Cells use various protein complexes to remodel membrane-bound organelles. In vitro reconstitution of the activity of one such complex, ESCRT-III, shows that it promotes membrane bending in an unconventional way.
Many transmembrane proteins — including those that span the cell membrane — are degraded within cellular organelles called lysosomes once their work is done. Reaching the lysosomes is not simple, however, and involves several membrane-vesicle trafficking steps1. On page 172 of this issue, Wollert et al.2 elucidate the role of the components of the ESCRT-III protein complex in one stage of this process. The authors' results are of interest not just for the insights they provide into the processes of intracellular trafficking and protein degradation, but also because ESCRT proteins mediate the budding of certain viruses, including HIV-1, as well as the separation of the two daughter cells at the end of cell division.
So what is the itinerary for the journey of a cell-membrane protein to the lysosomes? First, small portions of cell membrane invaginate and become detached to form organelles called endosomes, incorporating the proteins. The endosome further invaginates internally into its own lumen to form several intraluminal vesicles (ILVs). Endosomes with their ILVs are called multivesicular bodies, and eventually fuse with lysosomes, where ILVs and their contents are degraded by the cocktail of lysosomal enzymes.
Compared with other membrane-deformation events that occur in the cell, ILV formation is puzzling (Fig. 1). Vesicles generally bud from organelles (such as the Golgi complex) in the opposite direction to ILV formation: a protein coat assembles on the cytoplasmic side of the organelle membrane to form an outer shell and impose deformation towards the cytoplasm3. Once the vesicle has become detached from the membrane, the protein coat disassembles and is reused for another round of budding.
Given the opposite topology of ILVs, budding away from the cytoplasm, could it be that ESCRT-III sculpts the membrane from inside the bud? After all, induction of curvature from the inside has precedents: some proteins with convex shapes adhere to lipid membranes to form protrusions4, and some toxins, such as Shiga toxin, invade the cell by inducing their own uptake into cell-membrane invaginations5(Fig. 1). But the problem with ESCRT-III pertains to its recycling; if they act from within the bud, ESCRT-III proteins would be trapped in the ILV and so would be used only once.
Wollert et al.2 asked whether ESCRT-III might act outside the invagination, which would allow its subsequent recycling. To address this question, they used the well-established approach of biochemical reconstitution. In short, they added purified recombinant yeast ESCRT-III proteins to artificial, giant unilamellar (single-membrane) vesicles (GUVs) that resemble cellular membranes, and studied them using light microscopy. Their observations are most informative.
On incubation with three components of ESCRT-III (Vps20, Snf7 and Vps24), the GUVs become filled with internal vesicles that are clearly detached from the surrounding membrane. This finding suggests that these three proteins are sufficient to promote the entire process of ILV formation — from the induction of membrane curvature to the scission of the bud neck.
In a second set of experiments, the authors added two further ESCRT-III proteins to the mixture: Vps4, an ATPase enzyme that mediates the disassembly of ESCRT-III proteins, and Vps2, the final component of ESCRT-III, which connects Vps4 to the rest of the ESCRT-III proteins. This time, Wollert et al. found that a second wave of ILVs followed the first, being identified by having their contents stained with a green dye. So, although Vps4 is not involved in the formation of ILVs, its ATPase activity seems to allow the recycling of the ESCRT-III machinery.
Wollert and colleagues' strategy seems so straightforward that one wonders why it was not used before. But biochemical reconstitution with five pure proteins and a model lipid membrane is more difficult than it sounds. To prevent self-aggregation of each protein, the authors had to use specifically engineered proteins. Moreover, the absolute concentration of the proteins, their relative ratios and the lipid composition of the GUVs were all carefully chosen. Of course, a GUV is not a cellular organelle, but these precautions highlight the relevance of the effects observed.
The results2 support the idea that, indeed, ESCRT-III drives membrane invagination and fission while itself remaining at the door (or neck) of the forming ILV. But our pen trembles when we consider drawing an explicit model. The structures of some of the proteins involved in membrane deformation speak for themselves — take the triskelion shape of the clathrin protein, or the banana shape of BAR-domain proteins, for example. This is not, however, the case for ESCRT-III proteins, which share a simple elementary fold that can be envisaged as a small, flat jigsaw puzzle, with a basic charge well suited for interaction with the negatively charged membrane lipids6.
Recent work7,8,9,10suggests that ESCRT-III proteins assemble in an ordered manner at the surface of lipid membranes, where they form spiral filaments (Fig. 2). The spiral shape allows ESCRT-III to act as a fence encircling cargo proteins en route to the nascent ILV. But what drives membrane invagination?
If the spiral is smooth, and if the basic face of the monomers is parallel to the main plane of each turn (like a flat, coiled piece of rope; see Fig. 6 on page 176), it is hard to imagine how it could drive invagination: the spiral should happily sit on a flat, negatively charged membrane. But if the basic surface is tilted, and the spiral has some stiffness, each turn might come out of the previous one, both to escape mechanical constraint and to interact with the membrane (Fig. 2). Under these circumstances, a conical invagination should form. But the spiral should not invade the neck so deeply as to create a narrow, empty lipid tubule ready to undergo scission. Intriguingly, Vps24 seems both to act as a stop signal for the elongation of the spiral10 and to favour the ILV fission step2. Such speculations highlight the fact that ESCRT-III is so different from other protein machinery involved in intracellular trafficking that Wollert and colleagues' study2, and those of others7,8,9,10, are probably just the beginning of many surprises yet to come.
Hurley, J. H. & Emr, S. D. Annu. Rev. Biophys. Biomol. Struct. 35, 277–298 (2006).
Wollert, T., Wunder, C., Lippincott-Schwartz, J. & Hurley, J. H. Nature 458, 172–177 (2009).
Antonny, B. Curr. Opin. Cell Biol. 18, 386–394 (2006).
Mattila, P. K. et al. J. Cell Biol. 176, 953–964 (2007).
Romer, W. et al. Nature 450, 670–675 (2007).
Muziol, T. et al. Dev. Cell 10, 821–830 (2006).
Teis, D., Saksena, S. & Emr, S. D. Dev. Cell 15, 578–589 (2008).
Hanson, P. I., Roth, R., Lin, Y. & Heuser, J. E. J. Cell Biol. 180, 389–402 (2008).
Lata, S. et al. Science 321, 1354–1357 (2008).
Saksena, S., Wahlman, J., Teis, D., Johnson, A. E. & Emr, S. D. Cell 136, 97–109 (2009).
About this article
The Journal of Cell Biology (2016)
Developmental Cell (2009)
Critical Reviews in Biochemistry and Molecular Biology (2009)