Golgi-resident TRIO regulates membrane trafficking during neurite outgrowth

Neurite outgrowth requires coordinated cytoskeletal rearrangements in the growth cone and directional membrane delivery from the neuronal soma. As an essential Rho guanine nucleotide exchange factor (GEF), TRIO is necessary for cytoskeletal dynamics during neurite outgrowth, but its participation in the membrane delivery is unclear. Using co-localization studies, live-cell imaging, and fluorescence recovery after photobleaching analysis, along with neurite outgrowth assay and various biochemical approaches, we here report that in mouse cerebellar granule neurons, TRIO protein pools at the Golgi and regulates membrane trafficking by controlling the directional maintenance of both RAB8 (member RAS oncogene family 8)– and RAB10-positive membrane vesicles. We found that the spectrin repeats in Golgi-resident TRIO confer RAB8 and RAB10 activation by interacting with and activating the RAB GEF RABIN8. Constitutively active RAB8 or RAB10 could partially restore the neurite outgrowth of TRIO-deficient cerebellar granule neurons, suggesting that TRIO-regulated membrane trafficking has an important functional role in neurite outgrowth. Our results also suggest cross-talk between Rho GEF and Rab GEF in controlling both cytoskeletal dynamics and membrane trafficking during neuronal development. They further highlight how protein pools localized to specific organelles regulate crucial cellular activities and functions. In conclusion, our findings indicate that TRIO regulates membrane trafficking during neurite outgrowth in coordination with its GEF-dependent function in controlling cytoskeletal dynamics via Rho GTPases.

After receiving developmental signals, post-mitotic neurons differentiate into mature neurons and establish specific functional structures, including neurites. The area of the plasma membrane of a developing neuron is estimated to increase by 10,000-fold because of the formation of axons and dendrites (1,2). During this period, a large number of intracellular processes are initiated, including protein and lipid synthesis, cytoskeletal dynamics, membrane production, and trafficking. Cytoskeletal rearrangements and membrane trafficking are both required for neurite outgrowth, with the former providing the driving force for growth cone turning and elongation (3,4) and the latter providing membrane lipids and proteins at sites located far from the Golgi network (2,5). Within these highly regulated processes, Rho family GTPases, such as RAC1, CDC42, and RHOA, have been found to regulate cytoskeletal rearrangements (6 -8). RAC1 and CDC42 regulate neurite elongation and branch formation, and RHOA induces neurite retraction (9 -12). On the other hand, Rab family GTPases regulate membrane trafficking processes (2,(13)(14)(15), in which RAB6, 8, 10, 13, and 33 function in trans-Golgi network (TGN)-related 3 vesicles; RAB5, 7, 21, and 22 regulate early and late endosomes; and RAB4, 11, and 35 regulate recycling endosomes (13,16). The GTPase activities of both Rho and Rab are controlled by GTP-/ GDP-bound cycles, which are regulated by guanine nucleotide exchange factors (GEFs) and GTPase-activating proteins, respectively (8,13). The orchestrated interactions of these two families of small GTPases are thus essential for neurite development.
As a member of the Dbl family of GEFs, the triple functional domain (TRIO) is a large GEF with multiple functional domains that regulates Rho family GTPases, including RAC1, CDC42, and RHOA (17), and plays a key role in axon growth and guidance in Caenorhabditis elegans and Drosophila (18 -23). Deletion of TRIO in mice abolishes neurite outgrowth and axon guidance and produces multiple abnormalities, such as deformities in the skeletal muscle, disorganization of neuronal tissues, and defects in learning and memory (24 -28). TRIO also serves a role in excitatory synaptic transmission and long-term potentiation (29). In particular, several TRIO mutants have been identified in individuals with intellectual disability, schizophrenia, and autism (30 -33). Biochemically, the N-terminal GEF domain is responsible for the activation of the Rho GTPases RAC1 and RHOG (further activates CDC42), whereas the C-terminal GEF domain contributes to RHOA activation (34). The third functional domain is the C-terminal protein serine/threonine kinase domain, whose function remains unknown. The full-length protein containing this kinase domain is expressed at relatively low levels in the nervous system (31,35,36). The TRIO N terminus also contains spectrin repeats and a SEC-14 motif, whose functions are as yet undetermined. Because spectrin is associated to the Golgi apparatus, here we have determined the possible involvement of TRIO in the Golgi apparatus and the Golgi-derived membrane vesicles. The result shows that TRIO is located in the Golgi apparatus through the spectrin repeats in a complex with RABIN8, a common GEF for both RAB8 and RAB10. TRIO facilitates RABIN8 phosphorylation via a direct interaction with RABIN8. TRIO deletion leads to an increase in the frequency of the directional switch of RAB8-and RAB10-positive vesicles and hence a decrease in the traffic distance of the vesicles. However, the velocities of the vesicles are not affected. This observation revealed a novel role of TRIO in membrane trafficking, which is instructive to understand the process coupling of membrane trafficking and cytoskeleton rearrangement during neurite outgrowth.

