SQL-1, homologue of the Golgi protein GMAP210, modulates intraflagellar transport in C. elegans

Primary cilia are microtubule-based protrusions that can be found on the surface of almost all vertebrate cells, and have important sensory functions. Cilia dysfunction has been associated with a number of genetic diseases, collectively called ciliopathies (Hildebrandt et al., 2011). Many ciliopathies affect various tissues and are caused by defects in proteins that play a role in transport to the cilium or within the cilium. Often these mutations do not completely block ciliogenesis, but rather result in changes in cilia morphology and/or localization of signaling proteins in cilia.

A large diversity in cilia lengths and morphologies exists, probably reflecting the specialized functions cilia can have in different tissues and/or organisms. However, little is known about how this diversity is achieved. Different types of signaling molecules, including second messengers (Mukhopadhyay et al., 2008; Ou et al., 2009; Besschetnova et al., 2010; Abdul-Majeed et al., 2012), kinases (Berman et al., 2003; Wang et al., 2006; Burghoorn et al., 2007; Tam et al., 2007; Miyoshi et al., 2009; Omori et al., 2010), phosphatases (Kim et al., 2010; Clément et al., 2011; Abdul-Majeed et al., 2012) and G proteins (Burghoorn et al., 2010) have been shown to regulate cilia length and cilia morphology. Since these signaling pathways also regulate transport, both in and toward the cilium, it is likely that these transport processes play a role in regulating cilium length and morphology (Silverman and Leroux, 2009).

Ciliary proteins often originate from the Golgi apparatus (Pazour and Bloodgood, 2008; Emmer et al., 2010). A number of small GTPases, including ARF4, ARL6 and RAB8 (Nachury et al., 2007; Mazelova et al., 2009; Jin et al., 2010; Wiens et al., 2010), are involved in budding of vesicles from the Golgi apparatus and fusion at the ciliary base. In the absence of components of the BBSome complex, composed of seven Bardet–Biedl Syndrome (BBS) proteins (Nachury et al., 2007), the GPCRs (G protein coupled receptors) rhodopsin, SSTR3, and MCHR1 (Nishimura et al., 2004; Abd-El-Barr et al., 2007; Berbari et al., 2008), fail to reach the cilium, indicating that the BBSome functions in trafficking of (G-protein coupled) receptors to the cilium. However, whether the BBSome functions in vesicular transport, cargo selection, or vesicle docking and fusion is not completely clear.

Once inside the cilium, signaling proteins and structural proteins are transported by intraflagellar transport (IFT). IFT particles carry cargo along the microtubule axis of the cilium, towards the distal tip and back. Movement of IFT particles is powered by kinesin-2 motors (anterograde transport) and cytosolic dynein 2 (retrograde transport) (Rosenbaum and Witman, 2002). IFT particles are organized in two distinct subcomplexes, complex A and complex B (Cole and Snell, 2009). Complex A has been implicated in retrograde transport and complex B in anterograde transport. In addition, proteins that are part of the BBSome have been observed moving in the cilium, with similar velocities as components of the IFT complex (Blacque et al., 2004; Nachury et al., 2007), suggesting that the BBSome associates with or is part of the IFT complex.

GMAP210 was recently found to play a role in vesicular transport from the Golgi to the cilium (Follit et al., 2008). It belongs to the Golgin family, whose members are characterized by a large number of coiled-coil domains, and function in maintaining the organization and position of the Golgi apparatus (Short et al., 2005). GMAP210 localizes to the cis-Golgi network (Rios et al., 1994), and is important for Golgi integrity (Ríos et al., 2004). Interestingly, GMAP210 anchors the complex B protein IFT20 to the Golgi apparatus. IFT20 probably functions as an adaptor protein for targeting vesicles to the cilium. In line with its novel role in vesicular trafficking to the cilium, absence of GMAP210 results in reduced levels of ciliary polycystin-2 (Follit et al., 2008). GMAP210 knockout mice die at birth, most likely due to heart and lung defects (Follit et al., 2008). Both heart and lung are among the organs affected in ciliopathy patients (Cardenas-Rodriguez and Badano, 2009).

The sensory cilia of Caenorhabditis elegans amphid neurons provide an excellent model to study the regulation of cilia morphology. They can be divided in a middle segment, which has nine microtubule doublets, and a distal segment, which has nine microtubule singlets (Ward et al., 1975; Perkins et al., 1986). In the middle segment, the IFT complex is transported by two members of the kinesin-2 family, kinesin-II and OSM-3. At the end of the middle segment kinesin-II dissociates from the complex, which leaves OSM-3 to transport the IFT complex in the distal segment (Snow et al., 2004). We have previously shown that in gpa-3QL animals, which carry a dominant active mutant version of the sensory Gα subunit, gpa-3, the coordination of kinesin-II and OSM-3 is affected, and the sensory cilia are shorter (Burghoorn et al., 2010). Interestingly, exposure of animals to dauer pheromone (a continuously secreted compound that at high concentrations induces an alternative larval stage, the dauer larva (Golden and Riddle, 1982), affects the coordination of kinesin-II and OSM-3 very similarly as gpa-3QL (Burghoorn et al., 2010). IFT measurements on gpa-3 mutant animals exposed to dauer pheromone suggest that gpa-3 functions in the pathway by which the pheromone affects IFT. Together these results suggest that environmental cues transduced via GPA-3 can modulate IFT.

