common principles of protein translocation across membranes

Common Principles of ProteinTranslocation Across

MembranesGottfried Schatz* and Bernhard Dobberstein

Most major systems that transport proteins across a membrane share the followingfeatures: an amino-terminal transient signal sequence on the transported protein, a

targeting system on the cis side of the membrane, a hetero-oligomeric transmembranechannel that is gated both across and within the plane of the membrane, a peripherallyattached protein translocation motor that is powered by the hydrolysis of nucleosidetriphosphate, and a protein folding system on the trans side of the membrane. Thesetransport systems are divided into two families: export systems that export proteins outof the cytosol, and import systems that transport proteins into cytosol-like compartments.

membrane (6). These NH2-terminal se-quences have been termed signal-, leader-,targeting-, transit-, or presequences (2); wewill refer to them as signal sequences. Ingeneral, export signal sequences are hydro-phobic and import signal sequences morehydrophilic. Export signal sequences can

A protein's function depends critically onits correct subcellular location. Cells havetherefore developed elaborate systems formaintaining membrane-limited compart-ments endowed with specific proteins.Roughly one-third of a cell's proteins aremembrane proteins, even in Mycoplasmagenitalium that must get by with only 482genes (1). Also, many soluble proteinsmust travel across one or two membranesto reach their final location, either outsidethe cell or within an intracellular com-partment (2). Thus, almost half of theproteins of an average cell are transportedinto or across a membrane.

Export and Import Systems

There are two major types of protein trans-port systems (Fig. 1). One type is located inthe bacterial plasma membrane, the endo-plasmic reticulum, and the internal mem-branes of mitochondria and chloroplasts.We refer to it as an export system because itexports proteins from the cytosol into ex-tracytosolic compartments (the bacterialperiplasm, the lumen of the endoplasmicreticulum, the thylakoid lumen, or the mi-tochondrial inner membrane). The othertype of system is located in the outer andinner membranes of mitochondria andchloroplasts and in the peroxisomal mem-brane. We refer to it as an import systembecause it transports proteins into compart-ments that are functionally equivalent to,or evolutionarily derived from, the cytosol(the inner compartments of mitochondria,chloroplasts, and peroxisomes) (2, 3). Mi-tochondria and chloroplasts thus have ex-port as well as import systems.

G. Schatz is at the Biozentrum der Universitat Basel,CH-4056 Basel, Switzerland. B. Dobberstein is at theZentrum fur Molekulare Biologie der Universitat Heidel-berg (ZMBH), D-69120 Heidelberg, Germany.*To whom correspondence should be addressed.

The export and import systems are gen-eral systems that can transport many differ-ent proteins. Membranes may also containspecial systems that transport only a fewproteins, but these systems are quite differ-ent from the general systems and shall notbe considered here. We shall also not con-sider protein transport through the nuclearpores because the transported proteins nev-er leave the cell's cytosolic phase [see therelated article by Gorlich and Mattaj (4) inthis issue]. Finally, we shall treat the proteintransport system of peroxisomes separately,as it appears to work quite differently fromthe other systems.

Export and import systems operate by asimilar set of principles (2, 5): The trans-ported protein usually carries an NH2-terminal signal and is recognized by cyto-solic factors that deliver it to specific re-ceptors on the target membrane; the pro-tein usually remains partly unfoldedduring transport; the signal sequenceopens a hetero-oligomeric transmembranechannel that is gated both across andwithin the plane of the membrane; thechannel is coupled to a peripherally at-tached protein translocation motor that isdriven by the hydrolysis of nucleosidetriphosphate and transports the proteinacross the membrane; and the proteinfolds on the trans side of the membranewith the help of chaperones or foldingenzymes. In discussing these differentpoints, we will focus on general principlesrather than on details in order to present aunified view of how proteins are translo-cated across membranes.

Targeting Signals

Proteins destined to be transported into oracross a membrane are usually synthesizedwith an NH2-terminal sequence that is pro-teolytically removed on the trans side of the

SCIENCE * VOL. 271 * 15 MARCH 1996

Fig. 1. Protein translocation systems. Export sys-tems (green) export proteins from a cytosol to anextracytosolic compartment. Import systems(red-brown) import proteins into a space that isfunctionally equivalent to, or evolutionarily derivedfrom, a cytosol. The import systems of mitochon-dria and chloroplasts consist of coupled systemsin the outer (OM) and the inner membrane (IM).Reversible dissociation of the two systems allowsprotein import into the intermembrane space(IMS). In chloroplasts, some imported as well asorganelle-encoded proteins are inserted into thy-lakoids, and perhaps also into the inner mem-brane, by an export system. In mitochondria, vir-tually all organelle-encoded proteins are insertedinto the inner membrane by an unknown system,most likely an export system. The import system inthe single membrane of peroxisomes is poorlyunderstood. P, periplasm; M, matrix; S, stroma;TL, thylakoid lumen; TM, thylakoid membrane;ER, endoplasmic reticulum; Sec63c, Sec63 com-plex; Sec6l c, Sec6l complex; SRP, signal recog-nition particle; SR, SRP receptor; SecYEG, Tim,Tom, Oep, and lap, protein subunits of the trans-location systems.

1519

m

usually be interchanged within a given ex-

port system and can often be interchangedbetween different export systems. In con-

trast, import signal sequences can usuallyonly be interchanged within the same or-

ganellar system (Fig. 2).Signal sequences increase the specific

interaction of a protein with its appropriatetransport machinery by different mecha-nisms. Experiments in Escherichia coli haveshown that the signal sequence may act attwo distinct steps within the transport pro-

cess. In the first step, it may retard foldingof a protein, causing the protein to bindthrough its mature region to the cytosolicchaperone SecB (see below). SecB thenspecifically interacts with the SecA subunitof the transport machinery in the plasmamembrane, delivering the bound protein to

the membrane-linked SecYEG complex (a

Export signal sequenceSPase

_~~~~~~~2 to 40 aa 7 to 20 aa

Mitochondral import signal sequenceIMPPase

OH OH OH M

Chloroplast import signal sequence

OH OH OH +CPPase

Composite signal sequence

+ MPPase or SPaseCPPase +

Organelle Stop-transfer orsignal sequence export signal sequence

Peroxisomal import signal sequences

PTS1

PTS2

Fig. 2. Signal sequences. Export signal sequenc-es are in green, import signal sequences in red.Abbreviations: aa, amino acid; SPase, signal pep-tidase; MPPase, mitochondrial processing pepti-dase; CPPase, chloroplast processing peptidase;PTS1 and PTS2, COOH- and NH2-terminal importsignal sequences for peroxisomes; OH, hydroxy-lated amino acids; vertical lines, hydrophobic re-gion; zigzag lines, 3 strand; +, positively chargedamino acid; SKL, Ser-Lys-Leu. The short verticalbars within the NH2-terminal part of mitochondrialsignal sequence indicate the hydrophobic face ofan amphiphilic helix. The drawing of the compositesignal sequence does not indicate the special fea-tures of each substructure.

1520

complex of SecY, SecE, and SecG sub-units). In the second step, the signal se-quence interacts with and activates themembrane-associated export machinery. Inthis mechanism the signal sequence is not atargeting signal but a folding inhibitor and atag by which the membrane-linked trans-port machinery identifies a protein destinedfor transport (7, 8). If the signal sequence isremoved, export is inhibited 100- to 1000-fold, but export can be increased to almostnormal rates by mutations in the SecYEGsubunits (9). In other cases, however, thesignal sequence acts as a true targeting sig-nal because it is specifically recognized by acytosolic chaperone or targeting factor (10).

