Cytoskeletal Polymers and Dynamics

Watch one microtubule grow from a seed in a dish of purified tubulin and it will not settle down. It lengthens slowly for minutes, then shrinks many times faster, and sometimes starts growing again.

A crystal in its solution does not behave like this. This lesson asks what the cell’s three filament systems are made of, what decides whether an end grows or shrinks, and why two of the three spend nucleotide energy in a way that keeps them unstable.

Three polymers built for different jobs

The cytoskeleton is the set of protein filaments in the cytosol, the fluid part of the cytoplasm outside organelles. It has three classes.

Actin filaments are about 7 nm wide. Each subunit is a globular actin molecule that carries ATP. Subunits add head to tail, so the filament has two different ends, called the barbed end and the pointed end after the arrowhead pattern seen when myosin fragments decorate a filament in the electron microscope.

Microtubules are hollow tubes about 25 nm wide, built from α/β-tubulin heterodimers that bind GTP. Dimers stack head to tail into protofilaments, and usually 13 protofilaments lie side by side to close the tube. Thirteen is usual, not universal. Every dimer points the same way, so the tube has a plus end, where β-tubulin is exposed, and a minus end.

Intermediate filaments are about 10 nm wide and are built from a family of related fibrous proteins, such as vimentin in fibroblasts and keratins in epithelial cells. Two coiled-coil dimers line up antiparallel and half staggered to form a tetramer. A building block that points both ways gives a filament with no polarity. The proteins also bind no nucleotide and have no enzyme activity of their own.

In comparisons of purified filaments, microtubules are the stiffest in bending, which suits them to serve as tracks and struts. Intermediate filaments are the most flexible in bending, yet a single filament can be stretched beyond two or three times its resting length without breaking. Think of a rope beside a fishing rod: the rope bends at a touch and is far harder to snap. Flexible is not the same as weak, because bending and stretching are different tests.

Polymer arithmetic: critical concentration and treadmilling

Treat a filament end as a doorway. Subunits arrive at a rate proportional to the free subunit concentration C, and leave at a rate that does not depend on C:

net rate at one end (subunits per second) = kon × C − koff

Set the net rate to zero and you get the concentration at which the end neither grows nor shrinks:

critical concentration, Cc = koff / kon

Above Cc the end grows; below it, the end shrinks. In a tube of pure actin, filaments form until the free monomer falls to near Cc, and extra actin beyond that goes into polymer.

Now the role of the nucleotide. If subunits never hydrolysed their ATP, both ends would have to share one critical concentration, because a subunit makes the same bonds whichever end it joins. Actin does hydrolyse its ATP, after it has joined the filament, and phosphate is released slowly, so the older lattice holds ADP-actin. The two ends then see different mixtures of subunits and can have different critical concentrations. Pollard measured this directly in 1986: with purified actin at 22 °C, elongation at each end rose linearly with ATP-actin concentration up to 20 micromolar, and the slopes and intercepts gave separate rate constants for each end. For ATP-actin the barbed end has the lower critical concentration.

When free actin sits between the two critical concentrations, the barbed end grows while the pointed end shrinks. At steady state the two rates match, the length stays constant, and subunits flow through the filament from front to back. This is treadmilling. Think of a queue that people join at the back and leave at the front: its length holds while everyone moves. Single purified actin filaments watched by microscopy show barbed-end growth and pointed-end shortening at approximately equal rates at steady state.

Binding above the critical concentration drives assembly. Hydrolysis comes afterwards and makes the old polymer less stable, which turns the steady state into a flux.

Dynamic instability: a cap that is lost

Microtubule ends do something different. Each end switches between steady growth and rapid shortening. The switch from growth to shortening is a catastrophe; the switch back is a rescue. This is dynamic instability, a property of individual ends, and a different behaviour from treadmilling.

Mitchison and Kirschner proposed dynamic instability in 1984 from measurements on populations of microtubules in vitro: the way numbers and lengths changed over time fitted growing and shrinking microtubules that coexist and interconvert rather infrequently. Single microtubules were watched switching in real time soon after, by dark-field microscopy (Horio and Hotani, 1986) and by video light microscopy in Ted Salmon’s laboratory (Walker and colleagues, 1988). The Walker study used pig brain tubulin free of associated proteins, grown from axoneme fragments at 37 °C. Both ends were dynamic. Rapid shortening did not change with tubulin concentration over 7 to 15.5 µM, and more tubulin made catastrophes rarer at both ends. In cells, minus ends are often capped or anchored, for example at the centrosome.

