Dauer pheromone and G-protein signaling modulate the coordination of intraflagellar transport kinesin motor proteins in C. elegans

Cilia length and function are dynamically regulated by modulation of intraflagellar transport (IFT). The cilia of C. elegans amphid channel neurons provide an excellent model to study this process, since they use two different kinesins for anterograde transport: kinesin-II and OSM-3 kinesin together in the cilia middle segments, but only OSM-3 in the distal segments. To address whether sensory signaling modulates the coordination of the kinesins, we studied IFT protein motility in gpa-3 mutant animals, since dominant active mutation of this sensory Gα protein GPA-3QL) affects cilia length. In addition, we examined animals exposed to dauer pheromone, since dauer formation, which involves gpa-3, induces changes in cilia morphology. Live imaging of fluorescently tagged IFT proteins showed that in gpa-3 mutants and in larvae exposed to dauer pheromone, kinesin-II speed is decreased and OSM-3 speed is increased, whereas structural IFT proteins move at an intermediate speed. These results indicate that mutation of gpa-3 and exposure to dauer pheromone partially uncouple the two kinesins. We propose a model in which GPA-3-regulated docking of kinesin-II and/or OSM-3 determines entry of IFT particles into the cilia subdomains, allowing structural and functional plasticity of cilia in response to environmental cues.


Introduction
Cilia are cellular protrusions that are present on almost all postmitotic vertebrate cells. Cilia have important motility or sensory functions and are required for the organization of several signal transduction pathways. Structurally, cilia are dynamic. Ciliogenesis, cilium length and morphology and the organization of signaling pathways in cilia are regulated (Mukhopadhyay et al., 2008;Quarmby, 2004;Sorokin, 1962;Wang et al., 2006). Although genetic analyses have identified several kinases that modulate cilia length (Bengs et al., 2005;Berman et al., 2003;Burghoorn et al., 2007;Tam et al., 2007), little is known about the molecular mechanisms involved; however, intraflagellar transport (IFT) is a probable process (Blaineau et al., 2007;Dentler, 2005;Engel et al., 2009;Evans et al., 2006;Marshall et al., 2005;Pan and Snell, 2005). IFT is responsible for bi-directional transport of structural and signaling components of the cilia along a microtubular axoneme. Kinesin-II motors mediate anterograde transport from the transition zone, where the axoneme is anchored in the cell, to the tip of the cilium. Cytoplasmic dynein 1B moves the particles back. IFT particles consist of at least 18 proteins, arranged in two complexes, A and B.
The amphid channel cilia of C. elegans can be divided into a middle segment with nine doublet microtubules and a distal segment with nine singlet microtubules (Perkins et al., 1986). Two kinesin complexes, heterotrimeric kinesin-II and homodimeric OSM-3 kinesin (KIF17 in human), which both belong to the kinesin-2 family, mediate transport in the middle segments, whereas only OSM-3 mediates transport in the distal segments (Snow et al., 2004). This bipartite cilium structure in C. elegans could provide a mechanism for plasticity of sensory signaling during environmental or developmental changes, for instance by regulating the length of the cilia, or the localization of signaling molecules in the two segments.
Many environmental cues are detected by G-protein-coupled receptors (GPCRs) and relayed by heterotrimeric G proteins to intracellular responses. C. elegans has 21 G, two G and two G subunits (Cuppen et al., 2003;Jansen et al., 1999). To test whether these G-protein subunits play a role in cilia development, C. elegans with mutant genes for all these proteins have been tested for the uptake of fluorescent dyes in the amphid neurons (dye filling), a process that requires the presence of the sensory cilia (Jansen et al., 1999;Perkins et al., 1986;Zwaal et al., 1997). Only gpa-3QL animals, which carry a dominant active mutant form of the sensory G subunit gene gpa-3, showed a dye filling defective phenotype, suggesting that it might affect cilia development or maintenance (Zwaal et al., 1997). gpa-3 is expressed in ten pairs of amphid sensory neurons, in the PHA and PHB phasmid neurons and in the AIZ and PVT interneurons, and plays a role in various sensory processes. GPA-3 also plays a role in dauer development, a developmental switch that allows C. elegans to survive under harsh environmental conditions or overcrowding (Riddle and Albert, 1997). When exposed to dauer pheromone, a constitutively produced pheromone that probably serves as a measure of population density, gpa-3(lf) animals form fewer dauers than wildtype animals, while gpa-3QL animals constitutively form dauers (Zwaal et al., 1997). Interestingly, dauer development requires cilia and is accompanied by alterations in the position and structure of several cilia (Albert and Riddle, 1983).
