Acute inhibition of heterotrimeric kinesin-2 function reveals mechanisms of intraflagellar transport in mammalian cilia

SUMMARY

The trafficking of components within cilia, called intraflagellar transport (IFT), is powered by kinesin-2 and dynein-2 motors. Loss of function in any subunit of the heterotrimeric KIF3A/KIF3B/KAP kinesin-2 motor prevents ciliogenesis in mammalian cells and has hindered an understanding of how kinesin-2 motors function in cilium assembly and IFT. We used a chemical-genetic approach to generate an inhibitable KIF3A/KIF3B/KAP kinesin-2 motor (i3A/i3B) that is capable of rescuing WT motor function for cilium assembly and Hedgehog signaling in Kif3a/Kif3b double-knockout cells. We demonstrate that KIF3A/KIF3B function is required not just for cilium assembly, but also for cilium maintenance as inhibition of i3A/i3B blocks IFT within two minutes and leads to a complete loss of primary cilia within eight hours. In contrast, inhibition of dynein-2 has no effect on cilium maintenance within the same time frame. The kinetics of cilia loss indicate that two processes contribute to ciliary disassembly in response to cessation of anterograde IFT: a slow shortening that is steady over time and a rapid deciliation that occurs with stochastic onset. We also demonstrate that the kinesin-2 family members KIF3A/KIF3C and KIF17 cannot rescue ciliogenesis in Kif3a/Kif3b double-knockout cells nor delay the loss of assembled cilia upon i3A/i3B inhibition. These results demonstrate that KIF3A/KIF3B/KAP is the sole and essential motor for cilium assembly and maintenance in mammalian cells. These findings highlight differences in how kinesin-2 motors were adapted for cilium assembly and IFT function across species.

Graphical Abstract

eTOC blurb

Motors in the kinesin-2 family drive intraflagellar transport in cilia across species. Engelke et al. engineer a version of the heterotrimeric kinesin-2 motor KIF3A/KIF3B/KAP that can be acutely inhibited. They demonstrate that KIF3A/KIF3B/KAP is essential for the assembly and maintenance of primary cilia in Kif3a^−/−;Kif3b^−/− NIH-3T3 cells.

INTRODUCTION

Cilia are microtubule-based organelles that protrude from the surface of almost every cell in the body. The beating motion of motile cilia powers cell motility (e.g. sperm) or propels fluids over organ surfaces (e.g. respiratory tract). Primary cilia, in contrast, are largely immotile and function as cellular antennae to sense extracellular stimuli and integrate cellular signaling pathways [1, 2]. For example, in vertebrates, Hedgehog (HH) signaling critically depends on the presence of a primary cilium to direct cell proliferation and cell differentiation during development and adult homeostasis [3]. Given the central functions of cilia as motile organelles and cellular signaling hubs, it is not surprising that ciliary malfunction gives rise to a collection of disabling and sometimes life-threatening diseases and syndromes (the ciliopathies) with phenotypes such as developmental malformations, cystic kidney disease, mental retardation, retinal degeneration, and cancer [4].

Intraflagellar transport (IFT) involves the active transport of proteins through the ciliary shaft and is indispensable for the genesis, structural maintenance, and function of nearly all types of cilia [5]. At the base of the cilium, IFT proteins assemble into linear arrays of repeating IFT-A and IFT-B complexes that are named IFT trains based on their beads-on-a-string-like appearance in transmission electron microscopy and tomography images [5]. Molecular motor proteins of the kinesin-2 family bind to IFT trains and transport them along the microtubules of the axoneme to the tip of the cilium and dynein-2 motors drive transport back to the cell body [6].

Current models of IFT are based on invertebrate model systems such as C. elegans and Chlamydomonas, however, these models are inadequate to describe kinesin-2 functions in IFT and ciliary signaling in mammalian cells. For example, the homodimeric kinesin-2 motor OSM-3 is capable of building a full-length cilium in C. elegans but has no homolog in Chlamydomonas and its mammalian homolog KIF17 appears to be dispensable for cilium formation in mice [7–10]. On the other hand, the heterotrimeric kinesin-2 motor Klp20/Klp11/Kap is dispensable for cilium formation in C. elegans but is essential in Chlamydomonas [7, 11] and genetic ablation of any subunit of the homologous mammalian KIF3A/KIF3B/KAP complex results in embryonic lethality, presumably due to an absence of cilia [12-14].

The function of kinesin-2 motors in IFT, and of molecular motors in intracellular transport in general, has largely been studied using genetic methods such as knockout or knockdown of gene expression. However, these approaches are subject to off-target and indirect effects and are not compatible with the time scale of motor-driven transport events. Here, we use a novel chemical-genetic strategy [15] to generate an inhibitable version of the heterotrimeric KIF3A/KIF3B/KAP kinesin-2 motor by engineering of the KIF3A and KIF3B motor subunits. Our approach enables highly-specific and rapid inhibition of the engineered motor without the need for screening small molecule libraries and thus enables us, for the first time, to acutely inhibit the KIF3A/KIF3B/KAP motor and study its function in ciliated cells. We find that kinesin-2 inhibition results in a rapid cessation of IFT and complete disassembly of the primary cilium within eight hours. This is in contrast to dynein-2 inhibition, where primary cilia are largely intact after a 12-hour treatment. Expression of the homodimeric kinesin-2 motor KIF17 neither rescues ciliogenesis nor delays the breakdown of cilia following heterotrimeric kinesin-2 inhibition. These results provide the foundation of a comprehensive model that defines the critical features of IFT during ciliogenesis and maintenance of cilium structure in mammalian cells.

RESULTS

Kif3a;Kif3b-deficient NIH-3T3 cells cannot generate primary cilia

To study the function of heterotrimeric kinesin-2 in IFT in mammalian primary cilia, we generated NIH-3T3 cells that lack expression of both heterotrimeric kinesin-2 motor subunits, KIF3A and KIF3B (Kif3a^−/−;Kif3b^−/− cells), using CRISPR/Cas9-mediated genome editing. A cell line with a one base pair (bp) insertion and a one bp deletion in the Kif3a alleles and a one bp insertion and a 123 bp insertion in the Kif3b alleles was generated (Figure S1). All of the altered open reading frames code for premature stop codons, presumably resulting in nonsense-mediated mRNA decay. The loss of KIF3A expression in Kif3a^−/−;Kif3b^−/− cells was verified by western blotting and immunostaining (Figure S2 A,B).

Upon cell cycle arrest induced by serum starvation, primary cilia were observed in the parental NIH-3T3 cell line transfected with a control (mCherry) plasmid but not in the Kif3a^−/−;Kif3b^−/− cells (Figure 1 A,B,E). Expression of either KIF3A or KIF3B alone was not sufficient to rescue ciliogenesis, whereas co-expression of both KIF3A and KIF3B motor subunits fully rescued the ability to generate a primary cilium (Figure 1 C-E). We note that the fluorescently-tagged, full-length KIF3A and KIF3B subunits assemble into an auto-inhibited, soluble motor that is diffusely localized throughout the cytosol, similar to other kinesin motors [16, 17]. The ability of ectopic KIF3A and KIF3B expression to rescue the loss of cilia phenotype confirms that the knockout is on target and that there is no expression of a dominant-negative acting kinesin species from the CRISPR/Cas9-modified alleles.

