The kinesin motors are important in the regulation of cellular functions such as protein trafficking, spindle organization and centrosome separation. In this study, we have identified POPX2, a serine-threonine phosphatase, as an interacting partner of the KAP3 subunit of the kinesin-2 motor. The kinesin-2 motor is a heterotrimeric complex composed of KIF3A, KIF3B motor subunits and KAP3, the non-motor subunit, which binds the cargo. Here we report that the phosphatase POPX2 is a negative regulator of the trafficking of N-cadherin and other cargoes; consequently, it markedly influences cell–cell adhesion. POPX2 affects trafficking by determining the phosphorylation status of KIF3A at serine 690. This is consistent with the observation that the KIF3A-S690A mutant is defective in cargo trafficking. Our studies also implicate CaMKII as the kinase that phosphorylates KIF3A at serine 690. These results strongly suggest that POPX2 and CaMKII are a phosphatase-kinase pair that regulates kinesin-mediated transport and cell–cell adhesion.
The kinesin motors have diverse functions in eukaryotic cells. They are responsible for the transport and trafficking of vesicles, organelles and proteins along the microtubule network as well as for the separation of the centrosomes and the organization of the mitotic spindles during cell division. Thus far about 14 kinesin families and some orphan kinesins have been identified. Most of the kinesins are plus-end motors but some, such as kinesin-14 are reported to be minus-end motors (Endow et al., 2010). The activities of the kinesin motors in the cell are regulated to ensure that cargoes are directed to the right place at the right time. The KIF3 motor belongs to the kinesin-2 family. It is a heterotrimeric motor, which comprises the KIF3A and KIF3B motor subunits and a non-motor KAP3 (KIF-associated protein 3) subunit (Hirokawa, 2000). KAP3 contains armadillo repeats, which are thought to mediate protein–protein interaction and regulate binding with the cargo proteins. To date, N-cadherin, adenomatous polyposis coli (APC), β-catenin, fodrin and Rab11-containing vesicles have been identified as cargoes transported by the KIF3 motor (Takeda et al., 2000; Jimbo et al., 2002; Teng et al., 2005; Schonteich et al., 2008; Hirokawa et al., 2009).
The POPX (partner of PIX) phosphatases belong to the PP2C serine-threonine phosphatase family. So far, two members, POPX1 and POPX2, have been identified (Koh et al., 2002). The larger POPX1 is found predominantly in the brain and testis, whereas POPX2 is ubiquitous. The main substrates of POPX2 identified thus far are the p21-activated kinase (PAK) and calcium calmodulin kinases (CaMKs) (Ishida et al., 1998; Koh et al., 2002). POPX2 is found in a complex with PIX (PAK-interacting exchange factor), which is a guanine nucleotide exchange factor for RAC1 and CDC42 (Koh et al., 2002). The SH3 domain of PIX, however, binds strongly to PAK (Manser et al., 1998), suggesting that POPX2 is in close proximity of an activator (PIX) and effector (PAK) of the Rho GTPases. POPX2 has also been shown to interact with the formin protein mDia1 and can regulate mDia1- and RhoA-regulated transcription mediated by serum response factor (Xie et al., 2008). The levels of POPX2 are higher in invasive cancer (Susila et al., 2010). Overexpression of POPX2 also results in higher cell motility, which suggests that POPX2 regulates proteins other than PAK and CaMK. One other pathway regulated by POPX2 is the glycogen synthase kinase 3 (GSK3) pathway (Singh et al., 2011; Zhang et al., 2013). However, GSK3 is not a direct substrate of POPX2.
In an attempt to understand more about the physiological roles of POPX2 and the biochemical pathways that POPX2 regulates, a yeast two-hybrid screen was conducted to identify POPX2 interaction proteins. KAP3, a subunit of the KIF3 motor, was identified as an interacting partner of POPX2. Investigating the physiological significance of this interaction, we found that fibroblasts overexpressing POPX2 are impaired in KIF3 motor-mediated transport of N-cadherin to the cell periphery and are, consequently, defective in cell–cell adhesion. The transport of other KIF3 motor cargoes, such as the Rab11-containing vesicles, is also similarly affected by POPX2 in a phosphatase-activity-dependent manner. Our findings suggest that POPX2 determines the phosphorylation status of the KIF3 kinesin motor and, in turn, regulates cell–cell adhesion.
