Cilia proteins getting to work – how do they commute from the cytoplasm to the base of cilia?

Cilia are microtubule-based projections with both sensory and motility functions. Primary cilia transduce signals required in embryonic development and adult tissue homeostasis (Corbit et al., 2005; Goetz and Anderson, 2010; Huangfu et al., 2003), whereas motile cilia regulate left–right symmetry breaking and extracellular fluid flow (Brooks and Wallingford, 2014; Nonaka et al., 1998). Defects in cilia assembly or function result in a pleiotropic set of diseases collectively termed ciliopathies, which can arise from either failure of ciliary signaling, cilia beating or both (Hildebrandt et al., 2011; Legendre et al., 2021).

Cilia are also organelles with unique protein compositions (Chávez et al., 2015; Garcia-Gonzalo et al., 2015; Ishikawa et al., 2012; Jacoby et al., 2009; Mick et al., 2015; Pazour et al., 2005). Therefore, during assembly, hundreds of proteins must be trafficked from their sites of synthesis in the cell body to the base of cilia, and finally into the axoneme (Ishikawa et al., 2012; Li et al., 2004; Pazour et al., 2005). Even after ciliogenesis, ongoing protein trafficking maintains cilia structure and function. Tubulin, for example, regularly turns over at the distal tip of cilia (Marshall and Rosenbaum, 2001; Stephens, 1997), and such protein turnover is essential to maintain ciliary length (Marshall and Rosenbaum, 2001). In fact, data from Chlamydomonas indicates that, during a 6-h period, ∼40% of ciliary proteins exchange with newly assembled proteins (Song and Dentler, 2001). Likewise, ciliary signaling requires dynamic regulation of protein targeting to and from the cilium in response to physiological stimuli.

The mechanisms by which entry to the specialized compartment of the cilium is maintained has been extensively discussed. However, before ciliary entry, proteins must first be trafficked from their site of synthesis in the cytoplasm to the basal body, the modified centriole that serves as a microtubule-organizing center (MTOC) for the cilium. In this Review, we discuss the role of active motor-driven and passive diffusion-and-capture mechanisms, as well as the various adaptor proteins that mediate them during the cytoplasmic pre-ciliary transport of ciliary proteins.

Basal bodies are structurally and functionally related to centrosomes (Kobayashi and Dynlacht, 2011), and centrosomal trafficking is far more deeply characterized. During the cell cycle, centrioles transition between templating basal bodies and centrosomes. Although centrioles undergo some modifications in this transition, the overall structural and molecular composition is homologous (Gupta et al., 2015; Vertii et al., 2016). Because the trafficking patterns of centrosomal proteins are better defined, we will first broadly describe centriolar trafficking mechanisms and highlight well-characterized examples.

Proteins synthesized in the cytoplasm can traffic via the extensive network of cytoplasmic microtubules emanating from centrioles (Fig. 1, left). These microtubules serve as tracks for microtubule motors, which harness the energy from ATP hydrolysis to power cargo transport (Roberts et al., 2013; Verhey et al., 2011). Because microtubules nucleated at centrioles grow their plus-ends into the cytoplasm (Petry and Vale, 2015), trafficking towards the basal body generally involves the minus-end-directed motor cytoplasmic dynein 1 (Caviston and Holzbaur, 2006) (Fig. 1, center). For example, pericentrin, a conserved soluble protein required for microtubule organization, interacts with dynein 1 and is recruited to centrosomes in a dynein 1- and microtubule-dependent fashion (Doxsey et al., 1994; Purohit et al., 1999; Young et al., 2000). Additional components, such as Nek2, Pcm1 and Par6α (also known as Pard6a in mammals), also rely on cytoplasmic dynein and microtubules for centrosomal recruitment (Dammermann and Merdes, 2002; Hames et al., 2005; Kim et al., 2008; Kodani et al., 2010; Quintyne and Schroer, 2002). Importantly, all three also localize to ciliary basal bodies and centriolar satellites (Fan et al., 2004; Kim et al., 2015; Kubo et al., 1999; Spalluto et al., 2012), so their transport in ciliated cells likely involves active transport.

