Intraflagellar transport (IFT) provides a mechanism for the transport of cilium-specific proteins, but the mechanisms for linkage of cargo and IFT proteins have not been identified. Using the sensory outer segments (OS) of photoreceptors, which are derived from sensory cilia, we have identified IFT–cargo complexes containing IFT proteins, kinesin 2 family proteins, two photoreceptor-specific membrane proteins, guanylyl cyclase 1 (GC1, Gucy2e) and rhodopsin (RHO), and the chaperones, mammalian relative of DNAJ, DnajB6 (MRJ), and HSC70 (Hspa8). Analysis of these complexes leads to a model in which MRJ through its binding to IFT88 and GC1 plays a critical role in formation or stabilization of the IFT–cargo complexes. Consistent with the function of MRJ in the activation of HSC70 ATPase activity, Mg-ATP enhances the co-IP of GC1, RHO, and MRJ with IFT proteins. Furthermore, RNAi knockdown of MRJ in IMCD3 cells expressing GC1-green fluorescent protein (GFP) reduces cilium membrane targeting of GC1-GFP without apparent effect on cilium elongation.
The outer segment (OS) of vertebrate rods and cones forms developmentally from the membrane of a sensory cilium (1–3). In common with other motile and sensory cilia (4), membrane proteins that function in the OS are synthesized in the adjacent cell body or inner segment (IS), and must pass through the cilium during both initial development and turnover of the mature OS (reviewed in 3). Since photoreceptors in mammals replace about 10% of their OS daily (5), continuous transport of these proteins occurs at a prodigious rate throughout the life of mature photoreceptors. Viewing the OS as a sensory cilium has led to the idea that intraflagellar transport (IFT), which is required for the assembly and maintenance of sensory cilia, motile cilia, and flagella, is necessary for membrane protein transport into the OS (6–8).
A defining feature of IFT is a pair of multi-protein complexes called the A and B sub-particles that collectively comprise at least 16 different proteins (9–11). IFT particles are thought to serve as cargo adaptors for microtubule-based transport of proteins necessary for ciliogenesis (4). Thus, IFT particles and associated proteins move from the cell body to the tip of the cilium or flagellum, and back again (12). The anterograde movement of IFT particles is powered by heterotrimeric members of the kinesin 2 family (9,13–18), referred to here as the Kif3 complex, while transport in the opposite direction requires the cytoplasmic dynein Dhc1b/2 heavy chain (19–22). Additional anterograde motors also appear to play a role in IFT in C. elegans and in vertebrate photoreceptors (13,23,24). Homologues of IFT proteins are found widely among ciliated eukaryotes (25–28). Furthermore, IFT particles have been identified in mammalian systems (7,29) with properties remarkably similar to those in Chlamydomonas(9,11,30).
An essential role for IFT in photoreceptors is supported by the finding that both IFT proteins (7,29,31) and IFT motors (32–34) are present on the photoreceptor axoneme, and mice with a targeted disruption of the Kif3 complex subunit, KIF3A (35,36), or a hypomorphic mutation in IFT88 (7) exhibit progressive photoreceptor degeneration with mis-localization of the membrane protein, rhodopsin (RHO). Nonetheless, the molecular details of how IFT proteins bind to putative cargo in photoreceptors or other sensory cilia have not been determined.
Here we examine IFT–cargo complexes in photoreceptors that contain photoreceptor-specific membrane guanylyl cyclase 1 (GC1, Gucy2e) and rhodopsin (RHO), two key elements of the OS phototransduction cascade. These complexes also contain MRJ (mammalian relative of DNAJ, DnajB6) and heat shock cognate protein 70 (HSC70). The analysis began with the identification of MRJ, a member of the DnaJ class of chaperones, as an IFT88/polaris-binding protein, and led to the finding that MRJ also binds to a cytosolic domain of GC1. A role for MRJ and its co-chaperone, HSC70, is suggested by the finding that addition of Mg-ATP enhances the co-IP of GC1, RHO, and MRJ with IFT proteins. Finally, we stably transfected IMCD3 cells with photoreceptor GC1-green fluorescent protein (GFP) and found that it concentrates in the sensory cilium. RNAi knockdown of MRJ reduced GC1-GFP in the cilium without effects on cilium elongation. Our data support a model in which MRJ and HSC70 function in the formation or stabilization of IFT–cargo complexes.
MRJ interacts with IFT88
We screened a bovine retinal yeast two-hybrid library using a domain (aa 117–317) of mouse IFT88 containing a cluster of tetratricopetide repeats (TPR) referred to as IFT88(TPR1-3). We used IFT88 as bait because it is a constituent of the IFT sub-particle B (9), contains TPR protein interaction repeats (37), and is required for proper assembly of the sensory OS (7). The screen revealed multiple colonies containing plasmids encoding 198 residues of the C-terminal domain of MRJ (Figure 1A), a member of the DnaJ co-chaperone family. The interaction between IFT88 and MRJ was verified in yeast cells by retransformation. Cells transformed by GAL4-AD-MRJ alone did not survive on high stringency plates, but when co-transformed with GAL4-BD-IFT88TPR1-3 colonies formed and turned blue after 1 day of growth (Figure 1A). This was similar to the positive control in which cells were transformed with plasmids encoding two known interacting proteins (SV40 Large T-antigen and p53). In α-galactosidase quantitative assays we evaluated the interaction between MRJ and the original bait [IFT88(TPR1-3)] as well as a distinct C-terminal sequence (aa 392–727) containing additional TPR repeats [IFT88(TPR4-10)], and a sequence (aa 117–727) containing both N- and C-terminal arrays of TPR repeats with an intervening linker region (Figure 1B). IFT88(TPR1-3) and IFT88(1–10) were similar to the positive control (SV40 Large T-antigen and p53), while the interaction between MRJ and IFT88(TPR4-10) was substantially more robust (Figure 1B). Moreover, MRJ specifically interacts with IFT88, but not IFT57, IFT52 or IFT20, which lack TPR repeats (Figure 1C). These data indicate that in yeast MRJ binds to the TPR repeats of IFT88.
