Kinesin-1 Regulates Extrasynaptic Targeting of NMDARs and Neuronal Vulnerability Toward Excitotoxicity.

N-methyl-D-aspartate (NMDA) receptor (NMDAR) is highly compartmentalized in neurons, and its dysfunction has been implicated in various neuropsychiatric and neurodegenerative disorders. Recent failure to exploit NMDAR antagonization as a potential therapeutic target has driven the need to identify molecular mechanisms that regulate NMDAR compartmentalization. Here, we report that the reduction of Kif5b, the heavy chain of kinesin-1, protected neurons against NMDA-induced excitotoxicity and ischemia-provoked neurodegeneration. Direct binding of kinesin-1 to the GluN2B cytoplasmic tails regulated the levels of NMDAR at extrasynaptic sites and the subsequent influx of calcium mediated by extrasynaptic NMDAR by regulating the insertion of NMDARs into neuronal surface. Transient increase of Kif5b restored the surface levels of NMDAR and the decreased neuronal susceptibility to NMDA-induced excitotoxicity. The expression of Kif5b was repressed in cerebral ischemia preconditioning. Our findings reveal that kinesin-1 regulates extrasynaptic NMDAR targeting and signaling, and the reduction of kinesin-1 could be exploited to defer neurodegeneration.

Introduction

The prevalence of neuropsychiatric and neurodegenerative disorders is increasing and is putting increasing pressure on health systems. Despite the differences in the pathologies among various disorders, evidences suggest that N-methyl-D-aspartic acid (NMDA) receptor (NMDAR) dysfunction is one of the common causes (Paoletti et al., 2013, Zhou and Sheng, 2013). NMDARs belong to a group of glutamate-gated ion channels that are essential for mediating brain plasticity (Paoletti et al., 2013). These receptors have been known for decades to receive input from presynaptic neurons and convey these signals into different patterns of neuronal activity, which eventually lead to long-term changes that underlie higher cognitive functions (Paoletti et al., 2013). Pathophysiologically, the overactivation of NMDARs activates subsequent pro-death pathways such as calcium influx and the production of reactive oxygen species, which are the main causes of synaptic dysfunction and neuronal death underlying neurodegenerative disorders, like stroke and Alzheimer disease. Despite huge clinical potential, current attempts to translate NMDAR blockade into clinical applications have so far been unsatisfactory. The concern raised in inhibiting NMDARs might result from the complexity of NMDAR compartmentalization.

There are two distinct pools of NMDARs. Compared with the well-established role of the synaptic NMDARs in mediating synaptic plasticity (Paoletti et al., 2013, Thomas and Huganir, 2004), the understanding on the extrasynaptic NMDARs is elusive. It has been reported that extrasynaptic NMDAR is activated by excessively released and spillover glutamate, and consequently shuts off the pro-survival pathway elicited by synaptic NMDARs (Bading, 2013, Hardingham et al., 2002). The extrasynaptic NMDAR-induced excitotoxicity has been implicated in various neurodegenerative conditions, including stroke and Alzheimer disease (Bading, 2017, Molokanova et al., 2014, Okamoto et al., 2009, Sattler et al., 1999, Tu et al., 2010). As such, direct and overall inhibition of NMDARs might disrupt NMDAR signaling in all compartments, including the physiological NMDAR signaling that benefits the neurons (Bading, 2017, De Keyser et al., 1999, Sanacora et al., 2017). It is therefore urged to understand the mechanism that regulates the compartmentalization of NMDARs, which can be exploited to abate pathophysiological NMDAR signaling with minimal adverse physiological impacts.

Owing to their highly polarized structure, neurons rely heavily on intracellular transportation systems to maintain the highly compartmentalized signaling that integrates and processes information from different inputs and sub-neuronal compartments (Hirokawa et al., 2010, Terenzio et al., 2017). Kinesin-1 is a molecular motor complex consisting of two heavy and two light chains. One of the heavy chains, Kif5b, is composed of three functional domains: the head domain drives anterograde transport along microtubules powered by ATP hydrolysis, the stalk domain facilitates the dimerization of heavy chains, and the tail domain mediates cargo binding (Hirokawa et al., 2010, Verhey et al., 2011). Kinesin-1 with intact Kif5b is critical for development, as demonstrated by homozygous knockout of kif5b in mice, which causes embryonic death as early as 9.5 days in utero (Tanaka et al., 1998). Moreover, kinesin-1 dysfunction is implicated in the pathogenesis of various neurodegenerative disorders. For example, Kif5b-containing vast vesicles, an early hallmark of axonal transport defect, are found in the post-mortem brain of patients with Alzheimer's disease (Stokin et al., 2005), and reduced levels of kinesin heavy chains are found in the early stages of Parkinson disease, which precedes alteration of dopaminergic markers (Chu et al., 2012).

Here we identified kinesin-1 as a microtubule-dependent molecular motor that regulates the distribution and function of extrasynaptic NMDARs. Kinesin-1 binds with GluN2B NMDARs via their carboxyl tails. The reduction of kinesin-1 prevents extrasynaptic NMDAR targeting, inhibits calcium influx from extrasynaptic NMDARs, and protects neuron against NMDA-elicited excitotoxicity and ischemia-evoked neurodegeneration. Furthermore, the expression of kinesin genes, including kif5b, is reduced in the transcriptome of cerebral ischemia preconditioning, indicating the reduction of kinesin-1 as an intrinsic neuroprotective mechanism by regulation of the compartmentalization of signaling.

Results

Kif5b Binds with GluN2B-Containing NMDARs

To test the hypothesis that kinesin-1 regulates the compartmentalization of NMDARs, we first immunoprecipitated endogenous kinesin-1 from mouse brain lysate using antibodies against Kif5b and detected GluN1, GluN2A, and GluN2B NMDAR subunits in the precipitate (Figure 1A). Reciprocal immunoprecipitation using an antibody against GluN1 further confirmed the presence of this complex (Figure S1A). Moreover, Kif5b and GluN2B were found to share partial co-localization in both soma and dendrites as seen in labeled primary hippocampal neurons at 14 days in vitro (DIV) using corresponding antibodies (Figure 1B). These data revealed that kinesin-1 and NMDAR form complex in vivo.

Figure 1

Kif5b Directly Binds with GluN2B

(A) Representative western blot images of Kif5b or normal IgG immunoprecipitation from mouse brain lysate.

(B) Immunostaining of Kif5b and GluN2B in primary hippocampal neurons (14 DIV). Scale bar (white solid line), 10 μm. The arrows in the “merged” panel indicate the co-localization signals.

(C) Schematic figure of Kif5b fragment containing the motor domain (1–413 aa), the stalk domain (414–679 aa), the tail domain (679–963 aa), the microtubule sliding domain (892–915 aa), and the autoinhibitory domain (918–926 aa). The microtubule sliding domain is conserved from Drosophila melanogaster to Homo sapiens. The four positively charged amino acids are shown in red.