Perhaps the best-studied soluble ciliary protein known to undergo microtubule-mediated cytoplasmic transport is Gli2 (Kim et al., 2009). Gli2 is a transcriptional activator of Hedgehog signaling whose enrichment at the ciliary tip increases upon pathway activation (Haycraft et al., 2005; Jiang and Hui, 2008). Noting association of Gli2 with cytoplasmic microtubules (Zhao et al., 2009), Kim and colleagues sought to test the importance of this colocalization for the function and ciliary localization of Gli2 (Kim et al., 2009). Through vinblastine treatment, which disrupts cytoplasmic microtubules while leaving ciliary microtubules intact (Gerdes et al., 2007), they demonstrated a dose-dependent loss of Gli2 accumulation in cilia (Kim et al., 2009). This loss of ciliary signal also inhibited Hedgehog pathway activation, demonstrating the importance of ciliary localization for protein function (Kim et al., 2009). As kinesin-II, which powers anterograde intraflagellar transport (IFT; see Box 1), binds to Gli2 (Carpenter et al., 2015), it will be interesting to determine how the handoff of Gli2 between cytoplasmic and IFT motors occurs. Another class of cytoplasmic motors of interest are the minus-end-directed C-kinesins (Konjikusic et al., 2021; Sablin et al., 1998), which are implicated in regulating localization of peroxisomes to the base of cilia (Miyamoto et al., 2020).

Although some proteins rely on active transport to establish appropriate cellular localization and enrichment, for many proteins, passive diffusion is a sufficiently fast method of trafficking. In particular, diffusion is faster for smaller proteins, as the diffusion coefficient scales inversely with molecular size (Miller, 1924). Such diffusion distributes proteins throughout the cytoplasm, so creating local protein enrichment requires capture at defined cellular sites. If a target protein reaches its appropriate subcellular localization, it can be captured through molecular interactions (Lippincott-Schwartz et al., 2001) (Fig. 1, center). These molecular interactions, such as protein–protein interaction or phase separation, lead to local enrichment.

Interestingly, not all randomly oriented movements are pure diffusion. In addition to pure diffusion, which is driven by thermal motion, active diffusion relies on active processes, which create faster molecular movements (Brangwynne et al., 2009). ATP-driven active processes, such as motors walking along the cytoskeleton, create force fluctuations in the cytoplasm that ‘stir’ its contents and increase the speed of intracellular diffusion (Guo et al., 2014). However, unlike active transport, which is directed, molecules undergoing active diffusion still move in random directions. Interestingly, centrin foci, which help form centrioles, might undergo active diffusion. For example, although centrin foci display diffusive undirected motion, the timescale of pure diffusion would not allow for centrin foci to incorporate into daughter centrioles prior to cell division (Rafelski et al., 2011).

Several proteins display microtubule-independent recruitment to the centrioles, providing evidence for diffusion-to-capture (Azimzadeh et al., 2009; Khodjakov and Rieder, 1999; Kim et al., 2005). One example is BBS6, a member of the BBsome, a modified chaperonin complex required for transport of ciliary membrane proteins (Kim et al., 2005; Seo et al., 2010). The molecular interaction that leads to BBS6 capture and subsequent enrichment at centrioles is yet to be discovered.

The centrosome is surrounded by pericentriolar material (PCM), as well as cytoplasmic granules called centriolar satellites that also localize near basal bodies and play pivotal roles in ciliogenesis and ciliary protein trafficking (Odabasi et al., 2020; Prosser and Pelletier, 2020). Indeed, loss of centriolar satellites upon depletion of Pcm1 leads to ciliogenesis defects and altered ciliary protein content (Odabasi et al., 2019; Wang et al., 2016). Despite the clear role for centriolar satellites in regulating centriolar and ciliary protein targeting, the exact mechanism of this regulation is still unclear (Prosser and Pelletier, 2020). Centriolar satellites undergo dynein-mediated transport to centrosomes (Dammermann and Merdes, 2002; Kubo et al., 1999; Quintyne and Schroer, 2002). Therefore, one hypothesis suggests centriolar satellites could deliver ciliary proteins to the basal body through active transport (Odabasi et al., 2020). Alternatively, satellites could act as sequestration sites that moderate incorporation of proteins into the cilium (Odabasi et al., 2020).