As TPR repeats are present in many proteins, we determined whether MRJ interacts with another TPR-containing protein, conventional kinesin light chain (KLC1) that is not part of the IFT pathway. We cloned a domain (aa 176–414) containing 5 TPR repeats of KLC1 (Figure 1D) and detected a weak interaction with MRJ in quantitative assays (Figure 1D). Unlike the interaction with IFT88(TPR1-3) and IFT88(TPR1-10), however, this interaction was below that of the positive control. This suggests that while MRJ interacts strongly with IFT88, it may also associate with other proteins that contain TPR repeats. This is not surprising because MRJ is likely to have multiple functions as a co-chaperone in the heat shock protein 70 (HSP70)/HSC70 system (38).
GST-MRJ Binds both Recombinant and Endogenous IFT88
We next independently confirmed the interaction between MRJ and IFT88 using glutathione S-transferase (GST) pull-down assays (Figure 1E–F). First, we separately expressed and purified GST-MRJ and either His- (data not shown) or GST-tagged IFT88, and performed cell-free binding assays using different concentrations of IFT88; the GST-tag on IFT88 was cleaved during purification. GST-MRJ directly interacts with IFT88 in a concentration dependent and saturable manner, whereas beads alone do not bind IFT88 (Figure 1E). Densitometric analysis of bands in these assays relative to known concentrations of recombinant IFT88 permitted us to estimate an equilibrium dissociation constant for IFT88–MRJ binding of 0.8 μm. Second, GST-MRJ was used in pull-down assays with a tissue extract from retina (Figure 1F) where it specifically pulled down native IFT88. These pull-down assays also co-precipitated IFT57 and the Kif3 complex subunit, KIF3A (Figure 1F). We previously demonstrated that IFT88 exists predominantly in a complex with other IFT proteins in retinal extracts (29). Thus, the GST-MRJ pull-down assays suggest that MRJ can interact with native IFT88 in the context of an IFT particle.
MRJ and HSC70 Co-IP with IFT Proteins and KIF3A
Dnaj proteins such as MRJ are known to bind HSP/HSC70 family members and to activate their ATPase activity. HSP70 and HSC70 differ principally in whether their expression is constitutive (HSC70), or is induced by stress stimuli such as heat shock (HSP70). Their binding to Dnaj proteins suggests that both MRJ and HSP70/HSC70 could be associated with IFT protein complexes. Consistent with this, HSC70 was co-precipitated along with IFT88, IFT57 and KIF3A in GST-MRJ pull-down assays (Figure 1F). Throughout this paper we refer to the HSP/HSC70 family member associated with MRJ as HSC70 (Hspa8) because it binds a HSC70 specific antibody. However, the antibody used in the early stages of this work recognized both HSP70 and HSC70, and it is likely that MRJ can interact via its J-domain with multiple members of the gene family.
The pull down of HSC70 in Figure 1F could reflect the known direct binding of MRJ to HSC70 (38) rather than an association of HSC70 with IFT complexes. We therefore carried out IP experiments using retinal extracts and found that anti-IFT88 and anti-MRJ antibodies reciprocally co-immunopreciptate the other protein (Figure 2A). Furthermore, multiple antibodies against IFT proteins, MRJ and HSC70 co-precipitate MRJ and IFT88 (Figure 2A). In addition, antibodies to the IFT subunits and both chaperones (MRJ and HSC70) co-precipitate IFT88, KIF3A, HSC70 and IFT57 (Figure 2B). These data indicate the existence of native complexes containing both chaperones along with IFT and motor proteins. The co-IP of IFT88 with MRJ and HSC70 antibodies and the similarity of IPs using HSC70 and IFT antibodies for western blotting suggest that HSC70 is associated with most IFT complexes.
In our experiments antibodies to IFT88 often recognized a doublet of proteins at approximately 90 kDa MW (Figure 2B), but in many experiments the two bands were not resolved (see Figure 2A); resolution of only a single band was particularly evident when low percentage gels were used. The fact that either N- or C-terminal peptide antibodies recognize both bands (Figure 2B, lane 1) suggests that they are variants of IFT88 that could be generated through proteolysis, post-translational modification, or alternative splicing. Data base analysis suggests the existence of splice variants of IFT88 (Ensembl Gene; http://www.ensembl.org/Homo_sapiens/index.html) that would differ by as much as 1 kDa in size while retaining both N- and C-terminal epitopes. Interestingly, these two bands are differentially immunoprecipitated by MRJ and IFT antibodies. Antibodies to IFT52, IFT20 and MRJ IP both (Figure 2B, lanes 3, 4, and 7) while those to IFT57 and HSC70 (Figure 2B, lanes 2 and 5) IP predominantly the lower band. Such data suggest variability in the composition of two forms of IFT88 in IFT complexes.
One goal of this study was to determine if photoreceptor OS proteins are associated with IFT proteins in IFT–cargo complexes. Because IFT protein complexes could be pulled down with GST-MRJ (Figure 1F), we looked for photoreceptor-specific OS proteins in silver stained gels from GST-MRJ pull-down experiments similar to that in Figure 1F using isolated bovine OS. A prominent ≅ 115 kDa protein band detected in silver-stained gels (Figure 3A) was excised, digested with trypsin and subjected to MALDI-TOF analysis. Twenty-seven peptides matched bovine photoreceptor guanylyl cyclase 1 (Gucy2e), henceforth referred to as GC1. The top matches also included 10–14 peptides from human, mouse (4 sequences) and dog GC1. This identification was confirmed by western blotting with an anti-GC1 antibody (Figure 3B). In this experiment GC1 was pulled down, but RHO, a much more abundant OS protein, was not detected.
The identification of GC1 and IFT proteins in bMRJ pull-down assays (see Figure 3B and C) suggests the existence of an IFT–cargo complex containing GC1. However, it was also possible that MRJ directly interacts with both IFT88 and GC1 and that the two are pulled down independently. To test for a direct interaction between MRJ and GC1, we cloned a cytosolic segment (aa 494–844) of mouse GC1, and separately expressed, and purified GST- and His-tagged GC1498−844 proteins. AA 494–844 of GC1 was used because it contains the highly conserved kinase homology domain (KHD) whose function is poorly understood; this region of the human gene contains multiple mutations that cause photoreceptor degeneration (39–42). Pull-down assays show that GST-bMRJ specifically pulls down His-GC1 (Figure 3C), and in the reciprocal experiment GST-GC1 pulls down His-mMRJ (Figure 3D). Using 1.0 μm GST-bMRJ and varying concentrations of GC1494−844 (Figure 3E) we found that the MRJ/GC1 interaction was concentration dependent and saturated at 0.25–0.5 μm GST-bMRJ. Densitometric analysis of the bands relative to lanes with known concentrations of GC1494−844 showed significant binding at 0.03 μm (Figure 3E) compared with 0.25 μm in similar assays for IFT88 binding (Figure 1E).