POPX2 interacts with the KIF3 motor complex
KAP3, the non-motor subunit of the KIF3 motor complex, was found to be an interacting partner of POPX2 in a yeast two-hybrid screen. To confirm this interaction by immunoprecipitation, we used GFP-tagged human POPX2 cDNA to generate a stable NIH3T3 fibroblast cell line because no antibodies are available to detect mouse POPX2. We found that endogenous KAP3 coimmunoprecipitated with GFP–POPX2 (Fig. 1A). KAP3 also interacted with POPX1. Using deletion analysis, we mapped the KAP3-interacting region near the N-terminus of both POPX1 (residues 236–269) and POPX2 (residues 150–193) (Fig. 1B,C). This region of POPX also overlaps with the region responsible for the interaction of POPX with PIX (Koh et al., 2002). Because PIX, POPX2 and KAP3 can form a trimeric complex (supplementary material Fig. S1A), this suggests that PIX and KAP3 interact with POPX2 at different sites. We also tested whether POPX2 interacts with KIF3A and KIF3B because a dual-specificity phosphatase Dusp26 has been reported to interact with both KIF3A and KAP3 (Tanuma et al., 2009). We found that both KAP3 and KIF3A coimmunoprecipitated well with POPX2 (supplementary material Fig. S1B). However, little KIF3B was found in the POPX2 pull-down. The armadillo repeats on KAP3 are responsible for interacting with POPX2. There are three clusters of armadillo repeats on the KAP3 protein. Both clusters 1 and 3 showed binding to POXP2 (Fig. 1D).
Localization of N-cadherin to the cell periphery is affected by POPX2
To determine possible roles of the interaction between POPX2 and KAP3, we investigated whether transport of the KIF3 cargoes is affected by POPX2. Because N-cadherin has been reported to be a cargo of the KIF3 motor and is normally localized to the cell periphery (Teng et al., 2005), we decided to study its localization in control NIH3T3 cells and in the stable POPX2-overexpressing cell line (X2). N-cadherin failed to localize to the cell periphery in POPX2-overexpressing cells (Fig. 2A). Similarly, another reported cargo of the KIF3 motor, β-catenin (Teng et al., 2005), also mislocalized in POPX2-overexpressing cells. β-catenin no longer localized to the cell–cell junctions in POPX2-overexpressing cells (Fig. 2B). The levels of N-cadherin and β-catenin were not affected by the overexpression of POPX2 (Fig. 2C). However, the amount of N-cadherin found at the cell surface was much reduced in cells overexpressing POPX2 (Fig. 2D). It is unlikely that the presence of POPX2 affected the interaction between KAP3 and N-cadherin because similar amounts of KAP3 and N-cadherin were co-immunoprecipitated in both the control cells and POPX2-overexpressing cells (Fig. 2E). The presence of POPX2 also did not affect the interaction between KIF3A and KAP3 (Fig. 2F). These observations imply that POPX2 affects the transport of the KIF3 cargoes to their destinations in the cells.
We next investigated whether the phosphatase activity of POPX2 was required to perturb the transport of KIF3 cargo. NIH3T3 cells were transfected with GFP–POPX2 and a phosphatase-dead mutant, GFP–POPX2m (Koh et al., 2002). We found that localization of N-cadherin and β-catenin was affected by POPX2 but not POPX2m, suggesting that the phosphatase activity of POPX2 is responsible for the impairment of KIF3-motor-mediated transport (Fig. 3A,B).
POPX2-overexpressing cells show defective cell–cell adhesion
N-cadherin is one of the major components of the adherens junction in NIH3T3 cells. Because POPX2-overexpressing cells showed impaired N-cadherin localization, we determined whether cell–cell adhesion was also affected. Briefly, cells were suspended with and without calcium or EGTA (ethylene glycol tetraacetic acid) for a fixed time interval. The number of aggregates formed was then counted. As cadherins require calcium to mediate cell–cell interaction, incubation of the cells without calcium or with EGTA to chelate the calcium ions would result in loss of cell aggregation. Similarly if N-cadherin were not properly localized to the cell periphery, no cell aggregation would occur. As expected, POPX2-expressing cells did not form aggregates in the presence of 2 mM CaCl2. However, aggregation of control cells or POPX2m-expressing cells was not affected (Fig. 3C). In the presence of POPX2, the efficiency of cell aggregation mediated by calcium was reduced to about 30% compared with that in control cells (Fig. 3D). Thus, the aggregation assay strongly correlates POPX2-mediated inhibition of N-cadherin localization to the cell periphery with the loss of cell–cell adhesion.
Transport by the KIF3 motor is affected by POPX2
We next tracked the transport of a KIF3 cargo, N-cadherin using live-cell imaging. NIH3T3 cells were doubly transfected with N-cadherin–GFP and mCherry–POPX2 or mCherry–POPX2m cDNA constructs. The trafficking of N-cadherin–GFP was monitored. The movement of N-cadherin–GFP was retarded in POPX2-expressing cells but not in POPX2m-expressing cells or the control cells transfected with mCherry vector construct (Fig. 4A and supplementary material Movie 1). The mean velocity exhibited by these N-cadherin–GFP vesicles suggests that POPX2 severely affects the trafficking of N-cadherin (Fig. 4B).
