Introduction
Proteins are carried between membranous organelles in the cell by membrane-bound carrier vesicles. One of the protein-transport events in which vesicles are involved is endocytosis — the process by which cells uptake material from outside the cell, by means of invaginating and forming intracellular vesicles from sections of the plasma membrane. Early studies indicated that a large GTP-hydrolysing protein, or GTPase, named dynamin may be involved in the process of budding of new vesicles from the plasma membrane 1, 2, but its precise function in vesicle formation has been difficult to establish. Now three groups, writing in this month"s issue of Nature Cell Biology 3, 4 and in a recent issue of Nature5, report results that at first appear at odds with each other but, on closer inspection, may allow a more realistic picture of dynamin"s function to emerge.
The formation of carrier vesicles and the selection of the correct cargo proteins require a mechanism for coating the vesicles to separate their membrane from that of their source 6. However, an early indication that such coating mechanisms are insufficient to achieve carrier-vesicle formation from the plasma membrane was the discovery that a mutation blocking endocytosis in the fruitfly Drosophila affected dynamin, which was not part of any known coating mechanism 1, 2.
Dynamin clearly does something very important in endocytosis, but what is it? Clues have come from studies of the purified protein. The dynamin structure (Fig. 1) includes an assembly domain, through which the protein can oligomerize into rings and spirals in low salt concentrations 5, 7. Dynamin also has a GTPase domain, which exhibits an intrinsic rate of GTP hydrolysis (2–100 GTP molecules hydrolysed per minute) that is fast for a regulatory GTPase, but which becomes two to three orders of magnitude faster on oligomerization of dynamin. This is because dynamin"s assembly domain also contains a GTPase-effector domain 4, 5.
Figure 1: Domains of dynamin.
A tripartite GTP-binding domain (pink stripes, left) is at the amino terminus. The pleckstrin-homology domain (PHD) binds phosphoinositides. The GTPase-effector domain (GED, green) is also an assembly domain and is responsible for the dramatic increase in GTP-hydrolysis rate that occurs when dynamin oligomerizes into rings and spirals 5. It contains two regions (blue stripes) that may form coiled-coil structures and so has also been called the "coiled-coil" domain. The final, proline-rich domain (PRD) binds proteins, such as amphiphysin-1 and -2, that have Src-homology-3 domains. Point mutations are now known in all four domains. R837E has no effect on endocytosis, but rescues the inhibition resulting from the K535A mutation, which inhibits PHD function 14. K535A reduces binding to liposomes containing phosphoinositides. K694A inhibits the polymerization of dynamin subunits, while K725A inhibits the activation of GTPase by polymerization. The arrows indicate whether endocytosis was inhibited or enhanced. Numbers at top are amino-acid numbers.
Full size image (15 KB)Oligomerization of dynamin is stimulated even at physiological ionic strengths by the presence of negatively charged spherical liposomes, to which dynamin binds 8, 9. After they bind dynamin, the liposomes take on a remarkable morphology, consisting of cylindrical tubes surrounded by a spiral of dynamin (Fig. 2a). Binding to dynamin and tubulation of the liposomes does not require energy, can occur in the cold, and can be driven by both the GTP- and the GDP-bound forms of dynamin. The structures seen in the presence of GDP and GTP are different, however, and hydrolysis of GTP to GDP while the dynamin is attached can lead to fragmentation of the tubules into small vesicles 3, 8, 9.
Figure 2: The three experimental systems used to study dynamin twists.
a, Dynamin tubulates spherical liposomes. GTPase activity generates narrower tubules and causes vesiculation to dynamin-coated vesicles 3, 8, 9. b, In the presence of dynamin, lipid nanotubes acquire a helical coat which changes pitch on GTP hydrolysis 4. No vesiculation occurs. c, Brain cytosol added to nerve-terminal membranes with ATP at 37 °C generates long tubules, wrapped in helices of dynamin and amphiphysin, that are capped with clathrin 3.
Full size image (51 KB)So the focus turns to what is happening when oligomerized dynamin on a lipid tubule hydrolyses its GTP. In this issue of Nature Cell Biology, Stowell and colleagues 4 describe the oligomerization of dynamin on preformed lipid nanotubes. As expected, oligomerization causes a massive increase in the rate of GTP hydrolysis. Surprisingly, though, Stowell et al. also found that the pitch (inter-ring spacing) of the helical dynamin changes dramatically on GTP hydrolysis. GTP-bound dynamin forms a tight helix (with a pitch of 11 nm) around the lipid tubules; this helix extends on hydrolysis of GTP to about double the spacing (20-nm pitch) (Fig. 2b). The authors liken dynamin"s behaviour to that of a spring, and suggest a clever "poppase" model for the function of dynamin in vesicle formation — they propose that the extension of the dynamin helix, which wraps around the neck of a budding vesicle, can "pop" the vesicle off the end of the neck. The nanotubes used in this experiment did not vesiculate.
