Cortactin-binding protein 2 increases microtubule stability and regulates dendritic arborization

Neurons are highly differentiated cells composed of specialized subcellular compartments (i.e. axons, dendrites and synapses) that are supported by different types of cytoskeleton. For instance, an F-actin cytoskeleton is the major type of cytoskeleton that determines the morphology of dendritic spines, which are the location of excitatory synapses in mammals (Rácz and Weinberg, 2013). Microtubules, by contrast, are mainly present in dendritic and axonal shafts, although they also transiently invade dendritic spines (Gu et al., 2008; Hoogenraad and Bradke, 2009; Hu et al., 2008; Jaworski et al., 2009; Luo, 2002). Because axons, dendrites and dendritic spines are all neuron-specific subcellular structures, it is reasonable to speculate that there are neuron-specific regulators that control the behaviors of the F-actin and microtubule cytoskeletons of these neuron-specific compartments and that are involved in their formation, maintenance and regulation. Cortactin-binding protein 2 (CTTNBP2), a neuron-specific protein (Chen et al., 2012; Chen and Hsueh, 2012; Ohoka and Takai, 1998), has been shown to play such a role in dendritic spine formation.

CTTNBP2, also known as CortBP2 (Cheung et al., 2001) or CBP90 (Ohoka and Takai, 1998), interacts with the SH3 domain of cortactin through its C-terminal proline-rich domain (Chen and Hsueh, 2012; Ohoka and Takai, 1998). Cortactin is widely expressed in various cells, including neurons. In mature neurons, cortactin is highly enriched at dendritic spines, where it promotes F-actin branching and bundle formation and thus enlarges the dendritic spine heads (Hering and Sheng, 2003; Huang et al., 1997; Weaver et al., 2001). CTTNBP2 is also highly concentrated in dendritic spines (Chen and Hsueh, 2012). The neuron-specific expression of CTTNBP2 suggests that the interaction between CTTNBP2 and cortactin likely controls the neuron-specific function of cortactin. Indeed, CTTNBP2 is crucial for determining the mobility of cortactin in dendritic spines and thus controls dendritic spine formation and maintenance (Chen and Hsueh, 2012). In addition to cortactin, CTTNBP2 also binds to striatin and zinedin, the regulatory B subunits of protein phosphatase 2A (PP2A) in neurons through an interaction between the N-terminal coiled-coil (NCC) domains of both CTTNBP2 and striatin or zinedin (Chen et al., 2012). Similar to cortactin, the dendritic spine distribution of striatin and zinedin is also controlled by CTTNBP2 (Chen et al., 2012). CTTNBP2, therefore, regulates dendritic spine formation and signaling through interaction with cortactin and the PP2A complex.

While analyzing the expression pattern of CTTNBP2 in COS cells, we noticed that in addition to being present in the cell cortex, where CTTNBP2 colocalizes with cortactin and the F-actin cytoskeleton, CTTNBP2 is also associated with some filamentous structures (Chen et al., 2012; Chen and Hsueh, 2012). In this study, we identified these filamentous structures as microtubules and found that CTTNBP2 also interacts with microtubules and increases microtubule stability. Our study suggests that in addition to associating with the F-actin cytoskeleton, CTTNBP2 also regulates neuronal morphology through controlling microtubule stability.

Using a COS cell expression system, our previous studies have shown that CTTNBP2 colocalizes with cortactin and F-actin at the cell cortex (Chen et al., 2012; Chen and Hsueh, 2012). However, we noticed that, in addition to the cell cortex, CTTNBP2 proteins were also present in some filamentous structures that were not positive for F-actin (supplementary material Fig. S1A). The percentage of cells possessing CTTNBP2 filaments varied from batch to batch, ranging from ∼30% to ∼80% of CTTNBP2-transfected cells. Although there was variability in the percentage, such cells were always present in CTTNBP2-expressing cultures. We were therefore curious about what such structures were. After double immunostaining with anti-tubulin antibody, confocal analysis indicated that the CTTNBP2-positive filamentous structures colocalized with microtubules (supplementary material Fig. S1B). We also used three-dimensional structured illumination microscopy (3D-SIM) to further investigate the interaction between CTTNBP2 and microtubules. It was clear that CTTNBP2 attaches to microtubules and forms filamentous structures along microtubules in transfected COS cells (supplementary material Fig. S1C).

To confirm the association of CTTNBP2 with microtubules, we performed a nocodazole treatment and washout experiment in CTTNBP2-transfected COS cells. Nocodazole is an agent commonly used to induce microtubule depolymerization. If the CTTNBP2 associated filaments were indeed microtubules, nocodazole treatment would be expected to disrupt the filamentous structures. In cells treated with vehicle, ∼71% of CTTNBP2-expressing cells possessed CTTNBP2-positive filaments, which were colocalized with microtubules (Fig. 1A,B). After nocodazole treatment for 2 h, microtubule filaments were disrupted in COS cells, and the CTTNBP2 filamentous structures also disappeared (Fig. 1A,B). After washout of nocodazol for 2 h, microtubules reassembled into a filamentous network (Fig. 1A). CTTNBP2-positive filaments also reoccurred and associated with microtubules in ∼63% of transfected cells (Fig. 1A,B). Taken together, these cell biology analyses suggest that CTTNBP2 associates with microtubules.

