Postsynaptic density protein 95 (PSD-95) is transported by KIF5 to dendritic regions

Postsynaptic density protein 95 (PSD-95) is a pivotal postsynaptic scaffolding protein in excitatory neurons. Although the transport and regulation of PSD-95 in synaptic regions is well understood, dendritic transport of PSD-95 before synaptic localization still remains to be clarified. To evaluate the role of KIF5, conventional kinesin, in the dendritic transport of PSD-95 protein, we expressed a transport defective form of KIF5A (ΔMD) that does not contain the N-terminal motor domain. Expression of ΔMD significantly decreased PSD-95 level in the dendrites. Consistently, KIF5 was associated with PSD-95 in in vitro and in vivo assays. This interaction was mediated by the C-terminal tail regions of KIF5A and the third PDZ domain of PSD-95. Additionally, the ADPDZ3 (the association domain of NMDA receptor and PDZ3 domain) expression significantly reduced the levels of PSD-95, glutamate receptor 1 (GluA1) in dendrites. The association between PSD-95 and KIF5A was dose-dependent on Staufen protein, suggesting that the Staufen plays a role as a regulatory role in the association. Taken together, our data suggest a new mechanism for dendritic transport of the AMPA receptor-PSD-95.

KIF proteins transport a various molecules, including proteins, synaptic vesicles, and mitochondria along the microtubules cytoskeleton of an axon or dendrites to synaptic regions [16]. In particular, KIF5, which belongs to the recently classified kinesin-1 family, consists of three isoforms (A, B and C) [17]. It transports ribonucleoprotein complexes, synaptic vesicles, mitochondria, AMPA receptor vesicles, tyrosine receptor kinase B (TrkB)-containing vesicles [18], and γ-aminobutyric acid (GABA A ) receptor vesicles [19] in neurons, as well as slowly transported cargo proteins in axons [20]. KIF5 localizes AMPA receptor vesicles to the postsynaptic regions interacting with glutamate receptor interacting protein 1 (GRIP1)-a scaffolding protein similar to a MAGUK and a member of the PSD-95/SAP90/discs large homology (DLG)/zona occludens (ZO)-1 (PDZ)domain proteins, functioning as a synaptic scaffolding protein [21,22]. Several studies have reported that MAGUKs are transported to membrane regions by KIF13B [23,24].
In the present study, we revealed that KIF5 as a motor protein involved in PSD-95 dendritic transport. The Cterminal tail region of KIF5A was associated with the PDZ3 domain of PSD-95. The expression of the ADPDZ3 domain, which includes an NMDA receptorassociated domain (AD), significantly decreased levels of PSD-95 and surface glutamate receptor 1 (GluA1) at the postsynaptic site. Finally, we found that the KIF5A-PSD-95 complex colocalized with GluA1-immuopositive particles in dendritic regions, indicating that KIF5A mediates the transport of both PSD-95 and GluA1containing vesicles. Thus, we suggest that PSD-95 works as both a scaffolding protein in the excitatory synapses and an adaptor between a cargo and motor proteins.

Expression of a dominant-negative form of KIF5A reduces level of PSD-95 in dendrites
To examine the relevance of kinesin motor protein to dendritic transport of PSD-95, we expressed either green fluorescent protein (GFP), GFP-tagged wild type (WT) KIF5A (which is enriched in neurons) [17], or a dominant-negative mutant of KIF5A lacking the N-terminal motor domain (ΔMD) [25] in cultured hippocampal neurons. We then examined PSD-95 particles in dendrites. Consistent with our previous results [26], ΔMD expression significantly reduced the number and average size of PSD-95 particles by 78.48 and 61.71% (Fig. 1), respectively, indicating that inhibition of KIF5A functions reduces PSD-95 levels in dendrites. Interestingly, expression of the WT did not induce any significant change.