Notably, satellite proximity to the cilium is necessary for proper protein targeting (Aydin et al., 2020), and some centriolar satellite proteins dynamically redistribute during ciliogenesis. One example of such redistribution is BBS4, another BBSome subunit. Centriolar satellite levels of BBS4 decrease as their ciliary levels increase (Nachury et al., 2007). Likewise, Ccdc66 is required for normal organization of the centriolar satellites and for ciliogenesis, and this protein localizes to the centrosome and then redistributes to cilia following ciliogenesis (Conkar et al., 2017). Ccdc66 interacts with dynein and undergoes active transport towards centrioles, further demonstrating that active transport and centriolar satellites as important regulators of ciliary protein levels (Conkar et al., 2019, 2017). Altogether, centriolar satellite transport from the cytoplasm to the basal body is required for proper ciliary protein trafficking.

Localized translation of mRNAs is an important mechanism of intracellular protein distribution in an array of settings (Buxbaum et al., 2015), and many protein complexes require colocalization of mRNAs for co-translational assembly of polypeptides (Duncan and Mata, 2011; Shiber et al., 2018). Although many mRNAs encoding centrosomal components and even IFT proteins display random subcellular distributions (Kwon et al., 2021; Sepulveda et al., 2018), protein synthesis at or near centrioles does impact the localization of certain centrosomal and ciliary proteins (Fig. 1, right).

For example, several mRNAs encoding embryonic patterning molecules localize to the centrosome (Lambert and Nagy, 2002), and their asymmetric inheritance promotes the establishment of distinct daughter cell types during embryonic development (Shlyakhtina et al., 2019). Similarly, centrosomal purification and fluorescent in situ hybridization assays have identified several centriole-localized mRNAs that encode proteins with centriolar functions (Alliegro et al., 2006; Chouaib et al., 2020; Kwon et al., 2021; Lécuyer et al., 2007; Safieddine et al., 2021; Sepulveda et al., 2018). Because localized protein synthesis requires not just mRNA localization, but also translation, it is compelling that the 18S ribosomal RNA also associates with centrosomes (Alliegro et al., 2006). In addition, recent evidence indicates that polysomes actively translate centrosomal proteins as they traffic to the centrosome (Safieddine et al., 2021; Sepulveda et al., 2018).

Finally, one provocative paper recently reported that certain mRNAs, RNA-binding proteins and even the translation machinery are present in motile cilia of mammalian brain ependymal cells (Hao et al., 2021). Consistent with this possibility, several proteomic studies have demonstrated the presence of ribosomal subunits in motile cilia from the mammalian brain and airway, as well as from the flagella of Chlamydomonas (Ostrowski et al., 2002; Pazour et al., 2005).

Further study of local translation as a mechanism for ciliary protein localization is warranted, as is a broader analysis generally of the roles for both active and passive transport mechanisms in the targeting of proteins to cilia. In the following sections, we provide four examples that help to illustrate the different mechanisms by which ciliary proteins make their way from the cytoplasm to their sites of action in the cilium.

The mechanism of trafficking membranes to cilia have been extensively reviewed, as have the routes of ciliary membrane proteins such as the Hedgehog mediator Smoothened and diverse ciliary GPCRs traffic from the trans-Golgi network to the cilium via vesicular transport (Garcia et al., 2018; Jensen and Leroux, 2017; Long and Huang, 2020; Zhao et al., 2022). These topics nonetheless bear a brief mention here because of the interplay of membrane trafficking with the transport of soluble ciliary components.

