Cell biology

The specifics of membrane fusion

Article metrics

The ability to maintain a diverse set of intracellular compartments, with distinct complements of proteins, is a defining feature of eukaryotic cells. Substances can be transported from one membrane-encased compartment to another, but the compartments maintain their unique identities. Transport occurs in membrane-bounded containers called vesicles, and several protein families have evolved to mediate the budding of a vesicle from the donor compartment, and its transport to and fusion with the target organelle. One of the last steps in the fusion process is overseen by a set of proteins called SNAREs. These have been suggested to be the core machinery that mediates the fusing of two membranes, as well as ensuring that vesicles deliver their cargo to the right compartment1,2. Writing on pages 153, 194 and 198 of this issue3,4,5, Rothman and colleagues conclude — with some caveats — that SNAREs are indeed important in defining the specificity of vesicle targeting.

SNAREs contain structural features called α-helices or coils. During membrane fusion, four α-helices from SNAREs found on the vesicle and target membranes come together to form a stable, four-helix bundle or coiled-coil6 (Fig. 1a). The formation of SNARE complexes is essential for membrane fusion, so a tremendous amount of research has been dedicated to understanding how these complexes form, and what they do. Rothman and colleagues earlier developed an in vitro assay that measures the fusion of liposomes — artificial spheres surrounded by a lipid membrane — reconstituted with neuronal SNARE proteins7. This system is ideal for assessing the role of SNAREs in isolation. It has been used to show that, when present on liposomes representing the vesicle and the target, SNAREs — in the absence of other factors — can induce membrane fusion7.

Figure 1: SNARE proteins and vesicle targeting.

SNAREs are found on vesicle and target membranes, and between them contribute four ‘coil’ structures to a SNARE complex. a, Rothman and colleagues3,4,5 have constructed artificial membrane-encased spheres (liposomes) with different combinations of SNAREs, to assess the role of SNAREs in fusion specificity. They find, as shown previously6, that — for vesicle fusion with the plasma membrane — three Q-SNAREs (red, coil type A; green, coil types B and C) are required on the target and one R-SNARE (blue; coil type D) on the vesicle4,5. SNAREs confer some specificity to vesicle fusion3,4,5. b, Q-SNAREs and R-SNAREs are characterized by a glutamine (Q) or arginine (R) residue, respectively, in the central layer of the SNARE complex. Top, the SNARE complex required for fusion of a synaptic vesicle with the plasma membrane (modified from ref. 6); bottom, the underlying molecular interactions. Black, the hydrophobic layers that mediate the core interactions; orange, the Q/R layer. Details of the interactions are shown underneath. V, valine; I, isoleucine; Q, glutamine; R, arginine; K, lysine; D, aspartic acid. The ball-and-stick structures represent the indicated amino acids; the dotted lines represent hydrogen bonds or salt bridges that stabilize interactions between SNAREs.

As well as being involved in driving membrane fusion, SNARE proteins have been implicated in ensuring the accuracy of vesicle trafficking. There appear to be enough SNAREs, differently localized throughout intracellular membrane compartments, to confer specificity on the process. For example, a transport vesicle that buds from the endoplasmic reticulum (ER) is destined for the Golgi, but how does the cell make sure that it doesn't fuse with, say, the plasma membrane instead? The idea is that, in this case, the SNAREs on the ER-derived vesicle can form a complex only with those on the Golgi, ensuring specificity (Fig. 1a). Such SNAREs are said to be ‘cognate’.

Cognate SNAREs have been characterized for some vesicle-trafficking steps, and several themes have emerged from these studies, the most important of which is the composition of the SNARE complexes. A four-helix bundle always contains one coil from each of four families6,8, abbreviated here for simplicity as A, B, C and D coils (for specialist readers, A designates the Q-SNARE syntaxin, B and C the Q-SNAREs of the SNAP-25 family, and D the R-SNARE of the VAMP family; Fig. 1b).

Rothman and colleagues3,4,5 have now used their liposome-based system to investigate whether isolated yeast SNAREs have a role in ensuring fusion specificity. The power of this experimental set-up is that any combination of SNAREs can be put into the ‘vesicle’ and ‘target’ liposomes. The authors find that there is a great deal of specificity in terms of which liposomes fuse together. But this specificity comes from different sources.

For example, the physical arrangement (topology) of the SNAREs can confer specificity. The authors tested the SNAREs that (in cells) are involved in transport between the ER and the Golgi, and found that liposome fusion occurred only when the A, B and D coils (from yeast SNARE proteins Sed5, Bos1 and Sec22, respectively) were on one membrane and the C coil (Bet1) was on the other3. No other combination of ER–Golgi SNAREs tested resulted in fusion.

