Mapping the GRIF-1 Binding Domain of the Kinesin, KIF5C, Substantiates a Role for GRIF-1 as an Adaptor Protein in the Anterograde Trafficking of Cargoes*

γ-Aminobutyric acid, type A (GABAA) receptor interacting factor-1 (GRIF-1) and N-acetylglucosamine transferase interacting protein (OIP) 106 are both members of a newly identified coiled-coil family of proteins. They are kinesin-associated proteins proposed to function as adaptors in the anterograde trafficking of organelles to synapses. Here we have studied in more detail the interaction between the prototypic kinesin heavy chain, KIF5C, kinesin light chain, and GRIF-1. The GRIF-1 binding site of KIF5C was mapped using truncation constructs in yeast two-hybrid interaction assays, co-immunoprecipitations, and co-localization studies following expression in mammalian cells. Using these approaches, it was shown that GRIF-1 and the KIF5C binding domain of GRIF-1, GRIF-1-(124-283), associated with the KIF5C non-motor domain. Refined studies using yeast two-hybrid interactions and co-immunoprecipitations showed that GRIF-1 and GRIF-1-(124-283) associated with the cargo binding region within the KIF5C non-motor domain. Substantiation that the GRIF-1-KIF5C interaction was direct was shown by fluorescence resonance energy transfer analyses using fluorescently tagged GRIF-1 and KIF5C constructs. A significant fluorescence resonance energy transfer value was found between the C-terminal EYFP-tagged KIF5C and ECFP-GRIF-1, the C-terminal EYFP-tagged KIF5C non-motor domain and ECFP-GRIF-1, but not between the N-terminal EYFP-tagged KIF5C nor the EYFP-KIF5C motor domain and ECFP-GRIF-1, thus confirming direct association between the two proteins at the KIF5C C-terminal and GRIF-1 N-terminal regions. Co-immunoprecipitation and confocal imaging strategies further showed that GRIF-1 can bind to the tetrameric kinesin light-chain/kinesin heavy-chain complex. These findings support a role for GRIF-1 as a kinesin adaptor molecule requisite for the anterograde delivery of defined cargoes such as mitochondria and/or vesicles incorporating β2 subunit-containing GABAA receptors, in the brain.

␥-Aminobutyric acid, type A (GABA A ) receptor interacting factor-1 (GRIF-1) and N-acetylglucosamine transferase interacting protein (OIP) 106 are both members of a newly identified coiled-coil family of proteins. They are kinesin-associated proteins proposed to function as adaptors in the anterograde trafficking of organelles to synapses. Here we have studied in more detail the interaction between the prototypic kinesin heavy chain, KIF5C, kinesin light chain, and GRIF-1. The GRIF-1 binding site of KIF5C was mapped using truncation constructs in yeast two-hybrid interaction assays, co-immunoprecipitations, and co-localization studies following expression in mammalian cells. Using these approaches, it was shown that GRIF-1 and the KIF5C binding domain of GRIF-1, GRIF-1-(124 -283), associated with the KIF5C non-motor domain. Refined studies using yeast two-hybrid interactions and co-immunoprecipitations showed that GRIF-1 and GRIF-1-(124 -283) associated with the cargo binding region within the KIF5C non-motor domain. Substantiation that the GRIF-1-KIF5C interaction was direct was shown by fluorescence resonance energy transfer analyses using fluorescently tagged GRIF-1 and KIF5C constructs. A significant fluorescence resonance energy transfer value was found between the C-terminal EYFP-tagged KIF5C and ECFP-GRIF-1, the C-terminal EYFP-tagged KIF5C nonmotor domain and ECFP-GRIF-1, but not between the N-terminal EYFP-tagged KIF5C nor the EYFP-KIF5C motor domain and ECFP-GRIF-1, thus confirming direct association between the two proteins at the KIF5C C-terminal and GRIF-1 N-terminal regions. Co-immunoprecipitation and confocal imaging strategies further showed that GRIF-1 can bind to the tetrameric kinesin light-chain/kinesin heavy-chain complex. These findings support a role for GRIF-1 as a kinesin adaptor molecule requisite for the anterograde delivery of defined cargoes such as mitochondria and/or vesicles incorporating ␤2 subunit-containing GABA A receptors, in the brain.
