Assembly of primary cilia

Authors

  • Lotte B. Pedersen,

    Corresponding author
    1. Department of Biology, Section of Cell and Molecular Biology, University of Copenhagen, Copenhagen, Denmark
    • Department of Biology, Section of Cell and Developmental Biology, University of Copenhagen, Universitetsparken 13, DK-2100 Copenhagen OE, Denmark
    Search for more papers by this author
  • Iben R. Veland,

    1. Department of Biology, Section of Cell and Molecular Biology, University of Copenhagen, Copenhagen, Denmark
    Search for more papers by this author
  • Jacob M. Schrøder,

    1. Department of Biology, Section of Cell and Molecular Biology, University of Copenhagen, Copenhagen, Denmark
    Search for more papers by this author
  • Søren T. Christensen

    1. Department of Biology, Section of Cell and Molecular Biology, University of Copenhagen, Copenhagen, Denmark
    Search for more papers by this author

Abstract

Primary cilia are microtubule-based, hair-like sensory organelles present on the surface of most growth-arrested cells in our body. Recent research has demonstrated a crucial role for primary cilia in regulating vertebrate developmental pathways and tissue homeostasis, and defects in genes involved in primary cilia assembly or function have been associated with a panoply of disorders and diseases, including polycystic kidney disease, left-right asymmetry defects, hydrocephalus, and Bardet Biedl Syndrome. Here we provide an up-to-date review focused on the molecular mechanisms involved in the assembly of primary cilia in vertebrate cells. We present an overview of the early stages of the cilia assembly process, as well as a description of the intraflagellar transport (IFT) system. IFT is a highly conserved process required for assembly of almost all eukaryotic cilia and flagella, and much of our current knowledge about IFT is based on studies performed in Chlamydomonas and Caenorhabditis elegans. Therefore, our review of the IFT literature includes studies performed in these two model organisms. The role of several non-IFT proteins (e.g., centrosomal proteins) in the ciliary assembly process is also discussed. Developmental Dynamics 237:1993–2006, 2008. © 2008 Wiley-Liss, Inc.

INTRODUCTION

Primary cilia are microtubule (MT)-based solitary and non-motile organelles that emanate from the surface of most quiescent cells in our body. The MT cytoskeleton of the primary cilium, the axoneme, consists of MT doublets (A plus B tubules) organized in a 9+0 ultrastructural pattern that grows from and continues the nine-fold symmetry of the mother centriole of the centrosome, which develops into the basal body of the cilium as cells enter growth arrest (Satir and Christensen,2006). Recent research has demonstrated that the primary cilium operates as an antenna-like unit that relays and coordinates signaling pathways, which are critical in embryonic and postnatal development, as well as in tissue homeostasis in adulthood. The physiological significance of primary cilia and their importance as sensory organelles for human health and development has been reviewed extensively in recent years (Badano et al.,2006; Christensen et al.,2007; Ibanez-Tallon et al.,2003; Pan et al.,2005; Pazour and Rosenbaum,2002; Sloboda,2002). Here we will focus primarily on the molecular mechanisms involved in the assembly of primary cilia in vertebrate cells, including the molecular characterization of intraflagellar transport (IFT) components and the role of several non-IFT proteins (e.g., centrosomal proteins) in the assembly process. However, since IFT is highly conserved among ciliated organisms and because much of our present knowledge about IFT is derived from work in Chlamydomonas and Caenorhabditis elegans, our description of IFT also includes studies performed in these two model organisms.

EARLY STAGES OF PRIMARY CILIA ASSEMBLY

The assembly of primary cilia is tightly coupled to the cell cycle and occurs from the distal end of the mother centriole as cells enter growth arrest/G1, whereas the cilia are shed shortly before cells enter mitosis (Archer and Wheatley,1971; Pan and Snell,2007; Quarmby and Parker,2005). Based on electron microscopy (EM) analysis of fibroblasts and smooth muscle cells in organ cultures, Sorokin elegantly set out the three distinct early stages of primary cilium assembly (Sorokin,1962; step 1–3 in Fig. 1h). First, a Golgi-derived vesicle attaches to the distal end of the mother centriole, from which the nascent axoneme begins to emerge; the vesicle becomes invaginated as the centriole extends and accumulates accessory structures to become the distal basal body. Second, nearby vesicles fuse with new membrane forming at the ciliary base to create a sheath surrounding the elongating axonemal shaft in which MT pairs are quickly assembled. In the third stage of ciliogenesis, the membrane-surrounded axoneme reaches the cell surface and the ciliary membrane fuses with the plasma membrane, forming a cup-like structure known as the ciliary necklace, which consists of multiple strands of intramembrane particles that connect to the centre of the mother centriole (Gilula and Satir,1972).

Figure 1.

Stages in formation of the primary cilium analyzed by EM. Stage 1: A Golgi-derived centriolar vesicle (CV) localizes to the distal end of the centrosomal mother centriole (a) and invaginates the nascent axoneme (b). Stage 2: Nearby vesicles (secondary CVs) fuse with the new membrane (c and inset) to form a sheath that surrounds the elongating axonemal shaft (d). Stage 3: The membrane-bound axoneme reaches and fuses with the plasma membrane forming the ciliary necklace (e). The inset in e shows a transverse section of the proximal region of a cilium with its characteristic 9+0 MT ultra structure. Stage 4: After docking of the mother centriole at the cell surface and fusion of the sheath with the plasma membrane, the nascent axoneme is elongated by IFT to form the mature primary cilium (f, transmission EM) and (g, scanning EM). h: A schematic model of the four stages in formation of the primary cilium. a–e are reproduced from Sorokin (1962) with permission of the publisher. e (inset) is reproduced from Wheatley (1969) with permission of the publisher. f and g are courtesy of Denys N. Wheatley and were reproduced from Currie and Wheatley (1966) and Wheatley et al. (1996) with permission of the publisher.

The EM observations by Sorokin (1962) suggest that the assembly of the primary cilium may be initiated while the mother centriole is positioned at or near the Golgi apparatus, and close to the nucleus; i.e., prior to centriole migration and docking of the ciliary sheath to the plasma membrane. However, in some cases the centriole that becomes the basal body moves peripherically towards the apical cell membrane, where the whole ciliary assembly process takes place. Either way, the cilium can increase considerably in length, (e.g., kidney epithelial cells; Wheatley and Bowser,2000), and may arise directly from the cell surface (e.g., chrondrocytes; Poole et al.,1997), or from deeply inside the cell close to the nucleus (e.g., adrenal zona glomerulosa cells; Wheatley,1967).

