The kinesin-4 motor protein Kif7 regulates Hedgehog signalling at cilia in mammals by controlling the activity of Gli transcription factors. Kif7 is now found to inhibit microtubule growth to restrict and coordinate the length of axonemal microtubules at the ciliary tip. Such Kif7-mediated organization of the ciliary tip compartment regulates Gli activity and is proposed to be required for correct Hedgehog signalling.
Hedgehog (Hh) signalling regulates embryonic development and tissue homeostasis in metazoa ranging from flies to humans1. In mammals, Hh signalling relies on the primary cilium, a microtubule-based organelle at the cell surface. The kinesin-4 Kif7 is a ciliary protein that regulates the activity of Gli transcription factors, the key mediators of the response to Hh ligands2,3,4,5, but the mechanistic basis of its function is poorly understood. In this issue, He et al.6 report that Kif7 modulates the length of microtubules at the tip of primary cilia and promotes the localization and appropriate regulation of Gli and the inhibitory factor Sufu (suppressor of fused) at the ciliary tip. Kif7 is likely to control axonemal microtubules directly, because the purified motor can accumulate at microtubule plus ends and inhibit their growth in an ATP-hydrolysis-dependent manner. This study thus directly links microtubule-regulation-dependent morphogenesis of primary cilia to the control of a major signalling pathway.
The assembly and maintenance of cilia depend on intraflagellar transport (IFT), a process driven by kinesin-2 and cytoplasmic dynein 2 motors. Multiple studies have shown that IFT is critical for most aspects of Hh signalling in mammals, not only because it is necessary to generate the cilium, the specialized structure that harbours Hh signalling components, but also because IFT directly regulates the movement of Hh components in and out of cilia7,8. The regulation of Hh signalling is complex, involving multiple factors that traffic in a coordinated manner into or out of cilia in response to pathway activation by secreted ligands such as Sonic hedgehog (Shh). In the absence of Shh, the Shh receptor Patched (Ptc) localizes to the cilium and inhibits the activity of the transmembrane protein Smoothened (Smo). As a result, the main effectors of Hh signalling, the Gli2 and Gli3 transcription factors, are processed into their repressor forms. In the presence of Shh, Ptc exits the cilium concomitantly with Smo ciliary entry and activation, triggering a series of events culminating in dissociation of Gli2/3 from Sufu and generation of full-length forms of Gli2/3, which are transported to the nucleus to activate transcription1 (Fig. 1a,b).
Despite recent advances, the mechanisms by which the cilium regulates Hh signalling remain obscure. In Drosophila melanogaster, where Hh signalling is mostly cilium-independent, the kinesin-4 Costal2 (Cos2) plays a central role in relaying Hh signals from Smo to Ci (the Drosophila homologue of mammalian Gli) by forming a microtubule-associated complex with Ci and other proteins that regulate Ci processing and activity1. The vertebrate homologue of Cos2, Kif7, similarly regulates Hh signalling by affecting Gli processing and activity2,3,4,5. However, in contrast to Cos2, mammalian Kif7 function in the Hh pathway depends strictly on the primary cilium2,4. Tagged versions of Kif7 were shown to accumulate at ciliary tips following Hh pathway activation in a manner dependent on the Kif7 motor domain2,4, but the mechanism by which Kif7 regulates Gli processing remained unclear.
He et al.6 analyse the effect of Kif7 on primary cilia by using two previously described Kif7 mouse mutants, a null mutant3 and a mutant with a leucine-to-proline substitution in the Kif7 motor domain that behaves like a null2. The authors demonstrated by immunofluorescence and electron microscopy that cilia on Kif7-mutant embryonic neural progenitor cells and cultured fibroblasts are abnormally long and have structural defects at the axonemal tip. These observations are similar to those reported for fibroblasts from human patients with Kif7 mutations9. By using cell-based assays and time-lapse imaging, He et al.6 went on to demonstrate that the abnormally long cilia of Kif7-mutant cells are unstable but display apparently normal rates of IFT, indicating that the ciliary phenotype of Kif7-mutant cells is not caused by IFT defects.
