Research Article

Kinesin Superfamily Motor Protein KIF17 and mLin-10 in NMDA Receptor-Containing Vesicle Transport

See allHide authors and affiliations

Science  09 Jun 2000:
Vol. 288, Issue 5472, pp. 1796-1802
DOI: 10.1126/science.288.5472.1796


Experiments with vesicles containingN-methyl-d-aspartate (NMDA) receptor 2B (NR2B subunit) show that they are transported along microtubules by KIF17, a neuron-specific molecular motor in neuronal dendrites. Selective transport is accomplished by direct interaction of the KIF17 tail with a PDZ domain of mLin-10 (Mint1/X11), which is a constituent of a large protein complex including mLin-2 (CASK), mLin-7 (MALS/Velis), and the NR2B subunit. This interaction, specific for a neurotransmitter receptor critically important for plasticity in the postsynaptic terminal, may be a regulatory point for synaptic plasticity and neuronal morphogenesis.

In mammalian neurons, neurotransmitter receptors such as glutamate receptors, including NMDA receptors, are sorted dynamically and precisely to the dendrites of the cell (1). Although putative anchoring, sorting, and signaling molecules have been colocalized with the receptors (2), it is not yet known how the receptors are sorted. Transport of molecules to specific regions in eukaryotic cells is accomplished by molecular motors (3). In neurons, various microtubule-associated motor proteins have been shown to transport organelles such as synaptic vesicle precursors and mitochondria to specific regions of the cell; however, the mechanisms by which each motor recognizes its specific cargo are not known (3). Here, we report that KIF17 (4), a neuron-specific microtubule-dependent molecular motor, binds directly and specifically to a PDZ domain (5) of mLin-10 (6) and transports the large protein complex containing the NR2B subunit, which forms the NMDA receptor with the NR1 subunit (7). This complex transports vesicles along microtubules in neurons such as hippocampal pyramidal neurons.

Identification of KIF17. Members of the kinesin superfamily (KIFs) support diverse transport systems in cells (3). We cloned KIF17, a neuron-specific motor (Fig. 1A) (8), to investigate the motors responsible for the sorting of various molecules within neurons. Osm-3 (9), a putative dendritic motor for odorant receptors in Caenorhabditis elegans, and KIF17 constitute a family (Fig. 1B). KIF17 is similar to Osm-3 in the head and tail domains (Fig. 1C) and has two putative stalk domains that form an α-helical coiled coil (Fig. 1D) (10); Osm-3, however, has only one stalk domain. Antibody raised against amino acids 505 to 707 of KIF17 (anti-KIF17) (11) recognized the native KIF17 protein in the brain as a single 170-kD band in SDS–polyacrylamide gel electrophoresis (PAGE) (Fig. 1E).

Figure 1

Identification of KIF17 as a homodimeric microtubule-dependent motor protein. (A) Full-length 1038–amino acid sequence of KIF17 (the sequence has been deposited in the DDBJ/EMBL/GenBank database, accession numberAB001424). The motor domain probe used to screen is indicated in boldface, the middle domain used to make antibodies is in bold italics, and the tail domain used for the two-hybrid screen is underlined. Single-letter abbreviations for amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr. (B) Relation between KIF17 and other known KIFs, as shown by a phylogenetic tree. (C) Protein matrix analysis between KIF17 and Osm-3 shows similarity of their head and tail domains. (D) The central region of KIF17 was predicted to form two stretches of an α-helical coiled-coil structure. (E) Immunoblot of native KIF17. Abbreviations: Br, brain; Liv, liver; block, peptide block with recombinant KIF17 as a control. (F) Motility assay of recombinant KIF17 with axonemes. Arrow shows a frayed plus-end of an axoneme-microtubule bundle.

Although KIFs could be monomeric, homodimeric, heterotrimeric, homotetrameric, or heterotetrameric (3), native KIF17 has a sedimentation coefficient of 3.0 S and a Stokes radius of 170 Å, which suggests that the molecular weight of the native holoenzyme is 215 kD, about twice that calculated from the sequence of the protein (116 kD). The migration of KIF17 in native PAGE was similar to that of the recombinant full-length KIF17 protein (12). Thus, KIF17 probably exists as a homodimer (13).

To measure the direction and velocity of the KIF17 motor activity, we assayed its motility with the use of recombinant KIF17. Recombinant KIF17 could slide an axoneme toward microtubule minus-ends, indicating that KIF17 is a microtubule plus-end–directed motor (Fig. 1F). The microtubule gliding assay showed that the average speed was 0.8 to 1.2 μm/s (14). Thus, KIF17 can act without a coenzyme and can mediate fast intracellular transport.