To find out how GPA-3 regulates IFT and cilia length we performed a screen for suppressors of the gpa-3QL cilia defect. We identified and characterized one of the mutants, sql-1, and found that it encodes the C. elegans homologue of the mammalian Golgin protein GMAP210. We show that SQL-1 is ubiquitously expressed and localizes to the Golgi apparatus. In sql-1 mutants the Golgi structure seems disorganized, in line with the function of Golgin proteins. In addition, sql-1 affects the stability of the IFT machinery, resulting in a partial separation of kinesin-II and OSM-3, and transport of the IFT complex predominantly by OSM-3.

A subset of the amphid neurons of wild-type C. elegans can take up fluorescent dyes (e.g. DiO) from the environment (Hedgecock et al., 1985). However, animals with ciliary defects lose this ability. Among the mutations that result in such a dye filling defect is a dominant active mutation of the sensory Gα subunit gpa-3, gpa-3QL (Zwaal et al., 1997). We performed a forward genetic screen for mutations that suppress the dye filling defect of gpa-3QL animals. Using SNP-mapping (Wicks et al., 2001), the mutation in gj202, one of the mutants (Fig. 1A), was mapped to a region of chromosome III. Injection of a combination of long-range PCR fragments spanning 35 kbp restored the gpa-3QL dye filling defect. This region was predicted to encode three genes, Y111B2A.4, Y111B2A.5 and Y111B2A.26. Using RT-PCR we found that these three predicted genes were actually a single gene spanning 18 exons, which we named sql-1 (suppressor of gpa-3QL 1). sql-1(gj202) contains an A>T nonsense mutation in exon 8 that introduces a premature stop codon (Fig. 1B).

sql-1 has previously been annotated as the C. elegans homologue of GMAP210 (Gillingham et al., 2004; Follit et al., 2008), a member of the family of Golgin proteins (Short et al., 2005). At the amino acid (a.a.) level SQL-1 shows 19% identity (and 42% similarity) to human GMAP210. Both SQL-1 and GMAP210 consist of predominantly coiled-coil domains; a.a. 5–1116 in SQL-1 and a.a. 64–1768 in GMAP210. The remaining 153 a.a. polypeptide of mammalian GMAP210 contains a GRAB (GRIP-related Arf-binding) domain, which is necessary for Golgi localization and might bind small GTPases, as well as a GA-1 motif, with unknown function (Gillingham et al., 2004; Short et al., 2005). The last 201 amino acids of mammalian GMAP210 have also been shown to interact with γ-tubulin (Infante et al., 1999; Rios et al., 2004). C. elegans SQL-1 contains regions with high similarity to the GRAB (56% similarity) and GA-1 (61% similarity) domains (supplementary material Fig. S1), suggesting these domains are conserved. The first 38 a.a. of GMAP210 contain an ALPS (ArfGAP-1 lipid-packing sensor) motif (Drin et al., 2007), which can also mediate Golgi targeting (Cardenas et al., 2009), but this motif is not present in C. elegans SQL-1. Finally, mammalian GMAP210 binds IFT20, and the IFT20-binding site was mapped to residues 1180–1319 of GMAP210 (Follit et al., 2008). Follit and colleagues suggested that this binding site was not conserved in C. elegans (Follit et al., 2008). We found a region in SQL-1 with 39% similarity (16% identity) to the IFT20 binding domain of GMAP210. To determine whether SQL-1 interacts with IFT-20, we performed co-IP assays on animals expressing GFP-tagged IFT-20. However, anti-SQL-1 antibodies did not bring down IFT-20::GFP, and anti-GFP antibodies failed to bring down endogenous SQL-1 (supplementary material Fig. S2A,B). In addition, we determined the subcellular localization of IFT-20 using a full length ift-20::gfp fusion construct. In the ciliated neurons, we observed IFT-20::GFP in the cilia and at their base in the basal body remnants, and a weak, diffuse IFT-20::GFP signal in the dendrites and cell bodies, but no distinct Golgi localization (supplementary material Fig. S2C, compare with Fig. 2B). The localization pattern of IFT-20::GFP in sql-1 mutant animals was identical to that of wild-type animals. These data suggest that in C. elegans, SQL-1 and IFT-20 do not physically interact, and that IFT-20 does not localize to the Golgi apparatus. Possibly, IFT-20s function in the Golgi apparatus is not conserved in C. elegans. However, we cannot exclude the possibility that IFT-20 is transported to and from the Golgi apparatus but that the steady state level at the Golgi apparatus is low.

