Module 2: Microtubules

The microtubule subunit is an obligate heterodimer of α-tubulinand β-tubulin (each ~50 kDa, ~450 residues, ~40% sequence identity). Both bind a guanine nucleotide at orthologous sites: the α-subunit has a non-exchangeable GTP (the “N-site”) buried at the longitudinal interface, while the β-subunit binds an exchangeable guanine nucleotide (the “E-site”) at the outer surface of the heterodimer. It is the β-bound nucleotide that is hydrolyzed during polymerization and that defines the lattice state (GTP vs. GDP).

Heterodimers stack head-to-tail into polar linear protofilaments, and 13 protofilaments associate side-by-side to form a hollow cylinder 25 nm in outer diameter with a 15 nm-wide lumen. The lattice exhibits a seam: 12 of the 13 lateral α-α / β-β contacts are “B-lattice” homologous, but one interface is offset by three monomers (the “A-lattice seam”) to produce a 3-start helical pitch (Mandelkow, Mandelkow & Milligan 1991; Nogales, Whittaker, Milligan & Downing 1999). Occasionally microtubules polymerize with 12 or 14 protofilaments, andin vitro tubulin polymerized in the absence of GTP or with taxol can assemble 9-16 protofilament tubes; 13 is the dominant number in cells.

The microtubule is intrinsically polar. The β-tubulin face points toward the plus (+) end, the α-tubulin face toward the minus (−) end. Plus ends grow fast and display dynamic instability; minus ends in cells are typically tethered and stabilized at microtubule-organizing centers (MTOCs) by the γ-tubulin ring complex (γ-TuRC). This polarity dictates which end interacts with the kinesin and dynein motor families introduced in Module 4.

In solution, tubulin heterodimer concentration is roughly 15-25 μM in cycling cells. GTP-tubulin exchanges its E-site nucleotide on free dimers in seconds; once incorporated into the lattice, GTP is hydrolyzed stochastically and the dimer is locked into its GDP state until the lattice disassembles (Carlier 1982; Mitchison 1993). This irreversibility at the single-dimer level is what makes the microtubule a non-equilibrium polymer.

Each plus end is a 13-site landing pad: a new GTP-tubulin heterodimer binds a protofilament tip with apparent second-order rate constant \(k_{\text{on}} \approx 3\)\(\;\mu\)M\(^{-1}\) s\(^{-1}\) and dissociates with\(k_{\text{off}} \approx 24\) s\(^{-1}\) (Mitchison 1993; Gardner, Zanic & Howard 2013). Because the lattice has 13 protofilaments, the bulk subunit addition rate at a growing plus end is about 13\(\times\) larger than the per-protofilament rate. The minus end has much lower on-rates (\(\sim 10\times\) smaller) and is usually capped by γ-TuRC in cells.

GTP hydrolysis is not required for polymer addition—non-hydrolyzable analogs (GMPCPP, GTPγS) still support elongation—but hydrolysis is what destabilizes the lattice and sets up catastrophe. Crystallography (Löwe et al. 2001; Nogales, Downing, Amos & Löwe 1998) reveals that GTP-tubulin is “straight” and packs into a cylindrical lattice, while GDP-tubulin has a ~12\(^\circ\) kink that prefers an outward-curving “ram’s horn” protofilament—a stressed configuration stored in the body of a growing microtubule and released upon catastrophe.

The observed plus-end growth velocity at tubulin concentration \([T]\) is approximately linear:

Shrinkage proceeds by loss of GDP-tubulin from ram’s-horn tips at high rate; the polymer does not randomly depolymerize but unzips catastrophically once the GTP cap is lost. The ratio of shrink rate to growth rate is typically 7-10\(\times\).

In their 1984 Nature paper, Mitchison & Kirschner observed—contrary to the equilibrium polymerization paradigm of actin—that individual microtubules coexist in two populations: some are growing, some are shrinking, at the same monomer concentration. Individual microtubules stochastically flip between these two states:

• Growth at slow velocity \(v_g \approx 1{-}3\) μm/min.
• Catastrophe: stochastic transition to shrinkage at frequency \(f_c\).
• Shrinkage at fast velocity \(v_s \approx 10{-}20\) μm/min.
• Rescue: stochastic transition back to growth at frequency \(f_r\).