Molecules Involved in Early Stages of Ciliogenesis

Even in some of the earliest examples of the centriole (diplosome) relationship with the cell membrane (Currie and Wheatley,1966), several exacting details can be found, many of which are only now beginning to be understood in terms of the elements (nanobodies) involved, their dynamics and their macromolecular characteristics. These include the distal appendages of the mother centriole, which were proposed to be involved in its docking to the plasma membrane (Sorokin,1968). Indeed, when both alleles of the Odf2 gene were knocked out in mouse F9 cells, the distal appendages on the mother centriole were missing and the cells failed to form primary cilia (Ishikawa et al.,2005), confirming that these appendages are important for ciliogenesis. The recently identified centrosomal protein Cep164 (Andersen et al.,2003) may also be involved in assembly or function of the distal appendages because Cep164 localizes to these structures and siRNA-mediated depletion of Cep164 in human retinal pigment epithelial (hTERT-RPE1) cells results in failure to assemble primary cilia (Graser et al.,2007). While the available evidence strongly suggests a role for Odf2 and Cep164 in docking of the basal body to the apical plasma membrane, it is likely that several additional centrosomal proteins are involved in this process. Thus, it was not surprising that in two recent studies based on systematic siRNA-mediated knock-down in retinal pigment epithelial (RPE) cells of a number of centrosome proteins, several of these (e.g., Pericentrin, Centrin 2, PCM-1, and the Alström Syndrome protein Alms1) were found to be required for assembly of primary cilia (Table 1) (Graser et al.,2007; Mikule et al.,2007). Previous work had indicated a role for Pericentrin in the targeting or tethering of IFT components to the ciliary basal body region in RPE cells (Jurczyk et al.,2004), and PCM-1 had been shown to interact with the Bardet Biedl Syndrome (BBS) protein 4 (Kim et al.,2004), which in turn has been implicated in targeting of membrane vesicles to the cilia (Nachury et al.,2007; see Blacque and Leroux,2006, for a comprehensive review of BBS). Interestingly, BBS4 has also been found to interact with the p150Glued subunit of the dynactin complex in yeast two-hybrid and immunoprecipitation assays (Kim et al.,2004), suggesting a possible role for p150Glued in cilia assembly or function. Indeed, recent evidence from our laboratory suggests that the p150Glued binding partner, the small MT plus end-tracking protein EB1, is required for primary cilia assembly in mouse fibroblasts (Schrøder et al.,2007). Although the exact mechanism by which EB1 promotes ciliogenesis still remains to be elucidated, it is tempting to speculate a possible involvement of EB1 in vesicular trafficking to the ciliary basal body, given the association of p150Glued and PCM-1 with BBS4 (Kim et al.,2004). Alternatively, EB1 and p150Glued might be involved in stabilizing axonemal MTs as they elongate, since both have been found to promote polymerization and elongation of MTs in vitro (Ligon et al.,2003). More experiments are required to distinguish between these possibilities and to determine whether the observed effect of EB1 on cilia assembly (Schrøder et al.,2007) is linked to PCM-1 and BBS proteins.

Table 1. Proteins Demonstrated to Be Involved in the Assembly of Vertebrate Primary Cilia
Protein nameProposed cilia-related function(s)Reference(s)
  • a

    The role of BBS proteins in cilia assembly appears to be complex and not yet fully understood. See text for details and see Blacque and Leroux (2006) for a comprehensive review on BBS.

  • b

    Centrosomal proteins involved in centriole duplication (Bettencourt-Dias and Glover,2007) as well as those for which limited or conflicting functional data is available are not included here.

Kinesin-2  
 KIF3A, KIF3BAnterograde IFTLin et al.,2003; Marszalek et al.,1999,2000; Takeda et al.,1999
 KIF17Anterograde IFT, assembly of distal axonemal singlet MTs (?)Jenkins et al.,2006
Cyt. dynein 2  
 DYNC2LI1Retrograde IFTRana et al.,2004
 DYNC2H1Retrograde IFTMay et al.,2005
IFT complex A  
 IFT140IFT (?), no obvious effect on cilia when inactivated in zebrafishTsujikawa and Malicki,2004
IFT complex B  
 Polaris/IFT88IFTMurcia et al.,2000; Pazour et al.,2000,2002
 IFT46IFT, cargo attachment?Gouttenoire et al.,2007
 IFT20IFT, ciliary membrane transportFollit et al.,2006
 Wim/IFT172IFT, ciliary tip regulationHuangfu et al.,2003; Sun et al.,2004
 IFT80IFTBeales et al.,2007
 Hippi/IFT57IFTHoude et al.,2006; Tsujikawa and Malicki,2004
 NGD5/IFT52IFTTsujikawa and Malicki,2004
Polarity proteins  
 Par3Binds directly to KIF3AFan et al.,2004; Nishimura et al.,2004b; Sfakianos et al.,2007
 Par6Interacts with Par3Fan et al.,2004
 aPKCζInteracts with Par3, regulates GSK3βEtienne-Manneville and Hall,2003; Fan et al.,2004
 14-3-3ηInteracts with Par3Fan et al.,2004
 CRB3Interacts with Par3/Par6/aPKCζFan et al.,2004,2007; Omori and Malicki,2006
 pVHLBinds directly to KIF3A and aPKCζLolkema et al.,2007; Okuda et al.,1999; Thoma et al.,2007
 GSK3βInteracts with pVHL and aPKCζ, regulates flagellar lengthEtienne-Manneville and Hall,2003; Thoma et al.,2007
BBSa  
 BBS-1Vesicle transport (?)Kulaga et al.,2004; Nachury et al.,2007
 BBS-2Vesicle transport (?)Nachury et al.,2007; Nishimura et al.,2004a
 BBS-4Vesicle transport (?); interacts with PCM-1 and p150GluedKim et al.,2004; Kulaga et al.,2004; Mykytyn et al.,2004; Nachury et al.,2007
Meckel-Gruber 
 MKS-1Centriole/basal body migrationDawe et al.,2007b; Kyttala et al.,2006
 MKS-3 (meckelin)Centriole/basal body migrationDawe et al.,2007b; Smith et al.,2006
Centrosomalb 
 PericentrinBasal body localization of IFT particlesGraser et al.,2007; Jurczyk et al.,2004; Mikule et al.,2007
 OFD1NDFerrante et al.,2006
 ODF2Docking of basal body at plasma membraneIshikawa et al.,2005
 PCM-1ND; interacts with BBS4Graser et al.,2007; Kim et al.,2004; Mikule et al.,2007
 Cep164Docking of basal body at plasma membraneGraser et al.,2007
 ALMS1NDGraser et al.,2007; Li et al.,2007; Mikule et al.,2007
Others  
 EBIND; interacts with p150GluedSchrøder et al.,2007
 FleerTubulin polyglutamylation, OSM-3/KIF17 activationPathak et al.,2007
 FAPP2Transport of vesicles to ciliaVieira et al.,2006
 Importin βSorting of proteins at ciliary necklace/pore regionFan et al.,2007

