DLK-1/p38 MAP Kinase Signaling Controls Cilium Length by Regulating RAB-5 Mediated Endocytosis in Caenorhabditis elegans

Cilia are sensory organelles present on almost all vertebrate cells. Cilium length is constant, but varies between cell types, indicating that cilium length is regulated. How this is achieved is unclear, but protein transport in cilia (intraflagellar transport, IFT) plays an important role. Several studies indicate that cilium length and function can be modulated by environmental cues. As a model, we study a C. elegans mutant that carries a dominant active G protein α subunit (gpa-3QL), resulting in altered IFT and short cilia. In a screen for suppressors of the gpa-3QL short cilium phenotype, we identified uev-3, which encodes an E2 ubiquitin-conjugating enzyme variant that acts in a MAP kinase pathway. Mutation of two other components of this pathway, dual leucine zipper-bearing MAPKKK DLK-1 and p38 MAPK PMK-3, also suppress the gpa-3QL short cilium phenotype. However, this suppression seems not to be caused by changes in IFT. The DLK-1/p38 pathway regulates several processes, including microtubule stability and endocytosis. We found that reducing endocytosis by mutating rabx-5 or rme-6, RAB-5 GEFs, or the clathrin heavy chain, suppresses gpa-3QL. In addition, gpa-3QL animals showed reduced levels of two GFP-tagged proteins involved in endocytosis, RAB-5 and DPY-23, whereas pmk-3 mutant animals showed accumulation of GFP-tagged RAB-5. Together our results reveal a new role for the DLK-1/p38 MAPK pathway in control of cilium length by regulating RAB-5 mediated endocytosis.


Introduction
Primary cilia are evolutionarily conserved organelles that extend from the cell's surface and are used to sense cues in the environment. Cilia are present on nearly all cells of the vertebrate body and harbor specific receptors and other signaling molecules depending on the cell type. Cilia dysfunction is the cause of many diseases and can result in different symptoms including infertility, polydactyly, retina degeneration, mental retardation and kidney cyst formation [1].
All cilia contain a microtubule core, the axoneme. The axonemal microtubules are essential for a specialized transport pathway called intraflagellar transport (IFT) [2,3]. IFT transports ciliary building blocks and signaling molecules along the axoneme to the ciliary tip (anterograde) and back to the base (retrograde). Anterograde transport is mediated by kinesin-2 and IFT dynein transports particles in the retrograde direction. In addition to the motors and cargo, IFT particles contain many other proteins, including complex A and complex B and Bardet-Biedl syndrome (BBS) proteins, that are thought to form a scaffold between cargo and motor complexes.
The lipid and protein composition of the cilium differs from that of the plasma membrane to accommodate the cilium's specialized function [4,5]. To establish the unique protein and membrane composition, entrance of proteins and lipids is restricted at the base of the cilium by a barrier, called the transition zone [6,7]. The cilium receives components from multiple sources. One route, originates from the Golgi and involves the Golgi protein GMAP210 and the complex B protein IFT20 [8][9][10]. In addition, the cilium receives components from endocytic compartments which accumulate at the base of the cilium [11][12][13]. Disruption of endocytic gene function causes defects in targeting of ciliary transmembrane proteins to the cilium and expansion of ciliary membranes [11,14,15]. In mammalian cells, clathrin-dependent endocytosis at the ciliary base is important for the regulation of TGF-β and Notch signaling [16,17].
Several aspects of endocytosis are regulated by the small GTPase Rab5, including vesicle formation, fusion and motility of early endosomes. Rab5 activity is positively regulated by the guanine nucleotide exchange factor (GEF) Rabex-5 (in C. elegans RABX-5 and RME-6), while the GTPase activating protein (GAP, TBC-2 in C. elegans) inactivates Rab5 [18]. In addition, Rab5 membrane localization is regulated by GDP dissociation inhibitors (GDIs). GDI proteins extract the inactive form of most prenylated Rab proteins from membranes and these proteins can subsequently be delivered to target membranes where a new cycle of Rab activation can occur.