TRIO is able to localize in the Golgi apparatus
Because TRIO contains spectrin repeats, which are tightly associated with the Golgi apparatus (37), we speculate that TRIO may have the ability to localize at the Golgi apparatus, thereby potentially regulating Golgi-derived membrane trafficking. First, we prepared subcellular fractions of mouse cerebellum using OptiPrep density gradient. The result showed that TRIO protein was detected in the fractions that were positive for Golgi marker GM130 and early endosome marker EEA1 and RAB5, whereas the TRIO signal was much weaker in the fractions positive for ER marker BIP (Fig. 1, A and B). We next Figure 1. TRIO localizes to the Golgi apparatus in CGNs. A, P10 mouse cerebellum was homogenized, and the postnuclear supernatants were subjected to 2.5-30% OptiPrep density gradient for subcellular fractionation. Fractions were subjected to Western blotting with TRIO, GM130, BIP, EEA1, and RAB5 antibodies. B, relative distribution of blotted proteins in A, with the strongest intensity of each protein considered as 1. C, CGNs isolated from WT mice were cultured for 24 -48 h and then stained with the antibody against Golgi markers GM130, TGN46, VAMP4, and RCAS1, as well as the TRIO-N antiserum and DAPI. The scale bar in the left panel represents 20 m, and the scale bars in the right magnified panels represent 5 m. Scatter plots and the PCCs of the fluorescence intensities of red and green channels are also shown. D, scatter plots and bar graph showing the PCC of TRIO and Golgi markers fluorescence intensities quantified from 15 to 20 neurons in each group, from three independent experiments. MW, molecular weight.

TRIO regulates membrane trafficking
studied the subcellular localization of TRIO in the mouse neuroblastoma Neuro-2a cells. Ectopic expression of EGFP-fused TRIO9S and TRIO8, the two main isoforms expressed in cerebellum (35,36), displayed a strong EGFP fluorescence in the perinuclear region, whereas EGFP alone did not (Fig. S1). Staining the transfected Neuro-2a cells with Golgi marker GM130 or TGN38 suggested that the enriched EGFP fluorescence in the perinuclear region was highly associated with Golgi apparatus (Fig. S1).
To determine the endogenous TRIO localization in neurons, we prepared a TRIO-specific antiserum targeting the N-terminal spectrin repeats (Fig. S2A). The specificity of this antiserum was determined using Trio NKO (Trio flox/flox ;Nestin-Cre) (27) lysates by Western blotting (Fig. S2B) and cultured Trio WKO (Trio flox/flox ;Wnt1-Cre, see below) CGNs by immunofluorescence (Fig. S2C). In addition, fluorescence intensity derived from this antiserum was highly correlated with EGFP intensity in COS-7 cells overexpressing EGFP-TRIO9S or EGFP-TRIO8 (Fig. S2D). We then immunostained endogenous TRIO in cultured CGNs using this antiserum and found that endogenous TRIO was also enriched in the perinuclear region, with lower but still detectable intensity along the neurite and in the growth cone (Fig. 1C). This enrichment of TRIO fluorescence was cor-related with GM130 ( Fig. 1C), suggesting that the endogenous TRIO also localizes in Golgi apparatus in CGNs. To confirm this observation, co-localization analysis of TRIO with other Golgi markers, including TGN46, VAMP4, and RCAS1, was performed (Fig. 1C). Quantification of the co-localization indicated that TRIO protein was highly correlated with Golgi apparatus, especially trans-Golgi networks (Fig. 1D). To investigate the amino acids required for TRIO's localization in Golgi, we constructed a series of plasmids encoding truncated TRIO proteins fused with EGFP ( Fig. 2A). TRIO(1-230) encodes the SEC-14 domain; TRIO(208 -673), TRIO(446 -909), and TRIO(672-1295) encode the fragments of different spectrin repeats; and TRIO(1296 -1909) encodes the N-terminal GEF1 domain. After transfection with the plasmids, the COS-7 cells expressing EGFP-TRIO(208 -673) displayed EGFP fluorescence signals highly restricted in the GM130-positive region (Fig. 2B), whereas the COS-7 cells transfected with other plasmids, including the control EGFP plasmid, showed diffused EGFP signals in the cytosol or in the nuclei (Fig. 2B). EGFP-TRIO(208 -673) was also enriched in regions positive for other Golgi markers, including TGN46 (Fig. 2C), VAMP4 (Fig. 2D), and RCAS1 (Fig. 2E). Thus, TRIO is able to localize in the Golgi apparatus through the first spectrin fragment.