␥-Aminobutyric acid, type A (GABA A ) 4 receptor interacting factor-1 (GRIF-1) was initially identified from rat brain by a yeast two-hybrid screen searching for GABA A receptor clustering and trafficking proteins (1). It was shown to associate at least in vitro with GABA A receptor ␤2 subunits. GRIF-1 is highly expressed in excitable tissue, most notably in the brain and in the heart (1). It is the orthologue of the human protein, ␤-O-linked N-acetylglucosamine transferase (OGT) interacting protein 98 (OIP98, also termed ALS2CR3, huMilt2, or TRAK2), and it is the homologue of the protein OIP106 (also termed huMilt1 and TRAK1 (2)). GRIF-1 is also probably the orthologue of the Drosophila protein, Milton, a kinesin-associated protein that is involved in the transport of mitochondria to the synapses in retina (3,4). GRIF-1 and OIP106 have also been shown to aggregate mitochondria following overexpression in mammalian cells (5). Recently, the gene encoding OIP106 (TRAK1), Trak1, was identified as the mutated gene in hyrt mice, an animal model of hypertonia (6). hyrt mice were shown to have a deficit of GABA A receptors; wild-type OIP106 (TRAK1) was shown to co-immunoprecipitate with GABA A receptor ␣1 subunits in extracts of mouse brain stem and spinal cord (6). Although the mutant OIP106 was shown to still associate with GABA A receptors, it was concluded that OIP106 (TRAK1) may play a crucial role in regulating endocytic trafficking of receptors and dysfunction disrupts receptor homeostasis leading to hypertonia (6). Thus, GRIF-1, OIP106, and Milton belong to a newly identified family of coiled-coil proteins putatively involved in the trafficking of organelles and GABA A receptors to synapses.
Two additional proteins that associate with high affinity with both GRIF-1 and OIP106 have been identified. These proteins are the enzyme, OGT (2), 5 and, consistent with the role in trafficking, the molecular motor protein, kinesin (5). Kinesin was also shown to be immunoprecipitated from detergent extracts of Drosophila heads with anti-Milton antibodies, thereby dem-onstrating, albeit reportedly indirect, an association between Milton and kinesin (3). Kinesins belong to the kinesin superfamily, i.e. KIFs. KIFs are microtubule-based mechanochemical motors that are involved primarily in the anterograde transport of organelles and protein complexes (reviewed in Refs. 7 and 8). They are particularly important in transport processes in neurons where it is necessary to deliver defined cargoes from the cell body along axons and dendrites to pre-and post-synaptic sites. In brain, GRIF-1 associates predominantly with KIF5A, whereas in heart association is predominantly with KIF5B (5). Following overexpression in human embryonic kidney (HEK) 293 cells, GRIF-1 will co-associate with exogenous KIF5C (5). KIF5A, KIF5B, and KIF5C are members of the kinesin-1 family. KIF5B is ubiquitously expressed, whereas KIF5A and KIF5C are only expressed in neurons (7). They are conventional kinesin heavy chains belonging to the kinesin-1 family, and they share at least 60% amino acid identity. It is probable that each has a binding site for GRIF-1 and that GRIF-1 is promiscuous with regard to association with KIF5 subtypes.
