The Docking of Kinesins, KIF5B and KIF5C, to Ran-binding Protein 2 (RanBP2) Is Mediated via a Novel RanBP2 Domain* □ S

The Ran-binding protein 2 (RanBP2) is a vertebrate mosaic protein composed of four interspersed RanGTPase binding domains (RBDs), a variable and species-specific zinc finger cluster domain, leucine-rich, cyclophilin, and cyclophilin-like (CLD) domains. Functional mapping of RanBP2 showed that the domains, zinc finger and CLD, between RBD1 and RBD2, and RBD3 and RBD4, respectively, associate specifically with the nuclear export receptor, CRM1/exportin-1, and components of the 19 S regulatory particle of the 26 S proteasome. Now, we report the mapping of a novel RanBP2 domain located between RBD2 and RBD3, which is also conserved in the partially duplicated isoform RanBP2L1. Yet, this domain leads to the neuronal association of only RanBP2 with two kinesin microtubule-based motor proteins, KIF5B and KIF5C. These kinesins associate directly in vitro and in vivo with RanBP2. Moreover, the kinesin light chain and RanGTPase are part of this RanBP2 macroassembly complex. These data provide evidence of a specific docking site in RanBP2 for KIF5B and KIF5C. A model emerges whereby RanBP2 acts as a selective signal integrator of nuclear and cytoplasmic trafficking pathways in neurons. The small

The small nuclear GTPase, Ran, is a key regulator of protein (1)(2)(3) and RNA (4 -7) nuclear export and protein nuclear import (2, 6, 8 -11). In contrast to earlier proposals (12,13) and other GTPase-mediated processes (14), recent data support that nucleocytoplasmic trafficking of cargoes across the nuclear envelope is independent of nucleotide hydrolysis (15)(16)(17)(18)(19)(20)(21). Instead, a predicted Ran-GTP to Ran-GDP gradient from the nucleus to the cytosol is proposed to propel the nuclear transport across the nuclear pore in a vectorial fashion (2,22,23). This predicted Ran-nucleotide gradient, together with the selective compartmentalization of key RanGTPase modulators in the nucleus and cytosol, limits the pool of transporters available for polarized delivery of cargoes into and from the nucleus, as these carriers act as sensors of the nucleotide-bound state of Ran (for review see Refs. 24 and 25). Among several Ran-dependent nuclear transporters recently identified, CRM1/exportin-1 (26), a member of the importin ␤/karyopherin-␤ class (22,23), was shown to associate with Ran-GTP and mediate the nuclear export of substrates containing nuclear export signal sequences (27)(28)(29)(30)(31). Three key players with restricted subcellular compartmentalization are thought to play a pivotal role in maintaining the predicted Ran-nucleotide bound gradient across the nuclear envelope. The chromatin-associated Ran-nucleotide exchange factor, RCC1 (32), promotes the production of nuclear Ran-GTP via the exchange of RanGDP to RanGTP (33). In the cytosol, RanGTP hydrolysis seems to be mediated by the costimulation of RanGTPase-activating protein (34) and high affinity Ran-binding proteins (34,35). This mechanism presumably ensures that loading of cargo destined for nuclear import and unloading of nuclear exported substrates, and loading of cargo for nuclear export and unloading of nuclear imported cargo, respectively, are compartmentalized in the cytosol and nucleus.
The Ran-binding protein 1 (RanBP1) 1 (36) and Ran-binding protein 2 (RanBP2) (37)(38)(39)(40) are two cytosolic proteins with high affinity for RanGTP (34 -36, 40, 41). RanBP1 is well conserved from yeast to higher eukaryotes (24,42,43). Conversely, RanBP2 is a large scaffold protein with tissue-restricted expression (37,44), four highly homologous RanBP1 domains (RBD n ϭ 1-4), and no orthologs in yeast and Drosophila genomes (24,44). In humans, the RanBP2 gene is partially duplicated once in chromosome 2 and designated RanBP2L1 (45). This gene encodes the leucine-rich domain, two (RBD2 and RBD3) of the four RBDs of RanBP2, but lacks its zinc finger cluster, W1W2 tandem repeats, RBD4, cyclophilin, and the C-terminal part of CLD domains (44). RanBP2 is highly expressed in retinal neurons (37) and is also shown to localize at cytoplasmic fibrils emanating from the nuclear pore complex of liver cells (38 -40). Several reports seem to implicate RanBP1 and RanBP2 in terminal steps of nuclear export and possibly, initial steps of nuclear import (10,24,35). Nuclear export transporters, such as exportin-1, form stable complexes with their nuclear cargo in the presence of RanGTP (27,46,47). These competent nuclear export complexes are thought to transverse the nuclear pore complex and dock at cytoplasmic stations at the cytosolic face of the nuclear pore complex or its vicinity (24,25). In support of this model is the observation that the zinc finger-rich cluster domain of RanBP2 constitutes a specific docking site for exportin-1, and in contrast to the association of Ran-GTP with exportin-1, leptomycin-B does not affect this interaction (48).
