedbyacostlyincreaseinLe(le>0),andisfavoredbyanincreaseinLfp(iflfp"lhd<0,that is,theHDisthermodynamicallyunfavorable)andA(ifs>0,thatis,theremainingHDisundertension).BystanderSNAREcomplexesthatexertforce/

stressontheremainingHDrim(greencoloredline),therebyincreasinglhd,contributetotheexpansionoftherim-pore.

(8)

overcome the inter-membrane repulsion. The smaller this required energy, the closer the resulting distance betweentheopposingleaflets,andthusthelowerthefree energy of the formed stalk intermediate [31^!,3^!,20^!, 23^!,36].Allfusogenicconditions,suchasanegativespon- taneouscurvature,tension/osmoticpressure,andpositive membranecurvature(curvaturestress),lowertheenergyof leafletapproachandtherebylowertheinitialfusionbarrier [33^!–36].Thenewnotionthatitisnotthestalk,butrather thesplayedlipidstatethatformsthemainfusionbarrier, requires a reconsideration of previous explanations of fusogenicityintermsof stabilizationoftheoverallnega- tivelycurvedstalkstructure.Infact,thisnewratelimiting barrierseemstobemainlydeterminedbytheinter-membranedistance[33^!,31^!,3^!,20^!,23^!,36].

Second,whereasthetransitionstatesinmembranefusion are lipidic, experiments and moleculardynamics simu- lationssuggestthat theSNARE complexesare actively involvedin severalways: (i) Theformationof thestalk intermediateseemsfacilitatedbytheinherentpropensity oftheSNARETMDstodistortthepackingofthenearby lipid tails [12,38^!,37]. (ii) Expansion of the stalk [30^!!] seemstobedrivenbymechanicalstresswhichisreleased bythepartlyassembledSNAREcomplexes.(iii)SNARE complexesmightpreventboththeformationandexpan- sion of a metastable hemifusion diaphragm that would otherwiseimpedethesubsequentopeningof thefusion pore [64^!,66^!,35^!,26^!,25,50^!]. (iv) Both the mechanical stress storedin theSNARE complex and theresulting membranepenetration ofthenegatively chargedTMD C-terminiareassociatedwiththeopeningofafusionpore [12,55^!,54,52^!,53,58^!].

In summary, SNARE complexes seem to play a quite versatileandinvolvedroleduringallstagesoffusion.In additiontomerelytriggeringfusionbyforcingtheoppos- ingmembranesintocloseproximity,SNAREcomplexes arenowseentoalsoovercomesubsequentfusionbarriers and to actively guide the fusion reaction up to the expansion of the fusion pore. Suchfunctional diversity of just one core complex is staggering, but may also explain why so many more molecular components are recruitedin theprocess [38^!,37,55^!,54,52^!,53,58^!,4,56^!].

Acknowledgments

WethankReinhardJahn,MarcusMu¨ller,TimSalditt,SebastianAeffner andYuliyaSmirnovaforstimulatingdiscussionsandconstructivecomments.

FinancialsupporthasbeenprovidedbytheDFGundergrantSFB803/B2.

References andrecommendedreading

Papersofparticularinterest,publishedwithintheperiodofreview, havebeenhighlightedas:

! ofspecialinterest

!!ofoutstandinginterest 1.

!

KozlovskyY,KozlovMM:Stalkmodelofmembranefusion:

solutionofenergycrisis.BiophysJ2002,82:882-895.

Aninfluentialpaperwerethestalkisconsideredafusionbarrier.

2.

!

SchickM:Membranefusion:theemergenceofanew paradigm.JStatPhys2011,142:1317.

Aclearlywrittenandeasytounderstandreviewpaperaboutalternative pathwaysinmembranefusion.

3.

!

MarkvoortAJ,MarrinkSJ:Lipidacrobaticsinthemembrane fusionarena.CurrTopicsMembr2011,68:259-294.

A very comprehensive overview on membrane fusion molecular simulationstudies.Averycompleteliteratureoverviewofthemolecular simulationstudiesofmembranefusion.

