Pii: S0962-8924(99)01591-3

In my recent review of dynamin-related proteins, I had to admit that research on dynamin itself would continue to lead the way1. This proved true with the publication of three new articles on dynamin2–4. Two of the new articles support the role of dynamin as a mechanochemical enzyme that drives the release of clathrin-coated vesicles from the plasma membrane. This mechanochemical function was previously based on the observations that: • dominant–negative mutants blocked endocytosis before membrane scission occurs5; • dynamin formed spirals in vitro6 and at the necks of budding vesicles7, suggestive of a constriction mechanism; • purified dynamin could sever lipid tubules in vitro8. The new article by Takei et al. showed that amphiphysin, which is a major dynamin binding partner, enhanced the severing of membrane tubules by coassembly with purified dynamin4. Surprisingly, amphiphysin alone was also capable of forming membrane tubules, but those tubules were not severed unless dynamin was added too, which indicates that dynamin mediates scission in vitro, consistent with previous observations that were made with purified dynamin8. In this case, scission might result from constriction of the dynamin spiral, either by a crimping mechanism8 or by a twisting mechanism9. The results obtained by Stowell et al. also suggested that dynamin had a mechanochemical function, but the proposed mechanism of membrane scission was very different3. Stowell et al. observed that dynamin spirals were stretched to twice their original lengths by GTP hydrolysis. This led to their proposal that stretching of the dynamin spiral might be enough to pop vesicles off from the plasma membrane. The popping model requires a barrier at the plasma membrane that prevents lipid molecules from entering the membrane tubules because those tubules might otherwise yield to the stretching forces of dynamin. The nature of this barrier is not yet known. In a surprising turn of events, the other new article on dynamin, by Sever et al. provides evidence that dynamin might not be the driving force for pinching off clathrin coated vesicles2. Instead, dynamin might play a regulatory role, while other proteins would be needed to catalyse the physical separation of membranes. What led the authors to propose this role reversal? First, they discovered that a C-terminal domain, which they named the GTPase effector domain (GED; equivalent to the assembly domain), stimulates the GTPase activity of dynamin. This stimulation is cooperative, suggesting that the GED can function as a multimer. The authors mutated the arginine and lysine residues of the GED individually to see whether any of them might act like the catalytic arginine of Ras GTPaseactivating protein (RasGAP)10. They identified several residues within the GED in which mutations reduced the rate of GTP hydrolysis. This slowing of GTP hydrolysis might prolong the GTP-bound state of dynamin. The big surprise came when genes encoding dynamins with GED mutations were transfected into mammalian cells. Although the GED mutations slowed GTP hydrolysis in vitro, they caused an increase in the rate of transferrin uptake. This effect is opposite to that of mutations in the GTPase domain, which inhibit GTP binding and have long been known to decrease the rate of transferrin uptake. The discovery that GED mutations increase the rate of endocytosis argues against a mechanochemical role in scission because such a role would have predicted a positive relationship between the rates of endocytosis and GTP hydrolysis. Thus, the increase in endocytosis caused by slowing GTP hydrolysis suggests that the dynamin GTPase does not provide the mechanical force needed for membrane scission. The force for scission must therefore come from another source (an ATPase?), which might be recruited after the dynamin spiral is fully assembled. In this model, the dynamin spiral might be used to sense the diameter of the neck of a budding vesicle. When the appropriate diameter is attained, dynamin might trigger membrane scission by activating the putative severing molecules. If GTP hydrolysis is slowed down, then the dynamin collar would be kept longer in a recruiting mode, capable of recruiting larger numbers of scission enzymes. In this new model, dynamin hydrolyses GTP, and the dynamin spiral then falls apart, but only after scission takes place. Past studies provide arguments for both sides of the debate. For example, GTPgS induces long spirals of dynamin with no scission, suggesting that GTP hydrolysis is required for scission7, and dynamin spirals assembled in vitro can fragment lipid tubules by hydrolysing GTP8. However, the dephosphorylation of dynamin in neurons appears to activate dynamin by slowing the rate of GTP hydrolysis11, consistent with the model proposed by Sever et al.2 It is clear that more research needs to be done. The activating mutations generated in this study will play an important role in that future research. Other activating mutations analogous to those found in Ras and other small GTP-binding proteins have eluded the investigators studying dynamin so far because the sequence of the dynamin GTPase domain is too divergent. Also, no genetic screens have identified mutations that activate dynamin. The novel mutations discovered by the Schmid lab might be our best hope for analysing the complex interactions that occur during the dynamin cycle. Acknowledgement