Basal microtubules converge (yellow arrowhead) and are stabilized by end- to-side (red arrows) and side-to-side (pink arrows) interactions.

Epithelial microtubules (MTs) rely on the cortex and each other to create stable organized patterns, as shown by Reilein et al. (page 845).MTs are normally organized by centrosomes in animal cells. But in polarized epithelial cells, centrosomes are busy building the primary cilium under the apical membrane, leaving basal MTs to fend for themselves. Reilein and colleagues wanted to know how this dynamic yet stable basal network forms.

Epithelial cells are especially tall, so imaging their basal MT network is difficult. Reilein thus called upon an old technique that John Heuser referred to as “unroofing”—she got rid of the apical and lateral membranes to clear her view of the basal cytoskeleton.

What was uncovered was a network of mostly immobile MTs, with a few MTs growing or shrinking until they made contact with other MTs or with the cortex. Both the plus and minus ends of MTs were stabilized at these contacts, presumably by MT plus-end binding proteins (such as APC, Clip170, or EB1) or perhaps by γ-tubulin at the minus end.

Treadmilling, bundling, and MT motor–based movements—features that create MT networks in other noncentrosomal systems—were rare or absent in the basal networks. The authors thus supposed that MT–MT and MT–cortex interactions might be sufficient to create this organization.

To test their hypothesis, the group developed a computational model. They had already shown that cortex-associated adenomatous polyposis coli (APC) organizes new MTs, so they included in their model random sites of APC-mediated MT interactions, as well as MT–MT interactions and default MT dynamic instability. They found that the creation of a stable pattern from these inputs required only one additional parameter—that MTs contacting either another MT or the cortex be rescued from dynamic instability.

During their imaging, the group had noticed basal membranes whose MT network was somehow destroyed and then recreated de novo. They compared these actual forming networks with the model versions and got uncannily similar results. Now, the authors need to identify definitively the MT-stabilizing proteins that lie at MT–MT and MT–cortex contact sites.