We study the structural plasticity of cilia in the nematode Caenorhabditis elegans. C. elegans harbors cilia on the dendritic endings of a subset of neurons, which mainly function in chemosensation. The cilia of the amphid channel neurons are structurally divided in a middle and a distal segment [19,20]. In these cilia anterograde IFT is mediated by two kinesin-2 motor complexes; heterotrimeric kinesin-II and homodimeric OSM-3 [21,22]. Imaging experiments have shown that kinesin-II and OSM-3 travel together in the middle segment of the cilium at a velocity of~0.7 μm/s, while only OSM-3 enters the distal segment where it moves at a higher speed (~1.1 μm/s) [21].
As in other organisms, the structure and function of cilia of C. elegans are dynamically regulated [23,24]. Structural changes were observed in cilia of dauer larvae, an alternative larval stage that allows animals to survive for long periods without food [25]. We found a partial uncoupling of the two kinesins in cilia of larvae exposed to a pheromone that induces dauer formation: kinesin-II moved at 0.6 μm/s, while OSM-3 moved at 0.9 μm/s. Complex A and B proteins moved at intermediate speeds [26]. Dauer development involves the ciliary localized heterotrimeric G protein α-subunit GPA-3 [27]. We found that IFT in the cilia of gpa-3 mutant animals is altered similarly to dauer pheromone exposed animals. In addition, mutants overexpressing a dominant active version of GPA-3 (gpa-3QL) have short cilia [26,27]. It is likely that the uncoupling of the two motor proteins in the gpa-3QL(syIs25) mutant causes cilia shortening.
In this study, we show that mutation of several genes in the DLK-1/p38 MAP kinase pathway can suppress the short cilia phenotype of gpa-3QL animals. In addition, our results suggest that this pathway acts in cilium length control by regulating RAB-5 mediated endocytosis.

Results
Mutation of uev-3 suppresses the dye-filling defect of gpa-3QL The cilium defects of gpa-3QL(syIs25) animals result in diminished uptake of fluorescent dyes in the sensory neurons, a process called dye-filling [20,27]. To identify new proteins that play a role in this process, we performed a forward genetic screen for suppressors of the gpa-3QL (syIs25) dye filling defect. Using SNP-mapping we mapped the mutation in sql-4(gj204) (suppressor of gpa-3QL #4) to a region of chromosome I of approximately 280 kb. Sequencing of genes in this region identified a G to A mutation in the first codon of exon 6 of the uev-3 gene, resulting in a premature stop.
Expression and localization of GPA-3 is not affected by mutation of uev-3 Since GPA-3QL expression affects cilium length and dye-filling in a dose-dependent manner [26], we wondered whether the suppressor mutations affect GPA-3QL protein levels and/or localization, thereby restoring dye-filling. Therefore, we performed immuno-fluorescence (IF) using an anti-GPA-3 antibody. GPA-3 is expressed in ten pairs of amphid neurons, in the PHA and PHB phasmid neurons and in the AIZ and PVT interneurons [27,36]. In wild type animals, GPA-3 is mainly detected in the cilia (S1 Fig). uev-3(gj204) and uev-3(ju639) animals showed very similar localization of GPA-3 (S1 Fig). In gpa-3QL(syIs25) animals, GPA-3 localizes to cilia as well as to cell bodies and dendrites [26] (S1 Fig). In the suppressor strains, the localization of GPA-3 and the intensity of the signal was similar to what was observed in the gpa-3QL(syIs25) mutant (S1 Fig). These results suggest that GPA-3 levels and localization are unaltered in the suppressor strains, although quantitative conclusions cannot be drawn from these experiments. Scale bars 2 μm. Anterior is to the left. (B and C) Percentage dye-filling in the indicated strains. Error bars SD. Statistical analysis was performed using an ANOVA, followed by a Bonferroni post hoc test. Black *: statistically significant compared to wild type, red *: statistically significant compared to gpa-3QL (p<0.001). Mutation of DLK-1/p38 MAP kinase genes dlk-1 and pmk-3 suppress the dye-filling defect of gpa-3QL UEV-3 is an E2 ubiquitin-conjugating enzyme variant shown to directly bind the p38 MAP kinase PMK-3 [28]. To investigate whether PMK-3 is also involved in cilium length control, we tested if mutation of pmk-3 suppressed the dye filling defect. pmk-3(ok169); gpa-3QL(syIs25) animals showed dye-filling, indicating that pmk-3(ok169) is also a suppressor of gpa-3QL (syIs25) (Fig 1B).