Initially, two sql-1 deletion mutants were obtained from the NBP-Japan, sql-1(tm2409) and sql-1(tm2440). In the sql-1(tm2440) strain 192 bp of exon 4 was deleted, resulting in the loss of 64 a.a. of SQL-1 (Fig. 1B). Western blot analysis, using antibodies raised against the N- and C-terminal parts of the SQL-1 protein, showed that the sql-1(tm2440) strain indeed expressed a smaller SQL-1 protein (Fig. 1C; supplementary material Fig. S3A). sql-1(tm2440); gpa-3QL animals were dye filling defective, thus the deletion in sql-1(tm2440) did not suppress the dye filling defect of gpa-3QL animals (data not shown). The sql-1(tm2409) strain carries a 175 bp deletion, starting in intron 14 and ending in exon 15. RT-PCR showed that this resulted in the loss of exon 15, a frameshift in exon 16, and an early stop. We could not detect the truncated SQL-1 protein of sql-1(tm2409) animals on a western blot (expected molecular weight 137 kDa), or in immunofluorescence (Fig. 1C; supplementary material Fig. S3A,B), demonstrating that sql-1(tm2409) is a null allele. In sql-1(tm2409); gpa-3QL mutants the dye filling defect caused by gpa-3QL was suppressed. Recently, we obtained an additional null mutant, tm4995, which carries a 632 bp deletion that removes exon 1. Also the sql-1(tm4995) mutation suppressed the dye filling defect of gpa-3QL (data not shown). In addition to the loss-of-function mutants, we generated animals that contain extra copies of the sql-1 gene, sql-1XS. Western blot analysis showed that sql-1XS animals (gj2077) expressed approximately three times more SQL-1 than wild-type animals (supplementary material Fig. S3C). Overexpression of sql-1 did not affect dye filling. All loss-of-function and overexpression animals were healthy and showed no apparent phenotype.

To determine the expression pattern and subcellular localization of SQL-1, two different antibodies (Abs) were raised against SQL-1, one against an N-terminal part (a.a. 106–737) of SQL-1, and one against a C-terminal part (a.a. 519–1328). Immunofluorescence on wild-type animals showed a spotted pattern throughout the whole body (supplementary material Fig. S3B), suggesting SQL-1 is ubiquitously expressed. A similar pattern was seen in sql-1(tm2440) animals, but no immunoreactivity could be detected in sql-1(tm2409) (supplementary material Fig. S3B), confirming the specificity of the antibodies.

To confirm the localization pattern determined in the immunofluorescence experiments, we generated animals that expressed full length SQL-1 fused to GFP, p_sql-1::sql-1::GFP animals. In these animals we observed spots throughout their bodies (Fig. 2B), similar to the pattern observed with SQL-1 Abs. Since the homologue of SQL-1, GMAP210, localizes to the Golgi apparatus (Rios et al., 1994), we stained p_sql-1::sql-1::gfp animals with an antibody against the glucuronyl transferase SQV-8, which stains the Golgi apparatus (Hadwiger et al., 2010). This showed co-localization of SQL-1 and SQV-8 (Fig. 2B). Both the immunofluorescence and GFP fusion data confirm that SQL-1 localizes to the Golgi apparatus. Interestingly, also an N-terminal (a.a. 1–302) SQL-1::GFP fusion localized to spots throughout the animal, suggesting that perhaps SQL-1 contains an N-terminal region that mediates Golgi localization. However, this GFP-fusion was more abundant in the cytosol than the full length SQL-1 GFP-fusion (supplementary material Fig. S3D).

Mutation of sql-1 suppresses the dye-filling defect of gpa-3QL animals. Therefore, we analyzed the expression and subcellular localization of SQL-1 in neurons. We generated animals that expressed the Golgi marker AMAN-2, an alpha-mannosidase, tagged with YFP only in one pair of ciliated sensory neurons, the ASI neurons (p_gpa-4::aman-2::YFP), or in a subset of inter- and motor neurons (p_glr-1::aman-2::YFP). Staining of these animals with anti-SQL-1 Ab indeed showed co-localization of AMAN-2::YFP and SQL-1 in the cell bodies of glr-1 expressing neurons and the ASI neurons (Fig. 2C). In p_gpa-4::sql-1::GFP animals, we observed SQL-1::GFP in the Golgi apparatus in the cell bodies, and sometimes SQL-1 spots in the dendrites (Fig. 2D). These SQL-1 spots were immobile, and were never observed in close proximity of the cilium. Together these data show that SQL-1 localizes to the Golgi apparatus in C. elegans ciliated sensory neurons.

Knockdown of GMAP210 has been shown to result in the dispersion of the Golgi membranes (Ríos et al., 2004; Yadav et al., 2009). However, no defects were observed in the Golgi apparatus of GMAP210 mutant mice (Follit et al., 2008). We wondered whether the absence of SQL-1 affects the integrity of the Golgi apparatus in C. elegans.

To investigate this we visualized the Golgi apparatus specifically in the ASI neurons using AMAN-2::GFP in wild-type and sql-1(tm2409) animals. In wild-type animals we observed distinct AMAN-2::GFP structures in the ASI cells (Fig. 3A,B). However, in the majority of the sql-1(tm2409) animals the AMAN-2::GFP signal was more fragmented or even diffuse (Fig. 3A,B). Very similar effects were seen in sql-1(tm2409); gpa-3QL animals (Fig. 3A,B). In gpa-3QL animals we observed distinct AMAN-2 structures similar to those observed in wild-type animals (Fig. 3A,B). These results suggest that loss of function of sql-1 affects the organization of the Golgi apparatus, although we cannot exclude the possibility that the effect of sql-1 mutation is AMAN-2 specific.

