The ability of neurons to generate multiple arbor terminals from a single axon is crucial for establishing proper neuronal wiring. Although growth and retraction of arbor terminals are differentially regulated within the axon, the mechanisms by which neurons locally control their structure remain largely unknown. In the present study, we found that the kinesin-1 (Kif5 proteins) head domain (K5H) preferentially marks a subset of arbor terminals. Time-lapse imaging clarified that these arbor terminals were more stable than others, because of a low retraction rate. Local inhibition of kinesin-1 in the arbor terminal by chromophore-assisted light inactivation (CALI) enhanced the retraction rate. The microtubule turnover was locally regulated depending on the length from the branching point to the terminal end, but did not directly correlate with the presence of K5H. By contrast, F-actin signal values in arbor terminals correlated spatiotemporally with K5H, and inhibition of actin turnover prevented retraction. Results from the present study reveal a new system mediated by kinesin-1 sorting in axons that differentially controls stability of arbor terminals.
To connect with multiple target cells, neurons elaborate the axonal arbor by controlling growth and retraction during development (Gibson and Ma, 2011; Kalil and Dent, 2014). Previous in vivo (Portera-Cailliau et al., 2005; Hua et al., 2005; Meyer and Smith, 2006; Stettler et al., 2006; Nishiyama et al., 2007) and in vitro (Bastmeyer and O'Leary, 1996; Ruthel and Hollenbeck, 2000) studies have revealed the diverse plasticity of arbor terminals in a single axon. The competition between different terminals of a single axon in cultured neurons (Hutchins and Kalil, 2008) suggests that neurons possess an intracellular system that coordinates the branched axonal shape by regulating growth and retraction of arbor terminals in a region-specific manner. Despite intensive studies on the axonal arborization, little is known about the intracellular mechanisms mediating terminal-dependent control of growth and retraction in the axonal arbor.
In collateral branch formation, filopodial and lamellipodial protrusions, which contain F-actin bundles, emerge from actin patches found along the axon shaft. Newly formed branches are then invaded by microtubules to become mature axonal branches (Gallo and Letourneau, 1999; Gallo, 2011; Lewis et al., 2013). It is thought that the balance between stabilization and destabilization of F-actin and microtubules determines the formation and growth of axonal branches. For example, activation of the actin nucleation factors, Arp2/3 and Cordon-bleu (Cobl) is required for filopodia formation from the actin patch, and inhibition of this pathway results in decreased branch number (Strasser et al., 2004; Ahuja et al., 2007; Spillane et al., 2011, 2012). The microtubule-severing enzymes, spastin and katanin, which provide a branching point at the axonal shaft, also control the number of axonal branches (Yu et al., 2008). In contrast, neurons that lack Kif2a, a kinesin family molecule that depolymerizes microtubules, exhibit increased axonal branch length (Homma et al., 2003). Thus, local control of F-actin and microtubule dynamics is crucial for shaping the axonal arbor. Nevertheless, intracellular systems that regulate individual arbor terminals in the axon have not been fully demonstrated.
Ca2+ is a major intracellular signaling molecule that regulates axonal growth (Kater and Mills, 1991; Gomez and Spitzer, 1999; Tang and Kalil, 2005; Ageta-Ishihara et al., 2009). Previous studies have suggested that localized Ca2+ transients in restricted areas in the arbor enhance axonal growth, accompanied by retraction of other terminals (Hutchins and Kalil, 2008). Cyclic nucleotides are thought to be candidates that send long-range signals mediating the competition (Hutchins, 2010), given that in axonal specification, the activation of cAMP in a single neurite mutually inhibits cAMP in other neurites to form a single axon (Shelly et al., 2010). The molecular mechanisms whereby these signals coordinately regulate extension or retraction of different arbor terminals have not been elucidated.
Kinesin-driven anterograde transport along microtubules has been shown to play crucial roles in maintaining axonal morphology and function (Goldstein, 2001; Hirokawa and Takemura, 2005; Salinas et al., 2008; Maday et al., 2014). Conventional kinesin (kinesin-1) consists of heavy chain (Kif5a, Kif5b or Kif5c) dimmer and light chains. Kif5 proteins contains a motor domain (also called the head domain) that hydrolyzes ATP and moves along microtubules, whereas the C-terminal tail domain is required for cargo interactions. Because the tail domain inhibits motor function when not bound to cargos, deleting the tail domain results in constitutive activation of Kif5 (Coy et al., 1999). Intriguingly, the cleaved motor (head) domain of Kif5 (K5H) is selectively delivered to axons, but not dendrites (Nakata and Hirokawa, 2003; Jacobson et al., 2006), indicating that K5H has the ability to discriminate between axons and dendrites. The control of microtubule dynamics has been shown to depend on the region within the single cell (Witte et al., 2008), and microtubule dynamics affect tubulin states, such as GDP or GTP binding, acetylation and tyrosination or detyrosination, and decoration by microtubule-associated proteins (MAPs). Recent results from several reports, including from our group, have suggested that K5H recognizes differences between microtubules on axon and dendrites (Reed et al., 2006; Konishi and Setou, 2009; Hammond et al., 2010; Nakata et al., 2011). Nevertheless, very little is known about the regulation of axonal transport within the arbor. Therefore, in the current study, we asked whether there is any difference in efficiency of kinesin-1-dependent axonal transport between terminals of a single axonal arbor. Results showed a selective K5H accumulation in a subset of axonal terminals, which contributed to the shape of the axonal arbor by differentially regulating retraction.