To investigate the link between Kif7 and microtubules in more detail, the authors characterized in vitro the properties of the purified Kif7 fragment containing the motor and the first coiled-coil domain. Using a fluorescence-microscopy-based assay, they demonstrated that Kif7 directly binds to microtubules and preferentially associates with their growing plus ends, where it inhibits microtubule polymerization in a manner dependent on ATP hydrolysis. This result is not surprising given the established role of other vertebrate kinesin-4 motors, Kif4A (homologue of Xenopus laevis Xklp1) and Kif21A, in the inhibition of microtubule growth10,11. However, in contrast to Kif4A and Kif21A, which are both processive microtubule plus-end-directed motors, Kif7 displays no motility and is unable to even diffuse along microtubules. Furthermore, while Kif4A and Kif21A suppress catastrophes and thus essentially stabilize microtubule plus ends, Kif7 increases catastrophe frequency and acts as a microtubule-destabilizing factor. Therefore, Kif7 can directly and negatively regulate microtubule length, and this is likely to explain why Kif7-mutant cells have elongated cilia. Interestingly, two other families of microtubule-depolymerizing kinesins, kinesin-13 and kinesin-8, have also been implicated in ciliary length control, although their activity has so far been limited to motile cilia12,13. Recruitment of negative regulators of microtubule growth might be a general feature of the length control of the otherwise strongly stabilized axonemal microtubules. It is possible that each type of structure (that is, primary or motile cilia) has its own set of specific regulators. For example, Kif27, the closest mammalian homologue of Kif7, is present in motile cilia14 and may thus share the ability of Kif7 to regulate axonemal microtubule dynamics.
Abnormal elongation of primary cilia led to defects in the structure of their tips, with some outer microtubule doublets defective or missing6. Aberrant cilia structure is known to impair localization and function of Hh components8. When He et al.6 investigated the localization of core Hh components in Kif7-mutant cells, Gli2 and Sufu localized abnormally in puncta along the axoneme, whereas in wild-type cells they accumulated at ciliary tips, as expected1. Further, despite no obvious defects in IFT rates, the pattern of IFT particle movement was perturbed in Kif7-mutant cells. Stationary puncta of IFT proteins were observed along the axoneme of these cells, and the initiation of retrograde movement of IFT particles, which normally occurs at the ciliary tip, was also visible in the middle of the axoneme, suggesting that compartments with tip-like properties became distributed along the cilium. He et al.6 propose that the main function of Kif7 in Hh signalling is to coordinate growth of individual axonemal microtubules, leading to the creation of a single specialized compartment at the ciliary tip for the appropriate localization and processing of Gli–Sufu complexes. When Kif7 function is impaired, Hh pathway components that normally localize to the ciliary tip, such as Gli and Sufu, mislocalize to ectopic tip-like compartments along the axoneme, causing deregulation of signalling (Fig. 1c). In addition to regulating axonemal microtubule plus ends, Kif7 might also contribute to the organization of Hh signalling complexes at the tip of the cilium through direct interactions with Hh pathway components3,4,5. This idea is supported by the fact that Drosophila Cos2 forms a large complex with Hh signalling components1 and can partially substitute for Kif7 in zebrafish5, suggesting that the functions of the two kinesins are similar.
An unresolved question is how Kif7 localizes to the tip of the cilium. This localization depends on a functional Kif7 motor domain2,6, yet it does not display any motility along microtubules6. The intrinsic microtubule-end-binding activity of Kif7 might confer or at least contribute to tip localization. In vitro, Kif7 associates preferentially to GMPCPP-bound microtubules, which mimic the GTP-bound microtubule lattices, suggesting that similarly to the microtubule plus-end-tracking protein EB1 (Refs 15,16), Kif7 might recognize the GTP cap at growing microtubule ends. Consistent with this idea, EB1 accumulates at the tip of cilia17, indicating that such a mechanism might indeed be operational at axoneme tips in spite of their slow dynamics.
The preferential association of Kif7 with growing microtubule ends does not explain why its ciliary tip localization increases in a Smo-dependent manner following treatment of cells with Shh agonist2,4,6. It is possible that, similarly to Kif21a (ref. 11) and other kinesins, Kif7 is autoinhibited and that association with Hh signalling pathway components relieves this inhibition, either through physical association or post-translational modifications. More complex reconstitutions with the full-length Kif7 molecule might help to clarify this issue and further reveal the fascinating complexity of Hh signalling mechanisms.