(KIF2A) Functions in Suppression of Collateral Branch Extension

View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Elsevier - Publisher Connector Cell, Vol. 114, 229–239, July 25, 2003, Copyright 2003 by Cell Press Kinesin Superfamily Protein 2A (KIF2A) Functions in Suppression of Collateral Branch Extension

Noriko Homma, Yosuke Takei, pressed in postmitotic cells, such as juvenile neurons Yosuke Tanaka, Takao Nakata, (Noda et al. 1995; Morfini et al., 1997), and its function Sumio Terada, Masahide Kikkawa, in the brain is largely unknown. Yasuko Noda, and Nobutaka Hirokawa* During neuronal development, postmitotic neurons Department of Cell Biology and Anatomy extend long primary axons toward targets to form appro- Graduate School of Medicine priate connections. In addition to primary axons, neu- University of Tokyo rons sometimes establish connections with plural tar- Hongo Bunkyo-ku, Tokyo, 113-0033 gets by axonal collateral branches (reviewed by O’Leary Japan et al., 1990). During these processes, the growth cones, the motile portion of terminals, are known to properly orient axon formation (Tessier-Lavigne and Goodman, Summary 1996; Stoeckli and Landmesser, 1998; Keynes and Cook, 1995). For example, while a primary growth cone Through interactions with microtubules, the kinesin is predominantly extending, several collateral growth superfamily of proteins (KIFs) could have multiple roles cones, which are generated at future branching points, in neuronal function and development. During neu- remain short until conditions require them to extend ronal development, postmitotic neurons develop pri- (O’Leary and Terashima, 1998; Kalil et al., 2000). That mary axons extending toward targets, while other col- is, the primary growth cones extend, but the collateral lateral branches remain short. Although the process growth cones appear to be suppressed in the same of collateral branching is important for correct wiring neuron. of the brain, the mechanisms involved are not well What molecular mechanisms regulate growth and understood. In this study, we analyzed kif2aϪ/Ϫ mice, suppression of growth cones in the same neuron? It has whose brains showed multiple phenotypes, including become clear that the cytoskeleton plays a central role aberrant axonal branching due to overextension of in axonal growth as an intrinsic factor. In growth cones, collateral branches. In kif2aϪ/Ϫ growth cones, microtu- actin filaments and MTs are major cytoskeletal constit- bule-depolymerizing activity decreased. Moreover, uents whose organization changes dynamically (Tanaka many individual microtubules showed abnormal be- et al., 1995; Tanaka and Kirschner, 1991; Lin and havior at the kif2aϪ/Ϫ cell edge. Based on these results, Forschner, 1993). Recent studies showed that in addi- we propose that KIF2A regulates microtubule dynam- tion to the crucial role of actin filaments (Kuhn et al., ics at the growth cone edge by depolymerizing micro- 2000; Liu and Strittmatter, 2001), MT dynamics provide tubules and that it plays an important role in the sup- significant contributions to growth cone motility (Kalil pression of collateral branch extension. et al., 2000; Teng et al., 2001; Fukata et al., 2002). In this study, we generated kif2aϪ/Ϫ mice to elucidate Introduction KIF2A function in vivo. kif2aϪ/Ϫ mice died within a day of birth with multiple brain abnormalities. We focused The kinesin superfamily of proteins (KIFs) plays a signifi- on the mechanism underlying the development of abnor- cant role in transport of various membranous organelles mally long axonal collateral branches in kif2aϪ/Ϫ neu- and protein complexes on microtubules (Hirokawa, rons. The MT-depolymerizing activity in growth cones of 1998). Among KIFs, M-kinesins (also called KinIs) are kif2aϪ/Ϫ neurons was lower than that of kif2aϩ/ϩ neurons. unique in possessing a motor domain in the middle of the Moreover, individual MTs were overextended at the primary protein sequence (Aizawa et al., 1992; Hirokawa kif2aϪ/Ϫ cell edge. These data suggest that KIF2A depo- 1998; Hirokawa and Takemura, 2002; Kim and Endow, lymerizes MTs at the growth cone edge and suppresses 2000). In mice and humans, there are three KIF2 genes, the growth of axonal collateral branches. kif2a, kif2b, and kif2c (Miki et al., 2001). In addition to their unique primary sequence structure, previous stud- Results ies showed that some of them have a unique function in depolymerizing microtubules (MTs) in mitotic cells. kif2aϪ/Ϫ Mice Showed Multiple For example, the KIF2C subfamily of M-kinesins, includ- Brain Abnormalities ing MCAK (murine KIF2C) and XKCM1 (Xenopus KIF2C), We generated kif2aϪ/Ϫ mice by gene targeting using a depolymerizes MTs at MT tips in vitro (Wordeman and promoter trap strategy (Figure 1A). ES cell recombinants Mitchison, 1995; Walczak et al. 1996; Desai and Mitchi- were selected by genomic Southern blotting (Figure 1B). son, 1997; Maney et al., 2001, Hunter et al., 2003), and Disruption of KIF2A in these mice was confirmed by is involved in the cell cycle by regulating MT dynamics Western blotting (Figure 1C). kif2aϪ/Ϫ mice were born in cultured cells (Walczak, et al., 1996; Maney, et al., alive but died within one day of birth without sucking milk 1998, 2001; Kline-Smith and Walczak, 2002). Whereas (Figure 1D). Their brains showed multiple abnormalities XKIF2 (Xenopus KIF2A) has MT depolymerizing activity including ventricle enlargement (Figures 1E and 1F), in vitro (Desai et al., 1999), KIF2A is predominantly ex- laminary defects, and lack of nerve nuclei in normal localizations. Laminary defects were observed in the *Correspondence: [email protected] hippocampus (Figures 1G and 1H), cerebral cortex (Fig- Cell 230

ures 2A and 2B), and cerebellum (data not shown). The lack of nuclei was observed in the inferior olivary nucleus (Figures 1I and 1J), facial nerve nucleus (Figures 2F and 2G), and in many other nerve nuclei such as the basilar pontine nucleus, mesencephalic trigeminal nucleus, and trochlear nucleus (data not shown). These results suggest that KIF2A is involved in the lamination and organization of nerve nuclei during brain development.

