Bacterial dynamin: cellular function and molecular mechanism
Lead Research Organisation:
MRC Centre Cambridge
Department Name: LMB Structural Studies
Abstract
We have previously identified a molecule in a number of common bacteria that was thought to only exist in the cells of higher organisms such as yeasts, plants and animals, as well as humans. The molecule is involved in the shaping of the lipid membranes that engulf all living cells and that also form internal compartments inside some cells. Our finding re-emphasises the evolutionary link between all living organisms and opens up questions regarding the origins of processes that are very important to our brains, such as neurotransmitter uptake. It seems, that parts of these processes have already been invented by bacteria earlier. In the current proposal, we aim to eclucidate the exact molecular mechanism of thse membrane-shaping molecules using the bacterial versions. This is required since the human versions have proved to difficult to work with over recent years. An exact understanding will eventually make it possible to understand why, under certain circumstances, things go wrong and diseases develop. Apart from this, knowledge of what bacteria do with these molecules without having a brain or indeed intracellular organelles will be interesting.
Technical Summary
We will attempt to produce a high-resolution map of bacterial dynamin-like protein (BDLP) using cryo-electron microscopy and helical reconstruction techniques to be able to fit in our previous crystal structure. From preliminary work we anticipate a large conformational change to become obvious, stretching the molecule to over 16 nm. This will allow the hydrophobic 'paddle' to insert into one half of the lipid bi-layer, hence introducing membrane curvature. Experiments with engineered, isolated 'paddle' domains will confirm this. The fitting of the crystal structure will then allow us to interpret mutants and any nucleotide-induced conformational change at the atomic level, a first for any dynamin molecule and something that has been a major goal in the eukaryotic dynamin field for considerable time. Site-specific gold-labelling will hopefully enable us to produce models with absolute certainty because of the largely helical nature of the BDLP molecule. In addition to the mechanistic work, we also aim to obtain information about the biological role of BDLP in bacteria such as E. coli and B. subtilis. It is early days and we propose to use localisation techniques (GFP, IFM, gold-labelled sections, fractionation etc.) to gain first insights. The further route is unclear, depending on the outcome of these experiments but we will complement these with mutants of the GTPase function and KOs.
Organisations
Publications
Low HH
(2010)
Dynamin architecture--from monomer to polymer.
in Current opinion in structural biology
| Description | Final Report BB/F006284/1 - Bacterial dynamin: cellular function and molecular mechanism, 2008-2011 Jan Löwe, MRC Laboratory of Molecular Biology, Cambridge CB2 0QH, UK Introduction Dynamins belong to the family of large GTPases and are involved in a plethora of functions, unified by the ability of dynamins to act mechanically on membranes. Best known is probably vertebrate dynamin 1, which is involved in membrane constriction during endocytosis. Other dynamin functions in animal cells involve cell separation at the mid body, cell shape maintenance, centrosome cohesion but also more enigmatic roles such as restriction of envelopes viruses (Praefcke and McMahon, 2004). Until our discovery of bacterial homologues of dynamins in bacteria, it was thought that dynamins are eukaryotic signature proteins. The discovery had to be supported by a crystal structure of bacterial dynamin from Nostoc punctiforme. since sequence similarity is extremely low, especially in the helical parts of the molecule (Low and Löwe, 2006). The discovery of dynamin in bacteria was also supported by biochemical data, showing that BDLP1 (bacterial dynamin like protein) from Nostoc efficiently coats and tubulates liposomes in vitro, one of the hallmarks of all mechanical dynamin-like proteins (Low and Löwe, 2006). Results The grant proposed to investigate by cryo-electron microscopy (cryoEM) exactly how bacterial dynamin binds to membranes and how the curvature is generated. We knew this was possible because preliminary data showed amazing structures formed when membranes, GTP and BDLP1 were mixed: Figure 1. A) Crystal structure of the dynamin-like protein (BDLP) from Nostoc punctiforme. B) BDLP tubulates liposomes made from E. coli membrane lipids. The goal was to use cryoEM in order to resolve these large objects so that the crystal structure could be fitted. Earlier attempts with eukaryotic dynamins had been limited in scope and did not resolve the entire structure because no structural information was available except for dynamin's GTPase domains (Mears et al., 2007). First we needed to figure out if the BDLP1-constricted lipid tubes had helical symmetry; which they did, and we managed to index these despite a very small pitch of the helix making the indexing non trivial: Figure 2. Top: cryoEM image and indexed diffraction pattern of BDLP1-constricted liposomes. Bottom: low resolution reconstruction obtained by classical Bessel reconstruction showing the protein lattice on the outside. In order to obtain high resolution reconstructions we employed the help of Carsten Sachse (group of Roger Williams, MRC Laboratory of Molecular Biology), who wrote specialised software for iterative helical reconstruction that allows for small deviations of helix segments from the overall symmetry - essentially allowing for imperfections. Optimising the data collection strategy on our FEI G2 Polara at 300 keV, on film, optimisation of helical parameters, segment sizes and many other details are faithfully reported in the main publication resulting from this work in the Journal Cell (Low et al., 2009) (Figure S2). The resulting maps at around 1 nm (10 Å) resolution (FSC based) were of excellent quality and, for the first time, allowed the accurate description of a dynamin helix constricting a lipid tube at molecular resolutions. Figure 3. A) BDLP1:lipid density map at ~1 nm resolution, showing the outside