Dancing filopodia

How do cells do mechanical work to move, and divide? The filaments that make up the cytoskeleton, which provide the structure required for large-scale movement, are some of the most extraordinary assemblies in biology.  These filaments are dynamic (they form where they are needed, then dissolve again), and they’re amazingly long and thin: tens of microns long, but only tens of nanometers thick.  The pieces they are made of are only a couple of nanometers across.  Build a Lego tower four bricks wide and 10,000 bricks high; that’s a model of a filament.  Now quickly disassemble it all the way down to its component parts, and build it somewhere else.  The cell can’t afford to have permanent infrastructure: sometimes filaments are used as train tracks for motor proteins that transport important cargo to remote locations in the cell, sometimes they’re used to push the cell membrane forward so that the cell can move.  And the Lego tower assembles by itself, as if by magic, wherever it’s needed.  How?

This problem of collective protein behavior is one of the great unsolved problems in cell biology.  This is not a problem that’s likely to be solved by omics, but it seems to me that the less-fashionable field of biochemistry, combined with modern imaging, offers hope.  [Of course, my boss is an unrepentant biochemist, so take my opinions with a grain of salt.]  The Kirschner lab has just taken a very impressive stab at this problem by showing that important filament-driven structures called filopodia can be made to form in vitro (Lee, K., Gallop, J., Rambani, K., and Kirschner, MW. 2010. Self-assembly of filopodia-like structures on supported lipid bilayers.  Science 329 1341-1345 PMID: 20829485) and imaging their wild, dancing behavior.

Filopodia are little protrusions that are sent out from a migrating cell in the direction that it wants to move; they stick on the substrate, and the rest of the cell then follows them.  The filaments that give a filopodium structure are made out of actin, one of the three main cytoskeletal proteins.  The big question is, how does the actin know where to assemble and push out the filopodia?  Actin is everywhere in the cell, and yet it self-assembles only in very specific locations.  (A particularly nice way of looking at actin assembly is to study Listeria, which use actin tails as “jet propulsion” to rocket around the cell and punch through from one cell into another.  A great example of a pathogen subverting one of the cell’s basic mechanisms.) Previous work on the control of actin assembly, from the Kirschner and Mitchison labs and many others, has assembled a list of proteins and factors that can reconstitute the assembly of an actin tail on a well-defined synthetic lipid vesicle that contains the triggering lipid, PIP(2).  But, when this same set of proteins is added to a flat lipid bilayer — a better representation of the cell membrane, and also containing PIP(2) — you don’t get actin tails; instead a thin, flat layer of actin forms on the lipid bilayer and just lies there.  Something is missing.

When Marc Kirschner discovers something is missing in his biochemical reconstitutions, he has what amounts to a Pavlovian response: he turns back to the Xenopus egg extract system.  Xenopus egg extracts are essentially undiluted cytoplasm from Xenopus eggs; packed eggs are centrifuged at a rate that causes the egg to break and the nuclei to separate from the cytoplasm, then the cytoplasm is carefully sucked out of the tube.  These extracts can be nudged into performing several types of complex behaviors, including going through the whole cell cycle.  And if the extract can perform a behavior, fractionating the extract is a great way to identify the proteins involved.

The video of dancing filopodia above — more accurately, “filopodia-like structures”, or FLSs — shows the astonishing behavior you get when you put Xenopus extracts on a flat lipid bilayer containing the activating lipid PIP(2).  These highly dynamic structures are the right shape for filopodia, and their internal structure also looks right by electron microscopy.  There is, however, one obvious difference between these FLSs and real filopodia: real filopodia push the cell membrane outwards, while the FLSs are assembling on a membrane that’s supported on glass.  So instead of pushing the membrane out, they are themselves pushed backward by the membrane; what looks like the head of the filopodium in the movie is actually the tail.  The growth in the filament is happening where the filament touches the lipid membrane, not at the end that is waving in the breeze.

Lee, Gallop et al. ask, what is special about the spot at which a FLS forms?  It’s not a dramatic difference in the composition of the membrane: they look for areas that are enriched in the triggering lipid, PIP(2), using a protein that binds to PIP(2) and is labeled with GFP, and can’t find any.  It looks as if the formation of an FLS is triggered by self-assembly of proteins on a membrane that is permissive for self-assembly, but not driving it.

What assembles first?  The authors added fluorescently-labeled versions of all the candidate proteins they could think of to the Xenopus extracts, and used TIRF microscopy [which allowed them to monitor a very thin area, right next to the lipid membrane] to watch which proteins appeared first at spots that later formed FLSs.  Three proteins arrive before actin: the membrane-binding protein toca-1, and N-WASP and Arp2/3, which together activate actin polymerization.  There is a great deal of material here for an actin afficionado to chew on; I’ll skip lightly over the details to get to the punchline, which is a new model of how filopodia are initiated.  The key points are that the spot where the FLS is going to form accumulates actin-nucleating proteins very early; and that in the initial formation of the FLS, the bundling protein fascin is recruited last.

Until this work, there had been two main models of filopodium initiation.  In the first, actin filaments are formed randomly and then collected into filopodium-like structures by a bundling protein. In the second, a special set of proteins forms the tip and triggers formation of long filaments, which are bundled together later. Lee et al., being unusually diplomatic, suggest that both models have some correct elements.  Their new model is that signaling in a negatively charged membrane (for which PIP(2) is best, but other lipids will do) triggers the recruitment of nucleating proteins (such as toca-1) into a cluster that initiates the formation of short actin filaments.  These short filaments then get pulled together into a parallel bundle by bundling proteins, and as they grow the membrane is pushed outwards. In this new “clustering-outgrowth” model, several different proteins might have clustering or elongation activity, so the composition of filopodia might vary depending on circumstances.  Since Science relegated the figure showing the model to the supplementary material, I’m reproducing it here:

Now that we know all this, will it be possible to reconstitute the whole process using pure proteins?  Can we understand how the kinetic behavior of the individual proteins produces the dynamics and organization of these structures?  You’ll just have to wait for the next installment… which may well come from Jenny Gallop’s soon-to-be established lab at the Gurdon Institute, Cambridge UK.  Watch this space.

Lee K, Gallop JL, Rambani K, & Kirschner MW (2010). Self-assembly of filopodia-like structures on supported lipid bilayers. Science 329 (5997), 1341-5 PMID: 20829485