Kinesin-1 proteins are tetramers formed by the association of two kinesin heavy chains (KHC or alternatively KIF5) and two kinesin light chains (KLC). The KHC is formed from an N-terminal motor domain that contains the microtubule and ATP binding sites and a C-terminal non-motor domain. This domain includes a neck, a coiled-coil stalk region, and a cargo binding site in its C-terminal region. The KLC interacts with the KHC via the stalk region. Cargoes bind to either the KLC or to the KHC of kinesin-1 proteins. Increasing evidence suggests that this interaction is mediated by an adaptor protein. For example, mitochondria and syntaxin-1-containing vesicles are attached to the KHC cargo binding domain by the adaptor protein, syntabulin, for their transport to synapses (9); JIP-3, a c-Jun N-terminal kinase (JNK) signaling pathway protein, binds to a six tetratricopeptide motif in KLCs to transport the cargo, amyloid precursor protein (APP; 10); KIF17 forms a complex with mLin10 in the transportation of N-methyl-D-aspartate receptor NR2B subunits to the synapse (11,12); and glutamate receptor interacting protein 1 (GRIP1) is an adaptor protein linking the ␣-amino-3-hydroxy-5methylisoxazole-4-propionate (AMPA) subtype of glutamate receptor containing vesicles to KIF5 (13).
In this report, we have studied the interaction between GRIF-1 and the prototypic kinesin-1, KIF5C, and kinesin light chain. The results obtained provide supporting evidence that GRIF-1 is a kinesin adaptor protein involved in motor-dependent trafficking of organelles and/or proteins. Please note a preliminary report of some of these findings (14).
Anti-GFP and anti-KIF5C 938 -957 antibodies were from Abcam Ltd. (Cambridge, UK); anti-c-Myc clone 4A6 antibodies were from Upstate (Charlottesville, VA); and anti-mouse-Ig Alexa Fluor 633 antibodies from Invitrogen. Anti-FLAG antibodies were raised inhouse against the FLAG amino acid sequence, DYKDDDDK, with an N-terminal cysteine coupled to thyroglobulin and affinity-purified for use using CDYKDDDDK covalently coupled via the cysteine to thiopropyl-activated Sepharose.
Immunoblotting-Immunoblotting was performed as previously described using 25-50 g of protein/sample precipitated

TABLE 1 GRIF-1 interacts directly with the KIF5C non-motor domain: demonstration by yeast two-hybrid interaction assays
The yeast strain, L40, was co-transformed with bait and fish constructs and transformants grown on Ϫ2 SD, i.e. ϪW and ϪL, and Ϫ3 SD, i.e. ϪW, ϪL, and ϪH. Colonies from Ϫ2 plates were re-streaked onto new plates, and colony growth was determined as described under "Experimental Procedures."
Yeast Two-hybrid Interaction Assays-Yeast two-hybrid assays were carried out using a modified LexA system as described previously (5). The Saccharomyces cerevisiae strain, L40 (MATa his3⌬200 trp1-901 leu2-3112 ade2 LYS2::(4lexAop-HIS3) URA3::(8lexAoplacZ) GAL4), was transformed with combinations of GRIF-1 fragments in the activation domain (AD) plasmid, pGADT7, together with each of the KIF5C fragments in the DNA binding domain plasmid, pMBL33. Negative control transformations were carried out using empty AD or DNA binding domain vectors. Positive control transformations were carried out using AD-GRIF-1-(8 -633) with pMBL33␤2-IL, i.e. the DNA binding domain vector containing the residues 303-427 of the intracellular loop of the GABA A receptor ␤2 subunit. Resulting colonies were assessed for reporter gene activation by growth on nutritional selection agar lacking tryptophan, leucine, and histidine.