In terms of centriole migration to the site of primary cilia assembly at the apical cell surface, recent work has shed some light on the molecular mechanisms involved in this process, although many questions still remain to be answered (Dawe et al.,2007a). Some of the molecules that have been linked to centriole migration include the Meckel-Gruber Syndrome (MKS) proteins 1 and 3 (meckelin) (Dawe et al.,2007b). Since mutations in the genes encoding MKS1 and 3 are unlikely to account for all cases of MKS (Kyttala et al.,2006), it will be interesting to determine whether additional MKS genes are also involved in centriole migration.

INTRAFLAGELLAR TRANSPORT (IFT)

Once the basal body has docked onto the ciliary assembly site at the apical plasma membrane (step 3 in Fig. 1h), further assembly of the ciliary axoneme is mediated via IFT (step 4 in Fig. 1h; see also Fig. 2). IFT is a bidirectional motility system, localized between the axonemal outer doublet MTs and the flagellar/ciliary membrane, by which groups of large particles (IFT particles) are moved from the base of the cilium to its distal tip by kinesin-2 motors and are returned to the cell body by cytoplasmic dynein 2 (Fig. 2) (Rosenbaum and Witman,2002; Scholey,2003). The particles moving in the anterograde (base to tip) direction are associated with axonemal precursors destined for the assembly site at the tip, while the particles moving in the retrograde (tip to base) direction are associated with axonemal turn-over products that are being brought back to the cell body for recycling (Qin et al.,2004). IFT was first observed by differential interference contrast (DIC) microscopy in Chlamydomonas flagella, and EM observations of the particles seen to move by DIC showed that they are arranged in linear arrays that are sandwiched between the flagellar membrane and the axonemal outer doublet B MTs (Kozminski et al.,1993; Fig. 2, inset). IFT particle movement has subsequently been observed in cilia of C. elegans sensory neurons and in primary cilia of cultured IMCD3 and LLC-PK1 kidney cells using GFP-tagged IFT particle proteins and/or motor subunits, although the rates of particle movement measured in these cell types were slower than those measured in Chlamydomonas (Follit et al.,2006; Orozco et al.,1999; Snow et al.,2004).

Figure 2.

Model of canonical IFT. The model was adapted from Pedersen et al. (2006) and modified as described in the text. Step 1: IFT particles and motors localize to the peri-basal body region. Step 2: Kinesin-II mediates anterograde transport of IFT complexes A and B, axonemal precursors, and cytoplasmic dynein 2. Kinesin-II associates with complex A that in turns binds to B to which axonemal cargo proteins are attached. Step 3: Complexes A and B, axonemal precursors, and cytoplasmic dynein 2 are released into the flagellar tip compartment, then complex A and B dissociate from each other. Step 4: Complex A binds (directly or indirectly) to active cytoplasmic dynein 2 via the DYNC2LI1 subunit, and complex B then binds to complex A. Kinesin-II binds to DYNC2H1 independently of complex A and B and DYNC2LI1. Step 5: Cytoplasmic dynein 2 transports the other IFT components and axonemal turnover products back towards the cell body. Step 6: IFT components re-enter the cell body. The inset shows a transmission EM image of a longitudinal section of a Chlamydomonas flagellum with two groups of IFT particles (arrows). Courtesy of Stefan Geimer, University of Bayreuth.

The IFT particles have been purified from Chlamydomonas flagella using sucrose density gradient centrifugation and were found to sediment in two complexes, A and B, which collectively comprise about sixteen different polypeptides (Cole et al.,1998; Piperno and Mead,1997), including several that were known to be required for assembly of neuronal sensory cilia in C. elegans (Cole et al.,1998; Perkins et al.,1986). The amino acid sequences for all of these IFT particle proteins have been determined and for most of them the corresponding genes have been cloned (Cole,2003). Recent comparative genomics studies have shown that the genes encoding IFT components are highly conserved among ciliated eukaryotes, but absent from non-ciliated organisms such as higher plants and fungi (Avidor-Reiss et al.,2004; Li et al.,2004). Furthermore, functional analyses of individual IFT components performed in multiple organisms (see below) indicate that IFT is essential for the assembly of almost all eukaryotic cilia and flagella. The only exceptions are those flagella/cilia that assemble in the cytoplasm, e.g., the flagella of Drosophila sperm cells (Witman,2003).

Anterograde IFT Motors

The motors that power anterograde IFT belong to the kinesin-2 family. Two types of kinesin-2 motor complexes have been implicated in IFT: a heterotrimeric- and a homodimeric complex, respectively (Scholey,2003). Heterotrimeric kinesin-2 (also termed kinesin-II) was first purified from sea urchin eggs (Cole et al.,1992,1993) and has subsequently been characterized in numerous organisms, including Chlamydomonas, C. elegans, Tetrahymena, Drosophila, and mouse. The complex consists of two motor subunits of 90 and 85 kDa, respectively, and a 100-kDa non-motor subunit called Kinesin-Associated Protein (KAP) (Scholey,2003). In Chlamydomonas, the 90- and 85-kDa motor subunits of the kinesin-II complex are encoded by FLA10 and FLA8, respectively, while the KAP subunit is encoded by FLA3 (Cole,2003; Kozminski et al.,1995; Miller et al.,2005; Mueller et al.,2005; Walther et al.,1994). Chlamydomonas mutants defective in these genes were instrumental in defining the role of kinesin-II in anterograde IFT. For example, a null mutant in FLA10 fails to assemble flagella (Matsuura et al.,2002), indicating that kinesin-II is indispensable for flagellar assembly in this organism. Furthermore, studies using a temperature-sensitive FLA10 mutant indicated that kinesin-II is also required for maintaining flagellar length once the flagella are fully assembled (Huang et al.,1977; Kozminski et al.,1995). Subsequent motility in studies in C. elegans, by in vivo imaging of GFP-tagged motors and IFT particle proteins or by in vitro motility assays with purified motors, provided definite proof that kinesin-II physically moves IFT particles anterogradely along the axoneme (reviewed in Scholey,2008).