Catastrophe frequency in mammalian interphase cells is roughly \(f_c \approx 0.01{-}0.03\)s\(^{-1}\) (Walker et al. 1988; Gardner 2008); rescue is less frequent in vitro but is enhanced in cells by CLASPs and other plus-end factors. The phenomenon is purely of the plus-end cap; minus ends show little dynamic instability once capped by γ-TuRC.

The GTP cap hypothesis (Mitchison & Kirschner 1984) explains catastrophe as the loss of a protective layer of GTP-tubulin at the growing tip. Because GDP-tubulin has a kinked conformation that disfavors lateral contacts, a microtubule lattice made of GDP-tubulin is mechanically unstable and peels. Only the straight GTP-cap holds it together. When stochastic fluctuations shrink the cap to zero, GDP-tubulin is exposed and the microtubule catastrophes.

Treat the cap as an M/M/1 queue: GTP-tubulin is added at rate \(k_{\text{on}}[T]\)and aged (hydrolyzed) at rate \(k_{\text{hyd}}\) per subunit. In steady state, the mean cap length in dimers is

Because the first passage of a birth-death queue to zero is exponentially rare, the probability that the cap has shrunk to nothing at any given test is

Experimental fluorescence imaging of EB1 comets (Bieling et al. 2007; Akhmanova & Steinmetz 2008) reveals that the GTP (or GTP-γ-P_i) cap is ~0.5-1 μm long—a few hundred dimers—consistent with the M/M/1 prediction at physiological tubulin concentrations. The cap extends beyond the pure GTP region, indicating that the recognition structure is actually a lattice conformation, not the nucleotide itself (Maurer, Bieling, Cope, Hoenger & Surrey 2011).

Treat a population of microtubules as a two-state Markov chain. Let \(p_g(L,t)\)and \(p_s(L,t)\) be the probability densities of a microtubule of length\(L\) in growth or shrinkage at time \(t\). The master equation is

Dogterom & Leibler (1993, Phys. Rev. Lett.) solved this in the stationary, bounded regime where\(f_c\,v_s \,>\, f_r\,v_g\). The steady-state marginal length distribution is exponential:

If \(\lambda < 0\) (the unboundedregime), lengths grow without limit; the population drifts at velocity\(V = (v_g f_r - v_s f_c)/(f_c + f_r)\). Cells tune their tubulin concentration and the catastrophe regulators to keep individual microtubules in the bounded regime for interphase arrays, but switch to the unbounded regime transiently during mitotic spindle assembly to efficiently sample cytoplasmic space.

The fraction of microtubules in the growing state at steady state is\(p_g = f_r/(f_c+f_r)\), a two-state detailed-balance result; in the bounded regime this prediction is confirmed by fluorescence speckle microscopy in interphase cells.

Treating the microtubule as a homogeneous isotropic elastic rod of outer radius\(R_o \approx 12.5\) nm and inner radius\(R_i \approx 7.5\) nm, the flexural rigidity is\(EI \approx E\cdot \pi(R_o^4 - R_i^4)/4\) with Young’s modulus\(E \approx 1\) GPa (isotropic estimate). Thermal bending measurements (Gittes et al. 1993; Pampaloni et al. 2006) give

For cell-scale lengths (\(L \ll L_p\)), microtubules behave as essentially straight rigid rods—perfect for long-range intracellular transport tracks. However, Pampaloni et al. (2006) showed that the apparent persistence length depends on contour length, increasing from ~0.1 mm at \(L = 2\) μm to several millimeters at \(L = 50\) μm, reflecting anisotropic elasticity of the helical lattice (strong axial, weak lateral bonds).

Microtubules withstand pushing forces of up to ~5 pN per filament before stalling against obstacles (Dogterom & Yurke 1997); beyond the stall force, addition slows exponentially in the applied force.

Spontaneous microtubule nucleation from pure tubulin in vitro is extremely slow: the nucleus is a 2-D sheet of ~7 dimers that must bend into the cylindrical lattice—a high entropic barrier (Fygenson, Flyvbjerg, Sneppen, Libchaber & Leibler 1995). Cells bypass this barrier entirely by using a pre-assembled template.

