Introductory commentsToday we review the structural properties of biological membranes and discuss a

few of the ways in which these structures influence cytomechanics and biochemistry. We will then start to deal with how cells assemble their membranes, which are mainly composed of a mixture of phospholipids, sterols, and integral and peripheral proteins. What I hope will tie these themes together is an emphasis on the physical properties of phospholipid bilayers, and how these properties dictate the biology that happens in andaround them. Some understanding of these properties is essential if we are going to get into the guts of the membrane transport machinery, as we will begin to do at the end of today’s class and for the next two lectures.

A case can be made that membranes are as central to life as nucleic acids, and perhaps more important than proteins. Membranes separate in from out. This allows cells to not be at equilibrium with their surroundings, i.e., “dead”. Membranes do this by slowing down chemical diffusion and electrical charge transfer. In other words, membranes facilitate the formation of gradients, and thus potentials: electrical, chemical(pH, ionic, osmotic, redox). Work is required to make these gradients; conversely, work can be extracted as gradients are dissipated. Membranes allow macromolecules to be at high concentrations, thus they make practical the reversible assembly of macromolecular complexes -- on-rates will be high, and thus complexes can exist even with high off-rates. Membranes provide surfaces where complexes can assemble (Vale paper). Intracellular membranes allow the formation of subcompartments: cytoplasm, periplasm (in Gram negative bacteria), chloroplast, lysosome, nucleus, etc. so that poorly-compatible biochemical processes can be segregated.

Structure of lipid bilayers

S. White & W. Wimleyhttp://blanco.biomol.uci.edu/Bilayer_Struc.html

The most abundant components of eukaryotic membranes are phospholipids, sterols such as cholesterol, and proteins. Very many other, less abundant molecules also associate with membranes, including various drugs and toxins. You can find structures of various phsopholipids and sterols in any cell biology or biochemistry textbook. You should be familiar with the basic architecture of these molecules, and I’m assuming thatyou already are. If you’re not, look in the Alberts or Lodish texts listed in the syllabus.

Membrane assymetry occurs in two directions: normal to the plane of the bilayer (the two leaflets have different compositions), and within the plane of the bilayer. Assymetry normal to the bilayer is what most biologists mean when they say membrane assymetry. You find different lipids in the cis (cytoplasmic) leaflet than in the trans leaflet. This is because (1) lipid anabolism and catabolism are compartmentalized, and (2) because of flippases, enzymes that transfer lipids from one leaflet to the other. We know shockingly little about these enzymes, but (most of us think that) they will be of great importance.

Assymetry also means that integral membrane proteins have a specific orientation. For example, the epidermal growth factor (EGF) receptor has a ligand binding domain on the outside and a kinase domain in the cytoplasm, and in between them is a single alpha helix that spans the plasma membrane. We don’t fully understand how the cell manages to consistently put the EGF receptor in the membrane right-way-out; if you want to learn more about this, there are good experiments and reviews by Gunnar von Heijne, among others. It is worth thinking about what would happen if the orientation of membrane proteins is randomized, that is, made isotropic. This is not as unlikely as you might think it is: many biochemical studies rely on detergent solubilization and reconstitution of membrane proteins in liposomes. This randomizes their orientation.. What if a transporter is inserted backwards?

Integral membrane proteins come in two major flavors: alpha-helical proteins and beta-stranded barrels. The beta barrels are found in the outer membranes of gram negative bacteria, and mitochondria and chloroplasts (which began as endosymbiotic bacteria in the first place). Some of the protein toxins that insert into eukaryotic membranes also adopt the beta barrel structure. We’ll discuss how the alpha-helical membrane proteins are assembled later today.

A vast array of peripheral membrane proteins are also found in association with biological membranes. They have many modes of association with membranes (electrostatic, hydrophobic, stereospecific), and when we talk about the Vale paper we’llhave more to say about this.

