The development of the primary cilium, a microtubule-based organelle projecting from the surface of nearly all cells, has been proposed to be a consequence of evolved motor protein-based trafficking unique to eukaryotic cells3
. Primary cilia play important roles in sensory functions such as photoreception, renal functioning, and odorant sensing at single- and multicellular levels4–6
. Defective biogenesis or functioning of cilia causes a variety of human diseases, collectively termed ciliopathies7,8
, with pathological conditions including cystic kidney disease, brain malformations, and obesity.
Although able to respond to a variety of sensory stimulants, the basic structure of primary cilia is highly conserved. The core axoneme consists of a ring of nine doublet microtubules that extend from the mother centriole at the basal body1,9
. Ciliary construction and maintenance proceeds through IFT of ciliary components along the axoneme by kinesin and dynein motors9
. In C. elegans
, IFT requires the coordinated efforts of heterotrimeric kinesin-2 (KIF3A/KIF3B/KAP complex) and homodimeric OSM-3 motors10,11
. KIF17, the vertebrate homolog of OSM-3, has been shown to function as a ciliary motor in zebrafish photoreceptors and mammalian olfactory sensory neurons12–14
How kinesin motors and their cargos gain entry to the cilium is unknown. Ciliary entry is a selective process as analysis across several species has identified a unique ciliary proteome15
. Ciliary entry presumably requires the transport of proteins located near the basal body across the ciliary transition zone16
which may function as a diffusion barrier separating the cytoplasm from the intraciliary compartment. IFT cargo proteins have been observed around the basal body17
and transition fibers18
in the initial segment of cilia.
To study ciliary targeting of KIF17 in mammalian cells, we expressed mCitrine (mCit)-tagged KIF17 in cell lines that generate primary cilia. KIF17 accumulated at the distal tip of the primary cilium in all cell lines tested including neuronal (Odora rat olfactory sensory neurons19
), epithelial (MDCKII canine kidney and hTERT-RPE human retinal pigment epithelia) and fibroblast (NIH3T3) cells (). Localization to the distal cilium was confirmed by co-staining for acetylated and γ-tubulin to mark the cilium and basal body, respectively (). Ciliary localization of tagged KIF17 was observed regardless of the epitope (mCit, FLAG, or myc) or its position (N- or C- terminal) (data not shown).
Figure 1 The KIF17 CLS is necessary and sufficient for ciliary localization. (a) Odora, MDCK II, NIH3T3, and hTERT-RPE cells expressing full length KIF17-mCit (green) were fixed and stained for acetylated tubulin to mark cilia (red). Top row, images of entire (more ...)
To identify sequences in KIF17 required for ciliary localization, we created truncated forms of the motor ( and Supplementary Fig. S1a
). Deletion of the C-terminal tail domain abolished ciliary localization [KIF17(1-846), ], suggesting that the tail domain contains sequences required for ciliary targeting. Further C-terminal truncations also failed to localize to cilia (Supplementary Fig. S1
). Surprisingly, constructs containing the KIF17 stalk and tail domains [mCherry-KIF17(490-1029), ] or the KIF17 tail domain alone [myc-KIF17(801-1028), ] localized predominantly to the nucleus (). This suggests that similar mechanisms may control nuclear and ciliary targeting. Parallels between nuclear and ciliary import have been suggested in literature20–22
, yet no direct evidence exists to date.
To explore the possibility that ciliary entry of KIF17 utilizes mechanisms similar to nuclear import, we searched KIF17 for sequences resembling an NLS23
and identified two potential sites: aa767-772 (KRRKR) and aa1016-1019 (KRKK). To test whether these sequences are necessary for KIF17 ciliary localization, we mutated the relevant residues to alanines in the full length motor (). Mutation of residues 764-772 in the KIF17 stalk did not effect ciliary localization () whereas mutation of residues 1016-1019 in the KIF17 tail domain abolished ciliary localization (). Identical results were obtained in other cell lines (Supplementary Fig. S2
). These results indicate the KRKK sequence in the KIF17 tail domain acts as a CLS. The KIF17 CLS can also function as an NLS as mutation of residues 1016-1019 in the isolated KIF17 tail domain reduced the nuclear localization of this construct ().