TRIO regulates membrane trafficking TRIO regulates membrane trafficking during neurite outgrowth
To study the function of TRIO during neurite outgrowth, we prepared TRIO-deficient CGNs by crossing Trio floxed mice with Wnt1-Cre (27,38). Wnt1-Cre was expressed in the cerebellum and in cultured CGNs, as illustrated by the Rosa-mTmG reporter (39) (Fig. S2, A and B). Unlike the Trio NKO mice we reported previously, the resultant Trio WKO (Trio flox/flox ;Wnt1-Cre) mice live to adulthood, although the body weights were apparently smaller than control mice. We then performed a fluorescence recovery after photobleaching (FRAP) analysis to determine the potential functions of TRIO in membrane trafficking. EGFP-tagged human transferrin receptor (EGFP-hTfR) was transfected to neurons to label newly synthesized membranes (40,41). Because the membrane recovery from photobleaching reflects the processes of membrane trafficking and synthesis, there are investigators using this method to assess the membrane dynamics during hippocampal axonal growth (41). After photobleaching, Trio WKO neurites showed a slower recovery rate of the fluorescence intensity than the control neurites (p Ͻ 0.01) (Fig. 3, A and B). Quantitation of the area under curve after bleaching showed that the recovery percentage of TRIO-deficient CGNs was significantly decreased compared with the control neurons (p Ͻ 0.01) (Fig. 3C). Thus, the TRIO-deficient CGNs exhibited abnormal membrane recovery during the neurite growth.
Because TRIO is associated with Golgi apparatus, we then determined the role of TRIO in the directional trafficking of TGN-derived membrane vesicles. It has been reported that the directional trafficking of these membrane vesicles to the neu-ronal growth cone is primarily mediated by RAB8 and RAB10 during axonal growth (42,43), in which tdTomato fluorescent protein-fused RAB10 is applied to label RAB10-positive membrane vesicles and monitor the trafficking behaviors (42). We here used the tdTomato to label RAB8A and RAB10 to investigate whether TRIO affects RAB8-and RAB10-positive membrane trafficking. We observed the labeled membrane vesicles using time-lapse microscopy, generated the kymographs, and analyzed the trafficking behaviors using KymoAnalyzer (44). TRIO knockout in CGNs decreased the average distance traveled by RAB8A-positive vesicles in both the anterograde and retrograde directions, but the switch frequencies were significantly increased (Fig. 4, A-C, and Movies S1 and S2). However, the average velocities of vesicles traveling in both directions, and the percentages of time spent in motion states were not altered (Fig. 4, D and E, and Movies S1 and S2). In addition, RAB10-positive vesicles in TRIO-deficient CGNs showed a similar behavior to RAB8-positive vesicles in terms of the decreased distance traveled and increased switching frequency and comparable average velocities and percentage of time spent in motion states (Fig. 4, F-J, and Movies S3 and S4). This observation suggested an essential role of TRIO in regulating the switching frequency of the membrane vesicles.

RAB8/RAB10 activation is required for TRIO-mediated neurite outgrowth
The abnormal trafficking of RAB8-and RAB10-positive vesicle prompted us to hypothesize that RAB8 and RAB10 activities were altered in TRIO-deficient neurons. Three siRNAs specifically targeting Rab8a or Rab10 were introduced to knockdown these two GTPases (Fig. S4, A and E). The result showed that knockdown of Rab8a or Rab10 significantly inhibited neurite outgrowth (Fig. S4, B-D and F-H) as in previous reports (43,45), suggesting a required role for RAB8 and RAB10 in neurite outgrowth in CGNs. We then measured the levels of GTP-bound RAB8 and RAB10 in the developing cerebella. After pulldown with the purified GST-MICAL-L2-C protein (46), the GTP-RAB8 and GTP-RAB10 proteins from the cerebellum were measured by Western blotting. The results showed that the GTP-bound forms of both RAB8 and RAB10 were decreased in TRIO-deficient cerebellar tissues (Fig. 5, A and B), suggesting a requirement of TRIO for RAB8/RAB10 activation. The observation from Neuro-2a cells overexpressing EGFP-TRIO9S or EGFP-TRIO8 displaying increased levels of GTP-RAB8 (Fig. 5, C and D) also supports this conclusion. The activation of RAB10 could not be determined because of the low expression of RAB10 in Neuro-2a cells (data not shown). In parallel, we also performed GST-FIP3-RBD11 pulldown assay (47,48) and GST-Rabaptin5-R5BD pulldown assay (49) to determine the GTP-bound levels of RAB11 (a recycling endosome-associated Rab) and RAB5 (an early endosome-associated Rab), in TRIO-deficient cerebella or TRIO-overexpressing Neuro-2a cells (Fig. S5). No alteration of either GTP-RAB11 or GTP-RAB5 was observed, implying that TRIO activated RAB8/RAB10 in a selective manner. We then assessed the roles of RAB8 and RAB10 in the neurite outgrowth defect induced by TRIO deletion. After the introduction of RAB8A(Q67L), an active form of RAB8, into TRIO-deficient CGNs, the neurite these dominant-negative and constitutive active Rab mutants did not modulate the activity of Rho family GTPases, because overexpressing these mutants in Neuro-2a cells did not alter RAC1 activity (Fig. S6). Based on these results, we conclude that both RAB8 and RAB10 are required for neurite outgrowth in CGNs, and these two Rab GTPases may be regulated by TRIO.