(D) Representative western blot image of Kif5b tail immunoprecipitated with GluN2B intermediate tail (1,040–1,261 aa) from 293T cells overexpressing the indicated constructs.

(E) Representative western blot images of GST-tagged Kif5b tail fragment (900–935 aa) and its mutants with pull down of GluN1 and GluN2B from mouse brain lysate.

(F) Representative western blot image of GST-tagged Kif5b tail directly binding with in vitro transcribed and translated GluN2B intermediate tails.

(G) Sequence homology between GluN2A (1,045–1,255 aa) and GluN2B (1,044–1,261 aa). Red asterisk indicates the identical amino acid between GluN2A and GluN2B. Blue plus indicates amino acide with the same charge. Yellow highlight indicates the GluN2B tail region required for binding with Kif5b.

(H) Representative western blot image of GST-Kif5b tail (850–963 aa) binding with GluN2A and GluN2B tail fragments, as indicated.

The Kif5b protein consists of the head domain that acts as the motor for moving along the microtubule, the coiled-coil stalk domain that forms a motor complex with other heavy chains, and the tail domain that binds to cargo. Two functional sites within the tail domain have been identified, a microtubule sliding site and an autoinhibitory site (Kaan et al., 2011, Wong and Rice, 2010). We generated a series of Kif5b truncations conjugated with a glutathione S-transferase (GST) tag in a pull-down experiment to investigate this interaction in detail (Figure 1C). We found that the Kif5b tail (850–915 amino acid [aa]) including the microtubule sliding site, which was outside the KLC-binding domain, mediated binding to NMDAR (Figures S1B and S1C). This mapping result was confirmed by the co-expression of FLAG-tagged Kif5b fragments and the intermediate tail (1,040–1,261 aa) of GluN2B in a 293T cell line and co-immunoprecipitation by FLAG (Figure 1D). It is worth noting that deletion of the microtubule sliding domain (892–915 aa, ΔMs), but not the autoinhibitiory domain (918–926 aa, ΔAi), largely abolished these interactions, suggesting that this domain was important in mediating Kif5b binding with NMDARs (Figure 1D). By examining this interaction in further detail, we identified four positively charged amino acids that are conserved across species within this region (Figure 1C) and wondered whether these positively charged amino acids are required for binding with NMDAR. Mutations of either of the two positively charged amino acids (907–909 aa, RSK to SSS or AAA; 913–915 aa, RRG to SSS or AAA) abolished this binding, whereas mutations of other amino acids (910–912 aa, NMA to SSS or AAA) did not cause any disruption (Figure 1E). Furthermore, these were direct interactions, because GST-tagged Kif5b tail fragments were able to bind with in vitro transcribed and translated GluN2B intermediate tails (1,040–1,261 aa) in a cell-free system (Figure 1F).

To further examine the specificity of Kif5b interaction with GluN2B, we overexpressed green fluorescent protein (GFP)-tagged GluN2A/GluN1 and GluN2B/GluN1 separately in 293T cells where there is no endogenous GluN2A or GluN2B and found that Kif5b C-terminal fragment (850–963 aa) pulled down substantial amount of GluN2B/GluN1, but not GluN2A/GluN1, indicating that Kif5b might preferably bind with GluN2B over GluN2A (Figure S1D). We mapped that GluN2B (1,040–1,261 aa) sufficed binding with Kif5b (Figure 1F) and therefore compared the C terminal of GluN2B (1,040–1,261 aa) with that of GluN2A. GluN2A (1,045–1,255 aa) shared some similarity with GluN2B (1,044–1,261 aa, Figure 1G), but this similarity was not sufficient for the GluN2A fragment (1,046–1,255 aa) to bind with Kif5b C terminal (850–963 aa, Figure 1H). To further validate this finding, we did a series of C-terminal truncations on GluN2B and identified that GluN2B 1,160–1,180 aa was required for binding with Kif5b, as revealed that Kif5b C-terminal fragment pulled down GluN2B fragment 1,040–1,180 aa, but not 1,040–1,160 aa (Figure 1H). The above data combined reveal a complex consisting of kinesin-1 and GluN2B-containing NMDARs, indicating a role of kinesin-1 in mediating the function of GluN2B NMDARs.

Kinesin-1 Regulates Extrasynaptic NMDAR Distribution and Function

Kinesin-1 regulates transportation of cargo along microtubules through direct binding to cargo or through various adaptor proteins. This motor-cargo interaction determines the subcellular distribution of cargo and thus regulates their functions (Hirokawa et al., 2010). It has been previously reported that GluN2B-containing NMDAR is conveyed by Kif17 (Setou et al., 2000, Yin et al., 2011). Interestingly, depletion of kif17 did not completely abolish the movement of GluN2B in vivo (Yin et al., 2011), raising the possibility that multiple transport systems might contribute to regulate GluN2B-containing NMDAR trafficking. To examine the functional impact of kinesin-1 on GluN2B in vivo, we utilized a mouse model with heterozygous knockout of kif5b genes, as the homozygous knockout of kif5b causes embryonic lethality (Tanaka et al., 1998). As expected, besides the reduced levels of Kif5b, none of the other tested kinesins was found altered, including Kif5a, Kif5c, Kif17, and Kif3 (Kif5b, p = 0.0013, t = 4.4294, df = 10; Kif5a, p = 0.5934, t = 0.5535, df = 9; Kif5c, p = 0.2031, t = 1.3619, df = 10; Kif17, p = 0.6209, t = 0.5103, df = 10; Kif3, p > 0.9999, t = 0, df = 10; unpaired t test; Figure S2A). Kif5b hippocampus displayed reduced levels of the kinesin-1-GluN2B complex as revealed by immunoprecipitation of Kif5b (p = 0.0005, t = 5.603, df = 8; unpaired t test; Figure S2B). Because the level of Kif17 was not altered in our kif5b mice (Kif17, p = 0.6209, t = 0.5103, df = 10; unpaired t test; Figure S2A), it appears that kinesin-1 reduction does not simultaneously affect Kif17-mediated transport. As such, we hypothesized that Kif5b might serve as a motor necessary for GluN2B-containing NMDAR trafficking. The western blot analysis showed no differences in the overall levels of NMDAR subunits GluN1, GluN2A, and GluN2B in hippocampal lysate in wild-type and kif5b mice (GluN1, p = 0.5405, t = 0.6362, df = 9; GluN2A, p = 0.4666, t = 0.7602, df = 9; GluN2B, p = 0.3506, t = 0.9845, df = 9; unpaired t test; Figure S2C). However, surface biotinylation of hippocampal slices from kif5b mice showed significantly reduced levels of surface GluN2B-containing NMDARs (GluN1, p = 0.0005, t = 4.672, df = 12; GluN2A, p = 0.5826, t = 0.5727, df = 8; GluN2B, p < 0.0001, t = 6.647, df = 12; unpaired t test; Figure S2D). This was further confirmed by surface labeling of primary neurons transfected with GFP-tagged GluN2B, which showed both surface GFP puncta density and intensity were reduced in kif5b neurons at 14–16 DIV (Figures 2A–2C; C, left, density, p < 0.0001, D = 0.6448; right, intensity, p < 0.0001, D = 0.5961; Kolmogorov-Smirnov test). To test whether kinesin-1 affects dynamic turnover of GluN2B-containing NMDAR, we performed a surface antibody labeling assay on neuronal culture from kif5b mice and their control littermates. We first saturated the GFP-GluN2B transfected neurons with anti-GFP antibody raised in chicken. After thorough wash, the neurons were returned to 37°C for additional 2 h to allow GluN2B turnover. The neurons were further incubated with anti-GFP antibody raised in rabbit to distinguish the original surface GluN2B from the newly inserted GluN2B (Figure 2D). Compared with the control littermates, there was a significantly reduced amount of newly inserted GluN2B in the kif5b cultures (p < 0.0001, U = 22.5, Mann-Whitney U test; Figures 2E and 2F).

