|Home | About | Journals | Submit | Contact Us | Français|
The BBSome is a complex of Bardet-Biedl Syndrome (BBS) proteins that shares common structural elements with COPI, COPII and clathrin coats. Here we show that the BBSome constitutes a coat complex that sorts membrane proteins to primary cilia. Biochemically, the BBSome is the major effector of the Arf-like GTPase Arl6/ BBS3. In vivo, the BBSome and Arl6 localize to ciliary punctae and Arl6GTP is required to target the BBSome to cilia. Congruently, GTP-bound Arl6 and acidic phospholipids are sufficient to efficiently recruit the BBSome to chemically defined liposomes. Finally, ultrastructural analyses demonstrate that BBSome binding to liposomes produces distinct patches of polymerized coat. Since we establish that the ciliary targeting signal of somatostatin receptor 3 needs to be directly recognized by the BBSome to mediate targeting to cilia, we propose that trafficking to cilia entails the coupling of BBSome coat polymerization to the recognition of sorting signals.
Primary cilia are microtubule-based projections found on nearly every cell in the human body. Since primary cilia are required for phototransduction, olfaction, planar cell polarity and Hedgehog signaling and since the receptor for each of these signaling pathways has been localized to the primary cilium (Fliegauf et al., 2007), the targeting of signaling receptors to cilia is thought to be crucial for signal sensing and transduction. Yet, our understanding of membrane protein targeting to cilia remains fragmentary (Nachury et al., 2010).
The delivery of membrane proteins to cilia sequentially entails sorting and packaging into carrier vesicles, docking and fusion of vesicles with the base of the cilium and intraflagellar transport (IFT) from cilia base to cilia tip (Rosenbaum and Witman, 2002). The step of docking and fusion requires the GTPase Rab8 and its guanine nucleotide exchange factor (GEF) Rabin8 (Moritz et al., 2001; Nachury et al., 2007). The sorting step most likely relies on the recognition of a ciliary targeting signal (CTS) by a sorting complex and CTSs have been identified in several ciliary membrane proteins (Tam et al., 2000; Berbari et al., 2008b; Follit et al., 2010). While the GTPases Arf4 and Rab8 have been shown to recognize the CTSs of rhodopsin and fibrocystin respectively (Mazelova et al., 2009; Follit et al., 2010), it is expected that coat complexes resembling COPI, COPII and clathrin carry out the sorting of membrane proteins to cilia. Canonical coat complexes are recruited to membranes by phosphoinositides (PIPs) and –in most cases– by a GTP-bound Arf-family GTPase and the direct recognition of sorting signals by coat complexes ensures that coat polymerization packages transmembrane cargoes into a carrier vesicle (McMahon and Mills, 2004). Thus far, no coat complex has been identified for trafficking to cilia.
Recently, we discovered the BBSome, an octameric complex consisting of the seven highly conserved Bardet-Biedl syndrome (BBS) proteins BBS1, BBS2, BBS4, BBS5, BBS7, BBS8 and BBS9 and of the novel protein BBIP10 (Nachury et al., 2007; Loktev et al., 2008). Bardet-Biedl Syndrome is an autosomal recessive disorder characterized by retinal degeneration, polydactyly, kidney cysts and obesity that can be caused by mutations in any of 14 known genes and whose etiology is associated with cilium dysfunction (Fliegauf et al., 2007). Since the BBSome binds Rabin8 and associates with the ciliary membrane and since BBS5 binds PIPs on protein-lipid overlays, we have proposed that the BBSome functions in vesicular trafficking to the cilium (Nachury et al., 2007). In support of this hypothesis, the G Protein coupled receptor (GPCR) Somatostatin receptor 3 (SSTR3) fails to reach the cilium of hippocampal neurons in bbs2 and bbs4 knockout mice (Berbari et al., 2008a). However, a role for the BBSome in vesicular transport remains controversial, with alternative roles in microtubule anchoring (Kim et al., 2004), intraflagellar transport (Ou et al., 2005; Lechtreck et al., 2009) and ubiquitination (Gerdes et al., 2007) having been proposed. Regardless of the model considered, the definite molecular activity of the BBSome remains unknown.
Interestingly, BBS3 encodes the small Arf-like GTPase Arl6 which is not part of the BBSome and whose function remains uncharacterized (Fan et al., 2004; Chiang et al., 2004). We now show that Arl6GTP recruits the BBSome onto membranes to assemble an electron-dense coat and that the BBSome sorts SSTR3 to cilia by directly recognizing SSTR3’s CTS. Thus, the BBSome constitutes a coat complex that sorts membrane proteins to cilia.