To determine whether the defects we observed were due to the impairment of KIF3-motor-mediated transport, we included two other control experiments. HeLa cells were used to show that the phenomenon we observed is not cell-type specific but is due to the influence of POPX2 on the KIF3 motor. Because Rab11-containing vesicles are also transported by the KIF3 motor (Schonteich et al., 2008), we monitored the trafficking of dsRed–Rab11 in GFP–POPX2- and GFP–POPX2m-expressing cells. Similar to what was observed for N-cadherin, the trafficking of Rab11-containing vesicles was impaired in the presence of POPX2 but not POPX2m (Fig. 4C,D; supplementary materia Movie 2). In addition, POPX2 also caused the clustering of Rab11-containing vesicles (Fig. 4E,F). However, the trafficking of vesicles containing clathrin light chain A, which was not a cargo of the KIF3 motor, was not affected by POPX2 (Fig. 4G,H; supplementary material Movie 3). We have also included an additional control experiment to show that overexpression of POPX2 does not affect the trafficking of mitochondria (Fig. 4I), suggesting that POPX2 specifically affects cargoes transported by the KIF3 motor. Immunostaining using α-tubulin antibodies also suggested that POPX2 does not cause observable defects to the microtubule network (supplementary material Fig. S2A). We also confirmed that N-cadherin–GFP is localized to early endosomes and secretory vesicles by co-staining with EEA1 and chromogranin A, respectively (supplementary material Fig. S2B). This is consistent with earlier reports of localization of N-cadherin to the plasma membrane, Golgi and endocytic and secretory vesicles (Mary et al., 2002). Kinesin-2 has been detected in cell fractions containing both early and late endocytic vesicles, and addition of antibodies targeting kinesin-2 resulted in decreased directional movement of endosomes (Loubéry et al., 2008).
Serine 690 of KIF3A is important for regulation of KIF3 motor transport
Having found that POPX2 phosphatase profoundly affects the trafficking of KIF3 cargoes, we proceeded to determine the underlying mechanism. We hypothesized that POPX2 regulates the KIF3 motor through the control of KIF3 phosphorylation status. Our observations thus far suggest that dephosphorylated KIF3 motor complex is impaired in cargo transport.
Various phosphoproteomic-based analysis have identified the serine-690 residue on the C-terminal tail of KIF3A as a major phosphorylation site (Ballif et al., 2004; Dephoure et al., 2008; Pan et al., 2008; Tweedie-Cullen et al., 2009; Huttlin et al., 2010; Olsen et al., 2010; Wiśniewski et al., 2010). This serine residue falls within the putative consensus sequence R-x-x-S for CaMKII and this sequence is conserved among human, mouse and zebrafish KIF3A (Fig. 5A). Hence, we generated KIF3A-S690 mutants and monitored their effects on KIF3 cargo transport. Serine 690 was mutated to alanine (S690A) to preclude phosphorylation, or to aspartic acid (S690D) to mimic constitutive phosphorylation.
We first determined the consequence of mutating KIF3A-S690 on the distribution of N-cadherin. Stable cell lines overexpressing GFP–KIF3A-WT, GFP–KIF3A-S690A or GFP–KIF3A-S690D were generated by retroviral infection of the NIH3T3 cells. Immunofluorescence staining of the endogenous N-cadherin showed that overexpression of the KIF3A-S690A mutant led to a decreased localization of N-cadherin to the cell periphery and cell–cell junctions (Fig. 5B). We also observed that overexpression of KIF3A-S690A, but not KIF3A-WT and KIF3A-S690D, led to less N-cadherin at the cell surface (Fig. 5C). We then examined the effect of mutating KIF3A-S690 on the motility of N-cadherin-GFP. Long centrifugal movement of N-cadherin-GFP was observed in NIH3T3 cells overexpressing mCherry-KIF3A or mCherry-S690D mutant, indicating that overexpression of these proteins did not cause significant changes in the motility of N-cadherin vesicles (Fig. 5D and supplementary material Movie 4). However, movement of N-cadherin-GFP was impaired in cells overexpressing the mCherry-S690A mutant (Fig. 5D and supplementary material Movie 4). We noticed that this impairment occurred only in cells with very high expression of mCherry-S690A. Cells with low or moderate expression of mCherry-S690A did not show significant changes in the motility of the N-cadherin vesicles. This might be due to the basal activity of endogenous wild-type KIF3A. However, high expression of mCherry-KIF3A and mCherry-S690D did not have any apparent effect on the N-cadherin transport, indicating that inhibition of N-cadherin trafficking in mCherry-S690A-overexpressing cells was not due to off-target effects caused by high levels of protein expression. Calculation of mean velocity showed that the KIF3A-S690A mutant affected trafficking of N-cadherin (Fig. 5E).