In this study, we examined whether dauer pheromone and GPA-3 regulate IFT. Electron microscopy and fluorescence microscopy of GFP-labeled cilia revealed that the cilia of gpa-3QL animals are shorter, but cilia length in gpa-3(lf) animals was only slightly affected. Live imaging of IFT proteins fused with GFP showed uncoupling of the two kinesins in gpa-3QL and gpa-3(lf) animals and in larvae exposed to dauer pheromone. However, structural IFT particle proteins moved at speeds intermediate to the two kinesins. Our results show that an environmental cue, probably mediated by GPA-3, modulates the coordination of the two IFT kinesins. We propose that this mechanism allows the regulation of cilia length or the localization of proteins in specific ciliary subdomains. Zwaal et al. (Zwaal et al., 1997) have shown that animals that carry a dominant active mutation of gpa-3, gpa-3QL(syIs24) and syIs25, are dye-filling defective. To study the cilia morphology of gpa-3QL animals, we examined cross sections of the amphid channel cilia of wild-type, gpa-3(lf) and gpa-3QL animals, using electron microscopy (EM; Fig. 1). The cilia of wild-type animals looked as previously described (Fig. 1B-D) (Perkins et al., 1986;Ward et al., 1975) and we observed no defects in the cilia of gpa-3(lf) animals (results not shown). gpa-3QL animals exhibited a variety of morphological defects in the cilia (  S1). In the distal segments we could distinguish four or five cilia (instead of ten in wild-type animals). These distal segments contained much electron-dense material and some had smaller diameters. Only few microtubules could be discerned, perhaps because of the presence of the electron-dense material ( Fig. 1E; supplementary material Fig. S1). Further proximally, at the beginning of the middle segments, more cilia could be seen (five to nine), some of which had smaller diameters and contained only few microtubules ( Fig. 1F; supplementary material Fig. S1). Closer to the transition zone ten cilia could be distinguished, some of which had smaller diameters ( Fig. 1G; supplementary material Fig. S1). Similar defects were observed in the amphid channel cilia of four gpa-3QL animals (supplementary material Fig. S1). In one of these animals the socket cell surrounding the cilia distal segments seemed more electron dense. The significance of this observation is not clear.

gpa-3QL affects cilia morphology
The EM analysis suggests that the amphid channel cilia of gpa-3QL animals are shorter and/or posteriorly displaced, although not all cilia are affected. To determine which cilia were affected, GFP was expressed in several amphid sensory neurons, using promoters that drive expression in a single or a restricted set of cells. Four constructs were transcriptional fusions: p gpa-4 ::gfp [drives GFP expression in the ASI neurons (Jansen et al., 1999)], p flp-6 ::gfp [in ASE (Li et al., 1999)], p ops-1 ::gfp [in ASG (Sagasti et al., 1999)] and p srh-142 ::gfp [in ADF (Sarafi-Reinach and Sengupta, 2000)]. In these cases GFP enters the cilia and probably localizes to the lumen. In addition, the cilia of the ASI neurons and of ASH, ASK and ADL were visualized using translational fusions, p gpa-4 ::gpa-4::gfp and p gpa-15 ::gpa-15::gfp, respectively, in which GFP is fused in frame to the first 46 or 40 N-terminal amino acids of the heterotrimeric G proteins GPA-4 or GPA-15 (Jansen et al., 1999). Since the N-terminal amino acids of G proteins are required for their localization at the membrane, we expect these GFP fusions to localize at the ciliary membrane. Finally, we used a p gpa-4 ::tbb-4::mCherry construct to visualize the axoneme of the ASI cilia using a translational fusion between full length tbb-4 and mCherry. tbb-4 encodes a -tubulin, which is part of the middle and distal segments of sensory cilia of C. elegans, but not the transition zone (Bae et al., 2006).