In human and mouse, the two motor subunits of the kinesin-II complex are called KIF3A and KIF3B (Scholey,1996). Consistent with evidence from Chlamydomonas, genetic studies in mice have indicated an essential role for kinesin-II in the assembly of primary cilia. First, null mutations in either KIF3B or KIF3A cause embryonic lethality in mice and the homozygous knock-out mutant embryos lack nodal cilia and develop situs inversus (Marszalek et al.,1999; Nonaka et al.,1998; Takeda et al.,1999). Second, when KIF3A was removed specifically from mouse photoreceptors using Cre-loxP mutagenesis, large accumulations of membranes, opsin, and arrestin accumulated within the photoreceptor inner segment suggesting a defect in IFT in the connecting cilium (Marszalek et al.,2000). Finally, Cre-loxP-mediated inactivation of KIF3A in the mouse kidney results in failure to assemble primary cilia and kidney cysts develop (Lin et al.,2003). In addition to KIF3A and KIF3B, mammals contain a third kinesin-II motor subunit called KIF3C, which is preferentially expressed in neural tissue and associates with KIF3A/KAP (Yang and Goldstein,1998). However, the functional significance of KIF3C is not clear as Kif3c −/− mutant mice have no obvious phenotype (Yang et al.,2001).

Some ciliated organisms contain a homodimeric kinesin-2 motor that functions somewhat redundantly with kinesin-II during anterograde IFT (Snow et al.,2004). This homodimeric kinesin-2 motor was first characterized in C. elegans where it is known as OSM-3 (Shakir et al.,1993; Signor et al.,1999b), and it has subsequently been identified in cilia of Tetrahymena and some vertebrate cell types, where it is named Kin5 and KIF17, respectively (Awan et al.,2004; Setou et al.,2000). Genetic and motility studies in C. elegans have suggested that OSM-3 functions as an accessory anterograde IFT motor involved in the assembly or function of specific subsets of sensory cilia, thereby contributing to the generation of ciliary structural and functional diversity (Evans et al.,2006; Mukhopadhyay et al.,2007). For example, in the C. elegans amphid channel cilia, OSM-3 appears to cooperate with kinesin-II to drive anterograde IFT along the middle segment of the cilia, where the axoneme consists of doublet MTs; OSM-3 alone mediates anterograde transport along the distal segment, which contains singlet MTs (Snow et al.,2004). For a detailed review of OSM-3 in C. elegans, see (Blacque et al.,2008).

The mammalian OSM-3 homolog, KIF17, was recently implicated in the targeting of cyclic nucleotide-gated (CNG) channels to sensory cilia in rodent olfactory epithelium. Native KIF17 localizes to the olfactory sensory cilia in mouse and expression of a dominant-negative KIF17 construct in MDCK cells prevented targeting of CNG channels to the primary cilia. Furthermore, immunoprecipitation suggested that KIF17 interacts with CNG subunits in native rat olfactory epithelium (Jenkins et al.,2006). Whether the proposed KIF17-mediated targeting of CNGs to olfactory cilia is related to assembly of distal singlet MTs in these cilia is unknown, although in frog olfactory epithelia, CNGs were found to localize to the distal segment of the sensory cilia (Flannery et al.,2006), which consists of singlet MTs (Reese,1965).

Regulation of kinesin-2 motors.

The coordinated use of kinesin-II and OSM-3 in C. elegans amphid channel cilia is dependent on BBS proteins, which appear to be required for keeping the two motors together during anterograde IFT along the middle segment (Blacque et al.,2004,2008; Ou et al.,2005; Pan et al.,2006). Interestingly, in mutant mice that lack BBS-1 or -4, the olfactory cilia are either greatly reduced in length or absent and the mice are anosmic (Kulaga et al.,2004). Furthermore, BBS-4 and BBS-2 mutant mice display defects in the assembly of sperm flagella and photoreceptor outer segments (Mykytyn et al.,2004; Nishimura et al.,2004a), consistent with a role of BBS proteins in regulating anterograde IFT. Since respiratory cilia and kidney primary cilia appear to be unaffected by the loss of BBS proteins (Mykytyn et al.,2004; Nishimura et al.,2004a), the role of BBS proteins in IFT regulation appears to be limited to a specific subset of cilia, possibly those that express homodimeric kinesin-2. However, the role of BBS proteins in cilia assembly and function appears to be complex and is most likely not restricted to coordination of anterograde IFT motors (Blacque and Leroux,2006; Nachury et al.,2007).

Recently, the MAP kinase DYF-5 was implicated as an additional regulator of kinesin-2 motor function in C. elegans. Specifically, it was proposed that DYF-5 modulates the processivity of OSM-3 and serves to restrict kinesin-II to the middle segments (Burghoorn et al.,2007). Homologs of DYF-5 have also been identified and characterized in Leishmania (LmxMPK9) and Chlamydomonas (LF4). In these organisms, mutants lacking the DYF-5 homologous proteins were found to grow elongated flagella consistent with a possible role for DYF-5 in regulating IFT (Bengs et al.,2005; Berman et al.,2003). Whether DYF-5 homologs contribute to IFT/cilia length regulation in vertebrates remains to be determined.

Kinesin-2 and polarity proteins.

In vertebrate cells, there is evidence that certain cell polarity proteins interact with kinesin-II and are required for primary cilia assembly. The Par3/Par6/aPKCζ polarity complex localizes to primary cilia in MDCK and IMCD3 cells and a similar localization was reported in MDCK cells for the Par3-interacting partner 14-3-3η and the Crumbs1-like protein CRB3 (Fan et al.,2004,2007). Depletion of 14-3-3η, CRB3, or Par3 by siRNA- or shRNA-based techniques has been shown to abolish primary cilium formation in MDCK cells (Fan et al.,2004; Sfakianos et al.,2007). Mutations in crumbs genes in zebrafish have also been found to prevent ciliogenesis in a variety of tissues, suggesting a conserved role of polarity proteins in cilia assembly (Omori and Malicki,2006). Interestingly, MT-binding assays, immunoprecipitation, and GST pull-down assays have indicated that Par3 interacts directly with the C-terminal region of KIF3A (Fan et al.,2004; Nishimura et al.,2004b), suggesting that the effect of the Par3/Par6/aPKCζ polarity complex on ciliogenesis is mediated via heterotrimeric kinesin-II. Consistent with this idea, recent work showed that stable knock-down of Par3 in MDCK cells impaired formation of primary cilia, and that this phenotype could be rescued by transfection with full-length Par3 cDNA only, but not with truncated versions lacking the Kif3A binding site. Consequently, it was suggested that the Par3/Par6/aPKCζ complex acts as a scaffold/link between Kif3A and CRB3 and thereby mediates transport of the latter to the cilium (Sfakianos et al.,2007).

