MKS and NPHP modules cooperate to establish basal body/transition zone membrane associations and ciliary gate function during ciliogenesis

Primary cilia protrude from most mammalian cells and modulate sensory processes, including chemo-, mechano-, and photo-reception (Fliegauf et al., 2007). Cilia regulate various signaling pathways during embryonic development and are needed for normal postnatal tissue homeostasis (Gerdes et al., 2009). Mutations disrupting ciliary functions cause human disorders (ciliopathies) that collectively affect nearly all tissues/organs (Sharma et al., 2008). A nonexhaustive list of ciliopathies includes Meckel-Gruber syndrome (MKS), nephronophthisis (NPHP), Bardet-Biedl syndrome (BBS), Joubert syndrome (JBTS), Senior-Løken syndrome (SLSN), Leber congenital amaurosis (LCA), polycystic kidney disease (PKD), and oral-facial-digital syndrome (OFD). These disorders present with variable but overlapping clinical phenotypes that encompass polycystic kidneys, liver fibrosis, skeletal anomalies, sensory impairment, and brain/nervous system deformities (Fliegauf et al., 2007).

At least 35 loci have been identified in ciliopathy patients, some of which contribute to multiple seemingly distinct syndromes (Baker and Beales, 2009). Many of these genes encode proteins that localize to the basal body (BB)—a centriolar structure universally required for extending the microtubule-based ciliary axoneme—or to an adjacent domain, termed “transition zone” (TZ) in most cilia, or “connecting cilium” in photoreceptors (Horst et al., 1990; see schematic of BB-TZ-cilia structures in Fig. 1 A and relevant disease proteins in Table S1 A). Within the BB-TZ region are subdomains that include transitional fibers (TFs) and Y-links. TFs form a pinwheel-like structure, of unknown protein composition, that links the BB to the proximal ciliary membrane. The Y-links of the TZ connect—via high-affinity linkages—axonemal microtubules to the membrane at the ciliary necklace, a proteinaceous decoration of the TZ membrane (Muresan and Besharse, 1994). Together, the TFs and TZ are proposed to form a gate (Rosenbaum and Witman, 2002; Satir and Christensen, 2007) that excludes vesicles from cilia, prevents unwanted diffusion of membrane proteins into cilia, and selectively regulates protein ciliary entry and exit (Fig. 1 A).

Axoneme elongation is thought to initiate when the mother centriole docks with a membrane either at the cell surface or a ciliary vesicle in the cytosol (see Fig. 10 D; Sorokin, 1962). Full extension of the axoneme then relies on an intraflagellar transport (IFT) machinery that uses kinesin and dynein motors and associated subcomplexes (IFT-A, IFT-B, and BBSome) to traffic ciliary cargo from TFs to the cilium tip and back (Fig. 1 A; Silverman and Leroux, 2009). In contrast to the extensive characterization of IFT-mediated axoneme extension, the components and mechanism involved in BB/TZ membrane association and establishment of the ciliary gate remain virtually unknown. Based on knockdown studies in mammalian cells, the ciliopathy proteins MKS1 and MKS3 have been implicated in BB migration/docking and thus ciliogenesis (Dawe et al., 2007). However, these defects are absent from rodent Mks1 or Mks3 mutants (Tammachote et al., 2009); thus, the role of these and most other BB/TZ-associated ciliopathy proteins remains unclear. Recently, the ciliopathy protein CEP290 was localized to TZ Y-links, and its disruption in Chlamydomonas altered the ciliary composition of IFT components and other proteins (Craige et al., 2010); Caenorhabditis elegans NPHP-1 and NPHP-4 have also been proposed to act in ciliary gating (Jauregui et al., 2008). Whether additional TZ or IFT proteins are similarly involved in regulating ciliary gating, and the mechanism by which they perform these functions, is not known.

Here, we used C. elegans to elucidate the functions of eight conserved proteins, six of which are MKS/NPHP associated. We find that MKS-5/RPGRIP1L interacts with two distinct TZ functional modules, MKS/MKSR and NPHP, consisting of MKS-1/MKSR-1/MKSR-2/MKS-3/MKS-6 and NPHP-1/NPHP-4 proteins, respectively. Functional interactions between different MKS module components and the NPHP module are essential for an IFT-independent early stage of ciliogenesis, namely docking/anchoring of the BB/TZ to the membrane. Moreover, the two modules restrict inappropriate accumulation of membrane-associated proteins inside cilia. Our findings help to comprehensively define an interaction network of ciliopathy-associated proteins and allow us to propose for the first time a unified model for the function of diverse MKS/NPHP proteins, in which the MKS and NPHP modules altogether enable associations between microtubules and the ciliary membrane; this ciliogenic event coincides with construction of the ciliary gate that establishes the specialized compartment.