The cilia of the ADF, ASH, ASI, ASK and ADL neurons of adult gpa-3QL animals were shorter than those of wild-type adult animals (Fig. 2). Similar effects on ASI cilium length were observed with the transcriptional and the translational gpa-4::GFP fusions, which both visualize the complete cilium starting at the transition zone, and with the p gpa-4 ::tbb-4::mCherry construct, which visualizes only the axoneme ( Fig. 2; Table 1 and results not shown). The effect on cilium length was not fully penetrant; 14-82% of gpa-2078 Journal of Cell Science 123 (12) 3QL animals had shorter cilia (Table 1). By contrast, the cilia of the ASG neurons were longer, and the ASE cilia of gpa-3QL animals did not differ significantly from those of wild type ( Fig. 2; Table 1). Visualization of the ASH, ASK and ADL neurons using gpa-15::gfp showed that the middle segments and transition zones of their cilia were more spread out than in wild-type animals. In addition, the transition zones of the ASG and ADL cilia were displaced posteriorly (Fig. 2). There seems to be no general correlation between posterior displacement and cilium length, since ADL cilia were on average shorter, whereas ASG cilia were longer. Also loss-of-function of gpa-3 affected cilium length, but to a lesser extent than gpa-3QL: ASI cilia were slightly longer and ASH, ASK and ADL cilia slightly shorter (Table 1).
The two gpa-3QL strains had different length cilia: The ASI cilia were only shorter in the gpa-3QL(syIs25) animals, whereas ASH, ASK and ADL cilia were shortened in both strains (Table 1). These differences might be caused by lower gpa-3QL expression in gpa-3QL(syIs24) animals, however immunohistochemistry and western blot analysis using anti-GPA-3 antibodies revealed similar, approximately 60-fold overexpression in the two gpa-3QL strains (supplementary material Fig. S2). However, this analysis did not address possible cell-specific variations in expression levels. To determine if the phenotypic differences were caused by variations 2079 Sensory cues affect IFT kinesins   Data are the average lengths of cilia visualized with various GFP or mCherry fusion constructs ± the standard deviation (in µm); the number of measurements (n) and the percentage of short cilia (%).
Significant differences in length are indicated; compared with wild-type animals ( a in the expression of gpa-3QL, we injected the gpa-3QL construct at a range of concentrations into animals expressing the p gpa-4 ::gfp transcriptional fusion or the p gpa-15 ::gpa-15::gfp translational fusion and checked cilia length, morphology and posterior displacement of the basal bodies. The number and the severity of the ciliary defects correlated with the concentration of gpa-3QL injected (supplementary material Tables S1-S3), indicating that the variation in phenotypes is caused by differences in gpa-3QL expression levels.
Next, we examined whether activation of GPA-3 is important for its effect on cilium length. Overexpression of the wild-type gpa-3 gene had a weaker effect on dye filling (maximally 37% Dyf at 170 ng/l, compared with 85% for 170 ng/l gpa-3QL), did not significantly shorten cilia length (P>0.01), and only mildly increased the percentage of animals with short cilia (supplementary material Table S4). These results suggest that overexpression of wild-type gpa-3 affects cilia, but that activation of GPA-3 strongly enhances its effects.