The fluid mosaic model of Singer and Nicolson (1972, Science 175:720) rationalized a great deal of data about membrane structure and function (and it is worth noting that Singer’s work on lateral membrane organization in lymphocytes, done in the 70’s and early 80’s, prefigured a great deal of what followed in immunology and cell biology, from the late 90’s on). For the moment, it may be sufficient to say that the largest driving force in assembly of a phospholipid bilayer is the increase in aqueous solvent entropy that occurs when acyl chains - the greasy bits - on the lipids are dehydrated. The resulting “structure” – and by this we mean the time-averaged positions of the acyl chain, gylceride, and hydrophilic headgroup – is very stable. In addition, the bilayer is relatively inelastic and it is quite strong. What does this mean? When placed under

lateral pressure (by, say, osmotic turgor), membranes stretch very little – less than 3%! – before they rupture Nichol & Hutter, 1996, J. Physiol. 493:187–198) Membranes also bendrelatively freely (bending stiffness ~0.3 pN•nm). This is less than the magnitude of the thermal energy kT (~4 pN•nm @ 300K). It should go without saying that these are typical values, and of course can vary widely depending on temperature, composition of the membrane, salt concentration, pH, etc.

Now, the membrane is not just a bag of enzymes. Bits of it are attached to the cytoskeleton and in many cases also to extracellular structures including cell walls, extracellular matrix, and other cells. This brings us to our first paper, by Raucher et al.

Cell222

Figure 1. Tether Force Measurements of the Adhesion Energy between the Plasma Membrane and the Cortical Cytoskeleton

(A) DIC image of a tether force measurement using optical tweezers.

(B) Schematic view of the optical tweezers force measurement that defines the local adhesion energy term.

(C) Typical displacement trace, showing how far the center of the bead has been moved away from the center of the optical tweezers.

(D) The absence of F-actin in membrane tethers. A membrane marker FM1-43 staining of plasma membrane (left panel) and rhodamine

phalloidin staining of F-actin (right panel).

(E) The measured adhesion energy is independent of the mechanism of bead attachment.

(see Experimental Procedures). This measured adhe- GFP-PH(Akt) domain, which binds PI(3,4)P2, had no ef-

fect (Franke et al., 1997; Isakoff et al., 1998). Since PHsion energy was not dependent on whether beads were

coated with either IgG, IgM, or ConA (Figure 1E). Simi- domain expression may change levels of PIP2, this may

consequently alter baseline calcium or DAG concentra-larly, fibroblasts plated on glass, laminin, or polylysine

had the same plasma membrane to cytoskeleton adhe- tions. Therefore, we tested the effect of calcium and

DAGon adhesion energy. Addition of phorbolester, diac-sion energy (data not shown).

We investigated the cellular roles of PIP2 by express- ylglycerol (DiC8), thapsigargin, and a combination of

phorbolester and thapsigargin showed only a small re-ing pleckstrin homology (PH) domains that specifically

sequester PIP2 (Stauffer et al., 1998). From several PIP2- duction in adhesion energy. This suggests that the

marked reduction in adhesion energy by PH(PLC!) doesbinding PH domains that we tested (PH domains from

PLC", PLC!, and pleckstrin; data not shown), the green not result from a change in the baseline DAG or calcium

concentration (Figure 2C) but rather from a reduction offluorescent protein (GFP)–tagged PHdomain from phos-

pholipase C ! (PLC!) showed the highest ratio of plasma available PIP2.

membrane to cytosolic localized GFP-PH (Figure 2A).

This suggests that the PH domain from PLC! has a high Expression of Plasma Membrane Targeted 5!-SpecificPIP2 Phosphatase Decreases Adhesion Energyin vivo binding affinity for plasma membrane PIP2. As ex-

pected, a control construct with Lys30Asn and Lys32Asn While the experiments with the expressed PH domain

show that sequestration of plasma membrane PIP2 lip-mutations (GFP-PH*) that prevented PIP2-binding in

vitro showed no plasma membrane localization (Fig- ids byGFP-PH(PLC!) lowers adhesion energy,we testedthe connection between PIP2 concentration and adhe-ure 2B).

Expression of the PIP2-sequestering GFP-PH(PLC!) sion energy more directly by using an enzymatic ap-

proach. Our strategy was to target a PIP2-specific 5#-phos-construct dramatically reduced the adhesion energy

(Figure 2C). In control measurements, the adhesion en- phatase to the plasma membrane to selectively reduce

plasma membrane PIP2 concentration but not other in-ergy was the same in cells expressing GFP or GFP-

PH*(PLC!). Furthermore, the reduction in adhesion en- tracellular pools of PIP2 or other plasmamembrane phos-

phatidylinositol phosphates (Figure 3A). We achievedergy was specific for PI(4,5)P2 over PI(3,4)P2, since a

Raucher et al., Cell 100:221

This paper is really cool for a couple of reasons. First of all it deals with the molecular bases of a global mechanical parameter of cells - the adhesion energy between a membrane and the cytoskeleton (broadly defined!). This energy has big implications. For example, there are a number of experiments that show that increases in membrane tension promote membrane fusion. Conversely, increases in membrane fusion seem to inhibit endocytosis. We’ll talk about this a bit more during the next two lectures.

