Introduction

Axonal transport is fundamental to neuronal function (Hirokawa et al. 2010). Kinesin superfamily proteins (KIFs) were originally identified as molecular motors for axonal transport (Vale et al. 1985; Aizawa et al. 1992; Otsuka et al. 1991; Hall and Hedgecock 1991). Later studies have shown that KIFs serve as molecular motors for cell division as well (Scholey et al. 1985; Endow et al. 1990). Genomic data have suggested that mammals have more than 45 KIFs (Miki et al. 2001). Nearly half of the KIFs are expressed in neuronal cells and transport various membrane organelles in neurons, while some KIFs are ubiquitously expressed and involved in general intracellular transport (Hirokawa et al. 2009, 2010).

Cellular morphology is supported by the cytoskeleton. Microtubules are one of the major components of the cytoskeleton in eukaryotic cells (Alberts 2008). KIFs use microtubules as rails for their movement. Microtubules are polymers that are composed of α- and β-tubulin dimers. Purified tubulin can polymerize and form microtubules in vitro, but these microtubules are very unstable (Weisenberg 1972). In vivo, microtubules are stabilized by microtubule-associated proteins (MAPs) (Takei et al. 2000; Teng et al. 2001; Harada et al. 1994).

Recent works have shown that KIFs act as regulators for microtubule dynamics. In addition to the roles as sources of the force in intracellular transport and cell division, it has been revealed that Kinesin-8 and Kinesin-13 family members (such as KIF18A, KIF19A, and KIF2A-C) have microtubule-depolymerizing activities (Walczak et al. 2013). In contrast, Kinesin-4 and Kinesin-11 members (such as KIF21A and KIF26A) have microtubule-stabilizing activities (Zhou et al. 2009; van der Vaart et al. 2013). Recent studies have shown that these atypical KIFs play fundamental roles in cellular morphogenesis. In knockout mice in which these KIFs have been deleted, cellular morphology is strongly affected (Cheng et al. 2014; Zhou et al. 2009; Niwa et al. 2012; Homma et al. 2003; He et al. 2014). KIF2A and KIF21A have been identified as causes of human neuronal diseases (Poirier et al. 2013; Yamada et al. 2003).

In this review, I will mainly focus on the roles of KIFs in the regulation of microtubule dynamics and morphogenesis, because there are already many reviews that discuss the roles of KIFs in intracellular transport and cell division (Hirokawa et al. 2009, 2010; Cross and McAinsh 2014; Walczak et al. 2013). In particular, I will describe the roles of the KIFs that regulate microtubule dynamics and cellular morphogenesis. In addition, some interesting but unanswered questions and hypothetical answers to those questions will be discussed.

Microtubule-depolymerizing KIFs in neurons

KIF2A, a member of Kinesin-13, was cloned from brain cDNA by degenerate PCR (Aizawa et al. 1992). While other KIFs have their motor domains at the N-terminus or C-terminus, the motor domain of KIF2A is in the central part of the protein. It has been shown that KIF2A is strongly expressed in neurons and accumulates in growth cones (Noda et al. 1995, 2012). A closely related protein, MCAK (KIF2C, previously called XKCM1 and Kin I), was identified as a kinesin associated with the centromere in mitotic cells (Wordeman and Mitchison 1995). Depletion of MCAK by a specific antibody induces abnormally long microtubules during in vitro experiments using lysate from Xenopus oocytes (Walczak et al. 1996). More strikingly, purified recombinant KIF2A and MCAK have microtubule-depolymerizing activities at both the plus and minus ends of microtubules (Desai et al. 1999). Similar to the motor activities of other KIFs, the depolymerizing activity is ATP dependent. Thus, it is thought that an ATP-dependent conformational change of MCAK affects the microtubule structure and induces depolymerization (Ogawa et al. 2004). The function of KIF2A was revealed by analyzing KIF2A-knockout mice (Homma et al. 2003). While MCAK is important for proper cell division, KIF2A is not considered essential for cell division because knockout mice develop into neonates. However, KIF2A-knockout mice die after birth because of neuronal defects. Abnormal lateral axonal branches are observed in KIF2A-knockout neurons. Abnormal microtubule elongation is observed in the growth cone. Thus, KIF2A regulates the microtubule dynamics in the growth cone and the axonal length (Fig. 1). Consistent with this, a genetic study performed in humans has identified KIF2A as a cause of malformations in cortical development (Poirier et al. 2013). How, then, is KIF2A regulated? It has been suggested that KIF2A is activated by direct binding with phosphatidylinositol-4-phosphate 5-kinase alpha (PIPKα) (Noda et al. 2012). PIPKα is co-purified with KIF2A using a specific monoclonal antibody. When recombinant PIPKα is added, the microtubule-depolymerizing activity of KIF2A is enhanced in vitro. The formation of axonal branches is enhanced in PIPKα knockdown cells because microtubules are more stable, and these stable microtubules often extend to the plasma membrane in PIPKα-depleted neurons. In wild-type cells, microtubules are depolymerized in the growth cone and shrink before collision with the plasma membrane. These cellular morphology and microtubule phenotypes observed in PIPKα-depleted neurons are similar to KIF2A-knockout and -knockdown neurons, suggesting that PIPKα and KIF2A work together in vivo.