To answer the separate question of whether GC1 is found in IFT protein complexes, IP assays similar to those in Figure 2 were blotted for GC1. GC1, IFT88 and HSC70 were all co-precipitated using multiple antibodies to IFT proteins, MRJ and HSC70 (Figure 3F). The principal variation was that the anti-GC1 antibody co-precipitated a large amount of the GC1 and a lesser amount of IFT88 compared with the IFT and MRJ antibodies, indicating that only a small fraction of total GC1 is associated with IFT–cargo complexes. HSC70 and IFT88 were also present in each IP; the larger amount of HSC70 in the IP with the MRJ antibody was expected because of its known direct binding to MRJ. However, comparison of the relative blot intensities for GC1, IFT88 and HSC70 in the four IFT IPs (Figure 3F, lanes 6–9) suggests heterogeneity in the relative composition of cargo complexes containing GC1. Finally, in contrast to the GST-MRJ pull-down experiments (Figure 3B), a small amount of RHO was found in each IP (data not shown); RHO was therefore evaluated in subsequent experiments (see Figure 4). Our data indicate that GC1 (and RHO) are associated with IFT protein complexes in retinal extracts.
Mg-ATP enhances the co-IP of GC1 and RHO with IFT88
The association of both MRJ and HSC70 with IFT–cargo complexes (Figures 2B, 3E) and the fact that MRJ activates HSC70 ATPase activity raises the possibility that ATP would modulate the binding of cargo protein or chaperones. Therefore, we examined the effect of ATP and its non-hydrolyzable analogues, AMP-PNP and ATP-γS, on the association of GC1, RHO, HSC70 and MRJ with IFT88 by IP (Figure 4). First, anti-IFT88C in the presence of 1 mm ATP and 5-mm MgCl2 co-precipitates more GC1 from retinal extracts compared with control (Figure 4, compare lanes 3 and 4). In contrast, IPs carried out in the presence of Mg-AMP-PNP, a weakly hydrolysable ATP analogue (43), ATP-γS, a non-hydrolyzable ATP analogue, or ATP in the absence of Mg2+ had no effect (Figure 4, lanes 5, 6 and 8). Furthermore, the effect of Mg-ATP was blocked by addition of either ethylenediaminetetraacetic acid (EDTA) (lane 7) or NaN3 (lane 9). The later effects are consistent with the known requirement of Mg2+ in the hydrolysis of ATP by HSC70 (43), and the common use of NaN3 as an inhibitor of ATP hydrolysis (44). Interestingly, a similar Mg-ATP dependence of RHO's association with IFT88 was seen, although its interaction was not completely blocked in the presence of NaN3 (Figure 4, lane 9). Furthermore, this effect was specific for monomeric RHO. RHO typically forms oligomers in SDS-PAGE (see multiple bands in Figure 4, lane 10), but only the monomeric form co-precipitated. Finally, the pattern of MRJ binding paralleled that of GC1 binding in that it was easily detectible only in the presence of Mg-ATP (Figure 4, lane 4). These results indicate that the presence of Mg-ATP enhances the co-IP of MRJ, GC1 and RHO with IFT88. Since RHO was not identified in GST pull-down assays using MRJ (Figure 3B), it is likely that RHO binds to IFT–cargo complexes independently of MRJ and GC1.
The effect of Mg-ATP on the GC1 and RHO content in IPs using anti-IFT88C was concentration dependent (Figure 5A), and readily reproducible under a variety of experimental conditions. For example, in experiments evaluating effects of MgCl2 and NaCl concentration, 4-mm ATP had similar effects at 1 or 4 mm MgCl2 and at 150- or 250-mm NaCl at either MgCl2 concentration (Figure 5B). The only salt effect detected was an apparent enhancement of the ATP-dependent co-IP of RHO at 250 mm NaCl, and a reversal of this effect at 4 mm MgCl2. Interestingly, in this experiment a doublet of IFT88 was precipitated (see also Figure 2) in the 0 ATP condition under all salt conditions, while in the presence of ATP, the upper IFT88 band was predominant. A similar effect was seen in the ATP dose-dependent data (Figure 5A), suggesting that the IFT–cargo complexes recovered using the IFT88C antibody contain mainly the larger form of IFT88.
Unlike MRJ, the co-IP of HSC70 with IFT88C was abundant in all treatments with the possible exception of Mg-ATPγS (Figure 4, lane 6). Although addition of ATP analogues appeared to enhance the co-IP of HSC70, conditions expected to inhibit ATP hydrolysis also revealed abundant HSC70 (Figure 4, lanes 7–9). Furthermore, Mg-ATP concentration had little effect on the co-IP of HSC70 with IFT88 (Figure 5A). HSC70 detected with two different antibodies was relatively abundant in IPs at all ATP concentrations (Figure 5A). These results suggest that a significant amount of HSC70 is constitutively associated with IFT proteins in our experiments.
MRJ Localization in Photoreceptor OS and Kidney Epithelial Cilia
We next evaluated MRJ localization in retina and in kidney epithelial cells. Double label immunocytochemistry of bovine retina (Figure 6A–F) using a rabbit antibody to MRJ (green) along with a monoclonal antibody to K26 (red), a connecting cilium-specific glycoprotein (45), revealed that while MRJ is present in all retinal layers. It was most highly concentrated within the photoreceptor IS immediately proximal to the connecting cilium. Within the OS, MRJ was much less abundant and was diffusely distributed (Figure 6A and D). To evaluate the MRJ distribution within the OS at higher resolution, we used isolated mouse OS counter stained with a monoclonal antibody to acetylated α-tubulin (Figure 6G–L). In these preparations the axoneme (tubulin, red) was intensely stained proximally, but this staining was greatly attenuated distally. MRJ did not specifically co-localize with the axoneme, suggesting that it is more closely associated with the membrane or cytoplasmic compartments within the OS. Both HSC70 and HSP70 are known to be constitutively expressed in rat photoreceptors and to localize to both IS and OS (46). Consistent with this, HSC70 was also present in the IS and basal body region of mouse photoreceptors, and, similar to MRJ, was diffusely distributed in the OS (Figure 6M–Q).