TRIO recruits RABIN8 to traffic vesicles and hence facilitates RABIN8 activation
Because RABIN8 has been identified as the major GEF for RAB8 and RAB10 in neurite outgrowth (46,50), we hypothesized that TRIO might be necessary for RABIN8 activation. We first measured active RABIN8 (phosphorylated RABIN8) in the protein lysates from the fresh cerebellar tissues from TRIO WKO mice and their control littermates at postnatal day 10. The phosphorylated RABIN8 was immunoprecipitated and blotted with an anti-phospho-mitogen-activated protein kinase/cyclin-de-

TRIO regulates membrane trafficking
pendent kinase substrate antibody (51) (Fig. 6A). The levels of phosphorylated RABIN8 were dramatically decreased in Trio WKO cerebella (Fig. 6, A and B) and CGNs in culture (Fig. 6, C and D). Because TRIO has no kinase domain for RABIN8 phosphorylation, the effect of TRIO on RABIN8 phosphorylation may be mediated through ERK1/2 (51). We thus measured the ERK1/2 protein in the fractions of membrane vesicles and expectedly found that ERK and RABIN8 co-existed in the subcellular fractions that were positive for RAB8 and RAB10 (Fig.  6E).
To understand the interplays of TRIO and RABIN8, we respectively characterized RABIN8 localization in TGN and membrane vesicles. We immunostained COS-7 cells with anti-RABIN8 and anti-TGN38 antibodies. The RABIN8 signal was observed in the cytoplasm and was enriched in the TGN38positive region (Fig. S3A). In cultured CGNs, the RABIN8 sig-nal was also enriched in the TGN38-positive TGN. This localization was consistent with the observation from cultured hippocampal neurons (52), and the localization pattern of RABIN8 in TGN was not affected apparently by TRIO knockout (Fig. S3B). We then examined the localization of RABIN8 in membrane vesicles by immunostaining. RABIN8 was detected in the RAB8-positive vesicles along the neurite from control CGN (Fig. 6F), but TRIO knockout resulted in a significantly decreased portion of RABIN8 overlapping with RAB8 (0.44 Ϯ 0.03 versus 0.28 Ϯ 0.01 of Pearson's correlation coefficient, and 0.82 Ϯ 0.02 versus 0.65 Ϯ 0.03 of Manders' coefficient 2, p Ͻ 0.05) (Fig. 6, F and G), suggesting that TRIO deletion reduced the accessibility of RABIN8 to RAB8-positive vesicles. This result indicated that TRIO was necessary for RABIN8 recruitment to the membrane vesicles where RABIN8 phosphorylation might be possibly catalyzed by ERK1/2.