Figure 1. KIF3A/KIF3B-deficient NIH-3T3 cells are not able to generate primary cilia.

(A) Parental (NIH-3T3) and (B-D) double knockout (Kif3a^−/−;Kif3b^−/−) cells ectopically expressing (A-B) mCherry or (C-D) KIF3A-mNeonGreen and/or KIF3B-mCherry were serum-starved for two days then fixed and stained with DAPI (blue) and an antibody to a marker of the ciliary membrane (ARL13B; magenta). Transfected cells are indicated by a white outline. Primary cilia are marked by a white arrowhead. Scale bar, 10 μm. (E) Quantification of the percentage of transfected cells that generate a primary cilium. Each dot represents the average of one independent experiment. n ≥ 94 transfected cells per condition. See also Figure S1 and S2.

Characterization of inhibitable kinesin-2 motor constructs

We previously described two strategies to generate inhibitable kinesin motors [15]. For generating inhibitable forms of the kinesin-2 motor KIF3A/KIF3B/KAP, we chose to implement the B/B strategy based on the selectivity and applicability of the B/B homodimerizer and the fact that inhibition of KIF3A/KIF3B motors with this strategy will not affect KIF3A/KIF3C motors. The B/B strategy makes use of the ability of the small molecule B/B homodimerizer (a rapamycin derivative) to rapidly and specifically mediate the dimerization of DmrB domains (homodimerizing FKBP variant). By fusing the DmrB domains to the N-termini of KIF3A and KIF3B, the B/B homodimerizer will crosslink and thereby inhibit the stepping motion of the heterotrimeric kinesin-2 motor in a B/B (inhibitor)-dependent fashion (Figure 2 A). Inhibitable KIF3A and KIF3B constructs are referred to as i3A and i3B, respectively.

Figure 2. Characterization of inhibitable KIF3A/KIF3B motors in ciliogenesis.

(A) Engineering strategy to generate inhibitable KIF3A/KIF3B motors. (left) The WT heterotrimeric kinesin-2 motor (KIF3A, dark blue; KIF3B, light blue; KAP3, green) moves along microtubules (gray) and transports molecular cargo (not shown). (middle) The inhibitable motor comprises KIF3A and KIF3B subunits engineered with a Dmr-B domain (orange semicircle) fused to their N-termini. This motor moves along microtubules and transports cargo similar to the WT motor. (right) Addition of the B/B homodimerizer (red dumbbell) crosslinks the motor domains and thereby inhibits processive motor motility. (B) Alignment of the N-terminal residues of kinesin-1 (KIF5C), kinesin-3 (KIF1A), and kinesin-2 (KIF3A, KIF3B) (top). The cover strand (brown arrow and text) is the most N-terminal structural element of the motor domain. Orange text indicates residues that were truncated in KIF1A to generate an inhibitable motor [25]. To generate inhibitable kinesin-2 motors, the position of the DmrB-domain relative to the cover strand was modified by truncation of 11, 12, or 13 resides of KIF3A (middle) or truncation of 5, 6, or 7 residues of KIF3B (bottom). (C-E) Representative images of (C) parental NIH-3T3 cells expressing mCherry control or (D-E) Kif3a^−/−;Kif3b^−/− cells expressing (D) WT or (E) inhibitable kinesin-2 motors [KIF3A constructs tagged with mNG (green), KIF3B constructs tagged with mCherry (red)]. The cells were treated with ethanol (vehicle control) or B/B, serum-starved for two days, then fixed and stained with DAPI (blue) and an antibody against ARL13B (magenta). Transfected cells are indicated by a white outline. Primary cilia are marked by a white arrowhead. Scale bar, 10 μm. (F) Quantification of the percentage of transfected cells that generate a primary cilium. (*) NIH-3T3 cells transfected with mCherry control. Each spot represents the mean of one independent experiment. n ≥ 57 transfected cells per condition. See also Figure S2 and S3.

As a starting point for the optimization of engineering inhibitable motors, we fused the DmrB domains directly to the N-termini of the KIF3A and KIF3B motor domains, since this DmrB fusion site had yielded inhibitable motors for kinesin-1 [15]. Transfection of the resulting DmrB-KIF3A and DmrB-KIF3B fusion constructs rescued ciliogenesis in Kif3a^−/−;Kif3b^−/− cells but the addition of the B/B inhibitor had no effect (data not shown), suggesting that these are active motors but are not inhibitable. We thus considered that engineering of the kinesin-3 motor KIF1A revealed that only one DmrB insertion site yielded a functional, inhibitable motor [15]. To find an analogous insertion site for the kinesin-2 motor subunits, we compared the N-terminal amino acid sequences of kinesin-1, −2, and −3 motors and designed comparable DmrB insertion sites for KIF3A and KIF3B (Figure 2 B). We created a series of fusion constructs in which the DmrB domain was fused to the N-terminus of KIF3A truncated by 11 (i3A^Δ11), 12 (i3A^Δ12), or 13 (i3A^Δ13) amino acids. In similar fashion, we fused the DmrB domain to the N-terminus of KIF3B truncated by five (i3B^Δ5), six (i3B^Δ6), or seven (i3B^Δ7) amino acids (Figure 2 B).

We compared the ability of each i3A construct to pair with each i3B construct and generate primary cilia in the absence but not in the presence of B/B inhibitor. Fusion of the DmrB domain to the Δ13 truncation of KIF3A (i3A^Δ13) or to the Δ7 truncation of KIF3B (i3B^Δ7) resulted in a failure to generate primary cilia (i3A^Δ13/i3B^Δ5, i3A^Δ11/i3B^Δ7, and i3A^Δ12/i3B^Δ7 in Figure 2 E,F and data not shown), indicating that engineering the i3A^Δ13 and i3B^Δ7 constructs resulted in non-functional motors. Fusion of the DmrB domain to the Δ12 truncation of KIF3A (i3A^Δ12) resulted in functional motors when paired with the i3B^Δ5 or i3B^Δ6 constructs of KIF3B as expression of i3A^Δ12/i3B^Δ5 or i3A^Δ12/i3B^Δ6 rescued ciliogenesis in Kif3a^−/−;Kif3b^−/− cells to a similar extent as expression of WT KIF3A/KF3B (i3A^Δ11/i3B^Δ5, i3A^Δ11/i3B^Δ6 and WT/WT in Figure 2 E,F). Importantly, addition of B/B inhibitor completely blocked ciliogenesis in i3A^Δ12/i3B^Δ6-expressing cells and to a lesser extent in i3A^Δ12/i3B^Δ5-expressing cells (Figure 2 E,F). Fusion of the DmrB domain to the Δ11 truncation of KIF3A (i3A^Δ11) also rescued ciliogenesis when paired with the i3B^Δ5 or i3B^Δ6 constructs of KIF3B, albeit at a lower frequency than cells expressing the WT motors, and their function was blocked by B/B inhibitor (Figure 2 F). Because the i3A^Δ12/i3B^Δ6 combination was fully functional for ciliogenesis and exhibited the best inhibition, this combination was utilized throughout the remainder of this work to selectively block heterotrimeric kinesin-2 function. The inhibition of i3A^Δ12/i3B^Δ6 was sensitive to B/B inhibitor concentration in dose-responsive manner (Figure S3).