Kinesin-1 preferentially accumulates in a subset of axonal terminals
To study terminal dependency of kinesin-1-mediated axonal transport, we utilized cerebellar granule neurons (CGNs) that possess relatively simple axonal arbors (Bilimoria et al., 2010; Ito-Ishida et al., 2012). In culture, CGNs extend one or two axons that possess several arbor terminals (Bilimoria et al., 2010; Kubota et al., 2013) (Fig. 1A), thus are thought to be suitable for studying the diversity of arbor terminals. During the axonal arborization of CGNs in vitro, elimination of main axon terminal as well as massive extension of collateral branch can be observed (Fig. S1). In the current study, the analyses were carried out without distinguishing between the main axonal process and branches formed through collateralization or bifurcation.
CGNs introduced with an expression vector for K5H–EGFP (Kif5c head fused to EGFP) were fixed at 7 DIV. As previously described (Nakata and Hirokawa, 2003; Jacobson et al., 2006), K5H selectively accumulates in the axon terminals (Fig. 1A). The signal intensity of K5H in each terminal varied substantially, even in the same axonal arbor (Fig. 1A, arrowheads). To objectively evaluate K5H signal variation, signals at the distal terminal pairs were quantified and relative signal levels were calculated. Simultaneously, the relative length of terminal segment (from the branching point to the terminal end) was calculated as shown in Fig. 1B, and terminal segments were categorized merely based on the length. If K5H molecules are randomly delivered into two neighboring terminals, the distribution of relative signal values should follow a normal distribution. The quantile–quantile plot of sample quantiles of relative K5H values versus theoretical quantiles did not lie on a straight diagonal line, because some fractions revealed low (close to 0) or high (close to 1) relative K5H values (Fig. 1C), indicating that data were not normally distributed. The scatter plot of relative K5H values versus the relative length of the same terminal pair revealed a mild, but significant, positive correlation (Fig. 1D, r=0.62, P<0.001, n=64). These results indicate that K5H is preferentially delivered into a subset of arbor terminals that tend to be longer than neighboring terminals in the CGNs.
We also cultured CGNs at a low density (Kubota et al., 2013), and subjected the cells to immunocytochemistry (at 5 DIV) using antibodies specific to kinesin-1 (Fig. 1E) as previously described (Konishi and Setou, 2009). Results showed that endogenous kinesin-1 signals in longer terminal segments were greater than in shorter terminal segments (Fig. 1E, P<0.01, n=12), suggesting that results observed by using K5H represent, at least in part, the nature of endogenous kinesin-1-mediated axonal transport. In addition, when we disrupted kinesin-1-mediated axonal transport by expressing the cargo-binding (tail) domain (K5T), we found that the length from branching point to axonal terminal became shorter compared with control neurons introduced with empty vector or vector for stalk region (K5S) (Fig. 1F, P<0.05, n=37). These results support our notion that selective delivery of kinesin-1 in a subset of arbor terminals might coordinate the shape of axonal arbors by controlling the length of terminal segments.
Retraction is inhibited in axonal terminals with accumulated K5H
To determine whether kinesin-1-mediated transport efficiency in arbor terminals correlates with elongation or stability, we simultaneously observed changes in axonal morphology and K5H distribution within the axonal arbor by fluorescence time-lapse imaging. Although a majority of arbor terminals were stable, both growth and retraction occurred in a terminal-dependent manner at 5 DIV (Fig. 2A,B). K5H signals stably accumulated in a subset of arbor terminals in most cases, but signals occasionally changed location to the other terminals. The rate of terminal growth was determined by differentiating the length of the terminal segment at each time period. The values for elongation (i.e. positive growth) and retraction (i.e. negative growth) for each arbor terminal were then extracted. The scatter plot revealed that the average rate of elongation and retraction largely varied depending on the terminal (Fig. 2C,D). Comparisons of the average elongation rate between terminals with a lower amount of K5H (i.e. relative K5H value <0.1, n=77) and K5H-enriched terminals (i.e. relative K5H value >0.1, n=20) revealed a significant median shift (0.17 μm/h in low-K5H versus 0.75 μm/h in high-K5H terminals, P<0.001, Mann–Whitney U-test). Nevertheless, the relationship between K5H signal values and elongation rates remains unclear, because the a high rate of elongation frequently occurred even in K5H-deprived arbor terminals, and also was not remarkable in terminals that exhibited very high K5H signals (Fig. 2C). Additionally, some arbor terminals deprived of K5H exhibited no growth rate, because they were removed by retraction within a short period of time. Conversely, the average retraction rate was consistently low in K5H-enriched arbor terminals (Fig. 2D), with statistical significance differences in the median (0.90 μm/h in lower K5H versus 0.57 μm/h in higher K5H, P<0.05, Mann–Whitney U-test). Consequently, although K5H-deprived terminal segments tended to become shorter, K5H-enriched terminal segments became longer or kept their length (Fig. 2E). Following these observations, we decided to focus our analysis on the relationship between the differential terminal retraction and efficiency of kinesin-1-mediated axonal transport in the arbor.