Recently, the von Hippel-Lindau tumour suppressor protein (pVHL) was found to interact directly with KIF3A in vitro and in vivo in ciliated kidney cells of 4-day-old mice (Lolkema et al.,2007). Interestingly, pVHL has also been shown to interact directly with the regulatory domain of aPKCζ in HEK293T cells (Okuda et al.,1999) and to co-localize with aPKCζ in MDCK primary cilia (Schermer et al.,2006). aPKCζ inhibits GSK-3β activity by phosphorylation (Etienne-Manneville and Hall,2003) and, in turn, GSK-3β has been demonstrated capable of reducing anterograde transport of membrane-bound organelles in squid axoplasm through phosphorylation of kinesin light chains (Morfini et al.,2002). Since inhibition of GSK-3β affects the regulation of flagellar length in Chlamydomonas (Wilson and Lefebvre,2004), it is possible that GSK-3β may similarly control anterograde transport along the ciliary axoneme. However, the potential roles of pVHL, aPKCζ, and GSK-3β in regulating anterograde IFT may be complex, because in primary mouse embryonic fibroblasts and human renal proximal tubule epithelial cells, pVHL-mediated stabilization of the primary cilium is only required during concomitant inactivation of GSK-3β (Thoma et al.,2007). Furthermore, it should be noted that homozygous aPKCζ null mutant mice appear grossly normal with no obvious indication of ciliary dysfunction (Leitges et al.,2001). Therefore, the results from the cell-based studies described above should be interpreted with some caution.

The Retrograde IFT Motor

The retrograde motor for IFT is an isoform of cytoplasmic dynein called cytoplasmic dynein 2 (Rosenbaum and Witman,2002; for an update on cytoplasmic dynein nomenclature, see Pfister et al.,2005). Dyneins are multiprotein complexes that contain one or more heavy chains (HCs) of the AAA+ family of ATPases as well as several accessory proteins required for regulation of motor activity, structural integrity, and cargo binding (Pfister et al.,2006; Sakato and King,2004). To date, four different subunits of the cytoplasmic dynein 2 complex have been identified and characterized: an isoform-specific HC (DYNC2H1), a light intermediate chain (DYNC2LI1), a putative intermediate chain (IC)/WD repeat protein, and a light chain (LC) (Rompolas et al.,2007, and references therein). Genetic studies in Chlamydomonas revealed a role for the cytoplasmic dynein 2 complex in retrograde IFT as mutants in the gene encoding the DYNC2H1 homolog (called DHC1b in Chlamydomonas) were found to assemble stumpy flagella that were filled with IFT particles, suggesting a defect in retrograde IFT (Pazour et al.,1999; Porter et al.,1999). Using genetics and live imaging of GFP-tagged IFT particles and motor subunits, a similar role for the C. elegans DYNC2H1 homolog, CHE-3, in retrograde IFT was identified (Signor et al.,1999a). The mammalian DYNC2H1 localizes to cilia in various tissues and cell types, including ependidymal cilia of the brain and connecting cilia of photoreceptors (Mikami et al.,2002). Consistent with a role in retrograde IFT, lesions in the mouse Dync2h1 gene leads to assembly of stumpy, bloated cilia in the neuroectoderm and limb mesenchyme, presumably owing to the accumulation of IFT particles at the tip of the mutant cilia (May et al.,2005).

The DYNC2LI1 subunit was first identified in mammals and, subsequently, in Chlamydomonas and C. elegans (Grissom et al.,2002; Hou et al.,2004; Perrone et al.,2003; Schafer et al.,2003). Chlamydomonas mutants that are null in the gene encoding the DYNC2LI1 homolog (D1bLIC) display a phenotype similar to that of DHC1b null mutants in that the mutant cells assemble short flagella that are filled with IFT particles (Hou et al.,2004). Similarly, homozygous Dync2li1 null mutant mouse embryos fail to assemble full-length cilia and exhibit phenotypes that are consistent with ciliary dysfunction, including embryonic lethality (Rana et al.,2004).

A Chlamydomonas mutant, fla14, which contains a lesion in the gene encoding LC8, is defective in retrograde IFT (Pazour et al.,1998), but since LC8 is a constituent of multiple protein complexes within flagella, including axonemal dyneins (King et al.,1996), the exact involvement of LC8 in retrograde IFT has, until recently, been obscure. However, by using a variety of extraction and fractionation procedures, Rompolas and colleagues recently showed that Chlamydomonas LC8 co-fractionates with DHC1b/ DYNC2H1 and D1bLIC/ DYNC2LI1 as well as with the novel WD repeat protein, FAP133. Furthermore, biochemical and immunolocalization analyses of FAP133 in wild type and DHC1b, D1bLIC, fla14 (LC8), or fla10 (kinesin-II) mutant cells strongly suggested that FAP133 is a component of the retrograde IFT motor (Rompolas et al.,2007). FAP133 is closely related to the vertebrate WD34 protein and contains two consensus LC8 binding sites in its N-terminus (Rompolas et al.,2007). Although a direct role for FAP133 and its vertebrate counterpart(s) in retrograde IFT remains to be determined, siRNA-mediated depletion of the trypanosome FAP133 homolog leads to flagellar dysfunction (Baron et al.,2007), consistent with a role for FAP133 in flagellar assembly. A dynein light chain of the Tctex-1 family has been implicated in IFT in C. elegans (Efimenko et al.,2005). Surprisingly, the genome of C. elegans does not appear to encode an obvious FAP133/WD34 homolog, raising the possibility that FAP133/WD34 is specific for motile cilia and flagella (Rompolas et al.,2007). It will be interesting to characterize FAP133/WD34 in vertebrate cells to test this possibility further.