We first sought to identify effectors of Arl6 by affinity chromatography. Mutations were introduced into Arl6 to preclude GTP hydrolysis (Q73L) and to limit aggregation of the GTP-bound form (ΔN16). We chose retinal extract as a starting material because of the tremendous rates of membrane protein trafficking to cilia in photoreceptors. Remarkably, eight protein bands were recovered specifically in the eluate of the Arl6GTP column and were identified as the eight subunits of the BBSome (Figure 1A). Further, direct “in-solution” mass spectrometry analysis of the eluates failed to identify any Arl6 effector besides the BBSome subunits. TACT1, the only other protein specifically recovered in the Arl6GTP eluate, binds directly to the BBSome and likely binds to Arl6GTP indirectly (Figure 1B; TS and MVN, unpublished). Furthermore, immunoblotting showed that over 75% of the BBSome was depleted by the Arl6GTP column and recovered in the Arl6GTP eluate while no BBSome binding to Arl6GDP and GST was detected (Figure 1C). Thus, the BBSome is the major Arl6 effector in retinal extracts. We further confirmed the BBSome-Arl6GTP interaction by showing that BBS1 was the BBSome subunit most efficiently captured by Arl6GTP and by mapping the interaction domain to the N-terminus of BBS1 (Figure 1D). Since Arl6 is the only BBS gene besides the BBSome subunits to be universally conserved in ciliated organisms, these results tie all of the conserved BBS proteins into two connected biochemical units and suggest a conserved function for the BBSome/ Arl6GTP interaction.
The finding that the BBSome is the major effector of an Arf-like GTPase suggested that the molecular activity of the BBSome may be related to that of coat complexes. We therefore set out to validate the BBSome coat hypothesis at the biochemical, structural and functional levels. First, we explored the structural anatomy of the BBSome using sensitive structure prediction algorithms. We extended previous findings (Kim et al., 2004) and discovered that BBS4 and BBS8 are almost entirely comprised of TPR repeats (Jínek et al., 2004) and are therefore predicted to fold into extended rod-shaped α-solenoids (Figure 2A) . Meanwhile, BBS1, BBS2, BBS7 and BBS9 share a related β-propeller fold (Chaudhuri et al., 2008) in their N-termini (Figure 2A) and a domain distantly related to the immunoglobulin (Ig)-like β-sandwich fold of the γ-adaptin ear (GAE) motif in their C-termini (Figure 2B). In BBS2, BBS7 and BBS9, the GAE domain is further accompanied by an α/ β platform domain (Figure 2C). In several clathrin adaptors and in COPI, the GAE motif –either by itself of fused to the α/ β platform– constitutes the so-called appendage domain that recruits either regulators of coat assembly or factors that program the coated vesicle for subsequent targeting events (McMahon and Mills, 2004). In the BBSome, Rabin8 binds to the C-terminus of BBS1 and it is conceivable that Rabin8 serves as a BBSome accessory factor. Since rigid α-solenoids and β-propellers form the architectural scaffolds of COPII and clathrin cages (Stagg et al., 2007), the abundance of β-propellers, α-solenoids and appendage domains inside the BBSome suggests an ancient evolutionary relationship between the BBSome and canonical coat complexes (Figure 2D).
We next sought to identify the compartment(s) where the BBSome/ Arl6GTP interaction takes place. We raised a polyclonal antibody against Arl6 and validated its specificity by immunoblotting lysates of RPE1-hTERT (RPE) cells transfected with Arl6 siRNA (Figure S1A). RPE cells grow a primary cilium when switched into quiescence and we previously showed that the BBSome subunits BBS4 and BBIP10 localize to cilia in RPE cells. We extended these findings by showing that endogenous BBS1 (Figure 3C), BBS2 (Figure S2A) and BBS5 (Figure S2B) all localized to cilia. Thus, we conclude that the BBSome is present in mammalian cilia as a complex as was recently shown in Chlamydomonas (Lechtreck et al., 2009). Remarkably, our anti-Arl6 antibody stained cilia (Figure 3A) and cilia staining was lost after siRNA-mediated depletion of Arl6 (Figure 4A, left panels). To accurately determine the distribution of Arl6 and the BBSome within cilia, a structure whose 300-nm diameter cannot be resolved by conventional light microscopy, we resorted to structured illumination microscopy, a “super-resolution” technique that lowers the optical resolution to less than 50 nm (Schermelleh et al., 2008). Arl6 staining appeared in a pattern of punctae flanking the microtubule axoneme that likely correspond to small membrane-associated patches (Figure 3B and Movie S1). Further, deconvolution microscopy could resolve a discrete pattern of Arl6 staining within cilia that precisely mirrored the distribution of the BBSome subunits BBS1 and BBIP10 (Figures 3C, S1B and C). Thus, the interaction between Arl6 and the BBSome may take place within these intraciliary patches.