To gain insights into MKS/MKSR/NPHP protein functions, we first defined their respective subcellular localization in C. elegans sensory neurons. Using fluorescently tagged proteins, we detect MKS/MKSR/NPHP proteins in a region corresponding to the TZ (adjacent to where IFT proteins concentrate at the TFs/BB). This is evident for MKS-1, MKS-1 related-1 (MKSR-1)/B9D1, MKS-1 related-2 (MKSR-2)/B9D2, MKS-3/meckelin, NPHP-1, and NPHP-4 (Winkelbauer et al., 2005; Williams et al., 2008, 2010; Bialas et al., 2009), as well as MKS-5/RPGRIP1L and MKS-6/CC2D2A, previously uncharacterized in C. elegans (Fig. 1, B–D; Fig. 2). By transmission electron microscopy (TEM), the ∼0.8-µm long TZ region containing Y-links is distal to the TFs, which sit at the ciliary base just inside the dendritic tip (Fig. 1 A; Perkins et al., 1986). These data suggest that the TZ represents a common site of dysfunction in MKS/NPHP ciliopathy patients; however, the functions of MKS/MKSR/NPHP proteins at the TZ remain undetermined. Notably, based on Hidden Markov Model profiling and structure/fold predictions, several TZ proteins (MKS-1, MKSR-1, MKSR-2, MKS-5, and MKS-6) share a related C2/B9 motif (Fig. 1, E–G; Table S1 C). This motif is predicted to bind Ca²⁺/lipids and participate—similar to synaptotagmin—in membrane/vesicle trafficking and fusion (Nalefski and Falke, 1996). The presence of this motif in multiple-ciliopathy TZ proteins raises the possibility that they perform a shared function at the TZ.

To examine potential ciliary roles of the uncharacterized C2 domain–containing MKS-5 and MKS-6 proteins, we first analyzed mks-5 and mks-6 mutants (Fig. 3 A). Similar to mutations disrupting B9 domain genes (mks-1, mksr-1, and mksr-2; Williams et al., 2008; Bialas et al., 2009), mks-6(gk674) mutants have no overt cilia or IFT defects, based on: (1) normal uptake of hydrophobic fluorescent dye (DiI), indicating cilia are present and environmentally exposed (Fig. 3, B and D); (2) normal cilia length, observed by visualizing cilia with GFP-tagged IFT proteins (Fig. 4, B and C; Fig. 5, A, A′, B, and B′); and finally, (3) normal IFT rates, based on in vivo time-lapse microscopy (Table S1 D). TEM analysis verifies the absence of gross cilia ultrastructure anomalies in mks-6(gk674) mutants, except for occasional loss of axoneme distal segments (Fig. 6). mks-6(gk674) mutants have a weak osmotic avoidance phenotype comparable to nphp-4(tm925) mutants (Fig. 4 D) but do not exhibit increased lifespan, unlike other cilia gene (e.g., IFT) mutants (Table S1 E). Analysis of mks-5(tm3100) mutants uncovered moderate defects in dye filling and phasmid (tail) cilia length (Fig. 3, C and D); however, this was not accompanied by prominent changes in either osmotic avoidance, distribution of IFT proteins to BBs/cilia (Fig. 4, A and C; Fig. 5, A, A′, B, and B′), or IFT rates (Table S1 D).

Previously, we found that B9 domain proteins (of the MKS/MKSR module) functionally interact with NPHP-1 and NPHP-4 (NPHP module) to facilitate ciliogenesis (Williams et al., 2008). To determine if similar genetic interactions occur between MKS-5 and MKS-6 and the MKS/MKSR or NPHP modules, we examined mks-5 and mks-6 mutations in combination with mks/mksr and nphp mutations. Compared with wild-type, mks-5, mks-6, or nphp-4 single mutants, both mks-5;nphp-4 and mks-6;nphp-4 double mutants exhibit strong dye-filling (Dyf) and osmotic avoidance (Osm) phenotypes (Fig. 3, B–D; Fig. 4 D). However, dye filling is normal in mks-6;mks-1, mks-6;mksr-1, mks-6;mks-3, and mks-6;mksr-2 double mutants (Fig. 3, B and D). These data suggest that mks-6 genetically interacts with the NPHP module and is associated with the MKS/MKSR module. Interestingly, combining mks-5 mutants with mksr-1, mksr-2, or mks-6 mutations also results in a stronger Dyf phenotype (Fig. 3, C and D), suggesting that mks-5 genetically associates with not only the NPHP module but also the MKS/MKSR module.

To study in greater detail the ciliary defects of single and double mks/mksr/nphp mutants, we used GFP-tagged ciliary (IFT) markers. We uncovered in double mutants multiple phenotypes, including missing or shorter cilia (Fig. 4 C; Fig. 5, A, A′, B, and B′), misoriented cilia (not depicted), and improperly positioned BB/TF regions within the amphid channel and with respect to the phasmid cell body (Fig. 4 C; Fig. 5, C, C′, D, and D′; Williams et al., 2008). Although there is a low penetrance of these phenotypes in nphp-4 single mutants, they were consistently observed in mks-6;nphp-4, mks-5;nphp-4, and mks-6;nphp-1 double mutants (Fig. 4 C; Fig. 5).