We next tested whether primary cilia generated upon expression of inhibitable i3A^Δ12/i3B^Δ6 motors are competent to mediate Hedgehog (HH) signaling. We first tested HH pathway activation in response to Smoothened agonist (SAG) using a reporter construct in which luciferase expression is driven by a HH-responsive promoter [18]. Addition of SAG resulted in a ~4-fold induction in luciferase activity for Kif3a^−/−;Kif3b^−/− cells expressing WT KIF3A/KIF3B or the inhibitable i3A^Δ12/i3B^Δ6 motor (Figure 3 B), indicating that the inhibitable motor generates fully HH-responsive cilia. Addition of the B/B inhibitor diminished the HH response to SAG treatment in Kif3a^−/−;Kif3b^−/− cells expressing the i3A^Δ12/i3B^Δ6 motor but not in cells expressing the WT KIF3A/KIF3B motor (Figure 3 B). As a second approach to test the HH response in cells expressing the inhibitable kinesin-2, we used a constitutively-active version of the HH transcription factor GLI2 (GLI2ΔN) [19]. Previous work suggested that expression of GLI2ΔN ectopically activates HH target gene expression, and that a loss of primary cilia can result in greater pathway activation in this context [20]. Expression of GLI2ΔN in Kif3a^−/−;Kif3b^−/− cells expressing WT KIF3A/KIF3B or the inhibitable i3A^Δ12/i3B^Δ6 motor resulted in ~200-fold luciferase induction (Figure 3 C). Addition of B/B inhibitor resulted in a greater increase in luciferase activity induction only in cells expressing the i3A^Δ12/i3B^Δ6 motor, reflecting pathway hyperactivation only in cells that lack a functional primary cilium (Figure 3 C).

Figure 3. Inhibitable kinesin-2 motors generate hedgehog (HH)-responsive cilia.

(A-C) HH signaling assays. (A) Schematic of the experimental setup. Cells were seeded on day 1 and transfected on day 2 with plasmids for expressing a HH-responsive luciferase plasmid and a control beta-galactosidase plasmid together with (B) WT or inhibitable KIF3A/KIF3B or (C) GLI2ΔN and WT or inhibitable KIF3A/KIF3B. The following day, the cells were serum-starved and treated with Sonic HH agonist (SAG) and/or B/B inhibitor. Two days later, luciferase activity was measured and normalized to beta-galactosidase activity. (B) Quantification of HH signaling in response to SAG. No significant difference (n.s.) was observed in luciferase induction between cells expressing WT or inhibitable motors. Cells expressing inhibitable KIF3A/KIF3B and treated with B/B inhibitor showed a significant decrease in response to SAG (***p = 0.0001, Sidak's multiple comparisons test) as compared to cells expressing WT KIF3A/KIF3B and treated with SAG. (C) Quantification of HH signaling in response to GLI2ΔN expression. No significant difference (n.s.) was observed in luciferase induction between cells expressing WT or inhibitable motors. Cells expressing inhibitable KIF3A/KIF3B and treated with B/B inhibitor showed a significant increase in GLI2ΔN-induced luciferase expression (****p < 0.0001, Sidak's multiple comparisons test) as compared to cells expressing WT KIF3A/KIF3B. Error bars indicate SEM.

In summary, we find that expression of the i3A^Δ12 and i3B^Δ6 constructs in Kif3a^−/−;Kif3b^−/− cells results in a bona fide inhibitable kinesin-2 motor. The engineered i3A^Δ12/i3B^Δ6 motor is referred to as i3A/i3B throughout the rest of the manuscript.

Inhibition of KIF3A/KIF3B results in the loss of IFT motors and particles from the primary cilium

The generation of an inhibitable kinesin-2 motor enables us, for the first time, to directly examine the role of KIF3A/KIF3B/KAP motors during IFT in fully-formed cilia. To investigate this, Kif3a^−/−;Kif3b^−/− cells were transfected with plasmids for expressing the inhibitable i3A/i3B motor together with a mNeonGreen (mNG)-tagged subunit of the IFT-B complex (IFT88-mNG). Analysis of kymographs generated from live-cell imaging experiments revealed that IFT88-marked IFT trains moved processively towards the tip of the cilium, paused for variable durations, and then trafficked back towards the base (Figure 4 A), similar to what has been observed previously [21, 22]. Anterograde and retrograde IFT particles driven by i3A/i3B motors moved with speeds on the order of 0.7 μm/s, consistent with previously measured IFT velocities in mammalian cilia [22, 23]. To determine the impact of inhibition of KIF3A/KIF3B function on IFT, cilia expressing the inhibitable motor and IFT88-mNG were imaged and then B/B inhibitor, or ethanol vehicle as a control, was added and imaging of the same cilium was resumed. IFT motion persisted after treatment with the vehicle (ethanol), however, no IFT motion was observed in the anterograde or retrograde direction within five minutes of treatment with the B/B inhibitor (Figure 4 A). In the example shown in Figure S4, three IFT trains were observed to move within the cilium before all transport ceased and no more IFT trains entered the cilium within two minutes of B/B inhibitor addition.

Figure 4. Inhibition of heterotrimeric kinesin-2 stops IFT and results in a complete loss of IFT trains and motors from cilia.

(A-C) Imaging of the IFT protein IFT88. A) Live-cell imaging of Kif3a^−/−;Kif3b^−/− cells expressing inhibitable kinesin-2 (i3A^Δ12 and i3B^Δ6) motors and IFT88-mNG. An image of a representative cilium (left) and corresponding kymographs before treatment (middle) and after addition of ethanol vehicle (EtOH, top right) or B/B inhibitor (B/B, bottom right) are shown. Imaging frequency, 5 Hz; scale bar, 2 μm. (B,C) Kif3a^−/−;Kif3b^−/− cells transfected with inhibitable kinesin-2 (i3A^Δ12 and i3B^Δ6) motors were treated with ethanol vehicle or B/B inhibitor for the indicated times and then fixed and stained with antibodies to IFT88 (gray), the basal body marker glutamylated tubulin (not shown), and the ciliary marker ARL13B (red). (B) Representative images with the fluorescence signals offset by 6 pixels for clarity. “b” indicates the ciliary base. (C) Quantification of the average fluorescence intensity of IFT88 in the ciliary shaft (left) and at the ciliary base (right). (D,E) Localization of the kinesin-2 subunit KIF3A. Kif3a^−/−;Kif3b^−/− cells transfected with inhibitable kinesin-2 (i3AΔ12-mNG and i3BΔ6) motors were treated with ethanol vehicle or B/B inhibitor for the indicated times, and then fixed and stained with antibodies to glutamylated tubulin (not shown) and the ciliary marker ARL13B (red). (D) Representative images with the fluorescence signals offset by 6 pixels for clarity. “b” indicates the ciliary base. (E) Quantification of the average fluorescence intensity of i3AΔ12-mNG in the ciliary shaft (left) and at the ciliary base (right). In (C) and (E), colored circles indicate the average fluorescence intensity per cell and gray bars show the average fluorescence intensity of all cells in that condition over two independent experiments. * p = 0.0198, *** p = 0.0001, and **** p < 0.0001, Dunn's test. Scale bars, 2 μm. See also Figure S4.