Although progress has been made in identifying the components mediating nucleocytoplasmic transport, far less is known about how the cargoes are dispatched from (and to) the nuclear transport complexes once they exit the nucleus, disassembled/ assembled into such complexes, and captured by downstream effectors involved in substrate targeting to various subcellular destinations. Recent structure-function analysis of RanBP2 (41,48,49) suggests that selective domains of RanBP2 independently recruit specific molecular partners and provide a dynamic platform for the delivery (and/or reception) of cargoes to (and/or from) the downstream cytosolic components yet to be identified. RanBP2 contains several RBDs interspersed along its long primary structure. The zinc finger cluster domain (ZnF) of RanBP2 is flanked by the RBD1 and RBD2, suggesting that upon nuclear exiting and docking of exportin-1-RanGTPcargo complexes to ZnF (48), one or more of the neighboring RanBDs may promote the disassembly of nuclear complexes via RanGTP hydrolysis. Similarly, we previously found (49,50) that a domain with weak homology to the C-terminal cyclophilin domain of RanBP2, cyclophilin-like domain (CLD), which is flanked by RBD3 and RBD4, associates specifically and in a tissue-restricted fashion with components of the 19 S cap of the 26 S proteasome. This suggests that the predicted unfolding/ chaperone activity of the 19 S regulatory particle (51,52), in concert with RanGTP-mediated hydrolysis by the flanking RBDs, may also promote the disassembly of 19 S cap-associated cargoes, 19 S cap itself, and/or target some of the cargoes/ carriers for possible degradation. In addition, if the mechanisms and components mediating the release and loading of nuclear cargo at the cytosolic face or vicinity of the nuclear pore complex are tightly coupled, and possibly subjected to regeneration, then factors mediating the delivery to, and transport from, the nuclear pore may be also subjected to analogous coupling and regenerating cycles. Such factors may also interface with nuclear shuttling components. This process is hinted, for example, by data reporting the localization at the nuclear pore complex of "atypical" proteins, such as Sec13p and related Seh1p, which have been implicated in ER to Golgi trafficking (53,54).
To gain further insight into the role of RanBP2 in the integration of signaling and nucleocytoplasmic trafficking processes, we postulated whether the segment region between the RBD2 and RBD3 of RanBP2, JX2, constitutes a novel functional domain. Here, we show that this previously unrecognized domain associates specifically, directly, and in a neuronal biased fashion with two members of kinesin heavy chain in vitro and in vivo. The implications of these findings are discussed.
Analytical Screening Assay of Retinal JX2-binding Proteins-Analytical binding reactions of GST-JX2 (2.2 M) and competitor, free JX2 (11-22 M) proteins with retinal and other CHAPS-solubilized tissue extracts (ϳ2-3 mg) were performed at 4°C followed by affinity capture and washing of GST-bound complexes exactly as described previously (48 -50) for other RanBP2 screening assays. CHAPS extracts were prepared in the presence of EDTA-free protease inhibitor mixture (Roche Molecular Biochemicals). GST-bound complexes were resolved on SDS-PAGE and silver-stained as described (50).