To assess the nature of the ciliary ultrastructural defects in mks/mksr/nphp double mutants, we used TEM. Strikingly, we observed profound and largely indistinguishable phenotypes across all double mutants analyzed. Unlike wild-type animals and mks-6 or nphp-4 single mutants, which show normal TZs (Fig. 6, A and B; Fig. 7; Jauregui et al., 2008), mks-6;nphp-4 and mks-5;nphp-4 double mutants exhibit mispositioned and disrupted TZ regions, with clear disconnections between the BB/TZ region and membrane, accompanied by missing Y-links (Fig. 6 C; Fig. 7; Fig. S1, Fig. S2); hence, their BB/TZ regions are not properly anchored at the distal end of the dendritic membrane. We also observed missing, misplaced, or shorter axonemes, and variably missing TFs. In perhaps the most severe instance of morphological abnormality, mks-5;nphp-4 double mutants had incomplete TZ microtubule rings (Fig. S2). Phenotypes nearly identical to mks-6;nphp-4 mutants were observed in the mksr-1;nphp-4 strain (Fig. 7; Fig. S3). The various cilia defects uncovered by TEM are quantified in Table I. Notably, all TZ/cilia anomalies are in addition to open tubules observed in the nphp-4 mutant (Jauregui et al., 2008) and are distinct from those of IFT and other cilia mutants, in which TZ regions are intact (Perkins et al., 1986).

Given the severe ciliary phenotypes of mks-5;nphp-4 and mks-6;nphp-4 strains, we wondered if mks-1;nphp-4 and mks-3;nphp-4 mutants, which have milder Dyf phenotypes (Williams et al., 2008, 2010), would present with similar—yet partial or more specific—ultrastructure defects. Although similar ciliary defects were observed in the latter mutants (Fig. 7; Fig. S4, Fig. S5), some of their TZ regions are in normal proximity to the ciliary membrane but often lack connecting Y-links; this raises the possibility that MKS/MKSR/NPHP proteins comprise components of these structures or are required for their stability.

Our data indicate that MKS/MKSR/NPHP proteins are important for BB/TZ interactions with the membrane, which likely occur during ciliogenesis before IFT-dependent axoneme extension, suggesting a role in an early step of the ciliogenic pathway (Sorokin, 1968). To test if this early step depends on IFT, we looked for IFT defects using kymograph analyses in mks/mksr/nphp double mutants where a small proportion of cilia still form. Remarkably, IFT was largely unaffected in these double (as well as single) mutants (Table S1 D; Bialas et al., 2009), in striking contrast to C. elegans IFT mutants where disruption of anterograde or retrograde IFT occurs (Hao and Scholey, 2009), or BBS mutants (bbs-1/7/8), where stability/association of IFT subcomplexes A and B is compromised (Blacque et al., 2004; Ou et al., 2007). Because no TZ phenotypes observed in our study are apparent in IFT or BBS mutants (Perkins et al., 1986; unpublished data), and because IFT is largely unaffected, we propose that NPHP and MKS protein modules act before IFT-driven axoneme formation. From this model we predict that the TZ should be established properly when IFT is disrupted. This was validated in IFT and BBS mutants, where MKS-6 (Fig. 8 F) and other MKS/NPHP proteins (unpublished data) localize normally to the TZs of cilia lacking fully formed axonemes. Together, these data support a model where TZ proteins function in an early ciliogenic step (BB/TZ association with membrane) followed by a subsequent IFT/BBS-dependent step in axoneme elongation/trafficking.

Our TEM analyses provide fundamental insights into defects caused when TZ proteins (MKS-1, MKSR-1, MKS-3, MKS-5, and MKS-6) are disrupted jointly with NPHP-4 (Figs. 6 and 7; and Figs. S1–S5). The ultrastructural defects overlap greatly between mutants, implying a shared function for MKS/MKSR/NPHP proteins. Further support comes from biochemical data linking many ciliopathy proteins in shared macromolecular complexes (Table S1 A). To provide further evidence for shared/modular functions, we queried whether disrupting particular TZ proteins affected the localization of others. Using this approach, we rule out MKS-1, MKS-3, MKS-6, and NPHP-1 as critical docking proteins, as their disruption had no effect on localization of other TZ proteins (Fig. 8, B–F; Winkelbauer et al., 2005; Williams et al., 2008, 2010). In contrast, disrupting MKSR-1, MKSR-2, or MKS-5 results in TZ delocalization of MKS-6 (Fig. 8 B) and MKS-3 (Fig. 8 C, Fig. 9 L). Furthermore, mks-5 mutants failed to properly localize MKS-1, MKSR-1, and MKSR-2, suggesting a key role of MKS-5 in docking proteins at the TZ (Fig. 8 C). Notably, NPHP-1 and NPHP-4 localization was also altered in mks-5 mutants (Fig. 8 C), albeit differently; instead of failing to anchor at the TZ, NPHP-1 and NPHP-4 occupied a smaller region than wild-type animals (TZ length of 0.65 µm, n = 19 in mks-5 mutants vs. 1.05 µm, n = 23 in controls; t test P < 0.0001). MKS-5 was unaffected upon disruption of MKSR-2 or NPHP-4, which are otherwise required for TZ docking most other MKS/MKSR proteins and NPHP-1, respectively (Fig. 8 D; Winkelbauer et al., 2005; Williams et al., 2008). Moreover, MKS-5 is still TZ localized in mks-6;nphp-4 double mutants in which ciliary microtubule–membrane attachments are disrupted (compare Fig. 8 D with Fig. 8 E, in which transmembrane MKS-3 is predictably lost in the same double mutant). Thus, we conclude that MKS-5 localizes to the TZ independently of other TZ proteins tested and performs a central role as a scaffold for anchoring other MKS/MKSR and NPHP module proteins.