MRJ is also associated with sensory cilia in kidney epithelial cells. IC on mouse IMCD3 (not shown) and porcine LLC-PK1 (Figure 7A) cells readily detected cilia with the anti-acetylated α-tubulin antibody (green) and both IFT88 and IFT57 (red) showed the expected punctate pattern along the length of the cilium. MRJ (red) was also found in cilia in a punctate pattern similar to that seen for IFT proteins (Figure 7A). Unlike IFT proteins, however, MRJ was present in multiple cytoplasmic compartments, including junctional regions between adjacent cells.
To further study MRJ in epithelial cells, a mouse His-MRJ expression vector was constructed, and stably transfected IMCD3 lines were created using antibiotic selection. As shown by western blot analysis (Figure 7B), a ∼ 32 kDa protein recognized by an anti-His antibody was expressed at moderate levels compared with that seen in transient transfections. IC of His-mMRJ using the anti-His antibody in the same cells showed staining along the cilia (Figure 7C). As was true for endogenous MRJ (Figure 7A), His-mMRJ was also present within cytoplasmic compartments. Overall our results indicate that, while MRJ is present in cilia of both photoreceptors and kidney epithelial cells, it is most abundant in cytoplasmic compartments.
Knockdown of MRJ reduces GC1-GFP in the cilium
In order to establish a model system for evaluation of membrane protein trafficking into the sensory cilium, we created stably transfected IMCD3 cell lines expressing GC1 with GFP fused in frame at its C-terminus (GC1-GFP). Stable lines expressing moderate to low levels of GC1-GFP as judged by western blot analysis (Figure 8A) were most useful for evaluating GC1 localization in sensory cilia. In such cultures virtually all cells make sensory cilia stainable with anti-acetylated α-tubulin and these cilia all contain GC1-GFP (Figure 8B). In many cases the GC1-GFP in the cilium was organized in descrete puncta; it also concentrated at the base of the cilium (Figure 8C). Analysis of cultures stained with anti-GFP antibodies reveal a similar pattern of staining, and co-localization studies using Hoechst to stain nuclei show that virtually all cells in the culture make cilia-containing GC1-GFP (data not shown).
We used IMCD3 cells expressing GC1-GFP in a test of the idea that MRJ is required for GC1-GFP trafficking into the sensory cilium (Figure 9). In these experiments the GC1-GFP plasmid was co-transfected with control or an MRJ-specific shRNA plasmid prior to selection. We used the Super Array SureSilencingTMplasmid-based RNA interference system and quantitative polymerase chain reaction (PCR) to measure the level of knockdown. One shRNA plasmid (M1) reduced MRJ mRNA by 70% compared with the control vector encoding a scrambled RNA sequence not represented in the mouse genome (data not shown). Western blot analysis 5 days after plating shows that this shRNA reduces MRJ protein to virtually undetectable levels compared with the control vector (Figure 9A). In the control experiment GC1-GFP is found in cilia in virtually all cells (Figure 9B, D, left panels). In contrast, the shRNA vector directed at MRJ dramatically reduces GC1-GFP in cilia without affecting cilium length (Figure 9C, D, right panels). These experiments indicate that MRJ is necessary for GC1-GFP trafficking into the sensory cilium.
IFT–cargo complexes containing GC1, RHO and the MRJ/HSC70 chaperone pair
Our central finding is the identification of protein complexes containing IFT proteins, the KIF3A motor subunit, the MRJ/HSC70 chaperone pair and the photoreceptor membrane proteins, GC1 and RHO. Our previous work on photoreceptors demonstrating a 17S IFT particle and the co-IP of all three subunits of the Kif3 complex with IFT proteins (29) implies that we are dealing with IFT particles bound to the Kif3 complex. We propose that these protein assemblages are involved in the transport of GC1 and RHO into the OS, and accordingly call them ‘IFT–cargo complexes’. The key events involve the interactions of MRJ with GC1, IFT88 and HSC70 (Figure 10). In particular, MRJ's known function as an activator of HSC70 ATPase activity and the effects of Mg-ATP on the co-IP of both GC1 and RHO suggest that the HSC70 ATPase cycle plays a role in the formation or stabilization of IFT–cargo complexes. Both HSC70 and HSP70 (46) as well as MRJ (Figure 6) are constitutively expressed in photoreceptors, and may be particularly important factors supporting the extraordinarily high level of membrane protein trafficking into the OS. However, MRJ's expression in multiple tissue types (38,47) and its role in GC1-GFP trafficking in IMCD3 cells (see Figure 9) raise the possibility that it may play a similar role in multiple types of cilia.
In photoreceptors, MRJ and HSC70 could associate with IFT–cargo complexes within the IS, the connecting cilium and/or the OS. Our IP data revealing IFT–cargo complexes utilize cell extracts and does not specify the relevant compartment, but a growing body of evidence suggests that IFT complexes are initially formed in the vicinity of the basal body. This is consistent with the finding that MRJ is considerably more abundant within the cell bodies of both kidney epithelial cells and photoreceptors (Figure 6 and 9). Furthermore, both HSC70 and HSP70 are constitutively expressed at significant levels in photoreceptors with more in the IS (46).
