CAMSAP3 accumulates in the pericentrosomal area and accompanies microtubule release from the centrosome via katanin

Two kinds of microtubules (MTs), centrosomal and non-centrosomal, exist in animal cells, and play a role in many cellular processes, including polarity establishment and cell migration (Bartolini and Gundersen, 2006). In contrast to centrosomal MTs, which nucleate from the centrosome, non-centrosomal MTs are generated through several distinct mechanisms (Keating and Borisy, 1999). Time-lapse fluorescence microscopy has demonstrated that non-centrosomal MTs arise primarily through constitutive nucleation and release from the centrosome, or are generated from other non-centrosomal sites (Chabin-Brion et al., 2001; Keating et al., 1997; Meng et al., 2008; Mogensen and Tucker, 1987; Tanaka et al., 2012; Tassin et al., 1985). However, how these non-centrosomal MTs are generated remains largely unknown.

In recent years, many observations have provided deep insights into the release mechanism generating non-centrosomal MTs in epithelial cells. During these processes, centrioles migrate to the apical surface, the radial MT array is, for the most part, lost and an apico-basal MT array develops (Meads and Schroer, 1995; Mogensen, 1999; Moss et al., 2007). Further reports have noted that cytoplasmic dynein contributes to the transport of non-centrosomal MTs into the axon, and the release from centrosome is essential for non-centrosomal MT generation (Ahmad et al., 1998, 1999). However, how non-centrosomal MTs are generated in epithelial cells remains unclear.

In this study, we observed that CAMSAPs, a family of proteins that bind to non-centrosomal MT minus-ends, accumulate at the pericentrosomal area and accompanies the detachment of MTs from the centrosome. MT regrowth assays showed that dynein activity is responsible for the accumulation of CAMSAPs in the pericentrosomal area. We also found that katanin promotes the release process but that it also requires the accumulation of CAMSAPs in the pericentrosomal area, furthering our understanding of the MT release mechanism. In addition, depletion of either CAMSAPs or katanin prevented the process of MT release. Thus, our results show that CAMSAPs and katanin cooperate to help non-centrosomal MTs release from the centrosome in epithelial cells.

In a monolayer culture of DLD1 cells, an epithelial cell line, MTs are mainly non-centrosomal, and CAMSAP3 puncta are scattered in the cytoplasm, especially in the pericentrosomal area (Fig. 1A; Movie 1). To investigate the dynamics of CAMSAP3 in the pericentrosomal area, we treated confluent DLD1 cells with nocodazole to depolymerize MTs and then observed the MT regrowth process (Fig. 1B). After treatment, CAMSAP3 puncta became diffuse. Strikingly, at 5 min after washout, CAMSAP3 began to concentrate around the centrosome. CAMSAP3 signals subsequently became stronger with increased numbers of astral MTs. At 30 min after washout, CAMSAP3 distribution in the pericentrosomal area decreased (Fig. 1C). Additionally, we transfected DLD1 cells with Kusabira Orange (KOR)-tagged CAMSAP3 and PLEKHA7–GFP, which label the centrosome (Meng et al., 2008). Time-lapse movies showed that CAMSAP3 clusters distributed around the centrosome and moved away, suggesting that this distribution of CAMSAP3 is dynamic during the MT regrowth process (Fig. 1D; Movie 1). Notably, the expression level of CAMSAP3 did not change during the MT regrowth process (Fig. S1A). These findings indicate that the dynamics of CAMSAP3 in the pericentrosomal area correlate with those of the astral MTs. Its dynamic distribution in DLD1 cells was distinct from that for CAMSAP1 and CAMSAP2 (Fig. S1B).