One way to explain the suppression of the dye filling defect in sql-1(tm2409); gpa-3QL animals is a trafficking defect of dominant active GPA-3QL protein, caused by the absence of SQL-1. Therefore, we used immunofluorescence to visualize the localization of GPA-3. In both wild-type and sql-1(tm2409) animals GPA-3 can be observed at the ciliary membrane at comparable levels (Fig. 3C), showing that sql-1 is not required for ciliary localization of GPA-3. gpa-3QL and sql-1; gpa-3QL animals showed very strong anti-GPA-3 staining in the cilia, dendrites, and cell bodies. Especially high levels of GPA-3 were observed at the base of the cilia, possibly in the periciliary membrane compartment (Kaplan et al., 2012), making it difficult to visualize GPA-3 levels in the cilia of these animals. Nonetheless, we observed no obvious differences in the levels of GPA-3QL staining in the cilia of gpa-3QL and sql-1; gpa-3QL animals (supplementary material Fig. S4A). Also, a western blot showed no differences in GPA-3 levels between the two strains (supplementary material Fig. S4B). Together, these data suggest that suppression of the Dyf phenotype in sql-1(tm2409); gpa-3QL animals is not caused by an effect on the amount or the localization of GPA-3.

We have previously shown that the cilia of the ADF, ASH, ASI, ASK and ADL neurons of adult gpa-3QL animals are shorter than those of wild-type adult animals (Burghoorn et al., 2010), explaining the dye filling defect. To determine if sql-1 mutation affects cilia length of gpa-3QL animals, we visualized cilia using an ASI neuron specific gpa-4::gfp construct, and a gpa-15::gfp construct expressed in ASH, ASK and ADL neurons (Jansen et al., 1999). Cilium length of gpa-3QL animals was restored to approximately wild-type length in sql-1(tm2409); gpa-3QL animals (Fig. 4A,B).

To determine if changes in sql-1 levels themselves affect cilium length we visualized cilia of sql-1(tm2409) and sql-1XS animals. Cilium length in sql-1(tm2409) animals was comparable to cilium length in wild-type animals (Fig. 4A,B). Interestingly, the cilia of ASI neurons of animals that overexpress SQL-1, sql-1XS, were significantly longer than those of wild-type animals (Fig. 4A). To test if this effect on cilia length depends on either of the two kinesins we measured ASI lengths in osm-3; sql-1XS and kap-1; sql-1XS animals. Loss of osm-3 shortened the cilia of sql-1XS animals, but these cilia were still significantly longer than those of osm-3 animals (Fig. 4A), indicating that at least the middle segments are longer in sql-1 animals. However, loss of kap-1 completely suppressed cilia lengthening as a result of sql-1 overexpression (Fig. 4A), indicating that kinesin-II is required for this effect.

To further confirm the specificity of suppression of the dye filling defective (Dyf) phenotype in gpa-3QL animals by the loss of SQL-1 we reintroduced the sql-1 gene in sql-1(tm2409); gpa-3QL animals. This resulted in a decreased length of the cilia, comparable to the length in gpa-3QL animals (Fig. 4A). In addition, while expression of wild-type sql-1 specifically in the ASI neurons of sql-1; gpa-3QL animals did rescue ASI cilia length, it did not affect the length of the cilia of ASH, ASK and ADL (Fig. 4A,B). This suggests that sql-1 acts cell-autonomously.

We previously hypothesized that the decreased cilia length in gpa-3QL animals was caused by the observed partial uncoupling of kinesin-II and OSM-3 (Burghoorn et al., 2010). To determine whether sql-1 also affects the IFT machinery, we examined the motility of the two ciliary kinesin subunits KAP-1 (mammalian homologue KAP3) and OSM-3 (KIF17), two complex A proteins CHE-11 (IFT140), DAF-10 (IFT22), three complex B proteins OSM-1 (IFT172), CHE-13 (IFT57) and IFT-20, and the dynein motor subunit XBX-1 (D2LIC) in sql-1(tm2409) animals as well as in wild-type animals. In wild-type animals all IFT proteins moved at ∼0.7 µm/second in the middle segments (Table 1), and at ∼1.15 µm/second in the distal segments (supplementary material Table S1). These speeds are consistent with published data (Snow et al., 2004). However, in the middle segments of sql-1(tm2409) animals, kinesin-II moved at 0.59 µm/second, and OSM-3 at 0.89 µm/second (Table 1). Since the two kinesins moved at different velocities it is not possible that they were in the same complex at all times. CHE-11::GFP (complex A), OSM-1::GFP and IFT-20::GFP (both complex B), and XBX-1::GFP (dynein) moved at similar velocities as OSM-3 in the middle segments of sql-1(tm2409) animals (Table 1), whereas DAF-10::GFP (complex A) and CHE-13::GFP (complex B) moved at 0.67 µm/second (Table 1). In the distal segments of sql-1(tm2409) animals, all examined IFT markers moved at velocities similar to those in wild-type animals (supplementary material Table S1).

To determine if any of these effects were caused by changes in the intrinsic velocities of kinesin-II or OSM-3, we decided to measure how the absence of SQL-1 affects IFT motility in osm-3 and kap-1 mutant backgrounds. In sql-1(tm2409); osm-3 animals KAP-1::GFP (kinesin-II), DAF-10::GFP (complex A), CHE-13::GFP (complex B), and XBX-1::GFP (cargo) moved at 0.49–0.52 µm/second (Table 1), similar to the velocities observed in osm-3 animals. In kap-1; sql-1(tm2409) animals OSM-3::GFP, DAF-10::GFP (complex A), CHE-13::GFP (complex B), and XBX-1::GFP (dynein) moved at 1.00–1.19 µm/second (Table 1), similar to velocities observed in kap-1 animals. These data indicate that the intrinsic velocities of the kinesins are not altered in sql-1 mutant animals.