The molecule that the authors focus on is PIP2 (phosphatidylinositol-4,5-bisphosphate). As many of you know, PIP2 has many roles in cell signaling. When it is attacked by phospholipase-C, the molecule is hydrolyzed at the phosphodiester linkage,yielding the lipid DAG (diacylglycerol) and the soluble sugar inositol-1,4,5-triphosphate(IP3). Both of these products have major signaling activities: DAG recruits proteins withC1 domains such as protein kinase C, and IP3 triggers Ca2+ release from the endoplasmicreticulum.

But a lot of recent data in many different systems shows that PIP2 and similar lipids are themselves critical signaling molecules, because they serve as intracellular receptors - specific binding sites that facilitate the assembly of peripheral membrane proteins at particular membrane locations. These proteins contain domains (PH, ENTH, FYVE, PX, N-ERMAD, MARCKS-ED and others) that, often in combination with other binding activities, recruit them to membranes containing these cognate lipids. Raucher et al. argue that they are looking at the aggregate effect of many, many such interactions, at the whole-cell level. How this works at the level of specific proteins is the subject of our second paper.

Earlier, we mentioned assymmetry along the axis normal to the membrane. There also is asymmetry, or perhaps a better word is anisotropy, in the plane of the membrane. In fact it is obvious, because a membrane is made of distinct parts, that the membrane is not going to be the same everywhere. That’s just a statistical consequence of the fluid mosaic model. If you have a rare component, let’s say a not-very-abundant receptor, and you look at a small region of the membrane, sometimes there will be a couple of receptor molecules in your region of interest, sometimes only one, sometimes none. What is less obvious is that there seem to be membrane regions – domains – that have different statistical properties. And all you need to get these domains is the right combination of lipids; we know this because microdomains can be observed in liposomes that contain only two phospholipids and cholesterol:

Free-floating giant unilamellar vesicles of 20–50 !mdiameter were prepared in > 18 M! cm water as byAngelova et al. [14] with modifications to increase yieldand compositional uniformity and to detach the vesiclesfrom the substrate on which they were grown. Vesicleswere viewed using a Nikon Eclipse fluorescence micro-scope with a Photometrics Cool Snap FX camera.

Fluorescence micrographs of vesicles are shown in thetop row of Fig. 1. By changing composition and tempera-ture, three distinct phase morphologies are observed inthe thin shell of the bilayer membrane: solid-liquid coex-istence, coexistence of two liquid phases, and one uni-form liquid phase. These three phases appear the sameindependent of whether the vesicles contain di(14:0)PC,di(15:0)PC, di(16:0)PC, or di(18:0)PC. We observe thatsolid domains are noncircular, rotate as rigid bodies,occur at low cholesterol composition, and exclude thedye. When two solid domains collide on the surface ofour vesicles, they do not deform upon contact. The melt-ing temperatures of solid phases are recorded in Fig. 2.The coexistence of solid and liquid phases in vesicles, thedisruption of solid by higher cholesterol fractions, and thecorrelation of higher melting temperatures with longer

acyl chains observed here are consistent with previousresults [15].

As cholesterol concentration is increased in the vesiclesbeyond !10 mol%, the solid phase is replaced by liquid,which is more applicable to rafts in biological mem-branes. As in the solid phase, liquid domains are large(micron scale), stable, and self-assemble without the aidof proteins. Coexisting liquid phases are found fromapproximately 10 mol % to near 50 mol % cholesterol.The range of lipid compositions over which micron-sizedliquid domains are recorded here is significantly widerthan previously known [10,11]. We characterize liquidphases by their circular domains. When two liquid do-mains collide on the surface of our vesicles, they coalesceby continuously deforming into a larger circular domain.Given that our vesicles are expected to have the samecomposition in both leaflets of the bilayer and that we donot observe overlapping domains, we conclude that do-mains in each bilayer leaflet are in registration, as re-ported previously [10,11]. In our experiments, two phasesare observed, bright and dark (although it is alwayspossible that an additional phase is present that the probedoes not distinguish). Liquid domains that are cholesterol