Previous studies have shown that depletion of CAMSAP3 in Caco2 cells enhances centrosomal MTs (Tanaka et al., 2012), and the same result was observed using DLD1 cells here (Fig. S1C,D). Our time-lapse recordings also showed that centrosomal MTs were released from the centrosome together with CAMSAP3 clusters in a MT regrowth assay (Fig. 2A; Movie 2). Co-immunostaining for EB1 (also known as MAPRE1) showed that CAMSAP3 localized at the minus ends of the released microtubules (Fig. S1E). These results led us to doubt whether CAMSAP3 promotes centrosomal MT release from the centrosome. To investigate this further, an MT regrowth assay was performed using control or CAMSAP3-depleted cells. Immunostaining showed enhanced radial MTs in CAMSAP3-depleted cells. Remarkably, numerous short non-centrosomal MTs around the centrosome were observed in control cells but not in CAMSAP3-depleted cells (Fig. 2B). Immunostaining of cells stably expressing CAMSAP3–GFP showed decreased numbers of centrosomal MTs compared with control cells that stably expressed GFP (Fig. 2C). Taken together, these observations indicate that CAMSAP3 accumulates in the pericentrosomal area to aid release of MTs. In contrast, many centrosomal proteins, such as ninein, dynactin, CEP135, BBS4, PCM-1, Nude l and EB1 have been suggested to have a role in MT anchorage, and their depletion results in failure of centrosomal MT anchorage (Askham et al., 2002; Guo et al., 2006; Ibi et al., 2011; Louie et al., 2004; Quintyne et al., 1999). Future work will focus on the details of the mechanism of centrosomal MT anchorage.

Cytoplasmic dynein is essential for the transport of MTs from the centrosome into the axon in neurons (Ahmad et al., 1998), and MT transport toward the cell periphery is blocked when dynein activity is impaired in migrating cells (Abal et al., 2002). Given that we found that the accumulation of CAMSAP3 in the pericentrosomal area depends on MTs (Fig. 1B), we hypothesized that dynein may be involved in the pericentrosomal accumulation of CAMSAP3. To test this idea, we focused on the accumulation of CAMSAP3 in the pericentrosomal area during the MT regrowth process and found that depletion of the dynein intermediate chain retarded this accumulation (Fig. S2A). Next, we transfected DLD1 cells with p150^Glued (also known as DCTN1) tagged with GFP (p150–GFP), which impairs dynein-mediated functions (Abal et al., 2002; Kardon and Vale, 2009). As revealed by the MT regrowth assay, p150–GFP-transfected cells showed a significant reduction in the pericentrosomal distribution of CAMSAP3 compared with cells transfected with GFP alone (Fig. S2C,E). These data suggest that dynein regulates the accumulation of CAMSAP3 in the pericentrosomal area. We also confirmed this result using the dynein inhibitor ciliobrevin D and activator nordihydroguaiaretic acid (NDGA) (Arasaki et al., 2007; Firestone et al., 2012) (Fig. S2B,D,E). Taken together, our data suggest that the pericentrosomal distribution of CAMSAP3 depends on dynein. Furthermore, we found that excess CAMSAP3 could partially rescue the phenotype of ciliobrevin D treatment (Fig. S2F,G), suggesting that the molecular mechanisms of CAMSAP3 accumulation in the pericentrosomal area are complex, and these issues will be taken up in future studies.

It has been shown that katanin is responsible for severing MTs from the centrosome in neurons. Interestingly, recent studies have found that katanin interacts with CAMSAP2 and CAMSAP3 (Jiang et al., 2014). Therefore, we intended to explore the functions of katanin in the process of MT release. We found that katanin was prominently localized at the centrosome as previously reported (Fig. 3A) (Ahmad et al., 1999; McNally and Thomas, 1998), and it also colocalized with CAMSAP3 in the pericentrosomal area (Fig. 3B). In addition, katanin p60 (also known as KATNA1) showed the same dynamics as CAMSAP3 at the centrosome during MT regrowth process (Fig. 3C).