Export signal sequences mediate thetransport of protein across the bacterialplasma membrane, into the endoplasmicreticulum, into thylakoids, and into the mi-tochondrial inner membrane. They mayalso determine whether transport into theendoplasmic reticulum occurs co- or post-translationally, or whether transport intothylakoids is driven by adenosine triphos-phate (ATP) or by a pH gradient (10, 11)(see also below). Finally, if an export signalsequence is not cleaved off, and if its hy-drophobic core is about 20 residues long, itcan function as a signal-anchor sequencethat permanently anchors the protein with-in the membrane, generating a transmem-brane protein.A transmembrane protein can also be

generated by the sequential function of asignal sequence and a downstream hydro-phobic stop-transfer sequence composed ofabout 20 amino acids. The signal sequenceinitiates translocation, and the stop-transfersequence then probably triggers a lateralopening of the translocation channel. Byescaping from the translocation channel,the stop-transfer sequence arrests transloca-tion and becomes the transmembrane an-chor of an integral membrane protein (2).

Import signal sequences transport pro-teins into the mitochondrial matrix or thechloroplast stroma. They are usually be-tween 20 and 35 residues long, more hy-drophilic than export signal sequences,and rich in hydroxylated amino acids. Mi-tochondrial import signal sequences are,in addition, rich in basic amino acids,generally lack acidic ones, and can foldinto an amphiphilic ot helix or ot sheet (6).The basic and amphiphilic character ofmitochondrial import signal sequences isessential for their function. The importsignal sequences of chloroplasts are nei-ther strongly basic nor amphiphilic, andtheir distinguishing features are still some-what mysterious. Some import signal se-quences transport a protein both into mi-tochondria and chloroplasts (12). Such adual targeting shows that import signalsequences of mitochondria and chloro-


plasts are functionally related.There also exist composite signal se-

quences in which two signal sequences actin tandem. For example, the import ofproteins from the cytoplasm into thyla-koids, the mitochondrial intermembranespace, or the mitochondrial outer mem-brane is often effected by an NH2-terminalimport signal sequence followed by an ex-port signal sequence or a stop-transfer se-quence. In chloroplasts, the import signalsequence directs the protein to the stromawhere it is cleaved off; the export signalsequence then transports the protein intothylakoids. In mitochondria, the importsignal sequence directs the protein to mi-tochondria, and a stop-transfer sequencethen arrests translocation either in theouter or the inner membrane. A stop-transfer sequence in the mitochondrialouter membrane generates a protein span-ning the outer membrane, and in the mi-tochondrial inner membrane stop-transfergenerates a protein spanning the innermembrane, and proteolytic cleavage ofthis inner membrane protein may thengenerate a soluble protein of the inter-membrane space (13). Export signal se-quences and stop-transfer sequences act-ing within chloroplasts and mitochondriaare less hydrophobic than the correspond-ing sequences of bacteria or the endoplas-mic reticulum, perhaps because the or-ganellar sequences must be transportedcompletely across the organellar outermembrane.

Signal sequences specific for a givenmembrane transport system lack a strictconsensus sequence. In fact, up to 25% ofrandomly generated peptides can functionas signal sequences for the endoplasmicreticulum, mitochondria, and the bacterialplasma membrane (14). Thus, it is gener-ally assumed that signal sequences havehighly degenerate primary sequences andthat their group-specific properties reflecta common secondary structure or a similardistribution of charged and apolar resi-dues. However, the sequence degeneracyof naturally occurring signal sequencesmay be overestimated. Most authentic sig-nal sequences lead to almost quantitativetranslocation of the attached proteinacross the target membrane, whereas mostrandomly generated sequences are muchless effective. Also, a signal sequence me-diates not only recognition of a protein bythe membrane-linked transport system,but it can also have other functions. Asalready briefly mentioned, it can deter-mine the targeting pathway to the mem-brane, or it may function as a permanenttransmembrane anchor. Each of thesefunctions is probably dictated by subtle,yet specific sequence motifs that are stillunknown.

I 1105.'-Wlh M.M w1lalw4w

*M0 i'.: :).-9if L. .2S L..'2;: '2, ON ',X

Cytosolic Chaperones andGuanosine Triphosphatases

Most membrane systems can only transport

proteins that are at least partly unfolded(15). Some small proteins (such as the coat

protein of phage M13) can spontaneouslymaintain a loose conformation in the cy-

tosol and travel from the ribosome to thetarget membrane unassisted (8), but largeproteins, particularly those with hydropho-bic internal domains or hydrophobic signalsequences, require the assistance of cytoso-

lic chaperones or similar factors (16). Inter-action with a membrane receptor then re-

leases the cytosolic factor and directs theprotein's signal sequence into the translo-cation channel. This release reaction issometimes controlled by the hydrolysis ofATP or guanosine triphosphate (GTP)(Fig. 3). A few proteins with signal se-

quences may fully fold in the cytosol andthen be unfolded by the membrane-linkedprotein transport system (17).

Chaperones are proteins that interactwith nonnative conformations of other pro-

teins. By this interaction they prevent ag-

gregation of newly synthesized proteins untilthese have either folded correctly in thecytosol or have been transferred to a mem-

brane receptor. Some cytosolic chaperonesinteract mainly with nascent polypeptides,whereas others interact mainly with proteinsthat have been released from a ribosome.Examples of factors binding to nascentchains are nascent chain-associated com-

plex (18) and trigger factor (19); examplesof factors binding to released polypeptidesare SecB (7) and mitochondrial import-stimulating factor (MSF) (20). Chaperonesalso differ in their substrate specificity.Thus, 70-kD heat shock proteins (Hsp7Os)interact with a wide spectrum of nonnativeproteins, whereas other chaperones, such as

SecB and MSF, only interact with proteinsdestined for transport into or across a mem-

brane. Finally, chaperones also differ withrespect to partner proteins and cofactors.Release of a bound protein from SecB onlyrequires interaction with the appropriate ac-

ceptor, SecA, but efficient release fromHsp7O or DnaK requires interaction withthe co-chaperones DnaJ or GrpE as well as

ATP hydrolysis by the chaperone. An even

higher level of complexity and regulation isrepresented by a family of cytosolic ribo-nucleoproteins termed signal recognitionparticle (SRP) whose function is regulatedby binding and hydrolysis of GTP.

In E. coli, there exist at least two path-ways for targeting proteins to the exportmachinery in the plasma membrane. One ofthese pathways is mediated by SecB andSecA. SecB is a chaperone that has so faronly been found in prokaryotes and is ded-icated to protein export. SecB binds to basic

clusters and apolar regions of nonnativeproteins in an ATP-independent mannerand delivers these proteins to membrane-bound SecA (7, 21). The second targetingpathway is mediated by SRP and FtsY (22).Escherichia coli SRP is a cytosolic ribonucle-oprotein composed of a 4.5S RNA and asingle protein termed fifty-four homolog(Ffh) because it is homologous to the 54-kDsubunit of eukaryotic SRP. FtsY is a solubleprotein that is homologous to the a subunitof the SRP receptor on the endoplasmicreticulum (see below). Both the SRP sub-unit and FtsY are guanosine triphosphatases(GTPases). The bacterial SRP binds to thesignal sequence of a nascent secretory pro-tein, and the resulting nascent chain-ribo-some-SRP complex then binds to FtsY in aGTP-dependent manner. SRP and FtsY in-crease the targeting efficiency of a subset ofE. coli secretory proteins, but how theyachieve this is not clear. Each of these twobacterial targeting pathways mediates ex-port of only a subset of presecretory pro-teins. The choice of the pathway is probablygoverned by a precursor's signal sequence,by sequence motifs within the mature re-gion of the precursor, and by the competi-tion between folding and binding to SecB.