GTPases and tethering complexes that mediate loaded vesicle transport can deliver proteins directly to the periciliary membrane or to the plasma membrane, and from there, proteins such as the Hedgehog receptor smoothened can undergo lateral transport through the periciliary barrier with the aid of the BBSome and IFT complexes (Nachury et al., 2010). Many such membrane proteins contain ciliary-targeting sequences (CTSs), and disruption of these sequences impedes proper ciliary localization (Badgandi et al., 2017; Berbari et al., 2008; Corbit et al., 2005). Among these, smoothened has been shown to contain a hydrophobic and basic motif (Corbit et al., 2005; Mukhopadhyay and Rohatgi, 2014; Mukhopadhyay et al., 2010; Rimkus et al., 2016). This motif is conserved in other ciliary signaling proteins, such as somatostatin receptor 3 (Sstr3) and serotonin receptor 6 (Htr6) (Corbit et al., 2005).

These results are of special relevance here because data from Chlamydomonas suggest that some non-membrane ciliary proteins might also exploit the same mechanisms, reaching the cilium by ‘hitchhiking’ on membranes. For example, several ciliary structural proteins were identified by co-fractionation as present on the outer membrane of vesicles in the cytoplasm (Sedmak and Wolfrum, 2010; Wood and Rosenbaum, 2014; Wood et al., 2012). Such a vesicular interaction could allow for dual delivery of membrane, membrane proteins and associated structural components necessary for the formation of cilia (Wood and Rosenbaum, 2014).

Finally, certain soluble proteins associated with the IFT system, discussed below and in Box 1, appear to play key roles in cytoplasmic transport, especially of membrane proteins (Liem et al., 2012; Mukhopadhyay et al., 2010). In fact, the IFT subunit WDR35 (also known as IFT121) was recently found to coat vesicles, and Wdr35^−/− mutant mouse embryonic fibroblasts (MEFs) failed to fuse vesicles with the ciliary sheath, leading to an IFT defect and stunted cilia growth (Quidwai et al., 2021). Loss of WDR35 in retinal pigment epithelial-1 (RPE1) cells has also been shown to severely disrupt ciliogenesis and also prevents the maturation of Rab8-related vesicular structures (Fu et al., 2016). Thus, a better understanding of the overlap between membrane trafficking, membrane protein trafficking and soluble protein trafficking will be essential for a comprehensive understanding of cilia function and ciliogenesis.

In addition to their role in membrane protein trafficking, the IFT proteins (Box 1) provide an excellent entry point for understanding the many issues related to ciliary protein transport in the cytoplasm.

The mRNAs encoding IFT proteins display a random cytoplasmic distribution (Kwon et al., 2021), and synthesis of IFT proteins occurs in the cytoplasm. Following their synthesis, IFT proteins accumulate in a peri-basal body pool before entering the axoneme, and the majority of IFT proteins in the cell are in fact cytoplasmic (Cole et al., 1998; Deane et al., 2001; Vashishtha et al., 1996). Interestingly, recent evidence has identified IFT proteins within the centriolar satellite interactome (Aydin et al., 2020; Gupta et al., 2015), implicating these locations as potential cargo-binding sites for IFT. Recent work has begun to shed light on the mechanisms by which IFT proteins are recruited to the basal body prior to ciliary import.

It has been proposed that IFTs might be actively transported along cytoplasmic microtubules, but current evidence suggests that these microtubules are dispensable for recruitment of IFT proteins to the basal body. Knockout of the subdistal appendage protein galectin-3 (gal3^−/−; also known as LGALS3) disrupts the cytoplasmic microtubule network in mouse tracheal multiciliated cells (MCCs), yet IFT88 still properly localizes to gal3^−/− basal bodies (Clare et al., 2014). In addition, IFT proteins associate with centrioles throughout the cell cycle, and data from cell culture implicated a diffusion-to-capture mechanism for centrosomal recruitment of IFT88 (Robert et al., 2007).

Recently, we found that the dynamics of several different IFT proteins are not impacted by disruption of cytoplasmic microtubules though cold-shock and nocodazole treatment in Xenopus multiciliated cells (MCCs) (Hibbard et al., 2021). Although this manipulation disrupted the recruitment dynamics of Ccdc66, a known cargo of active transport to centrioles (Conkar et al., 2017), both IFT-A and IFT-B proteins displayed normal basal body recruitment dynamics (Hibbard et al., 2021). These data argue against an active transport mechanism, so we propose that IFT proteins localize to the basal body via a diffusion-to-capture mechanism.