KIF17 is a dendrite-specific motor protein. KIF17 appeared to be brain-specific, present in abundance in the gray matter (especially in the hippocampus) but not in white matter such as the optic nerve (Fig. 2A) (15). Hence, KIF17 is likely to act in the somatodendritic region of neurons rather than in the axons. Indeed, brain sections immunostained with anti-KIF17 showed KIF17 in dendrites of pyramidal neurons but not in the axons or nuclei in cerebral cortex (Fig. 2B), hippocampus (Fig. 2C), and olfactory bulb (Fig. 2D). Double staining of cultured hippocampal neurons with anti-KIF17 and an axonal marker [monoclonal antibody to phosphorylated neurofilament H (pNFH mAb)] showed that KIF17 is a dendritic motor (Fig. 2E) (16). The related protein Osm-3 is also known as a molecular motor localized to the dendrites of specific neurons inC. elegans (9).

Figure 2

Localization of KIF17 in dendrites. (A) KIF17 immunoblots in various tissues. Br, brain; Lu, lung; Ht, heart; Li, liver; Sp, spleen; Kid, kidney; Mu, muscle; OB, olfactory bulb; CC, cerebral cortex; PU, putamen; Hip, hippocampus; Ce, cerebellum; Spi, spinal cord; CE, cauda equina; Sci, sciatic nerve; Opt, optic nerve. KIF5B (below each lane) was used as a control. (B) Cerebral cortex pyramidal layer triple-stained with anti-KIF17, pNFH mAb as a marker for axons, and TOTO-3 as a marker for DNA and RNA. Scale bar, 15 μm. (C) Hippocampus CA3 region triple-stained by anti-KIF17, anti-MAP2 as a marker for dendrites, and TOTO-3. Scale bar, 15 μm. The peptide-blocked staining of anti-KIF17 as a negative control is shown in the small inset at the top of the KIF17 image. (D) Olfactory bulb mitral cell layer triple-stained with anti-KIF17, pNFH mAb, and TOTO-3. Scale bar, 50 μm. (E) Cultured hippocampal neuron double-stained with anti-KIF17 and pNFH mAb. Scale bar, 15 μm.

KIF17 conveys a membranous cargo. To identify the cargo transported by KIF17, we investigated KIF17-associated structures by immunoisolation. The organelles purified along with KIF17 were clear membranous vesicles with a radius of 50 nm on average, as observed by electron microscopy (Fig. 3A) (17). These vesicles were different from those transported by KIFC2, a central nervous system–specific dendritic motor protein (18); nor were they identifiable as mitochondria, synaptic vesicles, dense core vesicles, or multivesicular bodies.

Figure 3

Cargo of KIF17. (A) Electron micrograph of the vesicles immunoisolated with anti-KIF17. Scale bar, 20 nm. Histogram shows the vesicle count per 100 beads. (B) Nature of KIF17 binding to the vesicles assayed under different wash conditions (ppt, precipitant; sup, supernatant).

We then examined the association between KIF17 and the vesicles biochemically. KIF17 molecules were released from the vesicles in the presence of high salt or high pH (Fig. 3B) (19); this result suggests that KIF17 is peripherally associated with the membrane and should require a targeting mechanism for specific cargo vesicles.

mLin-10 is the receptor for KIF17 on the membrane. We identified adaptor molecules by yeast two-hybrid screening, using the COOH-terminal region of KIF17 (20). The KIF17 tail bound to a COOH-terminal PDZ domain of mLin-10, a sorting protein. The binding partner of this PDZ domain of mLin-10, which shows a very high degree of conservation from C. elegans to human, is unknown (6).

We further investigated the specificity of the binding between many KIFs and mLin-10 using mutated preys. The binding was very specific between the KIF17 tail domain and the first PDZ domain of mLin-10. Disruption of the first PDZ domain diminished the binding with KIF17, whereas disruption of the second PDZ domain or deletion of the phospholipid interaction (PI) domain did not alter the interaction (Fig. 4A). A single amino acid change or a deletion at the COOH-terminal end of KIF17 eliminated this interaction (Fig. 4B). The tails of KIF1A (400 to end), KIF1B (830 to end), KIF5A (806 to end), KIF5B (808 to end), and KIF5C (930 to end) were tested for mLin-10 binding. Only that of KIF17 bound with mLin-10.