Confocal Microscopy and FRET Efficiency Determinations-Transfected cells 20 -40 h post-transfection were rinsed three times with ice-cold phosphate-buffered saline followed by fixation for 10 min with 4% (w/v) ice-cold paraformaldehyde. Coverslips were rinsed three times with phosphate-buffered saline and mounted onto a microscope slide using 10 l of mounting solution containing an anti-fading agent (Citifluor, Citifluor Ltd., Leicester, UK). The coverslips were sealed and kept at 4°C until analysis. For visualization of cells transfected with pCM-VKLC, post-fixation coverslips were incubated with anti-c-Myc antibodies (1:4000) overnight at 37°C, washed as described (1), and incubated with anti-mouse-Ig Alexa Fluor 633, and again washed as previously described (1). Cells were imaged using an inverted LSM510 META Zeiss confocal microscope in the multitrack mode. EYFP was excited with the ϭ 514 nm laser line, and emitted light was collected with a long pass filter LP530. ECFP was excited at ϭ 458 nm, and light was collected with a band-pass filter BP 475-525. The absence of bleed-through was checked prior to each experiment by using the Meta software of the microscope. This software enables the decomposition of the mixed emission spectra (i.e. generated from cells co-transfected with both ECFP and EYFP) into the emission of the single fluorescent dyes, i.e. ECFP and EYFP, according to reference spectra generated from cells transfected with either ECFP or EYFP alone. Images were analyzed using the Image Browser software (Zeiss) available with the microscope.
For FRET analyses, again the Meta software mode of the LSM510 META Zeiss confocal microscope was used. FRET efficiency was measured by acceptor photobleaching as described previously by Liu et al. (15) and Nashmi et al. (16).
That is, co-transfected HEK 293 cells were firstly imaged with the ϭ 458 nm laser to visualize ECFP-tagged proteins. A defined area of co-localization within one cell was selected, and EYFP was photobleached for ϳ30 s using the ϭ 514 nm laser at full power. Cells were imaged post-photobleaching with the ϭ 458 nm laser. FRET was measured as the increase of ECFP fluorescence after photobleaching where values were taken at t ϭ 1.6 s prior to photobleaching and t ϭ ϳ32 s post-photobleaching and were corrected for background fluorescence (usually of the order of ϳ10%) determined by the imaging of untransfected HEK 293 cells. The relative FRET efficiency was calculated by the following: (1 Ϫ [pre-bleached intensity of ECFP/ post-bleached intensity of ECFP]) ϫ 100% (14). For each FRET experiment, pseudo-FRET was always determined by applying the photobleaching protocol to cells transfected with pECFP alone. Pseudo-FRET values (typically ϳ2.9%) were always subtracted from the calculated FRET efficiencies of the test samples. As a further control, an area of the cell that was not photobleached was also analyzed in parallel for FRET to ensure that the FRET efficiency values determined were not artifacts due to the photobleaching protocol. A positive control, an ECFP-EYFP tandem construct linked by two amino acids (17), and a negative control, co-transfection of the separate clones pECFP and pEYFP, were both used to validate the system.

GRIF-1 Associates with the KIF5C Non-motor Domain: Demonstration by Both Yeast Two-hybrid Interaction Assays and Immunoprecipitation Strategies
We have previously shown that GRIF-1 associates with endogenous kinesin in the brain and in the heart by co-immunoprecipitation of GRIF-1 and kinesin using anti-GRIF-1 874 -889 antibodies. Association was also demonstrated, again by co-immunoprecipitation, between exogenous GRIF-1 and exogenous KIF5C both co-expressed in HEK 293 cells. Further, yeast two-hybrid interaction assays using GRIF-1 and KIF5C as fish and bait, respectively, suggested that association between the two proteins is probably direct and involves the first coiled-coil domain of GRIF-1, GRIF-1-(124 -283) (5). To delineate the GRIF-1 binding domain of KIF5C initially, the KIF5C motor domain, i.e. KIF5C-(1-335), and KIF5C non-motor domain, i.e. KIF5C-(336 -957), were subcloned in-frame into either the yeast twohybrid DNA binding domain vector, pMBL33, or the modified mammalian expression vector, pCMVTag4a, to yield N-terminally c-Myc-tagged fusion proteins. The association between GRIF-1 and KIF5C was then studied by both yeast two-hybrid interaction assays and immunoprecipitation strategies following the co-expression of GRIF-1 and KIF5C in HEK 293 cells. The results are summarized in Table 1 and Fig. 2.