Purification and Identification of Retinal p120 -The bovine retinal p120 protein was purified by scaling up the analytical binding reactions ϳ1000-fold. GST-bound complexes were incubated twice with elution buffer (50 mM Tris⅐HCl, pH 7.5, 500 mM NaCl, 2 mM CaCl 2 , 2 mM MgCl 2 , 0.2% Triton X-100, 0.1% ␤-mercaptoethanol) at room temperature for 10 min, and eluted p120 was concentrated and the buffer exchanged with storage buffer (50 mM Tris⅐HCl, pH 8.0, 0.1% ␤-mercaptoethanol, 10% glycerol) in Centricon-100 (Millipore) at 4°C. Eluted samples were resolved on 7.5% SDS-PAGE, stained with Coomassie Blue, and destained, and the SDS-PAGE band containing retinal p120 was isolated. Sample and neighboring mock (blank) gel pieces were placed on prewashed Eppendorf tubes (0.1% trifluoroacetic acid, 60% acetonitrile) and washed in 50% acetonitrile, 50 mM ammonium bicarbonate and then 50% acetonitrile, 10 mM ammonium bicarbonate for 30 min on a nutator. Gel slices were dried in a SpeedVac followed by in-gel trypsin digestion (0.1 g of trypsin/ϳ15 mm 3 of gel in 10 mM ammonium bicarbonate) of the mock and protein samples at 37°C for 24 h. Tryptic peptide mixtures were divided into two pools. 5 (v/v) and 95% (v/v) of the sample were subjected to MALDI-MS and Q-Tof MS/MS, respectively, for peptide mass spectra analysis and determination of peptide sequences. Peptide fingerprint analysis against NCBI and EMBL nonredundant data base was performed with Profound and PeptideSearch using a mass tolerance of Ϯ0.012% for monoisotopic and Ϯ0.05% for observed average masses. For Q-Tof analysis, samples were cleaned with a small packed "ZipTip." MS/MS spectra were searched using the Sequest search program. Mass spectrometry analyses of retinal p120 were carried out at the Cancer Center Mass Spectrometry Resource Laboratory of Yale University.
CHAPS-solubilized bovine extracts of different tissues and bovine and human retinal homogenates were prepared and concentrations determined exactly as described previously (50,57). Western blot analysis of tissues extracts and all analytical binding reactions with various GST-fused constructs (2.2 M) were carried out exactly as described previously (48 -50) for silver-stain analysis of binding reactions. In vitro association assays between recombinant GST-fused domains of KIF5 isoforms and unfused JX2 of RanBP2 were performed under similar conditions in CHAPS incubation buffer as described.
Coimmunoprecipitation Assays-Antibodies were incubated with protein A-agarose beads in a phosphate-saline buffer, pH 7.4, for 30 min. The beads were precipitated and washed 3 times with 10 volumes of phosphate-buffered saline. Immunoprecipitation assays were carried out with retinal extracts solubilized in Nonidet P-40 homogenization buffer (1% Nonidet P-40, 150 NaCl, 50 mM Tris⅐HCl, pH 8.0) without ␤-mercaptoethanol as described previously (57). 10 g of anti-RanBP2 and 6 g and 2 l of kinesin monoclonal and polyclonal antibodies, respectively, were used per 8 l of protein A-agarose beads to immunocapture cognate retinal complexes. H1-bound immunocomplexes were indirectly captured with a rabbit secondary antibody against Fc region of mouse H1 monoclonal IgG1. Approximately 2-3 mg of retinal extracts were incubated with each of the antibodies separately for 60 min at 26°C on a nutator, loaded onto spin filters (CytoSignal, Irvine, CA), and the supernatant filtrated, and the beads were washed three times with 0.5 ml of Nonidet P-40 homogenization buffer followed by the elution of the immunoaffinity captured complexes with 40 l of 1ϫ SDS-sample buffer. Boiling of the eluates was carried out in the presence of 1 mM dithiothreitol. The immunocomplexes were resolved by SDS-PAGE and Western blot analysis carried out with antibodies against RanBP2, kinesins, and RanGTPase at concentrations described under "Immunochemistry and Tissue Extract and Homogenate Preparation."