The idea of chaperones associated with cilia and IFT is not entirely new. The Chlamydomonas radial spoke protein RSP16 has been identified as a type II DnaJ protein similar to MRJ (48), and it has been suggested that RSP16 protein serves as a co-chaperone with HSP70 in the final assembly of the radial spoke in the distal cilium. In addition, HSP70 has been identified within Chlamydomonas flagella (49,50) and Tetrahymena cilia (51), and in a discussion of an early study of an IFT protein complex (14) HSP70 was reported as a co-fractionating protein. Although this finding did not hold up in subsequent high stringency purification of IFT particles (9), HSP70 was identified in subsequent proteomic studies (27,28,52). Furthermore, recent expression studies reveal that HSP70a is distributed along the flagellum in a punctate pattern similar to that of the Chlamydomonas fla10 kinesin, suggesting that HSP70 is specifically associated with IFT in Chlamydomonas(50). Finally, both HSP70 and MRJ were identified in a mammalian motile cilium proteome (52) and in a mouse photoreceptor OS proteome (53).
Although photoreceptor HSC70, MRJ and putative cargo co-IP with IFT proteins, it is likely that binding of both chaperones and cargo is labile and transient. In previous work in Chlamydomonas IFT sub-complexes A and B have been characterized using sucrose density sedimentation and gel filtration chromatography (9–11), but cargo proteins and IFT motors did not co-fractionate using those techniques (9,30). Nonetheless, both IFT motors and components of the axoneme were found to be associated with IFT proteins by co-IP (30). It is likely that cargo proteins are identified by IP because the conditions are less stringent, permitting capture of complexes in which binding is more labile. Our results using mammalian retina is similar to that in Chlamydomonas in that we have not detected co-fractionating GC1, RHO or IFT motors in sucrose density gradients (unpublished data) that readily identify a photoreceptor 17S IFT B sub-particle (29). Although our IP protocols are of sufficiently low stringency to permit capture of IFT–cargo complexes, the enhancement of co-precipitating GC1, RHO and MRJ relative to IFT88 by Mg-ATP argues for a specific pathway regulating their binding.
MRJ is likely to play a critical role in ‘IFT–cargo complexes’ because it binds directly to GC1, HSC70 and IFT88. MRJ binds HSC70 via its J-domain (54), but the activation of HSC70 ATPase activity by N-terminally truncated MRJ suggests that broader interaction is possible (38). Our data indicate that the N-terminal 44 aa are not required for IFT88 or GC1 binding, but at present we do not know whether distinct binding domains would permit MRJ to bind IFT88 and GC1 simultaneously to form a ternary complex. Resolution of this issue requires additional fine mapping data.
While MRJ association with the IFT particle appears to be regulated by Mg-ATP in a manner similar to that of GC1 and RHO, HSC70 is present in IFT complexes under all IP conditions used in these studies. The only conditions obviously affecting HSC70 levels were addition of EDTA, NaN3 or ATP without Mg2+. Those treatments resulted in increased HSC70 without increased MRJ or cargo proteins. This favours a model in which HSC70 is constitutively present on IFT particles (Figure 10A), as opposed to one in which HSC70 is recruited to the IFT particle by MRJ (Figure 10B). In this regard, it is of interest that the HSC70 C-terminus is known to bind to proteins with TPR repeats (55,56). In addition to IFT88 several additional IFT proteins including IFT172 within IFT sub-particle B contain TPR repeats (10,26,57). Thus, it is possible that HSC70 can bind directly to IFT88 or to one or more additional TPR-containing proteins within the IFT particle. IFT88 along with HSC70 bound to the particle could in turn recruit MRJ-bound cargo proteins such as GC1.
A Role for the HSC70 ATPase Cycle in IFT Cargo Loading?
HSC70 and other HSP70 family members along with their Dnaj co-chaperones function in numerous biochemical and cellular pathways (58,59). These pathways generally begin with substrate binding in which specificity is conferred by members of the Dnaj family of proteins, which bind to both substrate and HSC70. For example, mitochondrial HSP70 along with a Dnaj protein and a nucleotide exchange factor is critical for both translocation of proteins across the inner mitochondrial membrane and for protein folding in the mitochondrial matrix (60,61). In another well-studied example, auxillin, a DnaJ protein, binds to clathrin-coated vesicles where it recruits and activates HSC70 to un-coat the vesicle (62) while a distinct DnaJ protein recruits HSC70 during coated vesicle endocytosis (63). Finally, it has been reported that HSP70 ATPase activity dissociates conventional kinesin membrane vesicle cargo (64). In each case the function of the chaperone pair can be best understood in relationship to an ATPase cycle in which HSC70 oscillates between an ATP-bound, ‘open’ conformation and an ADP-bound, ‘closed’ conformation (58,59). DnaJ proteins interact with HSC70 in its ‘open’ conformation and activate its inherently low ATPase activity. The ‘closed’ conformation is often associated with a higher affinity for substrate, and is reversed on exchange of ADP for ATP, which can be stimulated by nucleotide exchange factors. The enhanced binding of GC1 and RHO to IFT complexes in the presence of Mg-ATP suggests a role for the ATPase cycle of HSC70 in the formation of IFT–cargo complexes. Although the plausibility of this model is supported by the known enhancement of HSC70 ATPase activity by MRJ (38), further analysis of the role of Mg-ATP in stabilizing IFT cargo complexes is necessary because other pathways such as protein phosphorylation could also modulate critical steps of IFT–cargo complex assembly.
Ciliary Membrane Proteins as IFT Cargo
Our data extend prior genetic and cell biological studies implicating IFT in membrane protein trafficking in photoreceptors and other systems. For example, mis-localization of RHO in mice with a mutation in IFT88/polaris (7) or a deficiency in the KIF3A subunit of the Kif3 complex (35,36) have provided genetic evidence that RHO is transported by IFT. We have additional data showing that GC1 is also mis-localized in Tg737orpk mice (unpublished) in a manner similar to RHO. Although such mis-localization could be an indirect consequence of the mutations, our data provide a direct biochemical linkage between GC1 and RHO and the IFT pathway. Linkages to cilia and a role for IFT have also emerged for other membrane receptors including polycystins 1 and 2 (65–68), olfactory cyclic nucleotide-gated channels (69), and the smoothened (Smo) receptor for hedgehog signalling (70). Finally, it should be mentioned that the movement of TRPV channels in parallel with IFT particles has been directly imaged in living C. elegans sensory cilia (71).