Given the presence of C2 domains in MKS-5 and MKS-6 and related B9 domains in MKS-1, MKSR-1, and MKSR-2, we hypothesized that these proteins may be required for vesicle trafficking and/or docking/fusion of vesicles harboring ciliary cargo at the TZ (Nalefski and Falke, 1996). Thus, we used a well-established odorant receptor (ODR-10; Sengupta et al., 1996) as a functional marker for vesicular trafficking to cilia. In wild-type animals, ODR-10 (expressed under the str-1 promoter) concentrates at the branched AWB cilium membrane (Fig. 9, A and B). In single TZ mutants, no apparent defects in cilia structure or ODR-10 localization are seen (Fig. 9, C–G). Although AWB cilium morphology in TZ double mutants is compromised, ODR-10 localization to its ciliary membrane appears normal (Fig. 9, H–K). These findings suggest that TZ proteins do not play an essential role in ciliary cargo-associated vesicle trafficking to, and docking/fusion steps at the base of cilia.

Our analyses uncovered important roles for MKSR-1, MKSR-2, and MKS-5 in TZ localization of the transmembrane protein MKS-3 (Fig. 8 C; Fig. 9 L). Interestingly, in addition to being lost from the TZ in mksr-1, mksr-2, and mks-5 mutants, MKS-3 accumulates along ciliary axonemes (and at nearby dendritic tips; Fig. 9 L). This indicates that MKSR-1, MKSR-2, and MKS-5 restrict MKS-3 to the TZ membrane. Based on these observations, we wondered if other membrane-associated proteins (in particular, non-TZ-associated proteins) also accumulate in cilia when TZ proteins are disrupted. We examined RPI-2, the C. elegans orthologue of human X-linked retinitis pigmentosa 2 (RP2; Schwahn et al., 1998; Chapple et al., 2000). In wild-type C. elegans, GFP-tagged RPI-2 associates with membrane in sensory neurons, enriched as a ring-like pattern near or at BB/TFs (Blacque et al., 2005), as in trypanosomes (Stephan et al., 2007), but not within cilia (Fig. 9 M). In contrast, we observed that in TZ mutants, RPI-2 accumulates inside cilia (Fig. 9 M). We acquired almost identical results with the transmembrane protein TRAM-1a (Fig. 9 N), which in wild-type animals is found at dendritic tips and excluded from cilia (Bae et al., 2006). Together, our data showing abnormal accumulations of MKS-3, RPI-2, and TRAM-1a in TZ mutant cilia indicate that, in addition to their role in early ciliogenesis, most if not all TZ proteins normally function to maintain a boundary at the cilium base, establishing the TZ as a bona fide ciliary gate.

There are at least 35 genes associated with primary cilia disorders (Baker and Beales, 2009). Several, including BBS genes, are linked to IFT (Blacque and Leroux, 2006; Beales et al., 2007), but most lack clearly assigned molecular functions. Many of these components display genetic and/or physical interactions and are associated with overlapping clinical ailments, suggesting involvement in a common cellular process.

Our findings indicate that C. elegans proteins implicated in many ciliopathies, including MKS, NPHP, JBTS, and LCA, function at the TZ and are required for BB and TZ attachment to the membrane and establishing a ciliary gate early in ciliogenesis (Fig. 10 D). Assignment of individual proteins to a particular TZ module (MKS/MKSR or NPHP) simplifies our understanding of the associated ciliopathies. Although previously seen as diseases of complex and seemingly unrelated molecular etiology, our data indicate that MKS and NPHP are likely disorders of macromolecular complexes sharing a common biological function. This is similar to the finding that several BBS proteins are constituents of the multimeric BBSome, which functions as a ciliary coat complex (Nachury et al., 2007; Jin et al., 2010). Indeed, a collective function of TZ proteins is supported by fragmented but increasing biochemical data in mammalian systems (Fig. 10 A; Table S1 A). Other functional modules/complexes may have disparate yet essential ciliary functions that are associated with equivalent ciliopathies; one likely includes NPHP2/inversin, NPHP3, and NEK8, which localize at the Inv ciliary compartment just distal to the TZ (Shiba et al., 2010).