IFT components exhibit co-dependence for their stability and recruitment to basal bodies, and several IFT proteins cannot be expressed or purified without other complex members (Taschner et al., 2011, 2016). These proteins are insoluble or degraded when expressed on their own, indicating that subcomplex assembly is required for the stability of several IFT components (Taschner et al., 2011). Consistent with this in vitro data, western blot data of whole-cell extract demonstrates that cellular levels of multiple members of the IFT-B1 complex (see Box 1) are impacted in Ift46 and Ift52 mutants (Richey and Qin, 2012). Similar results have been demonstrated for IFT-A integrity (Behal et al., 2012; Zhu et al., 2017).

In addition, studies of protein dynamics using fluorescence recovery after photobleaching (FRAP) in Chlamydomonas and Xenopus demonstrated that functionally related IFT proteins display shared recruitment dynamics (Hibbard et al., 2021; Wingfield et al., 2017). For example, all members of the IFT-B1 subcomplex display shared kinetics, but these differ from the kinetics shared by members of the IFT-B2 subcomplex (see Box 1). Members of the IFT-A complex likewise exhibit similar recruitment kinetics, but these are distinct from those displayed by IFT-B (Hibbard et al., 2021; Wingfield et al., 2017).

More direct evidence of co-trafficking within the cytoplasm comes from work studying the interplay of the IFT-B proteins IFT52 and IFT46. In Chlamydomonas, a 95-amino-acid IFT52-binding domain in IFT46 is necessary and sufficient for its basal body targeting, and IFT52 mutant cells lose the normal basal body enrichment of IFT46 (Lv et al., 2017). In a complementary experiment in Xenopus, loss of Ift52 significantly disrupted the FRAP kinetics of Ift46 in the peri-basal body pool (Hibbard et al., 2021). Altogether, these data suggest an attractive hypothesis that IFT subcomplex assembly occurs in the cytoplasm. This model is consistent with recent cryo-EM data suggesting that IFT subcomplexes are separately recruited to the ciliary base where they interact to form trains (van den Hoek et al., 2022).

After identification of the peri-basal body pool of IFT proteins, efforts were made to identify the basal body sub-structures that were responsible for this accumulation. By using immunogold electron microscopy, distal appendages (also called transition fibers) were first implicated as being the site where IFT52 accumulates; this suggested that these structures could serve as the basal body-docking sites for IFT particles (Deane et al., 2001). Recently, super-resolution microscopy has provided better resolution of distal appendage and architecture of IFT particles (Hazime et al., 2021; Katoh et al., 2020; Shi et al., 2017; Yang et al., 2018, 2015). It has been suggested that the IFT proteins occupy the space between distal appendage blades, termed the distal appendage matrix (Katoh et al., 2020; van den Hoek et al., 2022; Yang et al., 2018).

Genetic approaches have confirmed that distal appendage proteins (DAPs) are important for IFT protein recruitment. For example, depletion of Cep164, a core structural component of distal appendages, results in defective basal body localization of IFT88 and IFT81 (Čajánek and Nigg, 2014; Schmidt et al., 2012). Likewise, other DAPs, such as Ttbk2, Fbf1 and Cep83 (also known as Ccdc41), are important for the centriolar localization of IFT proteins (Goetz et al., 2012; Joo et al., 2013; Tanos et al., 2013; Wei et al., 2013). There is also evidence of direct interactions between DAPs and IFTs, for example between Fbf1 and IFT54 (Wei et al., 2013), and between Cep83 and IFT20 (Joo et al., 2013). Finally, loss of the distal centriole proteins Ofd1 and C2cd3, which act upstream of distal appendage formation, leads to a disrupted recruitment of IFT81 and IFT88 to basal bodies (Singla et al., 2010; Thauvin-Robinet et al., 2014).