GRIF-1 interacts directly with the KIF5C cargo binding domain coil 3: demonstration by yeast two-hybrid interaction assays
The yeast strain, L40, was co-transformed with bait and fish constructs and transformants grown on Ϫ2 SD, i.e. ϪW and ϪL and Ϫ3 SD, i.e. ϪW, ϪL, and ϪH. Colonies from Ϫ2 plates were re-streaked onto new plates and colony growth was determined as described under "Experimental Procedures."
The results are representative of n ϭ 3 independent co-transformations.
The above experiments were also repeated for KLC (70 kDa). No association between GRIF-1 and KLC was detected by either yeast two-hybrid or co-immunoprecipitation assays (Table 1 and Fig. 2).

Characterization of ECFP-GRIF-1 and EYFP-KIF5C Constructs-
The following in-frame constructs were generated by molecular cloning: full-length N-and C-terminally tagged KIF5C-EYFP, an N-terminal EYFP-tagged KIF5C-(1-335) motor domain, a C-terminal EYFP-tagged KIF5C-(336 -957) nonmotor domain, and an N-terminal ECFP-GRIF-1 (Fig. 1). Each was characterized with respect to molecular size by immunoblotting following expression in HEK 293 cells using anti-GFP antibodies and anti-GRIF-1 or anti-KIF5C antibodies. This was to ensure that the respective EYFP-or ECFP-tagged proteins did not undergo proteolytic digestion, thus the fluorescent moieties remained attached to the target proteins. The results for the full-length KIF5C constructs and GRIF-1 are shown in Fig. 4 and for the KIF5C-(1-335) motor domain and KIF5C-(336 -957) non-motor domains in Fig. 5. In all cases, an increase in molecular weight for the tagged proteins was observed consistent with the addition of ECFP or EYFP. Further, a single band was found for all constructs in immunoblots using anti-GFP antibodies.    Fig. 4 (E-G) and for the KIF5C motor and nonmotor domains in Fig. 5 (C-F). In all cases except as expected for the KIF5C-(1-335) and GRIF-1 double transfectants, a specific association was found between the tagged-untagged or tagged-tagged GRIF-1/KIF5C combinations. Note that the results for the immunoprecipitations for the C-terminaltagged KIF5C constructs, i.e. KIF5C-EYFP, are not shown, because they gave identical results to those found for the immunoprecipitations of N-terminal-tagged KIF5C and GRIF-1/ECFP-GRIF-1.
Co-localization Studies-In transfected COS-7 cells, ECFP-GRIF-1 was found to be distributed throughout the cell cyto-plasm with some areas showing notable enrichment at sites close to the nucleus (Fig. 6, B and C). This localization pattern is similar to that reported for the expression of GRIF-1 in HEK 293 cells (1) where the enriched areas were shown to co-localize with aggregated mitochondria (5). EYFP-KIF5C and KIF5C-EYFP expressed alone in COS-7 cells showed the same distribution profile with fluorescence visible throughout the cell cytoplasm, and, when they are present, EYFP-KIF5C was concentrated at the ends of cellular extensions (Fig. 6E). When ECFP-GRIF-1 plus EYFP-KIF5C or KIF5C-EYFP was co-expressed, the distribution of ECFP-GRIF-1 was changed. In all cells imaged (n ϭ 35), the majority of ECFP-GRIF-1 was recruited to KIF5C-enriched areas, which were particularly apparent at the end of cellular extensions. In cells where no processes were evident, co-localization was seen adjacent to the cell membrane (supplemental Fig. S1). In some co-transfected cells (19 out of 35 cells; 54%), some ECFP-GRIF-1 was still found in the cell cytoplasm where it was not associated with EYFP-KIF5C (Fig. 6, G  and H). This may be a result of overexpression of ECFP-GRIF-1 in those particular cells. Fig. 6 also shows the distribution of mitochondria in COS-7 cells transfected with either ECFP-GRIF-1 (L-P), ECFP-KIF5C (Q-U), or ECFP-GRIF-1 plus EYFP-KIF5C (V-AA). Aggregated mitochondria are only seen in cells when ECFP-GRIF-1 was overexpressed; i.e. compare Fig. 6N (ECFP-GRIF-1 transfectant) with Fig. 6S (EYFP-KIF5C transfectant).