Local inhibition of kinesin-1-enhanced retraction rate of arbor terminals
We then applied chromophore-assisted light inactivation (CALI) to inhibit kinesin-1 activity in a terminal-dependent manner. KillerRed (KR) has been used as a genetically encoded photosensitizer for CALI (Destaing et al., 2010; Baumgart et al., 2012; Sano et al., 2014). To inhibit endogenous kinesin-1, CGNs were expressed with fusion protein consisting of KR and the kinesin light chain KLC1 (KR–KLC). KR–KLC fluorescence within a small circular area can be selectively bleached by illumination with a mercury lamp through the iris. An ∼60% and ∼80% loss of KR–KLC signal was observed within 30 s and 60 s of exposure, respectively (Fig. 3A,B). At the axonal terminal segment, KR and KR–KLC were distributed throughout the process (Fig. 3C,D). Unlike K5H, it is expected that KR–KLC proteins exist in various states (e.g. as free forms and as a kinesin complex binding to various type of cargos). We illuminated the axonal terminal and monitored fluorescence recovery. Recovery of the KR signal reached a maximum in the first time-point analyzed (15 min) after bleaching, at which time half of signal relative to the pre-illumination was detected, and subsequently decayed (Fig. 3C,E). In addition, a reduction in the KR signal was observed even outside of the illuminated area. We conclude that a part of signal was lost during the photobleaching and subsequent imaging, and fast recovery of KR represents fast diffusion of free KR proteins. By contrast, recovery of KR–KLC at 15 min was smaller compared than recovery of KR, then slowly increased and reached a similar level at 3 h (Fig. 3D,E).
We applied CALI to one side (longer terminal segment) of a terminal pairs of neurons expressing KR or KR–KLC together with EGFP. Subsequently, the effect of CALI application on the arbor terminals was examined by monitoring EGFP signals for 8 h (Fig. 3F; Fig. S2). Given that in a preliminary experiment, 60 s of illumination caused immortalization or disintegration of arbor terminal, we applied 30 s of exposure in this analysis to reduce photo toxicity. We measured the difference in length of both the target terminal segment and the neighboring un-illuminated terminal segment at 15-min and intervals of 15 min thereafter. Most arbor terminals exhibited a limited retraction after CALI exposure in KR-expressing CGNs, whereas in KR–KLC-expressing CGNs, intermittent retraction was only observed at the arbor terminal exposed to CALI (target terminal) (Fig. 3F; Fig. S2). Although there was no clear time window that the boost of retraction occurred, retraction was not remarkable in the last few hours of the analysis (Fig. S2). Given that the KR–KLC signal recovered to a similar level to that of KR in 3 h after photobleaching, we calculated the amount of retraction in 3 h (Fig. 3G). Quantification of terminal retraction values revealed greater retraction in KR–KLC-expressing CGNs (n=18) compared with KR-expressing control neurons (n=19) at target terminals (Fig. 3G, P<0.05, Mann–Whitney U-test). In contrast, retraction of the neighboring shorter terminal segment that did not receive photo illumination (neighboring terminal) was not greater in KR–KLC-expressing CGNs than in control neurons (P=0.11).
Microtubule turnover is differentially regulated between neighboring terminal segments
To determine whether K5H accumulation in a subset of axonal terminals reflects a variation in microtubule content between terminal segments, we expressed mCherry–tubulin and K5H–EGFP in CGNs because introduction of the K5H–EGFP plasmid requires a high-density culture, which makes it difficult to detect endogenous tubulin in a single axonal arbor by immunocytochemistry. In order to estimate the amount of microtubules with this method, signals from mCherry–tubulin that is incorporated in microtubules have to be separated from those of free mCherry–tubulin. To remove unpolymerized free tubulin, neurons were subjected to simultaneous extraction and fixation at 7 DIV (Fig. 4A). Signals at the terminal pairs were quantified, and relative signal levels were calculated (Fig. 4B). In some of the terminal pairs, the longer terminal segment enriched with K5H exhibited greater mCherry–tubulin signals (Fig. 4A). However, in other terminal pairs, greater K5H accumulation was observed in terminal segments with less mCherry–tubulin (Fig. 4A). Consequently, the overall correlation between K5H–EGFP values and mCherry–tubulin signals in the same terminal pair was not significant (Fig. 4B, r=0.30, P=0.33, n=44). These results indicate that factors other than microtubule content might contribute to the preferential accumulation of K5H in longer terminal segments.