Together with our biochemical data, the co-localization studies suggested that Arl6GTP may recruit the BBSome to cilia. Here, we found that the BBSome subunits BBS1 and BBIP10 failed to localize to cilia when we depleted Arl6 by siRNA (Figure 4A–C). While treatment with two different siRNAs targeting Arl6 dramatically decreased the abundance of Arl6 protein, the abundance of the BBSome subunit BBS4 remained unaffected (Figure 4B) and BBS4 still migrated as part of a 500 kDa complex on size exclusion chromatography in the absence of Arl6 (Figure S3A). Thus, Arl6 is specifically required for BBSome localization to cilia but not for BBSome assembly. Next, we generated clonal RPE cell lines expressing moderate levels of Arl6 variants to determine whether GTP binding and hydrolysis by Arl6 were required for targeting the BBSome to cilia (Figure 4E and Figure S3B). While Arl6-GFP and Arl6[Q73L]-GFP were both found inside cilia, Arl6[T31R]-GFP, a variant deficient in GTP but not GDP binding (Kobayashi et al., 2009), was absent from primary cilia (Figure 4D). To assess the contribution of Arl6-GFP to BBSome targeting to cilia, we selectively depleted endogenous Arl6 using an siRNA targeting the 3’ UTR of Arl6 mRNA (Figure 4E). While the localization pattern of the Arl6-GFP variants remained unaffected, only Arl6-GFP and Arl6[Q73L]-GFP supported BBSome targeting to cilia (Figure 4D and F). Furthermore, measurements of BBS1 immunoreactivity inside cilia showed that Arl6[Q73L]-GFP recruited a greater amount of BBSome to the cilium than Arl6-GFP (Figure S3C). We conclude that Arl6 and BBSome targeting to cilia both require Arl6 binding to GTP but not Arl6 GTPase activity. Conversely, depletion of the BBSome subunits BBS2, BBS4 and BBS5 resulted in a dramatic decrease of Arl6 staining within cilia (Figure S4A).
Interestingly, we noted that the fraction of cells with BBSome- or Arl6-positive cilia varied from 15 to 60% depending on the experiment (compare Figures 4C, 4F and S4A). While the source of the variability remains unknown, we hypothesized that the levels of Arl6 and the BBSome in most cilia fall below the detection threshold of our traditional immunofluorescence protocol. We therefore developed a method that decreases background staining while preserving the signals in cilia (see Supplemental Experimental Methods and Figure S4C–G); and we now found that nearly every RPE cilia stains positive for Arl6 and BBS1 (Figure S4G). This increase in the proportion of Arl6- and BBS1-positive cilia is not simply an artifact of the new staining procedure since Arl6 and BBS1 staining are still lost from cilia when Arl6 is depleted by siRNA. Together, these results demonstrate the interdependence of Arl6GTP and BBSome targeting to cilia and suggest that Arl6GTP and the BBSome synergize in binding to the membrane of the cilium.
We first tested whether Arl6 behaves like Arf1 and Sar1, i.e. binds to membranes upon GTP binding by exposing an amphipathic helix that inserts itself in the lipid shell and terminates in a basic collar that interacts with phospholipid headgroups (Gillingham and Munro, 2007). Helical representation of the N-terminus of Arl6 demonstrates the amphipathic nature of the N-terminal helix of Arl6 and its termination with several positively charged residues (Figure 5A and B). Since Arl6 is predicted not to be myristoylated (Gillingham and Munro, 2007), we expressed Arl6 in bacteria and purified it to homogeneity (Figure S5A). As mammalian cilia cannot be isolated in sufficient quantities to generate a pure lipid fraction, we conducted sedimentation assays with liposomes made from brain lipids (Folch fraction I). Here, we found that recombinant Arl6 efficiently bound to liposomes in the presence of the slowly hydrolyzable analogue GMP-PNP but not in the presence of GDP or when the N-terminal amphipathic helix was removed (Figure 5C). We therefore conclude that Arl6 conforms to the Arf1/ Sar1 paradigm and interacts with membranes through its N -terminal amphipathic helix when GTP-bound.