We also performed a biochemical study – a microtubule-binding protein spin-down assay – to investigate the association between CTTNBP2 and microtubules. In the assay, purified tubulins are used to form microtubules that are concentrated in pellets after polymerization and sedimentation, which are then incubated with cell lysates from COS cells transfected with CTTNBP2 or vector control. CTTNBP2 co-fractionated with microtubules in the pellets (Fig. 1C,D), confirming the association of CTTNBP2 with microtubules. In the absence of microtubules, CTTNBP2 was not noticeably precipitated by the centrifugation (Fig. 1D), supporting the specificity of the presence of CTTNBP2 in the microtubule fraction.

We then investigated which region of CTTNBP2 is involved in the association with microtubules. A series of CTTNBP2 fragments (Fig. 1C) were constructed and expressed in COS cells. After removing endogenous microtubules, the COS cell lysates were then used in a microtubule-binding protein spin-down assay. We found that in addition to the full-length CTTNBP2, the fragment containing the middle (Mid) domain alone was sufficient for microtubule association (Fig. 1C,D). Although the NCC fragment was also precipitated in the microtubule pellets, the presence of the NCC in the pellet was likely non-specific, because the solubility of the NCC did not seem comparable to that of other fragments. It was also present in a noticeable amount in the pellets even in the absence of microtubules (Fig. 1D). We thus suggest that the Mid domain is the major microtubule-interacting region of CTTNBP2.

In addition to using crude COS cell extracts for the microtubule-binding assay, we also expressed and purified glutathione S-transferase (GST)-fusion proteins containing the NCC, Mid or P-rich domains of CTTNBP2 from bacteria for the microtubule-binding experiment. In this experiment, similar to the NCC and P-rich domain fusion proteins and GST alone, the GST–Mid fusion proteins were not pulled down substantially by microtubule filaments in the spin-down assay (Fig. 1E). This suggests that the association between the Mid domain and microtubules might not be direct and that an unknown factor in COS cells regulates the association of the Mid domain of CTTNBP2 and microtubule filaments.

Because we consistently found two distinct distributions of CTTNBP2 in COS cells, it seemed likely that an intrinsic factor in the culture, such as the status of cell growth, controls the distribution of CTTNBP2. To explore this possibility, we used a re-plating experiment to investigate the role of cell–substratum interaction in the regulation of the association of CTTNBP2 with the cytoskeleton. At 24 h after transfection with HA-tagged CTTNBP2, COS cells were harvested and replated on new coverslips for 1, 2 and 4 h. We found that at 1 h after replating, in ∼90% of transfected cells, CTTNBP2 tended to form patches and distribute into the cell cortex, where CTTNBP2 colocalized very well with the F-actin cytoskeleton (Fig. 2A). At this stage, microtubule filaments were clearly presented in COS cells. However, they did not show noticeable colocalization with CTTNBP2 (Fig. 2A). The results of quantification showed that, at this stage, ∼37% of CTTNBP2-positive pixels colocalized with the signals of the F-actin cytoskeleton, whereas only 8% of CTTNBP2-positive pixels overlapped with microtubule filaments (Fig. 2B). After a further hour, the association of CTTNBP2 and microtubules became stronger. We found two populations of CTTNBP2-expressing cells at this stage. The first type of cells (∼40% of transfected cells) still had a prominent cell cortex distribution of CTTNBP2. However, in this type of cell, a degree of colocalization of CTTNBP2 and microtubules was already noticeable (Fig. 2A). In the second type of cell (∼60% of transfected cells), CTTNBP2 displayed a filamentous structure and showed fairly good colocalization with microtubules (Fig. 2A). Consequently, the colocalization coefficient of CTTNBP2 and microtubules was increased to ∼40%, whereas the colocalization of CTTNBP2 and F-actin was reduced to ∼20% at 2 h after replating (Fig. 2B). At 4 h after replating, the proportion of filamentous CTTNBP2 was even more apparent (Fig. 2A). The colocalization coefficient of CTTNBP2 and F-actin had dropped to ∼10% (Fig. 2B). Taken together, the results of the replating experiment suggest that the cell–substratum interaction regulates the association of CTTNBP2 with either the microtubule or actin cytoskeleton. When the cell–substratum interaction is more stable, the association of CTTNBP2 with microtubules is stronger.