Migratory Defects in kif2aϪ/Ϫ Mouse Brain To investigate the cause of the kif2aϪ/Ϫ brain abnormalities, we analyzed neuronal migration, since such abnormalities are often related to migratory defects (reviewed by Hatten, 2002). In the cerebral cortex, we performed birth-dating analysis (Takahashi et al., 1992) by labeling E14-born cortical neurons with bromodeoxyuridin (BrdU) and analyzing their distribution at P0. As shown in Figure 2C, E14-born cells in the kif2aϩ/ϩ brain finished migration at P0 and formed a tightly packed layer near the cortical surface (Figure 2C). In contrast, E14-born cells remained behind in the kif2aϪ/Ϫ brain (Figures 2D and 2E) suggesting a migratory defect of cortical granular neurons in kif2aϪ/Ϫ mice. We next performed quantitative analysis of the migration ratio of facial nerve cells (Terashima et al., 1993). We retrogradely traced their cell bodies using the DiI labeling method (Takei et al., 1995). When a small DiI crystal was applied to a peripheral facial nerve, neuronal processes and their cell bodies could be labeled and monitored for considerable distances due to diffusion of the dye along plasma membranes. In the kif2aϩ/ϩ brain, facial nerve cells completed migration and formed rounded nuclei at P0 (broken-line ellipse in Figure 2H). In contrast, kif2aϪ/Ϫ cells that could not be identified at a low magnification (Figure 2G) were scattered in the course of migration (Figures 2I and 2K). Statistical analysis of the migration ratio showed a migratory delay in the kif2aϪ/Ϫ brain (Figure 2M). These results showed that KIF2A deficiency caused neuronal migratory defects resulting in laminary defects and mislocalization of nerve nuclei in the brain. More detailed observations of DiI-labeled facial nerve Ϫ Ϫ Figure 1. kif2a / Mice cells showed that while the axon bundle was tightly (A) The targeting vector designed for kif2aϪ/Ϫ mice. The BglII- fasciculated during migration in the kif2aϩ/ϩ brain (arrow digested fragment for genomic Southern blotting is also indicated. in Figure 2J), it was extensively defasciculated (arrows ␤ ␤ B, BglII; S, SpeI; geo, SAIRES geo casette; pBS, pBluescriptII- in Figure 2K) in the kif2aϪ/Ϫ brain. This finding prompted SK(ϩ); and D, diphtheria toxin A fragment. Arrowheads, loxP se- Ϫ/Ϫ quence; empty arrowheads, PCR primers (see Experimental Proce- us to examine the axonal morphology in the kif2a dures). brain. (B) Genomic Southern blotting. The recombinant allele shows an additional band of 2.5 kb. Abnormal Axonal Collateral Branching (C) Western blotting. No band was detected from crude extracts Ϫ/Ϫ Ϫ/Ϫ in kif2a Brain In Vivo from kif2a sample at P0. ϩ/ϩ (D) kif2aϪ/Ϫ mice that did not suck milk. Asterisk, a stomach filled To observe the axonal morphology in kif2a and Ϫ/Ϫ with milk. kif2a brains, DiI crystals were applied to the axon (E–J) kif2aϪ/Ϫ brain abnormalities at P0. The lateral ventricles of bundle in the cerebral cortex and hippocampus (red kif2aϪ/Ϫ brain (F) are slightly larger than kif2aϩ/ϩ brain (E). Hippocam- stars in Figures 3C and 3H). In the cerebral cortex, axons ϩ ϩ pal pyramidal cells are tightly packed in a layer in kif2a / brain (G), were observed to run longitudinally along the cortical Ϫ/Ϫ but not in kif2a brain (H), particularly in CA1 (white arrowheads) plate (CP) in the kif2aϩ/ϩ brain (Figure 3A). In contrast, and CA3 (black arrowheads) field. Inferior olivary complex is clearly observed in kif2aϩ/ϩ brain (enclosed area by broken line in [I]) but many horizontally running neurites were observed in the Ϫ/Ϫ not in kif2aϪ/Ϫ brain (J). cc, corpus callosum; ac, anterior commisure; kif2a brain (arrows in Figures 3B). We confirmed these lv, lateral ventricle. CA1, CA1 field; CA3, CA3 field; and DG, dentate horizontal neurites as axonal collateral branches by re- gyrus. DAO, dorsal accessory olive; and MAO, medial accessory construction of confocal images of a neuron in 3D. As olive. shown in Figure 3F, a kif2aϪ/Ϫ neuron extended more collateral branches along the primary axon (white arrow- KIF2A Function in Growth Cones 231

Figure 2. Migratory Defects in kif2aϪ/Ϫ Brain (A and B) Laminary defects in cerebral cortex at P0 in kif2aϪ/Ϫ mice. Compared with kif2aϩ/ϩmice (A), abnormally packed clusters of neurons were observed in kif2aϪ/Ϫ mice (arrows in B). (C–E) Neuronal birth-dating analysis of cerebral cortex at P0. Many of the labeled neurons remained in CPi and SP of the cortex in kif2aϪ/Ϫ mice (D), but not kif2aϩ/ϩmice (C). (E) The distribution of BrdU-labeled neurons shows migratory defects in cerebral cortex in kif2aϪ/Ϫ mice (*p Ͻ 0.01 by Man-Whitney’s test). ML, marginal layer; CPs, supragranular part of cortex; CPi, infragranular part of cortex; SP, subplate layer. IZ, intermediate zone; and VZ, ventricular zone. (F–M) Migratory defect of facial nerve nuclei. (F–I) Bodian-stained sections of the facial nerve nuclei of kif2aϩ/ϩ (F and H) and kif2aϪ/Ϫ (G and I) brain at P0. (H) and (I) are images within rectangles in (F) and (G) at high magnification. In (H) and (I), the lines indicate the ventricular zone and the arrows indicate the migratory direction of facial nerve cells. (J and K) DiI labeling of the facial nerve nucleus of the kif2aϩ/ϩ (J) and kif2aϪ/Ϫ (K) mice at P0. Note that the axon barrels of kif2aϪ/Ϫ neurons are severely defasciculated. (L) Quantitative procedure for determining the standardized positions of facial nerve cells. Line Z passing through the internal genu (G) and the soma of facial nerve cells (F) are drawn. This line crosses the ventral margin of the brainstem (blue broken line) at point M. The lengths of GF (ϭ X) and GM (ϭ Y) are measured and migration ratio is calculated (MR ϭ X/Y ϫ 100). (M) The migration ratio versus cell population. Severe migratory delay is observed in kif2aϪ/Ϫneurons (*p Ͻ 0.01 by Man-Whitney’s test). All quantitative data are presented as the means Ϯ SEM from three serial sections each in five independent experiments.

heads in Figure 3F) than a kif2aϩ/ϩ neuron (Figure 3D). Absence of KIF2A Affects the Length of Collateral Next we traced an axon and its collateral branches from Branches but Not the Number of Branches their cell body using NeuronTracer software (Figures 3E We next examined how KIF2A is involved in abnormal and 3G) and found that horizontal neurites in the SP branching. There are at least three possibilities: KIF2A area of a kif2aϩ/ϩ neuron (empty arrowheads in Figure could be involved in the generation of collateral growth 3D) did not originate from the chosen neuron. The aver- cones, in the elongation of preexisting but unextended age number of collateral branches per 100 ␮m long axon collateral growth cones, or in both. To test these possi- of kif2aϪ/Ϫ neurons (means Ϯ SD) was almost four times bilities, we prepared low-density primary cultures of hip- (1.65 Ϯ 0.06, n ϭ 9) greater than that of kif2aϩ/ϩ neurons pocampal neurons (Goslin and Banker, 1998). Two days (0.38 Ϯ 0.13, n ϭ 10). The same phenotype was observed after plating, neurons were fixed and labeled with an in hippocampal pyramidal neurons in the CA1 region anti-tubulin antibody and FITC-conjugated phalloidin. In (Figures 3J and 3I). Figure 4A, a kif2aϩ/ϩ neuron is at stage 3 and predomi- Cell 232