To determine the minimal requirements for BBSome binding to liposomes, we next needed a highly purified BBSome fraction. Such a fraction was obtained by fractionating retinal extract over the Arl6GTP column (Figure 1A) followed by cation exchange chromatography (Figure 5D). The purified BBSome was nearly free of contaminants as assessed by silver staining and behaved as a monodisperse complex devoid of aggregates by rate zonal sedimentation (Figure S5B). We then performed sedimentation assays with purified BBSome, Arl6, guanine nucleotides, and liposomes made from brain lipids and found that efficient binding of the BBSome to liposomes required Arl6 and GMP-PNP (Figure 5E). We have thus reconstituted the recruitment of the BBSome to membranes from purified components in vitro and no protein factor other than GTP-bound Arl6 is required for this binding.
Given the robust interaction between Arl6GTP and the BBSome on one hand and between Arl6GTP and liposomes on the other hand, it was conceivable that the BBSome recruitment to liposomes was strictly indirect and did not involve any contact between the BBSome and lipid headgroups. However, COPI and COPII coats and clathrin adaptors have all been shown to directly contact acidic phospholipids or specific PIPs (Matsuoka et al., 1998; Spang et al., 1998; Bremser et al., 1999; McMahon and Mills, 2004) and those contacts are likely important in the sculpting of buds and vesicles by the polymerizing coat. To determine whether specific lipids are required for the binding of the BBSome to membranes, we made liposomes from synthetic phospholipids. Given that the lipid composition of mammalian cilia is not known, we started with a base mixture (dubbed “major mix”) that allows for the efficient capture of COPI, COPII and exomer coat complexes and that functions in COPI and COPII budding reactions (Matsuoka et al., 1998; Spang et al., 1998; Wang et al., 2006). The phospholipid part of the major mix contains 76 mol% of neutral phospholipids (53 mol% phosphatidylcholine [PC] and 23 mol % phosphatidylethanolamine [PE]) and 24 mol% of acidic phospholipids (8 mol% phosphatidylserine [PS], 5 mol% phosphatidic acid [PA], 11 mol% phosphatidylinositol [PI]). The cholesterol/ phospholipid molar ratio is 23/ 77 and all acyl chains are oleoyl (18:1) to keep fluidity high through the 4–30°C range of temperatures. To preclude any background signal stemming from protein precipitation during the course of the incubation with liposomes, we isolated the protein complexes bound to liposomes by buoyant density flotation on miniature iodixanol step gradients.
Initial experiments showed moderate binding of the BBSome to major mix liposomes in the presence of Arl6GMP-PNP (Figure 5F, lane 1). Since we have previously shown that the BBSome subunit BBS5 binds to PIPs on protein lipid overlays, we replaced a portion of the PI in the major mix with one of seven individual PIPs. While the specificity of BBSome binding for a specific PIP was somewhat variable (Figure S5C), multi-phosphorylated PIPs (in particular PI(3,4)P2) enhanced BBSome binding to liposomes by as much as 3-fold. This enhanced BBSome binding did not result from a general stickiness of PIP liposomes as BBSome binding to PI(3,4)P2 liposomes was strongly Arl6- and GMP-PNP-dependent (Figure 5G, lane 3 and 4; Figure S5D). We note that recombinant BBS5 exhibits a different PIP specificity on lipid blot overlays (Nachury et al., 2007) than the BBSome does on liposomes and conclude that BBS5 and individual lipids taken out of their physiological environments may not faithfully recapitulate the specificity of the BBSome complex for lipids in hydrated bilayers.
The preference for a lipid bearing a strong negative charge by the BBSome suggested that other acidic phospholipids in the major mix might participate in BBSome recruitment to membranes. We therefore tested Arl6- and GMP-PNP-dependent binding of the BBSome to liposomes made from neutral lipids. Since PC/ PE liposomes were less effective than PI(3,4)P2 liposomes at recruiting Arl6GMP-PNP, the molarity of Arl6 was adjusted to recover similar amounts of Arl6GMP-PNP regardless of the liposome composition. Despite the 5-fold increase in Arl6 molarity, the binding of Arl6 to liposomes was still strongly dependent upon the addition of GMP-PNP (Figure 5G, lane 1 and 2). While the amount of Arl6GMP-PNP recovered remained relatively unchanged, BBSome binding was drastically reduced in the absence of charges on the bilayer surface (Figure 5G, lane 1 and 3). Thus, Arl6GMP-PNP binds membranes through neutral and acidic phospholipids and Arl6GMP-PNP and acidic lipids (in particular multi-phosphorylated PIPs) synergize to recruit the BBSome to membranes.