Previous studies focused on neuronal polarity have suggested that kinesin-1 is preferentially transported in processes that contain acetylated and detyrosinated microtubules (Reed et al., 2006; Konishi and Setou, 2009; Hammond et al., 2010). Stable microtubules contain more detyrosinated and acetylated tubulins than dynamic microtubules (Janke and Bulinski, 2011). Therefore, we analyzed microtubule turnover in axonal arbors. We applied fluorescence recovery after photobleaching (FRAP) by expressing an α-tubulin tagged with EGFP (EGFP–tubulin). In CGNs, EGFP–tubulin was detected throughout axonal arbors (Fig. 4C). We focused on pairs of arbor terminals and applied photobleaching in a rectangular area proximal (approximately 10 μm) to the branching point (Fig. 4C). Recovery of fluorescence in each terminal segment was measured at 40-s intervals. Because free tubulins are rapidly recovered in a matter of seconds, fluorescence recovery over time reflects mostly the turnover of EGFP–tubulins that exist as microtubules. Notably, when FRAP analysis was applied on terminal segments that possessed varying segment lengths, we found that recovery of EGFP–tubulin in longer axonal terminal segments was slower than in shorter ones (Fig. 4D). The dissociation rate of EGFP–tubulin, which was obtained by a mathematical fitting to the model (Edson et al., 1993; Hush et al., 1994; see Materials and Methods), revealed that turnover of EGFP–tubulin was slower in longer terminal segments compared with shorter terminal segments (Fig. S3A, shorter terminal segment; koff=0.055±0.019/min versus longer terminal segment; koff=0.017±0.006/min, mean±s.e.m., n=4).
We also assessed the possibility that selective kinesin-1 transport directly correlates with microtubule stability in the terminal segments. In this experiment, FRAP was applied to CGN axons that simultaneously expressed EGFP–tubulin and K5H–mCherry. As descried above, the length of the K5H-enriched terminals tended to be longer than neighboring terminal segments (Fig. 1D). To investigate the direct correlation between K5H signal and turnover of EGFP–tubulin, we selected terminal pairs that were comparable in segment length (<1.5-fold difference). When we analyzed terminal pairs that possessed different amounts of K5H (>2-fold), we found no significant correlation between EGFP–tubulin recovery and K5H signal value (Fig. 4E,F). In addition, no significant correlation was observed between the K5H signal value and the dissociation rate of EGFP–tubulin (Fig. S3B).
Correlation between length of terminal segment and reporters of microtubule stability
We next performed immunocytochemistry by using antibodies that specifically recognize post-transcriptionally modified tubulins (Fig. 5A). The ratios of tyrosinated to detyrosinated tubulins and acetylated tubulins varied between neighboring terminal segments with different lengths. Consistent with a previous report (Robson and Burgoyne, 1989; Ahmad et al., 1993; Brown et al., 1993), the ratio of modification gradually changed depending on the distance from the branching point (Fig. 5A). To account for region-dependent differences in tubulin modification within the terminal segment, we quantified staining profiles for distances from axonal branching points and distances from terminal ends (Fig. 5B,C). Notably, the ratio of tyrosinated to detyrosinated tubulin was greater in shorter terminal segments compared with longer terminal segments at the same distance from a branching point. The ratio difference increased in a distance-dependent manner, but was significant even at 10 μm from the branching point (Fig. 5C, P<0.05, n=11). These results were consistent with the FRAP analysis. Similarly, the ratio of acetylated tubulin in shorter terminal segments was significantly less than the ratio at the same position in longer terminal segments relative to the branching point (Fig. 5C, P<0.05 at 10 μm, n=8). Conversely, in both tyrosination and detyrosination and acetylation, staining profiles versus distance from terminal ends were similar between shorter and longer terminal segments (Fig. 5C).
We further investigated the region-dependent frequency of microtubule growth in the arbor by introducing an end-binding 3 (EB3, also known as MAPRE3) plasmid that was fused to fluorescent proteins to label plus-ends of growing microtubules (Stepanova et al., 2003). EB3 particles were detected more frequently in distal regions compared with proximal regions, and shorter terminal segments contained more particles compared with longer terminal segments at the same distance from a branching point (Fig. 5D, P<0.05, n=9). Additionally, profiles of EB3 particle numbers versus distance from terminal ends were similar between shorter and longer terminal segments (Fig. 5D), which was consistent with analysis of tyrosination and detyrosination and acetylation of microtubules. Thus, differences in microtubule turnover between short and long terminal segments could be explained by a regional difference that is dependent on the distance from a terminal end. In a previous study, Seetapun and Odde (2010) provided a simple length-dependent model for selective accumulation of stable microtubules in axons during polarization, which does not require neurite-dependent control of a microtubule assembly. In this model, the growth rate of microtubules in minor and longer processes remained unaltered. However, because the traveling time depends on process length, the net turnover of microtubules is slower in longer processes (i.e. axons) compared with shorter minor processes. We measured microtubule growth in axonal arbors by simultaneously monitoring EB3 particles in shorter and longer terminal segments (Fig. 5E). We found no length-dependent difference in the velocity of EB3 particles (Fig. 5E). These results support the notion that microtubule turnover in terminal segments is differentially regulated and dependent on distance from the terminal end, rather than by segment-specific microtubule regulation.