PSD-95 colocalizes with KIF5
In order to specify which isoform of KIF5 interacts with the PSD-95, we examined an association between isoform of KIF5 and PSD-95. The results showed that the PSD-95 interacts with all isoforms of KIF5s (Additional file 1: Figure S1). We examined the localization of PSD-95 and KIF5 in cultured neurons. As shown in Fig. 2a and b, a significant number of puncta immunopositive for KIF5 were colocalized with puncta of PSD-95 (PSD-95/KIF5: 53.17% ± 3.86%) and many puncta immunopositive for PSD-95 were colocalized with puncta of KIF5 (KIF5/PSD-95: 62.55% ± 1.69%). To further visualize the colocalization of two proteins in vivo, we performed a proximity ligation assay (PLA), which also indicated an association between the two proteins. The results of PLA showed that endogenous KIF5 and PSD-95 interacts in dendrites (Fig.  2c). Supporting this result, our immunoprecipitation analysis using anti-KIF5 or anti-PSD-95 antibody and rat brain lysates also showed interaction between endogenous KIF5 and PSD-95 (Fig. 2d). Taken together, these results show that PSD-95 may interact with KIF5 in neurons. Although all KIF5 members had showed the interaction with PSD-95 (Additional file 1: Figure S1), we decided to focus on the role of KIF5A, having pan-neuronal expression pattern [17], to study detailed mechanisms of the interaction and dendritic transport.
Interaction between PSD-95 and KIF5A requires the PDZ3 domain of PSD-95 In order to identify which domains of PSD-95 and KIF5A are required for this interaction, we constructed a series of truncated mutants of PSD-95 and KIF5A. First, we examined whether FLAGtagged full-length KIF5A could interact with hemagglutinin (HA)-tagged truncated PSD-95 mutants. As shown in Fig. 3a and c, the third PDZ domain (PDZ3) was required for interaction with KIF5A. Next, we examined whether HA-tagged fulllength PSD-95 could interact with FLAG-tagged truncated KIF5A mutants. Consistent with previous reports [27,28], the results indicated that the tail region of KIF5A was required for the interaction (Fig.  3d). To identify the necessary regions more closely, we constructed plasmids to express either the PDZassociated domain of NMDA receptors (AD) or PDZ3 and examined the interaction with full-length KIF5A. The PDZ3 domain was sufficient for the interaction with KIF5A, although this interaction was weaker than that with ADPDZ3 ( Fig. 4). Thus, in the following experiments, we used ADPDZ3 to block the interaction between the two proteins. Our data indicate that the tail region of KIF5A and PDZ3 domain of PSD-95 are required for the interaction between KIF5A and PSD-95.

ADPDZ3 expression reduces PSD-95 level in dendrites
Because ADPDZ3 was required for the interaction between KIF5A and PSD-95 ( Fig. 4), we examined whether ADPDZ3 expression affects the dendritic level of PSD-95. The cultured hippocampal neurons were infected with Sindbis viruses encoding either GFP or GFP-ADPDZ3 and subjected to immunostaining using monoclonal anti-PSD-95 antibody or polyclonal anti-synapsin I antibody. The monoclonal PSD-95 antibody did not detect overexpressed ADPDZ3 particles (data not shown). ADPDZ3 domain expression significantly reduced the number of PSD-95 particles, but not synapsin I particles (Fig. 5a−c), suggesting that inhibition of the interaction between PSD-95 and KIF5A blocks PSD-95 dendritic localization. This reduction was more dramatic in distal dendritic regions than in proximal regions (Additional file 2: Figure S2). Accordingly, ADPDZ3 expression resulted in a significant reduction of synapse number, possibly due to the reduced dendritic transport of PSD-95 (Fig. 5a, d).

ADPDZ3 expression reduces surface GluA1 level in dendrites
Since expression of the ADPDZ3 domain reduced the level of PSD-95-a major scaffolding protein in the synapses of excitatory neurons [1,2,4,29], it is possible that expression of AMPA receptor at postsynaptic membranes is also affected. To test this idea, we examined whether ADPDZ3 expression also reduces the level of surface AMPA receptors. After ADPDZ3 expression, surface GluA1 was immunostained and evaluated by imaging analysis. Consistent with the results of our PSD-95 assay, ADPDZ3 expression significantly reduced the number of surface GluA1 particles (Fig. 6), supporting the idea that expression of AMPA receptor at synaptic membranes is dependent on KIF5A-mediated transport of PSD-95 to synaptic regions.

