Introduction
After the loss of heterotrimeric kinesin-II function, some membrane proteins are delivered normally to ciliary membranes10 and distinct parts of axonemes can still assemble11. Therefore, in contrast to the accepted single-pathway mechanism of IFT1, 2, 3, 4, it is possible that multiple pathways deliver ciliary cargo molecules to distinct sites to build discrete parts of cilia, reminiscent of the multiple transport pathways operating in axons7, 12, 13.
Sensory9 and motile14 axonemes are differentiated longitudinally into two domains: the 'middle segment' and the 'distal segment' consisting of 9 doublet and 9 singlet microtubules, respectively (Fig. 1). To investigate whether these two domains could be built by distinct IFT pathways, we used time-lapse fluorescence microscopy to observe the motility of specifically labelled IFT motors and IFT particles within sensory cilia on chemosensory neurons of C. elegans15. Previously, we observed IFT particles moving base-to-tip at unitary rates of 0.7 
m s-1 (refs 15, 16), but with improved optics and kymography17 we now find that IFT particles move in an anterograde direction at two rates: 0.7 m s-1 along the middle segment then accelerating to 1.3 m s-1 along the distal segment (Fig. 2; Table 1; also see Supplementary Information, Fig. S1). No significant difference in retrograde transport along the two segments was observed (Vmiddle = 1.17 
0.25 m s-1, n = 49; Vdistal = 1.09 0.25 m s-1, n = 43).
Figure 1: Sensory cilia on neurons of wild-type and mutant C. elegans.
(A) During chemotaxis, environmental chemicals are detected by chemoreceptors concentrated in the membranes surrounding the non-motile axonemes of sensory cilia on chemosensory neurons9, 19, 29, 30. Their axonemes are 7.5 m long, consisting of a 1-m-long basal body, the transition zone (TZ), a 4-m-long 'middle segment' consisting of nine doublet microtubules and a 2.5-m 'distal segment' consisting of nine singlet microtubules that extend from the middle segment A subfibres. GFP–IFT proteins (green) accumulate in the TZ and move in an anterograde direction along the middle and distal segments to the tip (arrows). (B) Fluorescence micrographs and corresponding schematic representations of the distribution of IFT particles (OSM-6–GFP) along sensory cilia (asterisk, TZ; arrowhead, middle segment; arrow, distal segment) in (a) wild-type (b) kap-1-mutant, (c) osm-3-mutant (missing distal segment) and (d) kap-1; osm-3 double mutant (missing middle and distal segments) animals. Numbers = length, with n = number of animals and m = number of cilia measured. Scale bar represents 3 m.
Figure 2: Anterograde IFT along middle and distal segments of sensory cilia.
(a) Motility of OSM-6–GFP within cilia of wild-type animals. Left, fluorescence micrographs with corresponding schematic showing lines used to generate kymographs along the middle (M1–M4) and distal (D) segments. Kymographs (M1–M4 and D) and corresponding lines (M1'–M4' and D') show that motility along distal segments is faster than along middle segments. Horizontal bars represent 2.5 m; vertical bar, 5 s. (b) Model showing sequential 'middle' (large arrows) and 'distal' (small arrows) IFT-pathways. After assembly of IFT machinery and its entry into cilium (Step 1), IFT-particles are moved in an anterograde direction by kinesin-II and Osm-3–kinesin at 0.7 m s-1 along the middle segment with kinesin-II exerting drag (Step 2). Kinesin-II undergoes turnaround (Step 3), and liberated Osm-3–kinesin-bound IFT particles move at 1.3 m s-1 to the distal tip (4) where Osm-3–kinesin turnaround occurs. Recycling involves retrograde transport (steps 6,7) and disassembly (8). In this model, the cargo of kinesin-II (K) and Osm-3–kinesin (O) are axoneme-stabilizing factors. Either factor can stabilize middle segment doublets, but O is essential to stabilize distal singlets.
Biochemical fractionation and localization suggested that these cilia contain two candidate anterograde IFT motors: heterotrimeric kinesin-II (consisting of one KRP85 motor, one KRP95 motor and one accessory KAP encoded by klp-11, klp-20 and kap-1) and homodimeric Osm-3–kinesin (consisting of 2
OSM-3)8; cilia in osm-3 mutants lack their distal segments9 suggesting a role in ciliogenesis, but the function of kinesin-II in this system is unknown. When the movement of these motors along middle and distal segments was observed in vivo, Osm-3–kinesin was found to move the full length of the axoneme at the same two rates as IFT particles, but kinesin-II moved only along the middle segment and never entered the distal segment (Table 1; also see Supplementary Information, Fig. S2).