To examine the effects of kinesin-2 inhibition across a large number of cells, we fixed and stained cells at various time points after B/B treatment. A significant loss of IFT88 from the shaft of the primary cilium was observed within five minutes of B/B treatment and no further loss of IFT88 signal in the cilium was detected after 10 minutes of treatment (Figure 4 B,C). Thus, blocking kinesin-2 function leads to a complete absence of IFT88-marked IFT trains within the primary cilium. In contrast, we observed an increase in IFT88 signal at the base of the cilium within the first two minutes after B/B treatment. However, this increase was not statistically significant and was transient as the IFT88 signal dropped below baseline conditions with longer B/B treatment (Figure 4 B,C).

To test whether inhibited KIF3A/KIF3B/KAP motors are immobilized within the cilium or removed from the cilium, we repeated the time course of B/B treatment with Kif3a^−/−;Kif3b^−/− cells expressing a mNG-tagged version of i3A/i3B motor itself (i3A-mNG). A significant loss in i3A-mNG signal in the cilium shaft was detected within 2 min of B/B treatment however, no change was observed in the amount of i3A/mNG motors at the base of the cilium (Figure 4 D,E). Taken together, these data show that all IFT motion in cells expressing the i3A/i3B motor is blocked within two minutes of B/B inhibitor addition, leading to a subsequent loss of both the IFT motor and particles from the primary cilium.

KIF3A/KIF3B-mediated IFT is essential for the maintenance of ciliary structure

How does an acute block in IFT affect the stability of the primary cilium? To address this question, cells expressing the i3A/i3B motor were allowed to generate a primary cilium and then the effects of B/B addition on cilium maintenance were quantitatively assessed (Figure 5 A,B). Treatment with B/B resulted in a decrease in the percentage of cells with a primary cilium as early as one hour after B/B treatment (63% ciliated vs 83% in vehicle-treated) (Figure 5 B). The percentage of ciliated cells continued to decline over time until only 2% of cells expressing the i3A/i3B motor retained a primary cilium after eight hours of B/B treatment (Figure 5 B). To begin to understand the mechanism of ciliary disassembly, we quantified the cilium length at each time point after B/B treatment. We found that cilia remaining after B/B treatment displayed only a modest decrease in cilium length, with the average length decreasing from 2.6 μm in vehicle-treated cells (ethanol) to 1.7 μm six hours after B/B treatment (Figure S5 A). Based on the rapid reduction in ciliated cells and the slow decrease in the length of the remaining cilia, we suggest that there are two processes that contribute to ciliary disassembly: a slow cilium shrinking, presumably from the tip, and a rapid cilium loss that occurs with stochastic onset (Figure 5 E).

Figure 5. Inhibition of heterotrimeric kinesin-2 results in the loss of cilia.

(A-B) Inhibition of heterotrimeric kinesin-2. (A) Schematic of experimental setup. Kif3a^−/−;Kif3b^−/− cells were serum-starved and transfected with inhibitable kinesin-2 (i3A^Δ12 and i3B^Δ6) motors. The cells were treated with ethanol (control) or B/B inhibitor at the indicated time points and then fixed and stained with antibodies against the axonemal marker acetylated α-tubulin and the ciliary membrane marker ARL13B. (B) Quantification of the percentage of cells that generate a primary cilium after treatment with B/B inhibitor for the indicated times. Each spot indicates the mean of one independent experiment. Error bars, SEM. ** p = 0.002; *** p < 0.0004; **** p < 0.0001 as determined by post hoc Sidak's test. (C-D) Inhibition of cytoplasmic dynein. (C) Schematic of experimental setup. NIH-3T3 cells were treated with ethanol (control) or ciliobrevin (CB) at the indicated time points. (D) Quantification of the percentage of cells with a primary cilium after treatment with Ciliobrevin D for the indicated times. ** p = 0.0016, Dunnett's test. (E) Model for cilium disassembly. In response to a block in anterograde IFT, cilia undergo a slow and steady shortening, presumably from the tip. At the same time, a rapid and active disassembly process is stochastically activated, resulting in deciliation. See also Figures S2 and S5.

To compare cilium maintenance in response to inhibition of kinesin-2 versus dynein-2 motors, we treated cells with Ciliobrevin D, an inhibitor of the retrograde IFT motor dynein-2 [24] (Figure 5 C,D). Little to no change in the percentage of ciliated cells was determined after eight hours of Ciliobrevin D treatment and only after 24 hours of Ciliobrevin D treatment was a significant decrease in ciliated cells observed (43% ciliated versus 79% in the ethanol control) (Figure 5 D). In the same time frame, the cilia underwent a slow and continuous decrease in length, with the average length decreasing from 2.8 μm in ethanol-treated cells to 1.9 μm in Ciliobrevin D-treated cells (Figure S4 B). Taken together, these results indicate that a continual supply of ciliary components by anterograde IFT is critical for the maintenance of primary cilia whereas retrograde IFT may play an indirect role. Furthermore, only the acute block to anterograde IFT activates the rapid cilium disassembly pathway.

The homodimeric kinesin-2 KIF17 cannot function in ciliogenesis or cilium maintenance

In addition to KIF3A/KIF3B/KAP, two other motor proteins of the kinesin-2 family, heterotrimeric KIF3A/KIF3C/KAP and homodimeric KIF17, are implicated in driving transport events within mammalian cilia [25-27]. We thus probed an involvement of these motors in ciliogenesis by ectopically expressing either KIF3A/KIF3C or KIF17 in Kif3a^−/−;Kif3b^−/− cells. We found that Kif3a^−/−;Kif3b^−/− cells expressing either motor were unable to generate cilia (Figure 7 A), demonstrating that KIF3A/KIF3B/KAP is the essential motor for the generation of primary cilia in NIH-3T3 cells.

Figure 7. Utilization of kinesin-2 motors for IFT and ciliogenesis differs across species.

