Neuronal adaptor FE65 stimulates Rac1-mediated neurite outgrowth by recruiting and activating ELMO1

Neurite outgrowth is a crucial process in developing neurons for neural network formation. Understanding the regulatory mechanisms of neurite outgrowth is essential for developing strategies to stimulate neurite regeneration after nerve injury and in neurodegenerative disorders. FE65 is a brain-enriched adaptor that stimulates Rac1-mediated neurite elongation. However, the precise mechanism by which FE65 promotes the process remains elusive. Here, we show that ELMO1, a subunit of ELMO1-DOCK180 bipartite Rac1 guanine nucleotide exchange factor (GEF), interacts with the FE65 N-terminal region. Overexpression of FE65 and/or ELMO1 enhances, whereas knockdown of FE65 or ELMO1 inhibits, neurite outgrowth and Rac1 activation. The effect of FE65 alone or together with ELMO1 is attenuated by an FE65 double mutation that disrupts FE65-ELMO1 interaction. Notably, FE65 is found to activate ELMO1 by diminishing ELMO1 intramolecular autoinhibitory interaction and to promote the targeting of ELMO1 to the plasma membrane, where Rac1 is activated. We also show that FE65, ELMO1, and DOCK180 form a tripartite complex. Knockdown of DOCK180 reduces the stimulatory effect of FE65-ELMO1 on Rac1 activation and neurite outgrowth. Thus, we identify a novel mechanism by which FE65 stimulates Rac1-mediated neurite outgrowth by recruiting and activating ELMO1.

FE65 is a brain-enriched adaptor protein with several protein interaction domains, including a WW domain and two C-terminal PTB 2 domains (see Fig. 1A). It is generally believed that FE65 acts as a "scaffold" protein to recruit various interactors to form a functional complex. Indeed, a number of FE65-interact-ing proteins (FIPs) are identified for its conserved domains. These findings have provided insights into the biological functions of FE65. For instance, FE65, via its PTB2 domain, is shown to interact with Alzheimer's disease amyloid precursor protein (APP) and to alter APP processing. Moreover, the FE65 PTB1 domain is found to interact with the small GTPase ARF6 (reviewed in Refs. 1 and 2).
Additionally, FE65 is implicated in neurodevelopment. FE65 knockout mice have been shown to have significant impairment in hidden-platform acquisition ability and reversal learning (3). Increased neonatal mortality and abnormal circling behavior are reported for FE65/FE65L1 double knockout mice. In fact, aberrant laminin organization of meningeal fibroblasts and axonal projection and pathfinding impairments in several brain regions have been observed in the mice (4). Moreover, neuromuscular junction abnormalities and long-term potentiation impairments are observed in the mice (5). We have also shown that FE65 stimulates neurite outgrowth, an essential process for neural network formation, by activating ARF6-Rac1 signaling (6). Because FE65 does not possess any guanine nucleotide exchange factor (GEF) catalytic activity, the underlying mechanisms by which FE65 stimulates the signaling pathway remain elusive. Nevertheless, the above findings underscore FE65 plays significant developmental roles in the nervous system.
Despite the fact that a number of interactors for the conserved domains are identified, FE65 N-terminal region (NTR) binding partners have not been properly determined yet. As an adaptor, it is possible that FE65 NTR is also involved in protein recruitment. To this end, we performed a yeast two-hybrid screen using FE65 NTR as "bait" and identified engulfment and cell motility protein 1 (ELMO1) as a novel FE65 interactor.
ELMO1 is a ubiquitously expressed protein containing an N-terminal Ras-binding domain, an ELMO inhibitory domain (EID), an ELMO domain, a pleckstrin homology domain, an ELMO autoregulatory domain (EAD), and a C-terminal PXXP motif (Fig. 1A) (reviewed in Ref. 7). ELMO1 is reported to be involved in various processes, including phagocytosis, cell migration, cancer metastasis, dendritic spine morphogenesis, and lymphocyte migration (reviewed in Refs. 8 and 9 -11). Many of these processes require dynamic remodeling of the actin cytoskeleton. The effects of ELMO1 are attributable, at least in part, to Rac1, a member of the Rho GTPases that regulate actin dynamics. Rac1 is activated by the GEFs, which promote the exchange of GDP to GTP in Rac1. By contrast, GTPase-activating proteins inactivate Rac1 by promoting GTP hydrolysis (reviewed in Ref. 12). In cells, ELMO1 complexes with DOCK180 to form a bipartite GEF in cells awaiting the yet to be identified activation signal(s) to mediate Rac1-dependent cytoskeleton remodeling (7,13,14). Here, we show that FE65 interacts with ELMO1 and in turn diminishes ELMO1 intramolecular autoinhibitory interaction. Moreover, FE65-ELMO1 interaction plays a role in stimulating Rac1-mediated neurite outgrowth.