4. JahnR,Grubmu¨llerH:Membranefusion.CurrOpinCellBiol 2002,14:488-495.

5. ChernomordikLV,KozlovMM:Protein-lipidinterplayinfusion andfissionofbiologicalmembranes.AnnuRevBiochem2003, 72:175-207.

6.

!

ChernomordikLV,KozlovMM:Mechanicsofmembranefusion.

NatStructMolBiol2008,15:675-683.

Aninfluentialreviewpaperthatdiscussestheconservedfusion-through- hemifusion pathway of merger between biological membranes and proposethat theentire progression, from the closejuxtaposition of membranebilayerstotheexpansionofafusionpore,iscontrolledby protein-generatedmembranestresses.

7. WeberT,ZemelmanBV,McNewJA,WestermannB,GmachlM, ParlatiF,So¨llnerTH,RothmanJE:SNAREpins:minimal machineryformembranefusion.Cell1998,92:759-772.

8. KnechtV,Grubmu¨llerH:Mechanicalcouplingviathemembrane fusionsnareproteinsyntaxin1A:amoleculardynamicsstudy.

BiophysJ2003,84:1527-1547.

9. DurrieuMP,BondPJ,SansomMSP,LaveryR,BaadenM:

Coarse-grainsimulationsoftheR-SNAREfusionproteininits membraneenvironmentdetectlong-livedconformational sub-states.ChemPhysChem2009,10:1548-1552.

10.

!

DurrieuMP,LaveryR,BaadenM:Interactionsbetweenneuronal fusionproteinsexploredbymoleculardynamics.BiophysJ 2008,94:3436-3446.

AtomisticmolecularsimulationstudyoftheSNAREcomplex.TheSNARE core complex behaves like a stiff rod, with limited conformational dynamics.

11. WuS,GuoH:Simulationstudyofprotein-mediatedvesicle fusion.JPhysChemB2009,113:589-591.

12. RisseladaHJ,KutznerC,Grubmu¨llerH:Caughtintheact:

visualizationofsnare-mediatedfusioneventsinmolecular detail.ChemBiolChem2011,12:1049-1055.

13. MarrinkSJ,RisseladaHJ,YefimovS,TielemanDP,deVriesAH:

TheMARTINIforcefield:coarsegrainedmodelfor

biomolecularsimulations.JPhysChemB2007,111:7812-7824.

14. MonticelliL,KandasamySK,PerioleX,LarsonRG,TielemanDP, MarrinkSJ:Themartinicoarsegrainedforcefield:extensionto proteins.JChemTheoryComput2008,4:819-834.

15.

!

SteinA,WeberG,WahlMC,JahnR:Helicalextensionofthe neuronalSNAREcomplexintothemembrane.Nature2009, 460:525-528.

Here,itisshownthatthezippingoftheSNAREsincludestheTMDs.

16. FasshauerD,OttoH,EliasonW,JahnR,Bru¨ngerA:Structural changesareassociatedwithsolubleN-ethylmaleimide- sensitivefusionproteinattachmentproteinreceptorcomplex formation.JBiolChem1997,272:28036-28041.

17. EllenaJF,BinyongL,WiktorM,SteinA,CafisoD,JahnR, TammLK:Dynamicstructureoflipid-boundsynaptobrevin suggestsanucleation-propagationmechanismfortrans- SNAREcomplexformation.ProcNatlAcadSciUSA2009, 106:20306-20311.

18.

!

BaoukinaS,TielemanDP:LungsurfactantproteinSP-B promotesformationofbilayerreservoirsfrommonolayerand lipidtransferbetweentheinterfaceandsub-phase.BiophysJ 2011,100:1678-1687.

Protein-mediatedfusionmechanismsstudiedwithanear-atomicmodel.

19.

!

StevensMJ,HohJH,WoolfTB:Insightsintothemolecular mechanismofmembranefusionfromsimulation:evidencefor theassociationofsplayedtails.PhysRevLett2003,91:188102.