Simplified cartoon representations of IFT in (A) mammalian, (B) Chlamydomonas, and (C) C. elegans amphid cilia. (A) At the base of the cilium, heterotrimeric KIF3A/KIF3B/KAP motors bind to preassembled IFT trains for transport along axonemal microtubules. Ciliary proteins and inactive KIF17 and dynein-2 motor proteins are cargoes of the IFT trains. Ciliary membrane proteins track intermittently with IFT. KIF17 cannot drive IFT but could transport other ciliary proteins (e.g. signaling proteins) independently or via its binding to IFT trains. At the tip of the cilium, cargo and motor proteins are unloaded and the IFT trains are transported back to the base of the cilium by cytoplasmic dynein-2. The kinesin-2 motors are returned to the base of the cilium partly by diffusion and in part as passive cargo of IFT trains. (B) In Chlamydomonas, which lacks a KIF17 homolog, IFT trains are transported in anterograde direction exclusively by heterotrimeric kinesin-2. This motor is returned to the base of the cilium by diffusion. (C) In the amphid and phasmid cilia of C. elegans, heterotrimeric kinesin-2 transports IFT trains through the transition zone but detaches from IFT trains in the proximal cilium and is successively replaced by the KIF17 homolog, OSM-3, which transports the IFT trains along the distal axoneme. Both kinesin-2 and OSM-3 are returned to the base of the cilium as inactive cargo on dynein-driven IFT trains.

The role of the homodimeric KIF17 motor in ciliary assembly and function has been perplexing as KIF17 binds to IFT-B particles and co-migrates with IFt in mammalian cilia [28, 29] yet genetic ablation of KIF17 expression in vertebrates has no measurable effect on ciliary assembly or function [9, 10, 26, 30]. Given the division of labor between the kinesin-2 homologs in C. elegans [7, 31], we reasoned that KIF17 may be dispensable for ciliogenesis in mammalian cells but could take over for IFT transport during cilia maintenance. To test this possibility, we co-expressed i3A/i3B and KIF17 at comparable levels in Kif3a^−/−;Kif3b^−/− cells (Figure S2 C,D). After cells were fully ciliated, the B/B inhibitor was added to block KIF3A/KIF3B/KAP function and the percentage of ciliated cells was quantified. Across a six-hour time frame, cells expressing the KIF17 motor displayed an identical loss of primary cilia (Figure 7 B) and decrease in ciliary length (Figure S6) as cells expressing only the i3A/i3B motor. The inability of KIF17 to support IFT upon inhibition of KIF3A/KIF3B/KAP function is not caused by a general blockage to transport along axonemal microtubules by the inhibited i3A/i3B motors as the inhibited motors were depleted from the cilium upon B/B treatment (Figure 5 A,B). These results suggest that despite its to bind to IFT proteins and co-traffic with IFT markers, KIF17 does not drive IFT in primary cilia of mammalian cells.

DISCUSSION

Delineating cellular processes with inhibitable KIF3A/KIF3B/KAP motors

We engineered an inhibitable version of the heterotrimeric KIF3A/KIF3B/KAP kinesin-2 motor that allows, for the first time, specific inhibition of this motor and delineation of its role in IFT in primary cilia. The successful engineering of inhibitable motors from the kinesin-1, kinesin-3 [15], and now kinesin-2 families corroborates the notion that our chemical-genetic approach can be used to generate inhibitable motors across the kinesin family, and likely other motor proteins of interest, without the need for labor-intensive small molecule screens. The inhibition of the engineered motors is highly specific as it relies on the well-characterized B/B homodimerizer, which has no endogenous cellular targets, and thus avoids off-target effects. The inhibition is also rapid, occurring on the timescale of the motor’s transport functions; this is in contrast to inhibition of transport function for a temperature-sensitive allele of kinesin-2 (fla10-1^ts) in Chlamydomonas where a full block to IFT is not observed until ~60 min after shifting cells to the non-permissive temperature [11].

The engineered i3A/i3B/KAP motor is fully functional as it restores ciliogenesis in Kif3a^−/−;Kif3b^−/− cells and generates HH-responsive cilia, an important criterion that has not been demonstrated for engineered motors generated by other chemical-genetic approaches [32]. An inability to fully replicate wild-type protein function has also been noted for the fla10-1^ts Chlamydomonas temperature-sensitive allele as fewer IFT trains are observed in cilia grown at the permissive temperature than in the WT strain [33]. While we have tested our i3A/i3B motor extensively in Kif3a^−/−;Kif3b^−/− cells, future work will show whether this motor indeed fully complements the WT protein function in an i3A/i3B mouse model. Importantly, fusion of DmrB domains to both motor-containing subunits, KIF3A and KIF3B, has an added advantage in that it allows a separation of the functions of KIF3A/KIF3B/KAP from those of KIF3A/KIF3C/KAP.

Inhibition of KIF3A/KIF3B/KAP results in a complete loss of IFT

We observed a rapid block to IFT upon inhibition of the i3A/i3B motor in mammalian primary cilia. In the absence of B/B inhibitor, IFT88 moved processively along the length of the cilium in anterograde and retrograde directions, whereas addition of the B/B inhibitor resulted in a rapid block to IFT (within two minutes). In the example in Figure S4, three IFT trains can be observed moving in the anterograde direction after the addition of the inhibitor. It seems likely that these trains are driven by i3A/i3B motors that have not yet been inhibited by B/B inhibitor molecules diffusing to their target.

The most dramatic effect of i3A/i3B inhibition is a rapid block to new IFT trains entering the cilium. This result suggests that B/B inhibitor crosslinks and stops i3A/i3B motors once they are activated by IFT cargo binding at the base of the cilium. An alternative possibility is that B/B inhibitor crosslinks the motor domains of soluble and autoinhibited i3A/i3B motors, thereby depleting the pool of active heterotrimeric kinesin-2 that can be recruited to IFT trains at the base of the cilium. These possibilities are not mutually exclusive and further work will examine the ability of B/B inhibitor to block actively transporting kinesin-2 motors and/or deplete the soluble pool of autoinhibited motors.

The inhibition of anterograde transport resulted in a transient increase in IFT88, but no change in i3A motor, located at the base of the cilium. These results indicate that IFT proteins and kinesin-2 motors are recruited separately to the base of the cilium. This is consistent with recent work in Chlamydomonas demonstrating that IFT trains are assembled at the basal body and kinesin-2 is recruited to the train just before it is injected into the cilium [34]. These results also indicate that the number of IFT and motor binding sites at the base of the cilium is limited, again consistent with the recent work in Chlamydomonas [34]. After the transient increase, IFT88 levels at the base dropped and plateaued at around 70% of the initial intensity values. One possible explanation for this observation is that there is a role for kinesin-2 in actively tethering IFT trains to the base of the cilium before their injection into the cilium. Alternatively, there may be a feedback mechanism that regulates the amount of IFT protein at the base in response to a cue reporting the status of the cilium [35-37]. A third possibility is that the decrease in IFT88 at the base of the cilium is due to kinesin-2-dependent maintenance of basal body organization or plasma membrane docking [30, 38]. Further studies are needed to address these possibilities.