Figure 4

Direct interaction between KIF17 tail and mLin-10 PDZ domain. (A and B) Identification of the interaction domain of KIF17 and mLin-10 using a yeast two-hybrid system. The amino acid residues of mLin-10 and KIF17 that were fused to the pLexA DNA binding domain are represented by boxes, and the corresponding amino acid positions are indicated (+, bound to KIF17 and mLin-10; –, not bound to KIF17 and mLin-10). ELGF is the known interaction consensus sequence for PDZ with the COOH-terminal of the binding partner. ILGV (residues 668 to 671 of mLin-10) → AAAA represents an ELGF motif that was mutagenized by site-directed PCR to all alanines in order to disrupt the first PDZ motif. The QLGF (residues 760 to 763 of mLin-10) → AAAA clone is another mutant with its second PDZ motif disrupted. KIF17 delta 3 a.a. denotes constructs of KIF17 with the three COOH-terminal amino acids deleted. KIF17 EPL* → EPA* denotes a construct with its wild-type COOH-terminal sequence EPL stop codon mutated to EPA stop codon. EPL* → EAL* and EPL* → EAA* are mutants constructed in the same way. (C) Plasmon resonance analysis of interaction of purified recombinant KIF17 with purified recombinant mLin-10 or control BSA. (D) Recombinant KIF17 pulls down native mLin-10 from brain lysate. Abbreviations: Wild, KIF17 939–1038; delta3, KIF17 939–1035; delta10, KIF17 939–1028; and KIFC2, KIFC2 1–423. (E) Immunoprecipitation of mLin-10 complex from brain.

We further examined this binding with the use of surface plasmon resonance measurement (Fig. 4C). The direct KIF17–mLin-10 interaction was confirmed to be one with strong affinity (dissociation constantK d = 3.0 × 10–6 M). This binding was sufficiently strong to mediate sorting in vivo (21).

We investigated the interaction with a pull-down assay and immunoprecipitation. mLin-10 was pulled down with the recombinant KIF17, but not with KIFC2 or the tail amino acids–deleted KIF17 (Fig. 4D) (22). Also, mLin-10 was coimmunoprecipitated with KIF17, whereas other known KIFs (KIFC2, KIF1A, KIF3A/B, and KIF5A/B/C) were not detectable in the immunoprecipitate; this result further confirmed the specificity of the interaction (Fig. 4E) (23). Thus, it appears that mLin-10 is the specific binding partner of KIF17 in vivo.

NR2B sorting vesicle carried by KIF17–mLin-10 complex. In Lin-10–mutated C. elegans, the glutamate receptor GluR1 is mislocalized (6). mLin-10 exists in a protein complex with mLin-2 and mLin-7, which binds to the NMDA receptor subunit NR2B (24). To test whether mLin-2, mLin-7, and NR2B were cargo molecules of the KIF17–mLin-10 complex, we examined the pattern of developmental expression of the cargo proteins. The expression of the candidate proteins was concurrent with that of KIF17 (Fig. 5A), whereas conventional kinesin (KIF5B) showed a different expression pattern (25).

Figure 5

KIF17–mLin-10 complex binds to NR2B-containing vesicles in a specific manner. (A) Developmental changes in the expression of KIF17, mLin-10, mLin-2, mLin-7, NR2B, and KIF5B in brain. Numbers above the lanes indicate embryonic days 11 to 17, postnatal days 0 to 6, and postnatal weeks 1, 2, and 3. (B) The Nycodenz floating assay fractions. The thick line indicates the density of the fraction; the thin line represents the protein concentration of the fraction. Cytochrome oxidase (Cox) and caveolin indicate other membranous proteins with different peaks. (C) Immunoprecipitation from brain lysate or vesicle fractions. (D) Colocalization of KIF17 and NR2B on vesicles along microtubules. Note that vesicle size and microtubule thickness are enhanced by the optical limitation of resolution. (E) KIF17 tail peptide block analysis. Red boxes represent numbers of NR2B+ vesicles. Error bars denote SEM (n = 3).

KIF17, mLin-10, mLin-2, and mLin-7 on the vesicles floated to the same low-density membranous protein fractions as NR2B, but not to other membranous protein fractions containing caveolin or mitochondrial cytochrome oxidase. Free KIF17, mLin-10, mLin-2, and mLin-7 were recovered in the heavier fractions. The ratio of free protein to membrane-bound protein increased according to the predicted binding order of the proteins to the membrane: KIF17, mLin-10, mLin-2, mLin-7, and NR2B (Fig. 5B) (26).