Next, we determined whether overexpression of the KIF3A-S690A mutant perturbs the motility of Rab11. Again, using live-cell imaging, we observed that the dsRed–Rab11 vesicles were less motile and accumulated more around the perinuclear region in cells that overexpressed GFP–KIF3A-S690A (Fig. 5F and supplementary material Movie 5). The KIF3A-S690A mutant also affected the mean velocity of Rab11-containing vesicles (Fig. 5G). The KIF3A-S690D mutant did not affect the trafficking of Rab11-containing vesicles. Taken together, our results suggest that the phosphorylation status of the KIF3A serine-690 residue is crucial in regulating cargo transport.
To remove the bias in manual tracking of N-cadherin or Rab11 vesicles, we used particle segmentation and tracking algorithms to track every fluorescent particle in the cells in all our analyses. Representative tracks in control and X2 cells are shown in supplementary material Fig. S3A,B. In addition, we also found that the average velocity of N-cadherin trafficking ranges from ∼0.2 to 0.36 µm/second depending on the cell type. Additional experiments were also conducted at different frame rates for the live imaging. We did not observe any differences arising from different frame rates used (1 frame per 3 seconds versus 2 frames per second, supplementary material Fig. S3C).
Phosphorylation of KIF3A-S690 does not affect cargo binding affinity
Previous studies have reported that phosphorylation of serine residues in the C-terminal tail of the kinesin motor modulates the affinity of cargo binding (Ichimura et al., 2002; Guillaud et al., 2008; Vagnoni et al., 2011). Thus, we sought to elucidate whether mutation of serine 690 of KIF3A affects binding of molecular cargo. Because N-cadherin associates with the KIF3 motor primarily through the KAP3 subunit, we determined whether the binding between KAP3 and the various KIF3A mutants is affected. GST–KAP3 was co-expressed with various FLAG-tagged KIF3A constructs and the cell lysates were subjected to FLAG immunoprecipitation. Wild-type KIF3A, KIF3A-S690A and KIF3A-S690D interacted equally well with GST–KAP3 in these assays and no interaction was observed between FLAG-KIF3A and the control GST (Fig. 6A). The lack of effect of mutating KIF3A serine 690 on KAP3 binding implies that the inhibition on cargo transport is not due to modulation of binding affinity between KIF3A and KAP3.
We also checked whether the interaction between the motor domain and the tail domain of KIF3A is affected by S690 phosphorylation. GST-tagged tail domain (aa 601–702) constructs of KIF3A-WT, KIF3A-S690A and KIF3A-S690D were tested for their interaction with the motor domain (aa 1–354). We found that all three tail domain constructs bound similarly well with the motor domain, suggesting that S690 mutation to A or D does not affect head to tail interaction of KIF3A (Fig. 6B). However, it is possible that S690 phosphorylation status affects KIF3 motor interaction with microtubules. Therefore, we immunoprecipitated KIF3A-WT, KIF3A-S690A and KIF3A-S690D from transfected HeLa cells, and determined their ability to bind to microtubules. We found that KIF3A-S690A showed weaker interaction with microtubules (Fig. 6C).
CaMKII and POPX2 regulate the phosphorylation status of KIF3A-S690
Because we have established that the phosphorylation status of S690 of KIF3A is important in the transport activity of the KIF3 motor, we tested the possibility that POPX2 directly dephosphorylates KIF3A. It is first necessary to ascertain that KIF3A-S690 is phosphorylated by CaMKII. The KIF3A tail domain could indeed be phosphorylated by CaMKII in an in vitro kinase assay (Fig. 6D, lane 3). The S690 residue is likely to be the major phosphorylation site within the tail domain because no observable phosphorylated band was detected for the KIF3A-S690A tail domain in the presence of active CaMKII (Fig. 6D, lane 4). CaMKII could be the specific kinase that phosphorylates the KIF3A tail domain because the active PAK did not phosphorylate the KIF3A tail (Fig. 6D, lane 1). In addition, trafficking of N-cadherin–GFP was impaired in the presence of the CaMKII inhibitors (KN-93, KN-62 and AIP), suggesting that CaMKII phosphorylation regulates the KIF3 motor complex (Fig. 7A–D, supplementary material Movie 6).
