Last time I promised you something inside your own cells that holds a fixed length while every molecule in it is marching. Here it is. The surprise is not the marching — that has been understood for fifty years. The surprise is what sets the pace.
An actin filament is a helical polymer about seven nanometres across, assembled from one of the most abundant proteins you own. In a crawling cell they crowd the leading edge in their thousands, a few millionths of a metre long, and the useful ones do not change length at all. Subunits lock onto one end and drop off the other at matching rates, so the filament stands still while its substance travels its whole length and departs. The name for this, coined in 1976, is treadmilling.
Two ends that disagree about how crowded the room is
Every polymer has a concentration of loose subunits at which joining and leaving exactly balance — below it the polymer dissolves, above it it grows. Actin has two such numbers, because its two ends are not chemically alike. They were measured in 1986 by growing filaments off a nucleating seed and counting subunits under an electron microscope: roughly 0.12 micromolar at the fast end, roughly 0.60 at the slow one.
Now hold the free-subunit concentration between those two figures. The fast end finds the room too crowded, and grows. The slow end finds the same room too empty, and shrinks. Same solution, same protein, opposite verdicts at opposite ends of one object, indefinitely.
The clock is inside each brick
What makes the ends disagree is a timer that every subunit carries in.
Actin arrives holding a molecule of ATP. Soon after it joins, the filament splits it — and here is the part that matters. The splitting is fast. But the severed phosphate then stays trapped in the subunit and leaves very slowly indeed, at something like five thousandths per second: a couple of minutes’ half-life. Nothing else in the cycle is remotely that leisurely.
So a subunit’s chemical state is a straight readout of how long it has been aboard — freshly loaded at the growing end, entirely spent by the time the far end works back to it. The filament is not merely a structure. It is a queue with a timestamp on every element, and its polarity, the thing that lets a cell tell forwards from backwards at all, is that gradient of age made physical.
This is No. 1’s candle flame with a direction added. A flame’s matter also streams through a persistent shape and leaves, but the leaving does no work and points nowhere. Here the throughput has a preferred end, and shoving against that end is how a cell crosses a dish.
And it is nowhere near fast enough
Now the trouble.
Purified actin, left alone in a tube, does everything described above — at a crawl. Filaments take tens of minutes to age and come apart. The leading edge of a moving cell rebuilds itself in seconds. The gap is about two orders of magnitude: the difference between a mechanism and an explanation.
The instinct is to hunt for something pushing harder at the growing end. That is the wrong end. Because the two rates have to match, the whole cycle is rate-limited by the slower one — a filament cannot take on more at the front than it sheds at the back. A crawling cell gets faster by getting better at destroying its own rear.
The demolition crew can only touch old work
That job falls to a protein called cofilin, and the elegant part is how it finds its targets. Cofilin binds by preference to subunits that have already let go of their phosphate — which is to say, the old ones. Working alongside a partner protein, it can accelerate loss from the slow end by as much as three hundredfold. The cell’s throttle is not on assembly. It is on forgetting.
Better still, cofilin also cuts filaments, manufacturing fresh ends — but recent single-filament work finds it will only do this to aged filament. Presented with newly assembled subunits, at every concentration the experimenters could reach, it stripped the end and never severed the shaft. The machinery that dismantles the cytoskeleton is chemically incapable of attacking what was just built.
That is the distinction owed to No. 34’s clot, which likewise builds and dissolves at once: a clot runs two opposing systems over the same mesh, and the outcome is a race between them. Here there is one gradient, and the demolition cannot reach the young end. Nor is it No. 39’s spider, which eats the whole web and spins another. Nothing here is ever taken down as a unit. The structure never once stops existing.
Nobody has actually watched it glide
One last honesty. When individual filaments were first watched at steady state in 2002 — rather than inferred from the average behaviour of a whole tube — they did treadmill at roughly the expected rate. They also jittered, lengthening and shortening by far more than one-subunit-at-a-time arithmetic permits. The tidy conveyor belt is a population statistic. Any particular filament is stumbling.
Which may be the truest thing about it. The cell is not running a machine. It is running a tendency — hard enough, and in enough copies at once, that a tendency becomes a direction, and a direction becomes a cell arriving somewhere.
≈
One loop I’m watching
Next: a sealed metal tube with no pump, no fan and no moving part anyone installed, which carries heat along its length hundreds of times better than solid copper. A liquid boils at the hot end, crosses as vapour, condenses at the cold end, and is walked back by nothing but the shape of the wall — a closed loop of matter moving something that is not matter, paid for entirely by the difference it is busy erasing. Next time.