In order to examine whether the severing function of katanin is required for this releasing process, we depleted katanin p60 and then performed a MT regrowth assay (Fig. 3D,E). About 30 min after nocodazole washout, radial MTs still remained in katanin-p60-depleted cells, whereas MTs had returned to normal in control cells (Fig. 3E). In addition, enhanced centrosomal MTs were also observed in cells that had been depleted of katanin p60 (Fig. 3F). Meanwhile, more CAMSAP3 was localized in the pericentrosomal area (Fig. 3G), suggesting that the distribution of CAMSAP3 is caused by the density of microtubules in the pericentrosomal area, as in a previous experiment (Fig. 1B). However, depletion of CAMSAP3 had no effect on katanin p60 distribution (Fig. S3A). These results suggest that katanin functions to sever MTs from centrosomes in epithelial cells.

Since both CAMSAP3 and katanin accumulate in the pericentrosomal area and negatively regulate the radial MT array (Fig. 3A-F), we next examined the relationship between CAMSAP3 and katanin in the severing process of MTs from the centrosome. We performed an MT regrowth assay using cell lines that stably expressed CAMSAP3–KOR or its deletion mutant CC1F–KOR, which lacks the domain required for binding to katanin p60 (deletion of residues 1–594 of CAMSAP3) (Akhmanova and Hoogenraad, 2015; Jiang et al., 2014). Although katanin p60 localized normally at the centrosome in both cell lines (data not shown), MT release from the centrosome in the CC1F–KOR cell line was significantly delayed relative to that in the CAMSAP3–KOR cell line (Fig. 3H), suggesting that the severing activity of katanin requires its interaction with CAMSAP3.

Next, we focused on U2OS cells, which have low CAMSAP3 expression and prominent centrosomal-centered MT arrays (Fig. S3B). We transfected U2OS cells with CAMSAP3–GFP and CC1F–GFP, and immunostaining revealed the loss of centrosomal-centered MT arrays in cells expressing CAMSAP3–GFP, whereas cells that expressed CC1F–GFP still possessed centrosomal-centered MT arrays (Fig. S3C). Moreover, there was a significant distribution of CC1F–GFP around the centrosome (Fig. S3C). These results support the centrosomal distribution of CAMSAP3 as an initial condition required for release of MT from the centrosome.

To verify that CAMSAP3 promotes MT release via katanin severing activity, we knocked down katanin p60 in U2OS cells that had been transfected with CAMSAP3–GFP. The centrosomal-centered MT array reoccurred in these cells (Fig. S3D). Taken together, these results support our model that CAMSAPs and katanin cooperate in the process of releasing MTs from the centrosome. However, depletion of CAMSAP3 did not affect the centrosomal distribution of katanin (Fig. S3A). How CAMSAP3 activates the severing activity of katanin remains a question open to future study.

Finally, to ask how the MT network forms in epithelial cells, we focused on the MT organization process of monolayer culture DLD1 cells, because single DLD1 cells exhibited mainly centrosomal-centered MT arrays while confluent cells did not (Fig. 4A).

To explore the mechanism behind MT organization under different conditions, we performed immunostaining in single and confluent cells. More CAMSAP3 signal was found in the pericentrosomal area in single cells (Fig. 4B), suggesting that the localization of CAMSAP3 and katanin is related to the condition of the cells. We then transfected CAMSAP3–KOR into single DLD1 cells and found that the centrosomal MT array was lost compared with CC1F–KOR or untransfected cells (Fig. 4C), consistent with previous reports (Jiang et al., 2014; Tanaka et al., 2012). Consistently, depletion of CAMSAP3 induced a centrosomal-centered MT array in confluent DLD1 cells (Fig. S1D). These data suggest that CAMSAP3 is a key factor controlling the MT reorganization process in epithelial cells.

Since cell–cell junctions are involved in several signaling pathways, such as outside–inside signaling mediated by E-cadherin to control contact inhibition and other phenomena (Meng and Takeichi, 2009), we speculated that cell attachment might have an effect on MT organization. We therefore depleted E-cadherin to disrupt Ca²⁺-dependent cell–cell junctions and subsequently found that E-cadherin knockdown enhanced centrosomal-centered MT arrays (Fig. 4D). Meanwhile, more CAMSAP3 and katanin p60 was distributed to the centrosome (Fig. 4D,E). These data suggest that cell attachment is involved in the organization of non-centrosomal MTs.