Figure 2

Kinesin-1 Regulates NMDAR Surface Expression

(A) Schematic figure showing antibody labeling the surface-GluN2B-containing NMDARs. Plasmid with extracellular GFP-tagged GluN2B was transfected into primary neuronal cultures at 14 DIV. After 24 h, neurons were incubated with anti-GFP antibody at 4°C for 1 h, followed by fixation and secondary antibody incubation.

(B and C) (B) Representative images and (C) quantification of surface GFP-GluN2B staining in primary neuronal cultures from kif5b mice (+/−) and their wild-type littermates (+/+). (C, left) Box plots show the number of puncta per 10 μm, (C, right) the curve plots show the relative intensity of each puncta (puncta number: n = 41 for wild-type and n = 39 for kif5b, five neurons were analyzed per genotype, Mann-Whitney U test; puncta intensity: six neurons were analyzed per genotype, at least 100 puncta were analyzed per neuron from three individual cultures, Kolmogorov-Smirnov test). Scale bar (white solid line), 20 μm.

(D) Schematic figure showing antibody labeling for the newly inserted GluN2B-containing NMDARs. After transfection, the neurons were saturated with chicken anti-GFP antibody for 10 min at room temperature and washed before returning to 37°C for additional 2 h. Then neurons were incubated with rabbit anti-GFP antibody for another 10 min at room temperature to label the newly inserted GluN2B. The neurons were washed, fixed, and permeabilized for secondary antibody labeling to distinguish between the newly inserted and the original GluN2B.

(E and F) (E) Representative images and (F) quantification of surface GFP-GluN2B staining in primary neuronal cultures from kif5b mice (+/−) and their wild-type littermates (+/+). (F) Box plot showed the number of puncta of newly inserted GluN2B (green channel) per 10 μm. (n = 15 per genotype; at least three neurons per genotype; for each neuron, at least three neurites were analyzed; Mann-Whitley U test). Scale bar (white solid line), 10 μm.

(G and H) (G) Representative images show surface staining of GFP-GluN2B puncta from neurons co-transfected with the indicated Kif5b constructs. Bar plots show (H, top) the average density of surface puncta and (H, bottom) the intensity of each puncta (H, top: n = 14 for FL, n = 13 for ΔMs, n = 19 for ΔAi, eight neurons analyzed for each group; H, bottom: n = 167 for FL, n = 128 for ΔMs, n = 137 for ΔAi, Kruskal-Wallis test followed by Dunn's post hoc test, 20 puncta/neuron analyzed for each group; three replicate experiments). Scale bar (white solid line), 20 μm. Values are presented as mean ± SEM, unless otherwise indicated (test as indicated; exact p value given if less than 0.05).

We further examined the effect of the identified domain on mediating the GluN2B-containing NMDAR distribution. The 3×FLAG-tagged full-length Kif5b, Kif5b with microtubule sliding site deletion, or Kif5b with autoinhibitory site deletion together with GFP-GluN2B were transfected into primary cultures from kif5b mice. Compared with neurons expressing Kif5b with microtubule sliding site deletion, both wild-type Kif5b and Kif5b with autoinhibitory site deletion showed increased levels of both GluN2B surface puncta density and puncta intensity (density, H = 27.96, p < 0.0001; FL versus ΔMs, p < 0.0001; FL versus ΔAi, p = 0.4242; ΔMs versus ΔAi, p < 0.0001; intensity, H = 217.0, p < 0.0001; FL versus ΔMs, p < 0.0001; FL versus ΔAi, p = 0.3565; ΔMs versus ΔAi, p < 0.0001; Kruskal-Wallis test followed by post hoc Dunn's multiple comparisons test; Figures 2G and 2H).

NMDAR acts as a calcium channel upon activation. The resulting calcium influx regulates synaptic strength (Bading, 2013, Volk et al., 2015) or causes the pro-death signaling cascade leading to neuronal death when it is overloaded (Bading, 2013, Tu et al., 2010). NMDAR-dependent calcium influx is associated with synaptic dysfunction and neuronal loss in both acute and chronic neurodegenerative conditions (Bading, 2013, Zhou and Sheng, 2013). To gain insight on the functional impact of kinesin-1 on NMDAR compartmentalization, we used Fluo-3 to quantify calcium influx. In kif5b neurons, we found that NMDA-induced calcium influx was significantly reduced (Genotype, F(1, 12,000) = 17,333, p < 0.0001; two-way ANOVA followed by Sidak post hoc test; peak, p < 0.0001, U = 68.00; Mann-Whitney U test; Figure S3A), possibly indicating NMDAR hypofunction or a more generalized defect in calcium homeostasis. To distinguish between these two possibilities, we depolarized the neurons with KCl (70 mM), which triggered a similar level of calcium influx in both genotypes (Genotype, F(1, 4,240) = 19.63, p < 0.0001; two-way ANOVA followed by Sidak post hoc test; peak, p = 0.1561, U = 289.0; Mann-Whitney U test; Figure S3B). This result reassured that the reduced calcium influx in kif5b was caused by NMDAR hypofunction.

