Loss of CAMSAP3 promotes EMT via the modification of microtubule–Akt machinery

Metastasis remains a major obstacle for cancer therapy. During metastasis, cancer cells invade nearby tissues, subsequently disseminating into the vascular system before invading secondary organs (Karlsson et al., 2017). The majority of cancer is derived from epithelial organs, and the epithelial-to-mesenchymal transition (EMT) of cancer cells is thought to contribute to their invasion and metastasis. The morphological change of epithelial cells to an elongated fibroblast-like shape, accompanied by looser cell–cell adhesion and higher motility, favors their invasion of nearby tissues (Godde et al., 2010; Lamouille et al., 2014). Over the past decade, an increasing number of studies have indicated that metastatic cancer cells acquire mesenchymal-like characteristics and that this change is relevant to poor clinical outcomes (Aktas et al., 2009; Iwatsuki et al., 2010; Spaderna et al., 2006).

EMT involves reorganization of the cytoskeleton. Although the role of actin and intermediate filaments in EMT is well-characterized (Haynes et al., 2011; Liu et al., 2015; Shankar and Nabi, 2015; Velez-delValle et al., 2016), the possible functions of microtubules in EMT or cancer behavior have only recently been elucidated. Microtubules, major components of the cytoskeleton, have been shown to govern intracellular trafficking, molecular signaling and directional cell movement through their dynamic properties (Etienne-Manneville, 2013; Kaverina and Straube, 2011; Parker et al., 2014). Microtubules are composed of α- and β-tubulin dimers, and display two ends, the minus- and plus-ends. The minus-end generally binds to microtubule-organizing centers (MTOCs), including the centrosome and Golgi complex, whereas the microtubule plus-end is involved in tubulin assembly or disassembly, causing dynamic instability (Dyachuk et al., 2016; Toya and Takeichi, 2016). These characteristics of microtubules are involved in the transitional stages of microtubule growth and shrinkage, affecting the interaction of microtubules with other cytoskeletal and cytoplasmic elements. Microtubule dynamics have been linked with various cellular behaviors and morphogenesis, both in tissue development and pathological diseases. For instance, an enhancement of microtubule stability in which tubulin acetylation becomes more enriched has been shown to drive metastasis in breast cancers (Boggs et al., 2015; Matrone et al., 2010). Cooperation of microtubules with other cytoskeletal elements and signaling molecules also contributes to the activation of downstream effector-related cell behaviors (Giustiniani et al., 2009; Jo et al., 2014; Wang et al., 2014).

In differentiated epithelial cells, the majority of microtubules are not anchored to the centrosome, but instead their minus-ends are stabilized by binding to a family of proteins, including calmodulin-regulated spectrin-associated proteins (CAMSAPs) in vertebrates, Patronin in Drosophila and PTRN-1 in Caenorhabditis elegans (Hendershott and Vale, 2014; Nashchekin et al., 2016; Richardson et al., 2014). The vertebrate CAMSAP family consists of CAMSAP1, CAMSAP2 (also known as CAMSAP1L1) and CAMSAP3 (also known as Nezha), and each plays roles in cell and tissue morphogenesis (Jiang et al., 2014; Nashchekin et al., 2016; Tanaka et al., 2012; Toya et al., 2015). In intestinal epithelial cells, CAMSAP3 is required to maintain proper microtubule organization and organelle polarization (Toya et al., 2015). The present study aimed to investigate the potential role of CAMSAP3 in the behavior of carcinoma cells. Surprisingly, knockout of CAMSAP3 in lung carcinoma lines upregulated various mesenchymal markers at the transcriptional level in parallel with activation of Akt proteins (also known as protein kinase B proteins, Akt hereafter), an EMT regulator. These findings suggested that CAMSAP3 functions to suppress EMT in lung carcinoma cells and that loss of CAMSAP3 leads to greater cell migratory abilities.

CAMSAP3 genes were knocked out in H460 cells, a large-cell lung carcinoma line, using the CRISPR/Cas9 system (Fig. S1A,B). In CAMSAP3-KO cells, termed H460/C3ko cells, no expression of CAMSAP3 proteins was detected (Fig. 1A). Remarkably, mesenchymal markers, including N-cadherin, Slug and ZEB1, were upregulated in these cells, whereas the epithelial marker, E-cadherin, was downregulated, suggesting that a type of EMT was induced as a result of CAMSAP3 knockout. Changes to EMT marker expression occurred at the transcriptional level, as the amounts of protein detected correlated with their mRNA levels, with the exception of Slug. The increase in Slug protein expression might have been through another mechanism (Fig. 1B). H460/C3ko cells also showed a morphological EMT, exhibiting a mesenchymal shape with a dramatic increase in actin stress fibers anchored to paxillin, a focal adhesion protein (Fig. 1C,D). In contrast, vimentin, an intermediate filament protein that is known to be upregulated after EMT (Ahmad et al., 2012; Myong, 2012), was not increased in H460/C3ko cells (Fig. S1C), indicating that CAMSAP3 loss did not upregulate all mesenchymal markers.