nantly extends a single primary axon (large arrowhead in Figure 4A). Several collateral growth cones were observed along the primary axon (small arrows in Figure 4A). However, kif2aϪ/Ϫ neurons developed aberrantly elongated collateral branches, which rebranched into tertiary and quaternary branches (small arrows in Figure 4B). Statistical analysis showed that among hippocampal neurons at stage 3, more than 70% of kif2aϪ/Ϫ neurons showed an abnormal architecture, defined as hav- ing tertiary branches of more than 20 ␮m long (kif2aϩ/ϩ ϭ 10.2 Ϯ 3.4%; Figure 4C, kif2aϪ/Ϫ ϭ 71.8 Ϯ 5.5%; Figure 4D, n ϭ 120). The abnormal architecture was consistent with the phenotype observed in vivo (Figure 3J) and it was rescued when exogenous KIF2A was transfected to kif2aϪ/Ϫ neurons (rkif2aϪ/Ϫ ϭ 12.2 Ϯ 2.8%, Figure 4E, n ϭ 120). Other morphological parameters were also compared. The average length of the collaterals of kif2aϪ/Ϫ neurons was longer than that of kif2aϩ/ϩ neurons (Figure 4H). The longest axons (Figure 4G) were also slightly longer in kif2aϪ/Ϫ neurons. However, there was no significant difference in the average number of collaterals (Figure 4F). The rescued kif2aϪ/Ϫ neurons (Figures 4F–4H, green bars) showed the same feature as that of kif2aϩ/ϩ neurons (Figures 4F–4H, red bars), suggesting that KIF2A is involved in controlling the elongation of collateral branches but not their generation.

kif2aϪ/Ϫ Cultured Neurons Lose Suppression of Collateral Branches Next we performed time-lapse analysis using cultured hippocampal pyramidal neurons to observe chronologi- cal changes of arborization pattern. The neurons were infected with recombinant adenovirus expressing GFP- tubulin to visualize neurites in detail (Nakata et al., 1998). Prior to collecting data, we confirmed that there was no significant effect of adenovirus infection on the development of neurons derived from kif2aϩ/ϩ and kif2aϪ/Ϫ mice (data not shown). In kif2aϩ/ϩ neurons, elongation of collaterals was basi- cally suppressed (arrows 1 and 2 in Figure 5A), and elongated collaterals (arrows 3 and 4 in Figure 5A, 41 hr to 53 hr) and trailing processes (arrow 6 in Figure 5A, 65 hr to 77 hr) actively shrunk. On the other hand, the cultured kif2aϪ/Ϫ neurons continued extending axonal collateral branches once they were generated (broken arrows in Figure 5B). Active shrinkage of collaterals and Figure 3. Abnormal Axonal Branching in kif2aϪ/Ϫ Brain trailing processes was not observed as frequently as in ϩ/ϩ Schema of cortical granular cells (C) and hippocampal pyramidal kif2a neurons (arrowheads in Figure 5B). Cell bodies Ϫ/Ϫ cells (H) were drawn. Red stars indicate DiI crystals, and pink ellipses of more than 70% of 5-day-old cultured kif2a neurons and lines are labeled cell bodies and their axons. stopped translocating at branching points (kif2aϩ/ϩ ϭ (A and B) Axons in kif2aϩ/ϩ and kif2aϪ/Ϫ cerebral cortex. More addi- 2.6 Ϯ 2.5%, kif2aϪ/Ϫ ϭ 71.3 Ϯ 3.1%). This phenotype tional horizontal neurites (arrowheads in B) are observed in CPi area was rescued by exogenous KIF2A (rescued kif2aϪ/Ϫ ϭ of kif2aϪ/Ϫ cortex than that of kif2aϩ/ϩ cortex (A). 5.2 Ϯ 2.1%). Thus, we suggest that KIF2A is an essential (D–J) A DiI-labeled neuron in vivo. 3D images of individual neurons in cerebral cortex (D–G) and hippocampus (I and J). Note that kif2aϪ/Ϫ factor for suppressing collateral branches (collateral neurons extend many collateral branches (white arrowheads in [F] suppression) and that a cause of migratory defects in and [J]) while kif2aϩ/ϩ neurons elongate primary axons straightly the kif2aϪ/Ϫ brain (Figure 2) could be abnormal collateral (white arrow in [D] and [I]). (E and G) 3D images of collaterals along elongation in vivo. the primary axon in cerebral cortex were traced using NeuronTracer. This analysis clarified that many horizontal processes observed in IZ of kif2aϩ/ϩ neuron were not derived from the chosen cell (empty Decreased MT Depolymerization in Growth Cones Ϫ/Ϫ arrowheads). The primary axon is light green and the secondary of Cultured kif2a Neurons branch is light blue. As shown in Figure 6A, the recombinant murine KIF2A protein showed MT-depolymerizing activity in an ATP- dependent manner in vitro similar to other members of KIF2A Function in Growth Cones 233

Figure 4. Abnormal Elongation of Collateral Branches in kif2aϪ/Ϫ Cultured Neurons (A and B) Representative hippocampal neurons 2 days after plating. Neurons were double-labeled with FITC-conjugated phalloidin for F-actin (green) and anti-tubulin antibody (red). kif2aϪ/Ϫ neuron extends more long collaterals (arrows in [B]) than kif2aϩ/ϩ ones (arrows in [A]). (C–E) Rescue analysis. Low-magnification views of kif2aϩ/ϩkif2aϪ/Ϫ, and rescued kif2aϪ/Ϫ neurons immunostained with anti-tubulin antibody. The abnormal branching of kif2aϪ/Ϫ neurons (D) are rescued by adenovirus-mediated expression of exogenous KIF2A (E). (F–H) Statistical analysis for the number and the length of neurites. Data are expressed as means Ϯ SEM from 20 neurons each in five independent experiments (*p Ͻ 0.01 by Man-Whitney’s test). Compared with the average number of collaterals (F), there is significant difference in the average length of the longest axon (G) and collaterals (H) between kif2aϩ/ϩ (red bars) and kif2aϪ/Ϫ (blue bars) neurons. The white bars represent the data of rescued kif2aϪ/Ϫ neurons. the M-kinesins as shown in the case of Xenopus (Desai small broken-line box in Figure 6B). The total fluores- et al., 1999). Since KIF2A was reported to be abundantly cence intensity in growth cones versus time is shown expressed in growth cones (Noda et al., 1995; Morfini in Figure 6G. The rationale of this experiment was as et al., 1997), we tested MT depolymerization by KIF2A follows: MTs and tubulin dimers are in a dynamic equilib- in neuronal growth cones. An experiment was designed rium at a growth cone (arrows in Figures 6E and 6F). based on the fluorescence loss in photobleaching (FLIP) Free tubulin dimers in the cytoplasm are diffusible. Once method. In this method, axonal areas immediately adja- photobleached, only bleached fluorescent tubulin di- cent to growth cones were consistently photobleached mers (black dots in Figure 6F) are transported from the (a large broken-line box in Figure 6B). Electron micros- axon to growth cones. Since fluorescent tubulin dimers copy confirmed no cytoskeletal damage by photo- (green dots in Figure 6F) derived from depolymerized bleaching (compare Figure 6D with Figure 6C). The loss MTs in growth cones are diffusible, their concentration of fluorescence in growth cone areas was monitored (a will rapidly reach equilibrium between the bleached Cell 234