Our data suggest that MEK-1, possibly acting redundantly with another MAP2K, acts directly upstream of PMK-3 and downstream of DLK-1 in the pathway regulating cilium length in gpa-3QL animals (Fig 2).
Mutation of dlk-1, pmk-3 and uev-3 restores cilium length of gpa-3QL animals cell autonomously Cilium length is reduced in adult gpa-3QL(syIs25) animals [26]. We tested whether cilium length is restored in the suppressor strains. First, we measured cilium length in the ASI neurons of uev-3, dlk-1 and pmk-3 single mutants using a p gpa-4 ::gfp construct, resulting in expression of GFP specifically in this pair of neurons. uev-3(ju639) animals had slightly shorter cilia than wild type, while the other single mutants showed wild type lengths ( Fig 3A). Cilium length in ASI neurons of the suppressor mutants was significantly longer than in gpa-3QL(syIs25) ( Fig  3A). In addition, cilia of the ASH, ASK and ADL cells, visualized using a p gpa-15 ::gfp construct, were restored in the uev-3(ju639); gpa-3QL(syIs25) mutant ( Fig 3B). However, the posterior displacement of the cilia, previously seen in the gpa-3QL(syIs25) mutant, was still observed in the suppressor strain (S2 Fig) [26].
To determine whether UEV-3 localizes to cilia, we expressed UEV-3::GFP from the ASI neuron specific gpa-4 promoter. This revealed that UEV-3::GFP was predominantly localized to the nuclei of the ASI neurons (Fig 4A), which fits with the presence of a nuclear localization signal (KKRRR) in the C-terminus of the protein. We also detected weak UEV-3::GFP fluorescence in the dendrites, axons and cilia of these neurons ( Fig 4A). It would be interesting to determine whether the nuclear localization signal in UEV-3 also plays a role in import into the cilium, as has been found for the kinesin-2 KIF17, where ciliary localization depends on a KRKK sequence and nuclear import proteins [38].
We observed strong PMK-3::GFP fluorescence in nuclei of neurons surrounding the second pharyngeal bulb, where the cell bodies of the ciliated amphid channel neurons localize ( Fig  4C). PMK-3::GFP is mainly present in the nucleus [40], but we also detected PMK-3::GFP in cilia and in the dendrites ( Fig 4D). To look more closely at the subcellular localization of PMK-3::GFP, we expressed this fusion protein specifically in ASI neurons. PMK-3::GFP localized mostly to the nucleus of these neurons, but was also visible in the cytoplasm, dendrites, axons and cilia ( Fig 4A).
To determine the localization of DLK-1, we expressed DLK-1::GFP from the ASI specific gpa-4 promoter. DLK-1::GFP mostly localized to the dendrites and axons of these neurons. The protein accumulated at the base of the cilium, but was only occasionally detected at very low levels inside the cilium (Fig 4A). DLK-1::GFP was present as a diffuse signal and as punctae in the dendrites as well as in the cytoplasm, in accordance with previous observations (Fig 4A) [40]. Some of these punctae were mobile in both the dendrite and cell body (S1 and S2 Videos). As a first step to identify the organelles that harbor these DLK-1::GFP punctae, we performed co-localization experiments combining DLK-1::GFP with mCherry::RAB-5 to visualize endosomes, or with immunofluorescence staining against the Golgi protein SQL-1 [10]. These experiments showed that the DLK-1::GFP punctae are distinct from mCherry::RAB-5 labelled Together these results show that UEV-3:: GFP and PMK-3::GFP localize to the nucleus, the cytoplasm, axons, dendrites and cilia and DLK-1::GFP localizes to the cytoplasm, axons and dendrites and at the base of the cilia. Whether the localization inside the cilium and/or at its base is important for their function in cilium length control remains to be determined.

Intraflagellar transport is altered in uev-3 and dlk-1 mutant animals
IFT is affected in gpa-3QL(syIs25) animals: kinesin-II subunit KAP-1 moves at a lower speed compared to wild type (0.6 μm/s), while OSM-3 has a higher velocity (0.9 μm/s) [26]. We proposed that in gpa-3QL(syIs25) animals three types of IFT particles exist; particles transported only by kinesin-II (0.5 μm/s), particles transported only by OSM-3 (1.1 μm/s) and particles transported by both motors (0.7 μm/s). Of these, the particles transported by OSM-3 only or by both motors can be transported into the distal segments by OSM-3, whereas particles transported by kinesin-II only will not be able to enter the distal segments, since they do not contain OSM-3. These effects could explain the short cilia observed in gpa-3QL(syIs25) animals.