FE65 interacts directly with ELMO1
From a yeast two-hybrid screen of a mouse brain cDNA library using FE65(1-217) fragment as "bait," a cDNA clone encoding ELMO1(626 -727), which contains part of the pleckstrin homology domain, EAD, and the PXXP motif, was isolated. Their interaction in yeast was confirmed using a ␤-gal colony-lift filter assay (Fig. 1B).
To further analyze the FE65-ELMO1 interaction, bacterially expressed GST and GST-ELMO1(626 -727) fusion proteins were used as baits to pull down FE65 from transfected cells. In this assay, ELMO1(626 -727), but not GST, pulled down FE65 (Fig. 1C). We next tested the FE65-ELMO1 interaction by using immunoprecipitation assays from transfected cells. FE65 was transfected either alone or co-transfected with Myc-tagged ELMO1. ELMO1 was then immunoprecipitated via the Myc tag. FE65 was detected in immunoprecipitates from the FE65 ϩ ELMO1 but not FE65 singly transfected control cells (Fig. 1D). To demonstrate that endogenous FE65-ELMO1 interaction, we immunoprecipitated ELMO1 from rat brains and probed the immunoprecipitate for FE65. As shown in Fig. 1E, both ELMO1 and FE65 were present in the immunoprecipitates.
The above biochemical assays indicate that FE65 interacts with ELMO1. However, they could not exclude the possibility that FE65-ELMO1 interaction is mediated by an intermediate molecule. To examine whether this interaction is direct, we incubated Escherichia coli purified His-FE65(1-217) with E. coli purified GST or GST-ELMO1(626 -727) baits. As shown in Fig. 1F, His-FE65(1-217) could be pulled down by GST-ELMO1(626 -727) but not GST. Taken together, these results demonstrate that FE65 interacts directly with ELMO1.
Next, we determined the regions in FE65 and ELMO1 that mediate their interaction. In addition to FE65(1-217), we further prepared three FE65 NTR yeast "bait" constructs. Yeast cells were co-transformed with ELMO1(626 -727) prey construct and different FE65 NTR baits and then grown on double and quadruple dropout SD agar plates. We found that the transformants could not grow on a quadruple dropout plate when FE65 amino acids 1-60 were deleted (Fig. 1G, top). This suggests that the first 60 residues of FE65 are essential for FE65-ELMO1 interaction. We then determined the residue(s) in FE65 that is critical for mediating FE65-ELMO1 interaction by generating a number of FE65 mutants and tested their abilities to interact with ELMO1 in co-immunoprecipitation assays. We found that FE65 Lys-48 and Arg-51 (FE65m) are crucial for FE65-ELMO1 interaction, as the interaction was markedly reduced by an FE65 K48A/R51A double mutation (Fig. 1H).
On the other hand, ELMO1(626 -680) could not interact with FE65(1-217) in yeast two-hybrid assay (Fig. 1G, bottom). This suggests that amino acids 681-727 of ELMO1, which contain EAD and PXXP, are required for the interaction. ELMO1 Met-692 and Glu-693 are two critical residues in EAD for mediating protein interaction, at least with ELMO1 EID (15). Therefore, we tested whether the two residues are essential for FE65-ELMO1 interaction. As shown in Fig. 1G, an ELMO1 M692A/ E693A (ELMO1m) double mutation abolished the interaction between ELMO1(626 -727) and FE65  in the yeast twohybrid system. We confirmed the importance of the two residues by a co-immunoprecipitation assay in which FE65-ELMO1 interaction was markedly reduced by an ELMO1m double mutation (Fig. 1I). Together, our result suggests that FE65(1-60) and ELMO1 EAD are required for FE65-ELMO1 interaction.

