The Molecular Mechanism of Intracellular Membrane Fusion Richard H. Scheller
The human brain contains approximately 10 15 connections between nerve cells (Figure 2). The specific formation and modulation of these connections, which are called synapses, underlies our ability to perceive sensations and to think. Modulation of the activity of synapses has been shown, by Eric Kandel and others, to underlie the formation of memories. Clearly understanding the cellular and molecular basis of how synapses work is of fundamental importance. Here I show two electron micrographs of typical synapses, one between a nerve terminal and a muscle fiber and one from the central nervous system (Figure 3). The presynaptic nerve terminal contains a number of small membrane vesicles that are filled with neurotransmitters, in this case acetylcholine and glutamate, chemical messengers of the neuromuscular junction and brain respectively. Classic studies of the neuromuscular junction revealed the vesicle cycle of neurotransmitter release. As an action potential reaches the nerve terminal channels open allowing an influx of ions including Ca ++. The Ca ++ ions trigger fusion of the vesicle membrane with the plasma membrane and release of neurotransmitter, which is then sensed by receptors on the post-synaptic membrane. Following the release process membrane is then recycled and new vesicles form (Figure 4). In the mid 1980 s I became interested in understanding the molecular mechanism of neurotransmitter release. How might one approach this problem? I trivially hypothesized that the proteins associated with synaptic vesicles would be involved in the process. It is possible to purify vesicles by a series of centrifugation steps followed by column chromatography that fractionates membrane organelles according to their size. This final stage of vesicle purification is shown in panel B of figure 5. Since the vesicles contain ATP they are easily identified. The resulting membranes are shown in panel C of figure 5, a large heterogeneous collection of membranes and the homogeneous 60 nm synaptic vesicles. We screened a cdna expression library made from the neurons of the electric lobe of the pacific electric ray Torpedo californica with an antibody raised against synaptic vesicles that was made by Reg Kelly from UCSF, and isolated a number of clones. We then made antibodies against the proteins encoded by these clones and one of them, encoding a protein we named VAMP1, clearly copurified with the synaptic vesicles. A Western blot, across the column, showing the VAMP1 immunoreactivity copurifies with the vesicle peak is shown in panel B, figure 5. VAMP1 is a relatively small protein with a carboxy terminal membrane anchor and about 100 amino acids oriented to the cytoplasm. Will have more to say about this protein later (Figure 6). The next sets of experiments were aimed at identifying proteins from the synaptic vesicle that interact with molecules on the plasma membrane. We used an antibody raised against p65 or synaptotagmin and immunoprecipated from detergent solubalized membranes. We identified a co-immunoprecipitating protein that we named syntaxin from the Greek word syntax meaning to put together in an ordered way (Figure 7). Immunohistochemical studies, first at the resolution of the light microscope and later using immunoelectoron microscopy demonstrated that syntaxin is largely localized to the plasma membrane. You can see the vesicles, stained here in red and either a green ring, or a crescent staining of presynaptic membrane active zone (Figure 8).