We propose that MKS/MKSR and NPHP proteins form a hierarchical network comprised of distinct modules with partial functional redundancy (Fig. 10, B and C). The rationale for this modular hypothesis stems from our findings that in C. elegans, synthetic ciliary defects result from mutations in at least two TZ genes—but not any two genes (Fig. 3, B–D; Figs. 4–7). For example, even triple B9 gene (mks-1, mksr-1, and mksr-2) or nphp-1;nphp-4 double mutants lack prominent (additive) ciliary defects (Jauregui et al., 2008; Bialas et al., 2009). In contrast, disrupting any B9 gene together with nphp-1 or nphp-4 results in severe cilia anomalies (Williams et al., 2008). Other mutant combinations such as mks-1;mks-3, or mks-6;mksr-2, do not noticeably disrupt cilia, whereas the same four individual mutants show strong genetic interactions with nphp-1 or nphp-4 (Fig. 3, B–D; Figs. 4–7). Thus, we group MKS-1, MKSR-1, MKSR-2, MKS-3, and MKS-6 into an MKS/MKSR module and NPHP-1 and NPHP-4 into an NPHP module (Fig. 10 C). Importantly, MKS-5 represents an exception to this module assignment. Disrupting mks-5 together with mksr-1, mksr-2, mks-6 (MKS/MKSR module), or nphp-4 (NPHP module) compromises cilia morphology (Fig. 3, B–D). Moreover, mks-5 mutants fail to localize MKS-1, MKSR-1, MKSR-2, MKS-3, and MKS-6 (i.e., the entire known MKS/MKSR module) to the TZ; the mks-5 mutation also partially alters NPHP-1 and NPHP-4 distribution at the TZ (summarized in Fig. 10 B). Based on our data, we propose that the MKS/MKSR and NPHP modules are functionally linked through MKS-5 (Fig. 10 C). Whether the modules are physically associated together at the TZ remains to be examined; however, the established biochemical interaction between mammalian MKS5/RPGRIP1L and NPHP4 (Roepman et al., 2005) indicates this is likely. Further, the likelihood that MKS-5 links multiple functional modules is supported by the finding that human RPGRIP1L mutations contribute to at least six distinct ciliopathies, including MKS and NPHP (Zaghloul and Katsanis, 2010).

The genetic interactions we observe in C. elegans also likely exist between TZ genes in mammals. For example, RPGRIP1L and CEP290 heterozygous variants are linked to increased phenotypic pleiotropy in some NPHP patients in which primary disease symptoms are caused by mutations in other genes (Tory et al., 2007; Khanna et al., 2009). Our findings may also help explain data reported by Hoefele et al. (2005), who screened human NPHP patients and determined that single heterozygous mutations in NPHP4 were over three times more prevalent than two recessive mutations. Based on our genetic interaction studies in C. elegans, we predict that the ailments in these patients result from interactions between mutations in other TZ module genes. With the growing number of available MKS and NPHP rodent models, such possibilities could be formally examined by assessing the consequences of combining mutant alleles of multiple TZ genes.

Intriguingly, proteins seemingly absent in C. elegans or Drosophila have recently been implicated in BB- or TZ-associated functions in other organisms. These include the Ahi1 ciliopathy protein, which binds RAB8 and is proposed to play a role in polarized membrane trafficking (Hsiao et al., 2009), as well as OFD1 and Talpid3, which may assist with docking the ciliary vesicle onto the centriole (Yin et al., 2009; Singla et al., 2010). Another such protein is CEP290, which binds MKS6 (Gorden et al., 2008) and thus likely functions in the MKS or NPHP module. CEP290 was assigned as a component of Y-links and ciliary gate in Chlamydomonas (Craige et al., 2010), and mutations in this gene occur in several ciliopathies. Whether CEP290 functionally interacts with other TZ proteins to facilitate BB/TZ membrane attachments in an early ciliogenic event—as we have shown in this study for several evolutionarily conserved TZ proteins—remains to be determined.

The TZ is an underappreciated ciliary subcompartment, often incorrectly presumed to be one and the same with the adjacent BB. This misconception was inadvertently strengthened by Keller et al. (2005), who identified constituents of the Chlamydomonas “basal body” proteome; however, in these analyses the TZ was co-isolated with the BB. Thus, some TZ proteins (e.g., CEP290, NPHP4) were ascribed to the BB, leading to the widespread belief that NPHP and MKS are diseases of BB dysfunction.

Confusion regarding differences between the C. elegans BB and TZ is also apparent. As C. elegans lacks classical BB microtubule architecture at the base of cilia (Perkins et al., 1986), similar to that observed in mature murine sperm (Manandhar et al., 1998), it has long been speculated that the nematode TZ is “analogous” to the BB. However, several observations suggest this is inaccurate. First, in daf-19 mutants, which lack TZs and cilia, typical centriolar pairs are observed unattached to distal dendritic membranes where cilia would normally form (Perkins et al., 1986). Second, TFs present at the base of C. elegans cilia (Fig. 1 A; Perkins et al., 1986) are evolutionarily conserved structures present at the distal end of all BBs (Silverman and Leroux, 2009). Third, the conserved centriolar/BB protein HYLS-1 localizes in C. elegans within the TF region, just proximal to NPHP-4 at the TZ (Dammermann et al., 2009). Importantly, HYLS-1 in C. elegans and vertebrates is dispensable for centriole function during cell division but critical for ciliogenesis (Dammermann et al., 2009). We propose that the C. elegans TF region, where different IFT proteins concentrate—as shown for IFT52 in Chlamydomonas (Deane et al., 2001)—functions as a bona fide BB that is adjacent to but distinct from the TZ, where MKS/MKSR/NPHP proteins localize (Fig. 1 A, B–D; Fig. 2). Further, our finding that Y-links are completely, or even selectively, lost in MKS/MKSR/NPHP double mutants leaves open the possibility that at least some of these proteins represent structural elements (e.g., Y-links) of the TZ. Indeed, structural and biochemical properties of various MKS/MKSR/NPHP proteins support this possibility; NPHP1 binds microtubules (Otto et al., 2003), the C2/B9 domain proteins are predicted to associate with the inner leaflet of the plasma membrane, and MKS3 is membrane spanning. Taken together, one begins to envision two or more MKS/MKSR/NPHP modules—likely joined by MKS5—as components of the Y-link/ciliary necklace structures that connect axonemal microtubules to the surrounding membrane.