On the neuron surface, NMDARs are distributed in postsynaptic densities (PSDs) as well as in extrasynaptic regions, and the compartmentalization determines their signaling and functions. Given the notion that GluN2B is predominately concentrated in the extrasynaptic regions in the adult mouse hippocampus (Tovar and Westbrook, 1999), we hypothesized that the kinesin-1-NMDAR complex would in a biased manner regulate the targeting and functioning of extrasynaptic NMDARs. To test this hypothesis, we immunostained surface GFP-GluN2B together with the PSD marker, PSD-95, to distinguish different NMDAR compartments at the same dendrite. Similar numbers of synaptic puncta containing GluN2B and PSD-95 were found in both wild-type and kif5b groups (p = 0.3767, t = 1.266, df = 58; two-way ANOVA followed by Sidak post hoc test), whereas we observed a significant reduction of surface GFP-GluN2B signals that were not co-localized with PSD-95 in kif5b neurons (p < 0.0001, t = 9.280, df = 58; two-way ANOVA followed by Sidak post hoc test; Figures 3A and 3B). These data showed that the surface targeting of extrasynaptic NMDARs was dampened by reducing kinesin-1 level. To examine the functional impact of kinesin-1 on extrasynaptic NMDAR, we monitored the calcium influx elicited by bicuculline, which increases neuronal activity by blocking a GABA A receptor, together with 4-aminopyridine (4-AP), a non-competitive NMDAR antagonist. We observed a calcium influx that was believed to be synapse specific. We then added MK-801, which targets NMDAR calcium channel opening, to block only the activated NMDAR (Hardingham et al., 2002). We did not detect any notable differences in calcium influx between the wild-type and kif5b neurons, suggesting that synaptic NMDAR-dependent calcium influx was not altered in kif5b neurons (p = 0.8232, U = 325.0; Mann-Whitney U test; Figures 3C and 3D). As MK-801 is demonstrated to irreversibly block extrasynaptic NMDAR, we re-stimulated the neurons after MK-801 blockade with NMDA to examine extrasynaptic NMDAR-dependent calcium influx, which showed a significant reduction in kif5b neurons (p < 0.0001, U = 14.0; Mann-Whitney U test; Figures 3C and 3D). Taken together, the above data showed that kinesin-1 binding with GluN2B mediates NMDAR compartmentalization by regulating the extrasynaptic targeting of NMDARs, and thereby regulates NMDA-elicited calcium influx. This further indicates a role of kinesin in determining cargo destination toward different neuronal sub-compartments.

Figure 3

Kinesin-1 Regulates Extrasynaptic NMDAR Surface Targeting and Function

(A and B) (A) Representative images and (B) quantification of surface GFP-GluN2B co-localized with PSD-95 (n = 16 for wild-type and n = 15 for kif5b, unpaired t test; six images per genotype; scale bar, 10 μm).

(C and D) (C) Time lapse and (D) peak response (box plot) of calcium influx by synaptic and extrasynaptic NMDAR stimulation of kif5b (+/−) and wild-type (+/+) neurons. Kif5b+/ (+/−) and wild-type (+/+) neurons evoked by bicuculline (Bic; 20 μM)/4-AP (4 mM) in the presence of nifedipine (Nif; 10 μM) and then blocked by MK-801 (10 μM), an irreversible NMDA receptor calcium channel blocker. The difference before and after MK-801 blockage was elicited by synaptic NMDARs. After 100 s of resting, NMDA and glycine were applied to activate the extrasynaptic NMDARs. The relative intensity [(F-F0)/F0] was calculated (n = 27 for wild-type and n = 25 for kif5b+/ from two replicate experiments; D, Mann-Whitney U test; scale bar, 20 μm). Values are presented as mean ± SEM, unless otherwise indicated (test as indicated; exact p value given if less than 0.05).

Kinesin-1 Reduction Attenuated NMDA-Elicited Excitotoxicity and Ischemia-Induced Neurodegeneration

Nitric oxide (NO) is one of the main downstream molecules generated from NMDAR overactivation and acts as one of the main mediators for NMDAR-induced excitotoxicity (Sattler et al., 1999). It is generated preferably from extrasynaptic NMDARs under certain neurodegenerative conditions (Molokanova et al., 2014). We used an NO-sensitive electrode to measure NO emissions from acutely prepared hippocampal slices from wild-type and Kif5b mice (Tjong et al., 2007). Hypoxia (N2) and L-arginine (L-Arg) induced significantly lower NO emissions in kif5b slices than in wild-type (L-Arg, p < 0.0001, t = 9.53, df = 9; N2, p < 0.0001, t = 15.53, df = 8; unpaired t test; Figure S4A). Activation of NMDARs by NMDA (50 μM) or L-glutamate (L-Glu; 4 μM) together with glycine (Gly; 20 μM) increased NO emissions in both genotypes, but the amount was significantly less in kif5b slices (L-Glu, p < 0.0001, t = 27.10, df = 17; NMDA, p < 0.0001, t = 29.26 df = 6; unpaired t test; Figures S4B and S4C). The NO emissions from both genotypes were greatly suppressed by NMDAR blocker MK-801 (100 μM) and NO synthase (NOS) inhibitor NG-Monomethyl-L-arginine (L-NMMA; 100 μM), indicating the NO emission was NMDAR and NOS dependent (Genotype × treatment, p < 0.0001, F(2,19) = 162.8; L-Glu versus L-Glu/L-NMMA, p < 0.0001, t = 28.97, df = 19; L-Glu versus L-Glu/MK-801, p < 0.0001, t = 30.68, df = 19; two-way ANOVA followed by Sidak post hoc test; Figure S4D). Furthermore, in slices with NMDARs blocked by MK-801 (10 μM), depolarization using KCl (50 mM) produced similar levels of NO emissions in both genotypes (KCl, p < 0.0001, t = 14.93, df = 8; KCl/MK, p = 0.1940, t = 1.559, df = 4; unpaired t test; Figure S4C). The reduced NO emission in kif5b slices might be caused by altered NO production machinery due to reduced kinesin-1, rather than by reduced NMDAR-mediated calcium influx. To exclude this possibility, we first performed a western blot analysis of neuronal NOS (nNOS) in wild-type and kif5b slices, which showed identical amounts in both genotypes (p = 0.3579, t = 0.9689, df = 9; unpaired t test; Figure S2C). Next, we treated wild-type and kif5b slices with A23187 (10 μM), which again showed similar NO emissions between genotypes (p = 0.9155, t = 0.1095, df = 8; unpaired t test; Figure S4C).