The effect of gpa-3QL is inducible, reversible and cell-autonomous
To determine whether gpa-3QL acts specifically during ciliogenesis, which occurs at the embryonic threefold stage (Fujiwara et al., 1999), we generated animals that carried a heat-shock inducible gpa-3QL construct. Since both gpa-3QL(syIs24) and syIs25 are dye-filling defective (Dyf), we first tested whether induction of gpa-3QL expression resulted in a dye-filling defect. Heat shock of mixed populations of animals resulted in dye filling defects in L1-L4 larvae and adult animals. We tested three independent strains, resulting in 91% (gjEx1457), 45% (gjEx1456) and 17% (gjEx1292) Dyf in adult animals (Fig. 3). The variation in Dyf phenotypes is probably caused by somatic mosaicism and variation in gpa-3QL expression between the strains. The induced Dyf phenotype was reversible since dye filling was restored 24 hours after heat-shock treatment (gjEx1457 6%, gjEx1456 1% and gjEx1292 2% Dyf; Fig. 3). Also cilium length of gjEx1457 animals was reduced 2 hours after heat shock, but recovered to normal length within 24 hours (average length before heat shock 4.91±0.31 m (n18; 0% short cilia), 2 hours after heat shock 4.32±0.63 m (n36, P<0.005 compared with before or 24 hours after heat shock; 42% short cilia) and 24 hours after heat shock 4.85±0.44 m (n30, 0% short cilia). These results indicate that the effect of gpa-3QL is not restricted to a specific developmental time-window, but affects a regulatory mechanism important for cilia structure throughout the life of C. elegans. gpa-3 is only expressed in ten pairs of amphid sensory neurons, in the PHA and PHB phasmid neurons and in the AIZ and PVT interneurons (Lans et al., 2004;Zwaal et al., 1997). To determine whether the effect of gpa-3QL on cilia length is cell autonomous, or whether expression in one pair of sensory neurons also affects cilia morphology of other neurons, we generated animals that express gpa-3QL specifically in the ASI neurons, using a p gpa-4 ::gpa-3QL construct. We observed significantly shorter ASI cilia in animals expressing p gpa-4 ::gpa-3QL. However, we did not observe shortening of cilia of the neighboring neuron ADL, ASK or ASH in two independent transgenic strains (Table 1). These results suggest that the effect of gpa-3QL on cilia length is cellautonomous, although we cannot exclude the possibility that expression is too low to affect ASH, ASK and ADL cilia length.

Mutation of gpa-3 affects the coordination of kinesin-II and OSM-3
Cilium length is thought to be regulated by modulation of transport in the cilium (Dentler, 2005;Engel et al., 2009;Marshall et al., 2005;Pan and Snell, 2005). Hence we set out to see if mutation of gpa-3 affects IFT. First, we determined whether the effects of mutation of gpa-3 on cilia length require either of the two kinesin-2 motors. kap-1; gpa-3QL(syIs25) double mutants had significantly longer ASI cilia than gpa-3QL(syIs25) animals and kap-1; gpa-3QL(syIs24) double mutants had longer ADL, ASH and ASK cilia and shorter ASG cilia than gpa-3QL(syIs24) animals (Table 1). osm-3; gpa-3QL(syIs25) animals had even shorter ASI cilia than osm-3 animals. Also the effect of gpa-3(lf) on the length of ASI cilia was suppressed in kap-1; gpa-3 animals. These results suggest that the effect of gpa-3 mutation on cilia length is partially mediated by kinesin-II and might also be mediated by OSM-3.
Second, we determined the localization of KAP-1::GFP and OSM-3::GFP to visualize the two kinesin-2 motors, the dynein light intermediate chain XBX-1::GFP, the complex A protein CHE-11::GFP (IFT140 in human) and the complex B protein OSM-1::GFP (IFT172 in human) (Qin et al., 2001;Schafer et al., 2003;Signor et al., 1999;Snow et al., 2004). All these proteins could be detected at their normal localization in gpa-3QL animals (results not shown). It must be noted that ciliary defects of gpa-3QL animals were difficult to visualize using these constructs, probably because they are expressed in all amphid channel cilia and because of the variability of the defects.
Third, we measured transport rates of IFT components ( Fig. 4; supplementary material Table S5; representative examples of kymographs are presented in supplementary material Fig. S3). In the middle segments of wild-type animals, kinesin-II and OSM-3 both traveled at 0.7 m/second. In the absence of kinesin-II, OSM-3 moved faster (1.17 m/second) and in the absence of OSM-3, kinesin-II moved slower (0.49 m/second). In the distal segments, OSM-3 moved at 1.07 m/second. These speeds are in agreement with previous reports (Ou et al., 2005;Snow et al., 2004). We found that in the middle segments of both gpa-3QL and gpa-3(lf) animals OSM-3::GFP-containing particles moved at approximately 1 m/second. KAP-1::GFP speeds were not as strongly affected, resulting in speeds of approximately 0.60 m/second. In double mutants between gpa-3QL or gpa-3(lf) and kap-1 or osm-3 the kinesins moved at approximately the same speeds as in kap-1 or osm-3 single mutants, suggesting that mutation of gpa-3 does not affect the velocity of the kinesins per se, but rather that the two Fig. 3. The effect of gpa-3QL on cilia is inducible and reversible. DiI dye filling of animals expressing gpa-3QL under control of the hsp-16.2 heatshock promoter and wild-type controls before treatment, 2 hours after heat shock (hs), and 24 hours after heat shock. Scale bars: 50m. kinesin motors do not travel together in the same IFT particles. We observed no effects on the speeds in the distal segments (supplementary material Table S5).