Implications for mechanisms of cilium disassembly

After the block in IFT in response to kinesin-2 inhibition, there was a rapid decrease in the number of ciliated cells, with a nearly complete loss of cilia by eight hours of kinesin-2 inhibition. Surprisingly, the average length of the remaining cilia decreased only slowly and moderately over the time course of kinesin-2 inhibition. Interestingly, similar changes were observed in Chlamydomonas strains expressing temperature-sensitive alleles of KIF3A (fla10-1^ts) or KAP (fla3^ts) where a shift to the non-permissive temperature resulted in stochastic onset of an abrupt shedding of cilia in the majority of cells and a slow and continuous decrease in cilium length in the remaining cells [39, 40]. We thus suggest that there are two superimposed processes that lead to the loss of cilia in response to a block in anterograde transport. First, there is a slow and continuous shortening of the cilium caused by either a decrease in IFT-mediated tubulin delivery to the cilium tip, according to the balance-point model [41], and/or continued shedding of the tip of the cilium, which has been reported to occur for ciliary disassembly in response to mitogenic signals [42].

Second, there is a rapid and active cilium disassembly process that is stochastically activated upon loss of anterograde IFT. It is possible that the block to anterograde IFT leads to the activation of microtubule-depolymerizing kinesins which would cause rapid shortening of the axoneme and resorption of the cilium [43-45]. An alternative possibility is that the block to anterograde IFT activates microtubule severing enzymes such as katanin which have been demonstrated to drive deciliation in Chlamydomonas [46]. In either case, it will be interesting to investigate whether loss of anterograde IFT activates signaling pathways involved in disassembly of the primary cilium during re-entry of cells into the cell cycle (reviewed in [47]).

The rapid cilium loss in response to kinesin-2 inhibition contrasts with the effects of inhibiting dynein function. Treatment of ciliated cells with the dynein inhibitor Ciliobrevin D causes both anterograde and retrograde IFT to stop within ~5 minutes [21], however, we observed only the slow and steady shortening of cilia for 12 hours after Ciliobrevin D treatment and even after 24 hours of treatment, more than 50% of the cells still had a primary cilium, consistent with other reports [24]. This phenotype is similar to that observed in Chlamydomonas mutants with a temperature-sensitive defect in the dynein-2 motor upon a shift to the non-permissive temperature [48, 49]. It is presently unclear why anterograde transport is abolished by both ciliobrevin inhibition of dynein and B/B inhibition of i3A/i3B and yet rapid cilium disassembly is observed only in the latter scenario. It is possible that ciliobrevin does not provide a complete block to IFT and the residual transport is sufficient for cilium maintenance. These discrepancies highlight the power of our approach to rapidly and specifically inhibit motor protein function.

Utilization of kinesin-2 motors to drive IFT varies across organisms

We propose a model for how IFT is accomplished in mammalian cilia as compared to studies in two model organisms, the nematode C. elegans and the green algae Chlamydomonas. In the ciliated endings of sensory neurons in the amphid and phasmid structures of C. elegans, IFT trains are imported into the cilium and carried along the proximal axoneme by heterotrimeric kinesin-2 (KLP11/KLP20/KAP) at a slow rate and then handed over to the homodimeric motor OSM-3 which solely transports the IFT trains along the distal axoneme at a fast rate ([31], Figure 7 C). Genetic and imaging studies have shown that loss of heterotrimeric kinesin-2 motor function has only minor effects on cilium structure as OSM-3 is capable of driving IFT and generating cilia in these sensory structures [7]. These findings provide the basis of current models in the field that heterotrimeric kinesin-2 is a slow and weak motor and its role is to aid in the navigation of IFT particles through the transition zone [6].

Chlamydomonas, however, does not express a homodimeric kinesin-2 [8], making heterotrimeric kinesin-2 (FLA10/FLA8/KAP) the sole and essential motor for IFT in this organism (Figure 7 B). Our data demonstrate a similar requirement for the heterotrimeric KIF3A/KIF3B/KAP kinesin-2 as the sole and essential motor for the import and transport of IFT trains along the axoneme in mammalian primary cilia (Figure 7A). We demonstrate that expression of neither the heterotrimeric motor KIF3A/KIF3C/KAP nor the homodimeric motor KIF17 was able to rescue ciliogenesis, consistent with the fact that loss of KIF3C or KIF17 expression has minimal phenotypes in mouse models [9, 10, 50]. We suggest that in mammals, the homodimeric KIF17 motor may play an accessory role in either transporting ciliary cargoes in an IFT-independent manner and/or as an adaptor to link specific ciliary cargoes to the IFT particles ([25, 51], Figure 7 A). Our findings call into question a theoretical model that KIF3A/KIF3B/KAP and KIF17 take turns in transporting IFT trains along the axoneme [27]. It is worth noting that even within an organism, tissue-specific adaptation of the kinesin-2-IFT machinery may occur, for example, in the cephalic male (CEM) neurons of C. elegans [52] and some ciliated cells in zebrafish [26].

Utilization of the IFT machinery also varies across species with respect to the recycling of kinesin-2 motors after delivery of IFT particles to the tip of the cilium. In C .elegans, the kinesin-2 motors have been observed to undergo retrograde IFT and to accumulate at the ciliary tips in dynein-2 mutants [53], suggesting an active process that drives recycling of kinesin-2 motors and IFT particles (Figure 7C). In contrast, in Chlamydomonas cilia, return of kinesin-2 motors to the base of the cilium does not rely on dynein but rather occurs by diffusion [35, 48, 54] (Figure 7B). . In mammalian cilia, kinesin-2 motors have been observed to undergo both retrograde transport and free diffusion back to the cell body [28, 55] (Figure 7A). Further work is required to identify the molecular mechanisms that mediate these species-specific differences.

Outlook

Future work with the i3A/i3B motor will allow us to directly determine the ciliary cargoes that depend on KIF3A/KIF3B/KAP activity for import into cilia and/or transport along the axoneme. We will also be able to directly test the roles of kinesin-2 in the steps that lead to ciliogenesis, such as the assembly of the subdistal appendages and the positioning of the basal body [47]. Furthermore, the i3A/i3B motor will be a valuable tool to dissect other dynamic cellular processes in which KIF3A/KIF3B/KAP has been implicated, such as axon specification in neurons [56].

STAR METHODS

Contact for Reagent and Resource Sharing

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Kristen Verhey (kjverhey@umich.edu). All plasmids encoding an mNeonGreen fusion protein are restricted from distribution through a MTA with Allele Biotechnology.

Experimental Model and Subject Details

NIH-3T3 Flp-In (female Mus musculus embryonic fibroblast) cells and cell lines derived in this study were cultured in D-MEM (Gibco) with 10% Fetal Clone III (HyClone) and GlutaMAX (Gibco) at 37°C and 5% CO₂. Flp-In cells were purchased from Thermo Fisher Scientific, have not been authenticated, and are regularly tested for mycoplasma contamination.