As it is known that extrasynaptic NMDAR activation would consequently lead to excitotoxicity, we tested whether the reduction of kinesin-1 would elicit neuroprotection against NMDA-induced excitotoxicity. The heterozygous knockout of kif5b gene had minimal impact on the viability of the neuronal culture, as the neuronal morphology was indistinguishable from that of its wild-type littermates. We treated the cortical cultures from both wild-type and kif5b mice with NMDA (25–200 μM) and a co-agonist glycine (10 μM) for 10 min, which sufficed to induce neuronal death. After NMDA washout, cortical neurons were cultured for an additional 24 h. Cell death was examined by propidium iodide (PI) staining or by lactate dehydrogenase (LDH) assay, followed by morphological analysis under phase contrast microscopy. The kif5b neurons treated with 100 μM NMDA and 10 μM glycine exhibited significantly fewer PI-positive neurons and less membrane leakage compared with the wild-type group (p = 0.0329, t = 2.758, df = 6; unpaired t test; Figures 4A and 4B; NMDA kif5b vs. kif5b, p < 0.0001, t = 8.575, df = 21; two-way ANOVA followed by Sidak post hoc test; Figure 4C). Morphologically, neurites in kif5b neurons were mostly intact indicating significantly less early neurodegeneration, whereas most neurons in the wild-type group showed fragmented neurites indicating synaptic loss leading to early neurodegeneration (p < 0.0001, t = 30.37, df = 6; unpaired t test; Figures 4A and 4B). Although NMDA elicited neurotoxicity in both genotypes in a dose-dependent manner, the neurotoxic effects were significantly lower in kif5b neurons at all tested concentrations (PI positive dose, F(2,17) = 22.34, p < 0.0001; degenerative dose, F(2,17) = 10.63, p = 0.0010; two-way ANOVA followed by Sidak post hoc test; Figures S4E and S4F). Moreover, treatment with memantine (50 μM), a potent NMDAR antagonist, significantly reduced NMDA-elicited neurotoxicity abolishing the differential neurotoxicity between kif5b and wild-type neurons (genotype × memantine, F(1,21) = 38.64, p < 0.0001; NMDA/memantine +/+ versus +/−, p = 0.9797, t = 0.7228, df = 21; two-way ANOVA followed by Sidak post hoc test; Figure 4C). Furthermore, selective inhibition of extrasynaptic NMDAR by a low dose of memantine (1 μM) (Okamoto et al., 2009) and GluN2B-specific inhibitor Ro25-6981 (5 μM) produced similar levels of NMDA-induced neurotoxicity in both wild-type and kif5b neurons (Mem, p = 0.4627, t = 0.7637, df = 10; Ro25, p = 0.7127, t = 0.3789, df = 10; unpaired t test; Figure S4G). As the control experiments, kainate (25 μM), a non-NMDA excitotoxin, produced similar neurotoxicity in kif5b and wild-type neurons (p = 0.4883, t = 0.7195, df = 10; unpaired t test; Figure S4H). Treatment with A23187, a calcium ionophore that increases the calcium-channel-independent level of intracellular calcium, elicited similar neurotoxicity in both genotypes (p = 0.8807, t = 0.1533, df = 12; unpaired t test; Figure S4H), indicating that kinesin-1-reduction-elicited neuroprotection is selective to NMDAR, at least in part, by preventing extrasynaptic NMDAR targeting and functioning.

Figure 4

Kinesin-1 Reduction Protects Neuron against NMDA-Induced Excitotoxicity

(A and B) (A) Representative images and (B, left) quantification of PI-positive neurons and (B, right) degenerative neurons of wild-type (+/+) and Kif5b (+/−) cultures (n = 4 per genotype, unpaired t test) after treatment with NMDA/glycine (Gly) for 10 min followed by washing for 24 h. The PI-positive fraction is the ratio of PI-positive neurons to PI-negative neurons, which indicates the portion of dead neurons. The degenerative fraction is the ratio of neurons with fragmented neurites to neurons with normal intact neurites, which indicates the portion of neurons with signs of early degeneration. Scale bar (white solid line), = 100 μm.

(C) Lactate dehydrogenase (LDH) leakage from NMDA-treated kif5b (+/−) and wild-type (+/+) neurons (n = 7 for wild-type and n = 6 for kif5b from three individual cultures, two-way ANOVA followed by Sidak post hoc test). The dead fraction is the ratio of LDH level in the drug-treated group to “all-kill” group.

(D and E) (D) Representative images and (E) quantification of dead and degenerative fractions of kif5b cultures transfected with GFP, GFP-2A-Kif5bFL, GFP-2A-Kif5bΔMT, and GFP-2A-Kif5bΔAi after treatment with NMDA/Gly or vehicle (Veh) for 10 min followed by washing for 24 h (n = 8 for each genotype from three individual cultures, unpaired t test). Scale bar (white solid line), 50 μm. Values are presented as mean ± SEM (test as indicated; exact p value given if less than 0.05).

To exclude the possibility that the reduced vulnerability to NMDA by reduced kinesin-1 level was due to secondary adaptation rather than the kinesin-1 reduction itself, we transiently overexpressed Kif5b with GFP separated by a 2A peptide in kif5b neurons. After NMDA treatment, kif5b neurons transfected with GFP had well-preserved morphology with robust neurites, whereas neurons transfected with the full-length Kif5b-2A-GFP displayed shrunken nuclei with significant loss of neurites (p < 0.0001, t = 10.59, df = 14; unpaired t test; Figures 4D and 4E). Moreover, we assessed the NMDA-elicited cell death in kif5b neurons expressing either Kif5b with microtubule sliding site deletion or Kif5b with autoinhibitory site deletion. Transfection of either Kif5b mutants did not lead to spontaneous neurodegeneration (Figures 4D and 4E), as revealed by the intact neuronal morphology. We found that only Kif5b with autoinhibitory site deletion had enhanced neuronal susceptibility to NMDA-elicited neurotoxicity, as shown by significantly more fragmented neurites (p < 0.0001, t = 9.079 df = 15; unpaired t test; Figures 4D and 4E). These results confirm that the neuroprotection observed in kif5b neurons was due to a direct kinesin-1-mediated cellular mechanism, possibly associated with intracellular transportation.

To examine if the kinesin-1 reduction would affect neuronal fate following noxious insult in vivo, we performed bilateral cerebral carotid artery occlusion (BCCAO) for 20 min on kif5b and wild-type mice. The kif5b mice were viable and fertile with no growth retardation (Genotype × time, male, F(5,18) = 0.05272, p = 0.9980, female, F(5,42) = 1.088, p = 0.3813; two-way ANOVA; Figure S5A) and no signs of spontaneous neurodegeneration, as seen by the similar gross hippocampal anatomy in both genotypes (Figure S5B). The induced transient global ischemia triggered robust neurodegeneration in the hippocampus (Tu et al., 2010). At 7 days after reperfusion, brain sections examined by TUNEL assay showed severe neuronal damage in the wild-type mice but only mild to moderate damage in the kif5b mice. There was a significant increase in TUNEL-positive signals detected in wild-type hippocampal CA1 and CA3 regions (CA1, p < 0.0001, t = 6.0483, df = 36; CA3, p < 0.0001, t = 10.7425, df = 22; unpaired t test; Figures S5C and S5D). Next, we performed middle cerebral artery occlusion (MCAO) for 120 min on kif5b and wild-type mice to induce focal cerebral ischemia (Lo et al., 2005, Tu et al., 2010). Consistent with BCCAO, we observed significantly improved neurological scores and reduced infarct volume in the kif5b mice (Genotype, F (1, 11) = 13.88, p = 0.0033; two-way repeated measures ANOVA followed by Sidak post hoc test; infarct volume, p = 0.0014, t = 4.256, df = 11; unpaired t test; neurological score, U = 6.000, p = 0.0408, Mann-Whitney rank-sum test; Figures 5A–5D).