EYFP-KIF5C-(1-335) expressed alone showed a diffuse distribution throughout the whole cell, including in 55% of transfected cells (16 of 29 cells), and the cell nucleus (Fig. 7B). In addition, in ϳ24% of transfected cells (7 of 29 cells), KIF5C-(1-335) was associated with filamentous structures thought to be the microtubules, consistent with the existence of a microtubule binding site within the KIF5C-(1-335) (supplemental Fig.  S2). This distribution pattern was not changed by co-expression of EYFP-KIF5C-(1-335) with ECFP-GRIF-1 (Fig. 7, E-G). EYFP-KIF5C-(336 -957) expressed alone was localized as filamentous structures in the cell cytoplasm (Fig. 7J). This distribution was similar to that reported by Navone et al. (18) following overexpression in CV-1 monkey kidney endothelial cells of a vesicular stomatitis virus-tagged KHC construct that contained the KHC C-terminal portion of the ␣-helical coiled-coil rod and the C-terminal tail. The filamentous labeling found by Navone et al. (18) co-localized with microtubules and a micro- tubule binding site within the non-motor domain was identified (18). In the presence of ECFP-GRIF-1 the distribution pattern of EYFP-KIF5C-(336 -957) was changed. It was always co-localized with ECFP-GRIF-1-rich regions close to the nucleus (Fig. 7, M and O). This pattern is reminiscent of the localization of GRIF-1 with aggregated mitochondria as reported for GRIF-1-transfected HEK 293 cells (5). Indeed, when COS-7 cells were transfected with pECFP-GRIF-1, pEYFP-KIF5C-(336 -957), and pDsRed1Mito to label mitochondria, all three fluorescent moieties were co-localized close to the cell nucleus (Fig. 7, Q-V).
FRET Measurements-To test whether ECFP-GRIF-1 associated directly with fluorescently tagged KIF5C, FRET studies were carried out in transfected HEK 293 cells. FRET efficiencies were measured by acceptor photobleaching. It was necessary to carry out these measurements in HEK 293 cells rather than COS-7 cells, because in the latter, the expression level of the fluorescent constructs was relatively low thus exposure to the appropriate laser for cell imaging resulted in the photobleaching of either EYFP or ECFP. In HEK 293 cells, the expression of the constructs was higher due to both a higher transfection efficiency and greater protein expression thus circumventing problems due to the photobleaching. Note that similar subcellular distribution profiles were found for the expression of single and pairwise GRIF-1/KIF5C combinations for both COS-7 cells and HEK 293 cells, i.e. GRIF-1/KIF5C and GRIF-1/KIF5C-(336 -957) both co-localized.

GRIF-1, KIF5C, and KLC Interactions
To investigate whether GRIF-1 associates with the assembled heavy chain plus light chain tetrameric kinesin complex, immunoprecipitation and confocal imaging experiments were carried out on cells co-transfected with GRIF-1, KIF5C, and KLC clones. The results are shown in Fig. 9.
Co-localization Studies-The same combinations of clones as described for the co-immunoprecipitation studies were used for the transfection of COS-7 cells. Representative confocal microscopy images are shown in Fig. 9. Points to note are: (i) overexpression of GRIF-1 did not change the distribution pattern of KLC and no areas of co-localization were found (Fig. 9, D-H); (ii) EYFP-KIF5C and KLC mostly co-localized within the cell cytoplasm (Fig. 9, I-M); (iii) it was noted that in the majority of the EYFP-KIF5C plus KLC double transfectants, no fluorescence was detected at the tips of cell extensions such as seen in Fig. 6E; (iv) in the triple transfections, co-localization of EYFP-KIF5C, KLC, and ECFP-GRIF-1 was seen both in the cell cytoplasm and later, as for both EYFP-KIF5C alone and EYFP-KIF5C plus ECFP-GRIF-1, concentrated at focal points at the tips of cell processes (Fig. 9, N-S).