The γ-tubulin ring complex (γ-TuRC) is a multi-subunit lock-washer-shaped template of ~13-fold symmetry that mimics a microtubule minus-end (Moritz et al. 2000; Kollman, Merdes, Mourey & Agard 2011Nat. Rev. Mol. Cell Biol.; Consolati et al. 2020). It contains γ-tubulin heterodimers plus γ-tubulin complex proteins (GCPs 2-6) that arrange in a helical ring matching the 13-protofilament lattice. Tubulin heterodimers dock onto the γ-TuRC with their β-tubulin face outward and grow plus-end outward while the minus-end remains capped by γ-TuRC.

In animal cells, γ-TuRC is localized to MTOCs:

• Centrosome: pair of centrioles surrounded by pericentriolar material (PCM) containing γ-TuRC, CDK5RAP2, pericentrin, ninein; the dominant MTOC in most interphase cells.
• Golgi apparatus: AKAP450-docked γ-TuRC; nucleates asymmetric arrays in neurons and secretory cells.
• Augmin complex: HAUS1-8 scaffold that recruits γ-TuRC to pre-existing spindle microtubules, generating branched MT amplification (Goshima, Wollman, Goodwin, Zhang, Scholey, Vale & Stuurman 2008).
• Spindle pole body (yeast) and chromatin-driven nucleation via Ran-GTP and TPX2.

The microtubule lattice is decorated by a large family of MAPs that regulate stability, length, bundling, nucleation, and motor attachment:

• MAP2 (neuronal dendrites, ~280 kDa): binds and bundles MTs, spaces them at a characteristic distance; expressed mainly in dendrites.
• MAP4 (ubiquitous, ~220 kDa): stabilizes interphase MTs; phosphorylated and released at mitosis.
• tau (neuronal axons, ~50-70 kDa): six splice isoforms; stabilizes axonal MTs; hyperphosphorylated tau forms the paired helical filaments (PHFs) of Alzheimer’s disease neurofibrillary tangles (Mandelkow 1994 Trends Cell Biol.; Goedert 1996).
• Stathmin / Op18: sequesters free tubulin dimers and promotes catastrophe; a master mitotic regulator.
• XMAP215 / ch-TOG (Stu2): processive polymerase at the plus end, accelerates growth up to 10x (Brouhard et al. 2008).
• Katanin (AAA+ ATPase, ~83 kDa): severs MTs; essential for meiotic spindle shortening.
• Spastin: severing; spastin mutations cause hereditary spastic paraplegia (Roll-Mecak & Vale 2008).
• Fidgetin and katanin-like: additional severing family members.

Plus-end tracking proteins (+TIPs)

A specialized subset of MAPs—the +TIPs—recognize the growing plus end and form comet-like accumulations that track elongation. The master +TIP is EB1 (and its paralogs EB2/EB3): a small dimeric protein with an N-terminal calponin-homology (CH) domain that binds the GTP/GDP-P_i lattice directly, and a C-terminal EB homology (EBH) domain that recruits downstream SxIP- or CAP-Gly-containing partners (CLIP170, p150Glued, CLASPs, MCAK, APC, ch-TOG). See Akhmanova & Steinmetz (2008, 2015 Nat. Rev. Mol. Cell Biol.).

EB1 comets move with the growing tip at the polymerization velocity and dissociate when the lattice matures to full GDP. Tracking EB1 in live cells is the standard way to visualize microtubule dynamics in vivo. Because EB1 also binds the lattice seam, it has been used structurally to identify seams in cryo-EM.

During mitosis the interphase MT array is torn down and a new bipolar spindle is built from two centrosomal asters. Kirschner & Mitchison (1986, Cell) proposed the search-and-capture model: microtubules grow and shrink stochastically (dynamic instability) in random directions, and those that happen to encounter a kinetochore are stabilized and captured. Because of the exponential length distribution, the probability that a given MT reaches a kinetochore at distance\(d\) is \(\sim e^{-\lambda d}\), and the expected capture time scales as \(\tau_{\text{cap}} \sim d \cdot \lambda / (\text{nucleation rate})\).