Figure 5

Kinesin-1 Reduction Protects Brain against Ischemia-Evoked Neurodegeneration

(A–C) (A) Representative images and (B and C) quantification of 2,3,5-triphenyltetrazolium chloride (TTC) staining of wild-type and kif5b mice after transient focal cerebral ischemia. Infarct volume of (B) the indicated slices and (C) whole brain (n = 6 for wild-type and n = 7 for kif5b mice; B, two-way ANOVA followed by Sidak post hoc test; C, unpaired t test).

(D) Neurological scoring revealed kif5b (+/−) mice had significant improvement in motor function after 2-h ischemia and 24-h reperfusion when compared with their wild-type littermates (+/+). (Neurological score: 0, freely moving; 1, reduced mobility; 2, partial paralysis; 3, no movement) (Mann-Whitney rank-sum test).

(E–G) (E) Representative images and (F and G) quantification of TTC staining of kif5b and kif5b;Camk2a-Cre mice after MCAO. Infarct volume for indicated slices (F) and whole brain (G) (n = 6 for kif5b and kif5b;Camk2a-Cre, respectively; F, two-way ANOVA followed by Sidak post hoc test; G, unpaired t test). Values are presented as mean ± SEM, unless otherwise indicated (test as indicated; exact p value given if less than 0.05).

To examine the possibility that the observed neuroprotective phenotype in kif5b mice is directly due to kinesin-1 reduction rather than developmental adaptation in vivo, we generated a conditional knockout model by crossing kif5b mice with Camk2a-Cre mice by kif5b excision starting in the third postnatal week to mimic postnatal Kif5b reduction, which should minimize developmental alteration caused by kif5b deletion (Tsien et al., 1996). As Camk2a-Cre mediates recombination only in a subset of excitatory neurons in the forebrain, it was not surprising to observe a modest reduction of Kif5b protein level at 6 weeks of age in both kif5b;Camk2a-Cre and kif5b;Camk2a-Cre mice (F (2,6) = 4.925, p = 0.0543; one-way ANOVA; Figure S5E). However, immunohistochemistry for Kif5b in the CA1 region, where Camk2a-Cre has the most impact, revealed a robust reduction in Kif5b protein level in kif5b;Camk2a-Cre mice when compared with their wild-type littermates (Figure S5F). These data confirm that Kif5b was reduced at 6 weeks age in a subset of neurons in the conditional knockout model. Consistent with our observation in kif5b mice, the Camk2a-Cre-mediated conditional kif5b knockout mice showed robust resistance to MCAO-induced neurodegeneration (F, genotype, F (1, 10) = 16.20, p = 0.0024, two-way repeated measures ANOVA; g, p = 0.0051, t = 3.565 df = 10; unpaired t test; Figures 5E–5G). Moreover, we observed a significant increase in NeuN immunoreactivity in the CA1 region of kif5b conditional knockout mice after 2-h MCAO and 24-h reperfusion when compared with their wild-type littermates (p = 0.0022; Mann-Whitney test; Figures S5G and S5H). These combined results show that kinesin-1 reduction protects neurons against NMDAR-elicited excitotoxicity and ischemia-induced neurodegeneration.

Reduction of Kif5b Expression Associated with Suppression of Cell Death Pathways

Ischemia preconditioning (IPC) elicited neural-activity-dependent neuroprotection by altering the transcriptome to adapt to and resist subsequent noxious insult (Stenzel-Poore et al., 2003). As previous findings show a robust neuroprotective effect by kinesin-1 reduction, we hypothesized that the alteration of kinesin genes, particularly kinesin-1, might be a core mechanism underlying the IPC-elicited neuroprotection. To test this, we analyzed a microarray dataset (GEO:GSE32529) that profiled gene expressions at 3, 24, and 72 h after cerebral ischemia preconditioning (IPC) by a 15-min occlusion of middle cerebral artery in mice. Significantly altered genes at 3, 24, and 72 h after IPC were filtered and analyzed by Ingenuity Pathway Analysis platform (Qiagen) and compared with sham control groups (Figure 6A). By combining and comparing results from the “Diseases and Functions” analysis at the three indicated time points, we found that a set of genes relating to apoptosis and cell death were robustly enriched at 3 h after IPC, whereas this set of genes was suppressed at 24 and 72 h after IPC (Figure 6B). The finding of enhanced genes relating to apoptosis and cell death at 3 h after IPC was unexpected, and the overall trend supports the notion that brief exposure to IPC reprograms gene expression to protect the brain against lateral noxious insult. Besides the expected trends of inhibited cell death pathways or genes at the extended times after IPC, we unexpectedly identified suppressed genes related to neurodevelopment at over 3 to 72 h (Figure 6B). This raises the possibility that the same set of genes that regulate neurodevelopment also plays a key role in mediating neuronal adaptation to stress, including noxious insult. Kinesin genes plays a critical role in the development of cells and tissues such as neurons, muscles, chondrocytes, and osteoblasts (Hsu et al., 2011, Muhia et al., 2016, Padamsey et al., 2017, Qiu et al., 2012, Wang et al., 2013). We therefore investigated the global expression profile of kinesin genes at 3, 24, and 72 h after IPC in mice compared with the sham control group. The cumulative distribution of kinesin genes in the differentially expressed population exhibited a robust shift toward upregulation at 3 h, but not 24 h, after IPC (left, D = 0.2779, p = 0.0009, Kolmogorov-Smirnov test; right, D = 0.1387, p = 0.0891; Kolmogorov-Smirnov test; Figure S6A). In contrast, the cumulative distribution of kinesin genes exhibited a significant shift toward downregulation at 72 h compared with the overall gene expression profile (D = 0.1583, p = 0.0347; Kolmogorov-Smirnov test; Figure 6C). This became more apparent when comparing the cumulative expression profile of kinesin genes at 3 and 72 h, which showed a significant downregulation shift (D = 0.2785, p = 0.0165, Kolmogorov-Smirnov test; Figure 6D). The results of the cumulative expression profile of kinesin genes suggests a time-dependent downregulation shift after IPC, which might indicate a kinesin-dependent mechanism underlying IPC-induced adaptation and neuroprotection. To identify which kinesin is the key player, we extracted probes for kinesin genes that were robustly altered at 72 h after IPC, which revealed two probes for kif5b. Moreover, we found that kif5b and kif1b, but not kif3a, were robustly repressed at 72 h compared with 3 h after IPC (kif5b, p = 0.0313; kif1b, p = 0.0313; kif3a, p = 0.1250; Wilcoxon matched-pairs signed rank test; Figure S6B), suggesting that the repression of Kif5b gene was a candidate event regulating neuronal adaptation to further ischemia insult.