Fig. 1
figure 1

Microtubule-regulating KIFs in the neuron. (a) Microtubules are depolymerized by KIF2A in the growth cone, (b) microtubules are bundled and stabilized by KIF26A, and (c) microtubule dynamics are suppressed by KIF21A

Full size image

The function of Kinesin-13 in neuronal morphogenesis is conserved in other species. While mammals have three Kinesin-13 family proteins, Caenorhabditis elegans (C. elegans) has only one Kinesin-13 family member, KLP-7 (Srayko et al. 2005). In the C. elegans klp-7 mutant, abnormal axonal elongation is observed in addition to abnormal cell division (Ghosh-Roy et al. 2012; Schlaitz et al. 2007; Srayko et al. 2005). Thus, it is thought that one gene plays multiple roles in C. elegans, while cell-specific isoforms are used in mammalian cells.

It is interesting to consider how the activity and the growth cone localization of KIF2A are regulated. Subcellular fractionation and immunoelectron microscopy suggest that KIF2A is associated with vesicular structures in neurons (Noda et al. 1995). Thus, one possibility is that KIF2A is transported to the growth cone by another motor protein. In cell division, several factors such as Aurora kinase, protein phosphatase 2A, and +TIP proteins regulate the activity and localization of MCAK and KLP-7 (Schlaitz et al. 2007; Andrews et al. 2004; Jiang et al. 2009; Lee et al. 2008). In addition to PIPKα, these factors would be good candidate regulators, considering the similarity between KIF2A, MCAK, and KLP-7.

Microtubule-stabilizing kinesins in neurons

It is well established that neuronal microtubules are stabilized by MAPs such as tau, MAP1A, MAP1B, and MAP2 (Cleveland et al. 1977; Kanai et al. 1989; Izant and McIntosh 1980; Kim et al. 1979). Microtubules are unstable in MAP-knockout neurons, which leads to abnormal neuronal morphology (Harada et al. 1994, 2002; Takei et al. 2000; Teng et al. 2001). Recent works have revealed that not only MAPs but also KIFs stabilize microtubules in neurons (Fig. 1). KIF26A is a member of the Kinesin-11 family. Unlike other KIFs, KIF26A has an atypical motor domain that can bind to microtubules but does not hydrolyze ATP (Zhou et al. 2009). Consistent with this, the motor domain of KIF26A does not show microtubule gliding activity in vitro. Similar to MAPs, KIF26A bundles and stabilizes microtubules when overexpressed in COS-7 cells that have unstable microtubules. KIF26A-knockout mice suffer from a megacolon that is caused by defects in the enteric nervous system. Consistent with this, KIF26A is strongly expressed in enteric neurons. Microtubules are less stable in KIF26A-knockout enteric neurons than in wild-type neurons. KIF26A-deleted enteric neurons have shorter axons, and the communication between enteric neurons is perturbed. Thus, it is considered that KIF26A works as a microtubule-stabilizing factor in enteric neurons (Fig. 1). In addition to axonal defects, the number of enteric neurons is changed in KIF26A-knockout mice (Zhou et al. 2009). Enteric ganglia are much larger and the number of neurons in each ganglion is much higher in KIF26A-knockout mice. It is well established that the number of enteric neurons is controlled by the RET tyrosine receptor kinase (Heanue and Pachnis 2007). RET is activated upon dimerization induced by the binding of glial cell line derived neurotrophic factor (GDNF) to the glycosylphosphatidylinositol-anchored co-receptor GDNF family receptor α1 (GFRα1) (Heanue and Pachnis 2007). Activated RET recruits Grb2, which stimulates both Akt and MAPK cascades. It has been shown that KIF26A physically binds to Grb2 and negatively inhibits the signaling. This negative regulation is disrupted in KIF26A-knockout mice, which causes the hyperactivation of the GDNF-induced RET signaling. As Akt and MAPK cascades stimulate cell proliferation, the number of enteric neurons is increased in KIF26A-knockout mice.