Phosphorylated KIF3A-S690 could be effectively dephosphorylated by POPX2 (Fig. 7E, lane 3). The control calf-intestinal phosphatase showed a much lesser extent of dephosphorylation when used at a higher concentration (Fig. 7E, lane 4). In the phosphatase assays using GST-tagged phosphatases pulled down from transfected cell lysates, POPX2 showed higher efficiency compared with PP2Cα and PP1 towards phosphorylated KIF3A-S690 (Fig. 7F; see supplementary material Fig. S2C for purity of GST-tagged phosphatases).
Silencing of endogenous POPX2 does not adversely affect the KIF3 motor
Silencing endogenous POPX2 using siRNA did not lead to an increase in N-cadherin trafficking (Fig. 8). N-cadherin distribution to the cell–cell junctions (Fig. 8A) and the trafficking or motility of N-cadherin appeared to be similar in the control and POPX2-knockdown cells (Fig. 8B,C). Because the antibody against POPX2 cannot detect mouse POPX2, we used RT-PCR to check the siRNA knockdown efficiency (Fig. 8D). As an additional control, we have also confirmed that silencing of POPX2 in HeLa cells does not affect the localization of N-cadherin to the cell periphery and motility of N-cadherin–GFP (supplementary material Fig. S4). These observations are consistent with the overexpression of KIF3A-S690D (Fig. 5E,G). However, inhibition of CaMKII phosphorylation of KIF3A resulted in impairment of cargo transport. Taken together, our data suggest that under-phosphorylation of KIF3A-S690 will lead to impairment of cargo trafficking. Because we were tracking the motility of N-cadherin and other KIF3 cargoes, it is implicit that only the phosphorylated KIF3 motor will move along the microtubules. Interestingly, overexpression of KIF3A-S690D could rescue the N-cadherin and β-catenin localization defects induced by POPX2 (supplementary material Fig. S5). This observation further suggests that the phosphorylation status of S690 is important in the regulation of KIF3 motor transport.
The mechanisms by which kinesin proteins transport their molecular cargoes to specific subcellular locations and how this transport is regulated are not clearly understood. Several modes of regulation have been suggested. In general, one mode of regulation involves modulation of the binding of kinesins to the microtubule tracks. An example of this is kinesin-1, whose phosphorylation in the motor domain (serine 176) reduces its ability to bind microtubules and to transport cargo (Morfini et al., 2009). A second mode of control involves regulation of the association between cargo and the motor. Phosphorylation of residues in the C-terminal tail region of kinesins has been shown to either promote or inhibit cargo binding. For instance, phosphorylation of kinesin light chain 2 on serine 575 stimulates binding of 14-3-3 cargo protein (Ichimura et al., 2002), whereas phosphorylation of KIF17 on serine 1029 disrupts the binding of Mint1 to the tail domain of the motor (Guillaud et al., 2008). Finally, a fundamental control of kinesin function comes from auto-inhibition. Soluble kinesins have been shown to assume a folded, inactive conformation, where the C-terminal region interacts with the motor domain, leading to the inhibition of the ATPase activity and the ability to bind microtubules (Friedman and Vale, 1999; Stock et al., 1999). This interaction is thought to be relieved by high ionic strength or binding of molecular cargoes, which generates an extended, active conformation. Nonetheless, cargo binding might not be the complete answer to motor activation. Kinesin-1 motors that are attached to membrane cargo can still be inactive (Woniak and Allan, 2006). Instead, phosphorylation has been shown to play an important role in relieving kinesin auto-inhibition. It has been reported that phosphorylation of threonine 937 of KIF11 (Eg5) in the inhibitory C-terminal tail increases the efficiency of microtubule binding (Cahu et al., 2008). In a second example, phosphorylation of the inhibitory C-terminal tail of kinesin-7 causes the motor to unfold and acquire higher processivity along the microtubules (Espeut et al., 2008).
In the current study, we have demonstrated that preclusion of phosphorylation on KIF3A serine 690 leads to impairment of cargo transport. Because serine 690 resides at the C-terminal tail domain and the KIF3 complex have been reported to adopt a folded conformation (Wedaman et al., 1996), our data suggest that phosphorylation of KIF3A on serine 690 is important for relieving auto-inhibition. Consequently, when phosphorylation of this site is inhibited in the KIF3A-S690A mutant, we observed a reduction in cargo trafficking (Fig. 5). We hypothesize that the KIF3 motor is auto-inhibited when S690 is not phosphorylated. The auto-inhibited confirmation could be similar to other kinesin motors owing to a head-to-tail interaction. The motor domain of KIF3A can still interact with the S690D C-terminal tail (Fig. 6B). However, we do not think that this observation contradicts our hypothesis that phosphorylation relieves auto-inhibition of the kinesin. It is likely that the S690D mutation does not represent a true phospho-mimic as replacement of serine with aspartic acid only resulted in an increased positive charge. This model could be tested by the determination of the structure of the head domain in complex with the C-terminal tail domain, which contains WT, S690A or S690D. It appears that the phosphorylation status of KIF3A-S690 might affect the binding of KIF3 motor to the microtubules (Fig. 6C). This is consistent with observations of another kinesin-2 motor KIF17, which adopts a folded conformation whereby the C-terminal tail domain interacts with the motor domain and prevents the motor from binding to the microtubules (Hammond et al., 2010). Future experiments will involve further characterization of the binding of KIF3 motor to the microtubules and will determine whether the ATPase activity of the motor is also regulated by phosphorylation.