That the CLS is necessary for ciliary localization of KIF17 () but is not sufficient for ciliary targeting when present in the isolated tail domain () indicates that there are likely to be several sequences in KIF17 that contribute to ciliary localization. To test whether the KIF17 CLS is sufficient for ciliary targeting of kinesin motors, we fused the tail domain (aa801-1028) onto the C-terminus of a non-ciliary kinesin, the Kinesin-1 subunit kinesin heavy chain (KHC) (). Fusion of the wildtype KIF17 tail domain resulted in ciliary localization of KHC whereas fusion of the mutant KIF17 tail did not (). This demonstrates that the KIF17 tail domain contains a CLS that is necessary and sufficient for ciliary targeting of kinesin motors.
Nuclear import involves recognition of NLSs by importin proteins, translocation through the nuclear pore complex (NPC), and dissociation of the NLS-importin complex in the nucleus by active GTP-bound forms of the small G-protein Ran24
. To investigate whether similar mechanisms regulate ciliary import of CLS-containing KIF17, we tested whether Ran-GTP is present in primary cilia and regulates trafficking of KIF17. Ran and importin proteins are present in ciliary proteomes from several species15,25
. We found that Ran is present in a ciliary fraction isolated from rat olfactory tissue (). Isolation of a ciliary-enriched fraction was confirmed by the presence of the ciliary protein adenylyl cyclase III () and scanning electron microscopy (SEM) of olfactory tissue before and after cilia removal (Supplementary Fig. S3a
). Immunohistochemistry of rat olfactory () and respiratory (Supplementary Fig. S3b
) epithelia also demonstrates that Ran is present in the cilia layer at the apical surface. The ciliary-localized Ran represents the active GTP-bound state of the protein as both the cilium () and nucleus (Supplementary Fig. S4a
) can be stained with an antibody that recognizes Ran-GTP but not Ran-GDP26
. These results are consistent with our proposal that a RanGTP/GDP gradient across the ciliary/cytoplasmic barrier regulates ciliary import.
Figure 2 Ran is present in the ciliary compartment. (a) Primary cilia were isolated from rat olfactory epithelium and the presence of Ran and adenylyl cyclase III (ACIII) in the ciliary (Cilia) and remaining deciliated (Decil.) fractions were determined by western (more ...)
To test whether ciliary Ran-GTP regulates KIF17 import, we co-expressed KIF17-mCit with myc-tagged Ran proteins (WT, constitutively active GTP-bound G19V mutant, and T24N mutant that cannot bind nucleotide)27
. We used serum-starved NIH3T3 cells in order to co-express the exogenous proteins after cilia formation and limit any effects of Ran overexpression on ciliogenesis. Cytoplasmic expression of WT or Ran(T24N) did not affect ciliary localization of KIF17 whereas expression of GTP-bound Ran(G19V) significantly reduced the number of cells with ciliary KIF17 without affecting cilia length (Supplementary Fig. S5
To alleviate concerns that cytoplasmic expression of Ran proteins could indirectly affect ciliary targeting of KIF17, we developed a method for fast upregulation of Ran protein expression. The various Ran constructs were tagged with a destabilization-domain (DD) which targets expressed proteins for rapid degradation. Addition of the cell-permeable ligand Shield-1 prevents protein degradation and allows rapid and continuous upregulation of protein levels28–30
. Our DD-Ran constructs were also tagged with the fluorescent protein Cerulean (Cer). To verify the rapid expression of Ran proteins, lysates of COS cells expressing the DD-Cer-Ran plasmids and exposed to Shield-1 for 0–8 h were analyzed by immunoblotting with a Ran antibody. Increasing exposure to Shield-1 resulted in increasing levels of DD-Cer-Ran with no change in endogenous Ran protein levels (). In live cells, upregulation of DD-Cer-Ran protein expression can be observed after only 1 h of incubation with Shield-1 ().