It would be interesting to see whether such topological restrictions are important in vivo, and whether they extend to other trafficking steps. They clearly don't apply in the case of the yeast plasma-membrane SNARE (Sec9), because the B and C coils are fused into one protein, preventing the C coil from being on the other membrane. Indeed, the authors show4 that fusion involving plasma-membrane SNAREs occurs when A, B and C coils are on the same membrane, with the D coil on the other. This arrangement also works for the vacuolar SNAREs tested4,5. One explanation for the unexpected ER- to-Golgi arrangement could be that Sec22 functions in transport from the Golgi to the ER, whereas the other three SNAREs function in transport from the ER to the Golgi9,10. So Sec22 may not contribute the correct D coil in this case3.

Specificity may also come from the interactions of cognate SNAREs. An intuitively appealing hypothesis was that only cognate SNAREs could form complexes. But this theory was called into question by in vitro studies showing that, although an A coil, a B coil, a C coil and a D coil were strictly required for complex formation, it did not matter from which SNARE protein these coils originated8,11. However, another study12 showed that, in a particular cell line, only cognate SNAREs, when added in solution, could compete with membrane-bound SNAREs and inhibit vesicle fusion — with a couple of exceptions, non-cognate SNAREs in solution could not. Rothman and colleagues4 again used the liposome assay to look at this problem. They found that one of each coil — A, B, C and D — are absolutely required. For many incorrect SNARE combinations (such as an A coil, two Bs and a D; or an A, a B and two Ds), fusion does not occur.

But the true test of specificity is to find out whether (for example) different D coils can substitute for each other, rather than for C coils. The authors find several examples in which cognate A, B, C and D coils (for example Vam3, Vam7, Vti1 and Nyv1) produce more fusion than non-cognate coils (Vam3, Vam7, Vti1 and either Snc1 or Sec22). They did not investigate what happens when A coils are swapped, leaving open the possibility that some non-cognate SNARE interactions might still result in fusion. Indeed, there are examples in which non-cognate A, B, C and D coils result in fusion. With the plasma-membrane A, B and C coils, fusion can occur when any D coil is used4. So isolated SNAREs cannot solely account for the specificity observed in vesicle trafficking.

As the authors say4, it is likely that specificity is not uniquely encoded in the inherent ability of SNAREs to form complexes with each other. Given that SNAREs bind to each other through their conserved features (Fig. 1b), this is perhaps not surprising. Other proteins required earlier in the fusion process — such as members of the Sec1 and Rab families — may help cognate SNAREs to recognize each other. And the organization of the cell probably adds to specificity. For example, an ER-derived vesicle is more likely to encounter the Golgi than the plasma membrane, so fusion with the plasma membrane is unlikely to occur, even if the ER and plasma-membrane SNAREs were able to form a complex in vitro . It seems that the accuracy of vesicle targeting is safeguarded not through a single lock-and-key interaction between SNAREs, but rather through several layers of constraints — a situation common to many biological processes.


  1. 1

    Bennett, M. K. & Scheller, R. H. Proc. Natl Acad. Sci. USA 90, 2559–2563 (1993).

  2. 2

    Söllner, T. et al. Nature 362, 318–324 (1993).

  3. 3

    Parlati, F. et al. Nature 407, 194–198 (2000).

  4. 4

    McNew, J. A. et al. Nature 407, 153–159 (2000).

  5. 5

    Fukuda, R. et al. Nature 407, 198–202 (2000).

  6. 6

    Sutton, R. B., Fasshauer, D., Jahn, R. & Brunger, A. T. Nature 395, 347–353 (1998).

  7. 7

    Weber, T. et al. Cell 92, 759–772 (1998).

  8. 8

    Yang, B. et al. J. Biol. Chem. 274, 5649– 5653 (1999).

  9. 9

    Spang, A. & Schekman, R. J. Cell Biol. 143, 589–599 (1998).

  10. 10

    Cao, X. & Barlowe, C. J. Cell Biol. 149, 55–66 (2000).

  11. 11

    Fasshauer, D., Antonin, W., Margittai, M., Pabst, S. & Jahn, R. J. Biol. Chem. 274, 15440–15446 (1999).

  12. 12

    Scales, S. J. et al. Neuron 26, 457–464 (2000).

Download references

Author information

Correspondence to Suzie J. Scales.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Scales, S., Bock, J. & Scheller, R. The specifics of membrane fusion. Nature 407, 144–146 (2000) doi:10.1038/35025176

Download citation

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.