RESULTS
The Connecting Segment between RBD2 and RBD3 of RanBP2, Jx2, Is a Novel Domain That Associates Specifically with a 120-kDa Protein in Brain and Retinal Extracts-We extended structure-function analysis of RanBP2 to its putative JX2 domain (Fig. 1a) by employing the same systematic protein screening binding assay previously developed and reported for the functional characterization of other RanBP2 domains (48 -50). In particular, we screened for retinal protein(s) interacting specifically with this RanBP2 domain moiety in the presence of GST-JX2 but not with the concomitant presence of excess of unfused JX2 (competitor). As shown in Fig. 1b, incubation reactions of GST-JX2 with retinal extracts lead to the specific association to the JX2 protein moiety of a protein with an apparent molecular mass of 120 kDa, p120 (Fig. 1b, lane 3), as this interaction could not be observed in incubation reactions containing excess of free (unfused) JX2 (Fig. 1b, lane 4). In contrast to other RanBP2-interacting molecular partners (41,49,50), the association between JX2 of RanBP2 and p120 is not significantly affected by the presence of nonhydrolyzable nucleotide analogs (Fig. 1b, lanes 5-7). Moreover, in analogy with the interaction between CLD of RanBP2 and components of the 19 S regulatory particle of the 26 S proteasome (49,50), the formation of the JX2-p120 binary complex was observed in the retina and brain (Fig. 1b, lanes 3 and 9) but not in other tissues tested such as liver, kidney, spleen, and skeletal muscle.
Large Scale Purification and Identification of Retinal p120 -To determine the identity of retinal p120, we carried out large scale incubation reactions (ϳ1,000-fold) with retinal extracts and GST-JX2, eluted p120 from GST-JX2 in the presence of 500 mM NaCl and resolved the purified p120 from putative contaminating proteins on SDS-PAGE (Fig. 1b, lane 1). The purified p120 comigrated exactly with the retinal protein, p120, identified initially on analytical incubation reactions (Fig. 1b,  e.g. compare lanes 1 and 3). After purification and tryptic digestion of p120, digested p120 was then split into a smaller pool (5% v/v) and two identical pools (47.5% v/v) for matrixassisted laser desorption/ionization time-of-flight (MALDI-TOF) and nanoelectrospray tandem mass spectrometry analysis, respectively ( Fig. 1 of Supplemental Material). First, the FIG. 1. a, schematic diagram of RanBP2 primary structure. RanBP2 consists of several structural modules such as four interspersed RBDs, a variable and species-specific zinc finger cluster domain (ZnF), leucine-rich (LD), cyclophilin (CY), cyclophilin-like (CLD), and tandem repeat domains (W1W2). The junction segment, JX2, between RBD2 and RBD3, is the focus of this study. b, a brain and retinal protein of 120 kDa associates specifically with JX2 domain of RanBP2. Silver-stain SDS-PAGE analysis of analytical binding reactions of GST-JX2 coprecipitates from various detergent-solubilized tissue extracts shows that a single protein of ϳ120 kDa (p120, arrow) interacts specifically with JX2 moiety of GST-JX2 in retinal (lane 3) and brain (lane 9) extracts but not in other tissues (lanes 8 and 10 -12). The interaction of p120 with GST-JX2 was highly specific because this interaction was inhibited by the presence of an excess of unfused JX2 (lane 4), and GST by itself did not associate with retinal p120 (see Fig. 5, a-d). This interaction was not significantly affected by the presence of nonhydrolyzable nucleotide-triphosphate analogs (ϳ400 M) (lanes 6 -7). The GST-JX2 protein runs as a doublet on SDS-PAGE as noted in the figure. An aliquot of purified retinal p120 (2% v/v) (lane 1) comigrates exactly with the counterpart protein identified in analytical binding assays (e.g. compare lanes 1 and 3). Sk. Muscle, skeletal muscle; ATP␥S, adenosine 5Ј-O-(thiotriphosphate). purified p120 was analyzed by MALDI-TOF. Twelve of the 40 bovine tryptic peptide masses obtained matched the human kinesin heavy chain (KHC), KIF5C, and covered 11% of its sequence (58). The murine KIF5C ortholog was published although this manuscript was in revision and reported to be neuronal specific (59). In light of the extremely high homology between KIF5C and the KHC-related members, KIF5A (56,60) and KIF5B (61, 62) (Fig. 2 of Supplemental Material), and potential heterodimerization between these members, we searched for tryptic peptide masses unique to these kinesin members by tandem mass spectrometry of the larger pool of p120 digested with trypsin. Analysis of the Q1 mass spectrum (first mass spectrum) identified over 20 masses that matched KIF5A, KIF5B,and KIF5C ( Fig. 1 of Supplemental Material). In addition, 10 and 9 masses were unique to KIF5B and KIF5C, respectively, and none could be uniquely matched to KIF5A (Fig. 1 of Supplemental Material). Moreover, all peptide masses ended with, and were preceded by, lysines or arginines as expected from a tryptic digest. Some individual peptides common and unique among these kinesin members were then isolated, fragmented, and resolved in a tandem mass spectrum by nanoelectrospray tandem quadrupole-TOF-MS. To this end, six peptide sequences were determined from the nested set of counterpart-fragmented peptides. Two were identical to KIF5A, KIF5B, and KIF5C, two unique to KIF5B, and the other two just to KIF5C (Figs. 1 and 2 of Supplemental Material). Altogether, these results confirmed the presence of two distinct kinesin members, KIF5B and KIF5C, in purified p120 sample.