IFT Particle Polypeptides

IFT particles were first purified from Chlamydomonas flagella and were found to consist of at least 16 different polypeptides that sediment in sucrose density gradients as two different 16-17 S complexes termed complex A and B, respectively (Cole et al.,1998; Piperno and Mead,1997). Complex A contains six subunits while complex B contains at least ten. The IFT particle proteins range in apparent molecular mass from 20 to 172 kDa and are, consequently, designated IFT20 through IFT172 (Cole,2003). The genes for almost all these polypeptides have been cloned in Chlamydomonas and sequence analyses have revealed that both complex A and B polypeptides are rich in domains/amino acid repeats that have been implicated in transient protein–protein interactions, including WD40 repeats, coiled coils, and TPR motifs (Cole,2003). The presence of such domains/motifs in IFT particle polypeptides is consistent with the proposed role of IFT in delivering axonemal precursors to the flagellar tip and bringing turnover products back to the cell body (Qin et al.,2004). Recently, five new putative IFT proteins (DYF-1, -2, -3, -13, and IFTA-1) were identified in C. elegans, and although homologous proteins exist in Chlamydomonas and humans, they remain to be characterized in these organisms (Blacque et al.,2008). However, a homolog of DYF-1 (Fleer) was recently characterized in zebrafish, and pronephric cilia of fleer mutant embryos were found to be consistently shorter than their wild type counterparts while olfactory cilia were completely absent in the mutants. Importantly, this effect on ciliogenesis could be correlated with the lack of axonemal tubulin polyglutamylation in the fleer mutants as well as structural defects in B tubule assembly characterized by a gap or discontinuity in the outer aspect of the B tubule. Finally, since the fleer mutant phenotype was phenocopied by knock-out of the gene encoding tubulin polyglutamylase Ttll6, it was suggested that DYF-1 might act as a cargo adaptor protein linking Ttll6 to the anterograde IFT machinery. Ttll6 would then catalyze polyglutamylation of axonemal tubulin, which in turn may be required for appropriate regulation IFT motor processivity (Pathak et al.,2007). Consistent with this idea, the velocity of kinesin-II and OSM-3 was reduced or arrested, respectively, in C. elegans DYF-1 mutant amphid channel cilia (Ou et al.,2005).

Although initial sequence analysis of IFT particle polypeptides yielded little clue as to the specific function of individual IFT particle proteins, detailed functional analyses of many IFT particle proteins have now been performed in Chlamydomonas as well as in other organisms, including vertebrates. These analyses have shown that in virtually all organisms studied complete loss of any complex B polypeptide (IFT20, IFT27, IFT46, IFT52, IFT57, IFT80, IFT81, IFT88, IFT172) generally leads to severely shortened or absent cilia (Beales et al.,2007; Brazelton et al.,2001; Cole,2003; Deane et al.,2001; Follit et al.,2006; Fujiwara et al.,1999; Haycraft et al.,2001,2003; Houde et al.,2006; Huangfu et al.,2003; Pazour et al.,2000; Perkins et al.,1986; Qin et al.,2001,2007; Sun et al.,2004; Tsujikawa and Malicki,2004). In addition, partial inactivation/depletion of some of the complex B proteins (IFT20, IFT27, IFT46, and IFT172) have yielded insights into the specific role of these polypeptides in the IFT process. For example, IFT20 has been demonstrated to localize to the Golgi in addition to cilia, and moderate knock-down of IFT20 in NRK cells resulted in reduced levels of polycystin-2 in the primary cilia of these cells, suggesting that IFT20 is important for transport of membrane proteins from the Golgi to the cilia (Follit et al.,2006). IFT27 was recently found to be involved in cell cycle control in Chlamydomonas, in addition to having a role in flagellar assembly. IFT27 is homologous to Rab-like small G proteins and partial depletion of IFT27 by RNAi in Chlamydomonas results in cytokinesis defects and a prolonged cell cycle, whereas complete depletion of IFT27 is lethal. Moreover, in cells that contain reduced levels of IFT27, the cellular levels of other IFT complex A and B proteins was also reduced, suggesting a role for IFT27 in stabilizing and/or regulating the expression of both complex A and B proteins (Qin et al.,2007). Importantly, in mutant Chlamydomonas cells that lack other complex B proteins (IFT46, IFT52, or IFT88), the cellular levels of IFT27 and complex A proteins are unchanged or elevated, suggesting a unique role for IFT27 in global regulation of IFT particle protein expression and/or stability (Hou et al.,2007; Qin et al.,2007).

IFT88/Polaris is one of the most studied IFT particle polypeptides. A Chlamydomonas mutant that lacks IFT88 fails to assemble flagella (Pazour et al.,2000) and the C. elegans IFT88 ortholog, OSM-5, is also essential for assembly of sensory cilia in that organism (Haycraft et al.,2001; Perkins et al.,1986; Qin et al.,2001). The mouse IFT88 ortholog, known as Polaris, is encoded by the Tg737 gene (Murcia et al.,2000; Pazour et al.,2000) and mutations in this gene result in defective cilia assembly in virtually all tissues studied, including kidney (Pazour et al.,2000), photoreceptors (Pazour et al.,2002), brain (Banizs et al.,2005; Chizhikov et al.,2007), embryonic node (Murcia et al.,2000), and pancreas (Cano et al.,2004). Another complex B protein, IFT46, was recently shown to be required specifically for transporting outer dynein arms into Chlamydomonas flagella (Hou et al.,2007). Finally, a Chlamydomonas mutant, fla11, which contains a point mutation in IFT172, was found to accumulate IFT particle proteins at the flagellar tip suggesting a role for IFT172 in regulating IFT turn-around at this site. Consistent with this idea, IFT172 was found to interact with EB1, which localizes to the flagellar tip (Pedersen et al.,2003,2005).

In contrast to the phenotype observed upon complete loss of complex B polypeptides (i.e., severely shortened or absent cilia in all organisms studies), the phenotype observed upon complete loss of complex A polypeptides varies among different organisms and cell types. Lack of the complex A protein IFT140 results in complete loss of flagella in Chlamydomonas (Cole,2003) while in C. elegans, loss of the complex A polypeptides IFT140/CHE-11 or IFT122A/DAF-10 leads to shortened cilia with accumulations of complex B polypeptides at the tips reminiscent of the stunted cilia observed in cytoplasmic dynein 2 mutants (Collet et al.,1998; Haycraft et al.,2003; Perkins et al.,1986; Qin et al.,2001; Schafer et al.,2003). Finally, in zebrafish, loss of IFT140 was found not to have any overt effect on cilia structure, at least not in the photoreceptors and the auditory and olfactory organs (Tsujikawa and Malicki,2004).