Trafficking and Photoreceptor Degeneration
Growing evidence indicates that some hereditary photoreceptor diseases result from defective intracellular protein transport. For example, the most common cause of retinitis pigmentosa (RP), a common hereditary retinal dystrophy, is mutation of the RHO gene (72), and several of its mutations are thought to involve trafficking (73). Similarly, mutations in GC1 are known to cause a range of different inherited diseases including Lebers Congenital Amaurosis and cone-rod dystrophies in humans (42,74,75), and deletion of both photoreceptor membrane cyclases (GC1 and GC2) in mice results in abnormal trafficking of several OS proteins (76). The most clearly understood of the human GC1 mutations are in dimerization and catalytic domains that are thought to result in loss of enzyme activity, and exact examples of mutations affecting trafficking have not been identified (74). Our data show that MRJ interacts with a GC1 peptide (aa 494–844) that includes a KHD of unknown function and a portion of the dimerization domain. The KHD of GC1 contains disease-causing mutations of unknown molecular consequence (42,75). Our data linking the IFT pathway directly to GC1 via binding of MRJ to GC1494−844 suggest that mutations in the KHD lead to defective trafficking and human disease.
Materials and Methods
Antibodies used for immunocytochemistry for IFT and MRJ proteins were gifts of Gregory Pazour (7) and Chig-Hwa Sung (38), respectively. For double labelling mouse monoclonal antibodies directed at acetylated-α-tubulin (Sigma) or the ciliary K26 antigen (77) were used. In western blotting and immunoprecipitation assays, goat antibodies IFT88N, IFT88C and IFT57N as well as KIF3A (Covance) and KIF3B (BD Biosciences) are as described previously (29). New affinity-purified antibodies directed at synthetic peptides from the C-termini of IFT57 (peptide CSEARERYQQGNGGVT, called IFT57C, and IFT52 (peptide CLQEGDENPRDFTTLF, called IFT52C) were produced at Bethyl Laboratories using goat as the host species. The affinity-purified IFT20 antibody used in this work was produced in guinea pig (Covance) using full-length recombinant IFT20 as the antigen. Alexander M. Dizhoor (Pennsylvania College of Optometry) and Robert Molday (University of British Columbia) provided the rabbit and monoclonal (6H2) antibodies to GC1, respectively. The anti-HSP70 antibody (SPA-820, Stressgen Bioreagents) was a mouse monoclonal antibody, which detects both HSP70 and HSC70. The HSC70-specific antibody (ab19136, Abcam) was a rat monoclonal antibody. The mouse monoclonal anti-polyhistidine antibody was from Sigma.
Yeast Two Hybrid Library Screen
The Matchmaker Gal4 Two Hybrid System 3 (Clontech) was used following the manufacturers instructions (Clonetch, PT3024-1). The bait plasmid pGBKT7-IFT88(TPR1-3) included the coding sequencing of the N-terminal TPR array of mouse IFT88 (aa 177–317 of Genbank acc # AAB59705) in frame with the GAL4 DNA-binding domain; additional IFT88 constructs used in the analysis encoded aa 392–727 [IFT88(TPR4-10)] and aa 176–727 [IFT88(TPR1-10)] of mouse IFT88. The bovine retinal cDNA library, cloned into the pACT2 vector, was provided by Dr Ching-Hwa Sung (38). Plasmids were introduced into AH109 yeast cells by Lithium acetate (LiAc)-mediated co-transformation and plated on high stringency (SD-adenine-histidine-leucine-tryptophan) plates. After 3 weeks at 30°C 100 colonies (≥2 mm in diameter) were selected, inoculated into 0.5 mL of SD-adenine-histidine-leucine-tryptophan medium, and grown to saturation at 30°C with shaking.
For plasmid DNA sequencing and restriction enzyme analysis, E. coli (DH5α) were transformed with plasmids from each yeast colony, and plated on LB-ampicillin plates for selection. Plasmids were harvested by standard alkaline lysis using a spin plasmid miniprep kit (Qiagen), double digested with XhoI and EcoRI, and examined on 1% agarose gels to verify the presence of cDNA inserts; inserts were sequenced using a dye terminator cycle sequencing kit and the ABI Prism 310 Genetic Analyzer (Applied Biosystems). After the initial step of the screen 100 colonies were selected, 9 of which were MRJ; 11 gene products remained after growth on high stringency plates. After exclusion of clones that self-activated in the absence of bait, only MRJ (Dnajb6) and one other gene product (steroid sensitive gene, Ccdc1) were identified.
Yeast Two-Hybrid Re-transformation and α-Gal Quantitative Assays
To verify that the proteins encoded by the cDNA inserts interact with the IFT88 (TPR1-3), plasmids were co-transformed with pGBKT7-IFT88-TPR1-3 plasmid or with the empty vector, pGBKT7, into yeast AH109 cells using a small-scale transformation method (Clontech, PT3024-1). Cells grown on low stringency (SD-leucine-tryptophan) plates for 3 days at 30°C were transferred to SD-adenine-histidine-leucine-tryptophan plates supplemented with 20 μg/mL X-α-Gal (5-Bromo-4-chloro-3-indolyl-α-d-galactopyranoside) for high stringency selection. The formation of yeast colonies and expression of galactosidase activity were examined after growth for 3 days.
Quantitative α-galactosidase (α-Gal) assays were also conducted to determine the relative strength of the interaction. The catalytic activity of α-galactosidase was monitored colorimetrically by measuring the rate of hydrolysis of p-nitrophenyl-α-d-galactoside (PNP-α-Gal) to p-nitrophenol, which displays a strong absorption band at 410 nm (Yeast Protocols Handbook, Clontech). Every retransformation sample was assayed in triplicate in one experiment, and three independent experiments were conducted.
GST Pull-Down Assays
The GST-bMRJ construct, generously provided by Dr Ching-Hwa Sung (Cornell University) encoded the C-terminal 198 residues of bovine MRJ 3′ to the GST open reading frame in the pGEX-5X-2 vector (38). It was expressed in E. coli BL21 cells, and large-scale production of GST-fusion and cleaved proteins followed the protocol of Guan and Dixon (78). The retinal extract for pull-down assays was prepared by the method of Aslanukov and colleagues (79). Briefly 20 (∼ 10 g) fresh-frozen bovine retinas (Emmpak Foods) were ground into a fine powder on dry ice followed by homogenization (30–40 strokes) in a glass homogenizer with 30 mL of cold homogenization buffer (1% Nonidet P-40, 20 mm Tris–HCl, pH 6.8, 250-mm NaCl, 2 mm 2-mercaptoethanol, 0.02% NaN3, 5% (v/v) glycerol and a cocktail of protease inhibitors). Homogenates were centrifuged at 10 000 g at 4°C for 20 min and precleared with 5.0 mL of swollen glutathione sepharose beads and 500 μg of recombinant GST for about 30 min at 4°C.