SRP-dependent and -independent tar-geting is also found with proteins destinedfor the endoplasmic reticulum (10). SRP-dependent translocation in eukaryotes oc-curs cotranslationally. The coupling be-tween translation and translocation may bethe reason why eukaryotic SRP is morecomplex than its prokaryotic counterpart.Eukaryotic SRP contains a 7S RNA and sixprotein subunits, one of which (termedSRP54) is a GTPase. SRP binds both to theribosome as well as to the signal sequence ofa nascent protein destined for the endoplas-mic reticulum. The specificity of this inter-action depends on the ribosome-associatedheterodimeric nascent chain-associatedcomplex, which preferentially binds to sig-

Nascent chain- Nobinding proteins

w---

Solublechaperones andregulators

Membrane

receptors111 _iiiii1 rIIiiii


nal sequence-free nascent chains and pre-vents them from binding to SRP (18); uponbinding to a signal sequence and the ribo-some, SRP binds GTP and arrests furtherelongation until the nascent chain-ribo-some-SRP complex docks onto the SRPreceptor (also termed docking protein) onthe endoplasmic reticulum. The SRP recep-tor is a heterodimer composed of two GTP-binding proteins: a membrane-integrated rsubunit and a peripherally attached ox sub-unit. Binding of SRP to its receptor triggersGTP hydrolysis on both SRP and on theSRP receptor, allows release of the signalsequence from SRP, and initiates interac-tion of the signal sequence with the trans-location channel. This GTPase cascade en-sures a tight temporal and spatial couplingbetween translation and membrane translo-cation: The growing chain passes directlyfrom the ribosome into the translocationchannel (cotranslational translocation)(23, 24). A GTP-binding SRP54 also par-ticipates in protein transport from the chlo-roplast stroma into thylakoids (25).

Proteins can also be targeted to the en-doplasmic reticulum by an SRP-indepen-dent, posttranslational pathway that hasbeen characterized best in the yeast Saccha-romyces cerevisiae. This pathway seems touse cytosolic Hsp7O and the DnaJ homologYdjlp (10, 26). Cytosolic Hsp7Os (probablytogether with cytosolic DnaJs) also help totarget some proteins to mitochondria andother organelles (16), but in no case has thespecific membrane receptor been identified.

In addition to the Hsp7O-Ydj Ip system,mitochondria also use a dedicated, ATP-controlled targeting system for importingsome of their proteins. The cytosolic chap-erone mitochondrial import-stimulating fac-tor (MSF), a heterodimer composed of a 30-and a 32-kD subunit, specifically binds mi-tochondrial signal sequences and therebyprevents or even reverses aggregation of mi-tochondrial precursor proteins in the cytosol

Fig. 3. Protein-targeting pathways.Proteins are targeted to a mem-brane by different pathways, someof which are regulated by ATP orGTP. Green indicates export systemproteins and red-brown, import sys-tem proteins. Chaperones, regula-tory proteins, or receptors can act atthe ribosome, in the cytosol, or onthe target membrane. Examples aregiven for each of these steps. Com-ponents that are regulated by nucle-oside triphosphate (NTP) aremarked by an asterisk. Red zigzagline, signal sequence; NAC, nascentchain-associated complex (18); TF,trigger factor (19); MSF, mitochon-

11111111 drial import-stimulating factor (20).

1521

(20). Binding of a mitochondrial precursorprotein triggers ATP hydrolysis by MSF. TheMSF-precursor complex specifically docksonto the Tom37p-Tom7Op receptor on themitochondrial surface (see below) and re-leases the precursor to the import channel inthe mitochondrial outer membrane.

Membrane Receptors

Protein-translocating membranes of eu-karyotes contain receptors that recognizeeither signal sequences or precursors boundto a cytosolic factor. The known receptorsare integral membrane proteins with cyto-solically exposed domains. Receptor func-tion can be regulated by the binding andhydrolysis of nucleoside triphosphate. Re-ceptors are usually not fixed to the translo-cation channel but interact with it dynam-ically. Deletion of a single receptor is oftennot lethal and inhibits transport of differentprecursors to different extents, perhaps be-cause a given precursor may be presented toa membrane system by different routes. Themultiplicity of receptors reflects the multi-plicity of cytosolic targeting systems andallows a membrane to regulate the targetingof different precursors differentially.A major protein translocation receptor

on the endoplasmic reticulum is the SRPreceptor discussed above. Membrane recep-tors for SRP-independent protein transportinto the endoplasmic reticulum (26) havenot yet been identified, but candidates forsuch a receptor are subunits of the Sec63complex, or those of the gated transmem-brane channel itself (24, 26).

Mitochondria from S. cerevisiae andNeurospora crassa contain three integralouter membrane proteins that function asprotein import receptors. These proteins aretermed Tom2Op, Tom22p, and Tom7Op,according to their apparent molecular mass-es (27). Saccharomyces cerevisiae also con-tains a fourth receptor protein, Tom37p.The four yeast proteins are usually isolatedas a Tom37p-Tom7Op heterodimer and aless well defined Tom2Op-Tom22p subcom-plex, but they appear to exist as a singlehetero-oligomeric receptor in vivo. Thehighly acidic Tom2Op-Tom22p subunits(termed acid bristle subunits) probably bindthe basic and amphiphilic mitochondrialsignal sequences. This interaction does notrequire nucleoside triphosphates (28). Incontrast, the Tom37p-Tom7Op subunitsrecognize precursors that are bound to theATP-requiring cytosolic chaperone MSF asdiscussed above (20). It is not clear whichfeature in the precursor-MSF complexTom37p-Tom7Op recognizes, but free MSFis not recognized. The precursor's signalsequence is then transferred to the acidbristle subunits Tom2Op-Tom22p and fromthere across the protein transport channel

in the outer membrane. As with SRP-de-pendent and -independent translocationinto the endoplasmic reticulum, the twomitochondrial receptor subcomplexes canpartly assume each other's function. IfTom2Op, Tom37p, and Tom7Op are delet-ed, either singly or pairwise, protein importinto mitochondria is only partially inhibit-ed and the cells remain viable. OnlyTom22p is essential for protein import andviability, presumably because this subunit isalso a component of the transport systemacross the outer membrane (see below). Sofar, there is no evidence that GTP has a rolein mitochondrial protein import.

The binding of precursors to the chloro-plast surface appears to be controlled byGTP. The two putative protein import re-ceptors of chloroplasts, Oep34 and Oep86(for outer envelope protein; also termedIap34 and lap86) have GTP-binding motifs,and the binding of precursors to chloro-plasts is sensitive to guanosine 5'-O-(3-thiotriphosphate) (GTP-,y-S) (29, 30). It isnot yet known whether these two putativereceptor subunits interact with differentprecursors or whether they recognize signalsequences rather than precursors bound tocytosolic chaperones. In E. coli, precursorsor precursor-SecB complexes bind directlyto the SecA subunit of the translocationmachinery. SecA recognizes the precursor'ssignal sequence, its unfolded mature re-gions, and SecB (21, 31 ).