Staufen modulates the association between PSD-95 and KIF5A
Previous studies demonstrated that AMPA receptor vesicles are transported to the dendrites by KIF5 through the interaction between glutamate receptor interacting protein 1 (GRIP1) and kinesin heavy chain [22,28]. Thus, we investigated whether KIF5 − associated with PSD-95 interacts with AMPA receptor vesicles. GFP-tagged PSD-95 was expressed in the cultured hippocampal neurons and immunostained using anti-KIF5 and GluA1 antibodies to analyze colocalization among PSD-95, KIF5, and GluA1. [ΔMD]: 78.48% ± 4.23%, n = 29 dendrites, 3287 μm; Kruskal-Wallis test: P < 0.0001; Dunn's multiple comparison test: ** P < 0.01, ns: not significant). c ΔMD expression significantly reduces the average size of PSD-95 particles (GFP: 100.00% ± 5.83%, n = 31 dendrites; KIF5A [WT]: 88.39% ± 4.27%, n = 31 dendrites; KIF5A [ΔMD]: 61.71% ± 4.00%, n = 29 dendrites; Kruskal-Wallis test: P < 0.0001; Dunn's multiple comparison test: *** P < 0.001, ns: not significant). N values denote the dendritic number from n neurons As shown in Fig. 7a and b, we found colocalization among these three proteins in dendrites, suggesting that AMPA receptor vesicles are transported to the dendrites by PSD-95 − KIF5A complexes. Next, we explored to identify whether any other linker or adaptor proteins are involved in the interaction between KIF5 and PSD-95. Because a previous study showed that Staufen functions as a linker protein for RNA granules, which are known as a cargo of KIF5 [30], and synaptic localization of PSD-95 depends on the expression or availability of Staufen [31], we examined whether Staufen expression modulates the association between PSD-95 with KIF5A. We gradually increased an amount of Myc-tagged Staufen to the cells expressing FLAG-tagged KIF5A and HA-tagged PSD-95, and measured interactions between KIF5A and PSD-95 using immunoprecipitation. Our results showed that Staufen expression increased the interaction in dosedependent manner (Fig. 7c, d), indicating that Staufen functions as a linker or adaptor between cargoes and KIF5.