Wound-healing and transwell migration assays demonstrated that cell migration was also enhanced in the absence of CAMSAP3 (Fig. 1E,F). To test if cell motility increased in individual cells, time-lapse images were collected of cells seeded at low densities (Movies 1 and 2). Analysis of cell trajectories showed that H460/C3ko cells changed their positions more vigorously than control cells (Fig. 1G), suggesting that individual cells acquired higher motility after CAMSAP3 knockout. Because EMT is thought to provide a survival mechanism for cancer cells in detached environments (Guadamillas et al., 2011; Wu et al., 2016), a colony-forming assay in soft agar cultures was performed. H460/C3ko cells exhibited an enhanced growth and an increase in the formation of colonies (Fig. 1H), but also exhibited a slightly slower growth rate than control cells in 2D cultures (Fig. S1D).

To confirm if the behavior and properties of H460/C3ko cells depended on the loss of CAMSAP3, rescue experiments were conducted. H460/C3ko rescue cells (C3WT) were generated in which CAMSAP3 was stably reintroduced (Fig. S2A). C3WT restored the original epithelial morphology (Fig. S2B). Furthermore, the original proportion of E-cadherin to N-cadherin was restored at both the protein and mRNA levels (Fig. S2C,D), and cell motility was also reduced in C3WT cells (Fig. S2E). All these findings support the hypothesis that CAMSAP3 is important in maintaining epithelial phenotypes in lung carcinoma cells.

No change in CAMSAP2 expression level was observed when CAMSAP3 was removed, suggesting that CAMSAP2 was not involved in CAMSAP3 KO-mediated EMT (Fig. S1E). Taken together, these observations suggest that CAMSAP3 is important to maintain the epithelial phenotypes of H460 cells. Consistent with this notion, in the lungs of fetal mice, CAMSAP3 was expressed in epithelial cells but not in mesenchymal cells, and CAMSAP2 was expressed in both cell types (Fig. S1F).

To test whether the CAMSAP3 deletion-dependent EMT can be observed in other lung carcinoma cells, we used A549 cells, a lung adenocarcinoma line. We depleted CAMSAP3 using specific siRNAs in these cells, and also in H460 cells for comparison. As a result of CAMSAP3 knockdown, A549 cells displayed EMT, assessed by changes in cell morphology (Fig. S3A), EMT markers (Fig. S3B), and cell migration speed (Fig. S3C), as observed in H460 cells, suggesting that the requirement of CAMSAP3 for epithelial phenotypes is conserved in other lung carcinoma cells.

Additionally, the changes induced by CAMSAP3 depletion were compared to those induced by TGF-β treatment, an established method to induce EMT (Fig. S3A). Upregulation of Slug and ZEB1, as well as downregulation of E-cadherin, were similarly observed in both CAMSAP3-depleted and TGF-β-treated A549 cells (Fig. S3B). However, CAMSAP3 was not reduced in TGF-β-treated cells, suggesting that EMT was induced through distinct signaling pathways between the CAMSAP3-depleted and TGF-β-treated cells. We did not use H460 cells for the TGF-β experiment because this cell line is not responsive to TGF-β (Finger et al., 2008; Yang et al., 2015).

Akt is a key player in EMT (Suman et al., 2014; Xu et al., 2015; Yan et al., 2012). We therefore assessed possible involvement of Akt in CAMSAP3 loss-mediated EMT. Western blot analysis showed that the phosphorylated form of Akt (p-Akt) greatly increased in H460/C3ko cells but that the total Akt level was not altered (Fig. 2A), suggesting that CAMSAP3 removal enhanced Akt activity. p-Akt upregulation was also observed in CAMSAP3-depleted A549 cells (Fig. S4A). Immunostaining showed that p-Akt exhibited punctate distributions in the cytoplasm of control cells and that p-Akt puncta substantially increased after CAMSAP3 knockout, resulting in coverage of lamellipodial edges of the cells (Fig. 2B). These findings suggested that CAMSAP3 normally acts to suppress Akt activity.