Separation of OSM-3 and kinesin-II can be due to (1) a separation of the complex A and B proteins, as described for bbs-7 and bbs-8 animals (Ou et al., 2005), (2) a failure of one of the motors to link to the complex A or B scaffold, as described for dyf-5 or dyf-1 animals (Burghoorn et al., 2007;Ou et al., 2005), or (3) by an unknown mechanism. To discriminate between these possibilities we measured the anterograde speeds of fluorescently tagged complex A protein CHE-11/IFT140, complex B protein OSM-1/IFT172 and dynein subunit XBX-1 ( Fig. 4; supplementary material Table S5). The velocities in the middle segments of wildtype animals were all 0.7 m/second. In gpa-3QL and in gpa-3(lf) animals these three proteins traveled only slightly faster than 0.7 m/second, at a rate intermediate to that of the two kinesins, suggesting that these proteins move together. These findings do not fit either of the two mechanisms described before (Burghoorn et al., 2007;Ou et al., 2005) and suggest a novel mechanism that regulates the coordination of IFT.

Exposure to dauer pheromone affects the coordination of the two kinesins
gpa-3 plays a role in dauer development: gpa-3QL animals show a dauer constitutive phenotype, whereas gpa-3(lf) animals are dauer defective (Zwaal et al., 1997). Interestingly, dauer development involves alterations in the position and structure of several cilia, including a posterior displacement of the cilia of the ASI and ASG neurons (Albert and Riddle, 1983). Peckol et al. (Peckol et al., 2001) have shown that five of the six pairs of cells that normally take up fluorescent dye do so, with the exception of ASI. Measurement of ASI cilia length using GPA-4::GFP showed that exposure to dauer pheromone did not result in shorter cilia (results not shown), confirming that the ASI dye-filling defect does not result from cilia shortening, but may be the result of posterior displacement or by another structural change of the cilia (Albert and Riddle, 1983;Peckol et al., 2001).
To determine whether dauer pheromone exposure affects IFT protein motility, we measured transport rates of KAP-1::GFP, OSM-3::GFP, OSM-1::GFP and CHE-11::GFP in L2 and L2d animals. In the middle segments of untreated wild-type L2 larvae KAP-1::GFP and OSM-3::GFP traveled at a speed of approximately 0.7 m/second, which is comparable with the speeds in young adult animals. Exposure of larvae to dauer pheromone decreased KAP-1::GFP speed to 0.56 m/second and increased OSM-3::GFP speed to 0.89 m/second, whereas complex A and B proteins moved at approximately 0.7 m/second ( Fig. 5; supplementary material Table S6). Thus, exposure to dauer pheromone uncouples kinesin-II and OSM-3 kinesin, very similar to gain-or loss-offunction of gpa-3.
To determine whether the effect of dauer pheromone on the speeds of the kinesins is mediated by GPA-3, we measured KAP-2081 Sensory cues affect IFT kinesins Fig. 4. Mutation of gpa-3 affects transport speeds. Average speeds ± the standard deviation of OSM-3::GFP, KAP-1::GFP, CHE-11::GFP, OSM-1::GFP and XBX-1::GFP in various genetic backgrounds. Statistically significant differences compared with speeds in wild type are indicated in black, compared with OSM-3::GFP speeds in the same genetic background in red, and compared with KAP-1::GFP in the same genetic background in blue (***P<0.001, **P<0.005). Speeds were determined in adult animals. Further details can be found in supplementary material Table  S5.  Table S6. 1::GFP and OSM-3::GFP transport rates in gpa-3(lf) L2 and L2d animals. Exposure to dauer pheromone did not affect the speeds of the two kinesins in the middle segments of gpa-3(lf) larvae: KAP-1::GFP traveled at a speed of approximately 0.6 m/second, whereas OSM-3::GFP moved at approximately 0.9 m/second ( Fig. 5; supplementary material Table S6). Since exposure to dauer pheromone and gpa-3(lf) have very similar effects on the kinesin speeds, and these effects are not cumulative, this suggests that gpa-3 functions in the same pathway as exposure to dauer pheromone.
