Cryo-electron tomography of human metaphase spindles shows that spindle microtubules form locally antiparallel bundles with a sharply preferred ~8 nm wall-to-wall spacing, organized by the kinesin-5 motor Eg5 and by density-driven steric interactions rather than by motor crosslinking (Conway et al., bioRxiv 2026). The mechanistic basis of this local self-organization is unresolved on three linked points: it is unknown whether Eg5 generates antiparallel alignment by capturing and aligning short, freshly nucleated microtubule stubs or by directly promoting microtubule nucleation itself; it is unknown how motor proteins can access and act on microtubules inside bundles whose ~8 nm spacing is narrower than the motors; and it is unknown what physical interaction (e.g., depletion forces from the crowded cytoplasm) produces the density-dependent bundling that sets the spacing. No discriminating evidence currently exists for any of these alternatives.

Background

The mitotic spindle is the bipolar machine that segregates chromosomes during cell division. In human cells it is built from microtubules — polar protein filaments that grow and shrink at their plus ends — together with hundreds of associated factors, most prominently motor proteins of the kinesin and dynein families, which use ATP to walk along microtubules and, when they bind two filaments at once, to crosslink and slide them. The kinesin-5 motor Eg5 (also called KIF11) is a homo-tetramer with motor domains at both ends of the molecule, so a single Eg5 can hold two microtubules and push antiparallel ones apart; inhibiting Eg5 with the small molecule monastrol collapses the spindle into a monopolar aster, whereas inhibiting dynein, the main minus-end-directed motor, defocuses the spindle poles. Classic models assumed that spindle microtubules nucleate at the two poles and grow inward to capture chromosomes, but structural studies have shown that human spindles contain many microtubules far too short to span the pole-to-chromosome distance, raising the question of how a bipolar spindle self-assembles from short, local building blocks.

Conway et al. (2026) used cryo-electron tomography (cryoET — three-dimensional electron imaging of vitrified, frozen-hydrated cells, here applied to thin lamellae of mitotic human cells prepared by focused-ion-beam milling) together with subtomogram averaging (averaging of repeating tubulin-lattice units extracted along each filament, which reveals the direction of each microtubule's plus end and hence its polarity) to map microtubule polarity across human metaphase spindles, with the bulk polarity cross-checked by second-harmonic-generation microscopy, an optical method that exploits the non-centrosymmetric structure of microtubules to report their average polarity. They report three principal findings. First, spindle microtubules are arranged in locally antiparallel bundles with a sharply preferred ~8 nm wall-to-wall (surface-to-surface) spacing between neighboring filaments. Second, this local antiparallel organization persists in monopolar spindles that still contain active Eg5 (produced by depleting centrioles with the PLK4 inhibitor centrinone) and is reduced when Eg5 is inhibited, so Eg5 organizes antiparallel overlap locally, independently of the spindle's overall bipolar geometry. Third, the spacing is the same for parallel and antiparallel neighbors, appears as a single peak in the bundle structure factor (the diffraction-like signature of lateral repeat distances within a bundle cross-section), and scales inversely with microtubule density — evidence that density-driven steric (excluded-volume) interactions, rather than motor-mediated crosslinking, set the spacing, with Eg5 and dynein acting on density in partly interchangeable ways.

Two nucleation pathways plausibly supply the short new microtubules that Eg5 acts on inside the spindle: the Augmin complex recruits the γ-tubulin ring complex (γ-TuRC, the principal microtubule-nucleating machinery) to existing microtubule lattices to nucleate branched filaments in the spindle bulk, and the RanGTP gradient around chromosomes releases assembly factors such as TPX2 that nucleate microtubules near chromatin. The alternative branch — that Eg5 itself promotes nucleation — also has support: in vitro, kinesin-5 stabilizes a lattice-competent tubulin conformation and thereby promotes microtubule nucleation and assembly (Chen et al. 2019), and in yeast and human cells the kinesin-5 Cut7/KIF11 interacts with γ-TuRC and counteracts kinesin-14-mediated inhibition of nucleation (Olmsted et al. 2014). How motor-driven sorting of dynamic microtubules self-organizes networks has been reconstituted in minimal in vitro systems (Roostalu et al. 2018), but those systems decouple nucleation from motor action. On the bundling-force side, depletion forces — entropic attractions that arise in a crowded solution because macromolecules too large to fit into the gap between two closely spaced filaments gain free volume when the filaments approach each other — are a physically plausible candidate for the density-dependent bundling interaction. Crowding-like attractive interactions are known to reshape active microtubule networks in vitro and in simulation (Chew et al. 2023), and cytoplasmic density has been shown to tune spindle architecture and scaling in cells (Kletter et al. 2025), though via nucleation regulation rather than bundling forces. Direct visualization of proteins inside cells at nanometer resolution is becoming possible with genetically encoded multimeric tags visible in cryo-EM (Fung et al. 2023), but this methodology has not yet been applied to motors inside dense spindle bundles.

The discussion of Conway et al. (2026) leaves the mechanistic interpretation of their structural findings open in three linked respects — the Eg5 alignment mechanism, motor access to filaments within ~8 nm-spaced bundles, and the physical origin of the density-dependent bundling.

Problem Statement

Determine the molecular and physical mechanism by which the kinesin-5 motor Eg5 organizes locally antiparallel microtubule bundles within the dense microtubule arrays of the human metaphase spindle, by answering — with evidence that discriminates among the stated alternatives — the following three linked questions:

(a) Antiparallel alignment: does Eg5 generate antiparallel microtubule overlap primarily by capturing and aligning short, freshly nucleated microtubule stubs (nucleated in the spindle bulk by Augmin or near the chromosomes by the Ran pathway), by directly promoting microtubule nucleation itself, or by a combination of both — for example as established in reconstituted systems with purified components (Eg5 together with Augmin/γ-TuRC- or Ran-pathway-mediated nucleation) or by targeted cellular perturbations?