When the endoplasmic reticulum or theouter membranes of mitochondria and chlo-roplasts are solubilized with nonionic deter-gents, channels and receptors usually do notcofractionate (10, 27), suggesting that theirinteraction in vivo is dynamic. A channelmay thus be served by several different re-ceptors. Dynamic interaction of receptorswith import channels probably enhances thecapture of precursors by translocation chan-nels in a similar way as the excess of differ-ent antenna pigments enhances the captureof photons by photosystems.

Transmembrane ProteinChannels

The question of how proteins with theirmany hydrophilic and charged amino acidside chains can move across a phospholipidbilayer has always been one of the mostfascinating and baffling aspects of proteinkinesis. It now appears that the translocat-ing polypeptide chain is moved through ahydrophilic, hetero-oligomeric transmem-brane channel composed of integral mem-brane proteins; that the channel is gatedacross the membrane by the signal sequenceand within the plane of the membrane bystop-transfer or signal-anchor sequences;and that translocation of the polypeptidechain is generally powered by a peripherally

attached protein translocation motor hy-drolyzing nucleoside triphosphate. In bacte-ria and mitochondria, these motors are aid-ed by a transmembrane electrochemical po-tential (2, 21). Protein transport into thy-lakoids appears to be an exception as importof some proteins can be driven by a pHgradient alone (32).

How a protein enters and moves througha channel is not understood. The path of anascent polypeptide through the endoplas-mic reticulum channel has been traced withthe aid of polypeptides that were tagged atdefined sites with light-activated cross-link-ers or environment-sensitive fluorescentprobes. The cross-linkers showed that theSec6l complex is the channel, and fluores-cent probes showed that the channel inte-rior is hydrophilic (23, 33). The signal se-quence was found to be in contact withSec6l subunits as well as with the mem-brane lipids, explaining why it is retainedon the membrane. The protein-translocat-ing channels of E. coli and yeast were alsocharacterized by powerful genetic methodsand by reconstituting purified protein frac-tions into liposomes (21, 34).

In the E. coli plasma membrane, thetransmembrane channel (the first proteintransport system to be elucidated in molec-ular detail) is a heterotrimer composed ofthe integral proteins SecY, SecE, and SecG.This complex, termed SecYEG, acts in con-cert with SecA and the membrane proteinsSecD and SecF (21, 31). The three subunitsof the protein channel across the endoplas-mic reticulum of mammals and yeast, theSec6l complex, are similar in sequence tothe three subunits of the bacterial SecYEGcomplex (35). In the mitochondrial innermembrane, the protein import channel ap-pears to contain the two integral membraneproteins Tim23 and Timl7 (36) and per-haps two additional subunits (37). Studiesin chloroplasts have implicated two integralproteins of the inner envelope (lap100 andlap36) as candidate components of the pro-tein transport channel across that mem-brane (29).

Mitochondria and chloroplasts also haveexport systems. The mitochondrial systemhas not yet been properly characterized, butit seems to insert mitochondrially encodedproteins into the inner membrane (2).There is no evidence that it exports pro-teins into the soluble intermembrane space.The chloroplast system is better character-ized and also more versatile: It acts not onlyon chloroplast-encoded proteins but also onimported proteins and transports them intothe thylakoid membrane, the thylakoid lu-men, and perhaps also into the inner enve-lope membrane (38). As chloroplasts con-tain homologs of bacterial SecA and SecY(39) as well as SRP54 (25), their exportsystem closely resembles that of bacteria.


As protein-transportingimembranes gen-erally do not allow the passive diffusion ofions, the protein transport channels mustopen up only in response to a translocatingpolypeptide. Several lines of indirect evi-dence suggest that such a gating across themembrane may be controlled by the appro-priate signal sequence. First, mutations in asignal sequence for the bacterial plasmamembrane can be suppressed by mutationsin the channel subunit SecY, indicating aspecific interaction between these two part-ners (40). Second, the Sec6l channel of themammalian endoplasmic reticulum effec-tively discriminates between functional andnonfunctional signal sequences (24). Third,electrophysiological studies have detectedlarge-conductance channels in the endo-plasmic reticulum and the E. coli plasmamembrane that are specifically blocked by atranslocating polypeptide chain or openedby a chemically synthesized signal peptide(41 ). However, it is not yet certain thatthese gated ion channels are indeed proteintransport channels. As already mentioned,protein-translocation channels are also gat-ed within the plane of the membrane, re-leasing hydrophobic stop-transfer sequencesinto the lipid bilayer. Some steps of thislateral release might well be the mirrorimage of the steps by which hydrophobicexport signal sequences interact with thechannel: The hydrophobic segment of thesignal sequence might first partition intothe hydrophobic core of the bilayer andthen enter the transport channel laterally.

Protein transport channels are not activetransporters but conduits that functionallyresemble the Fo portion of the proton-trans-locating F1 F,-adenosine triphosphatase(ATPase). In the FIFO-ATPase, active pro-ton transport is effected by coupling thegated proton channel F, to the peripheralproton pump F1, which is powered by the

hydrolysis of ATP (42). In transmembraneprotein transport, most protein channels arecoupled to a peripheral protein translocationmotor that is powered by the hydrolysis ofGTP or ATP and that either pushes or pullsthe protein across the membrane.

There exist at least three types of mem-brane-associated protein translocation mo-tors (Fig. 4). One is Hsp7O, which is boundto the trans side of the membrane. Thismotor provides an ATP-powered "pull" andeffects posttranslational import into the en-doplasmic reticulum, mitochondria, andprobably also into chloroplasts. Anothermotor is SecA that is bound to the cis sideof the membrane. This motor provides anATP-powered push and effects proteintransport across the bacterial plasma mem-brane and probably also the thylakoid mem-brane. A third type of motor might be aribosome that is bound to the cis side of themembrane. In principle, this motor couldprovide a GTP-powered push, but there iscurrently no compelling proof that ribo-somes can push a growing polypeptidechain against a resistance.

The ATP-powered pull by a member ofthe Hsp7O family has been studied in theendoplasmic reticulum of yeast and in mi-tochondria. In the endoplasmic reticulumof yeast, a Hsp7O (termed Kar2p or BiP) isbound to a lumenally exposed DnaJ-likedomain of Sec63p, a transmembrane sub-unit of the hetero-oligomeric Sec63 com-plex (43). A similar arrangement is foundin yeast mitochondria: Import across theinner membrane is mediated by a matrix-located Hsp70 (termed mHsp7O), which isbound to the cochaperone GrpEp and tothe inner membrane protein Tim44p. ThemHsp7O-GrpEp-Tim44p complex inter-acts loosely with the import channel inthe inner membrane (44). In both cases,the Hsp7O on the trans side of the mem-

brane interacts with the translocatingpolypeptide chain and, by hydrolyzingATP, pulls it across the membrane.