Our data indicate that in sql-1(tm2409) animals, kinesin-II and OSM-3 are partially separated. Based on the observed speeds, we calculated that ∼38% of OSM-3 motor proteins moved independently from kinesin-II and that 61% of kinesin-II motor proteins moved independently from OSM-3.

Of the five non-kinesin IFT-components tested, three moved at the same speed as OSM-3, and two moved at ∼0.7 µm/second. This suggests that in sql-1(tm2409) animals also the IFT complex has disintegrated, where CHE-11, OSM-1, IFT-20 and XBX-1 associate with OSM-3 particles, and together with DAF-10 and CHE-13 associate with the kinesin-II/OSM-3 complex. An alternative explanation would be that DAF-10 and CHE-13 associate with all three possible particles (slow kinesin-II particles, fast OSM-3 particles and the kinesin-II/OSM-3 complex), averaging in a velocity of ∼0.7 µm/second. To distinguish between these two possibilities we plotted the distributions of the velocities of CHE-13::GFP and DAF-10::GFP IFT events, where the presence of CHE-13 and DAF-10 in three particle types would be visible in the distribution plots as additional peaks, or at least a wider distribution. However, the distributions of the CHE-11::GFP and DAF-10::GFP IFT events in wild-type and sql-1(tm2409) animals were very similar (supplementary material Fig. S5), suggesting that CHE-11 and DAF-10 were only present in the kinesin-II/OSM-3 transported particles. In addition, these results suggest that in sql-1 animals a fraction of kinesin-II moves without further IFT components.

In summary, our data are consistent with a model in which the middle segments of the sensory cilia of sql-1(tm2409) animals contain three types of particles: complete IFT particles, empty kinesin-II and incomplete IFT particles consisting of OSM-3, but not kinesin-II, and at least CHE-11, OSM-1, IFT-20 and XBX-1, but not DAF-10 and CHE-13 (Fig. 5).

Since in gpa-3QL, gpa-3 and sql-1(tm2409) animals IFT is affected, we examined IFT in double mutants, to study their genetic interactions. In the middle segments of sql-1(tm2409); gpa-3 animals KAP-1::GFP and XBX-1::GFP moved at velocities similar to those observed in sql-1 and gpa-3 single mutant animals. OSM-3::GFP moved at the same speed as in sql-1(tm2409) animals (Table 1). These data suggest that, genetically, sql-1 acts downstream of gpa-3.

In the middle segments of sql-1(tm2409); gpa-3QL animals, KAP-1::GFP moved at a velocity similar to that observed in the middle segments of sql-1(tm2409) and gpa-3QL animals, OSM-3::GFP moved at a velocity similar to that observed in gpa-3QL animals and slower than in sql-1(tm2409) animals, XBX-1::GFP moved significantly slower compared to the velocities observed in sql-1(tm2409) and gpa-3QL animals (Table 1). Possibly, effects of the mutations in sql-1(tm2409) and gpa-3QL on IFT are additive.

Interestingly, in the distal segments of sql-1(tm2409); gpa-3QL animals all measured components of the IFT complex (OSM-3, DAF-10, CHE-13 and XBX-1) moved slower than the expected 1.15 µm/second, the velocity of the IFT complex in the distal segments of wild-type, sql-1(tm2409) and gpa-3QL animals (supplementary material Table S1). This could be explained by either entry of ‘slow’ kinesin-II into the distal segment, as previously observed in dyf-5 mutant animals (Burghoorn et al., 2007), or by a change in the intrinsic velocity of OSM-3. First, we examined the localization of KAP-1::GFP in sql-1(tm2409); gpa-3QL animals, but we did not observe any KAP-1::GFP in the distal segment (supplementary material Fig. S6), suggesting that the slower velocities are not caused by entry of kinesin-II into the distal segment.

Second, we examined whether the intrinsic velocity of OSM-3 was affected. Therefore, we measured the velocity of OSM-3 in kap-1, kap-1; sql-1, kap-1; gpa-3QL, and kap-1; sql-1; gpa-3QL mutant animals. In all these animals the IFT complex is transported only by OSM-3 in both segments. As expected, OSM-3::GFP moved at a velocity of 1.17–1.23 µm/second in both segments of kap-1, kap-1; sql-1 and kap-1; gpa-3QL animals (Fig. 6; Table 1; supplementary material Table S1). However, in kap-1; sql-1; gpa-3QL animals OSM-3::GFP moved significantly slower than in kap-1, kap-1; sql-1 and kap-1; gpa-3QL animals (Fig. 6; Table 1; supplementary material Table S1). These data indicate that in the absence of SQL-1 and presence of GPA-3QL the motility of OSM-3 is affected. The reduced speed of OSM-3 in sql-1; gpa-3QL double mutant animals likely affects the speed of the whole IFT machinery, and might explain the lower velocities observed in the sql-1(tm2409); gpa-3QL double mutants.