This gave rise to an initial model of the vesicle docking and fusion complex where VAMP1 and synaptotagmin on the vesicle membrane interact with syntaxin on the plasma membrane. Close proximity to the Ca ++ channel would then trigger membrane fusion and transmitter release upon Ca ++ influx. Interestingly none of these proteins were known in yeast at the time because synaptotagmin is not found in yeast and VAMP and syntaxin are duplicated genes so they did not show up in mutational analysis looking for genes involved in membrane trafficking. We also predicted at the time that these proteins would be the scaffolds upon which the soluble factors, alpha-snap and NSF, would assemble (Figure 9). Clearly we were headed in the correct direction as in 1993 Rothman s group demonstrated that, in addition to alpha-snap, 3 proteins associated with the ATPase NSF when locked in the ATP state. When ATP was hydrolyzed these proteins were released from NSF. The 3 proteins were VAMP, syntaxin, and a new player, a previously characterized protein that at the time had no known function called SNAP -25. These proteins became known as SNAREs (Figure 10). From these experiments it was not clear how the NSF complex assembled. Did the SNAREs independently bind to alpha-snap or NSF. Did they bind each other, and so on. In independent experiments Tom Südhof s lab as well as our group demonstrated that a homologue of the yeast protein sec1 bound with high affinity to syntaxin. Syntaxin could not form complexes with other SNARE proteins until sec1 was released from the high affinity state of syntaxin. This was clearly shown in our crystal structure of snytaxin bound to sec1. The next major advance came in a collaboration between the laboratories of Jim Rothman and myself. Here we studied the proteins that associate with syntaxin. We demonstrated that an ATP independent complex forms between VAMP, syntaxin, SNAP-5 and synaptotagmin, the 7S complex. As alpha-snap adds to the complex, again in an ATP independent fashion, synaptotagmin is displaced. NSF binds to the complex in the ATP state and upon hydrolysis the SNARE complex is dissociated. At the time, 1993, the order of these events in the actual fusion reaction was not at all clear so it was hypothesized that the ATP hydrolysis overcame the energy barrier and drove the fusion to a state that would rapidly release transmitter upon Ca ++ influx, perhaps a hemifused sate as drawn in figure 11. The next set of experiments began to better understand structural aspects of the SNARE complex. I believe Hugh Pellham was the first to ask the question about the orientation of the vesicle VAMP and target syntaxin SNARE proteins with respect to each other. If the alignment was anti-parallel, one might propose a role in vesicle docking whereas if the alignment was parallel one might suggest that the formation of the complex would actually bring the membranes together and mediate the fusion reaction itself. Our approach was to fluorescently label the amino and carboxy terminal ends of the interacting domains of VAMP and syntaxin and to conduct fluorescent (Forster) resonance energy transfer (FRET) studies. These investigations along with the higher resolution 3-D chrystallographic studies of Brunger and Jahn established the parallel alignment of the coil domains suggesting that complex formation may drive fusion (Figure 12).
Thus the mechanism of membrane fusion was proposed to be as follows. Initially sec1 is bound to syntaxin and sec1 is released. An initial paring of the SNARE proteins, which I called nucleation, followed by a zippering which would lead to the full formation of the SNARE complex which itself would drive the fusion event, possibly first via a hemifused state. The nucleation and zippering terminology arose from my previous experience with nucleic acid hybridization where long polymers of DNA first associate over a short region then zipper as a long DNA duplex structure is formed. The stability of the SNARE complex is truly remarkable for a noncovalent complex in that the proteins do not dissociate until heated to 95 deg C. Thus the idea of the free energy of complex formation driving the membrane fusion became established. The role then of alpha-snap and ATP hydrolysis by NSF is to dissociate this very stable complex so the SNARE proteins can participate in another round of fusion (Figure 13). How then might one study this hypothesis? We chose to use a cracked PC12 cell system invented by Tom Martin. In this system cells are loaded with 3 H norepinephrine and then squeezed through a small space which produces a single large tear in the plasma membrane. After washing, one is then left with a cell ghost comprised of docked vesicles at the plasma membrane (Figure 14). A view from the cytoplasm of these cells looking at the plasma membrane shows the docked vesicles. One can also observe clathrin coated pits, sites of active membrane recycling (Figure 15). Then in a two step process release is stimulated. A first step is a priming reaction with MgATP and cytosol and a second step is actually triggering the release with Ca ++ (Figure 16). But how does one study the SNARE proteins in the presence of the endogeneous molecules? This is accomplished by inactivating one of the SNARE proteins by proteolytically cleaving the molecule with one of the toxins produced by Clostridium botulinum. These are some of the most potent toxins known to man. This particular toxin cleaves the C-terminal coil region of SNAP -25 releasing a 26 amino acid peptide. Now you know the molecular mechanism of the drug used to eliminate wrinkles that is so popular with some people: Botox. We were then able to rescue secretion by adding back the full coil to the cracked PC12 cells and then triggering release (Figure 17). The panel on the left shows the inactivation of release is dose dependent on the amount of neurotoxin. After washing away the toxin we are able to rescue the release by adding back the appropriate SNARE coil. Other SNARE coils do not rescue. In a series of experiments we demonstrated that in this system SNARE complex does not form during the priming reaction, only after the addition of Ca ++. Release in this system is of course much slower than in neurons (Figure 18). The crystal structure of the SNARE coil domain demonstrated the presence of 16 so called layers. 15 of these layers bury hydrophobic amino acid side chains in the center of the 4 helical bundle as shown on top with 2 valine and 2 isoluceine residues. Interestingly a central layer is comprised of polar side chains, an arginine and three glutamines which hydrogen bond to each other (Figure 19). We mutated hydrophobic residues to alanine, and the polar layer as well, and studied both the effect on the stability of the complex and the ability to support secretion.