To confirm if CAMSAP3 knockout-dependent Akt overactivation is responsible for EMT, H460/C3ko cells were treated with LY294002, an inhibitor of phosphatidylinositol 3-kinase (PI3K), which activates Akt through phosphorylation of its serine 473 and threonine 308 residues. The expression levels of mesenchymal markers were reduced in LY294002-treated cells (Fig. 2C). Wound-scratch assays also showed that attenuation of Akt activation resulted in a reduction of cell migration rate (Fig. 2D). Furthermore, Akt knockdown using specific siRNAs downregulated mesenchymal markers and suppressed cell movement in CAMSAP3-depleted cells (Fig. 2E, Fig. S4B). Thus, these findings demonstrate that CAMSAP3 knockout promotes EMT via p-Akt upregulation.

The mechanism of the effect of CAMSAP3 on Akt activity were investigated by initially testing the possibility that CAMSAP3 might physically interact with Akt. His-tagged CAMSAP3 plasmids were introduced into H460 cells due to the lack of antibodies for endogenous CAMSAP3 useful for this study, and cells were immunostained for both endogenous Akt and His-tagged CAMSAP3. The data showed no overlapping between localization of these two proteins (Fig. S4C). In addition, immunoprecipitation experiments indicated no coprecipitation of CAMSAP3 with either Akt or p-Akt (Fig. S4D). These findings suggest that CAMSAP3 affects Akt activity indirectly.

As CAMSAP3 is a microtubule-binding protein, we next hypothesized that CAMSAP3 might regulate Akt via microtubules. Importantly, it has been reported that localization of Akt on stable or acetylated microtubules is important for sustaining Akt activity (Giustiniani et al., 2009; Jo et al., 2014). We therefore tested whether Akt indeed interacts with microtubules and, if so, whether this interaction is important for Akt activity in H460 cells. Microtubule sedimentation assays showed that at least a fraction of p-Akt or Akt coprecipitated with microtubules, and p-Akt in the pellet fraction was decreased in nocodazole-treated cells (Fig. S5A). Furthermore, nocodazole treatment reduced even the total level of p-Akt (Fig. S5B), and a similar reduction was also observed in nocodazole-treated H460/C3ko cells as revealed by immunoblot and immunofluorescence experiments (Fig. S5C,D). This reduction of p-Akt was rescued by adding Taxol, which stabilizes microtubules, to the nocodazole-treatment medium (Fig. S5C,D). These results confirmed that microtubules are required to maintain proper p-Akt levels and for CAMSAP3 loss-dependent upregulation of p-Akt.

The ability of CAMSAP3 to control interactions between p-Akt and microtubules was next investigated. Previous observations have reported that CAMSAP3 depletion causes changes in the posttranscriptional modification (PTM) of microtubules, such as detyrosination and acetylation (Nagae et al., 2013; Tanaka et al., 2012). Therefore, the effect of CAMSAP3 knockout on PTM was investigated. Tubulin acetylation was greatly enhanced in H460/C3ko cells (Fig. 3A,B) and in CAMSAP3-depleted A549 cells (Fig. S4A). In contrast, the levels of total tubulin and detyrosinated tubulin were not significantly altered (Fig. 3B). C3WT cells were used to confirm if the increased tubulin acetylation depended on CAMSAP3 deletion, and we found that acetylated tubulin level was decreased in correlation with p-Akt reduction in C3WT cells (Fig. 3C,D). Next, we closely observed the distribution of p-Akt and acetylated microtubules by coimmunostaining in mock-transfected H460 (H460/Ctrl) and H460/C3ko cells, finding that at least a fraction of p-Akt puncta overlapped with acetylated microtubules (Fig. 3E), which was consistent with the observation that p-Akt co-precipitated with microtubules (Fig. S5A). Quantitative analysis of the images indicated a small increase in the colocalization of p-Akt and acetylated tubulin in CAMSAP3-deleted cells (Fig. 3E). Moreover, the p-Akt reduction in nocodazole-treated H460/C3ko cells coincided with a reduction of acetylated tubulin (Fig. S5C).

To investigate if enhanced tubulin acetylation plays a role in Akt activation and EMT, we knocked down α-tubulin acetyltransferase 1 (αTAT1, encoded by ATAT1) using specific siRNAs (si-αTAT1) in H460/C3ko cells, as αTAT1 is the sole enzyme responsible for tubulin acetylation in mammals (Akella et al., 2010; Chien et al., 2016; Kalebic et al., 2013; Shida et al., 2010). Although the αTAT1 protein level could not be determined due to a lack of appropriate antibodies for αTAT1, ATAT1 mRNA levels were reduced in response to RNAi (Fig. 4A). Western blot analysis also confirmed that acetylated tubulin levels were decreased following si-αTAT1 transfection (Fig. 4B). Importantly, p-Akt level was also reduced in the αTAT1 knockdown cells, suggesting that Akt activation is promoted by tubulin acetylation. Furthermore, αTAT1 depletion also suppressed the enhanced cell migration and mesenchymal marker expression in H460/3ko cells (Fig. 4C,D). These results suggest that tubulin acetylation is required for CAMSAP3 loss-mediated EMT.