One remarkable feature of the KIF3 kinesin that distinguishes it from other kinesins is the presence of two different motor subunits, KIF3A and KIF3B. The physiological significance of having two different motor subunits is still unclear. Phosphoproteomic analysis has indicated that mouse KIF3B is phosphorylated on serine 181 and serine 681 (Huttlin et al., 2010; Wiśniewski et al., 2010). Serine 181 is located in the motor domain whereas serine 681 resides at the tail domain. Whether POPX2 regulates phosphorylation of these two sites is not known. In addition, the consensus sequence for CaMKII is not found within these sites. However, it is possible that phosphorylation of both KIF3A and KIF3B is required for full activation of KIF3 motor activity. Likewise, inhibition of merely one phosphorylation site might not totally abrogate the motor function. This could partly explain why the KIF3A-S690A mutant is able to reduce but not fully inhibit KIF3-mediated protein transport.
Our results clearly suggest the involvement of both kinase and phosphatase in the regulation of kinesin function. We have shown that KIF3A-S690 can be phosphorylated directly by CaMKII and that phosphorylation at this site might be one of the mechanisms for relieving the auto-inhibition of the KIF3 motor. Moreover, the trafficking of N-cadherin cargo is impaired in the presence of CaMKII inhibitors, suggesting the importance of CaMKII phosphorylation in KIF3-motor-mediated transport (Fig. 7A–D). Dephosphorylation of KIF3A by POPX2, either directly or indirectly by the inhibition of CaMKII, however, confines the motor in an auto-inhibited state. That CaMKII is involved in the regulation of KIF3 motors also raises the possibility that calcium signaling fine-tunes cargo trafficking by the kinesin motors. It is well documented that extracellular calcium regulates cell–cell adhesion through cadherin proteins. Intracellular calcium can also affect N-cadherin and related signaling. The influx of calcium into cells upon injury or activation of NMDA receptors is known to lead to the activation of calpain, a protease that is sensitive to calcium. Activated calpain then cleaves N-cadherin (Jang et al., 2009). The cleavage of N-cadherin further affects cell–cell adhesion, AKT signaling, as well as N-cadherin and β-catenin interaction. However, it has also been reported that upon brain injury, astrocytes increase intracellular calcium levels, which subsequently leads to upregulation of N-cadherin translation. Upregulation of N-cadherin, in turn, plays important roles in reactive astrogliosis and neuroprotection (Kanemaru et al., 2013). In this study, we report a new mechanism that controls the amount of surface N-cadherin by regulation of KIF3 motor trafficking by CaMKII. Because N-cadherin is not the only cargo transported by the KIF3 motor, we expect that localization of other KIF3 cargoes such as β-catenin, adenomatous polyposis coli protein and cell-polarity proteins (Hirokawa et al., 2009) will also be similarly affected by the calcium content in the cell and the activity of CaMKII. Hence, it is not unexpected that CaMKII and POPX2 might also regulate cell polarity, particularly in neuronal cell types.
The KIF3 motor complex is also linked to tumour suppression. Conditional knockout of KAP3 in mouse embryos leads to tumour-like growth in the brain (Teng et al., 2005). This has been attributed to the mislocalization of N-cadherin and β-catenin, which then impacts the Wnt signaling pathway. Higher levels of POPX2 are observed in more invasive breast cancer cell lines (Susila et al., 2010) and POPX2 has also been linked to the regulation of MAP kinases, ERK1/2 and glycogen synthase kinases, GSK3α and GSK3β (Zhang et al., 2013). Modulation of the KIF3 motor complex (this study) might add an additional complexity to the role of POPX2 in the regulation of cancer biology. POPX2 could regulate the behavior of cancer cells by inhibiting proper localization of N-cadherin and β-catenin to the cell periphery, leading to their accumulation in other cellular compartments, which, in turn, affects other signaling pathways. The resultant loss of cell–cell contact might also cause dissemination of cancer cells during metastasis.