and the helical path. B) Showing the inner membrane tube constricted by the protein around it. C) View along the lipid tube (red, middle), showing one asymmetric unit of the helix boxed in yellow. D) The asymmetric unit of the helix of BDLP1 consists of a dimer of two BDLP1 monomers. The globular GTPase domains can immediately be recognised at the top, forming the outermost protein layer and the dimer is formed by canocical GTPase domain dimerisation. One of the great surprises was the diameter of the membrane tube buried on the inside of the structure. It has a diameter of around 10 nm, which is very small for any membrane, meaning that the membrane really will have to be under immense tension, held in place only by the proteins around it. In order to make the membrane more visible, quantitative plots of projections were used, clearly showing the inner leaflet of the membrane (dark ring). The outer leaflet was less visible, probably because this is how BDLP1 bends membrane, by inserting the hydrophobic tip of the molecule into the outer leaflet, therefore bending it: Figure 4. A) Projection along lipid tube axis, shown in grey scale. The inner leaflet of the membrane is clearly visible, the outer leaflet (which would be around 5 nm away from the inner) less so. B) Close-up of A. We were then in a position to fit the BDLP crystal structure. This was done manually and carefully, resulting in a complete atomic description of the filament: Figure 5. A) BDLP1:lipid cryoEM density map with fitted atomic model of BDLP1 dimer. The dimer is formed primarily through the canonical GTPase domain dimer at the top (stereo view). Note that the helices sticking out on the top are discussed extensively in the publication (Low et al., 2009) and in reality most likely follow round the GTPase domain. They were left here as they were so we did only have to bend one bond in each of the two hinge regions (Figure 5). B) Stereo view of the GTPase domains fitted into the cryoEM density map. The biggest and most important surprise was that the crystal structure did not fit as crystallised. In fact the molecule had to be opened up, in order to fit the density. The GTPase domain had to hinge round, away from the neck, and the neck had to be bent away from the trunk. Note that the hinge regions contain conserved residues in all dynamin-like proteins and that in order to obtain this conformation, only two bonds, one in each of the hinge regions, had to be changed, making this arrangement highly plausible: Figure 5. Conformational changes required to fit the BDLP1 crystal structure into the cryoEM density map, obtained from constricted BDLP1-coated lipid tubes. How does BDLP1 bend the membrane? It inserts hydrophobic 'paddle' residues directly into the membrane. This is why the outer leaflet is less prominent and this is why the molecule needs to be so large in order to relate the large size of the GTPase domains to a very small insertion domain through large radial spokes (trunk and neck, both helical and thin). Force generation is therefore geared, with large forces at the membrane becoming smaller ones with more travel at the outside where the GTPase domains are located. This is all summarised in a cartoon, now including a model for the membrane as well, in order to visualise everything to scale: Figure 6. Model of the complete DBSLP1:lipid helical array. Looking more closely at the maps and models and at other dynamis, their partial structures and sequences, many more details could be discovered that we then assembled into a unified view of protein architecture for the whole dynamin protein family and this was published in a subsequent review article: Figure 7. Unified view of dynamin-like protein architecture. BDLP1 has the same basic architecture as other dynamins Taken from (Low and Löwe, 2010). Figure 8. Complete model of the BDLP1:lipid tube as reconstructed and interpreted with atomic models both of the protein and the membrane. Pink: tube surface where the GTPase domains are located and dimerise. Blue: BDLP1 atomic model that needed opening in order to fit the tube density. Orange and yellow: outer and inner leaflet of the lipid membrane tube, respectively. Finally, what does all this tell us? Firstly, BDLP1 really constricts membrane, just as eukaryotic dynamin, making it a bona fide mechanical enzyme. Secondly, it suggests that membrane curvature is facilitated by the insertion of a hydrophobic paddle into the outer leaflet of the membrane, like a wedge. Interestingly, Dynamin 1 has a PH domain in its place, recognising specific lipid head groups (see Figure 7, right). Thirdly, the membrane seems under a lot of tension. This suggests that GTPase turnover might regulate disassembly, leading to a unified view of membrane fusion and fission by these proteins: Figure 9. A 'passive', polymerisation/de-polymerisation model for fusion and fission. Schematic drawing showing the different stages of BDLP/dynamin induced fission and fusion. Polymerisation is induced by GTP binding and induces high curvature. Hydrolysis to GDP causes catastrophic disassembly and produces a transition state that can either go back (grey arrow) or resolve through the rearrangement of the membrane linkage (blue connections). If the two membranes belong to the same surface, this results in fission. If they belong to two different surfaces (two vesicles, for example), the process results in fusion. Publications resulting from the grant and acknowledging support Low, H. H., Sachse, C., Amos, L. A., and Löwe, J. (2009). Structure of a bacterial dynamin-like protein lipid tube provides a mechanism for assembly and membrane curving. Cell 139, 1342-1352. Low, H. H., and Löwe, J. (2010). Dynamin architecture--from monomer to polymer. Curr. Opin. Struct. Biol. 20, 791-798. Further references Low, H. H., and Löwe, J. (2006). A bacterial dynamin-like protein. Nature 444, 766-769. Mears, J. A., Ray, P., and Hinshaw, J. E. (2007). A corkscrew model for dynamin constriction. Structure 15, 1190-1202. Praefcke, G. J., and McMahon, H. T. (2004). The dynamin superfamily: universal membrane tubulation and fission molecules? Nat Rev Mol Cell Biol 5, 133-147. |
| Exploitation Route | See above. |
| Sectors | Other |
| Description | Published in scientific journal |