Figure 3 Fast upregulation of cytosolic Ran-GTP levels abolishes ciliary localization of KIF17. (a) COS cells expressing Cer-Ran (G19V, T24N, and WT), DD-Cer-Ran(G19V, T24N, or WT) or untransfected control cells were exposed to Shield-1 for 0 – 8 h (right (more ...)
We then tested whether a rapid increase in cytoplasmic DD-Cer-Ran affected ciliary targeting of KIF17. NIH3T3 cells coexpressing KIF17-mCit and DD-Cer-Ran proteins were treated with Shield-1 for 0–4 h and then fixed and stained with antibodies to acetylated and γ-tubulins (). After 4 h of Shield-1 treatment, increased expression of Ran(T24N) or WT Ran did not effect KIF17-mCit localization (). However, increased expression of GTP-bound Ran(G19V) abolished ciliary localization of KIF17-mCit (). Interestingly, at shorter times of Shield-1 addition and DD-Cer-Ran(G19V) expression, KIF17-mCit localized to more proximal segments of the cilium and/or to the basal body (, bottom row). Similar results were obtained upon live imaging of cells expressing DD-Cer-Ran and monomeric red fluorescent protein (mRFP)-KIF17 constructs (Supplementary Fig. S6
). Differences in DD-Cer-Ran(G19V) fluorescence intensity and nuclear localization between fixed () and live () cells are due to the methanol fixation/immunostaining procedure (Supplementary Fig. S4b
). The loss of KIF17-mCit ciliary localization upon increased DD-Cer-Ran(G19V) expression is not due to Ran-GTP effects on cilia per se as no effect on the presence or length of cilia was observed (). These results show that ciliary KIF17 is: 1) dynamic in its location and 2) mislocalized upon increased levels of cytoplasmic Ran-GTP. We suggest that cytoplasmic Ran-GTP abolishes ciliary entry of KIF17 while the dynamic process of IFT allows KIF17 already present in cilia to exit.
To directly test whether Ran controls ciliary entry of KIF17, we performed fluorescence recovery after photobleaching (FRAP) analysis of ciliary KIF17-mCit in the presence or absence of cytoplasmic DD-Cer-Ran(G19V). KIF17-mCit in the distal tips of Odora cilia was photobleached in a single confocal z-plane. In the absence of Shield-1, KIF17-mCit fluorescence in the distal tips of cilia recovered to pre-bleach levels within 20 min (). After 1 h of Shield-1-induced upregulation of DD-Cer-Ran(G19V), little to no recovery of KIF17-mCit fluorescence in the cilium was observed (). Comparison of fluorescence averages from multiple FRAP experiments revealed a drastic reduction in ciliary KIF17 recovery when Ran(G19V) levels were increased (). We conclude that the presence of KIF17 in primary cilia is a steady-state process in which motor is entering and leaving the cilium with a constant accumulation at the distal tip, and that entry of KIF17 can be prevented by high levels of cytoplasmic Ran-GTP.
Figure 4 Upregulation of cytosolic Ran-GTP levels prevents ciliary entry of KIF17. FRAP analysis of Odora cells coexpressing KIF17-mCit and DD-Cer-Ran(G19V) in the (a) absence or (b) presence of Shield-1. The cells were imaged (pre-bleach) and then the fluorescence (more ...)