In this study we identified and characterized SQL-1 (suppressor of gpa-3QL 1), the C. elegans homologue of mammalian GMAP210. Loss of sql-1 suppressed the ciliary length defect of gpa-3QL animals and, probably as a result of that, also the dye filling defect of these animals. Loss-of-function of sql-1 by itself did not affect cilia length, however overexpression of sql-1 increased cilia length. We found that SQL-1 is ubiquitously expressed and localizes to the Golgi apparatus. Loss of sql-1 affected the localization of the Golgi protein AMAN-2, suggesting a mild effect on Golgi integrity. Importantly, mutation of sql-1 affected the IFT machinery and partially uncoupled kinesin-II and OSM-3. Genetic epistasis analyses suggested that sql-1 and gpa-3 function in the same genetic pathway.

GMAP210 localizes to the Golgi apparatus. Two domains have been identified that mediate this localization, the N-terminal ALPS domain, and the C-terminal GRAB domain (Gillingham et al., 2004; Follit et al., 2008; Cardenas et al., 2009). SQL-1 also localizes to the Golgi apparatus, as determined using immunofluorescence and GFP fusion constructs. In the sensory neurons, SQL-1 is mostly restricted to the cell bodies, with some SQL-1 spots in the dendrite, but never in or close to the cilium. Interestingly, not only full length SQL-1 fused to GFP localizes to the Golgi apparatus, but also a fusion protein containing the 302 most N-terminal amino acids localizes there. Since SQL-1 does not seem to have an ALPS domain and the GRAB domain is not present in the shortened fragment, Golgi targeting of this fragment must be mediated by a different, still uncharacterized SQL-1 domain.

Several functions have been ascribed to GMAP210. One of these proposed functions is maintaining the structure of the Golgi apparatus. Fragmentation of the Golgi apparatus has been observed in cultured HeLa cells after RNAi against GMAP210, as well as in primary dermal fibroblasts isolated from GMAP210 −/− mice (Ríos et al., 2004; Yadav et al., 2009; Smits et al., 2010). Interestingly, others did not observe any structural Golgi defects in similar experiments (Gillingham et al., 2004; Follit et al., 2008). We visualized the Golgi apparatus of C. elegans specifically in the ASI neurons with AMAN-2::GFP, and observed more dispersed or even diffuse AMAN-2::GFP fluorescence in sql-1(tm2409). This suggests that in C. elegans SQL-1 is involved in maintaining Golgi integrity.

Follit and colleagues have shown that GMAP210 interacts with complex B protein IFT20 and plays a role in sorting and/or transport to the ciliary membrane (Follit et al., 2008). They found that mouse embryonic kidney cells lacking GMAP210 could still form cilia, showing that GMAP210 is not required for cilia assembly, although these cilia were slightly shorter. Consistent with a role in vesicular transport to the cilium, it was shown that in the absence of GMAP210 polycystin-2 levels in the cilium were lower. In C. elegans GMAP210 and IFT20 do not interact, since SQL-1 and IFT-20 did not co-immunoprecipitate and we could not detect any IFT-20::GFP at the Golgi apparatus, where mammalian IFT20 resides in cultured cells (Follit et al., 2006). In addition, Kaplan and colleagues have previously shown that the ciliary membrane protein ODR-10 still localizes to the sensory cilia of ift-20 mutant animals (Kaplan et al., 2010). Thus, it is very well possible that the role of IFT-20 in trafficking ciliary membrane proteins is not conserved in C. elegans. Mutation of sql-1 does not affect cilia length, where the absence of GMAP210 in cultured mammalian cells results in slightly shorter cilia. Interestingly, animals that overexpress SQL-1 do show a cilia length increase, indicating that also SQL-1 plays a role in the regulation of cilium length.

In the absence of sql-1 the velocities of different components of the IFT machinery are affected. These changes are consistent with a model in which the middle segments of the sensory cilia of sql-1(tm2409) animals contain three types of particles: complete IFT particles, empty kinesin-II and incomplete IFT particles consisting of OSM-3, but not kinesin-II, and at least CHE-11, OSM-1, IFT-20 and XBX-1, but not DAF-10 and CHE-13 (Fig. 5).

The presence of an incomplete IFT particle suggests a destabilization of the IFT complex in sql-1(tm2409) animals. CHE-11/IFT140, one of the proteins present in the incomplete IFT particle of sql-1(tm2409) animals, is part of the ‘core’ of complex A (Ou et al., 2007; Mukhopadhyay et al., 2010; Behal et al., 2012). In addition, CHE-11/IFT140 interacts with complex B protein OSM-1/IFT172, which is also present in the incomplete IFT particle of sql-1(tm2409) animals (Follit et al., 2009). Possibly, the destabilized IFT particles in sql-1 mutants reflect the “core” of the IFT machinery. Further analyses using dual colour imaging to visualize the motility of additional IFT components simultaneously is necessary to reveal the makeup of the IFT particles in sql-1 mutant animals.