Along the bottom of the figure is the thermal stability of the complex in deg C. The ability to support secretion correlated with the stability of the helical bundle complex, suggesting that it is in fact the free energy change in complex formation that is used to overcome the barrier to membrane fusion. Even a mutation of surface residues that increased the complex stability increased the rate of fusion. Mutation of the hydrophilic layer had little effect on the fusion (Figure 20). What then is the role of this highly conserved central ionic layer? We measured, in vitro, the rate of complex dissociation in over 70 mutants. Only one amino acid substitution effected this rate and that was mutation of the glutamine of the ionic layer of syntaxin (Figure 21). As shown in the previous slide these mutations had only a minor effect on the stability of the complex. Consistent with this observation, when we transfected a syntaxin mutant of this amino acid into cells we recovered a larger amount of complex than in cells transfected with the wild-type syntaxin. In other words, we had shifted the ratio of free SNARE proteins to complex due to the impaired ability of alpha-snap and NSF to dissociate the complex. From this we concluded that the ionic layer is important for the dissociation of the complex by alpha-snap and NSF perhaps being the site that initiates melting of the helical bundle upon ATP hydrolysis by NSF. Now I need to introduce a bit of nomenclature. The SNARE 4 helical bundle is comprised of 4 coil domains and to form a functional helical bundle one coil of each type is required. These coil domains became known as Qa for the syntaxin coil, Qb and Qc for the two SNAP-25 coils and R for the VAMP coil. The individual SNAREs are known as Qa, Qb, Qc and R SNARES. With the completion of the human genome sequence we were able to determine the number of different SNAREs in the genomes of various species. There are a total of 35 in humans and 21 in yeast with a distribution of the specific coils as shown in the table (Figure 22). Why are there so many of these proteins? Do they have functions outside of secretion? We characterized most of the SNARE proteins in mammals and showed that individual SNAREs are strictly localized to different compartments within the cell (Figure 23). For example, in collaboration with Judith Klumperman we demonstrated that rbet1 is localized to the vesicular tubular compartment between the ER and the Golgi, syntaxin 5 to the cisternae of the Golgi and VAMP4 to the trans Golgi network. From these and other studies it became clear that sets of SNARE proteins mediate intracellular membrane trafficking in all cells. That is for example transport from the ER to the Golgi, endocytosis, transport to lysosomes and so on (Figure 24). Using our cracked PC12 system we studied the specificity of SNARE complex formation using different Qc SNARE coils. Indeed only plasma membrane localized coil domains were able to rescue secretion in the PC12 system described earlier, not Qc SNAREs from the TGN or endosomes (Figure 25). The specific paring along with the distinct localization of these proteins within the cell determines, at least in part, which cellular membranes can fuse with each other. Thus the SNAREs comprise a component of the machinery used to organize the membrane compartments of cells. Finally, as the work progressed it became clear that not only are the proteins discussed in this talk used by vertebrates but that all eukaryotic organisms use this
evolutionarily ancient mechanism to fuse membranes. The yeast SNARE organization is shown on top and the mammalian SNARE organization is shown on the bottom of figure 26. Thus, as happens so often in biology, one organ, in this case the brain, has adapted a general mechanism for its special function, synaptic transmission. Many thanks are due to the folks who worked on these studies with me over the years. I would also like to thank the selection committee as well the Norwegian Academy of Sciences for this great honor (Figure 27).