For pull-down assays about 2 μm of the GST-fused protein was incubated with retinal extract containing 1 mm Pefabloc SC (Boehringer Mannheim) for 1 h in a nutator mixer followed by an additional incubation for 1 h with 60 μL of 50% glutathione Sepharose 4B beads (Amersham Pharmacia Biotech) in 50-mm Tris–HCl (pH 7.5), 100-mm NaCl, 2 mm MgCl2, and 1% NP-40. Beads were centrifugation (2–3 seconds at 10 000 g) and washed four times with 0.5 mL of cold 50-mm Tris–HCl (pH 7.5), 100-mm NaCl, 2-mm MgCl2, 0.2% Triton X-100.
For recombinant in vitro assays 1.0 μm of GST bMRJ was incubated with different concentrations of affinity-purified, recombinant IFT88 ranging from 0.25 to 2.0 μm for 45 min in 1% Nonidet P-40 homogenization buffer at 4°C. Then 60 μL of glutathione sepharose beads in incubation buffer was added, further incubated for 60 min at 4°C and washed four times. An additional set of in vitro assays was carried out using affinity-purified recombinant His- or GST-tagged mouse GC1 (Gucy2e) and GST- or His-tagged MRJ. The His-tagged proteins were captured using Ni-NTA agarose beads (Invitrogen). Quantitative analysis involved use of GC1 ranging from 0.03 to 0.5 μm in combination with 0.50 μm GST-bMRJ. Finally, the retinal co-precipitates were resuspended in 1X SDS sample buffer (62.5 mm Tris–HCl, (pH 6.8) 2% SDS, 5% 2-mercaptoethanol, 10% Glycerol) and boiled for 3–5 min, and resolved by SDS-PAGE. Proteins were analysed by western blotting. In the semi-quantitative analyses bMRJ binding to IFT88 (Figure 2A) and GC1494−844 (Figure 3E) bands from digital images were scanned using LabWorks (UVP) image analysis software, and protein concentration of pull-downs was estimated from scans of lanes on the same blot loaded with known concentrations of recombinant IFT88 or GC1494−844.
Isolation and Identification of GC1 by MS
GC1 was isolated in a pull-down assay using recombinant GST-bMRJ incubated with glutathione beads at 4°C for 2 h, followed by incubation for an additional 2 h with an extract prepared from a purified bovine rod OS fraction (80) rather than whole retinal extract. GST beads were collected by centrifugation, washed, boiled with SDS sample buffer and examined on SDS-PAGE with silver staining (81). At the Medical College of Wisconsin Protein and Nucleic Acid Core Facility, the sample band and neighbouring blank gel pieces were digested by trypsin and then subjected to Zip-Tip cleanup. Finally, the peptides were analysed with the Voyager-DE PRO MALDI-TOF mass spectrometer (Applied Biosystems). The Protein Prospector program (http://prospector.ucsf.edu) was used to search the database for peptide identification. The top match was bovine GC1 (score 1.89e14) with 27 of 68 peptides and 39% coverage. The top seven matches included human, mouse (4 sequences), dog and rat GC1 with 10–14 peptides each. Three proteins of unknown function with scores and coverage similar to the non-bovine GC1's were identified. They were Tacc2 (Q9JJGO), Ccdc46 (Q5PR68) and Zc3h3 (Q8IXZ2). For these proteins and the non-bovine GC1 proteins the score dropped dramatically to the range of 2-4e4.
Immunoprecipitation (IP) assays were carried as described by Aslanukov and colleagues (79). Briefly, for each IP reaction about 4 μg of antibody was used per 20 μL of Protein G Sepharose beads in PBS, pH 7.4; antibodies were incubated for 45 min at room temperature. Approximately, 200 μL of NP-40 (1% Nonidet P-40, 250-mm NaCl, 20-mm Tris–HCl, pH 6.8) solubilized retinal extract was incubated with the antibody-bound beads for 60 min at 4°C. The beads with retinal extract were then loaded onto spin filters (Millipore) and washed three times with 0.5 mL of the NP-40 buffer followed by the elution of the bound protein complexes with 60 μL of SDS-sample buffer. Eluants were boiled and resolved on SDS-PAGE gels, and analysed by western blotting. IPs in the presence of Mg-ATP and its analogues or inhibitors of ATPase activity were incubated for 1 h at 4°C with the additions prior to the addition of antibody attached to protein G sepharose beads.
Pig kidney tubular epithelial cells (LLC-PK1) or mouse IMCD3 cells were grown to confluency on coverslips in medium 199 (Gibco 12340-030) supplemented with 3% foetal bovine serum (Gibco 16000-036) and 1 × penicillin -streptomycin-glutamine (Gibco 10378-016) at 37°C with 5% CO2 in air atmosphere. They were fixed for 10 min in a mixture of methanol and acetone at −20°C, rinsed with PBS, and blocked and permeabilized for 1 h at 30°C in 3% foetal bovine serum and 0.5% Triton X-100 in medium 199. For double labelling the coverslips were immersed in 60 μL of blocking buffer containing mouse monoclonal anti-acetylated α-tubulin antibody (1:100, Sigma), rabbit anti-IFT88 antibody (1:200), rabbit anti-IFT57 (1:200) or rabbit anti-MRJ antibody (1:20) for 1 h at 30°C in a humid chamber. After rinsing with PBS, the coverslips were incubated for 1 h at 30°C with a 1:500 dilution of the corresponding secondary antibodies: Alexa 488 conjugated goat anti-mouse IgG or Alexa 594 conjugated goat anti-rabbit IgG (Molecular Probes, Inc.). Hoechst dye (2 μg/mL) was used to label the nuclei. Slides were coversliped using Fluoromount G (EMS) and observed using a Nikon epifluorescence inverted microscope.