The standard explanation is the GTP cap. A dimer joins with GTP on its β-tubulin, and the GTP is hydrolysed soon after, inside the lattice. GDP-tubulin is held in the straight lattice only by its neighbours; the end stays intact while a cap of not-yet-hydrolysed dimers remains. Picture a stack of compressed springs held by a lid built from the newest springs. If addition slows and hydrolysis catches up, the lid is lost and the stack flies apart, with protofilaments peeling outward from the end. Roostalu and colleagues tested this in 2020 with recombinant human tubulin carrying mutations that slow hydrolysis. Caps grew longer, and the binding of end-binding (EB) proteins, which recognize the cap region, suggested that the cap matures gradually, with a wild-type rate estimated near 0.2 per second. Exactly what the cap looks like, and what final event commits an end to shortening, is still argued.

A worked problem: reading one trace

The data below are made up for teaching. They describe one plus end grown from a seed in purified tubulin; length is measured from the seed.

• 0 s, 5.0 µm; steady growth to 180 s, 9.5 µm; catastrophe
• 192 s, 3.5 µm; rescue; steady growth to 312 s, 6.5 µm; catastrophe
• 325 s, 0 µm (back to the seed)

Step 1, growth velocity. (9.5 − 5.0) µm over 180 s (3 min) = 1.5 µm per min. The second growth phase gives (6.5 − 3.5) µm over 120 s (2 min) = 1.5 µm per min, the same.

Step 2, shortening velocity. 6.0 µm in 12 s = 0.5 µm per s = 30 µm per min, twenty times faster than growth. The second shortening phase, 6.5 µm in 13 s, gives the same 0.5 µm per s.

Step 3, switching frequencies. The end spent 180 + 120 = 300 s growing and had 2 catastrophes: 2/300 = 0.0067 per s, or 0.4 per min. It spent 12 + 13 = 25 s shortening and had 1 rescue: 1/25 = 0.04 per s.

Step 4, dimers per second. Assume 13 protofilaments and 8 nm per dimer, so one micrometre holds 125 × 13 = 1,625 dimers. Growth of 1.5 µm per min is 0.025 µm per s, times 1,625 = about 41 dimers per s.

Check against real data. With Walker’s plus-end constants (kon = 8.9 per µM per s, koff = 44 per s) at 10 µM tubulin, net growth is 89 − 44 = 45 dimers per s, about 1.7 µm per min by the same conversion. Their rapid shortening, 733 dimers per s, is about 27 µm per min. The made-up trace is in the right range. These constants come from one purified preparation; in cells, associated proteins change them.

Drugs as tests

A drug tests a mechanism only if you know its target, its dose and the system.

• Taxol (paclitaxel) binds β-tubulin in the assembled lattice. With purified microtubule protein, Schiff and Horwitz found in 1981 that taxol-treated microtubules resisted disassembly on dilution and that taxol drove extra assembly after an apparent steady state. On a trace, predict that shortening becomes rare.
• Colchicine binds free tubulin dimers; the drug-dimer complex adds to an end and blocks further growth there, so small amounts suppress dynamics.
• Nocodazole also binds tubulin dimers. Its effect depends on dose. In cultured newt lung epithelial cells, 4 to 400 nM introduced a paused state and slowed both growth and shortening (Vasquez and colleagues, 1997); in BSC-1 monkey kidney cells after an hour, catastrophes became less frequent. The micromolar doses commonly used on cultured mammalian cells remove most microtubules.
• Phalloidin binds along actin filaments and slows their disassembly, as shown with purified muscle actin. It enters living cells poorly, so fluorescent phalloidin is mainly used to stain filaments in fixed cells.
• Cytochalasin B binds filament ends. Brown and Spudich showed this in 1981 by changing the number of ends on actin-coated beads while holding total actin constant: binding followed the ends.
• Latrunculin A forms a 1:1 complex with actin monomers (apparent Kd about 0.2 µM in pure actin). Single-filament imaging in 2018 found it binds ATP-actin far more tightly than ADP-actin (0.1 versus 4.7 µM) and also speeds disassembly at filament ends, so sequestration is not the whole story.

Intermediate filaments have no standard poison. Withaferin A redistributes vimentin in human fibroblasts, but it is not selective and did not alter the assembly of purified vimentin, so genetic loss and rescue is the cleaner test (experiments that distinguish). In cells, Rho-family GTPases decide where actin assembles (that lesson), and neurons use both polar polymers in the growth cone and for neuronal polarity.

Design problem

In a cultured fibroblast, the total actin fluorescence in the lamellipodium, the thin sheet at the leading edge, looks constant over 5 minutes. Design an experiment to decide whether the filaments there are static or treadmilling.