Recent studies have shown that KIF21A suppresses cortical microtubule dynamics in neurons (Cheng et al. 2014; van der Vaart et al. 2013). KIF21A was originally identified as a motor protein that is expressed in neurons (Marszalek et al. 1999). It has been found that KIF21A mutations cause a neuronal disease, congenital fibrosis of the extraocular muscles type 1 (CFEOM1) (Yamada et al. 2003). KIF21A belongs to the Kinesin-4 family. Previous studies have shown that another Kinesin-4 member, Xklp1, suppresses microtubule dynamics in vitro and in vivo (Bieling et al. 2010; Bringmann et al. 2004). Similar to Xlkp1, purified KIF21A suppresses both microtubule polymerization and depolymerization in vitro, suggesting that KIF21A suppresses microtubule dynamics and stabilizes microtubules. Moreover, KIF21A is genetically associated with MAP1B, supporting the notion that KIF21A regulates microtubule dynamics. CFEOM1-associated mutations disrupt the autoinhibition of KIF21A, which results in the hyperactivation of KIF21A (Cheng et al. 2014; van der Vaart et al. 2013). Thus, it is considered that these mutations change the microtubule dynamics, and the axon guidance of occulomotor neurons is affected in CFEOM1.

One important question is: how are the activities of these KIFs regulated? When neurites are extending, microtubules need to be in a dynamic state. Once neurites reach their destinations, microtubules should be stabilized. Identifying factors that unlock the autoinhibition of KIF21A would be helpful because the disrupted autoinhibition causes axon guidance defects and CFEOM1. In the case of KIF26A, the regulation mechanism is totally unknown. It is important to test whether or not KIF26A is regulated by autoinhibition, as other KIFs are. KIF26A is expressed in brain neurons but its roles are totally unknown. It would be interesting to test whether or not microtubule dynamics as well as growth factor signaling are changed in the KIF26A-knockout brain.

