News & Views | Published:


A class act

Membrane fusion occurs in many situations in living organisms — when certain viruses enter host cells, for instance. Three crystal structures shed light on the protein rearrangements that bring about such fusion.

Many viruses, such as the dengue fever, influenza and human immunodeficiency viruses, are encased in a lipid membrane. To reproduce, such 'enveloped' viruses must enter a host cell by fusing their own membrane coat with that of the cell. Fusion is caused by specific proteins in the viral membrane, and at least two classes of these proteins have been identified. In both classes, tightly regulated conformational changes are involved in membrane fusion. But class I and class II fusion proteins have distinct structural features — raising the expectation that the details of the two fusion mechanisms would also be unique. Yet the crystal structures of the post-fusion forms of three class II proteins, described elsewhere in this issue1,2 and in EMBO Journal3, suggest an unanticipated mechanistic parallel between class I and class II fusion proteins. The new structures also provide the first view of a key region of the proteins (the fusion loop) in the conformation that is predicted to insert into the target membrane.

The haemagglutinin protein from influenza virus has provided the model for class I fusion machines, because the atomic structures of three different forms of this protein have been determined4. The class I model, in its essential features, applies to many unrelated virus families and their fusion proteins, including the HIV gp120 protein, the F proteins from paramyxoviruses, the retroviral SU/TM proteins and the Ebola virus Gp2 protein. For influenza haemagglutinin, a drastic refolding of the protein is observed between its pre-fusion and post-fusion forms (Fig. 1). Related refolding events have been inferred for the other class I proteins5.

Figure 1: Proposed mechanism for membrane fusion by class I fusion proteins.

a, The metastable conformation of a trimeric generalized fusion protein, with helical domain A in orange, helical domain B in pink, and the transmembrane domain in purple. b, After binding to a receptor on the cellular membrane, or on exposure to the low pH found in intracellular compartments (endosomes), the protein forms an extended conformation and the hydrophobic fusion peptide (red) inserts into the target membrane. c, Several trimers are thought to be involved. d, Protein refolding begins. The free energy thereby released causes the membranes to bend towards each other. e, Formation of a restricted hemifusion stalk allows the lipids in the outer leaflets of the membranes to mix. f, Protein refolding completes, forming the final, most stable form of the fusion protein, with the fusion peptide and transmembrane domain anti-parallel to each other but in the same membrane. Only a and f have been observed by crystallography, but biochemical data support many of the proposed steps.

The second class of viral fusion protein was postulated6 from the striking structural conservation of the pre-fusion forms of three particular proteins. Two of these, the E proteins of tick-borne encephalitis virus (TBE)7 and dengue virus8, are likely to be representative of a large and diverse family of RNA viruses, the Flaviridae, which also includes yellow fever virus, West Nile virus and hepatitis C virus. The third protein — the E1 protein of Semliki Forest virus6 — represents a family of RNA viruses called Togaviridae, of which the best known is the rubella virus, which causes German measles. The post-fusion structures of dengue virus E protein1, Semliki Forest virus E1 protein2 and TBE E protein3 have now been determined; they reveal a marked and surprising convergence of the class I and II fusion mechanisms.

To understand the details of these findings, we need a little more background. Class I fusion proteins are composed of three identical protein subunits, the functional forms of which are generated from a precursor that is cleaved into two pieces. The carboxy-terminal end of one piece is anchored to the viral membrane; the other end (the new amino terminus) has a characteristic stretch of 20 hydrophobic amino acids — the fusion peptide. During fusion, the three fusion peptides become exposed and are inserted into the target cell membrane (Fig. 1), generating a differently folded intermediate that is anchored to both cellular and viral membranes. This intermediate can be inhibited by short peptides from regions of the protein that form the final core of the post-fusion form: a new HIV-inhibiting peptide, Fuzeon (enfuvirtide), blocks HIV infection at this step.

In the absence of inhibitors, however, the protein further refolds into a trimeric hairpin shape, a process that is tightly linked to fusion. An essential feature of class I proteins is the formation at this time of a trimeric helical coiled-coil rod adjacent to the fusion peptide. This structure may act as a template for the refolding of protein segments — often helical — that are near to the segments anchored in the virus membrane. Refolding relocates the fusion peptides and transmembrane anchor domains to the same end of the coiled coil, bringing viral and cellular membranes together to catalyse fusion4,5,9,10.

Class II fusion proteins have distinctly different structural features: they are predominantly non-helical, instead having a β-sheet-type structure; they are not cleaved during biosynthesis; and the portion that inserts into the target membrane is thought to be an internal hydrophobic fusion loop. The proteins have a three-domain architecture, in which domain I begins at the amino terminus, domain II contains the internal fusion loop, and domain III is at the carboxy terminus (Fig. 2). In cryo-electron-microscopy reconstructions of virus particles, dimers of E or E1 proteins lie flat upon the viral surface11. At pH values equivalent to those in endosomes (the cellular compartments in which internalized viruses meet an acidic pH), the proteins assemble irreversibly into trimers. But for the truncated, soluble proteins this occurs only in the presence of membrane lipids, suggesting that interactions with the membrane are important.