To understand the importance of these differences, we sought loss-of-function mutants defective in kinesin-II and Osm-3–kinesin function. Seven osm-3 mutant alleles have long existed, and mutants in KRP95 and KAP subunits8 were recently obtained by the C. elegans gene-knockout consortium18. We sequenced the DNA of all these mutants to deduce their molecular lesions, and used standard 'dye-filling' and osmotic-avoidance assays to assess ciliary structure and function9, 19 (Table 2). Despite two 'non-conformists' (osm-3 alleles mn391 and n1540/n1545), these assays are useful indicators of ciliary performance (although we show later that direct assays of IFT can be more informative).
As shown previously, mutants containing the severe osm-3(p802) allele that lack ciliary distal segments show defective dye-filling and osmotic avoidance9, 19, and our sequencing indicates that this allele may encode a motor subunit that lacks the stalk-tail domains and is unlikely to bind to IFT particles. In contrast, animals carrying single KRP95 and KAP mutants exhibit no dye-filling or osmotic-avoidance defects. The lesion in the KRP95 mutant klp-11 (tm324) truncates only the tail domain, which would predict a weak phenotype (but note that a similar osm-3 mutant (e1811) shows defective dye-filling and osmotic avoidance). The kap-1(ok676) mutant, however, is severely truncated lacking all its protein-interacting armadillo repeats, which suggests that it is likely to be a severe loss-of-function allele, yet unlike Drosophila KAP mutants20, it apparently has normal cilia and thus exhibits no defects in dye-filling or osmotic avoidance.
Kinesin-II is multifunctional, driving critical organelle, protein and RNA transport events outside cilia21, 22, 23 and so it is possible that kinesin-II performs non-ciliary functions in C. elegans with Osm-3–kinesin substituting as an IFT motor. Alternatively, redundancy may be involved: some single kinesin-II subunit knockouts yield no phenotype24, 25, seemingly in one case because of functional redundancy between the two motor subunits of this heterotrimeric complex25. In C. elegans, it might be possible that the kinesin-II and Osm-3–kinesin holoenzymes8 have redundant ciliary functions.
To test this, we crossed osm-3(p802) and kap-1(ok676) worms to produce double mutants with impaired kinesin-II and Osm-3–kinesin function. We then examined the distribution of the IFT particle subunit OSM-6 fused to green fluorescent protein (GFP) along the cilia of single- and double-mutant animals (Fig. 1). In KAP single mutants, IFT-particle fluorescence extends the same distance (7–8 m) from the transition zone as in wild-type animals, suggesting that in the absence of KAP function, IFT particles are transported normally all along the axoneme. In osm-3 mutants lacking distal segments9, IFT-particle fluorescence extends only 4–5 m from the transition zone, consistent with the hypothesis that in the absence of Osm-3–kinesin function, kinesin-II controls IFT-particle transport along the middle-segment remnants. In kap-1; osm-3 double mutants however, IFT-particle fluorescence does not extend along the cilia at all, but instead accumulates at the transition zone (Fig. 1), indicating that in the absence of both motors, IFT ceases and IFT particles are not moved away from the transition zone9. Thus, we propose that kinesin-II and Osm-3–kinesin have redundant roles in controlling IFT along the middle segments of the cilia, whereas Osm-3–kinesin, but not kinesin-II, controls IFT along the distal segment.
Results from IFT motility assays were consistent with this idea (Table 1; also see Supplementary Information, Fig. S1). In an Osm-3–kinesin-mutant background, IFT particles moved only along the middle segments at the characteristic slow rate, and no transport was observed along the distal segments. As expected, in the kap-1; osm-3 double mutants, no IFT-particle motility was observed. However, it was notable that in kap-1 single mutants, IFT particles moved the full length of the axoneme as in wild-type, consistent with the OSM-6–GFP distribution, but at a constant base-to-tip rate of 1.3 m s-1, similar to the faster distal-segment transport in wild-type animals.
The subtle phenotype displayed by the kap-1 single mutant, that is, accelerated IFT along the middle segment, surprisingly suggests that in wild-type animals both Osm-3–kinesin and kinesin-II move the same IFT particles along the middle-segment, with the slower-moving kinesin-II (0.5 m s-1) exerting drag on the faster-moving Osm-3–kinesin (1.3 m s-1), giving rise to the intermediate rate of transport observed (0.7 m s-1). We propose that at the middle segment tip, IFT particles reorganize, allowing kinesin-II to unload its cargo and undergo turnaround, whereas Osm-3–kinesin-bound particles are liberated to move unrestrained to the distal segment tip at the faster rate characteristic of Osm-3–kinesin. This generates two sequential IFT pathways (Fig. 2b). In the 'middle segment' pathway, kinesin-II and Osm-3–kinesin function redundantly to transport IFT particles and either motor, but not both, is dispensable for this. Then, in the 'distal segment' pathway, Osm-3–kinesin functions alone to transport IFT particles to the distal tip.