Interestingly, dlk-1(tm4024) and uev-3(gj204) single mutants showed increased speeds of OSM-3::GFP in the distal segments of the cilia ( Table 1). The functional significance of this finding is not clear, but the increased speed could reflect a change in the composition of the IFT particles, resulting in less drag on the motor proteins and thus a higher speed. More in detail analysis of the composition of the IFT particles and their motility is required to address this issue.
Next, we determined the speeds of other IFT particle components, the complex A subunit DAF-10::GFP (mammalian IFT122) and the complex B subunit CHE-13::GFP (mammalian IFT57). In gpa-3QL(syIs25) animals, complex A and B proteins move at a speed intermediate to those of the two kinesins (~0.75 μm/s [26]). In the uev-3(ju639) single mutant and in the uev-3(ju639); gpa-3QL(syIs25) double mutant, DAF-10::GFP and CHE-13::GFP moved at 0.63-0.68 μm/s in the middle segments of the amphid channel cilia (Table 1). Thus, these complex A and B proteins moved at a speed slightly lower than that of wild type and gpa-3QL (syIs25) animals.   Taken together, the speeds of the IFT components in the uev-3 and dlk-1 single mutants and in the gpa-3QL double mutants are very similar to those in the gpa-3QL mutants, suggesting that suppression of the gpa-3QL induced cilium length defect by mutation of uev-3 or dlk-1 is not caused by changes in the IFT machinery.

Blocking endocytosis suppresses the dye-filling defect of gpa-3QL
To identify the mechanism by which PMK-3 regulates cilium length, we tested several downstream effectors of the DLK-1/p38 MAPK pathway. First, we looked at the downstream substrate of PMK-3 in axon regeneration and synapse formation; the MAP kinase-activated protein kinase MAK-2 [28,31]. mak-2(gk1110) did not suppress the dye-filling defect ( Fig 5A).
In C. elegans the p38 MAP kinase pathway was shown to influence AMPA receptor endosomal trafficking at central synapses, indicating that the role of p38 MAP kinase in the regulation of endocytosis is conserved in C. elegans [40]. This is corroborated by suppression of the gpa-3QL cilium defect by mutation of pmk-3 or the RAB-5 GEFs rabx-5 and rme-6. To establish this further, we tested whether blocking endocytosis by mutating the clathrin heavy chain can also suppress the dye-filling defect. Clathrin is essential for life, so we used a temperature sensitive allele to test its effect on dye-filling. Interestingly,~80% of chc-1(b1025); gpa-3QL(gjEx862) animals cultured at the restrictive temperature of 25°C for 5 days were dye-filling (Fig 5B), indicating that reducing endocytosis by mutating clathrin suppressed the dye-filling defect of gpa-3QL. chc-1(b1025) single mutants grown at the restrictive temperature were all dye-filling.

GFP::RAB-5 accumulates in pmk-3 and pmk-3; gpa-3QL animals
Our finding that reducing endocytosis can suppress the dye filling defect of gpa-3QL animals suggests that increased endocytosis might contribute to the shortening of cilia in these animals. To visualize endocytosis, we expressed GFP::RAB-5 specifically in the ASI neurons. In wild type animals, GFP::RAB-5 localized to the cell body, dendrite and axon. In the cell body, the protein was present in the cytoplasm in punctae and ring structures, presumably early endosomes (Fig 6A). Unlike previous reports, we also detected GFP::RAB-5 inside the cilium (Fig 6A) [11,15].