The variable phenotypes observed in different organisms upon loss of complex A polypeptides could be linked to the presence or absence of homodimeric kinesin-2 in the cilia. Homodimeric kinesin-2 appears to be expressed in most C. elegans sensory cilia as well as in zebrafish photoreceptors (Blacque et al.,2008; Luby-Phelps et al.,2007), but not in Chlamydomonas flagella (Cole,2005). Furthermore, in C. elegans sensory cilia that express homodimeric kinesin-2 (OSM-3) as well as heterotrimeric kinesin-II, there is evidence that during anterograde IFT, complex A is associated with kinesin-II while complex B is associated with OSM-3. In these cilia, complex A and B are linked together via BBS proteins such that in the absence of any one of the two kinesin-2 motors, both A and B are transported along the cilia by the remaining motor (Blacque et al.,2008; Ou et al.,2005; Snow et al.,2004). Therefore, in the absence of complex A, one would predict that complex B can still undergo IFT, mediated via OSM-3. If complex B associates with cargo proteins needed for building the ciliary axoneme, as suggested by a recent study in Chlamydomonas (Hou et al.,2007), a cilium would still be assembled in the absence of complex A. However, if complex A is required for retrograde IFT, as suggested by genetic and biochemical studies in both C. elegans and Chlamydomonas (Blacque et al.,2008; Cole,2003; Pedersen et al.,2006; Rompolas et al.,2007), complex B would accumulate near the tip of complex A mutant cilia, which is exactly what has been observed in C. elegans sensory cilia that express OSM-3 (Collet et al.,1998; Haycraft et al.,2003; Schafer et al.,2003).

In cilia that do not contain homodimeric kinesin-2, e.g., Chlamydomonas flagella (Cole,2005), anterograde transport of both complex A and B within the cilium is dependent upon heterotrimeric kinesin-II. If active kinesin-II associates with complex A in these cilia as proposed for C. elegans (Blacque et al.,2008), complex B would not be transported into the cilia in the absence of complex A and no cilia would form. This is consistent with the reported absence of flagella in a Chlamydomonas mutant lacking the complex A protein IFT140 (Cole,2003) and is also consistent with recent biochemical evidence from Chlamydomonas suggesting that complex A associates strongly with kinesin-II as well as cytoplasmic dynein 2 (Rompolas et al.,2007).

In summary, the available evidence suggests that complex B is essential for cilia assembly in all organisms studied, possibly because complex B carries the precursors needed for assembly of the ciliary axoneme. Complex A appears to associate with active kinesin-II as well as active cytoplasmic dynein 2. In cilia that contain a single kinesin-2 motor (kinesin-II), complex A is proposed to be essential for both anterograde and retrograde IFT, while in cilia that also contain homodimeric kinesin-2, complex A seems to be required for retrograde IFT only.

Model for IFT

Based on a large body of genetic and biochemical evidence from Chlamydomonas and C. elegans, we recently proposed a detailed six-step model for IFT in Chlamydomonas (Pedersen et al.,2006). Given that the structure and function of IFT components are highly conserved across species, this model is likely applicable for vertebrate primary cilia as well, except for those that contain homodimeric/OSM-3-type kinesin-2. A model for IFT in cilia that contain homodimeric kinesin-2 has recently been presented elsewhere (Blacque et al.,2008) and will not be described here. In Figure 2, a slightly modified version of the “canonical” IFT model proposed in Pedersen et al. (2006) is presented. This model can be summarized as follows:

Step 1. IFT complexes A and B, kinesin-II, cytoplasmic dynein 2, and axonemal precursor proteins gather around the basal bodies by an unknown mechanism.

Step 2. IFT complexes A and B, inactive cytoplasmic dynein 2, and axonemal precursors are transported in large groups by active kinesin-II from the base to the tip of the outer doublet B MT subfiber. Active kinesin-II is attached to complex A, which in turn is associated with complex B to which axonemal precursors are bound. The attachment site of inactive cytoplasmic dynein 2 during anterograde transport is not clear.

Step 3. Kinesin-II reaches the distal end of the B subfiber, axonemal cargo proteins and IFT particles are released into the ciliary tip compartment and complexes A and B dissociate from each other. Cytoplasmic dynein 2 is also released into the ciliary tip compartment, while kinesin-II remains associated with the B subfiber and becomes inactive. The exact timing of these events is unclear.

Step 4. Following dissociation of IFT complexes A and B at the ciliary tip, and release of inactive cytoplasmic dynein 2 into the tip compartment, complex A binds (directly or indirectly) to active cytoplasmic dynein 2 via the DYNC2LI1 subunit, and complex B then binds to complex A. How and when cytoplasmic dynein 2 gets activated is unknown. Kinesin-II may exit the cilium independent of IFT complexes A and B and DYNC2LI1, perhaps by associating with DYNC2H1 directly or via other subunits such as LCs.

Step 5. After binding (directly or indirectly) of complex A to active cytoplasmic dynein 2, association of complex B with A, and attachment of kinesin-II (directly or indirectly) to DYNC2H1, active cytoplasmic dynein 2 transports everything back from the tip to the cell body.

Step 6. The IFT cycle is completed when IFT components are returned to the cytoplasm for recycling.

The above model is identical to that of Pedersen et al. (2006) with two exceptions. First, we previously suggested that inactive cytoplasmic dynein 2 is associated with complex B during kinesin-II-mediated anterograde IFT (Pedersen et al.,2006). This was based on observations of C. elegans mutants suggesting that kinesin-II-mediated transport of DYNC2LI1 into cilia can proceed in the absence of IFT complex A (Schafer et al.,2003), as well as on immunoprecipitation results from Chlamydomonas indicating that cytoplasmic dynein 2 can associate with complex B independent of complex A (Pedersen et al.,2006). Recently, however, gel filtration of Chlamydomonas flagellar extracts showed that the majority of cytoplasmic dynein 2 co-eluted with IFT complex A in a fraction that also contained kinesin-II, but not IFT complex B (Rompolas et al.,2007), suggesting that our initial hypothesis, i.e., that inactive cytoplasmic dynein 2 attaches to complex B during anterograde IFT (Pedersen et al.,2006), may not be correct. However, since it is not possible to distinguish between active and inactive pools of cytoplasmic dynein 2 in the flagellar extracts analyzed in Pedersen et al. (2006) and Rompolas et al. (2007), more experiments are required to make definite conclusions on this issue. Therefore, the model shown in Figure 2 does not specify the exact IFT component(s) that cytoplasmic dynein 2 associates with during anterograde IFT. Second, in our previous model (Pedersen et al.,2006), IFT cargo proteins were omitted for simplicity. In light of recent evidence from Chlamydomonas suggesting that the IFT complex B polypeptide IFT46 is directly involved in binding to ciliary precursors such as axonemal outer dynein arm proteins (Hou et al.,2007), and because lack of complex B polypeptides in general leads to complete inhibition of cilia formation in virtually all organisms studied (see IFT Particle Polypeptides section), it is now proposed that complex B functions as a cargo-binding module to which ciliary precursors attach during anterograde IFT. For further details regarding the model shown in Figure 2 (see Pedersen et al.,2006).