MATERIALS AND METHODS
Antibodies and plasmid constructs
Mouse monoclonal antibody against FLAG-M2, rabbit polyclonal antibody against FLAG and rabbit polyclonal antibody against β-catenin were purchased from Sigma-Aldrich. Mouse monoclonal antibodies against KAP3 were obtained from BD Biosciences and SantaCruz Biotechnology, respectively. Mouse monoclonal antibody against KIF3A and mouse monoclonal antibody against N-cadherin were from BD Biosciences. Rabbit polyclonal antibody against GFP was from Life Technologies. Rabbit polyclonal antibody against GST was from Bethyl Laboratories Inc. (Montgomery, TX, USA). Rabbit polyclonal antibody against phospho-serine/threonine was from ECM Biosciences (Versailles, KY). Mouse monoclonal antibody against actin was from Chemicon (Millipore). Rabbit anti-sera against human POPX2 were from the Koh lab. Constructs encoding POPX2 and the phosphatase-dead mutant, POPX2m were described previously (Koh et al., 2002). Full-length FLAG-tagged and GST-tagged KAP3 constructs were generated in the pXJ40 vector. Deletion mutants of KAP3 were constructed by cloning the PCR fragment generated with primers targeting various regions. Constructs encoding ORF of KIF3A were cloned into the pXJ40 vector containing their respective tags. KIF3A-S690A and KIF3A-S690D mutants were generated using the QuikChange site-directed mutagenesis kit (Stratagene-Agilent Technologies, Santa Clara, CA). The motor domain of KIF3A (1–354) was amplified by PCR and cloned into the pXJ40 vector with FLAG tag. N-cadherin–eGFP (Addgene plasmid 18870) construct was from Addgene (Cambridge, MA). The cDNA constructs of CaMKIIα were made in the pXJ40 vector with their respective tags. The original cDNA for KAP3 was a gift from Yoshimi Takai (Kobe University, Kobe, Japan). dsRed–Rab11a and mCherry–clathrin-LCA were kind gifts from Boon Chuan Low (National University of Singapore, Singapore) and Lei Lu (Nanyang Technological University, Singapore), respectively. The cDNA of CaMKII was a gift from Thomas Leung (Institute of Molecular and Cell Biology ASTAR, Singapore).
The accession numbers of the various genes and proteins used in this study are listed below: KIF3A (human), BC045542; KIF3B (mouse), NM_008444.4; KAP3 (human), NM_014970; POPX2 (human), AF520615; POPX1 (human): AF520614.
Cell culture, transfection and retrovirus infection
COS7 and NIH3T3 cells were grown in DMEM high-glucose medium supplemented with 10% FBS. HeLa cells were maintained in minimal essential medium with 10% FBS. The cells were transiently transfected with Lipofectamine 2000 (Life Technologies) according to the manufacturer's protocol. For stable transfections, GFP-tagged POPX2, POPX2m, KIF3A, KIF3A-S690A and KIF3A-S690D cDNAs were subcloned into the retroviral vector pBABE-puro. The packaging line, Phoenix-Eco cells, was transfected with various pBABE-puro plasmids using Lipofectamine 2000. The cells were cultured for more than 24 hours and the supernatants were used to infect the NIH3T3 cells in the presence of 8 µg/ml polybrene. Infected cells were selected and maintained in medium containing puromycin.
Immunofluorescence staining and time-lapse imaging
Cells on coverslips were fixed in 4% paraformaldehyde 24 hours after transfection. The cells were then permeabilized in 0.2% Triton X-100 and blocked with 4% BSA. The cells were incubated with primary antibody in 1% Triton X-100 (1∶100) at 4°C overnight, followed by incubation with secondary antibody in 1% Triton X-100 (1∶100) at room temperature for 1 hour. Coverslips were dried and mounted using Vectorshield with DAPI to stain nuclei.
A Carl Zeiss axiovert microscope and Zeiss objective Plan Neo-fluar 100× /1.3 oil or Plan-Apochromat 63× /1.4 were used. The MetaMorph software program was used to capture images using a Roper Scientific CoolSNAP CCD camera. Time-lapse epi-fluorescence microscopy was performed on a Carl Zeiss axiovert microscope equipped with Zeiss objective Plan Neo-fluar 100× /1.3 oil-immersion objective, sample heater (37°C) and CO2 incubation chamber. Images were captured with a Roper Scientific CoolSNAP CCD camera. Two different frame rates were used: 1 frame/3 seconds (for all movies and figures except otherwise stated) or 2 frames/second (for Fig. 4I, Fig. 7B, supplementary material Fig. S4). An automated particle tracking software was used to analyse the images (see below). The playback rate for all movies is 7 frames/second.