First, we analyzed whether GFP::RAB-5 localization was affected by mutation of its GEFs or GAP. Mutation of the RAB-5 GEF rme-6 resulted in increased GFP::RAB-5 levels, whereas mutation of rabx-5 only slightly affected GFP::RAB-5 levels, although quantification of the fluorescence intensities did not reveal a significant difference between wild type and rme-6 or rabx-5 animals (Fig 6B and 6C). Inactivation of the RAB-5 GAP in tbc-2(tm2241) animals did not affect GFP::RAB-5 levels (Fig 6B and 6C). Formal proof that the GFP::RAB-5 construct is functional could not be obtained, because rab-5 loss-of-function is lethal and this construct is only expressed in the ASI neurons. However, expression of GFP::RAB-5 in the ASI neurons did not affect dye filling, cilium length or speeds of the kinesin-II subunit KAP-1::mCherry or OSM-3::mCherry (S5 Fig). Next, we tested whether gpa-3QL affects GFP::RAB-5. Interestingly, in gpa-3QL(syIs25) animals GFP::RAB-5 fluorescence intensity was reduced both at the dendritic ending and in the cell body (Fig 6A), although quantification of the fluorescence intensities did not reveal a significant difference between wild type and gpa-3QL animals (Fig 6D and 6E). However, these results are in accordance with the hypothesis that endocytosis is affected in gpa-3QL animals.
Next, we measured GFP::RAB-5 levels in pmk-3(ok169), uev-3(ju639) and uev-3(gj204) animals and in double mutants of these with gpa-3QL(syIs25). In pmk-3(ok169) animals, in which presumably GDI-1 is less active, resulting in inactive, membrane bound RAB-5, we would expect accumulation of GFP::RAB-5. Indeed, in these animals GFP::RAB-5 accumulated significantly at the base of the cilium, in the distal part of the dendrite and in the cell body (Fig 6A  and 6D and 6E). Similar increases in GFP::RAB-5 fluorescence intensity were observed in the AWB neurons (S6 Fig). Also pmk-3(ok169); gpa-3QL(syIs25) double mutant animals displayed accumulation of GFP::RAB-5 at the ends of the dendrites and in the cell bodies similar to the pmk-3(ok169) single mutant (Fig 6A and 6D and 6E). GFP::RAB-5 fluorescence intensities in uev-3(ju639) and uev-3(gj204) mutant animals and in double mutants of both uev-3 alleles with gpa-3QL(syIs25) were not elevated compared to wild type animals (Fig 7A and 7B), suggesting that mutation of uev-3 has less effect on RAB-5 recycling than mutation of pmk-3. However, both uev-3 mutations restored GFP::RAB-5 levels at the base of the cilium to wild type levels in gpa-3QL(syIs25) animals (Fig 7A and 7B), in line with the model that the dlk-1/pmk-3 pathway affects cilium length by reducing RAB-5 mediated endocytosis. Interestingly, in the cell bodies of the ASI neurons, GFP::RAB-5 levels were lower in uev-3(gj204); gpa-3QL animals than in wild type animals again suggesting that mutation of uev-3 has less effect on GFP::RAB-5 recycling than mutation of pmk-3. In addition, these results suggest that the uev-3(gj204) allele might be weaker than the uev-3(ju639) allele, although our other analyses did not reveal such a difference.
Since the animals harboring the dominant active version of GPA-3 showed a slight decrease in fluorescence intensity of GFP::RAB-5, we wondered whether the opposite happens in the gpa-3(pk35) loss-of-function mutant. However, gpa-3(pk35) animals showed wild type fluorescence intensity levels (Fig 6A and 6D and 6E).

Discussion
We have identified DLK-1/p38 MAP kinase signaling, as a novel pathway that plays a role in the regulation of cilium length. The core of this pathway consists of the dual leucine zipperbearing MAPKKK DLK-1, which phosphorylates the MAP2K MEK-1 and possibly another redundantly acting kinase, leading to activation of the p38 MAP kinase PMK-3, which binds the E2 ubiquitin-conjugating enzyme variant UEV-3, also required for the function of the pathway. Interestingly, we found that the DLK-1/p38 MAP kinase pathway regulates cilium length by acting on endocytosis. Together our results suggest that in gpa-3QL animals endocytosis is enhanced, and that cilium length in these animals can be restored by reducing endocytosis.