Figure 2: Proposed mechanism for fusion by class II proteins, incorporating the new results1,2,3.

a, The dimeric E protein binds to a cellular receptor (grey) and the virus is internalized to endosomes. Membrane fusion, which will release the virus into the body of the cell, takes place within endosomes. Domain I is in red, domain II in yellow, and domain III in light and dark blue (the lighter blue shows a predicted but unsolved structure). b, The acidic pH inside endosomes causes domain II to swing upward, permitting E monomers to rearrange laterally. c, The fusion loop (red dot) inserts into the outer leaflet of the host-cell membrane, enabling trimer formation. d, The formation of trimer contacts extends from the top to the bottom of the molecule. Domain III shifts and rotates to create contacts, bending the membrane. e, The formation of further contacts leads to unrestricted hemifusion. f, The final most stable form of the protein. Only a and f have been observed by crystallography but biochemical data support many of the other steps.

The crystal structures reported by Modis et al.1, Gibbons et al.2 and Bressanelli et al.3 clarify the class II fusion mechanism. Another cryo-electron-microscopy study12 had suggested that a large portion of β-sheet structure might insert into the membrane during fusion. But the new structures do not support this model: they show significant reorganization of the orientations of the relative domains, with relatively limited protein refolding and only the fusion loop inserting into the membrane. The reorientations are accompanied by disassembly of the dimer and formation of an extensive interface in the trimer that involves all three domains. This would result in the trimer lifting up from the virus surface and projecting the internal fusion loop towards the target membrane (Fig. 2). It seems most likely from the location of hydrophilic residues that the fusion loop inserts only into the outer half (leaflet) of the target membrane.

So far this is all very different from influenza haemagglutinin (and, by inference, from other class I proteins), in which refolding, rather than reorganization, is the key. But a closer look at the trimeric forms of the E and E1 proteins reveals some unexpected similarities to class I proteins. In these forms, domains I and II produce an extended, thick, rod-like structure, with the fusion loops positioned at the tips — analogous to the coiled-coil rods seen in class I proteins. Moreover, the displacement of domain III, and its relocation along the outside of the rod, produces a familiar hairpin-like structure. In the class II structures, an extended groove along the outside of the trimer is poised to interact with the segments connecting to the anchor domains in the viral membrane — again reminiscent of class I proteins. Overall, the new structures suggest that trimerization of class II proteins assembles a trio of hairpins that juxtapose fusion loops and anchor domains at the same end of a roughly 100-Å-long structure. The similarities to the class I pathway hint at evolutionary convergence upon a common mechanistic approach, but with unrelated protein architectures.

These structures1,2,3 open new approaches to the development of inhibitors of class II fusion proteins, analogous to existing class I inhibitors. The structures also — together with the pre-fusion conformations and electron-microscopic reconstructions — provide new insights into the assembly and regulation of class II fusion machines. For proteins from both classes, the transition to the post-fusion state is associated with an irreversible conformational change that probably provides the energy for membrane fusion. But assembly of the pre-fusion structures into metastable 'cocked' states that can be triggered appropriately is achieved very differently. The new structures suggest that, for class II proteins, target membranes might have a crucial catalytic or facilitating role in lowering the barrier to transition, presumably by orienting protein monomers at the membrane surface and so promoting the necessary rearrangements.


  1. 1

    Modis, Y., Ogata, S., Clements, D. & Harrison, S. C. Nature 427, 313–319 (2004).

  2. 2

    Gibbons, D. L. et al. Nature 427, 320–325 (2004).

  3. 3

    Bressanelli, S. et al. EMBO J. (in the press).

  4. 4

    Skehel, J. J. & Wiley, D. C. Annu. Rev. Biochem. 69, 531–569 (2000).

  5. 5

    Colman, P. M. & Lawrence, M. C. Nature Rev. Mol. Cell Biol. 4, 309–319 (2003).

  6. 6

    Lescar, J. et al. Cell 105, 137–148 (2001).

  7. 7

    Rey, F. A. et al. Nature 375, 291–298 (1995).

  8. 8

    Modis, Y. et al. Proc. Natl Acad. Sci. USA 100, 6986–6991 (2003).

  9. 9

    Melikyan, G. B. et al. J. Cell Biol. 151, 413–423 (2000).

  10. 10

    Russell, C. J., Jardetzky, T. S. & Lamb, R. A. EMBO J. 20, 4024–4034 (2001).

  11. 11

    Zhang, W. Z. et al. Nature Struct. Biol. 10, 907–912 (2003).

  12. 12

    Gibbons, D. L. et al. Cell 114, 573–584 (2003).

Download references

Author information

Correspondence to Theodore S. Jardetzky or Robert A. Lamb.

Rights and permissions

Reprints and Permissions

About this article

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.