Method Details

All experiments were carried out in 2-7 independent replicates as indicated in figure legends. No randomization, stratification, blinding, or estimations of sample-size were performed in this study. Cells expressing high levels of exogenous proteins were excluded from analysis.

Plasmids and oligonucleotides.

For CRISPR/Cas9-mediated genome-editing, the plasmid eSpCas9(1.1) was a gift from Feng Zhang ([57], Addgene plasmid #71814). sgRNAs targeting a site in the third exon of the Mm Kif3a gene were generated by annealing the primers 5’-CACCGAAGGCGTTCGAGCAGTACCG-3’ and 5’-AAACCGGTACTGCTCGAACGCCTTC-3’. sgRNAs targeting a site in the second exon of the Mm Kif3b gene were generated by annealing the primers 5’-CACCGGTGAAAACCACATCCGGGT-3’ and 5’-AAACACCCGGATGTGGTTTTCACC-3’. The resulting DNA pieces were subcloned via BbsI restriction digestion into eSpCas9(1.1). The plasmid PGK-puro was a generous gift from Dr. Thom Saunders (University of Michigan). Plasmids encoding mCitrine and HsKIF17-mCit [17], mCherry (mChe, Clontech Laboratories), and mNeonGreen (mNG) [58] have been described previously. To generate plasmids encoding full length MmKIF3A-mNG and MmKIF3B-mChe, the open reading frames (ORF) of KIF3A and KIF3B were synthesized (Life Technologies) and cloned in frame via NheI and BamHI into pmNeonGreen-N1 and pmCherry-N1, respectively. The plasmid encoding full length MmKIF3C-mChe was cloned by amplifying the ORF of KIF3C with the primers 5’-GCGTATACTAGTGCCACCATGGCCAGTAAGACCAAGGCCAG-3’ and 5’-CGCATAGGATCCATGTCATGGTCTACCACTGTTGCAGGG-3’ from a plasmid that was a gift from L. Goldstein (University of California at San Diego). The KIF3C ORF was then cloned via the compatible ends of NheI/SpeI and BamHI into pmCherry-N1. Inserts encoding DmrB-KIF3A-mNG, DmrB-Δ11KIF3A-mNG, DmrB-Δ12KIF3A-mNG, DmrB-Δ13KIF3A-mNG, DmrB-KIF3B-mChe, DmrB-Δ5KIF3B-mChe, DmrB-Δ6KIF3B-mChe, and DmrB-Δ7KIF3B-mChe were generated by Splice by Overlap Extension PCR [59, 60] and cloned into pmNeonGreen-N1 and pmCherry-N1, respectively. Primer sequences are found in Table S1. To generate the constructs encoding inhibitable motors without a fluorescent tag, the ORFs encoding DmrB-Δ12KIF3A and DmrB-Δ6KIF3B were cloned via NheI and AgeI into a pEGFP-N1 backbone from which the EGFP had been removed. The HH-responsive luciferase encoding plasmid ptchΔ136-GL3 [61], pSV-β-galactosidase (Promega, E1081), pCIG-GLIΔN [61] were used for luciferase assays. The plasmid IFT88-mNG was cloned by replacing mCit with mNG through restriction digestion with AgeI and BsrGI in the plasmid myc-IFT88-mCit described [62]. All constructs were verified by DNA sequencing.

Cell culture and genome engineering.

Parental NIH-3T3 Flp-In cells (Thermo Fisher Scientific; RRID: CVCL_U422) and derived Kif3a^−/−;Kif3b^−/− cells were cultured in D-MEM (Gibco) with 10% Fetal Clone III (HyClone) and GlutaMAX (Gibco) at 37°C an d 5% CO₂. Both cell lines were transfected using Lipofectamine 2000 (Life Technologies) according to the manufacturer’s protocol. Double knockout cells (Kif3a^−/−;Kif3b^−/−) were generated by sequentially engineering the Kif3a and Kif3b gene loci. NIH-3T3 Flp-In cells were co-transfected with the eSpCas9(1.1) encoding an appropriate sgRNA (see above) and PGK_puro, which confers Puromycin resistance. One day later, 10 μ9-ml Puromycin (Sigma, P8833) was added to the culture medium to select for transfected cells. Three days later, single cells were sorted into separate wells of a 96-well plate (Flow Cytometry Core, University of Michigan) and expanded. To identify cell clones with altered KIF3A or KIF3B alleles, the DNA of cells not able to generate cilia in response to serum-starvation was isolated using the DNeasy Blood & Tissue Kit (Qiagen) and the DNA region of interest was amplified by PCR using the primers listed in Table S1. The PCR product was cloned (TOPO TA Cloning Kit; Invitrogen) and transformed into competent TOP10 bacteria (Thermo Fisher Scientific). Plasmids were isolated (GeneJET Plasmid Miniprep Kit, Thermo Fisher Scientific) and sequenced using the primer 5′-CAGGAAACAGCTATGACC-3′.

Immunohistochemistry, microscopy, and western blotting.

Cells were fixed with 4% formaldehyde in PBS, treated with 50 mM NH₄Cl in PBS to quench unreacted formaldehyde and permeabilized with 0.2% Triton X-100 in PBS. Subsequently, cells were blocked in blocking solution (0.2% fish skin gelatin in PBS). Primary and secondary antibodies were applied in blocking solution at room temperature for 1 h each. Cells were incubated 3x for 5 min in blocking solution to remove unbound antibodies. Antibodies used: polyclonal antibodies reacting with ARL13B (1:1000, Protein Tech Group, 17711-1-AP), IFT88 (1:500, Protein Tech Group, 13967-1-AP), and monoclonal antibodies reacting with ARL13B (1:200, NeuroMAB, 73-287), acetylated tubulin (1:10.000, Sigma, T6793), KIF3A (K2.4, 1:300, Abcam, ab24626), and polyglutamylated tubulin (1:1000, Adipogen Life Sciences, GT335). Nuclei were stained with 10.9 μM 4′,6-diamidino-2-phenylindole (DAPI) and cover glasses were mounted in ProlongGold (Life Technologies). Only cells expressing low to medium levels of exogenous motor were selected for imaging and quantification (Figure S2 C,D). Images were collected on an inverted epifluorescence microscope (Nikon TE2000E) equipped with a 60x, 1.40 numerical aperture (NA) oil-immersion objective and a 1.5x tube lens on a Photometrics CoolSnapHQ camera driven by NIS-Elements (Nikon) software. Images of KIF3A-mNG in cilia were captured by a Nikon A1 confocal system with 60× oil immersion objective (NA 1.40) on a high sensitivity GaAsP detector. To measure IFT88-mNG or i3A-mNG content of cilia, a region of interest (ROI) was drawn around the cilium shaft (identified by ARL13B) and the base of the cilium (identified by glutamylated tubulin) in ImageJ (NIH) and the average fluorescence intensity in the cilium shaft and at the cilium base was measured and the population average was calculated using Excel. To determine the relative levels of expression of endogenous and exogenous kinesin-2 motors, cells were washed with PBS, fixed in ice-cold methanol, and stained with the monoclonal antibody to KIF3A. The fluorescence intensities were determined on a cell-by-cell basis by generating a 20-pixel-wide (~2 μm), doughnut-shaped ROI with the software CellProfiler (Version 2.2.0) and measuring the average fluorescence intensity for both exogenous motor (i3A-mNG or KIF17-mNG) and endogenous KIF3A. The presence of KIF3A in WT NIH-3T3 and Kifa^−/−;Kif3b^−/− cells was determined by western blot analyses on Amersham Hybond P0.45 PVDF membranes using monoclonal antibodies to KIF3A (K2.4, 1:2.000, Abcam, ab24626) and β-tubulin (1:5.000 clone E7; Developmental Studies Hybridoma Bank).