FIG. 2. Top: A series of fluorescence micrographs of vesicles and monolayers of 1:1 di(18:1)PC/di(16:0)PC with varyingcholesterol. Scale bars are 20 !m for vesicles and 10 !m for monolayers. From left to right, the phases are solid-liquid coexistence,coexisting liquid phases, and either one uniform liquid phase (in vesicles) or coexisting liquid phases (in monolayers). Bottom left:Miscibility transition temperatures for vesicles. Points at low cholesterol represent melting of the solid phase. Filled points markmiscibility transition temperatures for two liquid phases. Temperatures from (10–50) "C are experimentally accessible. Error barsrepresent standard deviations over multiple measurements. Bottom right: Miscibility transition surface pressures for monolayers.The change of contrast is marked by a vertical gray dashed line. For all domains # 10 !m, striping of domains was seen near thetransition. Striping may occur in smaller domains beyond our resolution. In all cases, curves are drawn to guide the eye and are notexplicit fits of the data.

VOLUME 89, NUMBER 26 P H Y S I C A L R E V I E W L E T T E R S 23 DECEMBER 2002

268101-2 268101-2

A series of fluorescence micrographs of vesicles and monolayers of 1:1 di(18:1)PC/di(16:0)PC with varying cholesterol.

S. Veatch & S. Keller, Phys Rev Lett 89:268101http://depts.washington.edu/chemfac/keller_pubs.html

What about cells, though? There are a lot of bad experiments in the “lipid raft” literature (bad is hard to interpret – as the physicist Wolfgang Pauli once sniped, “It’s not even wrong!”). Part of the problem is that lipid microdomains on living cells have been, really hard to convincingly demonstrate in intact, unperturbed cell membranes. The domains in cells seem to be smaller than ~200 nm, the limit of resolution of the optical microscope, and other methods have yielded conflicting or difficult-to-interpret results. The use of differential detergent solubility has been especially problematic. (A good critical review of these issues is: Munro S, 2003, Lipid rafts: elusive or illusive? Cell. 2003115(4):377-88.) The uncertainty is compounded by the fact that biological membranes have a lot of different components, and many “lipid raft” markers are proteins, not lipids. Moreover, commonly used methods to deplete cholesterol may have unexpected and decidedly nonphysiological effects such as causing the formation of gel or solid domains (Veatch and Keller, 2003, Biophys J 85:3074-83.)!

One thing that can be done is to ask whether the lipid raft hypothesis improves our understanding of how peripheral membrane proteins actually do their jobs. This requires careful, detailed enzymology, and that is what Klopfenstein et al. have given us in the second paper.

Now we move on to the subject that makes up the bulk of what I’ll be talking about: membrane assemlbly and transport. There’s so much to cover here, and even the things that we’ll talk about the most, we’re only scratching the surface of.

As you should all know, secreted proteins begin their life at the endoplasmic reticulum, specifically at the rough endoplamsmic reticulum. As we’ll discuss tomorrow, it was known by the mid-1960s that secretory proteins first appear at the rough ER. By the time he gave his Nobel lecture in 1974, George Palade could draw this cartoon. Note the docked ribosomes on rough ER and the addition of saccharides and formation of disulfide bonds in the ER lumen.

Intracellular Aspects of the Process of Protein Secretion 183

Fig. 4. Diagram of the segregation step.

newly synthesized chain separates with the microsomal vesicles and does not

appear in the incubation medium, which topologically is the in vitro equivalent

of the cell sol. Since it had already been established by Sabatini et al (3 1)

that the ribosomes are attached to the ER membrane by their large subunits

i.e., the bearers of nascent chains) (Fig. 3), it was concluded that segrega-

tion is the result of a vectorial transport of the newly synthesized polypeptide

from the large ribosomal subunit through the ER membrane to the cisternal

space.