We next examined whether importin proteins play a role in ciliary entry of KIF17. In nuclear import, NLS-containing proteins form complexes with α- and/or β-importins and are shuttled into the nucleus through NPCs24,31
. We hypothesized that CLS-containing proteins complex with importins for transport across the ciliary transition zone. This possibility is supported by the presence of α/β-importins in ciliary proteomes from several species15
and the interaction of the ciliary membrane protein Crumbs3 with importin-β1 during spindle assembly and ciliogenesis32
. In Odora cells, importin-β2 localized near the nuclear envelope, as expected, as well as near the basal body and in the proximal region of the cilium, consistent with a role for importins in ciliary import (). KIF17 interacts with importin-β2 as immunoprecipitation of Flag-KIF17 with anti-Flag antibodies resulted in coprecipitation of endogenous importin-β2 (). The KIF17 CLS is critical for this interaction as importin-β2 was not coprecipitated with the 1016-1019ala mutant (). That mutation of the KIF17 CLS (aa1016-1019) interfered with both ciliary entry () and importin-β2 binding (), indicates that interaction with importin-β2 is necessary for ciliary entry of KIF17. After crossing the ciliary transition zone, Ran-GTP in the cilium could dissociate KIF17 and importin-β2, freeing the motor for IFT. Indeed, addition of recombinant GST-Ran(G19V) to cell lysates prior to immunoprecipitation reduced the KIF17/importin-β2 interaction whereas addition of WT or Ran(T24N) had no effect ().
Figure 5 KIF17 forms a complex with importin-β2 that is CLS- and Ran-GTP-dependent. (a) Odora cells were fixed and stained with antibodies to importin-β2 and acetelyated tubulin. Top row, images of entire cells; scale bar, 10 μm. Bottom (more ...)
As the primary sequence of the KIF17 CLS is similar to classical NLSs that interact with importin-β1, the interaction of KIF17 with importin-β2 was surprising. However, no interaction between KIF17 and importin-β1 was observed via immunoprecipitation (). And similarities can be found between the KIF17 CLS and the basic-enriched/PY subclass of consensus sequences for importin-β2 binding23
. To directly compare roles of importins β1 and β2 in ciliary entry of KIF17, we replaced the KIF17 CLS with NLSs known to interact with either importin-β1 or importin-β2. KIF17-mCit still localized to primary cilia when its CLS was replaced with the M9 NLS from hnRNP A1, which interacts with importin-β2 (). In contrast, KIF17-mCit was targeted to the nucleus when its CLS was replaced with the NLS from the SV-40 large T antigen, which interacts with importins α and β1 (). These results suggest that importin-β2 alone is responsible for ciliary entry of KIF17.
In conclusion, we propose a model for ciliary import () in which cytoplasmic KIF17 interacts with importin-β2 (). This complex crosses the ciliary transition zone () and is dissociated by Ran-GTP in the proximal cilium (), allowing KIF17 to proceed with its role in IFT. Perturbation of the Ran-GTP/GDP gradient prevented ciliary entry of KIF17, presumably by inhibiting formation of KIF17/importin complexes before transport across the ciliary transition zone. These results provide the first direct evidence that ciliary and nuclear import pathways utilize similar mechanisms. In addition, Ran and importin proteins regulate the localization and activation of kinesin motors during spindle assembly in mitotic cells33,34
, and our work expands the role of Ran to include global regulation of kinesin compartmentalization in interphase cells.
How proteins gain access to the ciliary compartment has been unclear. Sequences critical for ciliary targeting of membrane proteins (e.g. VxPx, RVxP, Ax(S/A)xQ) are known13,35,36
but it is unclear how these sequences function as several pathways have been described for trafficking of membrane proteins to the cilium9, 37–39
. Here we describe a novel entry pathway for cytoplasmic kinesin motors analogous to nuclear entry of proteins. Whether importin- and Ran-regulated import pathways regulate entry of other motors and their cargoes at the ciliary transition zone requires further analysis.
It is interesting to note that the KIF17 CLS can function as a CLS or an NLS depending on protein context. Several NLS-like sequences have been found on the KIF3A/KIF3B/KAP complex, and the KAP subunit has been observed to redistribute from cilia nuclei during the mitotic cycle40
. It is likely that additional signals in KIF17 are required to promote ciliary rather than nuclear import. One possiibility is that KIF17 motor activity along cytoplasmic and/or centriole microtubules is needed to position the CLS-containing protein at the ciliary base rather than the nuclear envelope. Alternatively, cargo and/or membrane binding may be required for ciliary entry. Further experiments are required to test these possibilities as well as the global role of importins and Ran in ciliary entry.