growth cones (blue line in Figure 6G) was less than that in kif2aϩ/ϩ growth cones (red line in Figure 6G). This tendency was confirmed for all the time points of measurements by Man-Whitney’s test (p Ͻ 0.03), suggesting that MT-depolymerizing activity in kif2aϪ/Ϫ growth cones was lower than that in kif2aϩ/ϩ growth cones. As controls, we prepared nocodazol-treated, taxol-treated, and fixed samples. The nocodazol-treated samples showed rapid decrease in fluorescence intensity (green line in Figure 6G) since MTs had been depolymerized. The taxol-treated samples with stabilized MTs showed a slower decrease in fluorescence intensity than the nontreated neurons (black line in Figure 6G) in a dose-dependent manner (data not shown). Fixed samples, in which both polymerized and free tubulin dimers were immobilized, displayed only a slight decrease in fluorescence intensity (pink line in Figure 6G), showing that there was a slight loss of background fluorescence in the experiment. These results indicate that the decrease in fluorescence intensity reflects MT-depolymerizing activity in growth cones.

Overextension of MTs at the Edge of kif2aϪ/Ϫ Glial Cells Using FLIP, we measured the fluorescence intensity of MT polymer and free tubulin as a whole. Next we wanted to compare more specifically the dynamics of the MT polymer. However, we found that it was difficult to observe individual MTs in neuronal cultures, because MTs are arranged very close to each other in growth cones. Therefore, to compare the behavior of single MTs, we used cultured glial cells infected with an adenovirus carrying GFP-tubulin. Glial cells express KIF2A (data not shown) and individual MTs are visible at the cell edge because of their sparse distribution. This experiment showed us the abnormal behavior of MTs at the cell edge as the most notable phenotype of kif2aϪ/Ϫ cells (See Supplemental Movies S1 and S2 available at http:// www.cell.com/cgi/content/full/114/2/229/DC1/). As shown in Figure 6J, when MTs in kif2aϩ/ϩ cells reached the cell edge, MT tips stopped growing and appeared to be anchored, or they underwent repeated short polymerization and depolymerization cycles in the peripheral area Figure 5. Time-Lapse Images of Cultured Hippocampal Neurons (arrowheads in Figure 6J). However, MT tips in kif2aϪ/Ϫ Representative time-course images of kif2aϩ/ϩ (A) and kif2aϪ/Ϫ neurons (B). Numbers at the right upper corner of the images indicate cells were not anchored upon reaching the cell edge the time after plating (hr). GFP-tubulin expression in neuron is visible. but instead they continued growing. As a result, as they The extension of collateral growth cones is suppressed in kif2aϩ/ϩ approach the membrane, they turned away from the cell neurons (arrows 1 and 2 in [A]), but not in kif2aϪ/Ϫ neurons (broken edge (arrowheads in Figure 6K) or sometimes ran along ϩ/ϩ arrows in [B]). Moreover, in kif2a neuron, collateral branches the plasma membrane (arrows in Figure 6K). To quantify (arrows 3 and 4 in [A]) or trailing processes (arrow 6 in [A]) shrink this phenomenon, we focused on MT tips within 7 ␮m during migration, but not in kif2aϪ/Ϫ neuron (arrowheads in [B]). ϩ/ϩ Ϫ/Ϫ kif2aϪ/Ϫ neuron finally stopped migrating at a bifurcation (at 95 hr from the cell edge (kif2a , Figure 6H; kif2a , Figure in [B]). In terms of the number of collateral growth cones, only GFP- 6I). We divided these MTs into four groups based on tubulin-expressing neurites are detected, and therefore all short their behavior (Table 1). These results confirmed the collateral growth cones such as those shown in Figure 4A are not overextension behavior of MTs in kif2aϪ/Ϫ cells (p Ͻ 0.01, necessarily visible. by Man-Whitney’s test), suggesting that KIF2A regulates MT dynamics at the plus-end of MTs near the cell edge. axon and the nonphotobleached growth cone. On the other hand, fluorescent MT polymers (green line in Fig- Discussion ure 6F) in the growth cone area are not photobleached during the experiment. Thus, the amount of fluorescence In this study, we investigated the role of KIF2A in vivo, lost in growth cone areas will reflect the MT-depolymer- particularly in axonal growth during brain development. izing rate in growth cones. We analyzed the brain abnormalities in kif2aϪ/Ϫ mice As shown in Figure 6G, fluorescence decay in kif2aϪ/Ϫ and observed that kif2aϪ/Ϫ neurons overextended their KIF2A Function in Growth Cones 235

Figure 6. KIF2A Function as a MT Destabilizer In Vitro and In Vivo (A) Murine KIF2A depolymerizes MTs. Taxol-stabilized MTs (1.6 ␮M) were incubated for 60 min at room temperature (RT) with KIF2A (80 nM) in the presence of ATP. Note that tubulins are recovered from the supernatant in the presence of ATP as a result of MT depolymerization by KIF2A. Control: Taxol-stabilized MTs without KIF2A. (B–G) MT-depolymerizing activity in growth cones of kif2aϪ/Ϫ neurons is lower than that of kif2aϩ/ϩ neuron. MT-depolymerizing activity was analyzed using FLIP method (see Experimental Procedures for details). (B) Confocal image of a growth cone of cultured hippocampal neuron before photobleaching. (C and D) Electron micrographs of nonphotobleached (box C in [B]) and photobleached (box D in [B]) axons. There was no significant damage of MTs in photobleached axon. P, proximal; D, distal. (E and F) Experimental design of FLIP analysis (see Results for detail). Green line, fluorescent MTs; green dots, fluorescent tubulin dimers; black lines, photobleached MTs; black dots, photobleached tubulin dimers. (G) Fluorescence loss in growth cone area after photobleaching. Note that fluorescence decay in kif2aϪ/Ϫ growth cones (blue line) was lower than that in kif2aϩ/ϩ ones (red line). As controls, nocodazol-treated (green line), taxol-treated (black line), and the fixed (pink line) samples are examined. Data are presented as means Ϯ SEM from five growth cones each in twenty independent experiments (*p Ͻ 0.03, by Student’s t test). (H and I) Individual MT dynamics in kif2aϩ/ϩ (H) and kif2aϪ/Ϫ (I) glial cells. Individual MTs were visualized using GFP-tubulin. (J and K) Time-lapse images of white broken boxes in (H) and (I). Actual image times are illustrated on the top right corner of images. White arrowheads in J and K indicate MT tips. When MTs in kif2aϪ/Ϫ cells reached the edge, MT tips tend to turn around (arrowhead in [K]) or hit the membrane (arrow in [K]). (See Supplemental Movies S1 and S2 available at http://www.cell.com/cgi/content/full/114/2/229/DC1/). Cell 236