This conclusion provides a satisfactory explanation for the basic structural

features of the endoplasmic reticulum: a cavitary cell organ of complicated

geometry which endows it with a large surface. All these features make sense if

we assume that one of the main functions of the system is the trapping of

proteins produced for export. With the exception of Ca2+ accumulation in the

sarcoplasmic reticulum, i.e., the equivalent cell organ of muscle fibers, no

other recognized function of the endoplasmic reticulum (e.g., phosphatide-

and triacylglycerol synthesis, mixed function oxygenation, fatty acid desatura-

tion) requires compellingly and directly a cavitary organ, at least according to

our current knowledge. In detail, however, the forces and reactions involved

in the trapping operation remain unknown. The interaction of the large ribo-

somal subunit with the ER membrane is understood only in very general terms

(30), and precise information bearing on specific molecules involved in at-

tachment is still lacking. Segregation appears to be an irreversible step: the

nascent polypeptide is extruded in the cisternal space and, once inside, it can

no longer get out (Fig. 4).

The membrane of isolated microsomes was found to be highly permeable to

Now, unfortunately, we have to skip through some of the most wonderful biochemical analysis done in our field, and just show a cartoon, which depicts the recognition of a nascent signal sequence by the SRP; arrest of translation; docking of SRP and SR (SRP Receptor), dissociation of SRP-SR and receptor with concomitant resumption of translation, now into the preprotein translocase (Sec61/SecYE complex), stimulation of mutual GTP hydrolysis by SRP-SR, and completion of coupled translation/translocation.

PLoSBiology | www.plosbiology.org October 2004 | Volume 2 | Issue 10 | e3201578

Reciprocal Activation of Two GTPases

Key features include the idea that translocation itself takes place through an aqueous pore made by an integral membrane protein, the tranlocase, and that the polypeptide translocates in an unfolded state. There is still considerable debate about whether the

translocating polypeptide can have any secondary structure, particularly helical structure. But it’s clear that it can’t have any larger folded domains. As I say in the Syllabus, you really should read at least the key papers that led to this picture! Even more frustratingly, we are not even touching on equally beautiful work, including the reconstitution of preprotein translocases by Wickner, Rapoport, and others, or the dissection of mitochondrial protein import by W. Neupert, U Hartl, N Pfanner, and their colleagues.

However, for the preprotein translocase, we now have something wonderful: a crystal structure. So I’m going to spend a few minutes describing this structure, and then we can talk about today’s last paper, which relies on this structure. It has several key features. First of all, we can see that the translocase is rather bulky, and that it appears to have funnel-like structures leading to a central constriction at the center. In fact, in the crystal there is not a pore at all! The sides of the funnels are lined by hydrophilic side chains, consistent with the notion, derived from biochemical studies byBlobel, Wickner, Rapoport, and others, that the pore is water-filled. The central constriction – is plugged by this little green alpha helix in the left cartoon, conected by long loops:

philic residues (Fig. 4a, b), but there are few charges (not shown).The absence of charges is particularly remarkable for the cyto-plasmic funnel; its interior is uncharged, but its outer rim in thecytoplasm contains a large number of both positive and negativeresidues.

The pore ringThe nature of the constriction suggests a mechanism by which themembrane barrier can be maintained during protein translocation.At its narrowest point, the channel is lined by a ‘pore ring’ of six

hydrophobic residues (Ile 75, Val 79, Ile 170, Ile 174, Ile 260 andLeu 406) (Fig. 3b, c). In E. coli, the corresponding residues are allisoleucines, and this amino acid is frequently found as pore residuesin other species (Supplementary Fig. S1). The opening formed bythe ring is about 5–8 A. The hydrophobic residues may form agasket-like seal around a translocating polypeptide, and bulkyisoleucines may be particularly suitable seal residues. Differentamino acids must pass through the pore ring, and the requiredflexibility may be provided largely by lateral shifts of the helices towhich the pore residues are attached (see below). Although the sealis not likely to be perfect, it would significantly hinder the passage ofsmall molecules during translocation. This model is different fromprevious proposals, in which the ribosome–channel junction andthe binding of the lumenal protein BiP provide the seal in co-translational translocation inmammals3,27. Ourmodel explains howthe membrane barrier can be maintained in both co- and post-translational translocation pathways, and why a gap between theribosome and the channel, seen in EM structures of the eukaryoticribosome–channel complex15–18, may not compromise the mem-brane barrier.The hourglass shape of the open channel would minimize its