How the DLK-1/p38 MAPK pathway is activated is unclear. It could act downstream of GPA-3, since loss-of-function of the DLK-1/p38 MAPK pathway suppresses the dye filling defect of gpa-3QL, and loss-of-function of the DLK-1/p38 MAPK pathway and gpa-3 have similar effects on IFT. However, loss-of-function of pmk-3 results in accumulation of GFP:: RAB-5 at the base of the cilium, whereas loss-of-function of gpa-3 does not, suggesting that the The localization of GFP-fusions of full-length PMK-3, UEV-3 and DLK-1 in or at the base of cilia is consistent with a function in cilia or at their base to directly regulate endocytosis, as also reported by others [34,[40][41][42]. But since these proteins could also be detected in dendrites and cell bodies and PMK-3 and UEV-3 were most abundant in the nucleus we cannot exclude the possibility that the effects of the DLK-1/p38 MAPK pathway on endocytosis and cilium length are indirect and for example mediated by changes in gene expression. Previous studies have indicated that PMK-3 and its mammalian homologue contribute to stress-induced changes in gene transcription [45,46]. 14 animals were imaged per genotype. Peaks of sql-1(tm2409) (black *) and sql-1(tm2409); gpa-3QL(syIs25) (blue *) are significantly different from those of wild type (p<0.05). (C) Mean fluorescence intensities of dendritic endings of the indicated strains. At least 13 animals were imaged per genotype. sql-1; pmk-3 double mutant animals showed significantly lower fluorescence intensities than wild type (p<0.005). (D) Mean fluorescence intensities of ASI cell bodies of the indicated strains, corrected for cell size. At least 7 animals were imaged per genotype. sql-1(tm2409) and sql-1(tm2409); pmk-3(ok169) animals showed significantly different fluorescence intensities than wild type animals (black *, p<0.001; green *, p<0.05, respectively). Error bars SEM. Statistical analysis was performed using an ANOVA, followed by a Bonferroni post hoc test. It is unclear what role UEV-3 plays in the DLK-1/PMK-3 pathway. Although UEV proteins lack catalytic activity, it has been suggested that these proteins have a function in the ubiquitination pathway, probably by forming heterodimers with active E2 ubiquitin-conjugating (UBC) enzymes. For example, UEV-1 interacts with UBC-13 [47,48] and the mammalian homologues of these proteins catalyze K63 poly-ubiquitin chain formation [49,50]. K63 ubiquitination has a regulatory role in for example protein localization, protein-protein interaction and transcription. In this way, UEV-3 could regulate PMK-3 activity by regulating its localization. However, we did not observe any effects on PMK-3::GFP localization in uev-3 (ju639) mutant animals.
Previously, the DLK-1/p38 MAP kinase pathway was shown to be important for axon development, axon regrowth after injury and synapse formation [28][29][30][31][32][33]. Interestingly, the axon and the cilium share many common features. For example, both are cellular compartments that contain very stable microtubules [33,51] and both have specialized functions supported by their unique membrane and protein compositions. To maintain these differential compositions, specific proteins and lipids have to be delivered from the cell body. Transport of axonal and ciliary building blocks occurs via microtubule-based transport driven by motor proteins. Recently, it was shown that the DLK-1 pathway acts in axon regrowth after injury by acting on microtubule dynamics via the kinesin-13 family member KLP-7 and the post-translationally modifying enzymes CCPP and TTLL [33]. Although regulation of microtubule stability would nicely fit with regulation of cilium length, mutation of these genes did not suppress the cilium defect of gpa-3QL(syIs25) animals.
Instead, we found that the DLK-1/p38 MAP kinase pathway regulates cilium length by acting on endocytosis. Visualization of endosomes using GFP::RAB-5 showed that GFP::RAB-5 levels were lower in the gpa-3QL mutant compared to wild type, while they were higher in the pmk-3(ok169) and the pmk-3(ok169); gpa-3QL(syIs25) suppressor mutants. In addition, we showed that the effect of gpa-3QL on cilium length can be suppressed by reducing endocytosis, by inactivation of clathrin, or loss-of-function of the RAB-5 GEFs rme-6 or rabx-5. Together these results suggest that in gpa-3QL animals endocytosis is enhanced, and that cilium length can be restored by reducing endocytosis.