To test if exposure to dauer pheromone has an acute effect on IFT, we exposed L1 larvae for 4 hours to dauer-inducing concentrations of dauer pheromone, and measured speeds of KAP-1::GFP and OSM-3::GFP. In these animals, short exposure to dauer pheromone had a similar effect on OSM-3::GFP speeds as 24-28 hours exposure, but had no significant effect on KAP-1::GFP speeds ( Fig. 5; supplementary material Table S6). These results indicate that dauer pheromone has an acute effect on the regulation of IFT in larvae.
Finally, we exposed adult animals to dauer pheromone for 4 hours, and measured IFT speeds. However, there were no effects on kinesin speeds in these animals ( Fig. 5; supplementary material Table S6), suggesting that there is a specific developmental time window in which dauer pheromone can modulate IFT.

Discussion
We show that in gpa-3QL and gpa-3(lf) animals and in larvae exposed to dauer pheromone IFT is perturbed: kinesin-II speed is reduced and OSM-3 kinesin speed is increased, indicating that they move separately. The velocities of KAP-1::GFP and OSM-3::GFP in these animals suggest that this separation is not absolute. Based on the average speeds of the two motors we calculated that in the cilia of gpa-3(lf) animals approximately 55% of the motors move separately, in gpa-3QL(syIs24) approximately 60% and in gpa-3QL(syIs25) approximately 40%, leaving a significant background of non-separated complexes. Surprisingly, complex A and B proteins travel at rates that are not or only slightly different from those in wild-type animals, and that are intermediate to the rates of the kinesins in gpa-3 or pheromone-exposed animals. This particular combination of speeds for IFT components does not fit a separation as found in bbs-7 and bbs-8 mutants, where kinesin-II travels together with the complex A proteins and OSM-3 travels with the complex B proteins (Ou et al., 2005), or a docking defect as found in dyf-5 or dyf-1 animals (Burghoorn et al., 2007;Ou et al., 2005).
We propose two possible explanations for these IFT speeds. First, mutation of gpa-3 or exposure of larvae to dauer pheromone could alter the stoichiometry of kinesin motors on the IFT particles. Depending on its penetrance, this would change the molar ratio of motors on the IFT particles, which would affect their transport rates , or even result in a full separation, where IFT particles are loaded exclusively with kinesins-II or OSM-3. In both cases, the velocities of complex A and B proteins would be approximately the mean of the speeds of kinesin-II and OSM-3, because these proteins are transported by both kinesins (supplementary material Fig. S4A). The second possibility is that in gpa-3 mutant animals and in dauer pheromone-treated larvae the ability of the kinesins to dock onto the IFT particles is affected, resulting in pools of free kinesins. IFT particles, containing complex A and B proteins, are then only transported by kinesin-II and OSM-3 together, at approximately 0.7 m/seconds, while nonassociated kinesin-II and OSM-3 travel at 0.5 and 1.2 m/seconds, respectively (supplementary material Fig. S4B).
The first scenario suggests the presence of discernable subgroups of kinesin-II-or OSM-3-transported particles. However, plotting the speeds of OSM-1::GFP, CHE-11::GFP and XBX-1::GFP in the middle segments of gpa-3QL cilia did not reveal a clear bimodal distribution (supplementary material Fig. S5). It must be noted that a bimodal distribution might be masked by the large variance in the speeds of the particles, and by the sizable fraction of particles transported by both motor complexes. However, data presented by Imanishi et al. (Imanishi et al., 2006) suggest that OSM-3 exists in an autoinhibited state, which is thought to be relieved by binding to IFT particles. This finding would argue against movement of free OSM-3 and thus possibly against our second model. Further analysis is required to resolve this issue.