interactions with a translocating polypeptide chain. Contacts wouldbe confined largely to the constriction formed by the pore residues.Because the constriction consists of only a single layer of hydro-phobic side chains (Fig. 3b), interactions with a polypeptide at itscentre would probably be weak, even when hydrophobic aminoacids are passing through. The lack of charges in the cavities oneither side of the constriction may also help to minimize inter-actions between the channel and the translocating chain.The diameter of the pore ring seen in the X-ray structuremight be

sufficient to accommodate an extended polypeptide chain, but witha little expansion it could allow passage of an a-helix. An increase inpore size could result from shifts in the helices that line the channel,perhaps facilitated by rearrangement of the loop between TM4 andTM5, in which the conserved Gly residues of a Gly-Ile-Gly-Ser-Gly-Ile-Gly sequence could serve as hinges. TM10 could also move,facilitated by conserved Gly residues in the membrane-embeddedloop between TM9 and TM10, related by pseudo-symmetry to the4/5 loop (Supplementary Fig. S5). These shifts could allow a variablepore width, and might explain how even bulky residues attached tothe side chains of amino acids in vitro28 or a disulphide-bondedpolypeptide loop of 13 residues in a secretory protein29 can betransferred through the channel. The pore size would effectivelyprevent the passage of folded domains, consistent with experimen-tal data30,31. Unless the pores of several copies of Sec61p/SecYcomplexes fuse into a larger pore (see below), the structure seems

Figure 2 Comparison with the 2D crystal structure of the E. coli SecY complex. a, X-raystructure of the M. jannaschii SecY complex (coloured as in Fig. 1a), visually docked into

the electron density map of the E. coli SecY complex, determined by cryo-electron

microscopy of 2D crystals19. Shown is a 5 A slab, viewed from the top, with the map

contoured at 1 j. The M. jannaschii TMs are numbered. TMs of the E. coli complex with

no correspondence in M. jannaschii were fitted into the density as grey cylinders. The

diamond symbol indicates the axis of two-fold symmetry in the E. coli complex.

b, Modelled dimer of the SecY complex, in the same orientation as in a, with TM2b andTM7 coloured as in Fig. 1a. TM2a (plug) is shown in dark green. A cysteine introduced at

the position indicated by a red sphere results in efficient crosslinks (X) between two g-

subunits.

Figure 3 The channel pore. a, View from the top with TM2a (plug) coloured in dark green.

b, View from the side with the front half of the model cut away. The modelled plug

movement towards the g-subunit (magenta) is indicated. The hydrophobic pore ring is

shown by the side chains coloured in gold. c, Top view with the plug modelled in its open

position. TM2b and TM7 are coloured as in Fig. 1a. The star indicates the region where

introduced cysteines result in crosslinking between TM2a and the TM of the g-subunit.

articles

NATURE |VOL 427 | 1 JANUARY 2004 | www.nature.com/nature 39© 2003 Nature Publishing Group

van den Berg et al., Nature 427:36

The idea, and there is some evidence consistent with it, is that the plug swings out of the way when transocation begins (middle). Once the plug swings out there’s an hourglass shaped pore. At the constriction of the hourglass, we see, a set of leucines, the“pore ring” shown in yellow. It’s thought that these leucines form a narrow, greasy “gasket” that prevents the movement of aqueous solvent during translocation.

So this is really satisfying. But there’s a huge question that is unanswered: if you have a transmembrane protein with cytoplasmic and extracytoplasmic domains, how does the transmembrane-spanning polypeptide escape into the cytoplasm?

form by the association of two dimers side by side, similar to theirarrangement in the 2D crystal lattice19. Crosslinks between the TMof the E. coli g-subunit (SecE) and TM2 and TM7 of the a-subunit(SecY) have been observed49, but they do not fit with either front-to-front or back-to-back orientation of the monomers.Ribosome-associated Sec61p channels contain three or four

copies of the complex15–18. In view of their sequence similarity(Supplementary Fig. S1), it seems likely that the co-translationaleukaryotic and the post-translational bacterial complexes functionsimilarly. We therefore favour a model in which two dimers, with aback-to-back arrangement of the complexes, associate side by sidebeneath the ribosome. The front sides of all complexes would thenface outwards, and only one complex at any given time would formthe active pore and contain a translocating polypeptide. As theribosome is asymmetric and wouldmake different contacts with thefour copies of the Sec61p/SecY complex, the monomers may havedifferent conformations. Our model implies that the appearance ofa pore in low-resolution EM structures of ribosome–channel

complexes is simply an indentation between the subunits ratherthan a channel. The recent higher-resolution EM structures do nothave a pore, and the observed central indentation is offset from theexit site of the nascent chain from the ribosome17,18.