TRIO regulates membrane trafficking The spectrin repeats of the TRIO N terminus are required for the interaction with RABIN8
To test whether the recruitment of RABIN8 by TRIO is mediated by physical interaction, we performed a series of binding assays. The P21 cerebellar lysates that expressed TRIO9S and TRIO8 at equivalent levels were subjected to immunoprecipitation using anti-TRIO and anti-RABIN8 antibodies to determine their interaction in vivo. TRIO and RABIN8 were detected in the respective immunoprecipitates by anti-RABIN8 and anti-TRIO antibodies but not by control IgG (Fig. 7A). We also transfected Neuro-2a cells with pEGFP-TRIO8, together with or without a plasmid expressing FLAG-RABIN8 to verify this observation. 48 h after transfection, the cells were lysed, and protein lysates were subjected to immuno-precipitation using a FLAG antibody covalently conjugated to agarose beads. Western blots of these immunoprecipitated samples showed that TRIO8 was detected in the immunoprecipitates of the cells expressing FLAG-RABIN8 (Fig. 7B). This observation was also supported by GST pulldown assay for P21 cerebellar lysates by using purified GST and recombinant GST-RABIN8 fusing protein. Both main isoforms of TRIO were detected in the samples containing GST-RABIN8, but not in the samples containing GST alone (Fig. 7D).
To identify the region of TRIO required for RABIN8 interaction, we prepared two truncated TRIO fragments: TRIO(1-1295) containing the spectrin repeats and TRIO (1296 -1909) containing the GEF1 domain. Immunoprecipitation showed that EGFP-TRIO(1-1295) was detected in FLAG-RABIN8 pre-  's t test). *, p Ͻ 0.05; n ϭ 3. E, P10 mouse cerebellum was homogenized, and the postnuclear supernatants were subjected to 2.5-30% OptiPrep density gradient for subcellular fractionation. Fractions were subjected to Western blotting with RABIN8, RAB8, RAB10, and ERK1/2 antibodies. F, CGNs were cultured for 2 DIV and subjected to immunofluorescence using RABIN8 and RAB8 antibodies. The scale bar represents 5 m. Scatter plots and the PCC of the fluorescence intensities of red and green channels were also shown. G, quantification of RABIN8 and RAB8 co-localization in neurites of CGNs as shown in F. Pearson's correlation coefficient and Manders' coefficient M1 and M2 were analyzed. The error bars indicate S.E. (Student's t test). *, p Ͻ 0.05. Each group comprised three mice (n ϭ 3) and 18 neurons/group. A.U., arbitrary units; n.s., not significant; IP, immunoprecipitation; MW, molecular weight.

TRIO regulates membrane trafficking
cipitates, whereas EGFP-TRIO(1296 -1909) did not produce any signal (Fig. 7C). Thus, the N-terminal region rather than the GEF domain of TRIO is required for the interaction with RABIN8. This conclusion was also supported by the result from the application of TRIO GEF1 inhibitor ITX3 (53), which showed no alteration of RABIN8 phosphorylation level in the ITX3-treated CGNs, whereas in which, the GTP-RAC1 levels were decreased (Fig. 7, E and F). To determine whether TRIO directly bind with RABIN8, we next expressed and purified the His 6 -tagged RABIN8 and then incubated with GST or the GST-fused TRIO variants: TRIO(1-230), TRIO(446 -909), and TRIO(672-1295) ( Fig. 2A). We also expressed TRIO(208 -673) but failed in the purification because of the formation of a tight insoluble inclusion body. The result showed that the spectrincontaining variants were able to strongly bind to His 6 -RABIN8, and TRIO(1-230) containing SEC-14 domain bound weakly to His 6 -RABIN8, whereas GST alone did not show any binding (Fig. 7G). These results suggested that TRIO was able to physically bind with RABIN8 through the spectrin repeats. We then transfected COS-7 cells with either plasmid of pEGFP-C3, pEGFP-TRIO(1-230), pEGFP-TRIO(208 -673), pEGFP-TRIO(446 -909), pEGFP-TRIO(672-1295), or pEGFP-TRIO(1296 -1909), together with plasmid encoding FLAG-RABIN8 to investigate whether each TRIO fragment was able to induce RABIN8 phosphorylation. The results suggested that TRIO(672-1295) and TRIO(1296 -1909) decreased RABIN8 phosphorylation (Fig. 7, H  and I). Because only TRIO(208 -673) was observed in Golgi, these data suggested that TRIO(672-1295) or TRIO(1296 -1909) may affect RABIN8 phosphorylation in a dominant-negative manner. Thus, both the correct localization of TRIO to Golgi and correct interaction of RABIN8 and TRIO appear to be required for RABIN8 activation.