F-actin distribution in the axonal arbor spatiotemporally correlates with K5H
We next investigated the relationship between F-actin and K5H at arbor terminals. F-actin accumulates in the growth cone and is required to form filopodia and lamellipodia (Lewis et al., 2013). In axons of young CGNs (at 2 DIV), strong F-actin signals were observed in the growth cone by staining with phalloidin. However, at 5 DIV, a time point by which axons have become thinner and form multiple terminals, F-actin signals at the axonal tip were highly restricted (Fig. S4A). To analyze F-actin distribution in detail in CGN axons, we expressed Lifeact, a marker for the visualization of F-actin (Riedl et al., 2008). In the axonal arbor, Lifeact signals colocalized with phalloidin signals at arbor terminals and protrusions (Fig. S4B). When we simultaneously expressed K5H and Lifeact in neurons, they colocalized in the axonal terminal (Fig. 6A). Plotting of relative Lifeact signal values versus relative K5H signals in terminal pairs clearly revealed a positive correlation (Fig. 6B, r=0.83, n=20). Consequently, arbor terminals with greater K5H local levels also had significantly greater Lifeact signals (Fig. 6B, P<0.001, n=15). To analyze the temporal correlation, axonal arbors expressing K5H and Lifeact were subjected to time-lapse imaging analysis at 10-min intervals. In some arbor terminals, the K5H signal dramatically decreased or increased during the time-lapse period (Fig. 6C). In these arbor terminals, K5H signal changes were followed by changes in Lifeact signal values (Fig. 6C,D). We collected data from arbor terminals in which the K5H signal changed >twofold within 100 min. When the Lifeact signals between the two different time points (i.e. maximum K5H versus minimum K5H) were compared, we found a significant difference (Fig. 6D, P<0.001, n=5). These results suggested a spatiotemporal correlation between the K5H and Lifeact signals.
To determine whether F-actin regulation plays a role in the retraction that frequently occurs in K5H-deprived arbor terminals, we utilized Latrunculin A, which has been shown to inhibit axonal retraction in sensory neurons by disrupting actin polymerization (Ahmad et al., 2000; Gallo et al., 2002). Neurons that express K5H–EGFP and mCherry were treated with Latrunculin A prior to time-lapse imaging. Changes in axonal morphology and distribution of K5H–EGFP were simultaneously analyzed for 5 h to determine the primary effect of inhibiting actin turnover. We observed that Latrunculin A-treated arbor terminals (n=21) exhibited a lower average retraction rate compared with vehicle-treated control arbor terminals (n=26) (Fig. 6E, median; 0.80 μm/h in DMSO versus 0.49 μm/h Latrunculin A, P<0.05, Mann–Whitney U-test). Taken together, these results indicate the possibility that actin-dependent mechanisms are involved in the axonal retraction that occurs in kinesin-1-deprived arbor terminals.
The intracellular systems involved in how neurons differentially regulate growth and stability of arbor terminals within the same axon remain largely unknown. Previous results have indicated that intracellular Ca2+ mediates the competitive signal between different terminals of the same axon (Hutchins and Kalil, 2008). However, the mechanisms by which localized Ca2+ signals regulate the axonal structure in a region-specific manner remain to be elucidated. Other molecules that are differentially controlled between axonal terminals need to be identified. In this study, we found that K5H tagged by fluorescent protein, as well as endogenous kinesin-1, accumulate in a subset of axonal arbor terminals (Fig. 1). The quartile–quartile plot analysis of K5H signals in arbor terminals revealed that the differences in K5H signals between terminals are not due to random variation. We only analyzed terminal segments that had a length >20 μm to avoid collecting data form newly formed protrusions that had not yet been invaded by microtubules (Gallo, 2011). Furthermore, time-lapse imaging revealed that K5H signals occasionally and rapidly changed their distribution, but that local levels of K5H did not increase in the subset of arbor terminals that exhibited a long lifespan. These findings revealed selective sorting of kinesin-1 in the axonal arbor. Time-lapse imaging analysis revealed that arbor terminals with high local levels of K5H exhibited a small retraction (Fig. 2). Consistently, local inhibition of kinesin-1 by CALI enhanced retraction specifically at the illuminated arbor terminal (Fig. 3). Considering the function of kinesin-1-mediated axonal transport, the accumulation of cargo molecules in particular arbor terminals would be expected to result in the deprivation of molecules in neighboring terminals. Results from the present study suggest a potential intracellular system for competitive stabilization of arbor terminals via kinesin-1 sorting in the axonal arbor. Additionally, changes in the length of terminal segment could provide feedback to kinesin-1-dependent transport as described below.