How does such an ATP-driven pullwork? The molecular ratchet model (Fig. 5)proposes that Hsp7O binds to segments ofthe polypeptide as these emerge on thetrans side of the membrane by random os-cillation within the transport channel (45).Acting like Maxwell's demon, Hsp7O thusconverts a random Brownian motion into aunidirectional movement. According tothis passive capture model, the rate of post-translational transport of proteins with tight-ly folded domains should be limited by thespontaneous unfolding of such domains onthe cis side of the membrane barrier. How-ever, mitochondria can rapidly import pre-cursor proteins with tightly folded domainswhose spontaneous unfolding would requirehours (17). This predicament has spawnedthe translocation motor model (46) (Fig. 5).This model proposes that the translocatingpolypeptide binds to Hsp7O, that this bind-ing stimulates ATP hydrolysis by Hsp7O, andthat ATP hydrolysis causes a conformationalchange in Hsp7O, which actively pulls a

Brownian ratchet

Translocation motor

moFig. 4. The protein trans-location systems of theendoplasmic reticulum,the mitochondrial innermembrane, and the bac-terial plasma membrane.Export systems are ingreen, import systems inred-brown, NTP-cou-pled protein transloca-tion motors in red, cyto-solic spaces in light blueor gray, and extracyto-solic spaces in tan. The E, ~iIMtohnrotwo systems of the en-doplasmic reticulum represent the posttranslational system identified in yeast and the cotranslationalsystem. EF, ribosome-associated elongation factor of protein synthesis; Sec63c and SeoGi c stand forthe respective hetero-oligomeric complexes; Tim, transport components of the mitochondrial innermembrane. BiP (Hsp7O) may also be required for cotranslational translocation into the endoplasmicreticulum (50). The SecA domain surrounded by a dotted line is involved in the insertion-deinsertion cycleduring the motor function of this protein. N' indicates an NH2-terminus generated by proteolytic removalof the presequence.

Fig. 5. Two models for the function of Hsp7O asan ATP-driven protein translocation motor. (A)The molecular ratchet model proposes thatHsp7O (red) binds to the translocating chain as itemerges on the trans side (bottom) of the mem-brane barrier as a result of random oscillationwithin the translocation channel (blue). By pre-venting back-movement, successive rounds ofHsp7O binding convert a random into a unidirec-tional motion. This model predicts that the rate oftranslocation of folded proteins is determined bythe rate of spontaneous unfolding. (B) The trans-location motor model proposes that Hsp7O cap-tures the incoming chain and then undergoes anATP-driven conformational change that activelypulls a segment of the bound chain across themembrane. This model implies that Hsp7O canactively unfold precursors. Adapted from (46).

SCIENCE * VOL. 271 * 15 MARCH 1996 1 523

0

bftI

I 0

segmnent of the bound chain across the mem-brane. By generating an inward force, Hsp7Ocan thus act as an unfoldase for a proteindomain that has not yet crossed the mem-brane. Finally, GrpEp-triggered adenosinediphosphate-ATP exchange on mHsp7Owould restart the cycle with a downstreamsegment of the polypeptide chain. The trans-location motor model invokes a functionalanalogy to the myosin-actin interaction inmuscle contraction. These two models arenot mutually exclusive. Studies with yeastmitochondria containing mutant mHsp7Oproteins are compatible with the possibilitythat the molecular ratchet mode operateswith loosely folded precursors, whereas thetranslocation motor mode is called up for thetranslocation of tightly folded protein do-miains (47).

The ATP-powered push by SecA in bac-terial protein export is the best documentedexample of an ATP-driven protein translo-cation motor (48). SecA does not resembleHsp7O proteins structurally but, like Hsp7Os,is an ATPase whose activity is stimulated byunfolded polypeptide chains. UnlikeHsp7Os, SecA specifically hinds the cytoso-lic chaperone SecB as well as acidic phos-pholipids that are essential for maximalATPase activity. SecA is attached to thecytosolic surface of the plasma membranethrough acidic phospholipids and theSecYEG complex. Upon binding ATP,SecA interacts with the polypeptide des-tined for export and then undergoes amassive conformational change that push-es a 30-kD SecA domain all the way acrossthe plasma membrane. This transmem-brane movement of a SecA domain carriesa segment of the translocating polypeptidechain across the membrane. Hydrolysis ofthe bound ATP induces return of thetranslocated SecA domain to its groundstate on the cytosolic face of the plasma

membrane and releases the translocatingpolypeptide from SecA. The releasedchain is then translocated further by thetransmembrane electrochemical potentialuntil it is recaptured by the next round ofthe SecA cycle. The SecA motor thus hastwo power strokes, one fueled by ATP andthe other by a transmembrane potential.Why do bacteria push whereas the endo-plasmic reticulum and mitochondria pull?A reason may be the lack of ATP on thetrans side of the bacterial plasma mem-brane: Bacteria have put the motor wherethe fuel is.

Whether a translocating polypeptidechain can be pushed across a membrane bya membrane-bound ribosome is still un-proven. If such a push exists, it can only beexerted by the GTP-powered polypeptideelongation machinery that extrudes thegrowing polypeptide chain from the ribo-somal tunnel across the transmembraneprotein channel (6). In order for thismechanism to work, the ribosome must befixed to the membrane independently ofthe translocating polypeptide chain. Thiscondition is met. The ribosome is tightlybound to the endoplasmic reticulumthrough specific, salt-sensitive interac-tions with the Sec6l complex and perhapsadditional "ribosome receptors" (24, 49).However, the situation may be more com-plex because recent data suggest thatHsp7O (BiP) is required even for cotrans-lational protein translocation (50): BiPwas found to be essential for cotransla-tional transport of invertase into yeastmicrosomes. This BiP requirement was notobserved when transport was assayed in amicrosomal system that had been recon-stituted from purified components. Thisdiscrepancy is unexplained; the ribosomalpush may need the help of an Hsp7O-mediated pull, or Hsp7O may be needed to

remove the translocated chain from theexit site of the translocation channel.

The outer membranes of mitochondriaand chloroplasts pose special problems be-cause they lack ATP-driven protein trans-location motors or a significant transmem-brane potential. Yet most of the organellarproteins must be completely transportedacross the outer membranes to reach inter-nal compartments. Although it is not en-tirely clear how transport across the outermembranes is energized, preliminary evi-dence suggests that the precursor's signalsequence is driven across the membrane byassociation with one or more binding siteson the inner side of the outer membrane(51). In the yeast mitochondrial outermembrane, one of the trans binding sitesmay be contributed by the import receptorTom22p, which has a small acidic domainthat protrudes into the intermembranespace and has a high affinity for mitochon-drial signal sequences. In chloroplasts, afunctionally equivalent site may be provid-ed by a Hsp7O that is bound to the innerface of the outer membrane. Binding to thisHsp7O may explain why binding of precur-sors to the chloroplast surface requires lowconcentrations of ATP in the intermem-brane space (52). The signal sequence maythus be transported across the outer mem-branes of mitochondria and chloroplasts bya relay of binding sites on both faces of themembrane. A similar binding cascade isinvoked for the movement of proteinsthrough the nuclear pore complex (4).Once the signal sequence of a protein des-tined for internal compartments hasemerged in the intermembrane space, it maybe captured by the inner membrane systemwhose ATP-linked protein translocationmotor completes translocation. In mito-chondria, capture by the inner membranesystem is initiated by the electric potential

Table 1. Proteins of ex-port and import sys-tems. Proteins with simi-lar functions share thesame horizontal line,proteins with similar se-quence the same color.Mammalian homologs ofyeast proteins are in pa-renthesis. See Fig. 1 forabbreviations and colorcoding.