To confirm whether enhanced tubulin acetylation is sufficient to promote EMT, we examined the effect of GFP-tagged αTAT1 overexpression in wild-type H460 cells on p-Akt activation and mesenchymal marker expression. Because transfection efficiency was relatively low (30–40%), we decided to observe p-Akt and EMT markers in individual transfected cells using immunocytochemistry rather than immunoblotting of whole cell lysates. Under this experimental setup, only Slug was detectable by immunostaining as an EMT marker. As expected, overexpression of αTAT1, but not its mutant D517N αTAT1 with no enzymatic activity (Shida et al., 2010), increased tubulin acetylation (Fig. 5A). In the same way, p-Akt and Slug levels were increased in cells overexpressing αTAT1, but not D517N αTAT1 (Fig. 5B), confirming that tubulin acetylation is important for Akt activation and the resultant EMT. We also examined whether Taxol treatment was sufficient to induce EMT. However, cells treated with Taxol did not survive for the periods required for the EMT assays (24 h), and therefore we could not determine whether or not simple stabilization of microtubules is sufficient for EMT induction. Overall, these results suggest that CAMSAP3 loss-induced p-Akt upregulation and EMT promotion are mediated by elevated tubulin acetylation.

Finally, the mechanism by which acetylated microtubules promote Akt activation was examined. A previous study has reported that dynactin p150 (also known as DCTN1, dynactin hereafter), a cofactor of the microtubule dynein complex, mediates the interaction between Akt and acetylated (stable) microtubules to sustain Akt activation (Jo et al., 2014). Therefore, the involvement of dynactin in CAMSAP3-mediated activation of Akt was examined. Dynactin p150 was shown to be increased in H460/C3ko cells (Fig. 6A). When dynactin was depleted in H460/C3ko cells using siRNA, the p-Akt level was reduced, which was consistent with a previous report (Jo et al., 2014). Furthermore, dynactin depletion substantially attenuated the expression of mesenchymal markers in these cells (Fig. 6B), as well as their wound-healing rate (Fig. 6C). In addition, the interaction of dynactin with microtubules was promoted by CAMSAP3 knockout (Fig. 6D). These findings suggested that dynactin is involved in the CAMSAP3-dependent Akt-activation process.

EMT is a process that is thought to promote cancer invasion (Fenouille et al., 2012; Polireddy et al., 2016). The present study suggests that lung carcinoma cells have an intrinsic mechanism to suppress EMT. CAMSAP3 loss caused EMT-like changes along with upregulation of mesenchymal markers in these cells. Studies of mechanisms underlying this phenomenon showed that removal of CAMSAP3 promoted tubulin acetylation, which was consistent with previous studies (Tanaka et al., 2012). Concomitantly, CAMSAP3 knockout cells had a higher level of active Akt than wild-type cells. We showed that inhibition or removal of Akt abrogated CAMSAP3 loss-mediated EMT, and tubulin acetylation was required for Akt activation. Based on these observations, we propose that CAMSAP3 normally suppresses Akt activity by controlling tubulin acetylation and that this process is important to maintain epithelial phenotypes in lung carcinoma cells. Since the cell lines used in this study already expressed a certain level of mesenchymal markers, it is likely that CAMSAP3 loss altered cell behavior by promoting EMT rather than triggering it.

Microtubules are a major cytoskeletal component of cells, and their dynamic properties are regulated by a number of mechanisms, including PTM (Al-Bassam and Chang, 2011; Kadavath et al., 2015; Song and Brady, 2015; Zhang et al., 2015). Acetylation of α-tubulin, a type of PTM, occurs in stable or long-lived microtubules (Janke and Chloë Bulinski, 2011; Portran et al., 2017). Tubulin acetylation is mediated by αTAT1 (Akella et al., 2010; Chien et al., 2016; Kalebic et al., 2013; Shida et al., 2010), and the present study confirmed that this enzyme was required for CAMSAP3 loss-mediated upregulation of acetylation. The question of how CAMSAP3 suppresses αTAT1-dependent tubulin acetylation remains to be clarified. Our recent study suggested that CAMSAP3 maintains dynamic microtubules, and thereby preventing an increase in acetylation (Pongrakhananon et al., 2018). On the other hand, CAMSAP2 expression level did not change in correlation with EMT, suggesting that CAMSAP2 and CAMSAP3 have distinct biochemical or physiological functions.