Discussion
In this report, we reveal that Golgi-resident TRIO essentially regulates the directional membrane trafficking of the developing CGNs. At the trans-Golgi network and membrane vesicles, the pooled TRIO interacts with and activates RABIN8, which is necessary for RAB8 and RAB10 activation. RAB8 and RAB10 activation are key for the membrane vesicles trafficking from neuronal soma to the growth cone, and constitutively active RAB8 and RAB10 restored the impaired outgrowth of TRIOdeficient neurites. Based on these results, we concluded that TRIO mediated neurite growth by coordinating at least two processes: cytoskeletal rearrangement and membrane trafficking. This coordination was established by different pools of TRIO protein. We propose a working model of the role of TRIO in neurite growth. In this model, extracellular signals activate growth cone-resident TRIO and lead to the activation of Rho GTPases, which regulates cytoskeletal dynamics in the growth cone, thereby regulating neurite elongation. Simultaneously, Golgi-resident TRIO recruits RABIN8 to membrane vesicles for activation through a physical interaction of the spectrin repeats within the TRIO N terminus, and the activated RABIN8 drives membrane transport through the activation of Rab GTPases such as RAB8 and RAB10. The cytoskeletal rearrangements mediated by TRIO-activated Rho GTPases provide the driving force for neurite growth, and the membrane vesicles transport lipid cargoes to enable new membrane formation. This working model is further supported by multiple lines of evidence: (1) Deletion of TRIO leads to an impaired response to various signaling molecules, such as Netrin-1 and Semaphorin-6A (27). This implies a common regulatory scenario shared by these signals, similar to the membrane trafficking process required for neurite outgrowth. (2) A point mutation within the spectrin repeat also causes autism, schizophrenia, and intellectual disability (30 -33). Because several Dbl family proteins, such as Kalirin, contain spectrin repeats, we predict that these proteins might also follow the similar working model to achieve their corresponding functions. Importantly, because many proteins contain multiple domains and some of them are capable of binding to subcellular structures, the subcellular pool formation might be a general regulatory mechanism for the functional protein machines.
Recently, TRIO mutations were identified in patients with autism, schizophrenia, and intellectual disability (30 -33), and the mutations may occur both in TRIO GEFD1 and the spectrin repeat (N1080I) (32). These observations strongly suggest a critical role for TRIO in learning and memory. Based on our result, in addition to the altered cytoskeleton rearrangement mediated by TRIO GEF domains, the altered membrane trafficking mediated by the spectrin repeat may also essentially contribute to the phenotypes of these diseases. This may be instructive in developing therapeutic interventions for this condition. For example, we may use reagents to activate membrane trafficking to restore the impaired neurite growth in subjects carrying a mutation in the spectrin repeats.
In summary, we revealed a function of TRIO in regulating membrane trafficking during neurite outgrowth, in addition to the GEF-dependent function in regulating cytoskeletal dynam-ics by Rho GTPases. This function is accomplished by Golgi and the Golgi-derived vesicle pool of TRIO, whereby interaction with RABIN8 via the non-GEF domain. TRIO recruits and activates RABIN8 and subsequently activates RAB8 and RAB10 so as to regulate membrane trafficking by controlling the membrane vesicles switch frequency. Our findings enabled us to propose a molecular mechanism depicting the coordination of cytoskeletal rearrangements and membrane-trafficking during neurite growth.

Animals
The mice used in this study were Trio floxed mice (27)

Generation of TRIO N terminus-specific antiserum
The CT233 antigen was selected, which was previously used as rat spectrin 5 and 6, to raise a rabbit anti-TRIO polyclonal antibody (35). The codon-optimized DNA sequence encoding mouse TRIO spectrin repeats 5 and 6, from amino acids Val 674 to Arg 900 , was synthesized (Genscript, Nanjing, China) and cloned into the pGEX-5X-1 vector (GE Healthcare). The plasmid was expressed in DH5␣ cells, production of the GST fusion protein was induced with 0.1 mM IPTG, and the recombinant protein was purified using the batch purification method with GSH-Sepharose 4B (catalog no. 17-0756-01, GE Healthcare), according to the manufacturer's instructions. The eluted protein was then dialyzed against PBS overnight and quantified using a Bio-Rad protein assay kit (catalog no. 500-0006, Bio-Rad). GST was not cleaved, and the fusion protein was diluted to 1 mg/ml in PBS. It was mixed well with an equal volume of Freund's complete adjuvant (F5881, Sigma-Aldrich) by vortexing for 30 min at room temperature. 200 l of the mixture (100 g of fusion protein) were used to immunize BALB/c mice via intraperitoneal injections. These mice were immunized a second and third time by repeating the injections of 100 g of fusion protein mixed with an equal volume of Freund's incomplete adjuvant (F5506, Sigma-Aldrich) after 14 and 28 days, respectively. The mice were sacrificed 40 days after the first injection, and sera were collected, followed by the characterization of each antiserum by Western blotting. Of all the antisera collected, the most specific antiserum was selected and used in the present study.

Neuronal nucleofection
CGNs were prepared from P6-8 mice as described above, because a sufficient number of neurons can be obtained from mice at this age for nucleofection. CGNs from each cerebellum were suspended in 200 l of nucleofection buffer, and each 100-l cell suspension (ϳ2 ϫ 10 6 neurons) was used for a nucleofection reaction using an Amaxa Nucleofector 2b device (Lonza) with program O-005. For the nucleofection of plasmids, 5 g of each plasmid were used in a single reaction. The cells were then pipetted out of the cuvette in 1 ml of DMEM containing 10% FBS and aliquoted into three wells of a 24-well plate on poly-D-lysine-coated 8-mm ϫ 8-mm glass coverslips (for rescue experiments) or one PDL-coated 35-mm glass-bottomed dish (for live cell image), and the medium was changed after 4 h, as described above.