We next investigated the morphological consequences of BBSome binding to liposomes. Incubation of COPI, COPII, clathrin and exomer coat components with liposomes leads to the formation of coated profiles and, in the case of COPI and COPII, the budding of 50-nm diameter coated vesicles (Matsuoka et al., 1998; Spang et al., 1998; Bremser et al., 1999; Wang et al., 2006). Liposomes consisted of uni- and multilamellar structures with smooth bilayer surfaces when visualized by thin section electron microscopy. After incubation with BBSome, Arl6 and GMP-PNP, close to 10% of all liposomes showed coated surfaces (Figure 6A). The coated profiles appeared as continuous and well delineated patches clearly separated from noncoated surfaces (Figure 6A and Figure S6) and repeat units could be distinguished at high magnification (Figure 6B), suggesting formation of an ordered polymer. Similar to the exomer coat, no membrane deformation was seen with the BBSome coat proteins and coated profiles retained the normal liposome curvature. When GMP-PNP was replaced with GDP (Figure 6C and D) or when BBSome was omitted (Figure 6E and F), no coated profiles were visible. Thus, upon recruitment to membranes by Arl6, the BBSome appears to polymerize into an electron dense coat associated with the liposome surface.
The polymerization of the BBSome into a membrane coat strongly implied that the BBSome sorts specific membrane proteins (i.e. cargoes) inside the cell. A strong candidate for BBSome cargo is SSTR3 which is lost from cilia in bbs2−/− and bbs4−/−hippocampal neurons (Berbari et al., 2008a). We extended these results by showing that the number of SSTR3-positive cilia decreases dramatically when Arl6 is depleted from cultured hippocampal neurons by lentivirus-mediated shRNA (Figure 7A and B).
The hypothesis that SSTR3 is a bona fide BBSome cargo predicts that the BBSome directly recognizes the CTS of SSTR3, which was previously mapped to the third intracellular loop (i3) (Berbari et al., 2008b). This prediction was tested by expressing SSTR3i3 fused to GST in bacteria, and conducting a GST capture assay with purified retinal BBSome. While GST-SSTR3i3 efficiently captured the purified BBSome, GST alone or GST fused to the third intracellular loop of the closely related GPCR SSTR5 failed to recover detectable amounts of BBSome (Figure 7C). Further, each BBSome subunit expressed in HEK cell bound to SSTR3i3 but not to SSTR5i3 (Figure S7A), indicating that SSTR3i3 is not recognized by a single BBSome subunit but rather by the holo-BBSome which assembles around BBSome subunits expressed in HEK cells (Seo et al., 2010).
Sequence analysis of multiple GPCRs targeted to cilia has suggested that the CTS of ciliary GPCRs centers around the conserved motif AX[S/ A]XQ and mutating the first and fifth amino acids of this motif to phenylalanine within SSTR3i3 may prevent targeting of a SSTR5-SSTR3i3 chimera to cilia (Berbari et al., 2008b). Surprisingly, SSTR3i3[AQ-FF] bound to the BBSome more efficiently than its wild-type counterpart (Figure 7D). We therefore targeted alternative amino acids within the AP[S/ A]CQ motifs of SSTR3i3 (Figures 7E and S7B) and found that the cysteine at the fourth position was the only amino acid required for BBSome binding (Figure 7D). To then assess the CTS activity of the various SSTR3/ 5i3 variants, we spliced them into the cytoplasmic tail of the plasma membrane protein CD8α and transiently expressed the chimeras in IMCD3 cells. While CD8α and CD8α-SSTR5i3 failed to efficiently target to cilia, CD8α-SSTR3i3 was transported to cilia in more than 90% of transfected cells. Most importantly, the targeting of CD8α–SSTR3i3[C-A] to cilia was severely impaired compared to CD8α-SSTR3i3 (Figure 7F) while targeting of CD8α-SSTR3i3[AQ-FF] to cilia remained unaffected (Figure S7B). Thus, the molecular recognition of SSTR3i3 by the BBSome is essential for the full CTS activity of SSTR3i3.