Our analyses revealed that MKS/MKSR and NPHP modules are collectively required for two essential aspects of ciliogenesis, namely membrane anchoring of the BB/TZ and formation of an intact TZ region. Through these processes, the ciliary gate is established (Fig. 10 D). The observed axoneme extension defects in TZ double mutants probably arise as a secondary consequence of anomalies in these early ciliogenic events.

Largely normal IFT rates in TZ mutant strains, and the fact that IFT mutants do not display phenotypes comparable to those of TZ mutants by TEM analysis (Perkins et al., 1986), support the notion that IFT proteins play a role in ciliogenesis distinct from that of TZ proteins, namely building the axoneme and delivering ciliary cargo vs. microtubule-membrane association/stabilization (Fig. 10 D). This difference is reflected in rodent models in which IFT perturbation (cilia ablation) causes early embryonic lethality (at ∼E8.5–10.5), whereas disrupting MKS/MKSR/NPHP proteins leads to less severe ciliogenesis defects and developmental outcomes (Murcia et al., 2000; Jiang et al., 2009; Tammachote et al., 2009). TEM studies of mammalian cells forming cilia (Sorokin, 1962) are consistent with a possible role for the TZ in early ciliogenesis (Fig. 10 D). Specifically, an early ciliogenic step involves the interaction of a Golgi-derived “ciliary” vesicle (CV) with the distal end of the mother centriole/nascent BB to establish microtubule–membrane associations. A “ciliary bud” grows from the centriole, invaginating the CV, which itself appears to grow by fusion with secondary vesicles. In all likelihood, the ciliary bud, visible before BB docking/anchoring with the plasma membrane (Sorokin, 1962; Moser et al., 2010), represents a developing TZ; i.e., the first section of the ciliary axoneme (Fig. 10 D; Rohatgi and Snell, 2010). After docking and fusion with the membrane, the ciliary bud/TZ further extends to form the ciliary axoneme, a step dependent on IFT proteins. Alternatively, in other cell types, the centriole/TZ may dock directly with the plasma membrane, forgoing association with a CV; in this instance, microtubule–membrane associations mediated through the TZ also occur before axoneme extension (Fig. 10 D).

The ciliary gate was long thought to be important in cilia function (Rosenbaum and Witman, 2002), but is only now being studied at the molecular level (Jauregui et al., 2008; Craige et al., 2010). The BB/TZ region is thought to facilitate ciliary gate function as a docking site for proteins destined for the cilium, as a region of selective active transport, and as a diffusion barrier. We demonstrate that individual MKS/MKSR and NPHP-disrupted strains lack a normal ciliary barrier, as evidenced by accumulation of nonciliary proteins within cilia. Alternatively, these proteins may enter cilia at low levels normally and instead require NPHP and MKS proteins for efficient removal from the compartment. Regardless, our data indicate that MKS/MKSR/NPHP proteins establish the TZ as a ciliary gate, and we predict that these TZ proteins likely function in coordination with other mechanisms—including IFT, the Ran-importin system, and BBSome (Bae et al., 2006; Jauregui et al., 2008; Craige et al., 2010; Dishinger et al., 2010; Jin et al., 2010)—to control cilium composition and thus function.

Ciliopathies are genetically heterogeneous but have overlapping phenotypic presentations, suggesting a common cellular mechanism as the basis of their etiology. In this study, we show that eight C. elegans proteins jointly function in establishing connections between the ciliary membrane and axoneme at the TZ, and in formation of the ciliary gate to regulate ciliary membrane composition. These TZ proteins are highly conserved in ciliated organisms (Hodges et al., 2010), suggesting that our model will be widely applicable.

All strains (Table S1 B) used were maintained and cultured at 20°C using standard techniques (Brenner, 1974). Many procedures used in this study are summarized in Inglis et al. (2009). K07G5.3(gk674) and C09G5.8(tm3100) worms were obtained from the knockout consortium (http://celeganskoconsortium.omrf.org) and the National Bioresource Project (Japan), respectively, and outcrossed 5x to wild type. Because the gk674 deletion removes some neighboring xpa-1 gene sequence (Fig. 3 A), we confirmed that the nonciliary function of this DNA repair gene was not abrogated. First, the lifespan of the available xpa-1(mn157) mutant is shorter (P<X) than N2 worms, whereas gk674 mutants are wild type for lifespan (Table S1 E). Second, xpa-1 animals, but not gk674 or N2 worms, possess defective DNA repair as observed by UV irradiation dose-killing curves.