Oligomers of the Sec61p complex may also be required for post-translational translocation in S. cerevisiae14, but there is no evidencethat the signal sequence intercalates between TMs of differenta-subunits50. Thus, in all three translocation pathways, a singlecopy of the Sec61p/SecY complex in an oligomeric assembly mayform the translocation pore.

If the active pore is formed by a monomer, what is the role ofoligomerization? Allosteric interactions between the monomers,regulation of the binding of partners and display of sites forrecruitment of other factors (for example, signal peptidase, oligo-saccharide transferase, TRAM or TRAP) are possible answers.Nonetheless, the issue remains an important puzzle for future work.

A model for protein translocationThe X-ray structure allows us to propose the following refinedmodel for the translocation of secretory proteins. Initially, thechannel is closed because the plug blocks the pore (Fig. 7, stage1). Next, a channel partner binds; depending on the mode oftranslocation and the organism, this can be a ribosome, theSec62/63p complex or SecA (Fig. 7, stage 2 indicates the situationwith a ribosome). Although part of an oligomer, only one copy ofthe SecY/Sec61p complex forms the active pore. The closed state ofthe channel may be destabilized by a conformational change, butbinding of the partner alone is insufficient to completely open thechannel. In the next step, the substrate inserts as a loop into thechannel, with its signal sequence intercalated between TM2b andTM7, and with its mature region in the pore (Fig. 7, stage 3).Insertion requires a hinge motion to separate TM2b and TM7, anddisplacement of the plug to its open-state position close to theg-subunit. The mature region of the polypeptide chain is thentransported through the pore, and the signal sequence is cleaved atsome point during translocation (Fig. 7, stage 4). While thepolypeptide chain is moving from an aqueous cytoplasmic cavityto an external one, the pore ring forms a seal around the chain,hindering the permeation of other molecules. Finally, when thepolypeptide has passed through, the plug returns to its closed-stateposition (Fig. 7, stage 5). Membrane protein biosynthesis, althoughless well understood, might occur in a similar way, except that TMscould move from the lateral gate all the way into lipid, and cytosolicdomains would form by emerging through the gap between ribo-some and channel. Several of these points differ from conventional

Figure 7 Different stages of translocation of a secretory protein. See main text fordescription.

Figure 6 Signal-sequence suppressor (prl) mutations. a, Stereo view from the top with the

Cb atoms of the residues corresponding to E. coli prl mutations in the a-subunit (SecY)

and g-subunit (SecE) shown as pink and purple spheres, respectively. TM2a, TM2b and

TM7 are highlighted in colour. b, View from the front.

articles

NATURE |VOL 427 | 1 JANUARY 2004 | www.nature.com/nature42 © 2003 Nature Publishing Group

also be identified from a comparison with the M. jannaschiicomplex. The bacterial b-subunit (SecG) comprises two TMs, thesecond of which has the same position and orientation as the TMofSecb of M. jannaschii25 (Fig. 2a), suggesting that they may haveanalogous functions. The N-terminal TM of the bacterial b-subunit(SecG) has no correspondence in archaea and eukaryotes. Theg-subunit (SecE) of E. coli has two non-essential TMs at its Nterminus23, which correspond to the two helices that are far apartfrom the others in the 2D crystal structure of the E. coli protein(Fig. 2a). These helices are in approximately the same location as theN terminus of the M. jannaschii g-subunit in the X-ray structure(Fig. 2b).

Translocation pore and plugExperiments in different systems have shown that the SecY/Sec61pcomplex forms oligomers during translocation, but as we will arguelater, our structure suggests that a single copy of the SecY/Sec61pcomplex serves as a functional channel. A large, funnel-like cavitywith a diameter of 20–25 A at the cytoplasmic side of the SecYcomplex could serve as a channel entrance (Fig. 3a, b). It containsmany conserved residues (Supplementary Fig. S6), suggesting that ithas an important function. The funnel tapers to a close in themiddle of the membrane (Fig. 3b), indicating that the structurecorresponds to a closed channel, impermeable to polypeptides oreven small molecules.How might the channel open and how could it recognize signals?