Discussion
In the present study, we provide evidence that KIF5 (kinesin-1 family), also known as conventional kinesin, is a motor protein for PSD-95 dendritic transport. A previous study [15] reported that KIF1Bα (kinesin-3 family) Results of co-localization analysis. 53.17% ± 3.86% (n = 19 dendrites, 1354 μm) of KIF5immunopositve puncta are colocalized with PSD-95-immunopositve puncta and 62.55% ± 1.69% (n = 19 dendrites, 1354 μm) of PSD-95immunopositive puncta are colocalized with KIF5-immunopositve puncta. c Results of proximity ligation assay. Cultured neurons were infected with Sindbis viruses encoding GFP and subjected to PLA. Red dots in PLA indicate an interaction between the two proteins. Scale bar: 20 μm. d Results of IP assays using rat brain lysates. In total 500 μg of rat brain lysate was used in the IP assays using 3 μg monoclonal anti-KIF5 antibodies or polyclonal PSD-95 antibodies. These were analyzed by Western blotting using the antibody indicated. Asterisks indicate interaction bands in the Western blot assays associates with the C-terminal regions of PSD-95, while another [23] reported KIF13B (GAKIN) interacted with PSD-95 in epithelial cells. Lin et al. (2012) showed that KIF3A (kinesin-2 family) is involved in GluA2 trafficking in combination with GRIP 1 and PSD-95 in retina cells [32]. In the line with this idea, our results showed that inhibiting KIF5A by expressing a dominant-negative form (ΔMD) did not completely block PSD-95 localization (Fig.  1). In addition to KIF5, we also identified interaction of PSD-95 with KIF3A (data not presented). These results indicate that multiple motor proteins are involved in PSD-95 transport in various neuron types. Despite their somewhat overlapping expression level, each KIF displays a distinct expression pattern in the brain [33]. KIF5A is highly enriched in neurons of the cortex and the hippocampus [17,25,33], KIF1B in motor neurons in the medulla oblongata and the spinal cord [34], and KIF3A in granular cells of the cerebellum [35], while KIF13A in non-nerve tissues [1,24]. It is thus possible that dendritic transport of scaffolding proteins such as PSD-95 is regulated by distinct KIF isoforms in different brain areas.
Many previous studies have indicated that PSD-95 work as an adaptor between a motor protein and receptor−containing vesicle cargoes. For example, Mint1/X11 (mLin-10)-a PDZ domain−containing protein-interlinks NMDA receptor−containing vesicles and KIF17, which is a member of kinesin-2 family [21]. Glutamate [NMDA] receptor subunit 1 (GluN1) and GluN2 − containing vesicles are transported to dendrites with SAP97 and calcium/calmodulin-dependent serine protein kinase (CASK), while PDZ domain−containing MAGUKs are transported by KIF17 [36]. In addition, by directly interacting with GRIP1, which is another PDZ domain−containing scaffolding protein, KIF5C is reported to transport GluA2 − containing vesicles to dendrites [28]. Thus, GRIP1 interlinks N-cadherin and GluA2containing vesicles, transporting them into the dendrites [22]. Interestingly, Huntingtin−associated protein 1 (HAP 1) works as an adaptor molecule for the dendritic transport of GABA A receptor-containing vesicles in the inhibitory synapses [37]. Scaffolding and motor proteins carry out dendritic transport and synaptic localization of transmembrane proteins, such as receptors [38]. Our GRIP 1 directly interacts with KIF5 through PDZ6 and PDZ7 domain, functioning as an adaptor between KIF5A and GluA2-containing vesicles [28]. The protein mLin-10 also directly interacts with KIF17 via the PDZ1 domain [21]. It is likely that PSD-95 directly interacts with KIF5A, although we did not examine this possibility in the study. However, a previous study identified a putative PDZ interaction motif (class I: Ser/Thr − X − Val, S/TXV) [39] in the tail regions of KIF5A, and this PDZ interaction motif was only found in the KIF5A isoform [17]. The present study corroborated previous data [21,28], showing that PSD-95 binding to KIF5 may steer KIF5A to dendrites, as occurs when an adaptor binds to a motor protein. Another previous study suggested that PSD-95 is associated with Staufen in synaptic regions [31]. Staufen also works as an adaptor protein for KIF5 cargoes [30]. Consistent with these studies, in the present study, Staufen expression increased the association of PSD-95 with KIF5 ( Fig. 7c  and d), indicating that Staufen might modulate the PSD-95-KIF5 complex. Further studies should investigate the detailed molecular configuration of scaffolding protein−motor protein transport complexes modulated by Staufen.
Motor protein expression increases the levels of the corresponding cargo protein or associated protein in dendrites [22]. In the present experiment, even though the dominant mutant form of KIF5A (ΔMD) significantly decreased the level of PSD-95 in dendrites, KIF5A (WT) did not increase PSD-95 transport (Fig. 1). Considering for multiple motors to be involved in the transport of PSD-95, the role of single protein might not be considerable. Alternatively, neuronal activity may be required to increase motor activity. Indeed, increases in expression levels of cargo proteins require neuronal activity [26]. The present study has suggested a new mechanism in the dendritic transport of PSD-95 and receptor-containing vesicles in glutamatergic synapses (Fig. 7e).