To generate a null allele for K07G5.3, we used PCR to screen for imprecise excision events from ttTi17821 worms, which contain a Mos1 transposon inserted in K07G5.3 (Fig. 3 A). The mutagenesis protocol, modified from Boulin and Bessereau (2007), is as follows. Strain EG1642 (lin-15B(n765)X; oxEx166[HSP::MosTransposase + cc::gfp]) carrying the extrachromosomal Mostransposase under a heat-shock promoter was crossed into strain ttTi17821. 100 young adult worms (ttTi17821; nxEx166[HSP::MosTransposase + cc::gfp]) (P0) were heat-shocked (33°C; 1 h), allowed to recover for 1 h at 20°C, and heat-shocked again before removal to 20°C for 2 h. P0 worms were individually propagated and allowed to lay eggs (F1) for 24 h at 20°C. After removal of P0 worms, F1 worms were allowed to produce F2 progeny. DNA lysates from ∼50% of the F1–F2 worm mixture from each plate were prepared and PCR was performed using primers (5′-GCTACACGAAGACTAGTACTGTTC-3′ and 5′-GCGTTGATGAGAAGAACGAATG-3′) flanking the Mos insertion site. 100 worms from plates containing a K07G5.3 deletion were cloned and screened for the deletion and homozygosed. Using this scheme, we isolated five K07G5.3 alleles, including nx105 (1828-bp deletion + 10-bp insertion; Fig. 3 A). Before analysis, nx105 worms were outcrossed 5x to N2.

Live animals were anaesthetized using 10 mM levamisole (diluted in M9 buffer), mounted on 2% agar pads, and observed using epifluorescence or spinning-disc confocal microscopy performed on an inverted microscope (model 2000U; Nikon) outfitted with a spinning-disk laser apparatus (UltraVIEW ERS 6FE-US; PerkinElmer). Whole-mount immunostaining experiments were performed essentially as described previously (Bobinnec et al., 2000). In brief, gravid adults were cut in M9 buffer containing 15 mM levamisole to release gonads, intestine, and embryos. A coverslip was gently applied and the slide frozen in liquid nitrogen. The coverslip was then removed and the slide immersed in −20°C methanol for 5 min and air-dried for 5 min. Worms were rehydrated in PBS-BSA (1%) for 30 min, and incubated in 3% PBS-BSA with polyglutamylated tubulin antibody GT335 (1:2,500 dilution; a gift from Carsten Janke, Université Montpellier, Montpellier, France) for 1 h. After 3 × 10 min washes with PBS, anti–mouse secondary antibodies (Alexa 594; 1:1,000 dilution; Invitrogen) were applied in 3% PBS-BSA for 1 h. Slides were washed 3 × 10 min with PBS, mounted in ProLong gold (Invitrogen), and observed by confocal microscopy. All images were captured using OpenLab or Volocity (PerkinElmer).

Anterograde IFT rate analyses were performed as described previously (Bialas et al., 2009; Inglis et al., 2009). Individual worms containing GFP-tagged IFT-associated proteins were picked onto 1% agar pads and immobilized using 10–20 mM levamisole. Amphid/phasmid cilia were examined with a 100x 1.35 NA objective and an ORCA AG CCD camera mounted on a Zeiss Axioskop 2 mot+ microscope. Images were acquired using Openlab version 5.0.2 (PerkinElmer), with exposure rates ranging from 150 ms/frame (for OSM-6::GFP–containing strains) to 250 ms/frame (for CHE-11::GFP–containing strains). Openlab LIFF files were imported, with relevant metadata, into ImageJ (http://rsb.info.nih.gov/ij/) using the Bio-formats importer plug-in (Laboratory for Optical and Computational Instrumentation, University of Wisconsin-Madison, Madison, WI). Kymographs for IFT rate analyses were generated using the MultipleKymograph ImageJ plug-in (Inglis et al., 2009).

Ciliary analyses included dye-filling, lifespan, and osmotic avoidance assays.

In brief, fluorescent dye uptake was performed as described previously (Jauregui et al., 2008). L4 larvae were incubated in Vybrant DiI (Invitrogen; 1:1,000-fold dilution of 1 mM stock in M9 buffer) for ∼30 min, allowed to roam on a plate with bacteria for 1 h to clear intestinal dye, and observed by fluorescence microscopy.

Lifespan assays were performed exactly as described in Bialas et al. (2009). Each assay was performed twice with statistically significant results.

Osmolarity avoidance assays were performed as described in Culotti and Russell (1978), testing for the crossing of worms over a ring of high osmolarity solution (8 M glycerol); 50 worms/strain were assayed, and experiments were repeated three times with statistically significant results.

The mks-6(gk674) deletion mutant removes a small portion of the xpa-1 gene. We therefore tested if the mks-6 mutant has lifespan and UV sensitivity phenotypes indicative of xpa-1 dysfunction. Table S1 E reveals no lifespan defect for the mks-6 mutant, and UV assays detailed below reveal no statistically significant increase in UV sensitivity of the mks-6 mutant compared with wild-type. UV assays on N2, mks-6(gk674), and xpa-1(mn157) worms were based on a protocol by Astin et al. (2008). Staged 1-d-old adults were placed onto NGM plates containing no food, then irradiated with a 254-nm UV Stratalinker 2400 (Agilent Technologies). 1 h after UV irradiation (100 J/m²), worms were transferred to NGM plates containing OP50 bacteria and survival was scored thereafter every 24 h for 4 d. Each assay consisted of ∼90 worms and was repeated thseccree times.