The IFT machinery not only consists of motor proteins and adapter proteins, but also contains regulatory proteins, such as the Rab-like G proteins IFT27 and Rab8 (Nachury et al., 2007;Qin et al., 2007). This raises the question of whether GPA-3 is also part of the IFT machinery. Although GPA-3QL and GPA-3 localize to the cilia, GPA-3::GFP does not localize to IFT particles, nor could we measure motility of GPA-3::GFP in the cilia. We therefore favor the possibility that GPA-3 acts at the membrane to relay GPCR activation into intracellular signaling, which acts on the coordination of kinesin motors in the cilia.
Our finding that exposure to dauer pheromone affects the speeds of kinesin-II and OSM-3 in a very similar way as mutation of gpa-3 and that these effects are not cumulative, suggests that gpa-3 functions in the same pathway as dauer pheromone to regulate the coordination of IFT kinesins. Zwaal et al. (Zwaal et al., 1997) previously found that gpa-3 probably functions upstream in the DAF-7/TGF pathway to regulate dauer formation. It is unclear if the regulation of IFT depends on the DAF-7/TGF pathway or functions upstream or in parallel of this pathway. Preliminary data suggest that exposure to dauer pheromone and activation of GPA-3 on the one hand induces the dauer pathway via inactivation of DAF-11 and DAF-7, and on the other hand affects the coordination of IFT by the two kinesins. Further experiments are required to resolve this issue. It is puzzling that exposure to dauer pheromone, dominant active mutation of gpa-3 and loss-of-function of gpa-3 have similar effects on IFT, i.e. they partially uncouple the two kinesins, whereas dauer pheromone and gpa-3QL induce dauer formation and affect cilia morphology, and gpa-3(lf) suppresses dauer formation and does not affect cilia morphology. At present, we have no data that explain these findings. However, it might be that although activation and inactivation of gpa-3 have similar effects on the coupling of the two kinesins, they might have differential effects on cargo loading. Pan and Snell have shown that shortening of the cilia in Chlamydomonas is the result of increased IFT shuttling and reduced anterograde cargo loading (Pan and Snell, 2005). An intriguing possibility is that in gpa-3QL animals the subset of IFT particles transported only by OSM-3 does not contain cargo, whereas the kinesin-II-transported particles contain molecules that maintain the integrity of the cilia, resulting in shortening of the distal segments, but leaving the middle segments intact. In gpa-3(lf) animals, the loading of kinesins could be the other way around, thus leaving the cilia intact. To test this model one would have to be able to differentiate the subsets of IFT particles, for example by their cargo. Such analyses would also help to choose between the two possible IFT models. Thus far, we have not identified such cargo molecules.
Recently, Engel et al. (Engel et al., 2009) found an inverse correlation between IFT particle size and the length of Chlamydomonas flagella: shorter flagella contained larger IFT particles or trains. This finding suggests that cilia and flagella length might be regulated by modulating IFT particle size. Unfortunately, we were not able to analyze IFT particle size in our kymographs, because of the background fluorescence caused by expression of the GFP fusion constructs in other cilia and because of the incomplete penetrance of gpa-3QL.
What would be the physiological significance of uncoupling of kinesin-II and OSM-3 kinesin by dauer pheromone and G protein signaling? First of all, exposure to dauer pheromone affects cilia structure, since it affects ASI dye filling and results in posterior displacement of the ASI and ASG cilia (Albert and Riddle, 1983), although it does not result in shortening of the cilia. Over activation of this pathway, in gpa-3QL animals, strongly affects the morphology of the amphid channel cilia: many cilia are shorter, some are longer, and some are displaced posteriorly. Although no obvious changes in cilia structure have been observed in gpa-3(lf) animals, it is possible that loss-of-function of gpa-3 has an opposite effect, resulting in stabilization of the cilia. In addition to regulating cilia morphology, we expect that coordinated coupling of kinesin-II and OSM-3 also allows relocalization of receptors or other signaling molecules in the cilia. For example, signaling molecules that are normally transported to the distal end by OSM-3 can be cleared from the distal segment by excluding them from OSM-3transported particles. These effects may serve to alter the sensitivity of the animal to certain environmental cues and for example reinforce the choice to become a dauer.
Many components of the IFT machinery and the signaling pathways are remarkably well conserved in evolution. This suggests that similar mechanisms in which external cues can influence cilia structure also exist in other organisms. A recent report by Besschetnova et al. (Besschetnova et al., 2010) and our results illustrate this concept in organisms as diverse as mammals and worms.