Figure 6

Transcriptome Analysis of Mouse Cortex after Ischemia Preconditioning Reveals that the Suppression of Cell Death Pathways Is Associated with Reduced kif5b Expression

(A) Schematic figure of the transcriptome analysis. The genes significantly altered at 3, 24, or 72 h after brief cerebral ischemia preconditioning (IPC) were analyzed by Ingenuity Pathway Analysis (IPA) platform. The expression of kinesin genes (kifs) were extracted and compared with the overall expression profile.

(B) IPA results of the pathways categorized by diseases and functions.

(C) The expression profiling of kinesin genes (kifs) were compared with the overall gene expression profile at 72 h after IPC normalized to 72 h in the sham control (Kolmogorov-Smirnov test).

(D) The expression profiling of kinesin genes at 3 and 72 h after IPC (Kolmogorov-Smirnov test).

(E and F) (E) Representative western blot images and (F) quantification of Kif5b, Kif17, and MAP2 in the hippocampus of wild-type mice after IPC or sham operation (n = 5 for the IPC group and n = 4 for the sham operation group, unpaired t test).

(G) Representative western blot image and (H) quantification of Kif5b, Kif17, and MAP2 in the hippocampus of wild-type mice with kainate treatment (KA, 12 mg/kg, 2 h) or vehicle (n = 9 for the KA group and n = 9 for the vehicle group, unpaired t test). Values are presented as mean ± SEM (test as indicated; exact p value given if less than 0.05).

To validate the transcriptome analysis, we first examined the expression levels of transport-related protein expression in mouse IPC model induced by a 15-min occlusion of middle cerebral artery. At 48 h after reperfusion, the level of Kif5b, the heavy chain of the microtubule-dependent motor kinesin-1, was significantly reduced (p = 0.0069, t = 3.605, df = 8, unpaired t test). However, levels of Kif17 and microtubule-associated protein-2 (MAP2), a component of neuron microtubule, were not changed (Kif17, p = 0.4703, t = 0.7632, df = 7; MAP2, p = 0.5731, t = 0.5910, df = 7; unpaired t test; Figures 6E and 6F). Next, we tested whether downregulation of Kif5b was specific to IPC or whether this was a response to the global regulation of neuronal activity. We injected mice with kainite to induce temporal lobe epilepsy and enhance neuronal activity (Ramamoorthi et al., 2011). We found that the level of Kif5b, but not Kif17, was reduced at 48 h after kainate treatment (Kif5b, p = 0.005, t = 3.2533, df = 16; Kif17, p = 0.3552, t = 0.9522, df = 16; unpaired t test; Figures 6G and 6H), which indicates that the reduction of kinesin-1 was associated with increased neuronal activity. This suggests a role of the intracellular transportation system underlying the suppression of cell death pathways by IPC and indicates that kinesin-1-dependent alteration of NMDAR compartmentalization might be the core mechanism.

Discussion

The present study revealed a complex between kinesin-1 and GluN2B-containing NMDARs. This complex mediates the extrasynaptic targeting of NMDARs, the resulting pro-death signaling including calcium influx and NO production, and eventually neuronal vulnerability to NMDA-induced excitotoxicity and ischemia-evoked neurodegeneration. The transcriptome of IPC further implicates a role of kinesin-1 in mediating activity-dependent neuroprotection, probably via altering NMDAR compartmentalization and shifting the balance to highlight pro-survival pathways. Our findings demonstrate a role of kinesin-1 in mediating the compartmentalization of signaling pathways within neurons, which can be tuned down by suppressing kinesin-1 expression or by dissociating the interaction between kinesin-1 and GluN2B containing NMDAR when neurons sense a detrimental insult.

It has long been known that hyperactivation of NMDARs contribute to excitotoxicity and neuronal death. NMDAR-related excitotoxicity also acts as the major mechanism that causes neuronal death in both acute and chronic neurodegenerative disorders. However, attempts to translate NMDAR antagonization to clinical applications are far from satisfactory (De Keyser et al., 1999). The reasons might stem from the complexity of NMDAR-mediated signaling and subcellular compartmentalization (Zhou and Sheng, 2013). There are two distinct pools of NMDAR: synaptic NMDAR steadily accumulates at the PSD, whereas extrasynaptic NMDAR is more mobile and dynamic in extrasynaptic regions (Bading, 2017, Paoletti et al., 2013, Tovar and Westbrook, 1999). The downstream signaling and the functional consequence of these two distinct pools of NMDAR are also different. The activation of synaptic NMDAR activates mitogen-activated protein kinase signaling and modulates synaptic strength (Thomas and Huganir, 2004). The dysfunction of this pathway causes not only learning and memory deficits but also neuropsychiatric disorders such as schizophrenia (Cohen et al., 2015). The activation of synaptic NMDAR elicits the calcium-elicited pro-survival pathway (Hardingham et al., 2002). Our knowledge of extrasynaptic NMDAR is far less thorough compared with our understanding of synaptic NMDAR. The activation of extrasynaptic NMDAR was found to suppress the pro-death pathway mediated by synaptic NMDAR, which shifts the balance toward cell death in neurons. Furthermore, DARK1 is recruited to GluN2B and mediates the pro-death signaling toward excitotoxicity (Tu et al., 2010), whereas PGC-1α opposes extrasynaptic NMDAR function (Puddifoot et al., 2012).

One major question in this field is how the compartmentalization of signaling is regulated. Previous evidence indicates that extrasynaptic NMDARs are more mobile than the synaptic NMDARs (Groc et al., 2004), which makes the convergence of different transportation systems possible. In support of this notion, dynamin-dependent regulation is known to limit extrasynaptic synaptic activation (Wild et al., 2014). Our finding that microtubule-dependent transportation regulates extrasynaptic NMDAR fits into the current model in two ways. First, kinesin-1 may directly convey GluN2B-containing NMDAR to the extrasynaptic region. Kinesin-1 has been known to directly interact with Myosin Va (Huang et al., 1999), suggesting cooperation between microtubule- and actin-dependent transportation. Besides synaptic regions, actin filament dynamics are found to be enriched at the lateral zone and neck of the spine (Frost et al., 2010). Our model suggests that extrasynaptic NMDAR is first conveyed and targeted to extrasynaptic regions by microtubule-dependent transport and then moves between synaptic and extrasynaptic regions through the cooperation of microtubule- and actin-dependent transport systems. Second, kinesin-1 may convey internalized NMDAR from synaptic to extrasynaptic regions (Du et al., 2016). This is supported by the evidence that internalized NMDARs contain vesicles that are attached to microtubule-like filaments (Tovar and Westbrook, 1999), which further indicates a role of kinesin-1 in determining the fate of the internalized vesicles toward recycling or degradation.