Analyses of phasmid and amphid cilium lengths, dendrite lengths, and clustering of basal bodies was facilitated by visualizing the GFP-tagged IFT markers CHE-11 (IFT140) or OSM-6 (IFT52) in wild-type and mutant strains and quantitated using ImageJ; schematics depicting each of the four analyses are shown in Fig. 4, A–C; Fig. 5. In brief, ciliary axoneme lengths represent the distance between the distal end of the basal body and the tip of the cilium; dendrite lengths represent the distance between the beginning of a phasmid neuron cell body and its respective basal body; basal body clustering/distribution is the distance between the anterior-most and posterior-most basal bodies for each amphid channel. All measurements are repeated to statistical significance and reported in micrometers.

Wild-type or mutant worms were washed directly into a primary fixative of 2.5% glutaraldehyde in 0.1M Sorensen phosphate buffer. To facilitate rapid ingress of fixative, worms were cut in half under a dissection microscope using a razor blade, transferred to Eppendorf tubes, and fixed for 1 h at room temperature. Samples were then centrifuged at 3,000 rpm for 2 min, supernatant removed and washed for 10 min in 0.1M Sorensen phosphate buffer. Worms were then post-fixed in 1% osmium tetroxide in 0.1M Sorensen phosphate buffer for 1 h at room temperature. After washing in buffer, specimens were processed for electron microscopy by standard methods; in brief, they were dehydrated in ascending grades of alcohol to 100%, infiltrated with epon, and placed in aluminum planchettes orientated in a longitudinal aspect and allowed to polymerize at 60°C for 24 h. Using a Leica UC6 ultramicrotome, individual worms were sectioned either in cross or longitudinal section, from anterior tip (for cross sections) or the side of the worm (for longitudinal sections), at 1 µm until the area of interest was located as judged by examining by light microscopy the sections stained with toluidine blue. Thereafter, serial ultra-thin sections of 80 nm were taken for electron microscopical examination. These were picked up onto 100 mesh copper grids and stained with uranyl acetate and lead citrate. Using a Tecnai Twin (FEI) electron microscope, sections were examined to locate, in the first instance, the most distal region of the ciliary region and subsequently from that point to the more proximal regions of the ciliary apparatus. At each strategic point, distal segment, middle segment, and transition zone/transition fiber regions were tilted using the Compustage of the Tecnai microscope to ensure that the axonemal microtubules were imaged in an exact geometrical normalcy to the imaging system. All images were recorded at an accelerating voltage (120 kV) and objective aperture of 10 µm using a MegaView 3 digital recording system.

HMMER profile hidden Markov model software suite (http://hmmer.janelia.org) was used to identify all C2 and B9 domains in the C. elegans genome and query their possible evolutionary relatedness. C. elegans C2 domains were first extracted from known C. elegans genes containing C2 domains (Table S1 C), and their alignment (obtained with ClustalW) was used as input for the HMMER program hmmbuild. Resulting profile was calibrated using hmmcalibrate and used to search the C. elegans proteome (version WS180) with hmmsearch. Using the same approach, we searched for B9 domain–containing genes in C. elegans using as input the three known B9 domains from MKS-1, MKSR-1, and MKSR-2 derived from three nematode species (C. elegans, C. briggsae, and C. remanei; Table S1 C). Structure predictions of the B9 domains from the human proteins MKS1, B9D1, and B9D2 were performed using the mGenTHREADER fold recognition program (McGuffin and Jones, 2003; http://bioinf.cs.ucl.ac.uk/psipred).

Table S1 (excel file) contains (A) conserved transition zone proteins and reported physical interactions, (B) C. elegans strains used in this study, (C) genome-wide HMM search for C. elegans proteins containing B9- and C2-related domains, (D) intraflagellar transport rate analyses for wild-type and TZ protein-disrupted strains, and (E) lifespan measurements. Figs. S1–S5 show ultrastructural (TEM) analysis of (S1) mks-6;nphp-4, (S2) mks-5;nphp-4, (S3) mksr-1;nphp-4, (S4) mks-1;nphp-4, and (S5) mks-3;nphp-4 amphid channel cilia.

We thank N. Mahendran and N. Berbari for their assistance with the project, and the C. elegans gene knockout consortium, National Bioresource Project for C. elegans, and Caenorhabditis Genetics Center (CGC) for strains.

This work was funded by grants from the March of Dimes (to M.R. Leroux), Canadian Institutes of Health Research (CIHR; MOP-82870 to M.R. Leroux), Science Foundation Ireland (to O.E. Blacque), European Commission FP7 framework program (to O.E. Blacque), NIH P30 DK074038 Hepato/Renal Fibrocystic Diseases Core Center, NIH RO1 DK065655 (to B.K. Yoder), and NSERC (to N. Chen). C.L. Williams was supported by NIH T32 DK007545-22. P.N. Inglis was supported by a scholarship from MSFHR. N. Chen holds a scholar award from MSFHR and is a CIHR New Investigator. M.R. Leroux holds a senior scholar award from the Michael Smith Foundation for Health Research (MSFHR). O.E. Blacque is funded by the European Community's Seventh Framework Programme FP7/2009 under grant agreement no. 241955, SYSCILIA.