Finally, we wished to pinpoint the compartment where CD8α-SSTR3i3 accumulates in the absence of BBSome or Arl6 function. To this end, we generated a stable cell line expressing low levels of CD8α-SSTR3i3. While CD8α-SSTR3i3 was targeted to cilia in >95% of cells treated with a control siRNA, depletion of Arl6 or BBS4 led to a pronounced decrease in ciliary targeting of CD8α-SSTR3i3 (Figure 7G and Figure S7C). Importantly, the total levels of CD8α-SSTR3i3 remained unchanged by depletion of Arl6 or BBS4 (Figure 7H “Total”) thus supporting the interpretation that the BBSome coat sorts the synthetic cargo to cilia rather than stabilizes it. Interestingly, examination of CD8α-SSTR3i3 localization in cells depleted of Arl6 or BBS4 revealed significant plasma membrane staining not observed in cells treated with control siRNA (Figure S7D). To rigorously test whether CD8α-SSTR3i3 accumulated at the plasma membrane in the absence of BBSome or Arl6, we conducted surface biotinylation followed by capture on avidin beads. Remarkably, we found that the levels of surface-exposed CD8α-SSTR3i3 remained unchanged in the absence of Arl6 or BBS4 (Figure 7H “Surface”). Thus, we conclude that a prototypical BBSome cargo normally localized in the ciliary membrane accumulates at the plasma membrane in the absence of BBSome coat function.
Together, these results establish SSTR3i3 as an Arl6- and BBSome-dependent CTS and suggest that the BBSome carries out the trafficking of SSTR3i3-containing cargoes from the plasma membrane to the ciliary membrane.
Since the BBSome forms a coat and reads the sorting signals of its cargoes to direct them to the cilium, the decision to sort membrane proteins such as SSTR3 toward the cilium is almost certainly made on the compartment where the BBSome coat assembles. Since a synthetic BBSome cargo is detected at the plasma membrane in the absence of BBSome function, a plausible model would have BBSome cargoes diffuse laterally in the plasma membrane until their sorting signals become recognized by the BBSome. The ensuing assembly of a planar BBSome coat would cluster these cargoes into a patch that can be dragged through the periciliary diffusion barrier separating plasma and ciliary membranes (Nachury et al., 2010). Such a scenario may explain how transmembrane proteins such as Smoothened enter the cilium by lateral diffusion from the plasma membrane (Milenkovic et al., 2009).
Once inside the cilium, the patch of BBSome coat is predicted to become transported by the IFT machinery. In nematodes, Chlamydomonas and human cells, the BBSome has been shown to undergo intraflagellar motility at the same rates as known IFT polypeptides and it has been proposed that the BBSome functions as an adaptor between the IFT complexes and IFT cargoes (Blacque et al., 2004; Nachury et al., 2007; Lechtreck et al., 2009). The observation that the BBSome assembles a coat in the absence of IFT polypeptides suggests that polymerization of a BBSome layer could drive polymerization of an IFT layer. These BBSome/ IFT patches would possess one layer of BBSome coat connected to the membrane-bound cargoes and one layer of IFT -A and B complexes bound to the IFT motors kinesin II or dynein 1b. The existence of a bilayered IFT/ BBSome coat would explain why IFT-A and B complexes fail to maintain cohesion in the absence of BBSome function (Ou et al., 2005).
While Arl6 and the BBSome are generally not required for ciliogenesis (see Supplemental Discussion), IFT function is universally required for cilium formation. A possible explanation resides in the fact that BBSome only transports a specific set of transmembrane proteins to cilia while the IFT complexes are likely required for all transport processes inside cilia.
Unlike COPI and COPII coat formation, BBSome coat assembly in vitro did not sculpt membranes into buds and 50-nm vesicles. While the BBSome may strictly resemble the clathrin plaques that cluster cargoes at the plasma membrane without deforming membranes (Saffarian et al., 2009), technical issues may have prevented vesicle budding upon BBSome coat formation in our in vitro system. First, there may be specific soluble factors that assist the BBSome coat in deforming membranes. In the example of clathrin, the large GTPase dynamin and local actin polymerization are required for the invagination and scission of clathrin-coated pits and patches (Koenig and Ikeda, 1996; Saffarian et al., 2009). Second, the presentation of sorting signals on the surface of the lipid bilayer and their capture by the BBSome may be necessary for budding coupled to BBSome coat assembly, as is the case for COPI when using liposomes mimicking Golgi membranes (Bremser et al., 1999). Finally, the concentrations of BBSome used in the present study (50 nM) may not be sufficiently high to permit the assembly of a BBSome coat competent to deform membranes. COPI-mediated budding was conducted using concentrations approaching 1.5 µM of coatomer (Bremser et al., 1999).