Ciliogenesis rescue and disassembly assays.

To assess the effect of B/B inhibitor on ciliogenesis, cells were seeded on cover glasses and twelve hours later, the culture medium was switched to 1% Fetal Clone III (serum-starvation). Vehicle (0.1% ethanol final) or 50 nM B/B homodimerizer (or as indicated in Figure S1) (Clonetech, 635060) was added followed by the transfection complexes. Two days later the cells were fixed and stained.

To assess the effect of B/B inhibitor or Ciliobrevin D on fully-formed cilia, cells were seeded on cover glasses and serum-starved and transfected twelve hours later. Two days later, vehicle (0.1% ethanol final) or 50 nM B/B homodimerizer or 30 μM Ciliobrevin D (Sigma, 250401) was added for the indicated times and cells were fixed and stained. Image analysis was performed using ImageJ (NIH). Cells expressing mCherry, or mCherry- and mNeonGreen-tagged inhibitable motors were selected and analyzed for the presence of a cilium, as judged by an ARL13B-positive filament that was ≥ 10 pixel (≥1.1 μm). We noted that cells that express low levels of the KIF3B and KIF3A constructs had the highest probability to generate cilia. Thus cells expressing high levels of transfected proteins were excluded from analysis (Figure S2 C-D).

HH-responsive luciferase assay.

Cells were seeded onto gelatin-coated 24-well plates (5 × 10⁴ cells/well). Twelve hours later, cells were co-transfected with WT or i3A/i3B constructs, a luciferase reporter (ptchΔ136-GL3), and β-galactosidase control (pSV-β-galactosidase), or were co-transfected with i3A/i3B constructs and pCIG-GLIΔN, a luciferase reporter (ptchΔ136-GL3), and β-galactosidase control (pSV-β-galactosidase). Two days after transfection, cells were briefly washed with PBS and transferred to culture medium containing 1% Fetal Clone III, 1% penicillin/streptomycin and 50 nM B/B inhibitor and 500 nM SAG (Enzo Life Sciences, ALX-270-426-M001), as indicated. Luciferase and β-galactosidase activities were measured after two days of serum starvation using a Spectra-max M5e plate reader (Molecular Devices) and the luciferase assay kit (Promega, E1501) and β-galactosidase assay kit (EMD Millipore, 70979). To calculate the level of HH pathway activity luciferase values were divided by the β-galactosidase values, and data were reported as fold induction relative to either vector transfected or untreated values. All treatments were performed in triplicate.

Live-cell imaging.

Cells were seeded into glass bottom dishes (MatTek Corporation, P35G-1.5-14-C) and 12 hours later switched to culture medium containing 1% serum and transfected. Two days later, medium was changed to Leibovitz’s L-15 Medium (Gibco, 21083-027) and imaged on a Nikon Ti-E/B microscope equipped with a 100x, 1.49 NA oil-immersion TIRF objective, 1.5x tube lens, three 20mW diode lasers (488, 561, 640 nm) controlled via acousto-optical tunable filter (AOTF) (Agilent) and EMCCD detector (iXon X3 DU897, Andor). Time-lapse images were acquired with 488 nm excitation, 200 ms exposure at 5 or 2.5 frames per second in a humidified imaging chamber (Tokai Hit) at 37 °C. Kymographs were generated using the Multiple K ymograph plugin (J. Rietdorf and A. Seitz) for ImageJ (NIH).

Quantification and Statistical Analysis

Statistical tests were chosen based on the distribution of the data and applied using the software GraphPad Prism (Version 7, GraphPad Software). No other methods were used to determine whether the data met the assumptions of the statistical approach. The statistical significance between different treatments was first tested via one-way ANOVAs [Figure 3B-C (p < 0.0001), Figure 5B (p.0.0001), and Figure 5D (p = 0.0053)] or Kruskal-Wallis nonparametric ANOVAs [Figure 4C (shaft and base, p < 0.0001), Figure 4E (shaft, p < 0. 0001 and base, p = 0.4449)]. For Figure 6B, a two-way ANOVA found a significant effect of treatment time (p<0.0001) but no effect of the expressed protein (mCit vs. KIF17) and no interaction between the treatment time and the expressed protein. Post hoc statistical tests were then applied as described in each figure legend.

Figure 6. Neither the heterotrimeric KIF3A/KIF3C/KAP motor nor the homodimeric KIF17 can substitute for KIF3A/KIF3B/KAP in mammalian cilia.

(A) Test for rescue of ciliogenesis. Kif3a^−/−;Kif3b^−/− cells were transfected with KIF17, KIF3A and KIF3B (3A/3B), or KIF3A and KIF3C (3A/3C) and serum-starved to allow for ciliogenesis. The cells were fixed and stained with an antibody to the ciliary membrane marker ARL13B and the percentage of transfected cells with a primary cilium was quantified. (B) Test for rescue of cilia maintenance. Kif3a^−/−;Kif3b^−/− cells were co-transfected with inhibitable i3A^Δ12/i3B^Δ6 and KIF17 motors and serum-starved. After two days, the cells were treated with ethanol control (EtOH) or B/B inhibitor for 2, 4, or 6 hours and then fixed and stained. The percentage of cells with a primary cilium was quantified using the ciliary membrane marker ARL13B. At each time point, there was no significant difference in the percentage of ciliated cells when comparing cells expressing KIF17 and the mCit control. For both conditions (mCit and KIF17), there was a significant decrease in the percentage of ciliated cells at each time point compared to the control ethanol treatment (p<0.0001, Sidak's test). Error bars, SEM. See also Figures S2 and S6.

Highlights.

• KIF3A/KIF3B/KAP is the sole intraflagellar transport motor in mammalian cells

• Engineered, inhibitable KIF3A/KIF3B motors rescue wild-type motor function

• Acute inhibition of KIF3A/KIF3B blocks anterograde IFT and leads to cilium loss

• KIF3A/KIF3B/KAP is essential for ciliogenesis and cilium maintenance in mammals