Table 1. Statistical Analysis of Population of MTs Classified Based on Their Behavior

kif2aϩ/ϩ cells kif2aϪ/Ϫ cells No. of MTs analyzed 130 86 No. of pausing MTs 23 3 Overextended MTs 6.7 Ϯ 3.9% 70.8 Ϯ 11.7% MTs not shrinking 0.8 Ϯ 2.0% 0.0% MTs not growing 10.0 Ϯ 7.1% 1.2 Ϯ 2.6% MTs with growth and shrinkage 82.5 Ϯ 7.5% 28.0 Ϯ 11.9% No. of cells analyzed 6 5

Parameters of MT dynamics were measured at the edge of live cells. All data are presented as means Ϯ SEM. collateral branches. Our focus on the mechanism led to the main conclusion that murine KIF2A depolymerizes MTs and may suppress outgrowth of collaterals by regulating the dynamics of MT tips at the growth cone membrane. Our findings indicate a role for M-kinesins in neuronal development and elucidate a mechanism of growth cone regulation important to axonal growth.

KIF2A Depolymerizes MTs and Regulates Individual MT Dynamics at the Growth Cone Edge For clarification of KIF2A function in neurite extension, we reported here that recombinant murine KIF2A showed an MT-depolymerizing activity in vitro (Figure 6A) and in growth cones of cultured neurons (Figure 6G). Near plasma membranes, individual MT tips tended Ϫ/Ϫ to overextend in kif2a glial cells (Figure 6K) whereas Figure 7. KIF2A Function In Vivo ϩ/ϩ MT tips in kif2a cells decelerated, started a back- (A and B) Schema of MT dynamics in growth cones. (A) In the and-forth movement, and were often anchored near the presence of KIF2A, the growth cone extension is suppressed be- membrane (Figure 6J). We speculate that KIF2A might cause MTs maintain the dynamic equilibrium between polymeriza- depolymerize individual MTs at the tips and control their tion and depolymerization, particularly at the edge of the growth -/- dynamics at the growth cone edge, thereby suppressing cone. (B) In kif2a neurons, MT tips at the cell edge do not show controlled dynamics and hit the membrane. MT tips turn back (red growth cone extension in juvenile brains (Figure 7A). In arrowhead) or push the membrane forward (red arrow). Therefore, Ϫ/Ϫ the kif2a growth cone, since MT tips lose control of growth cones are released from collateral suppression and start to their speed and dynamics, they may push against the extend. growth cone edge (red arrow in Figure 7B), resulting in (C and D) The lack of collateral suppression in kif2aϪ/Ϫ mice. (C) In ϩ ϩ overextension of axons and branches. kif2a / mice, primary axons extend to the target, while collateral Although this is the most direct interpretation of the elongation is suppressed (red line in [C]). When projection errors of primary axons occur (Figure 7C, middle panel), the growth cones present results, it is also possible that KIF2A may trans- of interstitial collateral branches start elongating at right angles from port some molecules that regulate MT dynamics and the axon shaft up to a certain point (Figure 7C, bottom panel) and thus, might be indirectly involved in growth cone sup- compensates for the errors. (D) In kif2aϪ/Ϫ brain, however, since pression. A previous study showed that suppression collateral growth cones are released from suppression (green line of KIF2A in PC12 cells by antisense oligonucleotides in D), primary axons extend to the target with elongated collateral changed the distribution pattern of IGF-1R␤gc and in- branches (Figure 7D, middle and bottom). hibited NGF-induced neurite extension (Morfini et al., 1997). However, our immunohistochemical analysis us- and suppression of collateral branches. Since there was ing an antibody for IGF-1R␤gc (kindly provided by G. no difference in the KIF2A distribution pattern between Morfini) and other markers of organelles such as synap- primary and collateral growth cones in a neuron (data tic vesicles, lysosomes, ER, and mitochondria showed not shown), one possibility is that KIF2A activity may no significant difference in the staining pattern between be different in different growth cones. KIF2A may control kif2aϩ/ϩ and kif2aϪ/Ϫ neurons (data not shown). We do its own activity by tail inhibition as observed in KIF2C/ not exclude the possibility that KIF2A functions in axonal MCAK (Moore and Wordeman, 2001), or other extrinsic/ growth as an organelle transporter (Noda et al., 1995), intrinsic signals or proteins may regulate KIF2A activity but based on the present work, we prefer the simple at growth cones. Another possibility is that activities of and direct interpretation that KIF2A suppresses growth other MT destabilizers or MT stabilizers are different. In cone advance in collaterals through its MT-depolymeriz- this case, a difference in activity of MT destabilizers, ing activity. such as the stathmin family proteins (Di Paolo et al., One important question is what are the differences in 1996; Belmont and Mitchison, 1996; Ozon and col- growth cones that lead to extension of primary axons leagues, 1998, 2002) and katanin (McNally and Vale, KIF2A Function in Growth Cones 237