Search-and-capture alone is too slow for large cells; amplification mechanisms (augmin-driven nucleation, chromatin-driven RanGTP gradient, motor-driven transport) are required for rapid bipolar attachment. Wollman, Cytrynbaum, Jones, Meyer, Scholey & Mogilner (2005, Curr. Biol.) quantified the time budget and showed that spatial biases are essential.

Kinetochore-microtubule attachments mature from lateral to end-on, with the MT plus end embedded in the Ndc80-Mis12-Knl1 outer-kinetochore ring. Chromosome congression to the metaphase plate uses both polar ejection force from dynamic MTs and kinesin-7/CENP-E. Aurora B kinase destabilizes incorrect (syntelic, merotelic) attachments in the error correction cycle. The spindle assembly checkpoint delays anaphase until all kinetochores are bi-oriented.

Motile cilia and eukaryotic flagella are built on a stereotyped microtubule architecture: nine doublet microtubules (A-tubule with 13 protofilaments, B-tubule with 10 sharing a wall) around two central singlet microtubules—the famous 9+2 axoneme. The doublets are templated from the triplet microtubules of the basal body (a modified centriole).

Motility is powered by axonemal dynein:

• Outer dynein arms (ODA): repeat every 24 nm on each A-tubule; provide the bulk of the beat force.
• Inner dynein arms (IDA): heterogeneous family of seven isoforms; shape the waveform.
• Radial spokes and the central pair coordinate the beat asymmetry.
• Nexin / DRC link neighboring doublets and convert sliding to bending.

Loss of outer dynein arms causes Kartagener syndrome(primary ciliary dyskinesia): immotile respiratory cilia, chronic sinusitis, bronchiectasis, male infertility (immotile sperm), and situs inversus in about half of affected individuals because embryonic nodal cilia cannot break left-right symmetry.

Primary cilia and ciliopathies

Most vertebrate cells also project a single, non-motile primary cilium (9+0 axoneme, no dynein) that functions as a signaling antenna for Hedgehog, Wnt, and GPCR pathways. Mutations in ciliary proteins cause a spectrum of ciliopathies: Bardet-Biedl syndrome, Joubert syndrome, Meckel syndrome, polycystic kidney disease, and nephronophthisis. Intraflagellar transport (IFT) along the axoneme uses kinesin-2 (anterograde) and cytoplasmic dynein-2 (retrograde) and is covered in Module 4.

Because microtubules are absolutely required for mitosis, they are a major target for cancer chemotherapy. Tubulin-binding drugs fall into two classes: stabilizers that suppress dynamics at low doses and prevent disassembly at high doses, and destabilizersthat promote depolymerization.

At clinically relevant concentrations, these drugs do not shift net MT polymer mass dramatically; they suppress the dynamic instability (catastrophes and rescues) so that the spindle cannot bi-orient chromosomes, triggering the spindle-assembly checkpoint and mitotic-arrest-induced apoptosis (Jordan & Wilson 2004, Nat. Rev. Cancer).

Microtubules serve as tracks for two superfamilies of ATP-hydrolyzing motors, covered in detail in Module 4:

• Kinesins (~45 families; kinesin-1/KIF5, kinesin-2/KIF3, kinesin-5/Eg5, kinesin-13/MCAK, ...): mostly plus-end-directed processive motors; cargo trafficking toward the cell periphery, anterograde axonal transport, and spindle organization.
• Dyneins (cytoplasmic dynein-1 for cargo and mitosis, cytoplasmic dynein-2 for IFT, axonemal dyneins for cilia): minus-end-directed AAA+ ring motors, ~1.4 MDa complexes requiring dynactin and cargo adaptors (BICD2, Hook3, ninein) for processive motion (Reck-Peterson, Redwine, Vale & Carter 2018).

Because of microtubule polarity, kinesin-mediated transport is directed outward (axonal, secretory) and dynein-mediated transport is directed inward (endocytic, retrograde, nuclear envelope positioning). The balance of these opposing motors on the same cargo sets organelle distribution.