Immunological Characterization of JX2-associated Complex-Several antibodies have been generated against KIF5 isoforms. In particular, mAbs H1 and H2 were produced against kinesin heavy chains purified from bovine brain (55), and polyclonal Abs, Pcp42 and Bsp36 (56), were raised against the C-terminal coiled-coil domains of recombinant KIF5B and KIF5A, respectively. However, we decided to revisit the immunological characterization of these antibodies against all three recombinant KIF5 isoforms (KIF5A, KIF5B, and KIF5C) in light of their high homology and because at the time these antibodies were generated not all KIF5 isoforms had been molecularly defined. On Western blots (Fig. 2), mAb H1 recognizes KIF5A and KIF5C and mAb H2 immunoreacts with all three KIF5 isoforms. The polyclonal antibody Pcp42 recognizes KIF5B and KIF5C, whereas Bsp36 immunoreacts with KIF5A and KIF5C.
To confirm independently the identity of the kinesins isolated by our screening assay and mass spectrometry analysis, we carried out the same analytical incubation reactions of retinal extracts with GST-JX2 of RanBP2 as described previously, followed by Western blot analysis of the resolved GST-JX2 coprecipitates with mAbs, H1 and H2, and polyclonal Abs, Pcp42 and Bsp36 (Fig. 3, a-d). The JX2 domain of RanBP2 coprecipitated specifically KIF5s, which were immunoreactive to mAbs, H1 and H2 (Fig. 3, a and b, lane 6), and polyclonal Abs, Pcp42 and Bsp36 (Figs. 3, c and d, lane 6). The interaction between GST-JX2 and KIF5s is highly specific because the presence of excess free JX2 (Fig. 3, a-d, lane 7) completely impaired the interaction, and GST alone did not interact with KIF5s (Fig. 3, a-d, lane 8). Moreover, the KIF5s associating with GST-JX2 had exactly the same electrophoretic mobility as those observed in human and bovine retinal homogenates and extracts (Fig. 3, a-d; compare lane 6 with lanes 1-3).
Kinesin light chains (KLC) often copurify with KHCs (63) and seem to remain associated with these at the cargo binding tail domain (64). Moreover, KLCs appear to play an important role in the binding of cargoes to the KHC complex (65) and determining the subcellular distribution of KHC (66). Hence, we investigated whether the accessory protein, KLC, was part of the JX2-KHC complex. Immunoblot analysis of retinal GST-JX2 coprecipitates showed that the accessory KLC of KIF5B and KIF5C coprecipitated with these, and the formation of this quaternary complex was fully abrogated in a parallel reaction containing excess free JX2 competitor (Fig. 3e, compare lanes 6  and 7). The GST-JX2-associated KLC had also the same electrophoretic mobility as that observed in human and bovine retinal homogenates and extracts (Fig. 3e, lanes 1-3).