If the BBSome assembles a canonical coat that sculpts membranes into buds and vesicles, where would this budding reaction take place within the cell? Since the steady-state localization of all known coat complexes is the organelle where they sort cargo and deform membranes, one would predict that the BBSome buds endocytic vesicles off of the ciliary membrane. As the passage of vesicles between the ciliary lumen and the cytoplasm is topologically unfeasible (Nachury et al., 2010), it appears more plausible that the BBSome buds vesicles off of the base of the cilium to remove membrane proteins from cilia as suggested by Lechtreck et al. (2009). However to account for BBSome-mediated sorting of SSTR3i3 to cilia, one needs to invoke the budding of vesicles by the BBSome from a donor compartment that is distinct from cilia. Since we have never observed any BBSome or Arl6 staining on endomembrane compartments or at the plasma membrane in a number of different cell lines using well-validated antibodies, we would need to postulate that, unlike all known coat complexes, the steady-state localization of the BBSome does not correspond to the donor compartment for BBSome-mediated budding. Although unlikely, it is formally possible that the population of BBSome and Arl6 we observe within cilia corresponds to a slowly recycling pool that becomes injected into cilia after the fusion of BBSome-coated vesicles with the base of the cilium.
Regardless of the model considered, only one protein, the small GTPase Arl6, is necessary and sufficient for recruiting the BBSome to membranes. No other soluble protein and no membrane proteins are required for BBSome binding to liposomes. Most strikingly, the biochemical output of the BBSome binding to membranes is the assembly of a thin electron-dense coat on the bilayer surface whose morphology is clearly distinct from COPI, COPII, clathrin and exomer coats. While the process we have recapitulated in the test tube informs the minimal requirements for BBSome coat formation, BBSome coat assembly in vivo must occur with a high degree of spatial and temporal specificity. First, the loading of GTP onto Arl6 is likely rate-limiting for BBSome coat assembly in vivo and an Arl6GEF may locally activate Arl6 to enable BBSome coat assembly in vivo. Second, PCM-1, a major BBSome-associated protein, does not co-purify with the BBSome on Arl6GTP chromatography. PCM-1 may therefore prevent BBSome binding to Arl6GTP and removal of PCM-1 from the BBSome may be a prerequisite for BBSome coat assembly. Third, the requirement for acidic lipids –possibly specific PIPs– to efficiently recruit the BBSome to membranes in vitro suggests that local lipid composition may dictate where the BBSome coat assembles. While no PIPs have been detected in cilia thus far, the 5-phosphatase INPP5E that converts PI(3,4,5)P3 to PI(3,4)P2 is found in cilia and loss of INPP5E leads to several ciliopathies closely related to BBS (Jacoby et al., 2009; Bielas et al., 2009). Furthermore, loss of the PI(3,4)P2-binding protein tubby leads to obesity and retinal degeneration in mice (Santagata et al., 2001). Since tubby genetically interacts with bbs-1 in worms (Mak et al., 2006), it is tempting to speculate that tubby and the BBSome may function on the same membrane compartment.
The BBSome coat model suggests that the variety of symptoms found in BBS patients is likely to result from the failure to transport signaling receptors to the cilium. While the relevance of SSTR3 to the etiology of BBS is currently unknown (Einstein et al., 2010), the observation that the leptin receptor interacts with BBS1 (Seo et al., 2009) provides a tantalizing hypothesis for the molecular basis of obesity in BBS, namely that leptin signaling may take place within primary cilia in a BBSome-dependent manner. The discovery of signaling receptors transported by the BBSome promises to uncover the signaling defects that underlie BBS and to provide mechanistic insights into the interplay between ciliary trafficking and signaling pathways.
Arl6 antibodies were raised in rabbits and purified following standard protocols. All other antibodies are described in supplemental methods. All chemicals were purchased from Sigma unless otherwise indicated. 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC), 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE), 1,2-dioleoyl-sn-glycero-3-phospho-L-serine (DOPS), 1,2-dioleoyl-sn-glycero-3-phosphate (DOPA) were purchased from Avanti polar lipids. PI and PIPs were from Echelon or AG Scientific. Texas Red 1,2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine (DHPE) was from Invitrogen. Chemicals for electron microscopy were from EM Sciences.