Because CTTNBP2 is specifically expressed in neurons, we then examined the distribution of endogenous CTTNBP2 in neurons. Our previous studies had shown that, in mature neurons, CTTNBP2 resides stably in dendritic spines, which are mainly supported by F-actin (Chen et al., 2012; Chen and Hsueh, 2012). We speculated that CTTNBP2 associates with microtubules along dendrites before dendritic spines form. To explore this possibility, we first examined the protein expression of CTTNBP2 before dendritic spine formation in our neuronal cultures. The results of immunoblotting indicated that in addition to at 18 days in vitro (DIV), when dendritic spines typically form under our culture conditions, CTTNBP2 proteins were also expressed in immature neuronal cultures at 1, 7 and 14 DIV (Fig. 3A). Immunostainings were then analyzed by both confocal and 3D-SIM imaging to examine the distribution of CTTNBP2 in cultured hippocampal neurons. At 14 DIV, when mature dendritic spines had not yet formed, CTTNBP2 proteins were present in patches and attached to microtubule bundles along dendrites (Fig. 2B). We also found F-actin puncta in the dendritic shafts. However, the F-actin and CTTNBP2 signals only partially overlapped each other (Fig. 2B). When the distributions of CTTNBP2 at 14 and 18 DIV were directly compared (Fig. 3C), it was clear that, at 18 DIV, both CTTNBP2 and F-actin were usually not colocalized with microtubule bundles in the dendritic shafts and had a high degree of colocalization with each other (Fig. 3C, right panel confocal images). However, at 14 DIV, although both CTTNBP2 and F-actin formed patches and were attached to microtubules, there was only a partial colocalization of CTTNBP2 and F-actin (Fig. 3C, left panel confocal images). The 3D-SIM further illustrated that CTTNBP2 had a close interaction with microtubules at 14 DIV but not at 18 DIV (Fig. 3B,C).

We further quantified the distribution of CTTNBP2 in dendritic spines and dendritic shafts. A GFP construct was transfected into cultured hippocampal neurons at 8, 12 and 14 DIV to outline the neuronal morphology. Immunostaining using antibodies against GFP and CTTNBP2 was performed at 14, 18 and 20 DIV. The relative intensity of CTTNBP2 in dendritic shafts gradually reduced during dendritic spine maturation, whereas the dendritic spine intensity of CTTNBP2 increased. For dendritic spines, the change from 14 DIV to 18 DIV was particularly noticeable (Fig. 3D,E). In conclusion, these data suggest that the association of CTTNBP with the F-actin and microtubule cytoskeletons is developmentally regulated in neurons.

Next, we investigated the function of the association of CTTNBP2 with microtubules. We noticed that in COS cells, CTTNBP2 expression changed the organization and structure of microtubules. In untransfected cells, microtubule filaments form a radial pattern extending from the microtubule organization center to the cell edge (Fig. 1A; Fig. 4A). This pattern was disrupted by CTTNBP2 expression. The microtubules tended to circle round and form thicker bundles in CTTNBP2-expressing cells (Fig. 1A; Fig. 4A). We quantified the percentage of cells containing microtubule bundles. In the presence of CTTNBP2, more than half of the transfected cells contained microtubule bundles (mean±s.d., 55.7%±8.1, total 123 cells collected from three experiments), a result that was rarely found in GFP-expressing cells (mean±s.d., 7.4%±4.2, total 120 cells collected from three experiments). These results suggest that CTTNBP2 is able to influence microtubule organization. To further investigate this point, we performed transmission electron microscopy (TEM). Indeed, we found that microtubule organization is very different in CTTNBP2-expressing and GFP-transfected cells (Fig. 4B). In CTTNBP2-expressing COS cells, microtubules tended to align and form tight bundles (Fig. 4B, right panel, arrows). This kind of microtubule array was never found in cells transfected with GFP control (Fig. 4B, left panel). The TEM results, therefore, further indicate that CTTNBP2 promotes the formation of microtubule bundles.

We then investigated whether cortactin influences the microtubule-bundling activity of CTTNBP2. In COS cells, cortactin is typically localized to the cell cortex (Fig. 4C, upper panel). Expression of the cortactin-knockdown construct Cort300 (Hering and Sheng, 2003) noticeably reduced the protein levels of endogenous cortactin (Fig. 4C,D). CTTNBP2 still associated with microtubules and promoted microtubule bundling when the protein levels of cortactin were reduced (Fig. 4C,E). The percentages of cells containing microtubule bundles were comparable between cells expressing vector control and Cort300 (Fig. 4E). This result suggests that cortactin is not required for the microtubule-bundling activity of CTTNBP2. To further confirm this point, the CTTNBP2 PA1 mutant, which lacks efficient interaction with cortactin (Chen and Hsueh, 2012) was expressed in COS cells. Similar to the wild-type proteins, the PA1 mutant proteins still effectively promoted microtubule bundling (supplementary material Fig. S1D,E), further indicating that the interaction with cortactin is not crucial for the microtubule-bundling activity of CTTNBP2. Note that cortactin was recruited to the microtubule filaments in the presence of CTTNBP2 (Fig. 4C, middle panel), although cortactin is not required for the microtubule association of CTTNBP2. It, therefore, seems likely that CTTNBP2 might act as linker to bring microtubules and cortactin together.