The Akt signaling pathway is of paramount importance in several sporadic cancers, and it is thus a potential target for anticancer drug development (Altomare and Testa, 2005; Banerji et al., 2017; Hyman et al., 2017). Akt is known to have multiple targets, and EMT regulatory proteins are a target of the Akt signaling pathway, which controls them at both the transcriptional and posttranscriptional level (Suman et al., 2014; Xu et al., 2015). Phosphorylation of Akt at serine 473 and threonine 308 by upstream PI3K initiates its activity; however, its activated state needs to be sustained. Emerging evidence has shown that there is an association between microtubules and Akt activation, and, importantly, it has been suggested that acetylated microtubules are required for prolonged Akt stimulation (Giustiniani et al., 2009; Jo et al., 2014). The present data demonstrates that an increase in acetylated tubulin mediated by CAMSAP3 depletion leads to the upregulation of Akt phosphorylation, despite the lack of cytological colocalization of CAMSAP3 and Akt. This aberrant increase in Akt phosphorylation was diminished by treating cells with either αTAT1 siRNA or microtubule-destabilizing agents, indicating a noticeable influence on Akt function of tubulin acetylation induced by CAMSAP3 loss. Furthermore, αTAT1 overexpression was sufficient for Akt activation and mesenchymal marker upregulation. In addition, dynactin was involved in CAMSAP3 knockout-mediated processes for Akt activation and EMT induction, which agreed with the previous finding that microtubule-dependent Akt activation is mediated by dynactin or its associated proteins (Jo et al., 2014; Kunoh et al., 2010). These observations indicate that Akt activity is elevated via CAMSAP3 knockout-induced increased tubulin acetylation, thus prompting cells to undergo EMT.

EMT is mediated by the binding of various stimuli, such as TGF-β, to specific receptors, which initiates intracellular signaling that regulates the transcription of genes involved in cell–cell adhesion, reorganization of the cytoskeleton, and survival mechanisms to support metastasis (Heerboth et al., 2015; Lotz-Jenne et al., 2016). However, our results suggest that distinct signaling pathways are used for EMT in CAMSAP3-deficient and TGF-β-treated cells. Because these cells share upregulation of some EMT markers, it is likely that both pathways merge in some signaling steps, but further studies are required to understand these mechanisms. Of note, changes of EMT markers in CAMSAP3-deficient cells occurred at the transcriptional level, suggesting that CAMSAP3 loss-dependent EMT is not induced by simple reorganization of the cytoskeletal system but instead involves complex gene regulation that may depend on signaling events downstream of Akt. In the present study, we assayed EMT using 2D-culture dishes, but it is possible for cells to migrate through 3D environments in vivo. Different types of cell attachment and migration may affect EMT signaling, and therefore it would be important to test the role of CAMSAP3 in EMT under 3D conditions in future studies.

The present study suggests that if CAMSAP3 is mutated in vivo during cancer progression, it would lead to overactivation of Akt, thereby promoting EMT. In contrast, a recent study has shown that although CAMSAP3 mutation or depletion interferes with the intracellular architecture in mouse intestinal cells (Toya et al., 2015), it does not induce EMT in these cells. It is possible that carcinoma cells are more susceptible to CAMSAP3 loss or EMT-inducing signals than normal epithelial cells. In conclusion, the present findings provide evidence for the remarkable function of CAMSAP3 in Akt signaling via a tubulin acetylation-dependent mechanism, highlighting a cooperation between these proteins that influences cancer cell behavior.