The mechanisms by which kinesin-1 gets sorted into a subset of arbor terminals from the axon remain to be clarified. Our results showed that microtubules in longer terminal segments are more stable than those in neighboring shorter terminal segments when compared in the area proximal to the branching point (Fig. 4). Consistently, posttranslational modifications of tubulins abundant in stable microtubules (i.e. detyrosination and acetylation) are greater in longer axonal terminal segments (Fig. 5). Thus, kinesin-1-dependent terminal selection could be explained by preferential recruitment to stable microtubules as described in other paradigms (Reed et al., 2006; Hammond et al., 2010; Dunn et al., 2008). In a previous study, we have reported that inhibition of tubulin tyrosination in immature hippocampal neurons disrupts polarized K5H transport into axons (Konishi and Setou, 2009). We also found that inhibition of tubulin tyrosination by siRNA for tubulin tyrosination ligase (TTL) tended to inhibit the terminal-dependent accumulation of K5H in CGNs, although no statistical significance was obtained (data not shown). Thus, the function of tyrosination–detyrosination alone might be limited in this paradigm. It is possible that other differences (e.g. MAPs, GTP- or GTP-binding state, microtubule numbers) could also contribute to this process, because they are possibly affected by microtubule dynamics. Studies focused on neural polarity have shown the contribution of multiple factors that can be differentially coordinated dependent on the situation and neuronal types (Barnes and Polleux, 2009). Further studies are required to fully demonstrate the molecular mechanisms of kinesin-1 sorting in the axonal arbor.
Microtubules undergo repetitive growth and shortening until they reach the axonal terminal, where most microtubules undergo shrinkage (catastrophe). According to the polarization model by Seetapun and Odde (2010), because of increased traveling time from the soma, microtubule turnover is slower in the axon due to process length. Under certain conditions, the region closer to the terminal contains a larger number of growing microtubule plus-ends than regions further from the terminal. Consequently, the longer process contains older microtubules than shorter minor processes. Although microtubules in mature axons are segmented (Yu and Baas, 1994), our observations suggest that this model can be adopted to explain microtubule difference in axonal arbors. First, there is a positive correlation between the length of the terminal segment and microtubule turnover. Second, the distribution of reporters for microtubule turnover depends on distance from terminal end. Third, there is no segment-dependent difference in velocity of EB3 particles. Obviously, other factors could also regulate microtubules, either stochastically or in a region-specific manner through cleavage or stabilization (Yu et al., 2008; Homma et al., 2003; Peris et al., 2009). Indeed, we found only a moderate correlation between K5H signals and length of terminal segment, and K5H distribution can be suddenly altered prior to a change in length of terminal segment. These factors might play crucial roles in switching between a stable and unstable state in arbor terminals.
As described above, longer terminal segments contain more stable microtubules and tend to contain greater K5H signals. Previous studies have reported that kinesin-1 transports molecules involved in the stabilization of microtubules (Kimura et al., 2005). However, microtubule turnover in terminal pairs with comparable segment lengths, but differences in K5H signals, was not remarkably different. This observation contradicts the notion that kinesin-1-dependent inhibition of retraction is mediated by selective stabilization of microtubules. Conversely, there was a spatiotemporal correlation between K5H and F-actin signals in arbor terminals, and disruption of axonal turnover with Latruculin A significantly inhibited retraction (Fig. 6), even in arbor terminals with low K5H signals. Actin regulation has been shown to play an important role in axonal growth, as well as retraction (Giannone et al., 2009). Retraction cues induce growth cone collapse by destabilizing F-actin in the axonal terminal. Conversely, RhoA- and myosin-II-driven contractility along intra-axonal F-actin mediates axonal retraction (Gallo et al., 2002; Gallo, 2006). Therefore, reorganization of F-actin is crucial for the retracting axon to change its structure. Although detailed molecular mechanisms remain to be elucidated, our current study raised a possible link between the differential control of terminal stability and actomyosin-mediated axonal retraction.
MATERIALS AND METHODS
Dissociated CGN culture was prepared according to the previous study (Konishi et al., 2004) with modifications. In brief, cerebella isolated from Slc:ICR mice (postnatal days 5–6) were digested with trypsin, and dissociated neurons were placed in Basal Medium Eagle (BME; Sigma-Aldrich, St Louis, MO) supplemented with 10% calf serum (Thermo Fisher Scientific, Waltham, MA), 1 mg/ml penicillin, 1 mg/ml streptomycin, 2 mM glutamine and 25 mM KCl. Neurons were spread on a plate that has been coated with poly-L-ornithine (Sigma-Aldrich). Low-density CGN cultures were prepared by co-cultivating with high-density CGN culture as described previously (Kubota et al., 2013). Animals were treated according to the institutional ethical guidelines of University of Fukui.