Given that the Mid domain alone is sufficient for microtubule association (Fig. 1C,D), we also examined whether the Mid domain alone is sufficient for microtubule bundling. When the Mid domain was expressed in COS cells, the radial microtubule pattern was disrupted; however, the pattern was not similar to that in cells expressing full-length CTTNBP2 (Fig. 4A). In the Mid-domain-expressing cells, the global microtubule signals were reduced. In addition, there were almost no thicker microtubule bundles in the Mid-domain-expressing cells (Fig. 4A, mean±s.d., 1.5±1.3%, total 123 cells from three experiments). These data suggest that the binding to microtubules by the Mid domain is not sufficient for microtubule bundling and that another region of CTTNBP2 is required for microtubule bundling. Previously, we have shown that CTTNBP2 is able to form hetero-oligomers with striatins through the NCC domain (Chen et al., 2012). It therefore seemed possible that to promote microtubule bundling, CTTNBP2 forms a homo-oligomer, or at least a homodimer, through a region outside of the Mid domain. Thus, a CTTNBP2 dimer or oligomer would be able to bind two or more microtubule filaments and promote bundling. The large punctate pattern of CTTNBP2 along dendritic microtubules (Fig. 3B,C) also suggests that there is oligomerization of CTTNBP2. To examine the possibility of the oligomerization of CTTNBP2, we first performed a co-immunoprecipitation experiment to investigate the homophilic interaction of CTTNBP2. Myc- and HA-tagged full-length CTTNBP2 (the schematic structure of CTTNBP2 constructs is shown in Fig. 5A) were co-transfected into COS cells. The cell lysates were then immunoprecipitated with anti-Myc antibody. We found that anti-Myc antibody not only precipitated Myc-tagged CTTNBP2 from the lysates but also brought down HA-tagged CTTNBP2 (Fig. 5B). The co-immunoprecipitation was specific because the anti-Myc antibody did not precipitate HA-tagged CTTNBP2 in the absence of Myc-tagged CTTNBP2 (Fig. 5B). These results support the hypothesis that there is oligomerization of CTTNBP2.

We then identified the oligomerization domain of CTTNBP2 by investigating the interaction between HA-tagged full-length CTTNBP2 and a series of Myc-tagged domain deletion constructs of CTTNBP2. The construct NCCM, which is missing the C-terminal proline-rich domain, was still able to interact with full-length CTTNBP2 (Fig. 5B), suggesting that the proline-rich domain is not involved in the oligomerization of CTTNBP2. The Mid domain alone did not bring down full-length CTTNBP2 (Fig. 5B), also suggesting that the Mid domain is not the oligomerization domain. The NCC construct containing the extreme N-terminal region and the following coiled-coil domain co-immunoprecipitated with full-length CTTNBP2 (Fig. 5B). These data indicate that the NCC region is the oligomerization domain.

To further characterize these domain deletion constructs, we next examined the subcellular distribution of these constructs in COS cells. The NCC construct tended to form aggregates and did not show any obvious colocalization with microtubules in COS cells (supplementary material Fig. S2). This is consistent with the results of biochemical study that showed that the NCC fragment had a lower solubility (Fig. 1D). The radial pattern of microtubules was preserved in NCC-expressing cells (supplementary material Fig. S2). The NCCM construct also formed aggregates in COS cells; however, a fraction of NCCM colocalized with microtubules (supplementary material Fig. S2). We noticed that, similar to the full-length CTTNBP2, the radial pattern of microtubules was also missing in NCCM-expressing COS cells (supplementary material Fig. S2). The proline-rich domain alone was evenly distributed in cells and did not noticeably influence the microtubule organization in COS cells (supplementary material Fig. S2). The results of immunostaining also support the hypothesis that the region containing the NCC is sufficient to induce self aggregation and that the Mid domain mediates the association between oligomers and microtubule filaments.

Based on the above results, it is reasonable to hypothesize that oligomerization through the NCC domain allows CTTNBP2 oligomers to have multiple binding sites for microtubule filaments and thus promotes microtubule bundling. If this hypothesis is correct, disruption of CTTNBP2 oligomerization would be expected to disrupt the function of CTTNBP2 in promoting microtubule bundling. To investigate this possibility, CTTNBP2 was co-transfected with different CTTNBP2 domains into COS cells. In the presence of the NCC domain, the percentage of cells containing microtubule bundles dropped to ∼10% (Fig. 5D). When the NCCM domain was co-transfected with CTTNBP2, NCCM colocalized with full-length CTTNBP2 and was distributed along the microtubules. It even further promoted microtubule bundling in COS cells (Fig. 5C,D). The proline-rich domain alone had no effect on CTTNBP2-dependent microtubule reorganization (Fig. 5C,D). For the Mid domain, although it slightly reduced the microtubule bundling induced by full-length CTTNBP2, the difference was not significant. It also did not impair the filament formation of CTTNBP2 (Fig. 5C,D). The data suggest that the Mid domain cannot efficiently inhibit the association of CTTNBP2 with microtubules in COS cells. Perhaps, this is because the full-length CTTNBP2 forms at least a dimer through the NCC region (Fig. 5B). In COS cells, both CTTNBP2 and the Mid fragment were overexpressed; hence, the monovalent Mid domain might not efficiently compete with bivalent full-length CTTNBP2.

Taken together, these co-transfection experiments also suggest that the NCC domain is involved in oligomerization of CTTNBP2, and that the oligomerization of CTTNBP2 through the NCC domain is also required for microtubule bundling.