NCI-H460 and A549 cells were purchased from American Type Culture Collection (ATCC; Manassas, VA, USA). The cells were not authenticated, but they were tested for contamination. H460 cells were cultured in RPMI medium, and A549 cells in DMEM, both incubated at 37°C with 5% CO₂. Both media were supplemented with 10% fetal bovine serum, 1% L-glutamine and 1% penicillin/streptomycin. Cell culture media, supplements and Phalloidin were obtained from Invitrogen. Taxol, nocodazole, LY294002 and DAPI were purchased from Sigma. The antibody against CAMSAP3 was generated as previously described (Tanaka et al., 2012) and used at 1:500. Other antibodies were purchased as follows: rabbit anti-CAMSAP2 (1:1000 for immunoblotting; Proteintech, 17880-1-AP), rabbit anti-Slug (1:1000 for immunoblotting, 1:200 for immunofluorescence; Cell Signaling Technology, 9585), rabbit anti-ZEB1 (1:1000 for immunoblotting; Cell Signaling Technology, 3396), rabbit anti-E-cadherin (1:1000 for immunoblotting; Cell Signaling Technology, 3195), rabbit anti-N-cadherin (1:1000 for immunoblotting; Cell Signaling Technology, 13116), rabbit anti-phosphorylated Akt (S437; 1:1000 for immunoblotting, 1:25 for immunofluorescence; Cell Signaling Technology, 9271), rabbit anti-Akt (1:1000 for immunoblotting, 1:200 for immunofluorescence; Cell Signaling Technology, 9272), mouse anti-GAPDH (1:1000; Cell Signaling Technology, 97166), mouse anti-tubulin (1:5000 for immunoblotting, 1:1000 for immunofluorescence; Sigma, T6199), mouse anti-acetylated tubulin (1:5000 for immunoblotting, 1:1000 for immunofluorescence; Sigma, T7451), mouse anti-paxillin (1:500 for immunofluorescence; BD Biosciences, 610051), mouse anti-His (1:1000 for immunoblotting, 1:500 for immunofluorescence; MBL, D291-3), mouse anti-dynactin (1:2000 for immunoblotting; BD Biosciences, 610473), rat anti-ECCD2 (1:400 for immunofluorescence; Shirayoshi et al., 1986) and rat anti-α-tubulin (1:5000 for immunoblotting, 1:1000 for immunofluorescence; Millipore, MAB1864). The secondary antibodies used were as follows: goat AlexaFluor 488-, 568-, 555- and 647-conjugated anti-mouse, -rabbit or -rat IgG (1:000 for immunofluorescence; Invitrogen), sheep HRP-conjugated anti-mouse and anti-rabbit IgG (1:5000 for immunoblotting; Cell Signaling Technology).

sgRNA sequences were designed to target CAMSAP3 exon 1 by an online tool (http://crispr.mit.edu/) as follows: sgRNA#1, 5′-CACCGACTAGAAAGGTCCTCCGCAG-3′; and sgRNA#2, 5′-CACCGAGCCCAGCCCAGTCCGAGCG-3′.

The CRISPR/Cas9 plasmid was constructed as described previously with some modifications (Ran et al., 2013). Plasmids expressing sgRNA were cloned by annealing each DNA oligo and ligating into pSpcas9-2A-puro (Addgene #48139, Cambridge, MA, USA) at the BsbI site. The sgRNA sequencing of vectors was conducted using the U6F primer. Transfection was performed using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Briefly, 2 μg of each plasmid in optiMEM was mixed with 6 μl of Lipofectamine 2000. After 30 min, the mixture was added to cells in culture, and cells were incubated at 37°C for 6 h. The addition of puromycin antibiotics was performed the next day. The expression of CAMSAP3 was confirmed by immunoblotting and real-time qPCR. The CAMSAP3 knockout cells were designated H460/C3ko, and the control mock transfectant cells were designated H460/Ctrl.

For construction of the CAMSAP3-expressing plasmid, full-length CAMSAP3 cDNA with a His-tag sequence on its 3′ end was cloned into pCANw. To generate αTAT1 plasmids, αTAT1 wild-type and D157N cDNA were amplified from pEF5B-FRT-GFP-αTAT1 and pEF5B-FRT-GFP-αTAT1-D157N (Addgene), respectively. A Not1 site and Kozak sequence at the 5′ terminus as well as a Sal1 site at the 3′ terminus were added to the cDNAs, and cDNAs were then inserted into the pCA-Sal-GFP vector. Transfection was performed using Lipofectamine 2000 according to the manufacturer's instructions. Briefly, 3 μg of plasmid in optiMEM media was mixed with 6 μl of Lipofectamine 2000. After 30 min, the mixture was added to cells in culture, and cells were incubated at 37°C for 6 h. Stable transfectants were isolated by culturing cells in a medium supplemented with antibiotic G418 (400 µg/ml) for at least 7 days followed by subsequent cloning. Expression of exogenous DNA was confirmed by immunoblotting or immunofluorescence assay.

Cells were transfected with siRNAs specific for target proteins by Lipofectamine RNAiMAX according to the manufacturer's protocol (Invitrogen). Stealth RNAis targeting CAMSAP3 or α-TAT1 with the following sequences and control siRNA were purchased from Invitrogen: siCAMSAP3, 5′-ACAGUGGCAGCAGUUCUCCUGUCUU-3′; siα-TAT1#1, 5′-ACCGCACCAACTGGCAATTGA-3′; and siα-TAT1#2, 5′-GAGCCAUUAUUGGUUUCCUCAAAGU-3′.