E. coli

Export systems

Endoplasmicreticulum

Chloroplast(thylakoid)

Import systems

Mitochondrion Chloroplast

Cytosolic Trigger factor; SecB NAC Pasl Op (Pas 8p); Pas7pchaperonesor regulators tffsYGTP _ MSF-ATP

Membrane Fs-A _ ;R Tom2Op-Tom22p Pas2Opreceptors Tom37p-Tom7Op Oep86, 34-GTP

$WoY le1p8 - 5l y Tim17p,23p ap100, 36

Channels Swi

Sec63p, 62p, 71p, 72p Tim44p ?SecD, SecF

Translocationmotors

FSe-cAlRibosomeIKar2p (BIP)

Outer membranechannels of organelles

|mHsp7O0 IchHsp7ITom4Op, 6p, 7p, 8p Oep (lap) 75


Peroxisome

I

1 524

across that membrane that electrophoreti-cally pulls the positively charged signal se-quence across the inner membrane (53).

Protein translocation across differentmembranes is thus effected by a limited setof machines composed of functionally anal-ogous sets of components. However, mostcomponents of the export systems are struc-turally quite different from their functionalcounterparts in the import systems (Table1). The only exceptions are the members ofthe Hsp7O family that participate in bothtypes of systems, but Hsp7Os participate inmany different aspects of protein foldingand are not dedicated to protein transportacross membranes. All export systems sharesome highly conserved components such asSRP, SRP receptor (FtsY), or the subunitsof the translocation channel. In contrast,the membrane-embedded proteins of theimport systems of mitochondria and chloro-plasts are not only different from each otherbut also from all known proteins. Theseimport machines might have evolved fromas yet undiscovered peptide or metabolitetransporters of prokaryotic ancestors. In-deed, the protein channel subunit Tom4Opof the outer membrane of fungal mitochon-dria has a predicted 3 sheet structure that istypical of pore-forming proteins of bacterialouter membranes, although even in this casethe primary sequence of the mitochondrialprotein is completely different from those ofits bacterial counterparts (54).

Chaperones and FoldingCatalysts

As most membranes can only transportloosely folded proteins, exit of a proteinfrom the transport channel is followed byfolding on the trans side of the membrane.Efficient folding is usually mediated by abattery of chaperones and other foldinghelpers (16). Most of these can interactwith a wide spectrum of different nonnativeproteins, but some act on only a few, oreven a single protein. Folding may requirethe successive action of several chaperonesor folding catalysts.

The chloroplast stroma and the mito-chondrial matrix contain homologs of bac-terial GroEL and GroES (termed Hsp6Oand HsplO, or Cpn6O and CpnlO) thatfacilitate the folding of imported as well asorganelle-synthesized proteins. It was firstthought that these organellar chaperoninsare required for the folding of all importedproteins, but it is now clear that they areonly required for a subset of them. In mito-chondria, several monomeric proteins withsimple folding pathways fold rapidly with-out the help of Hsp6O; some fold by meansof an intramitochondrial DnaJ or a cyclo-sporin-sensitive proline rotamase, and oth-ers may fold by means of a subfraction of

mHsp7O that is soluble in the matrix (55).This soluble mHsp7O also acts in qualitycontrol by binding misfolded proteins anddelivering them to the mitochondrial Lonprotease for degradation (56).

Refolding of exported bacterial proteinsin the periplasm is mediated by severalchaperones with very narrow substrate spec-ificity, by disulfide isomerases, and by atleast one proline rotamase (57). The endo-plasmic reticulum lumen contains severalwell-characterized chaperones such as BiP,Hsp9O, and calnexin, which cooperate witheach other and which probably all facilitatethe folding of imported proteins. However,a role in protein folding has so far only beenclearly established for BiP (58), disulfideisomerase, and proline rotamases (57). BiPand calnexin also act as quality controls,allowing only properly folded proteins to betransported along the secretory pathway.Improperly folded proteins are retained inthe endoplasmic reticulum and eventuallytargeted for proteolytic degradation (58).Some chaperones, such as mHsp7O or BiP,thus perform at least three different func-tions in protein transport across mem-branes: translocation, folding, and qualitycontrol. It is usually difficult to assay one ofthese functions independently of the others.

And Peroxisomes?

Most proteins destined to be importedinto peroxisomes carry a COOH-terminalSer-Lys-Leu (or closely similar) tripeptidemotif collectively termed PTS1 (for per-oxisome-targeting signal). A few other im-ported proteins carry a transient NH2-terminal signal (termed PTS2) with aloose 11-residue consensus motif, and stillothers contain unidentified signals (59).PTS1 signals are recognized by a proteintermed Pas8p or PaslOp (depending on theyeast species), but there is no agreementon whether this protein is located in thecytosol or on the peroxisome membrane.PTS2 signals are recognized by the proteinPas7p, but again there is disagreement onwhether this protein is in the cytosol or inthe peroxisomal matrix. Pas8p (or its ho-molog PaslOp) appears to be recognized bythe peroxisomal integral membrane pro-tein Pas2Op (60). Peroxisomes may thusresemble the other membrane systems inusing a cytosolic-targeting system coupledto a membrane receptor, but no channelsubunits have yet been identified. We alsodo not know how protein transport intoperoxisomes is energized, how peroxisomescan import large folded proteins, andwhether proteins are imported into uni-dentified peroxisomal precursor structuresin vivo. Peroxisomes may well blend intothe general picture once their transloca-tion system is better understood, but at the


moment they are difficult to integrate intoa unified view of membrane-linked pro-tein transport.

Outlook

In this article we have depicted proteintranslocation across membranes with abroad brush, de-emphasizing details and mi-nor exceptions in order to highlight com-mon principles. The uniformity of basicmechanisms that has emerged from recentwork places protein translocation firmlywithin the framework of other membrane-linked transport systems. Of course, some ofthe principles described here may not standthe test of time. It is still prudent to seeksimplicity, and then to distrust it.

REFERENCES AND NOTES

1. S. G. Oliver et al., Nature 357, 38 (1992); B. Dujon etal., ibid. 369, 371 (1994); C. M. Fraser et al., Science270, 397 (1995).

2. G. Blobel, Proc. Natl. Acad. Sci. U.S.A. 77, 1496(1980); K. Verner and G. Schatz, Science 241, 1307(1988); G. Attardi and G. Schatz, Annu. Rev. CellBiol. 4, 289 (1988).

3. L. Margulis, Origin of Eukaryotic Cells (Yale Univ.Press, New Haven, CT, 1970).

4. D. Gorlich and 1. W. Mattaj, Science 271, 1513(1996).

5. G. Schatz, Nature 321, 108 (1986).6. G. Blobel and D. D. Sabatini, in Biomembranes, L. A.

Manson, Ed. (Plenum, New York, 1971), vol. 2, pp.193-195; C. Milstein, G. G. Brownlee, T. M. Harri-son, M. B. Matthews, Nature New Biol. 239, 117(1972); G. Blobel and B. Dobberstein, J. Cell Biol. 67,835 (1975); G. von Heijne, J. Mol. Biol. 189, 239(1986); G. von Heijne, EMBO J. 5, 1335 (1986); D.Roise and G. Schatz, J. Biol. Chem. 263, 4509(1988); S. M. Theg and F. J. Geske, Biochemistry 31,5053 (1992).

7. S. J. S. Hardy and L. L. Randall, Science 251, 439(1991); D. L. Diamond, S. Strobel, S.-Y. Chun, L. L.Randall, Protein Sci. 4, 1118 (1995).