Given that microtubule bundling is highly associated with increased microtubule stability (Fong et al., 2013; Kanai et al., 1992; Kanai et al., 1989) and that CTTNBP2 expression induced formation of thick microtubule bundles, it is likely that microtubule stability is increased in the presence of CTTNBP2. To address this question, two experiments were performed. First, we performed immunostaining to monitor the acetylation and tyrosination of microtubules. Tubulin acetylation indicates stable microtubules; in contrast, tyrosinated tubulin suggests unstable microtubules (Fukushima et al., 2009; Garnham and Roll-Mecak, 2012; Takemura et al., 1992). Compared with neighboring untransfected cells and GFP-transfected cells, the tubulin acetylation levels in CTTNBP2-expressing cells were noticeably increased, whereas the tubulin tyrosination levels were decreased (Fig. 6A,B). In contrast, expression of the Mid domain reduced the tubulin acetylation levels, although it did not noticeably influence the levels of tubulin tyrosination compared with the GFP control (Fig. 6A,B). These results indicate that the presence of CTTNBP2 influences the post-translational modification of microtubules.

Second, we examined the sensitivity of microtubules to nocodazole. Nocodazole was added to transfected COS cells at final concentrations of 0, 1, 4 and 10 µM for 30 min. In cells transfected with GFP control, 10 µM nocodazole disrupted microtubule filaments in almost all the cells (Fig. 6C,D). When the concentration of nocodazole was reduced to 4 or 1 µM, ∼10 or 60% of GFP-transfected cells, respectively, still processed microtubule filaments (Fig. 6D). In CTTNBP2-transfected cells, nocodazole had no obvious disruptive effect on microtubules at the concentration of 1 µM. Even at concentrations of 4 and 10 µM, ∼60% of CTTNBP2-expressing cells still contained microtubule filaments (Fig. 6C,D), suggesting that the presence of CTTNBP2 resulted in resistance to nocodazole treatment. In contrast, the Mid-domain-expressing cells were highly sensitive to nocodazole (Fig. 6C,D); 1 µM of nocodazole was sufficient to disassemble microtubules in more than 95% of the Mid-domain-expressing cells (Fig. 6D). These data also suggest that CTTNBP2 expression increases microtubule stability, whereas the Mid domain alone reduces microtubule stability.

The above data suggest that the interaction with cortactin is not involved in the microtubule bundling mediated by CTTNBP2. The NCC domain, mediating oligomerization, and the Mid domain, interacting with microtubules, are required for microtubule bundling, which then increases the microtubule stability. Along these lines, we expected that the NCCM fragment would be sufficient to induce microtubule bundling and stabilization. To examine this possibility, we investigated the function of the NCCM fragment in the regulation of tubulin acetylation. Indeed, the tubulin acetylation level was higher in the NCCM-transfected cells compared with cells transfected with GFP (supplementary material Fig. S3A,B). In addition, expression of the NCCM fragment was also sufficient to cause an increase in nocodazole resistance. More than 40% of the NCCM-expressing COS cells still processed microtubule filaments after treatment with 10 µM nocodazole for 30 min (supplementary material Fig. S3C,D). These data indicate that the NCCM fragment is sufficient to induce microtubule stabilization.

We then investigated whether CTTNBP2 was able to regulate post-translational modification of tubulin in cultured hippocampal neurons. Cultured hippocampal neurons were transfected with a construct expressing artificial microRNA (miRNA) against CTTNBP2 (Chen and Hsueh, 2012) at 8 DIV and subjected to immunostaining with acetylated tubulin antibodies at 14 DIV, a time at which CTTNBP2 associated with microtubules in cultured hippocampal neurons (Fig. 3B,C). The quantitative analysis indicated that the acetylation levels of tubulins in dendrites were reduced in CTTNBP2-knockdown neurons (Fig. 7A,B). Similarly, expression of the Mid or NCC domain also reduced the tubulin acetylation (Fig. 7C,D). These results suggest that tubulin acetylation in neurons was influenced by CTTNBP2.

Because microtubules make up most of the cytoskeleton that supports dendrites, the influence of CTTNBP2 on microtubule stability in dendrites suggests that it likely controls dendritic arborization. To investigate this possibility, we investigated whether knockdown of endogenous CTTNBP2 or overexpression of the Mid and NCC fragments has a negative effect on dendritic growth in cultured hippocampal neurons. Compared with vector control, expression of the CTTNBP2 miRNA knockdown construct (BP2-KD) noticeably reduced the protein levels of endogenous CTTNBP2 (supplementary material Fig. S4A). Moreover, expression of BP2-KD impaired dendritic arborization, as shown by Sholl analysis that showed that CTTNBP2-knockdown neurons obviously had fewer intersections (Fig. 8A–C). This indicates that CTTNBP2-deficient neurons have less-complex dendritic arbors. To rule out the possibility of off-target effects of BP2-KD, a CTTNBP2 silent mutant (WT-res) that is resistant to BP2-KD was co-transfected with the BP2-KD construct in cultured neurons. We found that WT-res effectively rescued the defects of dendritic arborization induced by BP2-KD (Fig. 8A,C), confirming the specificity of BP2-KD. As well as WT-res, the CTTNBP2 PA1 mutant (PA1-res) that is resistant to BP2-KD was also able to rescue the defects (Fig. 8A,C). This is consistent with the observation that the interaction between CTTNBP2 and cortactin is not required for microtubule stabilization promoted by CTTNBP2. We also found that overexpression of CTTNBP2 did not further induce dendritic growth, as dendritic arborization in neurons transfected with CTTNBP2 was comparable to those transfected with vector control (supplementary material Fig. S4B,C).