Cells were fixed in PBS containing 4% paraformaldehyde for 5 min at 37°C, followed by extraction with cold methanol at −20°C for 5 min and processing for immunofluorescence as described (Nachury et al., 2007). Nuclei were stained with Hoechst 33258. Unless otherwise indicated in the figure legend, stacks of 24 z-sections were acquired at 0.25 µm interval with a 100x/ 1.45 NA objective, deconvolved by constrained iterative, and the section containing the cilium was selected for each figure panel. Enhanced immunofluorescence and CD8 staining are detailed in Supplemental Experimental Procedures.
Affinity chromatography onto immobilized GST-Arl6ΔN16 was performed following (Christoforidis and Zerial, 2000) with modifications. Bovine retinas (50 g) were thawed in 150 ml NS500 (25 mM Tris pH 8.0, 500 mM KCl, 5 mM MgCl2, 1 mM DTT) supplemented with 250 mM sucrose and protease inhibitors (1 mM AEBSF, 10 µg/ ml each of Leupeptin, Pepstatin A, Bestatin), homogenized by douncing and centrifuged for 2 h at 184,000 × gave in a Ti70 rotor. The retinal extract was loaded onto 1 ml GSTrap HP columns (GE) previously saturated with GST fusion proteins, columns were washed with 20 ml NS500 containing 50 µM nucleotide, and eluted at 22°C with 4 column volumes of EB (25 mM Tris pH 8.0, 2.5 M NaCl, 10 mM EDTA, 5 mM MgCl2, 1 mM DTT, protease inhibitors). Eluates were run on SDS-PAGE or concentrated by methanol/ chloroform precipitation for nanoscale microcapillary reverse phase liquid chromatography with electrospray ionization tandem mass spectrometry (LC-MS/ MS) as previously described (Haas et al., 2006). For BBSome purification, the eluate of the Arl6GTP column was then dialyzed against four successive buffers of decreasing ionic strength for 45 minutes each (final buffer: 25 mM HEPES pH 7.0, 100 mM NaCl, 5 mM MgCl2, 1 mM DTT) and cleared by ultracentrifugation. The dialyzed Arl6 eluate was then fractionated on a Mono S PC 1.6/ 5 column (GE) equilibrated in buffer H5 (25 mM HEPES pH 7.0, 50 mM NaCl) and developed with a gradient of 50 mM to 500 mM NaCl spanning 9 column volumes.
GST pull-downs are described in supplemental methods.
Liposomes were prepared according to (Matsuoka et al., 1998) with minor modifications detailed in supplemental methods. Liposomes were extruded through polycarbonate filters (100 nm pore size for brain lipids and 400 nm pore size for synthetic lipids) and kept at 4°C. All liposomes made from synthetic lipids contained 23 mol% of cholesterol and 77 mol% phospholipids, and Texas Red DHPE was included to normalize lipid concentrations across all liposome stocks and to normalize the rate of liposome recovery in flotation assays. Liposome pelleting and flotation assays were conducted in HKSM buffer (20 mM Hepes pH 7.0, 150 mM KOAc, 250 mM Sorbitol, 3.5 mM MgCl2) supplemented with nucleotides. Liposome flotation assays were conducted as detailed in supplementary methods. Ultrastructural analysis of protein/ liposome mixtures by thin section electron microscopy was preformed as described (Matsuoka et al., 1998).
We thank Suzanne Pfeffer and Scott Seeley for comments on the manuscript; Randy Schekman, Sean Studer, Susan Hamamoto, Ken Matsuoka and Chris Fromme for advice on liposome preparation and processing for EM; John Perrino and Jon Mulholland for assistance with EM; members of Dick Tsien’s laboratory for help with hippocampal neuron culture; Brian Kobilka and Juan Jose Fung for help with liposome preparation; Val Sheffield for the BBS1 antibody; Melanie Tallent for the SSTR3 knockout; Joachim Seemann for the CD8α cDNA; and Applied Precision for OMX imaging. This work was supported in part by grants to M.V.N. from the American Heart Association (AHA-0930365N), the March of Dimes (5-FY09-112), NIH/ NIGMS (R01GM089933), a Sloan Research Fellowship (BR-5014) and a Klingenstein Fellowship and by an NIH/ NIGMS fellowship to S.R.W (F32GM089218).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.