TARGETING OF PROTEINS TO THE CILIARY COMPARTMENT

Although IFT takes place in the cilium, the principal location of IFT particle proteins and motor subunits is around the basal bodies near the transition fibers that link the distal end of the basal body to the plasma membrane (Deane et al.,2001). It is thought that this region constitutes a selective filter or pore that allows only a specific subset of cellular proteins to enter the flagellar compartment (Rosenbaum and Witman,2002). Structural and functional similarities between the ciliary necklace/pore region and the nuclear pore complex (NPC) have previously been suggested (Christensen et al.,2007; Jekely and Arendt,2006). This is supported by recent findings that importin β-1, an essential component of the NPC that interacts with Ran to control nuclear import and export, localizes to centrosomes and primary cilia of MDCK cells, and expression of dominant-negative importin β-1 inhibits cilia formation in these cells (Fan et al.,2007). Recently, nephrocystins have been suggested to be part of the ciliary pore complex by virtue of their localization to the ciliary transition zone and because genetic lesions in genes coding for nephrocystins have been associated with ciliary disease (von Schnakenburg et al.,2007). However, more experimental evidence is required to confirm the possible role of nephrocystins in the trafficking/sorting of proteins to the ciliary compartment.

Membrane proteins destined to the ciliary compartment appear to be transported in vesicles from the Golgi apparatus to the ciliary necklace region where exocytosis occurs, and further transport into the ciliary compartment is mediated via IFT (Reiter and Mostov,2006; Rosenbaum and Witman,2002). For some ciliary proteins, specific ciliary targeting motifs or sequences have been identified (reviewed in Christensen et al.,2007) and some of the players involved in Golgi-to-cilia trafficking have also been identified. These include IFT20 (Follit et al.,2006), FAPP2 (Reiter and Mostov,2006), and possibly Bardet Biedl Syndrome (BBS) proteins and their binding partners (Nachury et al.,2007) (see Molecules Involved in Early Stages of Ciliogenesis section).

CONCLUSION

Since the initial EM observations of primary cilia development in the 1960s and 1970s (e.g., Sorokin,1962), significant progress has been made towards elucidating the molecular mechanisms by which these fascinating and physiologically important organelles are assembled. During early stages of primary cilia assembly, the basal body migrates towards the apical cell surface where it docks onto the cell membrane. Molecules involved in these events include MKS1 and 3 (basal body migration) and the centrosomal proteins Odf2 and Cep164 (basal body docking). Other centrosome-associated proteins that have been found to be involved in primary cilia assembly include Pericentrin, Centrin 2, PCM-1, and Alms1. PCM-1 interacts with BBS proteins, which in turn interact with the p150Glued subunit of the dynactin complex. It is possible that the latter proteins play a role in transport of proteins/vesicles to the ciliary base, a process that might also involve the MT-binding protein EB1. Future efforts should be directed towards elucidating the structural and functional relationships between these centrosome-associated proteins and the mechanisms by which they function in ciliary assembly.

Following docking of the basal body to the apical plasma membrane, the ciliary axoneme is elongated via IFT. Since its initial discovery in Chlamydomonas 15 years ago (Kozminski et al.,1993), research in IFT has increased exponentially in recent years due to the discoveries that IFT is essential for assembly of virtually all eukaryotic cilia and flagella, and that defective IFT may lead to severe human diseases and developmental defects. Almost all the IFT components (anterograde and retrograde IFT motor subunits, IFT particle polypeptides) have now been identified and characterized in multiple organisms, including vertebrates such as mouse and zebrafish. The available evidence suggests a model for IFT in which heterotrimeric kinesin-II is associated with IFT complex A and functions as a canonical anterograde IFT motor that transports IFT complex B-associated ciliary building blocks towards the tip of the axoneme for assembly. Axonemal turnover products are subsequently returned to the cell body via cytoplasmic dynein 2, a multimeric motor complex consisting of isoform-specifc HCs, light intermediate chains, LCs, and a newly identified WD repeat protein/IC. In certain subsets of cilia with specific sensory capacities, a homodimeric kinesin-2 motor (OSM-3/KIF17) functions somewhat redundantly with heterotrimeric kinesin-II during anterograde IFT and contributes to the generation of ciliary structural and functional diversity. How widely KIF17 is expressed in different vertebrate cell types is not yet clear and the exact contribution of KIF17 to global ciliary structure and function in vertebrates remains to be determined.

Finally, recent research has begun to shed some light on the mechanisms by which IFT, notably anterograde IFT, is regulated. The regulatory mechanisms involve BBS proteins, MAP kinases, tubulin polyglutamylation, and possibly polarity proteins such as the Par3/Par6/aPKCζ complex, pVHL, and GSK-3β. Comparatively little is known about how the retrograde IFT motor (cytoplasmic dynein 2) is regulated, but with improved methods for purifying this motor complex and identifying its component subunits and interaction partners (Rompolas et al.,2007), studies on its regulation should be possible in the future.

Acknowledgements

We thank D. N. Wheatley for helpful discussions regarding early stages of ciliogenesis and S. Geimer for the EM picture of Chlamydomonas IFT particles (Fig. 2). This work was supported by grants from the Danish Natural Science Research Council (272-05-0411), The Novo Nordisk Foundation, and The Lundbeck Foundation to L.B.P. and S.T.C. J.M.S. is supported by a PhD fellowship from the Faculty of Science, University of Copenhagen, and I.R.V. is the recipient of a scholarship from Novo Nordisk.

Ancillary