Although distal appendages are required generally for IFT recruitment to the peri-basal body pool, it is likely that IFT-A and IFT-B each require additional independent factors for their recruitment. Analysis of the ciliogenesis and planar cell polarity effector (CPLANE) complex have demonstrated that IFT-A and IFT-B are recruited independently. CPLANE comprises at least five distinct proteins, all of which are essential for ciliogenesis in Xenopus and mice (reviewed in Adler and Wallingford, 2017). The CPLANE complex assembles in a hierarchical manner at basal bodies, where it colocalizes with the distal appendage protein Cep164 (Toriyama et al., 2016). Moreover, in Xenopus MCCs, CPLANE proteins are specifically required for the recruitment of a specific subcomplex of IFT-A proteins (the peripheral proteins; see Box 1). The remainder of IFT-A, as well as IFT-B and other ciliary proteins, are recruited normally (Brooks and Wallingford, 2012; Toriyama et al., 2016). These data suggest that CPLANE might contribute to the ‘capture’ mechanism of IFT-A, but it is worth noting that one recent study suggests that the CPLANE protein Rsg1 could be dispensable for IFT-A recruitment in mice (Agbu et al., 2018). Further study will be of interest, as examination of allelic variants suggests that loss of IFT-A recruitment might underlie human ciliopathies (Toriyama et al., 2016).

The previous section dealt with the cytoplasmic transport mechanisms of core components required for assembly and homeostasis of all cilia. But cilia are remarkably diverse, and in this section, we will highlight the mechanisms by which specialized cilia transport specialized machinery. Perhaps the most well-defined category of specialized cilia are the motile cilia, whose oriented beating generates fluid flows that are critical for development and homeostasis in many tissues in animals and that propel sperm cells and many single-celled organisms (Fig. 2A).

The beating of motile cilia is driven by a complex set of axoneme-specific dynein motors that drive sliding of the ciliary microtubule doublets. Based on their relative positions within the axoneme, these motors are classified as either outer dynein arms (ODAs) or inner dynein arms (IDAs) (Fig. 2B), with the former driving ciliary beating generally, and the latter tuning the waveform (Kamiya and Yagi, 2014; King, 2018). Allelic variants in genes encoding ODA or IDA subunits are the major cause of the motile ciliopathy syndrome known as primary ciliary dyskinesia (PCD; MIM 244400). This genetic disease results in repeated sinopulmonary disease, bronchiectasis, cardiac defects and situs anomalies, as well as infertility (Horani et al., 2016; Mitchison and Valente, 2017; Wallmeier et al., 2020). How the axonemal dyneins are transported into cilia is therefore a critical question.

Axonemal dynein motors are known to pre-assemble in the cytoplasm prior to their deployment to cilia (Fowkes and Mitchell, 1998), and a large body of work has now elucidated the complex repertoire of cytoplasmic chaperones and assembly factors required in this process (reviewed in Desai et al., 2018; Fabczak and Osinka, 2019). These factors are sequestered together in liquid-like organelles termed ‘DynAPs’ in the cytoplasm of motile ciliated cells (Huizar et al., 2018) (Fig. 2A). However, how assembled motors are transported from the cytoplasm to the base of cilia remains poorly understood.

Insights into the process are beginning to emerge from studies of the ODA-specific IFT adaptor molecule Oda16 (known in vertebrates as Daw1 or Wdr69). Like the IFT proteins, Oda16 is present in both the cytoplasm and the axoneme, and Oda16 serves as an adaptor linking ODAs to the N-terminus of Ift46 (Ahmed et al., 2008; Hou et al., 2007; Hou and Witman, 2017; Taschner et al., 2017). Genetic analysis in Chlamydomonas revealed a specific role for IFT46 in the transport of ODAs, but not IDAs (Ahmed et al., 2008; Hou et al., 2007). Studies of these axonemal transport mechanisms ultimately led to insights into the cytoplasmic transport of ODAs; in Chlamydomonas mutants lacking the N-terminus of IFT46, ODA subunits accumulate at the basal body. However, this accumulation is lost in Oda16 mutants, demonstrating a role for Oda16 in cytosolic transport (Dai et al., 2018). Notably, the vertebrate ortholog of Oda16 is also essential for ODA transport and ciliary beating in zebrafish (Gao et al., 2010).