Our finding that kinesin-1 mediates transport of NMDAR adds complexity to the anterograde transport of NMDAR. It has been proposed that Kif17 drives NR2B-containing NMDA receptors to synaptic sites (Setou et al., 2000, Yin et al., 2011), which is in parallel with our observation that knockdown of Kif5b reduced only extrasynaptic targeting and functioning. It is widely recognized that neurons may use different transport systems to convey cargo to distinct sub-domains, at least in the case of mitochondria transport, which relies on both Kif5 and Kif1b (Nangaku et al., 1994, Wang and Schwarz, 2009). When reaching the site with high levels of calcium, the mitochondrial membrane protein Miro mediates detachment or folding of kinesin-1, which traps the mitochondria (Wang and Schwarz, 2009), whereas Kif1b uses a KBP-dependent mechanism that is distinct from kinesin-1 (Wozniak et al., 2005). NMDAR exists as both diheteromer, consisting of GluN1/GluN2A or GluN1/GluN2B as the majority, or triheteromer consisting of GluN1/GluN2A/GluN2B in vivo. The functional properties of these two forms of NMDAR are different (Hansen et al., 2014). The triheteromer is presumably the dominant form of NMDAR in the adult forebrain, consistent with our data showing that kinesin-1 co-immunoprecipitates with both GluN2B and GluN2A (Figure 1A). However, it has also been reported that GluN2B containing NMDAR is the major form of NMDAR in the extrasynaptic site (Tovar and Westbrook, 1999). It is plausible that kinesin-1 has greater affinity to the diheteromer consisting of GluN1/GluN2B, which therefore enriches GluN2B-containing NMDAR in the extrasynaptic region, whereas the Kif17 has a greater impact on triheteromer, which is enriched in the synaptic region. Supporting this notion, our analysis on the surface distribution NMDAR reveals the reduction of GluN2B, but not GluN2A, in the surface (Figures 1G and 1H). To understand the function and compartmentalization of NMDARs in more detail, especially its role in activity-dependent plasticity and disease pathology, we need to examine the regulatory or coordinating mechanisms among the different kinesins and between different transportation systems.

It is well established that the hierarchical assembly of synapses and neuronal contacts is highly dependent on intracellular transportation systems (Hirokawa et al., 2010). Our findings provide an intrinsic-activity-dependent mechanism for the regulation of NMDARs, which protects neurons against excitotoxic insult. Reduction of NMDAR activity could be regarded as a form of protective feedback that limits excess NMDAR-dependent calcium influx triggering excitotoxicity. This feedback can be through a number of factors, such as NMDAR subunit composition (Martel et al., 2012), the association of scaffolding proteins (Forder and Tymianski, 2009), and neuronal activity (Groc et al., 2004, Wild et al., 2014). Furthermore, our findings combine and unify the associations between kinesin level, elicited neuroprotection, and subsequent impact on extrasynaptic NMDAR, which supports the notion that regulating components beyond the synapse might be essential to induce neuroprotection. Consistent with our current findings, reduced mTORC1 activity and the resulting induction of autophagy has been proposed as another neuronal-activity-dependent mechanism to induce tolerance. Interestingly, this mechanism might function closely with intracellular transportation systems in eliciting neuroprotection (Fu et al., 2014, Papadakis et al., 2013). Altogether, these findings provide a potential therapeutic strategy that could delay degenerative processes by extending the period and effect of intrinsic neuroprotective mechanisms.

Inhibiting kinesin has been proposed as a promising approach to treating cancer, and some small molecules have already been developed. For example, the Kif11 motor protein mediates mitosis and drives glioblastoma invasion, proliferation, and self-renewal. Inhibition of Kif11 by a specific small molecule could stop the progression of glioblastoma (Venere et al., 2015). The perturbation of intracellular transport may not only treat malignant dividing cells such as cancer but also show promising effects on the central nervous system (CNS). Epothilone D, a microtubule stabilizer that can reduce microtubule dynamic and microtubule-dependent transport, was able to ameliorate the impaired exploratory behavior in a Rett mouse model (Delépine et al., 2016). Epothilone B, a US Food and Drug Administration-approved microtubule stabilizer that can limit microtubule-dependent transport, was able to reactivate neuronal polarization and promote axon regeneration after CNS injury (Ruschel et al., 2015).

Together with literatures, our findings suggest that targeting of intracellular transport, including kinesin-1, could be potential targets for deferring neurodegeneration. Extrasynaptic NMDAR has been implicated in both acute and chronic neurodegenerative diseases (Bading, 2017). This makes Kif5b/kinesin-1 a potential therapeutic candidate. Our study utilized an acute stroke model to examine the neuroprotective effect by reducing kinesin-1. It would be of translational interest to test the effect of kinesin-1 reduction on chronic neurodegenerative models, like those of Alzheimer and Parkinson diseases. Moreover, suppressing Kif5b expression, particularly by disrupting the Kif5b-GluN2B interaction, might open an alternate therapeutic avenue. As the structure for kinesin-1 cargo-binding motif has been resolved (Pernigo et al., 2013, Yip et al., 2016), it would be tempting to resolve the structure for kinesin-1-GluN2B complex and use this structure as the basis to predict or design molecules that could halt extrasynaptic NMDAR-mediated neurodegeneration. As such an approach would mimic intrinsic changes induced by brief and beneficial stimulations, it might be tolerated by brain and thus minimize potential side effects. This would also prolong the period before the brain becomes vulnerable to detrimental insult, enabling the brain to adapt or develop resistance (Stenzel-Poore et al., 2003), delay functional impacts like memory or motor deficits, and allow for efficient interventions (Armstead et al., 2018).

Limitations of the Study

We demonstrate that a panel of kinesin family proteins is reduced in the transcriptome of a mouse cerebral ischemia preconditioning model. Although our data show that Kif5b reduction is neuroprotective, the causal link between kinesin-1 reduction and ischemia-preconditioning-induced neuroprotection remains to be studied. Moreover, the current understanding of extrasynaptic NMDARs in physiological condition is still elusive. Although we find that the heterozygous knockout of kif5b gene reduces extrasynaptic NMDARs, but not synaptic ones, and the kif5b mice appear to be tolerant, additional experiments are needed to characterize the physiology and behavioral states of neuronal kinesin-1 reduction in vivo.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.