Cells were trypsinized in HBS (with or without 2 mM CaCl2), followed by two washes in the respective HBS supplemented with 1% BSA and passed through a 27G needle three times to obtain single cells. Cells were then diluted to 5×105 cell/ml in the same buffer and were incubated in 1.5 ml tubes at 37°C with gentle rotation. After 30 minutes, the number of cell particles was counted to determine the aggregation index (n0–n30)/n0, where n30 is the total particle number after 30 minutes incubation and n0 is the total particle number at the start of incubation (Ozawa et al., 1990). Images of cell particles on the haemocytometer were captured with phase-contrast microscopy.
In vitro kinase and phosphatase assays
FLAG–CaMKII was expressed in COS7 cells and subjected to FLAG immunoprecipitation. Activation of CaMKII was performed by adding kinase buffer (50 mM Tris-HCl, pH 7.5, 10 mM MgCl2, 5 mM MnCl2, 1 mM dithiothreitol and 0.05% Triton X-100) supplemented with 1.2 µM calmodulin, 2 mM CaCl2 and 200 µM ATP. The activation reaction was incubated at 30°C for 10 minutes. Kinase assays were performed by adding 8 µg of purified GST-fusion protein to the activated FLAG–CaMKII, in 30 µl of kinase buffer supplemented with 200 µM ATP. The kinase reaction (30°C, 30 minutes) was stopped by the addition of SDS sample buffer. For phosphatase assays, GST–KIF3A-Tail phosphorylated by CaMKII, was incubated with bacterially expressed POPX2 or CIP (New England Biolabs, Ipswich, MA) for 30 minutes at 30°C. The reactions were stopped by the addition of SDS sample buffer.
Binding of KIF3A and its mutants to microtubules was assayed using the microtubule-binding protein spin-down assay kit (Cytoskeleton Inc., Denver, CO). KIF3A proteins were first immunoprecipitated and eluted using low-pH glycine buffer, followed by neutralization with Tris buffer. Same amounts of eluted KIF3A proteins were incubated (as indicated by the input blot) with in vitro polymerized microtubules at room temperature for 30 minutes. Microtubule-bound and soluble proteins were separated by centrifugation. The microtubule fraction (pellet) was collected and analyzed by western blotting.
Automated particle tracking
The particle segmentation and tracking algorithms used in this paper are described previously (Lowe, 2004; Jaqaman et al., 2008). These algorithms are implemented through the Trackmate plug-in in Fiji (Schindelin et al., 2012). Briefly, the particles within cells in each image of the time-lapsed sequence are segmented and identified using the difference of Gaussians approach (Lowe, 2004). These particles are then filtered based on quality and mean intensity. Next, the segmented particles are tracked by linking the individual particles from frame to frame using the linear assignment problem (LAP) method (Jaqaman et al., 2008), with modifications to the linking cost calculations with respect to both quality and mean intensity. The velocities of individual particles are then obtained from the particle tracks within the time-lapsed images. The particle tracks were then filtered based on their directional persistence in movement to eliminate particles that were not persistently moving in a particular direction. At any point in time, the mean squared displacement of a tracked particle with two-dimensional coordinates x and y, is denoted as (r2), where r2 = x2+y2. We describe the movement of each particle with (r2) ∝ tγ, where t is the time elapsed since the start of particle tracking and γ is denoted as a particle movement persistence factor. When γ = 1, then (r2) ∝ t, where the displacement is proportional to the square root of the elapsed time and the particle is said to exhibit a random movement behavior. When γ = 2, then (r2) ∝ t2, where the displacement is proportional to the elapsed time and the particle exhibits directed movement persistence. Hence, particle movement persistence can be quantified using γ, and particles with γ ≥1.3 were retained.
↵* These authors contributed equally to this work.
The authors declare no conflict of interest.
H.Q.P., J.L.H. and S.K.L. performed the experiments and analyzed the data. Y.Z. and K.H.C. analyzed the data for particle tracking. H.Q.P., H.Y.L. and C.G.K. conceived and designed the experiments. C.G.K. wrote the manuscript.
This study was supported by the Academic Research Fund Tier 2 [grant number MOE2012-T2-1-046 to C.G.K.]; and the Biomedical Research Council, Singapore [grant number 10/1/22/19/659 to C.G.K.]. We also thank the National Research Foundation, Singapore.
Supplementary material available online at http://jcs.biologists.org/lookup/suppl/doi:10.1242/jcs.126482/-/DC1
- Received January 2, 2013.
- Accepted November 15, 2013.
- © 2014. Published by The Company of Biologists Ltd