The negative effect of the Mid and NCC fragments on dendritic arborization was also investigated. As expected, overexpression of the Mid and NCC fragments noticeably impaired dendritic arborization (Fig. 8D,E). These results further indicate that microtubule stabilization controlled by CTTNBP2 is involved in the regulation of dendritic arborization.

CTTNBP2 was originally identified as a cortactin-binding partner (Ohoka and Takai, 1998) that can control the synaptic distribution of cortactin (Chen and Hsueh, 2012). In this report, we further show that CTTNBP2 also associates with microtubules and regulates microtubule stability. CTTNBP2 uses different protein domains to achieve these two functions. The C-terminal proline-rich domain of CTTNBP2 interacts with the SH3 domain of cortactin. To associate with microtubules, the Mid domain alone is sufficient. An analysis of the amino acid sequence did not reveal any known protein motif in the Mid domain of CTTNBP2. The microtubule-binding protein spin-down assay further indicated that an unknown factor is involved in the interaction between CTTNBP2 and microtubules. Although it is unclear how the Mid domain associates with microtubules, our data suggest that to promote microtubule bundling and increase microtubule stability, both the NCC and the Mid domain are required. Because CTTNBP2 forms at least a dimer or oligomer through the NCC domain, the CTTNBP2 dimer or oligomer could then bind multiple microtubule molecules through an unknown mediator. Through this physical link, CTTNBP2 might then induce bundle formation of microtubules and increase the stability of microtubules. It is interesting that expression of the Mid domain actually has a dominant-negative effect on the maintenance of microtubule filaments in COS cells. It is possible that the Mid-domain-associating protein that directly interacts with microtubules can also be recognized by other microtubule regulators in COS cells. The presence of the Mid domain would then compete with the endogenous microtubule regulators in COS cells and destabilize the microtubule filaments.

In this study, our data suggest that the associations of CTTNBP2 with microtubules and F-actin are regulated by the cell–substratum interaction in COS cells and the developmental stage in neurons. When the interaction between COS cells and substratum had not stabilized (1 h after replating), CTTNBP2 had entered the cell cortex and colocalized with F-actin. When the COS cells had established a stable interaction with the substratum (such as at 4 h after replating), CTTNBP2 was preferentially associated with microtubules. In cultured neurons, when mature spines formed, CTTNBP2 was targeted to dendritic spines and colocalized with the F-actin cytoskeleton. Before mature spines formed, CTTNBP2 associated with microtubules. Although it is unclear what signal regulates this partner choice, our results clearly show that the interaction with cortactin did not play a crucial role in the regulation of the microtubule association of CTTNBP2. Thus, these two associations are separated and independent. However, we cannot rule out the possibility that CTTNBP2 can function as a bridge to link microtubules and F-actin under specific circumstances. More investigations are needed to examine which signaling pathway regulates the association of CTTNBP2 with different types of cytoskeleton and whether CTTNBP2 can act as a bridge to link these two types of cytoskeleton together.

We have previously shown that CTTNBP2 controls dendritic spine formation and maintenance (Chen and Hsueh, 2012). The negative effect of CTTNBP2 knockdown on dendritic spine density was rescued by the CTTNBP2 silent mutant that was resistant to the CTTNBP2 miRNA construct but not by the CTTNBP2 PA1 mutant without the ability to interact with cortactin (Chen and Hsueh, 2012). These data suggest that cortactin is crucial for CTTNBP2-dependent dendritic spine formation (Chen and Hsueh, 2012). Here, we report a second function of CTTNBP2 in neuronal morphogenesis, which is to regulate dendritic arborization. This second function is obviously independent of cortactin binding, because the PA1 mutant still effectively rescued the defects caused by CTTNBP2 knockdown. In addition, a previous study has also shown that knockdown of cortactin does not influence dendritic arborization (Jaworski et al., 2005). Our data showed that overexpression of the NCC or Mid domain impaired the dendritic arborization. Given that the NCC domain is required for dimerization or oligomerization of CTTNBP2, and that the Mid domain is involved in microtubule association, CTTNBP2 dimers or oligomers likely interact with an unknown microtubule-binding protein(s) and promote bundle formation of microtubules. The bundling mediated by CTTNBP2 then increases microtubule stability and thus controls dendritic arborization. Indeed, the stability of microtubules, the major type of cytoskeleton that supports dendrite morphology, has been shown to be crucial for dendrite morphology (Chen et al., 2011; Chen et al., 2013; Ori-McKenney et al., 2012; Stewart et al., 2012), although in the previous studies, stabilization of microtubules was not shown to occur through microtubule bundling.