Immunofluorescence, neurite outgrowth assay, and co-localization assay
Neurons and COS-7 cells grown on coverslips were washed with PBS, fixed with 4% paraformaldehyde (pH 7.4 in PBS) at room temperature for 5 min, and then washed with PBS three times for 5 min each. Neurons were then permeabilized with 0.5% Triton X-100 in PBS for 15 min, blocked with 1% BSA at room temperature for 1 h, and incubated with primary antibodies overnight at 4°C. The primary antibodies used in the present study were: rabbit anti-TAU (1:200 For the neurite outgrowth assay, images were acquired at 512 ϫ 512 pixels using an Olympus FV1000 confocal microscope (Olympus, Japan) by an experimenter who was blinded to the genotypes and/or experimental conditions, and the neurite lengths (i.e. the distance between the center of the soma and the tip of the longest neurite) were measured manually using National Institutes of Health ImageJ software. Only neurites that met the criteria described else-where were measured (55): (1) the neurite emerged from an isolated neuron, (2) the neurite was not in contact with other neurons or neurites, and (3) the neurite was longer than the diameter of the soma. For other analyses of immunofluorescence staining, the images were captured using a Zeiss LSM880 confocal microscope (Zeiss, Germany). Co-localization was quantified using the JACoP plugin (56) in ImageJ, including the Pearson's correlation coefficient (Pr) and Manders' coefficients (M1 and M2), using Costes' automatic threshold. Fluorescence intensity scatter plot was generated using the ScatterJn (57) plugin in ImageJ.

Live-cell imaging and kymograph analysis
Nucleofected neurons were cultured in phenol red-free Neurobasal medium (catalog no. 12348-17, Life Technologies) supplemented with 2% B-27 (catalog no. 17504-044, Life Technologies), 1ϫ GlutaMAX (catalog no. 35050061, Life Technologies), and 1ϫ penicillin/streptomycin (catalog no. 15140122, Life Technologies) for 24 -30 h and observed under a GE del-taVision Elite microscope (GE healthcare) at 37°C and 5% CO 2 . Time-lapse images were acquired every 100 ms for a total of 30 s using a 60ϫ objective lens (PlanApo NA1.42, oil immersion, Olympus). A pixel resolution of 0.108 m/pixel was generated using this system, and deconvoluted images were generated. Kymographs of RAB8A-and RAB10-positive vesicles were generated using the KymoResliceWide plugin in ImageJ software, with a line width setting of 10 pixels, and the movement parameters, including distance traveled, switch frequencies, velocities, and percentage of time in motion, were semiautomatically analyzed using the KymoAnalyzer macro (44) in ImageJ software.

FRAP analysis
Neurons grown in 35-mm glass-bottomed dishes for 1 DIV were transfected with pEGFP-hTfR with the calcium method using the CalPhos kit (Clontech), according to a previously described procedure (58), and the cells were allowed to express the fusion protein for another 24 -30 h. FRAP experiments were performed using a Zeiss LSM880 confocal microscope with a controlled temperature of 37°C in a humid chamber with a 5% CO 2 atmosphere. Five scans of prebleaching images were initially acquired to obtain a baseline intensity measurement. Bleaching was performed on selected neurites expressing EGFP-hTfR using the 488-nm argon laser line at 100% power with 20 iterations. The intensity during fluorescence recovery was measured by acquiring 150 additional scans at 2-s intervals with a 2% laser power at a resolution of 512 ϫ 512 pixels and an optical zoom of 2. Background correction and data normalization were performed using previously described methods (41). Briefly, a group of bleached neurites, a nonbleached region, and background region were first determined, and the intensity of the background region was subtracted from the intensity of the bleached neurites and nonbleached region at each time point for background correction. The corrected intensity of the bleached neurites was divided by the intensity of the nonbleached region and normalized to the average intensity of the five prebleaching scans.