For Akt and dynactin knockdown experiments, the following siRNAs targeting Akt and dynactin were synthesized and annealed: siAkt, sense 5′-GGAGAUCAUGCAGCAUCGC-3′ and anti-sense 5′-GCGAUGCUGCAUGAUCUCC-3′; si-mismatch control, sense 5′-GGGAAUCAUAAAGCAUUUC-3′ and anti-sense 5′-CCGGGGCUGCAUAAACUUC-3′; siDynactin, sense 5′-GACUUCACCCCUUGAUUAA-3′ and anti-sense 5′-UUAAUCAAGGGGUGAAGUC-3′; and si-mismatch control, sense 5′-GCUACUUCGUCCAAUCAUA-3′ and anti-sense 5′-UAUGAUUGGACGAAGUAGC-3′.

Briefly, 100 nM siRNAs in optiMEM were incubated with Lipofectamine RNAiMAX mixture for 15 min at room temperature and then added to cells, and cells were incubated at 37°C for 6 h. At 72 h after transfection, cells were subjected to western blot analysis or immunofluorescence assay.

For the wound-healing assay, 1.5×10⁴ cells were seeded into each well of a 24-well culture plate. After cells had reached confluence, wound scratches were generated using a pipette tip, and detached cells were removed by washing with PBS. Images were acquired at indicated time points, and wound area was quantified by ImageJ software. For transwell migration assays, 5×10⁴ cells in a serum-free medium were placed in the upper chamber with a 0.8 μm pore-sized membrane, and complete culture medium was added into the lower chamber. After 24 h, cells on the upper side of the membrane were swabbed out, and those on the lower side of the membrane were fixed with cold methanol at ­−20°C for 5 min followed by incubation with DAPI for 10 min. Migrating cells were imaged randomly using an Olympus IX51 fluorescence microscope.

Cells were cultured in chamber slides (Lab-Tek II, Thermo Fisher Scientific) overnight. Images were captured every 16 min by an Olympus FluoView FV10i confocal laser-scanning microscope, with 5% CO₂ at 37°C for 6:40 (h:min). The displacement of individual cells was tracked by Fiji software using the TractMate plug-in and analyzed by Microsoft Excel-executable DiPer program (Gorelik and Gautreau, 2014). The total trajectories of cell migration from the start to the end position were plotted using Plot At Origin macro.

Two-thousand cells were seeded into each well of 96-well plates with at least five replicates. After the indicated times, medium was replaced with the MTT solution (0.5 mg/ml), and plates were incubated at 37°C for 4 h. Dimethyl sulfoxide (DMSO; 100 μl) was added to dissolve formazan products, and the absorbance was measured at 570 nm using a microplate reader. Cell proliferation was calculated relative to the initial time point.

A 24-well plate was coated with 0.3% agarose in complete medium as the bottom layer. After solidification, 10³ cells were suspended in 0.5% agarose in medium and seeded onto the bottom layer. Cells were incubated for 14 days, and medium was added every 2 days to prevent dryness. Cells were then stained with 0.01% crystal violet in 10% ethanol for 30 min at room temperature. After washing several times with deionized water, colonies were collected and counted, and the colony number was presented as a relative number to the control cells using ImageJ software with the particle analysis plugin.

Cells were suspended in TMN lysis buffer (20 mM Tris-HCl, pH 7.5; 1 mM MgCl₂; 150 mM NaCl; 20 mM NaF; 0.5% sodium deoxycholate; 1% nonidet-40; 0.1 mM phenylmethylsulfonyl fluoride; and protease inhibitor cocktail, Roche) for 45 min on ice. The protein content was measured by BCA Protein Assay Reagent Kit (Thermo Scientific). Proteins were separated by SDS-PAGE and transferred to PVDF membranes. Blots were blocked with 5% skim milk in TBST (Tris buffer saline with 0.075% Tween-20), incubated with a specific primary antibody overnight at 4°C and incubated with a corresponding secondary antibody for 2 h. The protein expression levels were visualized by the enhanced chemiluminescence system using SuperSignal West Pico (Thermo Scientific) and Immobilon Western (Millipore).