Another cofactor, ODA8 (also known as Lrrc56), is also interesting. This protein is essential for cilia beating in humans, trypanosomes and Chlamydomonas, essential for ODA assembly and present at very high levels in the cytoplasm but only at very low levels in the axoneme (Bonnefoy et al., 2018; Desai et al., 2015). Strikingly, live imaging in trypanosomes suggests that ODA8 is not stably incorporated into axonemes but instead undergoes bidirectional trafficking similar to IFT (Bonnefoy et al., 2018). Thus, both Oda8 and Oda16 appear to link ODA pre-assembly in the cytoplasm to ODA transport into the axoneme via IFT. Transport of inner dynein arms in the axonemes also requires IFT (Jiang et al., 2017; Viswanadha et al., 2014).

Complementary insights into dynein transport in multiciliated cells have emerged from studies of Ccdc103, which is associated with deployment of axonemal dyneins in zebrafish and humans (Panizzi et al., 2012). Fascinatingly, in human patients with variant alleles of CCDC103, the ODA subunit DNAH5 not only failed to enter axonemes but also accumulated strongly in the peri-nuclear region, suggesting a role for CCDC103 in the cytoplasmic transport of at least some ODA subunits (Panizzi et al., 2012). Ccdc103 is present in both the cytoplasm and the axonemes of Chlamydomonas and zebrafish, and oligomerization of the protein is central to its function (Panizzi et al., 2012; Shoemark et al., 2018). Ccdc103 is enriched in cytoplasmic foci in Chlamydomonas (King and Patel-King, 2015), although it remains unclear whether these foci are related to DynAPs. However, formation of membrane-less organelles can be driven by multivalent protein–protein interactions (Banani et al., 2017), and because Ccdc103 contains multiple self-interaction sites, further investigation of Ccdc103 in the formation of biomolecular condensates is warranted.

Finally, although the transport of ODAs is beginning to come into focus, there are myriad of other specialized components in motile axonemes, and their transport is far less well-defined. For example, IDAs and ODAs are each tethered to the axoneme by a specialized machinery. While the IDA tether Cfap44 is present in cytosolic foci closely associated with DynAPs (Drew et al., 2020), the ODA tether, known as the ODA docking complex, is not (Huizar et al., 2018). Another multi-protein machine, the dynein regulatory complex (DRC) is also pre-assembled in the cytoplasm (Bower et al., 2018), and entry of the DRC into the axoneme is IFT dependent (Wren et al., 2013). By contrast, the radial spokes that connect the microtubule doublets to the central pair microtubules in motile cilia (Fig. 2B) appear to be transported into axonemes as individual subunits, which are then assembled at the ciliary tip (Gupta et al., 2012). Function of motile cilia requires proper assembly and transport of these protein complexes from their site of synthesis to their docking location within cilia, so it will be important to elucidate these cytoplasmic transport mechanisms and how they contribute to overall cilia function.

Cilia are remarkably diverse, not just across evolution but also across cell types within animals. For example, motile cilia, primary cilia and photoreceptor cilia in vertebrates are each specialized for a distinct function, and these specialized cilium types require the transport of a specialized machinery. This topic is far too extensive to tackle comprehensively in this Review, but below, we provide three examples that illustrate this diversity.