The Association of KIF5s with JX2 Domain Is Independent of the Presence of the RanBP2 Flanking Domains, RBD2 and RBD3, and RanGTPase-We investigated whether the domains upstream and downstream of JX2 of RanBP2, RBD2, and RBD3, respectively, and association of RanGTPase to these domains affected the interaction of KIF5s with JX2 domain of RanBP2 in GST pull-down assays with retinal extracts. As seen in Fig. 4, the GST-RBD2-JX2-RBD3 construct associated with retinal KIF5s (Fig. 4, a and b, lanes 3 and 4) at similar levels to those observed for GST-JX2 (Fig. 4, a and b, lanes 1 and 2) and independently of the presence of nonhydrolyzable GTP␥S (Fig.  4, a and b, compare lanes 3 and 4). However, in stark contrast to GST-JX2, the GST-RBD2-JX2-RBD3 associated with retinal RanGTPase (Fig. 4c, lanes 3 and 4) and this interaction was strongly enhanced by the presence of nonhydrolyzable GTP␥S (Fig. 4c, compare lanes 3 and 4). Finally, excess of unfused JX2 domain was sufficient to abolish completely the interaction between the KIF5s and GST-RBD2-JX2-RBD3 (Fig. 4, a and b;  compare lanes 5 and 3) but did not affect the association of RanGTPase with GST-RBD2-JX2-RBD3 (Fig. 4c, compare  lanes 5 and 3).
Expression Profile of KIF5A, KIF5B, KIF5C, and RanBP2L1-We then looked at the tissue expression profile of KIF5s with several cognate antibodies (Fig. 5, a-d), and we investigated whether the purified retinal p120 was immunoreactive to these antibodies and had the same electrophoretic mobility as the KIF5s observed in other tissues. As shown in Fig. 5, Western blots with H1 antibody showed that KIF5A and/or KIF5C (Fig.  5a) were extremely abundant in brain and retina but expression was also detected in the retina pigment epithelium and heart, and a lower electrophoretic mobility isoform seems to be highly expressed in skeletal muscle. Traces of KIF5A and/or KIF5C expression seem to also be detected in other tissues except liver. The extreme abundance of KIF5A and/or KIF5C in retina and brain was also confirmed with the polyclonal antibody, Bsp36 (although this antibody did not detect the skeletal muscle and heart isoforms). In contrast, the Pcp42 polyclonal antibody, which recognizes KIF5B and KIF5C, detected similar expression levels of KIF5 in all tissues tested (Fig. 5c). This may be explained by the uniform and ubiquitous expression of  lanes 1 and  3), which was also strongly increased by the presence of GTP␥S (c, compare lanes 2 and 4 and lanes 3 and 4). Purified retinal p120 was immunoreactive to all KIF5 antibodies and had the same electrophoretic mobility of KIF5s expressed in retina and brain extracts (lane 10). e, in addition to RanBP2 (see Fig. 8a, lanes 1  and 2), the JX2 antibody recognizes a 95-kDa protein, RanBP2L1 (large arrow), in all, but spleen, bovine tissue extracts (50 g) tested. A 93-kDa RanBP2L1 isoform (small arrow) was also expressed in liver, spleen, kidney, and lung. Note that in the retina an additional major immunoreactive 75-kDa isoform (arrowhead) is generated upon solubilization of the retina with CHAPS (lane 3, 100 g) but not Nonidet P-40 (lane 2, 100 g) or RIPA buffer (lane 1, 100 g). These isoforms are likely RanBP2L1 or their processed products. KIF5B in different tissues. Intriguingly, the H2 monoclonal Ab, which detected all recombinant isoforms of KIF5s (Fig. 2c), recognized predominantly high expression levels of KIF5s in the retina and brain (Fig. 5b). Conversely to mAb H1 (Fig. 5a), mAb H2 also recognized higher expression levels of KIF5s in kidney, spleen, and lung and no expression in skeletal muscle (Fig. 5b). Finally, p120 was immunoreactive to all the antibodies tested and had the same electrophoretic mobility as those observed in neuronal tissue extracts (Fig. 5, a-d, lane 10).