Immunoblotting
Total lysates of 80 worms were solubilized with SDS-PAGE sample buffer. After electrophoresis on a 15% SDS-PAGE gel, proteins were transferred to membranes and probed with antibodies against GPA-3 [clone AYW9, affinity purified (Lans et al., 2004), secondary antibody Amersham DRab HRP, detected with ECL, Amersham] and against -tubulin (Invitrogen clone B512, secondary antibody Li-COR DMouse Infrared 795, detected with an Odyssey infrared imager). GPA-3 protein levels were estimated by averaging GPA-3 band intensities (normalized against -tubulin) of three different exposures of four blots each. An example illustrating GPA-3QL overexpression is given in supplementary material Fig. S2.

Microscopy
The location of fluorescent proteins and cilia was examined using a Zeiss LSM 510 confocal microscope. Cilia lengths were measured using a Zeiss Imager Z1 microscope, by measuring the length from the transition zone to the distal tip of the cilium. In all cases cilium length was determined in adult animals. The percentage of short cilia, which were considered to be smaller than the intersection between the distribution plots of wild-type and osm-3 cilia lengths (4.04 m for GPA-4::GFP and 5.18 m for GPA-15::GFP), was determined. Dye filling was performed using 0.1 mg/ml DiI (Molecular Probes) as described previously (Perkins et al., 1986). Immunofluorescence using polyclonal rabbit antibody against GPA-3 was performed as described previously (Lans et al., 2004). For EM, animals were fixed in 3% glutaraldehyde for 16 hours, followed by post-fixation in 1% osmium tetroxide for 2 hours. Subsequently, animals were orientated in special moulds and embedded in Epon according to standard procedures. Ultrathin sections were cut with a Reichert ultramicrotome, stained with uranyl acetate and lead nitrate and examined with a Philips CM100 electron microscope at 80 kV.

Heat shock
Mixed populations of well-fed animals were heat-shocked at 30°C for 15 hours, allowed to recover at room temperature for 2 hours and subsequently analyzed using dye filling or by measuring cilium length as described above.

Live imaging of IFT particles
Live imaging of the GFP-tagged IFT particles was carried out as described previously Snow et al., 2004). Images were acquired on a Zeiss LSM 510 confocal microscope with a 63ϫ (NA 1.4) objective. Worms were mounted on an agarose pad and anaesthetized with 10 mM levamisole. Kymographs were generated in ImageJ with the kymograph plugin, written by J. Rietdorf.

Dauer analysis
Dauer pheromone was isolated as described previously (Golden and Riddle, 1982). Adult animals were allowed to lay eggs for 3 hours on plates containing dauer pheromone, at a concentration resulting in approx. 90% dauers in wild-type animals. Larvae were analyzed after 24-28 hours (L2 or L2d larvae) or after 52-58 hours (L4 or dauer animals) at 25°C.

Statistics and calculations
Statistical significance was determined using an ANOVA, followed by a Bonferroni post-hoc test.
The observed osm-3::gfp and kap-1::gfp speeds are each composed of the speed of the fraction that moves together with the other kinesin (at 0.70 m/seconds, osm-3::gfp speed in osm-3 animals, or kap-1::gfp speed in kap-1 animals) and the fraction that moves separately (at 1.17 m/seconds, osm-3::gfp speed in kap-1 animals, or 0.49 m/seconds, kap-1::gfp speed in osm-3 animals). The formula used to calculate the fraction of osm-3 that moves separately (x) is: x osm-3 (v gpa-3v together )/(v max -v together ), where v gpa-3 is the osm-3::gfp speed measured in a gpa-3 mutant, v together is the osm-3::gfp speed measured in osm-3 animals and v max is the osm-3::gfp speed measured in kap-1 animals. The formula used to calculate the fraction of kap-1 that moves separately (x) is: x kap-1 (v gpa-3 -v together )/(v max -v together ), where v gpa-3 is the kap-1::gfp speed measured in a gpa-3 mutant, v together is the kap-1::gfp speed measured in kap-1 animals and v max is the kap-1::gfp speed measured in osm-3 animals.