8. W. Wickner, Biochemistry 27, 1081 (1988).9. A. I. Derman, J. W. Puziss, P. J. Bassford Jr., J.

Beckwith, EMBO J. 12, 879 (1993).10. P. Walter and A. Johnson, Annu. Rev. Cell Biol. 10,

87 (1994).11. A. M. Chaddock et a/., EMBO J. 14, 2715 (1995); C.

Robinson et al., ibid. 13, 279 (1994).12. E. C. Hurt, N. Soltanifar, M. Goldschmidt-Clermont,

J.-D. Rochaix, G. Schatz, ibid. 5, 1343 (1986); J.Huang, E. Hack, R. W. Thornburg, A. M. Myers,Plant Cell 2, 1249 (1990); S. Brinks et al., J. Biol.Chem. 269, 16478 (1994); G. Creissen, H. Reyn-olds, Y. B. Xue, P. Mullineaux, Plant J. 8, 167 (1995).

13. K. Keegstra, Cell 56, 247 (1989); B. S. Glick, E. M.Beasley, G. Schatz, Trends Biochem. Sci. 17, 453(1992); F.-U. Hartl and W. Neupert, Science 247,930 (1990); M. Nguyen, A. W. Bell, G. C. Shore, J.Cell Biol. 106, 1499 (1988); K. Hahne, V. Haucke, L.Ramage, G. Schatz, Cell 79, 829 (1994).

14. A. Baker and G. Schatz, Proc. Natl. Acad. Sci.U.S.A. 84, 3117 (1987); C. A. Kaiser, D. Preuss, P.Grisafi, D. Botstein, Science 235, 312 (1987); A. Vas-sarotti, R. Stroud, M. Douglas, EMBO J. 6, 705(1987); B. D. Lemire, C. Fankhauser, A. Baker, G.Schatz, J. Biol. Chem. 264, 20206 (1989).

15. M. Eilers and G. Schatz, Nature 322, 228 (1986).16. J. P. Hendrick and F. U. Hartl, Annu. Rev. Biochem.

62, 349 (1993); S. J. Landry and L. M. Gierasch,Annu. Rev. Biophys. Biomolec. Struct. 23, 645(1994); L. L. Randall and S. J. S. Hardy, TrendsBiochem. Sci. 20, 65 (1995).

17. B. S. Glick, C. Wachter, G. A. Reid, G. Schatz, Pro-tein Sci. 2, 1901 (1993); R. A. Stuart et al., Eur. J.Biochem. 220, 9 (1993); B. D. Gambill et al., J. Cell

1525

Biol. 123,109 (1993).18. B. Wiedmann, H. Sakai, T. A. Davis, M. Wiedmann,

Nature 370, 434 (1994).19. G. Stoller et al., EMBO J. 14, 4939 (1995); 0. A.

Valent et al., ibid., p. 5494; T. Hesterkamp, S.Hauser, H. Lutcke, B. Bukau, Proc. Natl. Acad. Sci.U.S.A., in press.

20. N. Hachiya et al., EMBO J. 12, 1579 (1993); N.Hachiya et al., Nature 376, 705 (1995).

21. W. Wickner, A. J. M. Driessen, F.-U. Hartl, Annu.Rev. Biochem. 60,101 (1991).

22. J. Luirinketal., EMBOJ. 13,2289(1994); J. D. Miller,H. D. Bernstein, P. Walter, Nature 367, 657 (1994).

23. K. S. Crowley, G. D. Reingart, A. E. Johnson, Cell 73,1101 (1993); R. Gilmore, ibid. 75, 589 (1993); A. E.Johnson, Trends Biochem. Sci. 18, 456 (1993).

24. B. Jungnickel and T. Rapoport, Cell 82, 261 (1995).25. A. E. Franklin and M. E. Hoffman, J. Biol. Chem. 268,

22175(1993); X. Li, R. Henry, J. Yuan, K. Cline, N. E.Hoffman, Proc. Natl. Acad. Sci. U.S.A. 92, 3789(1995).

26. S. C. Ogg, M. A. Poritz, P. Walter, Mol. Biol. Cell 3,895 (1992).

27. M. Kiebler, K. Becker, N. Pfanner, W. Neupert, J.Membr. Biol. 135, 191 (1993); T. Lithgow, B. S.Glick, G. Schatz, Trends Biochem. Sci. 20, 98(1995); This review uses the new uniform nomencla-ture for membrane proteins of the mitochondrial pro-tein import machinery [N. Pfanner et al., Trends Bio-chem. Sci. 21, 51 (1996)]. In brief, proteins withtranslocase function of the outer and inner mem-brane are designated as Tom and Tim, respectively(for translocase of outer or inner membrane), fol-lowed by a number giving the approximate mass inkilodaltons. The new designations are as follows:Tom70 (previously Mas70 or Mom72); Tom37(Mas37); Tom22 (Masl 7 or Mas22 or Mom22);Tom20 (Mas2O or Moml 9); Tom40 (Isp42p orMom38); Tom6 to Tom8 (Isp6 to Isp8p); Tim44(Isp45p or Mim44 or Mpil); Tim23 (Mas6p or Mim23or Mpi3); and Timl 7 (Mimi 7 or Smsl or Mpi2).

28. V. Haucke, T. Lithgow, S. Rospert, K. Hahne, G.Schatz, J. Biol. Chem. 270, 5565 (1995); A. Mayer,F. E. Nargang, W. Neupert, R. Lill, EMBO J. 14,4204(1995); V. Haucke, M. Horst, G. Schatz, T. Lithgow,EMBO J., in press.

29. D. J. Schnell, F. Kessler, G. Blobel, Science 266,1007 (1994); F. Kessler, G. Blobel, H. A. Patel, D. J.Schnell, ibid., p. 1035.

30. S. Hirsch, E. Muckel, F. Heemeyer, G. von Heijne, J.Soll, ibid., p. 1989; M. Seedorf, K. Wegemann, J.Soll, Plant J. 7, 401 (1995); J. Soll, Bot. Acta 108,277 (1995).

31. M. Douville, A. Price, J. Eichler, A. Economou, W.Wickner, J. Bio. Chem. 270, 20106 (1995); K.-i.Nishiyama, M. Hanada, H. Tokuda, EMBO J. 13,3272 (1994).

32. C. Robinson, R. B. Klosgen, R. B. Herrmann, J. B.Shackleton, FEBS Lett. 325, 67 (1993); K. C. Cline,R. C. L. Henry, J. Yuan, EMBO J. 12, 4105 (1993).

33. B. Martoglio, M. Hoffmann, J. Brunner, B. Dobber-stein, Cell 81, 207 (1995).

34. D. Gbrlich and T. A. Rapoport, Cell 75, 615 (1993).35. E. Hartmann et al., Nature 367, 654 (1994).36. N. Pfanner, E. A. Craig, M. Meijer, Trends Biochem.

Sci. 19, 368 (1994); J. L. T. Emtage and R. E.Jensen, J. Cell Biol. 122,1003 (1993); P. J. T. Dek-ker et al., FEBS Lett. 330, 66 (1993); K. R. Ryan, M.M. Menold, S. Garrett, R. E. Jensen, Mol. Biol. Cell 5,529 (1994).

37. J. Berthold et al., Cell 81, 1085 (1995); K. Tokatlidis,B. S. Glick, T. Junne, G. Schatz, N. Kronidou, per-sonal communication.