Immunoprecipitation with anti-KIF17 (27) was detected by antibodies to mLin-10, mLin-2, mLin-7, and NR2B, but not by antibodies to PSD-95, which binds to NR2B at the postsynaptic density (2) (Fig. 5C). GABAAreceptor β2, glycine receptor, and substance P receptor were not detected (28). Immunoprecipitation with mLin-10 mAb also confirmed the interaction among KIF17, mLin-10, mLin-2, mLin-7, and NR2B (Fig. 4E). NR2B, mLin-10, mLin-2, and mLin-7 were detected in the isolated vesicles by anti-KIF17, but not by anti-KIF1A (Fig. 5C).

Because NR2B vesicles are transported transiently, the ratio of NR2B to other components of the complex is reduced in the immunoprecipitate. Some of the NR2B-containing vesicles attach to microtubules in a nucleotide-dependent manner, and KIF17 is located on these vesicles (Fig. 5D). This association of NR2B+ vesicles to microtubules was blocked by KIF17 tail peptides fused to glutathione S-transferase (GST-KIF17 939–1038) (Fig. 5E). This peptide does not block the microtubule-binding activity of the KIF17 motor domain (14), but binds native mLin-10 (22). Therefore, KIF17 tail peptide competes with the native KIF17 for mLin-10 binding and will decrease the number of NR2B-containing cargoes on the microtubules in this assay (29).

Thus, these proteins form a protein complex on the vesicle. We cannot exclude the possibility of another mechanism for recruitment of mLin-10 to the vesicle, but it would be reasonable to regard this mLin-10 → mLin-2 → mLin-7 → NR2B → NR1 cascade as a good candidate vesicle-binding mechanism for KIF17.

Colocalization of KIF17, mLin-10, and NR2B in the same dendrites. We immunostained brain sections and cultured hippocampal neurons. Double staining with anti-KIF17 and NR1 mab, anti-NR2B, and anti–mLin-10 revealed colocalization of these proteins in the same dendrites (Fig. 6) (30).

Figure 6

KIF17 colocalizes with mLin-10, NR2B, and NR1 in dendrites of neurons. (A) Coronal section of cerebral cortex double-stained with anti-KIF17 and NR1 mAb. (B) Coronal section of cerebral cortex pepsin-treated and double-stained with anti-KIF17 and anti-NR2B (αGluRε No59). (C andD) Cultured hippocampal neuron double-stained with anti-KIF17 and anti–mLin-10. Scale bar, 15 μm.

Visualization of NR2B transport by KIF17. Native vesicles, purified from brain with anti-KIF17 beads, moved in a plus-end–directed manner on microtubules when incubated with full-length KIF17 (1–1038) (Fig. 7, A and B) (31). Recombinant KIF17 (1–938), lacking the mLin-10 interaction domain, could not move these vesicles on microtubules, and only Brownian movement of the vesicles in the buffer was observed. Because this deletion mutant has motor activity in the microtubule gliding assay (14) and vesicles were being released (31), it appears that the mLin-10 binding site is necessary for KIF17 to transport these vesicles. Direct fixation and immunodetection by fluorescence in light microscopy (Fig. 7C) and gold labeling in electron microscopy (Fig. 7D) revealed that NR2B is on these active vesicles. Thus, cargoes containing NR2B are transported by KIF17, and the mLin-10 binding tail is necessary for this function of KIF17.

Figure 7

Visualization of NR2B transport by KIF17. (A) Movement of the KIF17 cargo vesicle on axoneme shows plus end–directed motility. Arrow shows the vesicle, which touches on the microtubule and moves, then goes out. [A quicktime movie can be viewed at (B) Two sequences of time-lapse images of movement of the KIF17 cargo vesicles on microtubules. (C) Immunofluorescent detection of NR2B on the vesicles moved by KIF17 on microtubules. (D) Electron micrograph showing immunocytochemistry of the KIF17-bearing vesicles on a microtubule. NR2B is detected by gold particles (diameter 10 nm). Scale bar, 100 nm. (E) Model of the NR2B transporting machinery.

These data collectively suggest that KIF17, a neuron-specific molecular motor with microtubule plus-end–directed motility interacts directly with a mLin-10 PDZ domain, resulting in the transport of NR2B in dendrites. We propose this motor-cargo complex as the sorting machinery for NR2B.

  • * To whom correspondence should be addressed. E-mail: hirokawa{at}


Stay Connected to Science

Navigate This Article