(b) Motor access in dense bundles: how do motor proteins and other microtubule binders physically operate within bundles whose ~8 nm wall-to-wall spacing between adjacent microtubules is narrower than the dimensions of the motors themselves — i.e., which of the proposed access modes actually occurs: localization at lattice gap defects (sites where a microtubule is missing from the bundle array, leaving room to maneuver), binding restricted to an outer shell of the bundle (a coaxial-cable-like architecture with an older, motor-inaccessible core), or crosslinking of non-adjacent microtubules in the second or higher coordination shells (rings of next-nearest and further neighbors)? Discriminating evidence could come from direct visualization of motors interacting with microtubules inside dense bundles, in vitro and/or in situ, for example with genetically encoded cryo-EM-visible tags for small binders such as Eg5 or PRC1 and direct tomographic detection for larger motors such as dynein.

(c) Bundling forces: what physical interaction produces the density-dependent bundling that sets microtubule spacing — in particular, are depletion forces from the crowded, viscous cytoplasm responsible? This could be addressed, for example, by quantitative theory of depletion-mediated (or alternative) spacing in dense, dynamic microtubule arrays and by experiments in cells with perturbed cytoplasmic viscosity or crowding that test whether the observed ~8 nm spacing and its inverse scaling with microtubule density follow from the proposed interaction.

An answer must resolve each part at a level that discriminates the stated alternatives (not merely illustrate one of them), and must remain consistent with the established structural constraints: locally antiparallel organization independent of spindle bipolarity, identical spacing for parallel and antiparallel neighbors, a single spacing peak in the bundle structure factor, and inverse spacing–density scaling across conditions.

References

  1. Conway W, Bobe D, Zimyanin V, Fabig G, de Marco A, Redemann S (2026). Eg5 activity and density-driven bundling organize the human metaphase mitotic spindle independently of spindle bipolarity. bioRxiv preprint 2026.01.16.699769, doi:10.64898/2026.01.16.699769, https://doi.org/10.64898/2026.01.16.699769 (full text also deposited as Europe PMC PMC12871229).
  2. Chen G-Y, Cleary JM, Asenjo AB, Chen Y, Mascaro JA, Arginteanu DFJ, Sosa H, Hancock WO (2019). Kinesin-5 Promotes Microtubule Nucleation and Assembly by Stabilizing a Lattice-Competent Conformation of Tubulin. Current Biology 29, 2259-2269, doi:10.1016/j.cub.2019.05.075, https://doi.org/10.1016/j.cub.2019.05.075
  3. Olmsted ZT, Colliver AG, Riehlman TD, Paluh JL (2014). Kinesin-14 and kinesin-5 antagonistically regulate microtubule nucleation by γ-TuRC in yeast and human cells. Nature Communications 5, 5339, doi:10.1038/ncomms6339, https://doi.org/10.1038/ncomms6339
  4. Roostalu J, Rickman J, Thomas C, Nédélec F, Surrey T (2018). Determinants of Polar versus Nematic Organization in Networks of Dynamic Microtubules and Mitotic Motors. Cell 175, 796-808, doi:10.1016/j.cell.2018.09.029, https://doi.org/10.1016/j.cell.2018.09.029
  5. Chew W-X, Henkin G, Nédélec F, Surrey T (2023). Effects of microtubule length and crowding on active microtubule network organization. iScience 26, 106063, doi:10.1016/j.isci.2023.106063, https://doi.org/10.1016/j.isci.2023.106063
  6. Kletter T, Muñoz O, Reusch S, Biswas A, Halavatyi A, Neumann B, Kuropka B, Zaburdaev V, Reber S (2025). Cell state-specific cytoplasmic density controls spindle architecture and scaling. Nature Cell Biology 27, 959-971, doi:10.1038/s41556-025-01678-x, https://doi.org/10.1038/s41556-025-01678-x
  7. Fung HKH, Hayashi Y, Salo VT, Babenko A, Zagoriy I, Brunner A, Ellenberg J, Müller CW, Cuylen-Haering S, Mahamid J (2023). Genetically encoded multimeric tags for subcellular protein localization in cryo-EM. Nature Methods 20, 1900-1908, doi:10.1038/s41592-023-02053-0, https://doi.org/10.1038/s41592-023-02053-0
  8. Steckhahn D, Betterton MD (2026). Feedback between filament spacing, crosslinker binding, and self-organization in cytoskeletal bundles. bioRxiv preprint 2026.07.14.738536, doi:10.64898/2026.07.14.738536, https://doi.org/10.64898/2026.07.14.738536
  9. Mullin-Bernstein Z, van Wierst S, Garrison C, Cho N, Dumont S (2026). Midzone bundles of the mammalian anaphase spindle are mechanically coupled both locally and globally. bioRxiv preprint 2026.04.08.716168, doi:10.64898/2026.04.08.716168, https://doi.org/10.64898/2026.04.08.716168
  10. Romeiro Motta M, Grisam C, Grünewald M, König B, Meißner L, Niese L, Diez S, Hamant O, Schaedel L (2025). MAP65-1/PRC1 reinforces microtubules through nucleation. bioRxiv preprint 2025.09.14.676082, doi:10.1101/2025.09.14.676082, https://doi.org/10.1101/2025.09.14.676082