Because CTTNBP2 is specifically expressed in neurons (Chen et al., 2012; Ohoka and Takai, 1998), it might play a unique role in controlling neuronal morphology. Interestingly, recent human genetic studies have identified several missense mutations in the CTTNBP2 gene in patients with autism spectrum disorders (Iossifov et al., 2012; Parikshak et al., 2013). Because autism spectrum disorders are recognized as neurodevelopmental disorders, our study revealing the role of CTTNBP2 in regulating neuronal morphology might also potentially provide a link between CTTNBP2 and psychiatric disorders.

The following antibodies were used in this study: mouse monoclonal anti-Myc (9B11; Cell Signaling Technology, Danvers, MA); rabbit polyclonal anti-Myc (06-549, Millipore, Billerica, MA); chicken polyclonal anti-GFP (ab13970, Abcam, Cambridge, MA); rat monoclonal anti-HA (3F10, Roche, Indianapolis, IN); mouse monoclonal antibodies against β-tubulin, α-tubulin, acetylated tubulin and tyrosinated tubulin (TUB 2.1, B-5-1-2, 6-11B-1 and TUB-1A2, respectively; Sigma-Aldrich, St Louis, MO); rabbit monoclonal anti-Tuj1 (TUJ 1-15-79, Emeryville, CA); and rabbit polyclonal anti-cortactin (H-191, Santa Cruz Biotechnology, TX). The rabbit polyclonal anti-CTTNBP2 antibody was generated as previously described (Chen and Hsueh, 2012).

All animal experiments were performed with the approval of the Academia Sinica Institutional Animal Care and Utilization Committee. For primary culture, pregnant rats were sacrificed by CO₂ inhalation; embryonic day (E)18–E19 fetal pups were then isolated and sacrificed by decapitation.

Myc-tagged CTTNBP2, BP2-miR and Ctrl-miR were constructed as previously described (Chen and Hsueh, 2012). Full-length or partial cDNA sequences corresponding to different regions of CTTNBP2 (NCC, amino acids 1–259; CC, amino acids 118–259; CCM, amino acids 118–522; Mid, amino acids 260–522; P-rich, amino acids 523–625) were amplified and cloned into vectors pGW1-CMV-Myc, pGW1-CMV-HA or pGEX-4T-1 in order to be expressed in COS cells or bacteria. The cortactin-knockdown construct Cort300 was a generous gift from Morgan Sheng (Genetech, CA).

COS cells expressing CTTNBP2 fragments were lysed in 50 mM HEPES pH 8, 320 mM sucrose and 1% Triton X-100, then centrifuged at 16,000 g for 30 min to remove insoluble debris. Prior to the microtubule-binding assay, endogenous microtubules were cleared by incubation with 20 µM taxol at 35°C for 30 min and centrifugation at 100,000 g, and the microtubule-free supernatant was kept for analysis. Various GST fusion proteins were purified from bacteria and also subjected to the spin-down assay. The microtubule-binding assay was performed according to the manufacturer's instructions (BK029, Cytoskeleton, Denver, CO). In brief, purified tubulin was incubated in general tubulin buffer (80 mM PIPES, pH 7, 2 mM MgCl₂, 0.5 mM EGTA and 1 mM GTP) at 35°C for 20 min. Taxol (20 µM) was then added to stabilize the microtubules. The microtubule-free cell lysate was then incubated with or without microtubules in General Tubulin Buffer and 20 µM taxol at room temperature for 30 min. Samples were placed onto a 100-µl cushion buffer (60% glycerol in general tubulin buffer) and centrifuged at 100,000 g for 40 min at room temperature. The supernatant and pellet were collected and analyzed by immunoblotting. Microtubule polymers and associated proteins are present in the pellet; free tubulins and non-associated proteins are present in the supernatant.

Rat hippocampal neurons from E18–E19 embryos were dissociated and cultured as previously described (Chen and Hsueh, 2012). The transfection of neurons was performed using the calcium phosphate precipitation method.

COS-1 cells were originally obtained from the Bioresource Collection and Research Center, Taiwan, and maintained in Dulbecco's modified Eagle's medium (DMEM) with 10% fetal bovine serum. The transfection of COS cells was conducted with Lipofectamine 2000 reagent according to the manufacturer's instructions (Invitrogen, Taipei, Taiwan).

Cells were seeded on ACLAR Embedding Film at ∼90% confluence. After rinsing with wash buffer (0.1 M cacodylate buffer, pH 7.2, 4% sucrose and 0.05% CaCl₂) at room temperature, cells were fixed with fixation solution (0.1 M cacodylate buffer, pH 7.2, 2.5% glutaraldehyde and 1% tannic acid) at 37°C. After extensive rinsing with wash buffer, cells were post-fixed in 1% OsO₄ solution buffered with 0.1 M cacodylate buffer, pH 7.2, at room temperature followed by staining with 1% uranyl acetate at room temperature. After dehydration with graded alcohol, cells were filtrated with a series of solutions as follows: (1) ethanol:Spurr's resin (Electron Microscopy Sciences, Hatfield, PA) at 1∶1 for 30 min; (2) ethanol:Spurr's resin = 1∶2 for 40 min; and (3) pure Spurr's Resin for 1 h. Samples were then polymerized at 70°C for 20 h. After slicing, the images of ultrastruture were captured by Tecnai G2 Spirit TWIN electron microscope (FEI) equipped with a Gatan CCD Camera (794.10.BP2 MultiScan) and acquisition software DigitalMicrograph (Gatan).