Plasmid DNAs were introduced in CGNs by the calcium phosphate method. Prior to the transfection, CGN culture was washed twice with Dulbecco's modified Eagle's medium (DMEM; Sigma-Aldrich), then placed in DMEM and incubated at 37°C in a CO2 chamber. DNA solution containing 250 mM CaCl2, was gradually mixed with same amount of 2× HBS solution (270 mM NaCl, 9.5 mM KCl, 1.4 mM NaH2PO4, 15 mM glucose, 42 mM Hepes). After 15 min, the DNA mixture was added to cells and the cells were incubated for 15 min in CO2 incubator. Finally, the CGN culture was washed twice with DMEM and placed in original medium. In the case of low-density culture, dissociated neurons were placed in DMEM that contained plasmids, and exposed to the square electric pulses (140 V, 5 ms for two times) before plating as described previously (Kubota et al., 2013). To construct a plasmid for KR–KLC, full-length KLC1 cDNA was amplified by PCR, and inserted into pKillerRed-C vector (Evrogen, Moscow, Russia). Plasmids for EGFP–tubulin (#30487, gift from Tso-Pang Yao), mCherry–tubulin (#26768, gift from Torsten Wittmann), tdTomato–EB3 (#58090, gift from Michael Davidson) and pLifeAct-mTurquoise2 (#36201, gift from Dorus Gadella), were obtained from Addgene. Plasmids for EGFP–EB3, K5H–EGFP and dominant-negative Kif5s were as described previously (Konishi and Setou, 2009; Okamoto et al., 2015). K5H–mCherry was constructed by inserting the K5H sequence into pmCherry-N vector (BD Biosciences, Franklin Lakes, NJ). To knockdown TTL, a stealth siRNA against TTL (#RSS301756, Thermo Fisher Scientific) were transfected into CGNs together with plasmids for K5H–EGFP and mCherry by the calcium phosphate method as described above.
For immunocytochemistry, neurons were fixed for 15 min at room temperature with paraformaldehyde solution (4% paraformaldehyde in PBS) followed by permeabilization with 0.2–0.4% Triton X-100 in PBS for 15 min, and incubated further (1 h) with blocking solution (5% goat serum, 3% BSA and 0.02% Tween 20 in PBS). Cells were then incubated with primary antibodies in blocking solution at 4°C for overnight, followed by incubation with secondary antibodies (2–3 h, at room temperature). To stain the nucleus and F-actin, Hoechst 33258 (Sigma-Aldrich) and Rhodamine–phalloidin (Cytoskeleton Inc., Denver, CO) were used, respectively. Monoclonal antibody against α-tubulin (1:1000; 12G10) was obtained from Developmental Studies Hybridoma Bank of University of Iowa. Monoclonal antibodies against β-tubulin (1:1000; Tub2.1, Cy3-conjugated, #C4585), tyrosinated tubulin (1:2000; TUB-1A2, #T9028) and acetylated tubulin (1:10000; 6-11B-1, #T6793) were obtained from Sigma-Aldrich. Antibodies against detyrosinated tubulin (1:1000; #AB3201) and kinesin heavy chain (1:50; clone H2, #MAB1614) were purchased from Merck Millipore (Darmstadt, Germany). For secondary antibodies, goat anti-mouse- or anti-rabbit-IgG conjugated to Alexa Fluor dyes (1:1000; Thermo Fisher Scientific) was used.
Cell imaging and data quantification
Images of fixed neurons were obtained under an Axiovert 200 M microscope equipped with Axiocam MRm digital camera (Carl-Zeiss, Oberkochen, Germany), and were analyzed using ImageJ software (National Institutes of Health). In some experiments, a confocal laser scanning microscope (LSM 5 PASCAL; Carl-Zeiss) or ApoTome.2 (Carl-Zeiss) was used to obtain sectionized images (see below). In the analysis of arbor terminals, processes that were more than 20 μm in length were selected to exclude short protrusions.
Time-lapse imaging of the axonal arbor
For time-lapse imaging, neurons were cultured on glass bottom plates (Iwaki, Chiba, Japan) attached by flexiperm (Sarstedt, Nümbrecht, Germany). Minimal Essential Medium (MEM; Thermo Fisher Scientific) that did not contain Texas Red was used instead of BME. Images were acquired at 10–20 min intervals by using an Axiovert 200 M equipped with MRm monochromatic digital camera (Carl Zeiss). During the imaging, neuron culture was kept at 36.7°C in the stage top incubator (ZILCS; Tokaihit, Shizuoka, Japan), which was supplied with 5% CO2, set on the scanning stage. Stage position and camera were controlled by AxioVison software (Carl Zeiss). Images were collected randomly from neurons whose axonal morphology was clearly observed, and degenerating neurons were excluded from the analysis. Axonal growth rate was calculated by differentiating the value of the segment length in each frame. A positive and negative change was defined as ‘elongation’ and ‘retraction’, respectively. In the inhibition of actin turnover, Latrunculin A (WAKO, Osaka, Japan) was added to the medium prior to placing the culture in the chamber.