Fresh bovine retinal sections were incubated with rabbit anti-MRJ and K26 antibodies, which were detected with goat anti-rabbit or goat anti-mouse IgG conjugated with Alexa 488 or Alexa 594 (Molecular Probes, Inc.). The monoclonal antibody K26, which detects a unique epitope on the connecting cilium of bovine photoreceptor cells, was used as a marker for connecting cilium (45). Mouse photoreceptors were isolated by shaking freshly dissected retinas in PBS on microscope slides. Cells were fixed with 4% paraformadehyde and stained as above.
Construction of Mouse MRJ, KLC1 and GC1 expression vectors
To obtain full-length mouse MRJ sequence RNA from C57Bl/6 mouse retina was isolated using Trizol according to the manufacturer's instruction (Life Technologies) and reverse transcribed using a reverse transcription system (Promega). After separating the PCR product on a 1% agarose gel, the band with the expected length (∼ 750 bp) was excised and the DNA was purified using a GENECLEAN TURBO kit (Bio 101, Inc). This DNA fragment was cloned into the mammalian expression vector pcDNA4/HisMax (Invitrogen) for expression of His-tagged mouse MRJ protein.
The sequence encoding 5 TPR repeats of KlC1 (kinesin light chain 1, aa 176–417) was amplified from C57Bl/6 mouse brain total RNA using Accu script RT-PCR kit (Stratagene) and cloned in pGBKT7 vector (82). The full-length sequence of mouse IFT88 (Genbank Acc# AAB59705) kindly provided by Gregory Pazour (83) was cloned into pTrcHis-TOPO and pGEX-4T-1 to obtain pTrcHis/mIFT88 and pGEX-4T-1/mIFT88 for recombinant IFT88 protein expression in E coli. The sequence encoding aa 494–844 of mouse photoreceptor GC1 (Gycy2e) was cloned into pTrcHis-TOPO and pGEX-4T-1 (Amersham Pharmacia Biotech) to obtain pTrcHis/mGC1F2 and pGEX-4T-1/mGC1F2. The eGFP tagged GC1 was constructed by amplifying full-length mouse GC1 using Stratascript RT-PCR system (Stratagene). The DNA fragment was then cloned in frame in pEGFP-N3 vector (Clontech). The resulting clone expresses eGFP fused with the C-terminus of GC1. The accuracy of clones was verified by DNA sequencing using a dye terminator cycle sequencing kit (Applied Biosystems). The shRNA plasmids for MRJ with an insert sequence called M1 (CGTGACGCACTTCCTGTTTGT) and a Negative control with insert sequence (GGAATCTCATTCGATGCATAC) were from Superarray Biosciences.
Creation of Stable His-MRJ Cell Lines
Mouse kidney inner medullar collecting duct (IMCD3) cells were grown in DMEM/F12 medium (Gibco 11330-032) supplemented with 10% foetal bovine serum and 1X Penicillin-streptomycin-glutamine at 37°C with 5% CO2 in air atmosphere. Cells at ≅ 90% confluency were transfected with 5 μg of pcDNA4/His-mMRJ plasmid using LipofectAMINE 2000 (Life technologies) and OptiMEM media (Gibco) according to the manufacturer's protocol. Thirty hours after transfection, cells were subcultured in 10-cm tissue culture dishes at a density of 10 000 cells/dish in normal growth media. Forty-eight hours after transfection, cells were supplied with fresh medium containing 100 μg/mL Zeocin (Invitrogen), and kept in the selection medium until colonies were formed. Colonies were expanded and western blotted with mouse monoclonal anti-polyhistidine antibody (Sigma) to check for the expression of His-tagged MRJ protein. The localization of His-MRJ in the stable cell line was examined by immunocytochemistry using anti-polyhistidine antibody (Sigma).
Preparation of GC1-GFP stable cell lines and MRJ knockdown
Mouse IMCD3 cells were grown up to 90–95% confluency as above in DMEM medium (Invitrogen, 11965-092) with 10% foetal bovine serum and 1X Penicillin-streptomycin-glutamine at 37°C with 5% CO2 in air atmosphere, and were transfected with 2.0 μg of GC1-GFP alone or with 3.0 μg MRJ shRNA or negative control plasmid by using LipofectAMINE 2000 reagent (Invitrogen) and OptiMEM media (Gibco). After 48 h at 37°C in a C02 incubator cells were subcultured in 10-cm dishes with fresh medium. Forty-eight hours after plating, the cells were given fresh medium supplemented with 200 μg/mL of Geneticin or G418 (Invitrogen) and maintained in this medium until colonies formed. After further growth colonies were subjected to western blotting and IC for identification of positive clones. Colonies were subjected to RT-PCR and western blotting to estimate the level of MRJ knockdown. For western blotting cells that had been transfected with GC1-GFP, GC1-GFP/-ve Control and GC1-GFP/MRJ shRNA plasmid were washed with ice-cold PBS, and lysed on ice in HMDEK buffer (9) containing 1% Triton-X 100. Cells were homogenized by repeated passage through a 22-gauge needle and centrifuged at 10,000 ×g for 15 min at 4°C. Protein concentrations were determined using the BCA method (Pierce), and samples (∼ 150 μg) were separated by SDS-PAGE and transferred to PVDF membranes.
The authors thank Dr Ching-Hwa Sung of Weill Medical School of Cornell University for graciously providing the bovine retinal yeast library, antiserum to the MRJ protein and an expression plasmid for bovine MRJ. We thank Gregory Pazour of the University of Massachusetts for providing the rabbit antibodies to IFT88 and IFT57 and a plasmid encoding mouse IFT88. We thank Robert Molday of the University of British Columbia for providing GC1 monoclonal antibodies, and Alexander Dizhoor of Pennsylvania College of Optometry for a rabbit antibody to GC1. The authors also thank Win Sale, Greg Pazour and Wolgang Baehr for their comments on an early version of the manuscript. This work was supported by NIH research grant EY03222 (JCB), NIH Core Grant for Vision Research P30-EY01931 and Development Funds from the Medical College of Wisconsin. SAB and CI were supported in part by an NIH Training Grant in Vision Science, T32-EY014537.