Super-resolution microscopy was performed using an Elyra PS.1 microscope (Carl Zeiss) equipped with a 63×/NA 1.4 oil (Plan-Apochromat; Carl Zeiss) objective lens and iXon 885 EMCCD (Andor Technology) at room temperature. The Z series was set at an interval of 0.11 µm. The 3D structure illumination was processed using Zen 2011 software at the following settings: 488-, 555- and 647-nm channels were aligned in the x, y and z axes using 100-nm and 500-nm TetraSpec Fluorescent Microspheres (Invitrogen), a noise filter parameter value of −3, and a SR frequency weighting value of +1. The 3D-reconstruction modeling was further conducted using Imaris image analysis software.

Cells were fixed with 4% paraformaldehyde and 4% sucrose in PBS, followed by permeabilization with 0.2% Triton X-100 in PBS. After blocking with 10% BSA, the cells were incubated with primary antibodies diluted in PBS containing 3% BSA at 4°C overnight. Following PBS washes, the cells were incubated with secondary antibodies conjugated to Alexa Fluor 488, 555 and/or 647 (Invitrogen) for 2 h. The images were acquired at room temperature using a confocal microscope (LSM700; Carl Zeiss, Oberkochen, Germany) equipped with a 10× NA 0.3 or 63× NA 1.4 oil (Plan-Apochromat; Carl Zeiss) objective lens and Zen 2009 (Carl Zeiss) acquisition and analysis software, or an upright microscope (Microscope Axio Imager, M2, Carl Zeiss) equipped with a 63× NA 1.4 oil (Plan-Apochromat; Carl Zeiss) objective lens, EMCCD camera Rolera EM-C² (QImaging, Surrey, BC) and Zen 2011 program (Carl Zeiss) acquisition software. The z-series images were then projected for analysis. Some of the image acquisition and quantification were performed blindly to minimize the effect of bias. For publication, the images were processed using Photoshop without any modification or with minimized adjustment of brightness and/or contrast applied to the entire images.

After incubation overnight, HA-tagged CTTNBP2 transfected COS cells were trypsinized and replated in a new 12-well dish (1×10⁵ cells/well). At 1, 2 and 4 h after re-plating, cells were fixed for immunostaining using antibodies against HA and α-tubulin and Alexa-Fluor-647–phalloidin. Images were then acquired using a microscope.

All the quantification analyses of the fluorescence images were performed using either ImageJ (v.14.7, NIH) or Zen 2009 analysis software (Carl Zeiss). For the analysis of the dendritic shaft and spine distribution of CTTNBP2, the fluorescence intensities of CTTNBP2 in the dendritic shafts and spines were measured to determine the enrichment coefficient for dendritic shafts and spines as follows: the average intensity at dendritic shaft or spine/the average of total intensity containing both dendritic spine and dendritic shaft. To quantify the colocalization of CTTNBP2 with microtubules and F-actin, Zen 2009 analysis software was used with the threshold set at higher than 100 for each channel. The coefficient indicates the percentage of CTTNBP2-positive pixels colocalized with microtubules or F-actin: the coefficient equals the number of pixels with both CTTNBP2 and F-actin or MT signals/total number of pixels with CTTNBP2 signals. To define microtubule bundles, given that microtubule thickness was usually about 0.5 µm in confocal images of untransfected or control vector-transfected cells, cells containing at least two microtubule bundles of >1.5 µm in thickness and >5 µm in length were recognized to have microtubule bundles. Because the centrosome has microtubule aggregation, this measurement was only performed outside of the area containing the centrosome. For CTTNBP2 filaments, cells that contained at least three CTTNBP2 filaments that were longer than 5 µm, were counted. To define cells with a microtubule network, cells had to contain at least ten microtubule filaments and each filament had to be longer than 10 µm. The thickness of microtubules was not taken into consideration.

All the data were processed using Prism version 5.03. For Fig. 4D,E and Fig 5D, all of the columns were compared with the GFP Ctrl, for Fig. 7B, and supplementary material Fig. S1E, S3B and S4A, Student's t-test was used. For Fig. 3E, Fig. 6B and Fig. 7D, the data were analyzed by one-way ANOVA with the Bonferroni test. For Fig. 8C,E and supplementary material Fig. S4C the data were analyzed by two-way ANOVA. Mean±s.e.m. or s.d. and the number of cells and/or experiments (n) are presented in all the figures or figure legends. *P<0.05; **P<0.01; ***P<0.001; ns, not significant.

We thank Morgan Sheng for the cortactin-knockdown construct and Wen-Li Pong at the Imaging Core of the Institute of Molecular Biology, Academia Sinica, Taiwan, for technical support.