At present, it is unclear how mutation of sql-1 affects the stability of the IFT complex. Our data and those presented by Follit et al. (Follit et al., 2008) indicate that SQL-1/GMAP210 only resides in the Golgi apparatus and not in the cilium. Therefore, it is likely that the IFT defect is caused by a sorting defect of vesicles destined for the cilium; e.g. vesicles could exit the Golgi apparatus prematurely, or fail to transport some proteins required for IFT or maintenance of the stability of the IFT complex. Other proteins required for the stability of the IFT complex include BBS and NPHP proteins. However, mutation of these proteins causes uncoupling of complex A and B (BBS proteins) (Ou et al., 2005; Ou et al., 2007) or mainly affect the complex B protein OSM-6/IFT52 (Jauregui et al., 2008). Finally, the presence of other kinesin such as KLP-6 (Morsci and Barr, 2011) or another still unidentified ciliary kinesin, could also explain the differences in the velocities observed in sql-1 mutant animals.

We identified sql-1 as a suppressor of the dye filling defect of gpa-3QL animals. In agreement with the effect on dye filing, mutation of sql-1 also suppressed the effect of gpa-3QL on cilium length. sql-1 did not suppress the effects of gpa-3QL on the IFT machinery, but instead introduced additional defects. Still, the suppression of the cilia length defect in gpa-3QL animals by sql-1 can be explained by its effect on the IFT machinery. In gpa-3QL animals, kinesin II and OSM-3 were partially separated, while complex A and B particle proteins moved at an intermediate velocity (Burghoorn et al., 2010). These data are consistent with a model where the middle segments of gpa-3QL animals contain complete IFT particles, IFT particles lacking kinesin-II and IFT particle lacking OSM-3 (supplementary material Fig. S7). We speculate that this results in a decreased delivery of particles to the distal segment and reduced ciliary length. In sql-1 mutant animals, IFT particles are predominantly transported by OSM-3. Therefore, inactivation of sql-1 in gpa-3QL animals would cause cargo to be transported predominantly by OSM-3, resulting in an increase of particles reaching the distal tip, and thus suppression of the cilium length defect.

From our and previously reported data it is clear that SQL-1/GMAP210 functions at the Golgi apparatus in routing of proteins to the cilium. Absence of SQL-1 leads to alterations in the IFT machinery, uncoupling of the two kinesins and destabilization of complex A and B. How these effects come about is unclear, but it seems likely that absence of SQL-1 in the Golgi affects the routing of one or more proteins that are required in the cilium for proper IFT.

Worms were cultured using standard methods (Brenner, 1974). Wild-type animals were C. elegans Bristol N2. Alleles used in this study were sql-1(gj202)III, sql-1(tm2440)III, sql-1(tm4995)III, sql-1(tm2409)III, gpa-3(pk35)V, gpa-3QL(syIs25)X, gpa-3QL(syIs24)IV, ift-20(gk548)I, osm-3(p802)IV, kap-1(ok676)II. sql-1 alleles tm2409, tm2440 and tm4955 were obtained from NBP-Japan. Published transgenes used were: gpa-4::gfp, gpa-15::gfp, osm-3::gfp, kap-1::gfp, che-11::gfp, osm-1::gfp, che-13::gfp, xbx-1::yfp, p_glr-1::aman-2::yfp, p_arl13::gfp::rab-8 (Jansen et al., 1999; Signor et al., 1999; Qin et al., 2001; Rolls et al., 2002; Haycraft et al., 2003; Schafer et al., 2003; Snow et al., 2004; Kaplan et al., 2010). Also, several new constructs were generated. osm-3::gfp was created by fusing a 1 kb osm-3 promoter fragment to osm-3::gfp (gift from Piali Sengupta), kap-1::gfp was created by fusing a 0.8 kb kap-1 promoter fragment to kap-1::gfp (gift from Piali Sengupta), P_osm-3::daf-10::gfp was generated by replacing the osm-3 ORF in osm-3::gfp with the daf-10 cDNA (gift from Piali Sengupta), ift-20::gfp was generated by fusing a 3.5 kb ift-20 fragment, containing 2.5 kb upstream sequence and the ift-20 gene, in frame with gfp using fusion PCR (Hobert, 2002). The translational sql-1::gfp and p_gpa-4::sql-1::gfp constructs were made by fusing the sql-1 promoter fragment (2.2 kb) or the gpa-4 promoter fragment (2.5 kb) with the sql-1 cDNA (obtained by combining three sql-1 cDNA fragments) in pPD95.77 (gift from Andrew Fire). The sql-1 constructs were made by introducing a stop codon in the translational sql-1::gfp construct, at the 3′ end of sql-1. A P_gpa-4::aman-2::yfp was made by fusing the gpa-4 promoter to the translation start site of aman-2 (gift from Tom Rapoport). A P_gpa-4::aman-2::gfp construct was made by subcloning a PCR product of aman-2 into P_gpa-4::gfp in frame with gfp. Sequences of primers used to generate these constructs are given in supplementary material Table S2. Microinjections were performed as described (Mello and Fire, 1995). IFT constructs were injected at concentrations ranging from 20 ng to 50 ng/µl and analyzed for correct expression by imaging IFT particles in wild-type animals. The sql-1 construct was injected in wild-type animals at concentrations up to 150 ng/µl to generate transgenic lines that overexpress sql-1 approximately two- to threefold.