To exclude the possibility that CAMSAP3, katanin p60 or E-cadherin depletion affect centrosomal nucleation, we performed immunostaining for EB1 and found no change in the intensity of EB1 in the pericentrosomal area after knock down (Fig. S4A,B). In conclusion, our experiments demonstrate that CAMSAP3 promotes an MT-releasing process through the severing activity of katanin in response to distinct conditions in epithelial cells (Fig. S4C). However, no evidence suggests that CAMSAP3 is directly involved in the MT severing process. CAMSAP3 may indirectly affect the severing activity of katanin via MT modification (Sudo and Baas, 2010) or it may competitively bind to katanin (Sudo and Maru, 2008). Moreover, how exactly MTs are released from the centrosome is not yet understood. Previous reports suggest that overexpression of Lis1 could promote MT release (Smith et al., 2000), but the detailed mechanism must still be elucidated. Finally, we think CAMSAP3-mediated MT release partially contributes to a change in MT arrangement because alterations in the stability of MT minus ends also could contribute to the observed phenotypes in part.

Human katanin p60 was obtained from a human brain cDNA library by performing PCR amplification and inserted into a pEGFP vector (Clontech). Mouse CAMSAP3–GFP and CAMSAP3–KOR were generated as described previously (Meng et al., 2008). siRNAs were used to knock down CAMSAP3 (Invitrogen) and katanin p60: si-CAMSAP3: 5′-CCAUGUCCAUGAGCGUCGAU-3′; si-Katanin-p60: 5′-GGCUCGAUUUUAUUCUCCATT-3′; negative-control siRNAs were obtained from Invitrogen.

Rabbit polyclonal antibody anti-mouse CAMSAP3 has been described previously (Meng et al., 2008). The specificity of antibodies was checked by knocking down the protein of interest and then immunostaining or western blotting to confirm the specificity of signals. Primary antibodies were: mouse anti-α-tubulin (Sigma, T6074, 1:2000) and rat anti-α-tubulin (Millipore, MAB1864, 1:1000); mouse and rabbit anti-γ-tubulin (Abcam, ab93867, ab11317, 1:500); mouse and rabbit anti-GFP (MBL, M048-3, 598, 1:300); rabbit anti-CAMSAP2 (Sigma, HPA027302, 1:250); mouse anti-katanin p60 (RD, MAB7100, 1:200); mouse anti-EB1 (BD Biosciences, 610535, 1:250); mouse anti-GM130 (BD Biosciences, 610822, 1:3000). Secondary antibodies: goat Alexa-Fluor-488-, -555- or -647-conjugated anti-mouse or anti-rabbit IgG (Invitrogen); CF488A-conjugated anti-rat IgG (Biotium).

Cell culture and immunofluorescence staining were as described previously (Meng et al., 2008). Cells at approximately 70% confluence were transfected by use of Lipofectamine 2000 or RNAiMAX (Invitrogen). For isolating stable transfectants, cells were exposed to 500 µg/ml G418.

Cells were treated with 10 µM nocodazole on ice for 30 min. After removing nocodazole, cells were washed with ice-cold 1×PBS twice, and then incubated in a 1:1 mixture of Dulbecco's modified Eagle's medium and Ham′s F-12 supplemented with 10% FBS, and subjected to fixation or live-cell imaging.

Time-lapse imaging was performed using a Deltavision microscope (Applied Precision). Sequential images were acquired at 3 s or 5 min intervals at 37°C. Fluorescence intensity was measured by using ImageJ. All statistical analysis was performed by using Student's t-test (two-tailed) with GraphPad Prism. Representative data from three independent experiments is presented.

We thank Zhiheng Xu, Shilai Bao and Yingchun Wang, as well as members of the Meng laboratory, for advice.