Cells were treated with 1 µM Taxol for 30 min at 37°C and lysed by a microtubule stabilizing buffer (MTB) containing 80 mM PIPES, 80 mM K-1,4-piperazinediethanesulfonic acid (pH 6.8), 1 mM EGTA, 1 mM MgCl₂, 0.5% (vol/vol) nonidet P-40, 20 mM NaF, 0.5% sodium deoxycholate, 10 mM Taxol, 0.1 mM phenylmethylsulfonyl fluoride, and protease inhibitor cocktail (Roche) for 5 min at 37°C in dark. The lysate was fractionated into pellet and supernatant by centrifugation at 17,400 g for 15 min at 30°C. The pellet was washed with MTB lacking detergent and resuspended with MTB in an equal volume of supernatant. Both pellet and supernatant fractions were boiled at 95°C with sampling buffer and subjected to western blot analysis.

Total RNA was isolated from cells using GENEzol reagent (Geneaid Biotech, Shijr, New Taipei, Taiwan). RNA (1 μg) was reverse transcribed to cDNA using ProtoScript II Reverse transcriptase (New England BioLabs) as described by the manufacturer's instructions. mRNA expression levels of CAMSAP3, E-cadherin, N-cadherin, Snail, Slug and ZEB1 were measured by a Bio-Rad T100 Thermal Cycler using 2× iTaq Universal SYBR Green Supermix (Bio-Rad). The primer pairs used are indicated in Table S1. The thermocycling conditions were as follows: 95°C for 10 min; and 35 cycles at 95°C for 30 s and 60°C for 30 s. The data were calculated using the ΔΔC_t method (Livak and Schmittgen, 2001). Each sample was performed in triplicate.

Cells were fixed with 4% paraformaldehyde in PBS for 20 min at room temperature in the dark and permeabilized in 0.1% Triton X-100 in PBS for 10 min. For microtubule staining, cells were fixed with cold methanol for 5 min at −20°C. Nonspecific signals were blocked by treatment with 3% BSA for 30 min or longer followed by incubation with primary antibodies overnight at 4°C. Cells were washed with PBS and incubated with secondary antibody for 2 h at room temperature in the dark. After washing with PBS containing DAPI, coverslips were washed with deionized water and subsequently mounted using FluorSave (EMD Millipore). Confocal images were acquired by either a Zeiss LSM880 through a Plan-Apochromat 63×/1.40 N.A. or a Leica TCS SP8 with 100× oil immersion objective lens. Image preparation and analysis were performed using ImageJ software. Colocalization of p-Akt and acetylated tubulin was quantified using ImageJ with JaCoP plugin, and presented as Manders' colocalization coefficient (Bolte and Cordelières, 2006; Bravo-Sagua et al., 2016).

The mouse experiment was performed in accordance with the protocol(s) approved by the Institutional Animal Care and Use Committee of RIKEN Kobe Branch. For lung tissue preparation, lungs were removed from embryonic day 18.5 C57BL/6N mice and fixed with 2% paraformaldehyde in PEM buffer with 75 mM sorbitol for 1 h at room temperature. Tissues were frozen in OTC solution (Sakura Finetek) and sectioned at a 7 µm thickness prior to immunostaining. Sections were incubated in HistoVT One (Nacalai) at 95°C for 5 min to retrieve antigens and then incubated in 3% BSA and 3% FCS in PBST for blocking. Sections were then incubated in the blocking solution containing primary antibody at 4°C overnight. Sections were then incubated in blocking solution containing a fluorochrome-conjugated secondary antibody for 2 h at room temperature. Samples were mounted on MAS-coated glass slides (Matsunami) using FluorSave reagent (Calbiochem). Images were acquired by a Zeiss Axioplan2 through a Plan-Apochromat 63×/1.40 N.A. Image preparation and analysis were performed using AxioVision software (Zeiss).

Cells were lysed in lysis buffer (20 mM Tris-HCl, pH 7.5; 1 mM MgCl₂; 150 mM NaCl; 20 mM NaF; 10 mM EGTA; 0.5% sodium deoxycholate; 1% nonidet-40; 0.1 mM phenylmethylsulfonyl fluoride; and protease inhibitor cocktail, Roche) for 45 min on ice. The supernatant was collected by centrifugation at 20,000× g for 20 min at 4°C, precleared with protein G-conjugated Sepharose beads (GE Healthcare) for 1 h, and incubated with specific antibody or IgG as control overnight. Protein complexes were pulled down by incubation with Protein G-conjugated Sepharose beads for 1 h at 4°C and then washed five times with lysis buffer. Precipitates were boiled with sample buffer at 95°C for 5 min and analyzed by immunoblotting.

All data are presented as the mean±s.e.m. obtained from at least four independent experiments. Statistical analysis was performed using unpaired Student's t-test or Mann-Whitney U-test using Prism 7 (GraphPad). P-values less than 0.05 were considered statistically significant.

We thank Hiroko Saito for lung tissue preparation.