The Coiled-coil/Tail Domains of KIF5B and KIF5C but Not KIF5A Interact Directly with JX2 Domain of RanBP2-Mass spectrometry data did not identify any tryptic fragment unique to KIF5A suggesting that this KIF5 isoform does not interact with RanBP2. Moreover, it is possible that a putative retinal protein may bridge the interaction between the JX2 domain of RanBP2 and KIF5s and/or posttranslational modifications in KIF5s determine the formation of the binary complex. To investigate these possibilities and to determine the domains of  1-2) and immuno-coprecipitates (lanes 3-7) was performed with an antibody against JX2, a domain shared by RanBP2 and RanBP2L1. Conversely to the antibody against the unique ZnF of RanBP2 (lane 3), the anti-JX2 antibody (lane 4) immunopurified both RanBP2 (arrow) and RanBP2L1 (arrowhead). Both of these were also detected in aliquots of retinal extracts, and RanBP2L1 seems to be more abundant than RanBP2 (lanes 1 and 2). KIF5s interacting with RanBP2, we separately expressed and purified the coiled-coil/tail and motor domains of KIF5s and performed cell-free binding assays between these and the unfused (free) JX2 domain of RanBP2. As shown in Fig. 7a, the JX2 domain of RanBP2 interacted only with the coiled-coil/tail domains of KIF5B and KIF5C but not KIF5A or their motor domains. KIF5B seems to exhibit a higher affinity toward the JX2 domain of RanBP2 than KIF5C. We then investigated some properties of the association between KIF5B and JX2 of RanBP2. As shown in Fig. 7b, the coiled-coil/tail domain of KIF5B bound to JX2 domain of RanBP2 in a concentration-and saturation-dependent fashion. DISCUSSION We report the identification of a novel structural and functional domain, located between RBD2 and RBD3, in RanBP2. This domain interacts in vitro and in vivo, specifically and directly, with the coiled-coil/tail domains of neuronal abundant KIF5C and ubiquitous KIF5B. This is supported by the following observations: (i) the JX2 domain of RanBP2 interacted in retinal and brain extracts, but not other tissues tested, with retinal p120 KHCs; (ii) purified p120 was composed of three classes of tryptic peptides, those identical to KIF5A, KIF5B, and KIF5C, those unique to KIF5B and KIF5C, but none unique to KIF5A; (iii) purified p120 was immunoreactive to antibodies against KIF5s; (iv) KIF5B and KIF5C coimmunoprecipitated with anti-RanBP2 antibodies and likewise RanBP2 coimmunopurified with antibodies against these kinesins; and (v) the coiled-coil/tail domains of recombinant KIF5B and KIF5C but not KIF5A associated directly with recombinant JX2 of RanBP2 in cell-free binding assays. KIF5B and KIF5C association with RanBP2 is also further supported by the specific co-association of the KLC with RanBP2. RanGT-Pase is also part of the RanBP2-kinesin complex in vivo. Finally, the ubiquitously expressed and putative RanBP2L1 protein, a partially duplicated isoform of RanBP2 that shares the JX2 domain (44), did not associate in vivo with KIF5B and KIF5C.
In light of RanBP2 association with the KIF5B and KIF5C in retina and brain, ubiquitous and neuronal abundant expression of KIF5B and KIF5C, respectively, and independent interaction in vitro of KIF5C and KIF5B with JX2 domain of RanBP2, our data indicate that KIF5B and KIF5C may interact independently with RanBP2 in the tissues tested. The tissue expression profile of KIF5A, KIF5B, and KIF5C and the immunological characterization of the anti-KHC mAbs, H1 and H2, presented in this report are in disagreement with those reported by Kanai et al. (59) and are more in line with earlier reports of Bloom and colleagues (55,67). This is because KIF5B was found to be immunoreactive against mAb H2 and immunorelated kinesins in non-neuronal tissues/cells cross-reacted with mAb H2 (and H1). Moreover, mAb H2 had an intriguing immunological behavior as it recognized all recombinant KIF5 isoforms, and yet the expression of these seems to be predominant in the retina and brain. One possibility may be that the expression levels of the neuronal abundant KIF5s (KIF5A and/or KIF5C) are significantly higher than that observed for KIF5B in such tissues and/or posttranslational modifications may skew the detection of some kinesins in non-neuronal tissues.
Our data also raise important questions regarding the function of KIF5 members. Kanai et al. (59) proposed these KIFs have redundant functions because kif5C Ϫ/Ϫ mice exhibit a mild central nervous system phenotype and either ectopically expressed KIF5A or KIF5C seems to rescue the gross phenotype of abnormal perinuclear condensation of mitochondria in culture cells derived from embryonic lethal kif5B Ϫ/Ϫ mice (59). However, the retinal phenotype of kif5C Ϫ/Ϫ mice remains unknown. Also, the sharp contrast of kif5C Ϫ/Ϫ and kif5B Ϫ/Ϫ phenotypes together with our data (e.g. KIF5A did not interact with RanBP2) and large numbers of other kinesins expressed in the retina (68) indicate that the roles of KIF5A, KIF5B, and KIF5C are likely not to be redundant.