gpa-3QL(syIs25)X animals were mutagenized with 50 mM ethyl methanesulfonate (EMS) to generate random mutations. 100 cultures were started with 12 EMS-treated animals each. The F2 and F3 progeny of the mutagenized animals were subjected to dye filling and 12 animals that took up fluorescent dye were individually picked onto new culture dishes. The progeny of these putative suppressor mutants was subjected to dye filling. This procedure identified nine independent suppressor mutants, including sql-1(gj202). Using SNP mapping (Wicks et al., 2001), we mapped sql-1(gj202) to a region of ∼120 kb on chromosome III. To identify the mutated genes, gpa-3QL(syIs25)X; sql-1(gj202) mutant animals were injected with long range PCR fragments spanning the 35 kbp region containing the predicted genes Y111B2A.4, Y111B2A.5 and Y111B2A.26. The dye filling defect was only restored in animals injected with PCR fragments spanning all three genes, not with each of the genes individually. Sequencing identified an A to T mutation in the predicted gene Y111B2A.5. We used RT-PCR to characterize the ORF of wild-type and sql-1 mutant animals. Rescue was confirmed by injecting 5 ng/µl sql-1 or sql-1::GFP construct into gpa-3QL(syIs25)X; sql-1(tm2409) mutant animals.

Animals were permeabilized, fixed and stained according to standard methods (Finney and Ruvkun, 1990). For SQL-1 rabbit polyclonal antibody production, the N-terminal cDNA of sql-1 (exon 3–8) and the C-terminal cDNA of sql-1 (exon 7–18) were cloned into pGST-Parallel [pGEX4T1 derivate (Sheffield et al., 1999)]. Polyclonal antibodies were generated by collecting serum from rabbits immunized with either the N-terminal or the C-terminal, E. coli produced, purified SQL-1 proteins fused to glutathione S-transferase (immunization performed by Harlan Laboratories). Antibodies were affinity purified prior to immunofluorescence staining (dilution was 1∶200 for both antibodies). Secondary antibodies were goat-anti-rabbit Alexa Fluor 594-conjugated (Molecular Probes, Eugene, OR; 1∶800). Specificity of the antibody was confirmed by the absence of immunoreactivity in sql-1(tm2409) mutant animals. SQV-8 antibody (used at a 1∶100 dilution) was a gift from Jon Audhya. GPA-3 antibody staining was performed as described (Lans et al., 2004). Dye filling was performed using DiO (Molecular probes) as described (Perkins et al., 1986). Antibody staining, localization of fluorescent proteins, cilia morphology and cilia lengths were measured using a Zeiss Imager Z1 microscope with a 100× (NA 1.4) objective.

For immunoblotting lysates of 40 worms were solubilized with Laemmli loading buffer. After electrophoresis on a SDS-PAGE gel, proteins were transferred to membranes and probed with a primary antibody, ECL-horseradish peroxidase-conjugated secondary antibody, and detected with ECL Plus Western blot detection system (Amersham). The following primary antibodies were used: rabbit polyclonal anti-SQL-1 (1∶1000), anti-GPA-3 (1∶1000), and anti-GFP (NeuroMab, 1∶40). The following secondary antibodies were used: ECL-horseradish peroxidase-conjugated anti-rabbit (Dako, 1∶10,000) and horseradish peroxidase-conjugated anti-mouse (GE Healthcare, 1∶10,000).

Strains used were Bristol N2 and gj2180[ift-20::gfp]. Worms were collected from nine full 10-cm plates and washed three times with M9, once with cold 100 mM NaCl, and resuspended in homogenization buffer (15 mM HEPES pH 7.6, 10 mM KCl, 1.5 mM MgCl2, 0.1 mM EDTA, 0.5 mM EGTA, 44 mM sucrose, 1 mM DTT, protease inhibitor). Worms were lysed by sonication, and after clearing of the lysates by centrifugation, they were loaded onto antibody coupled protein A Sepharose CL-4B beads (GE Healthcare). Rabbit polyclonal anti-SQL-1 and rabbit polyclonal anti-GFP (Abcam) antibodies were bound according to manufacturer's instructions. Lysates were incubated overnight at 4°C, washed 5 times with immunoprecipitation buffer (20 mM Tris-HCl pH 7.5, 150 mM NaCl, 5 mM MgCl2, 5% glycerol, 0.1% NP40, protease inhibitor). Afterwards, lysate, supernatant, and bound fraction were analyzed by immunoblotting.

Live imaging of the GFP-tagged IFT particles was carried out as described (Snow et al., 2004). Images were acquired on a Zeiss confocal microscope CLSM510 with a 63× (NA1.4) 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.

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 µm/second, OSM-3::GFP speed in osm-3 animals, and KAP-1::GFP speed in kap-1 animals) and the fraction that moves separately (at 1.17 µm/second, OSM-3::GFP speed in kap-1 animals, or 0.49 µm/second, KAP-1::GFP speed in osm-3 animals). The formula used to calculate the fraction of OSM-3 that moves separately (x) in sql-1 mutants is: x_osm-3 = (v_sql-1−v_together)/(v_max−v_together), where v_sql-1 is the OSM-3::GFP speed measured in a sql-1 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_sql-1−v_together)/(v_max−v_together), where v_sql-1 is the KAP-1::GFP speed measured in a sql-1 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.

We thank A. van der Vaart for helpful comments on the manuscript; NBP-Japan, the Caenorhabditis Genetics Center, P. Sengupta, J. Audhya, A. Fire, T. Rapoport, O. Blacque, and N. Tavernarakis for C. elegans strains, constructs and antibodies.