These findings on dynamin are all first class and give wonderful insights
into the conformations that the purified protein can adopt, but what do they
tell us about how dynamin works in vivo? One important caution sounded
by Stowell et al.4 is that the ability to tubulate liposomes
is not unique to dynamin. It can be induced by the oligomerization of other
proteins, such as RNA polymerase or amphiphysin, a protein that, like dynamin,
is concentrated in nerve terminals 3. But credibility is lent
the liposome experiments by the observation that dynamin, in the presence
of the non-hydrolysable GTP analogue GTP-
S, can also wind itself around
natural membranes, even nerve-terminal plasma membrane, to generate tubules 10 (Fig. 2c). Breathtakingly elegant electron
micrographs show tips of dynamin-coated tubules capped with structures made
of clathrin, a coat protein that may be required for vesicle endocytosis.
Such images suggest a model in which clathrin coats nucleate a budding event,
after which dynamin wraps itself around the neck of budding vesicles to carry
out the final vesicle-fission event 11.
The clathrin-nucleation model lost much of its appeal when liposome tubulation was found to occur in vitro in the presence of dynamin but absence of clathrin 8, 9. The second part of the model, the function of dynamin in a fission event, is, at first glance, supported by the liposome-vesiculation data 8, 9. However, we have to be a little concerned that vesiculation by dynamin is not found in vivo in the absence of coating mechanisms. Furthermore, vesiculation by coating proteins in the absence of added dynamin has been reported, casting doubt on this simple model 12.
It is always tempting to try to explain an entire cellular phenomenon with a single protein, especially if we were the ones to purify it! The properties of dynamin, however, give us strong clues that more factors than just clathrin and dynamin are involved in endocytosis. One of these clues is dynamin"s pleckstrin-homology domain (Fig. 1), which is known to bind phosphatidylinositol phosphates and to be required for dynamin"s endocytotic function 13, 14. The in vitro liposome-binding and -vesiculation processes, though, do not require the presence of such lipids 8, 9.
A second hint is the presence of dynamin"s carboxy-terminal, proline-rich domain (PRD). This domain is also necessary for dynamin"s function in vivo 14, 15, but not for the liposome-tubulation reaction 8. Amphiphysin, one of the proteins containing a domain known as the Src-homology-3 (SH3) domain that bind to the PRD of dynamin, has been implicated in endocytosis by transfection and microinjection experiments 16, 17. Although dynamin"s PRD is not required for liposome tubulation, Takei et al. report in this issue 3 that the structure of the dynamin helix is significantly altered when the SH3 domain of amphiphysin binds dynamin. Complexes of dynamin and amphiphysin can form characteristic rings in the absence of liposomes even at physiological ionic strength 3, which dynamin alone cannot. Thus the formation of helices in lysed nerve terminals probably results from recruitment of amphiphysin–dynamin complexes, not of dynamin alone as thought at first 10.
We need to wait for further studies to be performed before we know how dynamin"s function is modulated by its PH and PRD domains in vivo. In the meantime, we can ask whether the three remarkable properties of purified dynamin — its ability to polymerize into rings and helices, its oligomerization-sensitive GTPase activity, and its conformational change on GTP hydrolysis — cause membrane fission directly.
One canonical approach in cell biology is to knock out the function of individual domains in vitro and study the effect on cell behaviour. In an astonishing set of experiments, reported in Nature, Sever et al.5 have now done exactly that, introducing one mutation into dynamin (K694A) that inhibits its ability to self-assemble and another (K725A) that prevents self-assembly from enhancing GTP hydrolysis. If the GTPase-stimulated conformational change in dynamin is required for membrane fission, as has been proposed, then these mutations should inhibit endocytosis. Instead, overexpression of the mutant dynamins in fibroblasts enhances the rates and extent of endocytosis 5. Oligomerization and the resulting increase in GTPase activity appear, from this study, to be inhibitory to endocytosis, a simple conclusion but one difficult to reconcile with the suggestion that dynamin uses its GTPase-induced conformational change to stretch or break the necks of clathrin-coated vesicles.
To explain their data, Sever et al.5 suggest that dynamin is a regulator rather than a "pinchase" (earlier data indicated that dynamin might function by "pinching" the vesicle neck) and that the role of dynamin oligomerization is to ensure that a neck of the correct diameter forms before fission can begin. The enhanced GTPase activity seen after dynamin forms a ring or spiral must somehow act to inhibit endocytosis, as the mutations in the GTPase-effector domain enhance endocytosis. The conformational changes in dynamin might act to prevent access of the fission machinery until, for example, the correct cargo is assembled.
While we cell biologists amuse ourselves trying to create coherent models from the abundance of new data, we might also re-evaluate some old beliefs. Many of our previous ideas are based on studies of the collared pits that appear in Drosophila nerve terminals when dynamin function is blocked 18. Are the collared pit structures really made up of dynamin, as is so often assumed? Might dynamin be involved in invagination as well as vesiculation, as suggested by studies of Drosophila mutants 19? Dynamin has gracefully removed one or two of her veils, but several more must fall before all is revealed, and what we see might be very surprising.