A large body of data in the literature allows us to propose specificmodels for these and other properties. TM2a, which we call the‘plug’, blocks the bottom of the funnel about halfway across themembrane, and separates the cytoplasmic side from the externalaqueous space (Fig. 3a). We propose that the channel opens for

polypeptide translocation by displacement of the plug. Consistentwith this hypothesis, a cysteine introduced into the plug at residue67 of the E. coli a-subunit (SecY; corresponding to M. jannaschiiThr 61) can form a disulphide bridge in vivo with a cysteineintroduced at residue 120 of the E. coli g-subunit (SecE;M. jannaschii residue 64)26. These residues are more than 20 Aapart in the closed state of the channel (Fig. 3a). Disulphide bridgeformation results in a dominant-negative phenotype, as would beexpected if the channel were locked into a permanently open state.Another combination of cysteines (E. coli residue 64 in the a-sub-unit (SecY) and 124 in the g-subunit (SecE)) is also lethal, whereascombinations of neighbouring residues are not26, suggesting thatthe helical structures of the plug and the g-subunit are maintainedduring movement. Given that the TM of the g-subunit is acontinuous helix with one hydrophobic side, we assume that itremains stationary while the plugmoves as a rigid body into a cavityon the external side of the channel next to the C terminus of theg-subunit (Fig. 3b, c). This displacement requires a translation of,22 A towards the back of the molecule, as well as a shift of about12 A towards the external side of the membrane. The hinges for themotion could be the Gly residues at positions 49 and 68. Althoughnot universally conserved, all species have small residues in thisregion that could serve this function. Movement of the plug wouldopen the pore (Fig. 3b, c), resulting in an hourglass-shaped channelwith aqueous funnels that taper to a constriction in the middle ofthe membrane (Fig. 3b).

The funnel-like cavities on both sides of the constriction wouldcreate an aqueous channel across the membrane (Fig. 4a, b). Thisfeature is consistent with fluorescence life-time measurements3,which suggest that a translocating polypeptide is in an aqueousenvironment. The walls of both funnel-like cavities contain hydro-

Figure 1 General architecture of the SecY complex. a, Stereo view from the cytosol (top).

The a-subunit is coloured blue to red from the N to the C terminus with the TM segments

numbered; the b-subunit is shown in pink and the g-subunit in magenta. b, View from the

back, with the phospholipid head group and hydrocarbon regions of the membrane

shown in blue and grey in the background. Cytosolic loops probably involved in ribosome

and SecA binding are indicated. c, Top view with the N- and C-terminal halves of the

a-subunit in blue and red, respectively. d, Top view sliced through the middle of the

membrane. The solid lines connect the TMs from the N to the C terminus in the two halves;

the dotted arrow shows the axis of internal symmetry. e, Slab views from the front and

back, with the foreground removed and TM1 (dark blue), TM2a and TM2b (sky blue)

highlighted.

articles

NATURE |VOL 427 | 1 JANUARY 2004 | www.nature.com/nature38 © 2003 Nature Publishing Groupvan den Berg and colleagues point out that the protein can be seen as having two

major domains, and that at the interface between the TM6-10 domain and the TM1-5 domain – where I’ve drawn a green arrow on the left panel – the structure is topologically open, with mainly hydrophobic residues at that interface. The idea is that the transocase is a clamp that opens (or gates) laterally into the plane of the membrane. During transocation, it’s going to be dynamic, and the translocating polypeptide is thermally vibrating. So the gate is opening and closing, and the whole time the translocating polypeptide is sampling the (mostly) aqueous pore, and the more hydrophobic domain interface, and the really hydrophobic membrane, deciding where it “wants” to be. That is, the tranlocase allows the polypeptide to decide if it’s going to partition into the membrane cor or the aqueous phase.

This is the context into which our third paper falls, and it is worth going back, and looking at the first figure in these notes as you think about the paper from von Heijne’s group.

On Wednesday, we’ll start to think about how the secretory protein that’s just gone through the translocase can be packaged, transported to the cell surface, and secreted, and the membrane processes that make this transport possible, as well as inbound transport through the endocytic pathway.