1993; Ahmad, et al., 1999; Quarmby, 2000), or MT stabi- of KIF2A (Santama et al., 1998). The ES cell line J1 was transfected lizers, such as MT-associated proteins (MAPs)(Teng et with the targeting vector and screened for homologous recombi- al., 2001), may generate different MT dynamics in differ- nants with targeting efficiency of 10% following the standard procedures (Harada et al., 1994). Male chimera mice were backcrossed ent growth cones. In both cases, the regulation of KIF2A with C57BL/6J females in a specific pathogen-free environment, and its substrate MTs play a crucial role in the regulation and mouse lines were maintained in the same environment. Two of growth cone extension. independent mouse strains were used for analysis. kif2aϩ/ϩ alleles were detected by PCR using the following primers: 5Ј-GCT ATA AAG TTT TGC-3Ј and 5Ј-ACT TCC AGT CTG CCC ATA CG-3Ј. neo Collateral Suppression during Brain Development transgene was amplified as described previously (Harada et al., In this study, we refer to the mechanism that keeps 1994). KIF2A expression was analyzed by Western blotting of crude collateral branches short during primary axon targeting extracts of whole brain at P0 (Teng et al., 2001) using monoclonal as “collateral suppression.” KIF2A is an important factor antibody against the N-terminal region of KIF2A (Noda et al., 1995). for the suppression because the loss of KIF2A causes elongation of collateral growth cones (Figures 4B and Histological Studies 4D) and because this phenotype is rescued by exoge- For histopathology and birth-dating analysis, we prepared Bodian- stained brain sections and BrdU-labeled brain sections as described nous KIF2A (Figure 4E). previously (Teng et al., 2001) Previous studies showed the importance of collateral suppression for the development of a proper neuronal Quantitative Procedures for Birth-Dating Analysis network (reviewed by Kalil et al., 2000). It seems that of Cerebral Cortex and for Migration Ratio collateral growth cones remain short until the primary Analysis of Facial Nuclei axons arrive at their correct targets. If targets are We quantified BrdU-labeled nuclei of cerebral cortex as previously reached successfully, collaterals remain short. How- described (Takahashi et al., 1992; Teng et al., 2001). The distribution of labeled facial nerve cells after Bodian staining of kif2aϩ/ϩ and ever, if targeting fails (Figure 7C, middle panel), collater- kif2aϪ/Ϫ mice was quantified as described previously (Terashima et als start elongating in order to compensate for the error al., 1993) with slight modifications. The standardized position of (Figure 7C, bottom panel). These mechanisms would each stained neuron was determined as illustrated in Figure 2L. The be important for forming topographically appropriate distribution of facial nerve cells in cortical slice was analyzed using corticospinal connections from the earliest stages of NIH image software (W. Rasband, Research Service Branch, Na- development (O’Leary et al., 1990; Kuang and Kalil, tional Institutes of Health, Bethesda, MD). 1994; O’Leary and Terashima, 1998). In kif2aϪ/Ϫ brain, DiI Labeling Analysis collateral growth cones might elongate uncontrollably For labeling cerebral and hippocampal neurons in vivo, P0 mice (Figures 3B, 3F, 3J and 7D). As these long collaterals were perfused and fixed with 4% paraformaldehyde. Brains were can sometimes block migration of the cell body (Figure sliced into 200-␮m-thick frontal sections in PBS, and small DiI crys- 5B), failure in the establishment of normal neuronal cir- tals (Molecular Probe) were placed in the axon barrel of each neuron cuits could result. This could potentially explain several (red star in Figures 3C and 3G). After incubation for 12–18 hr at Њ of the observed phenotypes, such as early death and 30 C, the sections were mounted onto glass slides and observed under confocal laser scanning microscope (LSM510, Carl Zeiss Co., the inability to suck milk (Figure 1D). Thus, we propose Ltd.). The Z stack images were deconvoluted using Huygens soft- that the role of KIF2A in collateral suppression is impor- ware (Carl Zeiss Co., Ltd.) and reconstructed using IMARIS (Carl tant for basic mechanisms in brain development. Zeiss Co., Ltd.). For branch analysis in the cerebral cortex, the In addition to the axonal growth phenotype, several deconvoluted images were analyzed using software NeuronTracer other phenotypes were observed in kif2aϪ/Ϫ brain (Figure (Bitplane AG, Zurich). For labeling the facial nerve nucleus, P0 mice 1). These include laminary defects (Figures 1H and 2B) were perfused with 4% paraformaldehyde in PBS, and the proximal end of the peripheral facial nerve near the skull was cut after remov- and mispositioning of nerve nuclei (Figures 1J and 2I) ing the facial skin. A DiI crystal was applied to the cut edge. After due to neuronal migratory defects (Figures 2E and 2M). a 2-day incubation at 37ЊC, the brains were removed and sliced in As mentioned above, abnormal collateral extension in 100-␮m-thick frontal sections. They were observed as described the kif2aϪ/Ϫ brain could be one of the causes of migratory above. defects (Figure 5B). A previous study showed that MT dynamics itself contributes to the displacement of the Hippocampal Neuron Culture and Immunofluorescence Microscopy nucleus and cytoplasm of migrating neurons (Rakic et Cultured hippocampal neurons were prepared as described pre- al., 1996). It is possible that KIF2A might be involved in viously (Goslin and Banker, 1998) except using supplements (B27, this process. KIF2A deficiency may also change MT GibcoBRL) instead of coculture with glial cells. For tubulin and actin dynamics around the cell body and directly contribute staining, 2-day-old cultured neurons were stained as described pre- to the migratory defect. Neuronal migration is a funda- viously (Teng, et al., 2001). The primary antibodies used were: mono- mental process in brain development, and our kif2aϪ/Ϫ clonal anti-tubulin DH1A (1:200, Sigma). The secondary antibodies mouse will be a useful tool for elucidating its basic mech- were: rhodamine-conjugated goat anti-rabbit immunoglobulins (1:100, Cappel Labs.). F-actin was stained by FITC-conjugated phal- anism. loidin (1:300, Sigma-Aldrich). The neurons were observed under confocal laser scanning microscope (LSM510, Carl Zeiss Co., Ltd). The Experimental Procedures number and lengths of neurites were measured using NIH image software (W. Rasband, Research Service Branch, National Institutes Gene Targeting of kif2A of Health, Bethesda, MD). For time-lapse imaging, coverslips were Mouse kif2a genomic clones were obtained from ␭EMBL3 genomic coated with laminin for 2 hr at 37ЊC after poly L-lysine coating. library of mouse ES cell line J1 (Harada et al., 1994) A 0.9 kb coding Cultured hippocampal neurons were infected with recombinant ade- region of the ATP binding consensus P loop of the motor domain novirus, which expresses GFP-tubulin under the control of the CAG was replaced by a floxed pIRES␤ geopolyA cassette for promoter promoter (Nakata et al., 1998), for 30 min on the day of plating. trapping (Figure 1A). This replacement was designed to induce a Time-lapse images were obtained every 2 hr for 5 days using confo- frame shift and also to disrupt KIF2A␤, which is a splicing isoform cal laser scanning microscope (Biorad MRC-1024, Biorad Labora- Cell 238

tories) on an Axiovert100 equipped with a Kr/Ar laser (Zeiss) and a age.” Likewise, “MTs not growing” are at rest and shrinkage and 40ϫ water immersion lens (Zeiss). “MTs not shrinking” show growth and at rest.