Chromophore-assisted light inactivation
CGNs were introduced with expression vectors for KR or KR–KLC together with an EGFP plasmid at 2 DIV, and subjected to CALI at 4 or 5 DIV. For CALI experiments, cells were placed in the stage-top incubator, and a small circular area defined by the iris (∼100 μm diameter) was illuminated with a 100 W mercury arc lamp (HBO 100) through a 20× objective lens (Plan-Apochromat, NA 0.8) and a band pass filter (Ex BP/565/30) (Carl-Zeiss). To analyze the stability of the arbor terminal, more than half the area from the terminal end of the longer terminal segment was illuminated for 30 s, and monitored at 15-min intervals for 8 h. Time-lapse images of all axonal arbors that did not undergo degeneration were quantified. To remove any position shift, a time-series of axonal arbor images was aligned using StackReg plugin (Philippe Thévenaz, Swiss Federal Institute of Technology Lausanne) of ImageJ before measuring an image-to-image difference in the length of terminal segment.
Analysis of microtubule content
To analyze microtubule content, mCherry–tubulin was expressed in neurons. Neurons were then simultaneously fixed and permeabilized as described previously (Witte et al., 2008) in a solution containing 60 mM Pipes pH 6.9, 25 mM Hepes, 5 mM EGTA, 1 mM MgCl2, 0.25% glutaraldehyde, 3.7% paraformaldehyde, 3.7% sucrose and 0.1% Triton X-100, for 20 min at room temperature in order to remove unpolymerized tubulins. The mCherry–tubulin signals in each terminal segment (in the 20 μm from the branching point) were quantified.
FRAP analysis of tubulins
For FRAP analysis of tubulins, a confocal microscope (LSM 5 PASCAL, Carl Zeiss) equipped with a 488-nm argon laser was used. CGNs transfected with expression plasmids for EGFP–tubulin and with or without K5H–mCherry was placed in the stage-top incubator as described above. A rectangular region in the pair of axonal terminal segments that was ∼10 μm distal from the branching point was analyzed. After photobleaching the region of interest with the 63× objective lens, fluorescence recovery was monitored every 40 s. Data were quantified by using ImageJ. The fluorescence recovery rate was calculated from a single exponential function, as described previously (Hush et al., 1994) by following Eqn 1, where frap(t) is the recovered fluorescence at each time point and Ceq is the total amount of complex. The dissociation constant, koff was calculated from plots by fitting and used to calculate the half-time of recovery (2). (1) (2)
CGNs (5 DIV) transfected with EB3 plasmids by the calcium phosphate method were analyzed. For the fixation, neurons were immediately placed in cold methanol for 20 min and rehydrated with PBS. For the live-cell imaging, neurons were placed in the stage-top incubator, and fluorescence images of EB3 particles was obtained with the ApoTome.2 microscope and MRm camera at 3-s intervals under the 63× objective lens. Velocity of EB3 spot was quantified by using Multiple Kymograph plugin (J. Rietdorf and A. Seitz, European Molecular Biology Laboratory) of ImageJ.
Compiled data in bar graphs are expressed as mean±s.e.m. We used the two-tailed Student's t-test for statistical analysis unless otherwise stated. The levels of significance are indicated as follows: *P<0.05, **P<0.01, ***P<0.001.
We are grateful to Dr Kazuyuki Murase (University of Fukui) and Dr Mitsutoshi Setou (Hamamatsu University School of Medicine) for support, and Dr Tokuichi Iguchi (University of Osaka) for the advice with EB3 experiments. We also thank Drs Tso-Pang Yao, Torsten Wittmann, Michael Davidson and Dorus Gadella for providing constructs.
The authors declare no competing or financial interests.
T.S., K.M., N.S. and Y.K. analyzed axonal morphology and microtubules, and T.I. and M.K. analyzed F-actin. H.T. contributed to the data analysis, and M.S. and Y.K. designed the experiments and wrote the paper.
This work was supported by a Ministry of Education, Culture, Sports, Science, and Technology (MEXT) KAKENHI Grant-in-Aid for Scientific Research on Innovative Areas, Nanomedicine Molecular Science [grant numbers 24107509 and 26107707 to Y.K.].
Supplementary information available online at http://jcs.biologists.org/lookup/doi/10.1242/jcs.183806.supplemental
- Received November 23, 2015.
- Accepted August 2, 2016.
- © 2016. Published by The Company of Biologists Ltd