It is likely that the functional significance of having two IFT pathways is that each builds distinct parts of the ciliary axoneme. Thus, in the absence of kinesin-II, Osm-3–kinesin can assemble, and move along, a full-length axoneme; in the absence of Osm-3–kinesin the distal segment is missing, but an apparently normal middle segment remains, plausibly built by kinesin-II, which drives IFT along it; in the absence of both kinesin-II and Osm-3–kinesin, IFT is abolished, IFT particles accumulate at the transition zone and no axoneme is built, in accordance with the phenotype of IFT-particle mutants9. Thus, kinesin-II and Osm-3–kinesin function redundantly to build the middle segment, whereas Osm-3–kinesin is essential for distal-segment assembly. This could be accomplished if, for example, the cargo of kinesin-II and Osm-3–kinesin (Fig. 2b; K and O) are axoneme-stabilizing factors, for example, microtubule-associated proteins, with either cargo being capable of stabilizing the middle segment, but the Osm-3 cargo being uniquely required for distal-segment stabilization. Another possibility is that one or both motors could deliver distinct signalling molecules that control axoneme length, similar to the 'long flagella' (LF4) mitogen-activated protein kinase or the Chlamydomonas aurora-like kinase (CALK)26, 27.
This is the first demonstration that two distinct anterograde IFT-motor holoenzymes participate in IFT, raising the question of its generality. The cilia of Tetrahymena contain homologues of KRP85, KRP95 and OSM-3 (refs 25 and 28), but unlike in C. elegans, their oligomerization state and ability to move along cilia are unexplored. In sea urchin embryos11 and Chlamydomonas6, kinesin-II seems to be the dominant IFT motor, but Osm-3–kinesin could assemble the procilium that initiates ciliogenesis11 and drive the transient elongation of the distal segments of the axoneme during mating14.
Thus, C. elegans sensory cilia use two sequential IFT pathways: a 'middle segment' pathway dependent on kinesin-II and Osm-3–kinesin, and a 'distal segment' pathway dependent on Osm-3–kinesin alone. The subtle manner in which the two motors are deployed, having distinct but partially overlapping functions, is noteworthy. Functional cooperation and redundancy between mitotic motor holoenzymes is common, but to our knowledge this is the first such example pertaining to kinesins involved in intracellular transport. Moreover, this indicates that IFT may be more complex and fascinating than is currently appreciated; given the large range of cargo that must be delivered to cilia, perhaps two IFT pathways operating in a single cilium is the 'tip of the iceberg', and additional pathways may contribute to the delivery of ciliary components.
Methods
Fluorescence microscopy.
Intraflagellar transport was assayed as described15, 16, 17. Images were collected on an Olympus microscope equipped with a 100, 1.35 NA objective and an Ultraview spinning disc confocal head17 at 2–3 frames s-1 for 2–3 min, with the transgenic worms anaesthetized with 10 mM levamisole, mounted on agarose pads and maintained at 21 °C. Kymographs and movies were created using Metamorph software. The length distribution of OSM-6–GFP along sensory cilia was determined by densitometry of confocal images of live transgenic animals.
Creation and maintenance of transgenic and mutant animals.
We periodically requested kinesin-II mutants from the C. elegans gene knockout consortium, who produced and provided us with mutants in the KRP95 (klp-11(tm324)) and KAP (kap-1(ok676)) subunits created by UV–trimethyl-psoralen mutagenesis18. They were outcrossed multiple times and these, the osm-3 mutant strains, and the GFP–IFT-protein-expressing strains were maintained using standard procedures16, 17.
Characterization of mutants.
Genomic sequences were obtained by PCR amplification from single worms or genomic extracts. cDNAs were prepared by RT–PCR on total RNA extracts. For each allele, molecular lesions were confirmed by multiple sequence alignment from 3–4 independent PCR reactions using Pileup and Pretty sequence analysis software. Dye-filling and osmotic avoidance assays were performed as described9, 19.
Genetic crosses.
To create double motor mutants expressing IFT-particle–GFP transgenes, males expressing OSM-6–GFP were crossed with homozygous-mutant kap-1 hermaphrodites, and the F1 heterozygotes were self-mated to obtain kap-1(ok676)III; mnIs17[osm-6::GFP] progeny. Osm-3(p802)IV; mnIs17[osm-6::GFP] males were crossed with homozygous kap-1 hermaphrodites, and the F1 heterozygotes were selfed to obtain kap-1(ok676)III; osm-3(p802)IV; mnIs17 [osm-6::GFP] progeny. Homozygosity was confirmed for osm-3(p802)IV by dye filling and for kap-1(ok676)III by single-worm PCR.
Note: Supplementary Information is available on the Nature Cell Biology website.
