Vesicle Budding, Targeting and Fusion

A yeast mutant called sec1 grows at 25 °C. At 37 °C it keeps making secretory enzymes, but they stay inside: the internal pool rises up to eightfold, packed into membrane-bound vesicles that never reach the surface.

Novick and Schekman reported that result in 1979. The vesicles had formed, so budding worked. Something later had failed. This lesson asks what a vesicle needs at each stage: a way to pick its cargo and bend off a membrane, a way to recognize the right target, and a way to merge with it. It also asks how anyone worked out which protein acts at which stage.

A few terms first. The cytosol is the fluid of the cell outside organelles; the lumen is the inside of a membrane-bound compartment. A vesicle is a small membrane sphere, often 50 to 100 nm across, that carries lumenal cargo and membrane proteins from a donor compartment to a target. Because the vesicle’s lumen is topologically outside the cytosol (see membrane topology), everything that steers it has to act from the cytosolic face.

Two problems, choosing cargo and bending membrane: coats

A coat is a lattice of cytosolic proteins that assembles on a patch of membrane. Think of a coat as a cookie cutter with hands: the cutter shapes the bud, and the hands grab the cargo that belongs inside. Both jobs are done by the same proteins.

Each coat is started by a switch. At the ER it is Sar1, a small GTPase (the GTP switch cycle, with its GEFs and GAPs, is covered in GTPase switches). Sar1-GTP binds the ER membrane and recruits the Sec23/24 complex, which in turn recruits Sec13/31, the outer cage. Together these make COPII. With purified yeast proteins, Matsuoka and colleagues showed in 1998 that Sar1, Sec23/24 and Sec13/31 alone bud coated vesicles from chemically defined liposomes that contain PI4P or PI(4,5)P2, so the coat can bend membrane without help from other ER proteins. Cargo choice sits in Sec24: yeast mutant studies support a model in which Sec24 has several independent cargo-binding sites.

At the Golgi, Arf1-GTP recruits coatomer, a seven-subunit complex, to form COPI vesicles. COPI works within the Golgi and from Golgi back to ER. Its α- and β′-COP subunits bind cytosolic dilysine (KKXX) motifs, so ER membrane proteins that escape are fetched back.

At the plasma membrane, clathrin forms the outer lattice. Clathrin does not bind cargo itself; the adaptor AP2 does. Crystal structures of recombinant AP2 show a locked form in which part of the β2 subunit covers both cargo-binding sites. On membranes containing PI(4,5)P2, a phosphoinositide concentrated in the plasma membrane (see lipid anchors and addresses), AP2 opens and exposes them. The lipid and the cargo together tell the adaptor where it is.

Two cautions keep the cartoon honest. In cultured BSC-1 cells, clathrin lattices began to curve only at around 70% of their final clathrin content, and raised membrane tension stalled the change, so a coat does not always curve from the start. And in mammalian cells, whole-cell imaging in 2021 found ER export through an interwoven tubular network, not only through separate small spheres.

Getting an address: Rabs and tethers

Once the coat is shed, a vesicle must find its target among many membranes. Part of the answer is Rab proteins, a large family of small GTPases anchored to membranes by prenyl groups. A GDP-bound Rab is held in the cytosol by GDI (GDP dissociation inhibitor). A GEF on the right membrane loads GTP, and the GTP-bound Rab recruits effectors. Picture a Rab-GTP as a lit sign over one loading bay.

Among the effectors are tethers: long coiled-coil proteins or large complexes that capture a vesicle before its SNAREs engage. In yeast, the tether Uso1p works with the Rab Ypt1p on ER-to-Golgi vesicles. On the yeast vacuole, the six-subunit HOPS complex binds the Rab Ypt7; the purified complex is about 43 nm tall, and a structural working model places its Vps39 and Vps41 subunits at the Ypt7 contacts. HOPS also contains an SM protein (Sec1/Munc18 family), Vps33, which with Vps16 binds SNAREs.