DISCUSSION
We have previously shown that GRIF-1 associates with the molecular motor, kinesin (5). In this report, we have used four experimental approaches to map the GRIF-1 binding domain of the prototypic kinesin-1, KIF5C. Immunoprecipitations and yeast two-hybrid studies both mapped the GRIF-1 binding region of KIF5C to the C3 cargo binding domain. Further, in the yeast two-hybrid interaction assays it was found that the C3 cargo binding domain specifically associated with GRIF-1-(124 -283) in agreement with the previous mapping of the kinesin binding domain of GRIF-1 (5). Overexpression of EYFP-and ECFP-tagged KIF5C and GRIF-1, respectively, in either HEK 293 cells or COS-7 cells showed that GRIF-1 co-localized with KIF5C (Fig. 6, H-J) and in refined studies, with the KIF5C non-motor domain, KIF5C-(336 -957). FRET experiments established that the interaction between the two proteins was direct. The fact that a significant FRET value was found between the C-terminal-tagged KIF5C and ECFP-GRIF-1, the C-terminaltagged KIF5C-(336 -957) and GRIF-1, but not between the N-terminal-tagged KIF5C nor the motor domain, EYFP-KIF5C-(1-335), confirmed the direct association between the two proteins at the KIF5C C-terminal and GRIF-1 N-ter- minal regions. Finally, GRIF-1 was shown to form a ternary complex with KHC and KLC. These findings substantiate a role for GRIF-1 as an adaptor protein linking kinesin to its cargo in the anterograde trafficking processes in neurons. A schematic model summarizing the possible interactions between GRIF-1 and kinesin is shown in Fig. 10.
The concept of the selective transport of proteins and/or organelles via adaptor proteins linking kinesin motor proteins to their cargoes is an emerging feature of trafficking mechanisms in neurons. As described in the introduction, there are now several examples of such kinesin-associated adaptors and their cargoes, i.e. mitochondria and syntaxin-1-containing vesicles are attached to the KHC cargo binding domain by the adaptor protein, syntabulin, for their transport to synapses (9); threonine residues of protein substrates. This O-glycosylation post-translational modification occurs in the cell cytoplasm, and it is thought to regulate protein function and to have a reciprocal relationship with post-translational modification and regulation via phosphorylation, the so-called "YinOYang" relationship (29). The YingOYang neural network predictions for O-␤-GlcNAc attachment sites in eukaryotic protein sequences (www.cbs. dtu.dk/services/YinOYang) predicts several sites for O-glycosylation within the non-kinesin binding GRIF-1 C-terminal domain. It is known that GRIF-1 and OIP106 are O-glycosylated in vivo (2), although the residues that are actually O-glycosylated have not been identified. Thus perhaps the association of GRIF-1 and OIP106 with OGT is to regulate GRIF-1-KHC interactions rather than OGT per se being a cargo. In summary, the work described herein maps the GRIF-1 binding domain of KIF5C and shows that GRIF-1 forms a ternary complex with KIF5C and KLC that results in the aggregation of mitochondria. These findings consolidate the role of GRIF-1 as an adaptor protein involved in motor-dependent anterograde transport. Defects in these transport mechanisms, especially in the transport of mitochondria and GABA A receptors, purported to be GRIF-1 cargoes, may contribute to the pathology of neurodegenerative diseases such as Alzheimer disease, amyotrophic lateral sclerosis, hypertonia, and hereditary spastic paraplegia (30).