38. J. D. Rochaix,Annu. Rev. Cell Biol. 8,1 (1992); N. W.Giliham, J. E. Boynton, C. R. Hauser, Annu. Rev.Genet. 28, 71 (1994).

39. N. Nakai, A. Goto, T. Nohara, D. Sugita, T. Endo, J.Biol. Chem. 269, 31338 (1994); J. Yuan, R. Henry,M. McCaffery, K. Cline, Science 266, 796 (1994); V.Laidler, A. M. Chaddock, T. G. Knott, D. Walker, C.Robinson, J. Biol. Chem. 270,17664 (1995).

40. S. Emr, S. Hanley-Way, T. Silhavy, Cell 23, 79 (1981).41. 5. M. Simon and C. Blobel, ibid. 65,371 (1991); ibid.

69, 677 (1992).42. E. Racker, A New Look at Mechanisms in Bioener-

getics (Academic Press, New York, 1976).

43. S. J. Sanders, K. M. Whiffield, J. P. Vogel, M. D.Rose, R. W. Schekman, Cell 69, 353 (1992); J. L.Brodsky and R. Schekman, J. Cell Biol. 123, 1355(1993); S. Panzner et al., Cell 81, 561 (1995).

44. S. T. Hwang and G. Schatz, Proc. Natl. Acad. Sci.U.S.A. 86, 8432 (1989); N. Kronidou et al., ibid. 91,12818 (1994); H.-C. Schneider et al., Nature 371, 768(1994); J. Rassowetal., J. Cell Biol. 127, 1547 (1994).

45. W. Neupert, F.-U. Hartl, E. A. Craig, N. Pfanner, Cell63, 447 (1990); S. M. Simon, C. S. Peskin, G. F.Oster, Proc. Natl. Acad. Sci. U.S.A. 89, 3770 (1992).

46. B. S. Glick, Cell 80, 11 (1995).47. W. Voos et al., EMBO J., in press.48. A. Economou and W. Wickner, Cell 78, 835 (1994);

Y. J. Kim, T. Rajapandi, D. Oliver, ibid., p. 845; R.Schekman, ibid., p. 91 1; A. Economou, J. A. Pogli-ano, J. Beckwith, D. B. Oliver, W. Wickner, ibid. 83,1171 (1995).

49. B. Jungnickel, T. A. Rapoport, E. Hartmann, FEBSLett. 346, 73 (1995); A. J. Savitz and D. I. Meyer, J.Cell Biol. 120, 853 (1993).

50. J. L. Brodsky, J. Goeckeler, R. Schekman, Proc.Natl. Acad. Sci. U.S.A. 92, 9643 (1995).

51. A. Mayer, W. Neupert, R. Lill, Cell 80, 127 (1995); L.Bolliger et al., EMBO J. 14, 6319 (1995).

52. L. J. Olsen and K. Keegstra, J. Biol. Chem. 267, 433(1992).

53. J. Martin, K. Mahlke, N. Pfanner, ibid. 266, 18051(1991); D. Roise, Proc. Natl. Acad. Sci. U.S.A. 89,608 (1992).

54. C. A. Manella, A. Neuwald, C. E. Lawrence, J. Bioen-erg. Biomembr., in press.

Eukaryotic cells have an elaborate networkof organelles, many of which are in constantand bidirectional communication through aflow of small transport vesicles. For eachorganelle a specific mechanism exists to cap-ture and package certain proteins and lipidsthat are destined for transport to a receivingcompartment. In return, the receiving com-partment accepts proteins that are meant toremain, or to be passed to another station,and then retrieves for recycling other pro-teins that belong in the donor organelle.Among the recycled proteins are structuralcomponents of the traffic pathway that mustbe used repeatedly to sustain transport. The

R. Schekman is in the Department of Molecular and CellBiology and Howard Hughes Medical Institute, Universityof California, Berkeley, CA 94720-3202, USA. L. Orci isat the Department of Morphology, Faculty of Medicine,University of Geneva, Geneva 4, CH-1 21 1 Switzerland.

*To whom correspondence should be addressed.

55. N. Rowley et al., Cell 77, 249 (1994); S. Rospert etal., EMBO J., in press; A. Matouschek, S. Rospert,K. Schmid, B. S. Glick, G. Schatz, Proc. Natl. Acad.Sci. U.S.A. 92, 6319 (1995); J. Rassow et al., Mol.Cell. Biol. 15, 2654 (1995).

56. I. Wagner, H. Arnt, L. VanDyck, T. Langer, W. Neu-pert, EMBO J. 13, 5135 (1994).

57. J. R. Gaut and L. M. Hendershot, Curr. Opin. CellBiol. 5, 589 (1993); S. J. Hultgren, F. Jacob Dubuis-son, C. H. Jones, C. I. Branden, Adv. Protein Chem.44, 99 (1993); P. Klemm, Res. Microbiol. 143, 831(1992); C. Wulfing and A. Pl0ckth0n, Mol. Microbiol.12, 685 (1994); J. C. A. Bardwell, K. McGovern, J.Beckwith, Cell 67, 581 (1991).

58. J. F. Simons, S. Ferro-Novick, M. D. Rose, A. Hele-nius, J. Cell Biol. 130, 41 (1995).

59. R. A. Rachubinski and S. Subramani, Cell 83, 525(1995); P. Heinemann and W. W. Just, FEBS Lett.300,179 (1992); S. Huhse and W.-H. Kunau, in ColdSpring Harbor Symp. Quant. Biol., (Cold Spring Har-bor Laboratory Press, Cold Spring Harbor, NY, 1996).

60. Y. Elgersma, thesis, University of Amsterdam, Am-sterdam, Netherlands (1995).

61. We thank M. Horst, K. Keegstra, J. Soll, W. Kunau, A.Matouschek, S. Rospert, T. Silhavy, and W. Wicknerfor advice and comments, and Y. Cully and M. Jaggifor the artwork. Our own research was supported bygrants from the Human Capital and Mobility Programof the European Economic Community, the Deut-sche Forschungsgemeinschaft, and the HumanFrontier Science Program Organization to B.D., andfrom the Swiss National Science Foundation to G.S.

most remarkable feature of this process isthat selectivity is achieved in spite of thefluid nature of the membrane. In the absenceof specific mechanisms to recognize and se-quester proteins destined for transport andretrieval, communicating organelles wouldquickly lose their identity, succumbing tothe lateral diffusional mobility of membraneproteins embedded within the bilayer. Theevidence that we summarize in this reviewsuggests that membrane identity is main-tained by the selective capture into coatedvesicles of proteins destined for transport.

Three Paradigms of VesicleBud Formation

Three models have contributed to our un-derstanding of the mechanism of vesiclebudding. The first is fashioned on the ex-ample of enveloped viruses that bud from


Coat Proteins and VesicleBudding

Randy Schekman* and Lelio Orci

The trafficking of proteins within eukaryotic cells is achieved by the capture of cargo andtargeting molecules into vesicles that bud from a donor membrane and deliver theircontents to a receiving compartment. This process is bidirectional and may involvemultiple organelles within a cell. Distinct coat proteins mediate each budding event,serving both to shape the transport vesicle and to select by direct or indirect interactionthe desired set of cargo molecules. Secretion, which has been viewed as a defaultpathway, may require sorting and packaging signals on transported molecules to ensuretheir rapid delivery to the cell surface.

1 526

common principles of protein translocation across membranes

Documents