TRIO regulates membrane trafficking Subcellular fractionation
Subcellular fraction was prepared using OptiPrep Axis-Shield density gradient medium according to the procedure of the application sheet S24 (Alere Technologies AS, Oslo, Norway) with modifications. Briefly, the cerebellum was isolated from P10 mouse and homogenized in 1 ml of homogenization medium (0.25 M sucrose, 1 mM EDTA, and 10 mM HEPES⅐NaOH, pH 7.4) by Dounce tissue grinder and five passes through a fine syringe needle. Postnuclear supernatant was prepared by centrifuging the homogenate at 1500 ϫ g for 10 min. Discontinuous OptiPrep gradient was prepared using underlayering technique by layering 2 ml of the following solutions: 2.5, 5, 10, 15, 20, 25, and 30% iodixanol in homogenization medium. Approximately 700 l of postnuclear supernatant was uploaded to the gradient and centrifuged at 88,000 ϫ g at 4°C for 16 h using SW32 rotor (Beckman, German). Twenty-eight fractions (500 l each) were collected from the top, supplemented with 4ϫ Laemmli buffer, boiled at 95°C for 10 min, and subjected to Western blotting.

GST pulldown assay
The P21 mouse cerebellum was lysed in lysis buffer (50 mM Tris⅐HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, pH 8.0, 1% Triton X-100, 1 mM PMSF, 10 g/ml leupeptin, and 10 g/ml aprotinin). The lysate was cleared by centrifugation at 12,000 rpm for 10 min, and equal amounts of cleared lysate were incubated with 40 l of a 50% slurry of GSH-Sepharose 4B and 20 g of purified GST or GST-RABIN8 overnight at 4°C. The beads were then washed three times with 500 l of lysis buffer prior to the addition of 2ϫ Laemmli buffer. Samples were boiled at 95°C for 10 min and subjected to Western blotting.

Purification of the His 6 -RABIN8 and direct binding assay
His 6 -RABIN8 were expressed in BL21(DE3) cells that had been induced with 1 mM IPTG at 37°C for 4 h. Bacteria pellets were then sonicated in PBS with 1 mg/ml lysozyme. Proteins were then pelleted by centrifugation at 9000 rpm for 10 min and dissolved in PBS with 8 M urea. Cleared supernatant were then incubated with nickel-nitrilotriacetic acid beads (Life Technologies) at room temperature for 1 h. The beads were first washed by gravity flow with 100 mM Tris⅐HCl, pH 8.0, 6 M urea, and 20 mM imidazole and then washed another three times by the same wash buffer in which the urea concentration was 4, 2, and 0 M. The purified His 6 -RABIN8 was eluted with 500 mM imidazole in 100 mM Tris⅐HCl, pH 8.0, and dialyzed against buffer containing 20 mM HEPES⅐NaOH, pH 7.4, and 150 mM NaCl. The concentration was determined using a Bio-Rad protein assay kit (catalog no. 500-0006, Bio-Rad). Direct binding of His 6 -RABIN8 to GST-TRIO variants was assayed by diluting 10 g of purified His 6 -RABIN8, 20 g of purified GST-TRIO variants, and 40 l of a 50% slurry of GSH-Sepharose 4B (catalog no. 17-0756-01, GE Healthcare) in 600 l of binding buffer (20 mM HEPES⅐NaOH, pH 7.4, 150 mM NaCl, 0.1% Triton X-100). The samples were incubated overnight at 4°C. The supernatants were collected, and the beads were washed three times with binding buffer and boiled at 95°C for 10 min after the addition of Laemmli buffer. For Western blotting, 1/10 of pellet samples and 1/100 of supernatant samples were loaded to SDS-PAGE.

TRIO regulates membrane trafficking
Healthcare) at 4°C for 4 h, followed by three washes with 400 l of lysis buffer. The bound GTP-Rabs was eluted by boiling with Laemmli buffer and was detected by Western blotting.

RAC1 activation assay
RAC1 activation assays were performed using a previously described procedure (59), with minor modifications. Neuro-2a cells were lysed in lysis buffer (50 mM Tris⅐HCl, pH 7.2, 1% Triton X-100, 500 mM NaCl, 10 mM MgCl 2 , 1 mM PMSF, 10 g/ml leupeptin, and 10 g/ml aprotinin), and the cleared lysates were incubated with 20 g of GST-PAK-GBD and 40 l of a 50% slurry of GSH-Sepharose 4B (catalog no. 17-0756-01, GE Healthcare) at 4°C for 45 min. The beads were then washed three times with 400 l of lysis buffer and were boiled to elute the bound GTP-RAC1. For experiment in Fig. 7E, the supernatant was collected and directly used for RABIN8 immunoprecipitation to detect phosphorylated RABIN8 level.

Statistical analysis
All quantification analysis of Western blotting was performed using ImageJ software. In all experiments using mice, n represents the number of mice used. The data are presented as the means Ϯ S.E. The statistical analysis was performed using Prism 6.05 software. As indicated in the figure legends, Student's t test or one-way ANOVA with Bonferroni's test was used for tests with only one variable, and a two-way ANOVA with Tukey's test was used for tests with two independent variables. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001; ****, p Ͻ 0.0001.