Immunocytochemistry and proximity ligation assay
For the immunocytochemistry, cultures were fixed using a fixative (4% paraformaldehyde, 4% sucrose, pH 7.2) and permeabilized using PBT (0.1% TritonX-100, 0.1% BSA in PBS). In the case of surface GluA1 immunocytochemistry, no permeabilization step was performed. The cultures were pretreated using the preblock solution (2% BSA, 0.08 TritonX-100 in PBS) for 1 h and each antibody was directly added to the preblock solution for 2 h. For PLA using Duolink® In Situ-Fluorescence (Sigma-Aldrich), the cultures were infected with Sindbis viruses encoding GFP to visualize whole dendritic structures and then fixed as described above; rabbit polyclonal anti-PSD-95 antibodies (Cell Signaling) and mouse monoclonal anti-KIF5 antibodies (See figure on previous page.) Fig. 7 Complexes of PSD-95-KIF5 colocalized with GluA1 particles in dendrites. Cultured hippocampal neurons were transfected with PSD-95-GFP constructs and incubated for days. The cultures were immunostained with monoclonal anti-PSD-95 antibody and polyclonal GluA1 antibody; they were subsequently stained with C3-conjugated anti-mouse IgG and Alexa−Fluor® 647 anti-rabbit IgG antibody. a Representative images of immunostaining in the first row. Each image was merged to show colocalization in the second row. A colocalized image of PSD-95 and KIF5 was collated with the image of GluA1. The colocalized points appeared white. Scale bar: 20 μm. b Boxed dendrites were enlarged to see the colocalization of GluA1 with the complex of PSD-95-KIF5A. c Staufen expression increased the association of PSD-95 and KIF5A. HA-PSD-95 and FLAG-KIF5A were cotransfected with 1, 2, or 3 μg of Myc-Staufen, or without Myc-Staufen as a control. After immunoprecipitation using anti-FLAG antibody or mouse IgG, immunoprecipitates were analyzed by Western blotting using anti-HA antibody. The lower panel shows expression of each group. d Quantified data of Western blot analyses (0 μg of Staufen: 100.0% ± 0.00%, n = 3; 1 μg of Staufen: 141.6% ± 16.96%, n = 3; 2 μg of Staufen: 192.3% ± 6.59%; 3 μg of Staufen; 274.4% ± 42.30%, n = 3). N values indicate the number of independent experiments. e Schematic diagram showing GluA1-containing vesicle transport mediated by PSD-95-KIF5A complex in dendrites. TARP: transmembrane AMPA receptor regulatory protein (Clone H2, Millipore) were then used. All procedures were performed according to the manufacturers' instructions. The nucleus of each neuron was stained with 4′,6-diamidino-2-phenylindole (DAPI, Sigma-Aldrich). Immunostaining and PLA were visualized using confocal microscopy (Zeiss 710; Carl Zeiss, Oberkochen, Germany).

Image analysis
Secondary or tertiary dendrites with a similar diameter were selected from acquired neuron images and straightened using a plugin of ImageJ program (ver 1.47; National Institute of Health, Bethesda, VA, USA). The images of straightened dendrites were transited to threshold images. The number and size of PSD-95 or GluA1 particles were measured using the particle analysis plugin. Colocalization was measured from the threshold images using colocalization plugins and represented using either Pearson's correlation coefficient (R − value) or a percentage. All image analyses were performed by blind experiment.

Statistical analysis
Normality of the data was assessed using either the Kolmogorov-Smirnov test or the D'Agostino and Pearson omnibus normality tests. If the data followed Gaussian distribution, a Student's t-test was performed to determine statistical significance between two groups, while analysis of variance (ANOVA) was performed among three or more groups, with Newman Keul's analysis used as a post hoc analysis. If the data did not follow Gaussian distributions, the nonparametric Mann-Whitney test was performed to determine statistical significance between two groups, while the Kruskal-Wallis test combined with Dunn's multiple comparison test was performed among three or more groups. All statistical analyses were performed using GraphPad prism (ver 5.02; GraphPad Software, San Diego, CA, USA).
Additional file 1: Figure S1. All isoforms of KIF5 interact with PSD-95. Cultured HEK cells were transfected with plasmids of HA-tagged PSD-95 and FLAG-tagged KIF5A or FLAG-tagged KIF5B, or FLAG-tagged KIF5C. The lysates were used for IP using monoclonal anti-FLAG antibody (2 μg, Clone M2, Sigma-Aldrich) and then the precipitates were analyzed by Western blotting assay using monoclonal anti-HA antibody (1:2000, Clone HA-7; Sigma-Aldrich). The bottom blots show expression of each protein used in immunoprecipitations.