Acknowledgments Sedimentation Analysis Tubulin was purified from porcine brains as described previously We thank Dr. M. Yamakado (Department of Anatomy, Jichi Medical (Shelanski et al. 1973). In order to express recombinant KIF2A pro- School) for valuable discussion; Dr. G. Morfini (Institute Investiga- tein, full-length KIF2A DNA was inserted into the ApaI and NotI sites cio´ nMe´ dica Mercedes y Martı´n Ferreyra) for providing the anti-IGF- of pFastBacHT (Bac-to-Bac Baculovirus Expression Systems, Gibco 1R␤gc antibody, Dr. K. Nakajima (Jikei University) for the anti-reelin BRL) according to manufacturer’s instruction with slight modifica- antibody, Dr. T. Nakagawa for a cDNA clone for KIF2A, Dr. M. Setou tions. The Kozak sequence for eukaryotic expression was intro- for the KIF2A-expressing virus, and Dr. Hsueh (Institute of Molecular duced immediately before the initial ATG codon of murine KIF2A Biology, Academia Sinica) for the anti-Tbr1 antibody; Drs. A. Harada, cDNA by PCR. The expressed KIF2A protein was purified from R. Harada, Y. Okada, S. Takeda, K. Chijiiwa, Y. Kanai, S. Nonaka, 540 ml of Hi-5 cells as previously described (Desai et al., 1999). The J. Teng, H. Sato, N. Onouchi, and M. Sugaya-Otsuka for technical purified KIF2A protein showed MT-dependent ATPase activity (data assistance; Drs. C. Zhao, Y. Xu, and all members of the Hirokawa not shown). The depolymerization activity was measured by MT laboratory for valuable discussions and suggestions. This work was sedimentation analysis (Desai et al. 1999). supported by a Center of Excellence grant from the Ministry of Education, Culture, Sports, Science and Technology of Japan to N. MT Dynamics Assay Hirokawa. To analyze MT-depolymerizing activity in growth cones, we used fluorescence loss in photobleach (FLIP) method. Cultured hippo- Received: October 15, 2002 campal neurons were prepared to express GFP-tubulin as described Revised: June 2, 2003 above and observed after 3 days for the full incorporation of GFP- Accepted: June 25, 2003 tubulin into MTs. We selected growth cones randomly, and axon Published: July 24, 2003 areas at the neck of growth cones were repeatedly photobleached and the loss of fluorescence in growth cone areas was monitored. References FLIP largely eliminates the concern that the recovery properties are due to damage at the bleached spot, as all measurements are made Ahmad, F.J., Yu, W., McNally, F.J., and Baas, P.W. (1999). An essen- in areas that are not bleached. Time-series images were acquired tial role for katanin in severing microtubules in the neuron. J. Cell every 3 min for 15 min using a confocal laser scanning microscope Biol. 145, 305–315. (MRC-1024, Biorad Lab.). In all experiments, data analysis began Aizawa, H., Sekine, Y., Takemura, R., Zhang, Z., Nangaku, M., and at the time when we started photobleaching an axon and hence Hirokawa, N. (1992). Kinesin family in murine central nervous system. was designated as t ϭ 0. The total fluorescence intensity of GFP- J. Cell Biol. 119, 1287–1296. tubulin in the minimum square including growth cones was quantified using NIH image software (W. Rasband, Research Service Belmont, L.D., and Mitchison, T.J. (1996). Identification of a protein Branch, National Institutes of Health, Bethesda, MD). We also calcu- that interacts with tubulin dimers and increases the catastrophe lated the background intensity in the square and subtracted the rate of microtubules. Cell 84, 623–631. background intensity from the total fluorescence intensity. The in- Di Paolo, G., Pellier, V., Catsicas, M., Antonsson, B., Catsicas, S., and tensity versus time was plotted. As controls, cultured hippocampal Grenningloh, G. (1996). The phosphoprotein stathmin is essential neurons were exposed to the following drugs for 30 min before for nerve growth factor-stimulated differentiation. J. Cell Biol. 133, taking images: 10 ␮M taxol (Sigma, St. Louis, MO) and 33 nM noco- 1383–1390. dazole (Sigma). Nocodazole-treated cultures were placed on ice for Desai, A., and Mitchison, T.J. (1997). Microtubule polymerization 10 min to depolymerize their MTs. In addition, we prepared fixed dynamics. Annu. Rev. Cell Dev. Biol. 13, 83–117. samples with paraform-glutaraldehyde for detecting the artificial loss of intensity. In each genotype, at least 20 neurons from five Desai, A., Verma, S., Mitchison, T.J., and Walczak, C.E. (1999). Kin mice were examined. We plotted data with SEM from an average I kinesins are microtubule-destabilizing enzymes. Cell 96, 69–78. of 5 mice. After the assay, some neurons were fixed for 20 min in Fukata, Y., Itoh, T.J., Kimura, T., Menager, C., Nishimura, T., Shiro- Karnovsky’s fixative and further processed for conventional electron mizu, T., Watanabe, H., Inagaki, N., Iwamatsu, A., Hotani, H., and microscopy as described previously (Teng et al., 2001) to check the Kaibuchi, K. (2002). CRMP-2 binds to tubulin heterodimers to pro- MT organization in the photobleached area. mote microtubule assembly. Cell Biol. 4, 194–195. For individual MT dynamics imaging, GFP-tubulin-expressing glial Goslin, K., and Banker, G. (1998). Rat hippocampal neurons in low- Њ cells were randomly selected. These cells were maintained at 37 C density culture. In Culturing Nerve Cells, 2nd edition, G. Banker and in an incubator 400ASI 5% CO2 stream (Olympus) on a 1X-70 inverted K. Goslin, eds. (MIT Press: Cambridge, MA). 339–370. microscope equipped with a 60/1.45 NA Plan Apo oil immersion Harada, A., Oguchi, K., Okabe, S., Kuno, J., Terada, S., Ohshima, object. They were excited by a 514 nm Argon laser and an image T., Sato-Yoshitake, R., Takei, Y., Noda, T., and Hirokawa, N. (1994). was acquired at 100 ms exposure time by cool-SNAP HQ equipped Altered microtubule organization in small-calibre axons of mice lack- with a Gen TV intensifier (Hamamatsu). MT dynamics were measured ing tau protein. Nature 369, 488–491. as described by Kline-Smith and Walczak (2002), with a slight modifi- cation. Briefly, MT tips were tracked using the “track points” function Hatten, M. (2002). New Direction in Neuronal Migration. Science of Meta-Morph software (Universal Imaging). All data were pro- 297, 1660–1663. cessed in Microsoft Excel, in which pixel measurements were con- Hirokawa, N. (1998). Kinesin and Dynein Superfamily Protein and verted to distance measurements. Individual MT behavior was plot- the Mechanism of Organelle Transport. Science 279, 519–526. ted as the change in coordinates of the MT tip. Changes in the Hirokawa, N., and Takemura, R. (2002). Kinesin superfamily proteins. position of a MT tip that measured Ͻ1 ␮m between two consecutive In “Molecular Motors” edited by M. Schliwa. Wiley-VCH GmbH, time points were not counted as growth or shortened excursions, Weinheim. pp 79–109. which is the same guideline used by Waterman-Storer and Salmon (1998). Pauses ware defined as slopes of Ͻ1 ␮m/min. We counted Hunter, A.W., Caplow, M., Coy, D.L., Hancock, W.O., Diez, S., Wor- MTs whose tips located within 7 ␮m from the edge of a cell for deman, L., and Howard, J. (2003). The kinesin-related protein MCAK more than 1 min. Among them, we counted pausing MTs first that is a microtubule depolymerase that forms an ATP-hydrolyzing com- moved Ͻ 1␮m and next overextended MTs that extended more plex at microtubule ends. Mol. Cell 11, 445–457. than 7 ␮m during the observation period. The remaining MTs were Kalil, K., Szebenyi, G., and Dent, E.W. (2000). Common mechanisms classified into three groups: MTs that show both growth and shrink- underlying growth cone guidance and axon branching. J. Neurobiol. age within 7 ␮m were classified into “MTs with growth and shrink- 44, 145–158. KIF2A Function in Growth Cones 239