No single molecule is the address. Rab, phosphoinositide, tether and SNARE each narrow the choice, and the evidence fits a combination better than any one label.

Fusion: SNAREs zipper, NSF resets

Two bilayers resist merging because their water-coated surfaces repel at close range. SNAREs supply the force. Each SNARE has a SNARE motif of about 60 to 70 residues next to a membrane anchor. Four motifs, classed Qa, Qb, Qc and R, wind into a parallel four-helix bundle. At the synapse, syntaxin-1 gives Qa, SNAP-25 gives both Qb and Qc, and synaptobrevin gives R.

While the motifs come from two membranes, the bundle is a trans-SNARE complex. It zippers from the N-terminal end toward the membrane anchors, drawing the bilayers together until they fuse. Afterwards all four sit in one membrane as a cis-SNARE complex. NSF (Sec18 in yeast), an ATPase, binds that complex through its adaptor SNAP (Sec17) and uses ATP to pull it apart for reuse.

How much can SNAREs do alone? In 1998 Weber and colleagues put recombinant v-SNAREs (vesicle) and t-SNAREs (target) into separate liposomes, and the docked vesicles fused at physiological temperature. That shows SNAREs are sufficient for fusion of those liposomes. It does not show they supply specificity or speed in a cell. Purified yeast HOPS makes the gap visible: truncated forms lacking parts of Vps11 or Vps18 still tethered Ypt7 liposomes but failed to promote fusion. NSF’s role has also grown. In a 2024 yeast vacuole reconstitution, Sec18 promoted fusion even with Sec17 fixed to the membrane, beyond its job of recycling SNAREs.

Ordering a pathway with mutants and test tubes

The 1980 Novick, Field and Schekman screen enriched dense yeast cells on a Ludox gradient and recovered 188 clones that trapped invertase at 37 °C. They fell into 23 complementation groups (see complementation and pathways). That is the screen’s yield, not the number of genes secretion needs, and the SEC numbers carry no information about order. Order came from double mutants in 1981.

Here is the logic with made-up numbers. Invertase is measured in arbitrary units after 1 h at 37 °C.

fraction secreted = external / (external + internal)

Step 1. Wild type: external 180, internal 20, total 200, so 180/200 = 90% secreted. Mutant secX: external 12, internal 188, total 200, so 12/200 = 6%. The block is real.

Step 2. Is the trapped invertase dead protein or a stalled intermediate? Shift secX back to 25 °C with cycloheximide to stop new synthesis. After 60 min external is 150 and internal 50, total still 200. The pool released is (188 − 50)/188 = 138/188 = 73.4%. External rose by the same 138 units. The trapped material was deliverable.

Step 3. Electron microscopy at 37 °C shows secX accumulates ER, secY accumulates Golgi-like stacks (Berkeley bodies), and secZ accumulates 80 to 100 nm vesicles. The secX secZ double mutant accumulates ER, so X acts before Z. The secY secZ double accumulates Golgi-like membranes, so Y acts before Z. ER precedes Golgi on the secretory route, which places X before Y; a secX secY double accumulating ER would test that directly. Order: ER (X), Golgi (Y), vesicles (Z), surface. The assumption is a linear pathway with distinguishable arrest states; a gene product could still act early and be needed only later.

Test tubes split the steps further. In Barlowe’s 1994 yeast assay, vesicles made with GMP-PNP formed but did not target or fuse with Golgi, so budding and delivery are separate. In 1998 Cao, Ballew and Barlowe added GDI at different times: once vesicles had docked, fusion no longer cared. Timed addition places a step.

Design problem

You have a cell-free assay with donor yeast ER membranes, acceptor Golgi, cytosol and ATP. You want to know whether GDI, which strips Rabs from membranes, acts before or after vesicles dock, and whether a purified tether is enough for fusion. Design the experiment.