First, the light-detecting outer segments of vertebrate photoreceptors in the retina are modified cilia, and accordingly, retinal degeneration is a common presentation in human ciliopathies. In addition to ubiquitous ciliary transport mechanisms, such as IFT, photoreceptors use several specialized mechanisms (Sánchez-Bellver et al., 2021), most of which have been identified through studies of human retinal degeneration. For example, Usher syndrome is the most frequent cause of combined blindness and deafness in humans and is caused by genetic variants in any of several proteins, which play key cytoplasmic transport roles in the photoreceptor (Reiners et al., 2006). The Usher syndrome proteins form a multi-protein complex and collectively are required to link cytoplasmic rhodopsin transport to the IFT machinery at the base of the outer segment (Maerker et al., 2008). In the cytoplasm, Usher syndrome proteins interact with dynein, kinesins and even myosin motors, suggesting the possibility of active transport (Sánchez-Bellver et al., 2021). Likewise, retinitis pigmentosis type 2 is caused by genetic variants in the RP2 gene, which is essential for prenylated protein transport in the mouse retina (Zhang et al., 2015). Like the chaperones that facilitate dynein assembly in motile ciliated cells (above), photoreceptors employ at least two specialized cytosolic chaperones, Unc119b and Pde6d, which aid in the deployment of lipidated cargoes to the outer segment (Constantine et al., 2012; Zhang et al., 2007).

A second example is the Chlamydomonas flagellum, which is well known as a model for ciliary beating, but also plays critical sensory and signaling roles. Indeed, like animal photoreceptors, Chlamydomonas flagella contain rhodopsin proteins whose light-dependent localization to the axoneme requires IFT (Awasthi et al., 2016). In addition, interactions between gametes of distinct mating types require the flagellar localization of the integral membrane protein Sag1, which normally resides in the plasma membrane (Belzile et al., 2013; Ranjan et al., 2019). However, unlike many membrane proteins in vertebrates, its re-localization to the cilium does not require of IFT. Instead, Sag1 is transiently internalized from the plasma membrane and transported to the base of cilia by cytoplasmic microtubules (Belzile et al., 2013; Ranjan et al., 2019). An even more specialized mechanism has been described during Chlamydomonas flagellar regeneration; here, dramatic ATP-dependent contraction of fibers at the base of the flagellum (called rhizoplasts) are thought to drive rapid transport of flagellar precursors to the basal body (Salisbury et al., 1987).

A final important example is provided by the specialized cilia of Caenorhabditis elegans sensory neurons. These nematodes have evolved an array of bizarre ciliary morphologies that enable precise chemo- and mechano-sensation (Bae and Barr, 2008). These ciliated neurons have a specialized MTOC at the dendrite tip, which assembles parallel microtubules that transport ciliary cargoes (Harterink et al., 2018), and several specialized transport mechanisms have been described. For example, a protein called CHB-3 controls the active transport of ciliary guanylyl cyclases in dendrites (Fujiwara et al., 2010). Interestingly, CHB-3 contains a Zmynd domain, which is also found in the cytoplasmic ODA assembly factor Zmynd10 (Mali et al., 2018). Even more curious is the transmembrane protein organic solute transporter α-like protein 1 (OSTA-1), which is specifically required for establishing the complex morphologies of certain neuronal cilia, where it localizes to the base of cilia and in foci along the dendrite and is thought to control membrane trafficking (Olivier-Mason et al., 2013). The morphology of some sensory cilia is also controlled by the tubulin glutamylase Ttll11-B, which is also actively transported in dendrites (O'Hagan et al., 2017). Thus, each of the many specialized types of cilia in animals presents a microcosm of the overarching challenge to understanding the cytoplasmic transport of ciliary proteins.

The transport of proteins within cilia has been extensively studied, as have the mechanisms that govern entry to the privileged space of the cilium. By contrast, the delivery of ciliary proteins from their site of synthesis in the cytoplasm to the basal body has received far less attention, but it is no less critical for ciliogenesis and ciliary signaling. As we review here, these cytoplasmic transport routes are at least as complex and tightly regulated as the better-studied mechanisms that act subsequently to them. Both active and passive transport mechanisms are involved, and in many cases the proteins that mediate these processes are implicated in human ciliopathy. Finally, we note that the defined repertoire of ciliary proteins provides a biomedically relevant, yet focused paradigm for deeper investigation into far more general biological problems related to active and passive transport of materials in the cytoplasm (e.g. Brangwynne et al., 2009; Kural et al., 2007; Lau et al., 2003; Lin et al., 2016; Parry et al., 2014).