How can a change in endocytosis affect cilium length? The removal and supply of ciliary components has to be in balance to maintain cilium length. Removal of ciliary components is probably mediated by active export mediated by the IFT machinery and by endocytosis at the base of the cilium [11,13,52]. Thus, increased endocytosis in the gpa-3QL(syIs25) mutant would result in excessive removal of ciliary membranes and cilia shortening (Fig 10). Reducing endocytosis by reducing RAB-5 activity (by inhibiting DLK-1/p38 MAP kinase signaling and concomitantly GDI activity, or inactivation of the RAB-5 GEFs RME-6 or RABX-5), or by inactivation of clathrin, restores cilium length because the balance of ciliary membrane and protein supply/removal is re-established.
Previously, we identified another suppressor of gpa-3QL(syIs25), sql-1 [10]. SQL-1/ GMAP210 probably functions in sorting and/or targeting of vesicles from the Golgi to the cilium. To maintain cilium length, both supply and removal of ciliary (membrane) components need to be regulated. Our results suggest that the DLK-1/p38 MAP kinase pathway regulates the removal of ciliary components via endocytosis, while SQL-1 regulates the supply of these components (Fig 10). Indeed, overexpression of SQL-1 leads to an increase in cilium length [10]. In addition, we show that GFP::RAB-5 accumulates in sql-1 animals, suggesting that endocytosis is affected in these animals. Homeostasis of all membrane enclosed organelles are tightly linked and disruption of one organelle can affect another, for example because membrane dynamics or lipid synthesis is altered [53]. Therefore, it is possible that disruption of the Golgi complex in the sql-1 mutant has an effect on endocytosis. Our finding that GFP::RAB-5 levels are reduced in sql-1; pmk-3 mutants, or similar to wild type in sql-1; pmk-3; gpa-3QL animals, while they are strongly increased in sql-1 or pmk-3 single mutants, are consistent with the hypothesis that SQL-1 and PMK-3 affect GFP::RAB-5 levels via different mechanisms. The clathrin heavy chain, a subunit of the clathrin coat, localizes at the plasma membrane, the Golgi complex and the endosomal compartments where it functions in the formation of transport vesicles [53]. Thus, the suppression of the gpa-3QL dye-filling defect by inhibition of chc-1 could be caused by decreasing endosome formation, but also by altering the formation of clathrin-coated transport vesicles from the Golgi.
Based on our previous work, we proposed that dauer pheromone in the environment of the animal, detected by receptors in the cilia that activate the heterotrimeric G protein α subunit GPA-3, can modulate cilium length by changing the coordination of the two kinesins that mediate anterograde transport in the cilia, resulting in reduced transport of ciliary proteins into the distal segments. Our results presented here suggest that in addition in gpa-3QL animals the balance of supply (via the Golgi) and removal (via endocytosis) of membrane and protein components of the cilia is disturbed, resulting in shorter cilia (Fig 10). It will be interesting to determine if similar mechanisms contribute to structural plasticity of cilia in other organisms, allowing the regulation of sensory capacity in response to environmental signals.
Quantification of intensity measurements was done using ImageJ.

Live imaging intraflagellar transport
Live imaging of the GFP-tagged IFT motor proteins was performed as described [10]. Animals were mounted on a 2% agarose pad and anaesthetized with 10 mM levamisole. Images were acquired imaged at RT using a Nikon Ti Eclipse microscope with Spinning Disc unit (CSU-X1, Yokogawa), 100x Plan APO TIRF objective (n.a. 1.49), Photometrics QuantEM:512C EM CCD camera and Metamorph imaging software. Exposure time 300 ms, 150 frames. Kymographs were generated using ImageJ with the kymograph plugin written by Dr. I. Smal (Erasmus MC).
(TIF) S1 Video. DLK-1::GFP motility in the dendrite of an ASI neuron. Time-lapse video (7 frames/second) of the end of the dendrite of a wild type animal expressing DLK-1::GFP in the ASI neurons, revealing punctae moving along the dendrite. Video was acquired using a spinning disk microscope with a 100x objective. Exposure time 300 ms, 150 frames. Anterior is to the right.
(AVI) S2 Video. DLK-1::GFP motility in the cell body of an ASI neuron. Time-lapse video (7 frames/second) of the cell body of a wild type animal expressing DLK-1::GFP in the ASI neurons, revealing punctae moving in the cell body. Video were acquired using a spinning disk microscope with a 100x objective. Exposure time 300 ms, 150 frames. (AVI)