The restricted interaction of RanBP2 with KIF5B and KIF5C in the retina and brain but not liver, the failure to detect even trace expression levels of KIF5A and KIF5C in liver and very high expression levels of KIF5A and/or KIF5C in neuronal tissues, poses interesting questions about the specific role of KIF5A, KIF5B, KIF5C, and RanBP2 in non-neuronal versus neuronal tissues. Some scenarios are worth considering. First, the high expression levels of KIF5s and RanBP2 in the retina may directly reflect the prodigious rate of membrane turnover and de novo synthesis of opsin (and associated transduction components) to be delivered to the most distal subcellular compartment, the outer segments of photoreceptors cells (69), without compromising other housekeeping protein kinesis processes. Consequently, the demand for the availability of factors that mediate the processing and targeting of opsin, surveillance of this process, and possibly recycling of some of such factors is exceptional. To this effect, it is important to note that  1-4) and binding to JX2 saturates at ϳ0.6 M (compare lanes 4 and 5). Note that like its GST parent counterpart (Fig. 1), the unfused JX2 protein (e.g. lane 1, 40 ng) runs as a doublet in SDS-PAGE. the C-terminal RBD4-cyclophilin domains of RanBP2 selectively associate with specific red/green opsin conformers (41,70). Second, another key feature relates to the necessity of certain cargoes, many likely associated with kinesin(s), to travel often extremely long distances in neuronal cells such as the retinal ganglion cells, whose axons synapse distantly in the visual cortex of the brain. This is supported by data reporting that the delivery of substrates from the ganglion cell bodies to the synapses is dependent on kinesin motor-driven processes (71,72). Third, the possible existence of scaffold proteins assembling signaling biological units with transiently coupled factors is likely to play a significant contribution on sustaining such kinetic and/or cargo burden imposed on some retinal cells. To this end, RanBP2 could mediate channeling of substrates by promoting the efficient signal integration of distinct kinetic and biological events through the recruitment and transient coupling of factors involved in nuclear and kinesin (cytosolic)mediated trafficking processes. Conversely, tissues in which RanBP2 does not interact with KIF5s and/or with less demanding metabolic needs may do well without tightly integrated trafficking pathway(s).
Another important issue pertains to the role of RanBP2 and KIF5B/KIF5C in protein kinesis and biogenesis. In particular, the question arises whether KIF5B and KIF5C mediate the delivery of cargoes to and/or from RanBP2 and what those cargoes might be. The observations that KHC alone (73) and KHC-KLC (74, 75) undergo folding reactions and move poorly on microtubules suggest a transition between two kinesin conformational states: a folded conformer that binds microtubules and moves slowly and an "unfolded"cargo-stabilized counterpart that moves fast. In addition, KLC seems to inhibit microtubule binding by KHC (76). Hence, co-association of KLCs with RanBP2-KIF5⅐KIF5C complex and docking of exportin-1 onto ZnF of RanBP2 is suggestive that KIF5B/KIF5C is primed for transport of cargoes possibly delivered by exportin-1. But it is also possible that RanBP2 represents the final destination for KIF5B/KIF5C cargoes. Whereas some of these may be destined for nuclear import and/or contribute to the disassembly of nuclear exported cargoes, others could be candidates for cytoplasmic delivery of such cargoes via a kinesin-independent mechanism. Such a cross-talk mechanism could be mediated by interaction of KIF5C with actin-based transport motors as reported by Huang et al. (77). Whatever scenario may turn out to unfold, RanBP2 emerges as a key checkpoint for integration of signaling and trafficking processes. Piecing together the components mediating the dynamic assembly of the RanBP2 macroassembly complex will help determine the reciprocal roles of the RBDs of RanBP2, RanGTPase, and kinesins in cargo release and/or loading, the role of RanBP2 in kinesin motor activity, and the nature of the endogenous cargoes delivered by this macroassembly complex. In addition, it will define broadly the RanBP2-mediated signaling and trafficking pathways in neuronal and non-neuronal cells.