Keynes, R., and Cook, M.W.G. (1995). Axon guidance molecules. bly in the absence of added nucleotides. Proc. Natl. Acad. Sci. USA Cell 83, 161–169. 70, 765–768. Kim, H., and Endow, S.A. (2000). A kinesin family tree. J. Cell Sci. Stoeckli, E.T., and Landmesser, L.T. (1998). Axon guidance at choice 113, 3681–3682. points. Curr. Opin. Neurobiol. 8, 73–79. Kline-Smith, S.L., and Walczak, C.E. (2002). The microtubule-desta- Takahashi, T., Nowakowski, R.S., and Caviness, V.S., Jr. (1992). bilizing kinesin XKCM1 regulates microtubule dynamic instability in BUdR as an S-phase marker for quantitative studies of cytokinetic cells. Mol. Biol. Cell 13, 2718–2731. behavior in the murine cerebral ventricular zone. J. Neurocytol. 21, Kuang, R.Z., and Kalil, K. (1994). Development of specificity in corti- 185–197. cospinal connections by axon collaterals branching selectively into Takei, Y., Harada, A., Takeda, S., Kobayashi, K., Teradak, S., Noda, appropriate spinal targets. J. Comp. Neurol. 344, 270–282. T., Takahashi, T., and Hirokawa, N. (1995). Synapsin I deficiency Kuhn, T.B., Meberg, P.J., Brown, M.D., Bernstein, B.W., Minamide, results in the structural change in the presynaptic terminals in the L.S., Jensen, J.R., Okada, K., Soda, E.A., and Bamburg, J.R. (2000). murine nervous system. J. Cell Biol. 131, 1789–1800. Regulating actin dynamics in neuronal growth cones by ADF/cofilin Tanaka, E.M., and Kirschner, M.W. (1991). Microtubule behavior in and rho family GTPases. J. Neurobiol. 44, 126–144. the growth cones of living neurons during axon elongation. J. Cell Lin, C.H., and Forschner, P. (1993). Cytoskeletal remodeling during Biol. 115, 345–363. growth cone-target interactions. J. Cell Biol. 121, 1369–1383. Tanaka, E., Ho, T., and Kirschner, M.W. (1995). The role of microtubule dynamics in growth cone motility, axonal growth. J. Cell Biol. Liu, B.P., and Strittmatter, S.M. (2001). Semaphorin-mediated axonal 128, 139–155. guidance via Rho-related G proteins. Curr. Opin. Cell Biol. 13, 619–626. Teng, J., Takei, Y., Harada, A., Nakata, T., Chen, J., and Hirokawa, N. (2001). Synergistic effects of MAP2 and MAP1B knockout in Maney, T., Hunter, A.W., Wagenbach, M., and Wordeman, L. (1998). neuronal migration, dendritic outgrowth, and microtubule organiza- Mitotic centromere-associated kinesin is important for anaphase tion. J. Cell Biol. 155, 65–76. chromosome segregation. J. Cell Biol. 142, 787–801. Terashima, T., Kishimoto, Y., and Ochiishi, T. (1993). Musculotopic Maney, T., Wagenbach, M., and Wordeman, L. (2001). Molecular organization of the facial nucleus of the reeler mutant mouse. Brain dissection of the microtubule depolymerizing activity of mitotic cen- Res. 617, 1–9. tromere-associated kinesin. J. Biol. Chem. 276, 34753–34758. Tessier-Lavigne, M., and Goodman, C.S. (1996). The molecular biol- McNally, F.J., and Vale, R.D. (1993). Identification of katanin, an ogy of axon guidance. Science 274, 1123–1133. ATPase that severs and disassembles stable microtubules. Cell 75, 419–429. Walczak, C.E., Mitchison, T.J., and Desai, A. (1996). XKCM1: A Xeno- pus kinesin-related protein that regulates microtubule dynamics Miki, H., Setou, M., Kaneshiro, K., and Hirokawa, N. (2001). All kinesin during mitotic spindle assembly. Cell 84, 37–47. superfamily protein, KIF, genes in mouse and human. Proc. Natl. Acad. Sci. USA 98, 7004–7011. Waterman-Storer, C.M., and Salmon, E.D. (1998). How microtubules get fluorescent speckles. Biophys. J. 75, 2059–2069. Moore, A.T., and Wordeman, L.G. (2001). Regulation of mitotic centromere-associated kinesin. Mol. Biol. Cell 12 (suppl), 437a. Wordeman, L., and Mitchison, T.J. (1995). Identification and partial characterization of mitotic centromere-associated kinesin, a kinesin- Morfini, G., Quiroga, S., Rosa, A., Kosik, K., and Caceres, A. (1997). related protein that associates with centromeres during mitosis. J. Suppression of KIF2 in PC12 cells alters the distribution of growth Cell Biol. 128, 95–104. cone nonsynaptic membrane receptor and inhibits neurite extension. J. Cell Biol. 138, 657–669. Nakata, T., Terada, S., and Hirokawa, N. (1998). Visualization of the dynamics of synaptic vesicle and plasma membrane proteins in living axons. J. Cell Biol. 140, 659–674. Noda, Y., Sato-Yoshitake, R., Kondo, S., Nangaku, M., and Hirokawa, N. (1995). KIF2 is a new microtubule-based anterograde motor that transports membranous organelles distinct from those carried by kinesin heavy chain or KIF3A/B. J. Cell Biol. 129, 157–167. O’Leary, D.D., Bicknese, A.R., De Carlos, J.A., Heffner, C.D., Koester, S.E., Kutka, L.J., and Terashima, T. (1990). Target selection by cortical axons, alternative mechanisms to establish axonal connections in the developing brain. Cold Spring Harb. Symp. Quant. Biol. 55, 453–468. O’Leary, D.D., and Terashima, T. (1998). Cortical axons branch to multiple subcortical targets by interstitial axon budding; implications for target recognition and “waiting periods.” Neuron 1, 901–910. Ozon, S., Byk, T., and Sobel, A. (1998). SCLIP, a novel SCG10-like protein of the stathmin family expressed in the nervous system. J. Neurochem. 70, 2386–2396. Ozon, S., Guichet, A., Gavet, O., Roth, S., and Sobel, A. (2002). Drosophila stathmin, a microtubule-destabilizing factor involved in nervous system formation. Mol. Biol. Cell 13, 698–710. Quarmby, L. (2000). Cellular Samurai, katanin and the severing of microtubules. J. Cell Sci. 113, 2821–2827. Rakic, P., Knyihar-Csillik, E., and Csillik, B. (1996). Polarity of microtubule assemblies during neuronal cell migration. Proc. Natl. Acad. Sci. USA 93, 9218–9222. Santama, N., Krijnse-Locker, J., Griffiths, G., Noda, Y., Hirokawa, N., and Dotti, C.G. (1998). KIF2␤, a new kinesin superfamily protein in nonneuronal cells, is associated with lysosomes and may be implicated in their centrifugal translocation. EMBO J. 17, 5855–5867. Shelanski, M., Gaskin, F., and Cantor, C. (1973). Microtubule assem-