KIFs and the mechanism of ciliary length control

Cilia have highly developed microtubule networks modified by multiple MAPs and dynein motors (Kamiya 2002). Intraflagellar transport (IFT) is required for the formation of cilia (Rosenbaum and Witman 2002). The KIF3A-KIF3B-KAP3 hetero tetramer, called Kinesin-2, is a molecular motor for IFT (Cole et al. 1998; Kozminski et al. 1995; Cole et al. 1993; Yamazaki et al. 1996). It has been proposed that the regulation mechanisms of IFT determine the length of cilia (Ishikawa and Marshall 2011). Recent works have suggested that KIFs that regulate microtubule dynamics are important determinants of the length of cilia. For example, a member of Kinesin-8, KIF19A, regulates the ciliary length (Niwa et al. 2012). Kinesin-8 members have both motor and microtubule-depolymerizing activities and determine the spindle length in cell division (Walczak et al. 2013). Interestingly, Kinesin-8 family members depolymerize microtubules mainly from the plus end, which is different from Kinesin-13 family members, as they depolymerize microtubules from both ends. KIF19A also has both motor and microtubule-depolymerizing activities. KIF19A-induced microtubule depolymerization is plus-end dominant. KIF19A is localized to the tips of motile cilia that have 9 + 2 axonemes (Fig. 2). In KIF19A-knockout mice, motile cilia are longer than in the wild type. The motility of these longer cilia is altered and the metachronal wave of cilia is disrupted. As a result, these cilia cannot generate proper fluid flow. Consistent with this, KIF19A-knockout mice suffer from infertility and hydrocephalus, which are common symptoms of ciliary defect (Ishikawa and Marshall 2011). Interestingly, primary cilia that have 9 + 0 axonemes are not affected in KIF19A-null mice. These results have suggested that other mechanisms that regulate the microtubule dynamics would exist in primary cilia. Actually, a recent study has shown that KIF7, a member of Kinesin-4, regulates the length of primary cilia (He et al. 2014). In KIF7-mutant mice, primary cilia are longer than in the wild type. Rather than depolymerization, KIF7 accumulates at the tips of cilia and suppresses the microtubule polymerization (Fig. 2). The answers to several interesting questions remain elusive. Why are different mechanisms used in primary cilia and motile cilia? This may be related to their different cytoskeletal structures. Are these mechanisms conserved in other organisms? In Trypanosoma, Kinesin-13 family members regulate ciliary length (Chan and Ersfeld 2010). In Chlamydomonas, Kinesin-13 is required for the absorption of cilia, but ciliary length is not affected in Kinesin-13-depleted cells (Piao et al. 2009). Kinesin-8 family members have not been identified in these unicellular organisms. In mice, cilia phenotypes have not been reported in Kinesin-13-knockout mice (Homma et al. 2003). Thus, analyzing cilia in Kinesin-13-knockout mice or -knockdown cells will give us important information. In other model organisms, such as C. elegans and Drosophila melanogaster (D. melanogaster), KIF7 and KIF19A homologues have not been identified yet. However, there are some candidate molecules in their genomes.

Fig. 2
figure 2

Microtubule-regulating KIFs in cilia. KIF19A depolymerizes microtubules in motile cilia while KIF7 suppresses the polymerization of microtubules in primary cilia. Disruption of these mechanisms changes the microtubule dynamics and the cilia lengthen

Full size image

Conclusion

The involvement of KIFs in the regulation of microtubule dynamics and cellular morphology has been clarified, as described in this review. Unraveling the molecular mechanisms that allow these KIFs to regulate the microtubule dynamics would be an interesting research aim. This could be achieved through the structural analysis of the motor domains of these KIFs, as performed in KIF2C (Ogawa et al. 2004). Analysis of the cooperation and competition between microtubule-depolymerizing KIFs and microtubule-stabilizing KIFs would also provide important results. It would also be useful to explore the relationship between these microtubule-regulating KIFs and MAPs because it is now well established that MAPs are fundamental to microtubule dynamics.

In vitro reconstitution using purified motors and microtubules will be required, as will the generation and analysis of double mutants. Moreover, the next important step will be the identification of factors that regulate the activity of these KIFs. Autoregulation and phosphorylation are important factors for other molecular motors (Verhey and Hammond 2009). Small GTPases are also good candidate regulators. The roles of many small GTPases are yet to be clarified. Genetics is the most powerful tool for studying regulating mechanisms. Recent advances in whole-genome sequencing technology will lead to the identification of more human mutations that cause similar symptoms to KIF2A and KIF21A mutations. Genetics using model organisms such as C. elegans and D. melanogaster would be also useful because many signaling pathways, molecular motors, and their regulators have been identified in these organisms via mutant screening (Hall and Hedgecock 1991; Maeder et al. 2014). It will therefore be interesting and essential to test whether or not candidate KIF7, KIF19A, and KIF21A homologues predicted in C. elegans and D. melanogaster genomes are involved in the regulation of neuronal and ciliary microtubule dynamics in these organisms.