FE65 and ELMO1 promote neurite outgrowth and Rac1 activation
Previously, we have found that FE65 stimulates Rac1-dependent neurite outgrowth (6). Because ELMO1 also enhances Rac1 activation and is widely expressed in different cell types, including neurons ( Fig. 2A), we anticipated that FE65-ELMO1 interaction promotes neurite extension. To test this, 2 days in vitro rat primary cortical neurons were transfected with FE65, ELMO1, and FE65 ϩ ELMO1 together with enhanced GFPexpressing plasmid (pEGFP-C1), which serves as a morphology marker for the transfected neurons. The expressions of FE65 and/or ELMO1 were confirmed by both Western blotting and immunofluorescence analyses (Fig. 2B, top right and bottom). The lengths of the longest neurite from the growth cone tip to the periphery of the cell body of the transfected neurons were measured 24 h post-transfection. This is a widely employed neurite measurement approach (e.g. see Refs. 6,16,and 17). Similar to our previous report (6), overexpression of FE65, but not FE65m, stimulated neurite extension (Fig. 2B, top left). Notably, overexpression of ELMO1 also promoted neurite outgrowth (Fig. 2B). As we have shown that ELMO1 interacts with FE65 ( Fig. 1) and FE65 is expressed in neurons ( Fig. 2A), the effect of overexpression of ELMO1 on neurite outgrowth might be due to the transfected ELMO1 that interacted with endogenous FE65 in the neurons. In fact, a more pronounced stimulatory effect was observed in FE65 ϩ ELMO1 co-transfected neurons, and such an effect was not observed in FE65m ϩ ELMO1 (Fig. 2B). As FE65m could not interact with ELMO1 properly, our findings suggest that FE65-ELMO1 interaction could promote neurite outgrowth.
Rac1 is a key regulator of actin cytoskeleton dynamics that controls a number of developmental events in neurons, including neurite outgrowth (for a review, see Ref. 18). We therefore investigated whether FE65-ELMO1 interaction influences Rac1 activity. As shown in Fig. 3A, overexpression of FE65 or ELMO1 could activate Rac1. However, FE65m could not stimulate Rac1. Of note, Rac1 was strongly activated in FE65 ϩ ELMO1, but not FE65m ϩ ELMO1, co-transfected cells (Fig. 3A). Because FE65m has reduced binding ability FE65-ELMO1 complex stimulates neurite outgrowth toward ELMO1, our findings suggest that FE65-ELMO1 plays a role in promoting Rac1 activation.
To further dissect the roles of FE65 and ELMO1 in neurite outgrowth, an siRNA knockdown approach was employed. Akin to our previous finding (6), knockdown of FE65 reduced the basal neurite extension significantly (Fig. 2C). A similar effect was observed in ELMO1 knockdown neurons (Fig. 2C). These observations further suggest that both FE65 and ELMO1 are essential for neurite outgrowth. Intriguingly, the stimulatory effect of overexpression of ELMO1 on neurite outgrowth and Rac1 were markedly reduced in FE65 knockdown cells (Figs. 2D and 3B). This suggests that FE65 is required, at least in part, for the effects of ELMO1 on neurite outgrowth and Rac1 activation. On the other hand, overexpression of FE65 could not stimulate neurite growth in ELMO1 knockdown neurons (Fig. 2E), and its effect on Rac1 activation was also decreased markedly in ELMO1 knockdown cells (Fig. 3C). Collectively, our findings suggest that FE65 plays a role, at least in part, in stimulating Rac1-mediated neurite outgrowth by interacting with ELMO1.

DOCK180 is required for FE65-and ELMO1-mediated Rac1 activation and neurite outgrowth
As stated before, ELMO1 is a subunit of the ELMO1-DOCK180 bipartite Rac1 GEF complex (13). We therefore determined whether FE65 alters the GEF complex formation. To do this, FLAG-DOCK180 was precipitated by anti-FLAG M2 antibody from DOCK180 ϩ ELMO1-and DOCK180 ϩ ELMO1 ϩ FE65-transfected cell lysates. As shown in Fig. 4A, the amounts of ELMO1 in both immunoprecipitates were similar. Notably, FE65 was also detected in the immunoprecipitate from ELMO1 ϩ DOCK180 ϩ FE65-transfected cell lysate (Fig.  4A). Moreover, we were able to determine a tripartite complex of DOCK180, ELMO1, and FE65 at the endogenous level by immunoprecipitating DOCK180 from rat brain lysate (Fig. 4B). Taken together, both FE65 and DOCK180 can bind simultaneously to ELMO1 to form a tripartite complex.
Because DOCK180 is the catalytic subunit of ELMO1-DOCK180 bipartite Rac1 GEF complex (13), we tested whether DOCK180 was essential for FE65-mediated Rac1 activation. Similar to our previous observation, overexpression of FE65 activated Rac1 (Fig. 4C). However, the effect of FE65 on Rac1 activation was reduced significantly in DOCK180 knockdown cells (Fig. 4C). Similarly, knockdown of DOCK180 decreased the effect of ELMO1 on Rac1 activation (Fig. 4C). The combined effect of FE65 and ELMO1 on Rac1 was also markedly lowered in DOCK180 knockdown cells (Fig. 4C).
We also determined the effect of DOCK180 knockdown on FE65-ELMO1-mediated neurite extension. As shown in Fig.  4D, knockdown of DOCK180 reduced the stimulating effects of FE65, ELMO1, and FE65 ϩ ELMO1 on neurite outgrowth. Together, our results suggest that DOCK180 is required, at least in part, for FE65-ELMO1-mediated Rac1 activation and neurite outgrowth.

FE65 and ELMO1 co-localize on the plasma membrane and neuronal growth cones
To interact in cells, FE65 and ELMO1 must be located within the same subcellular compartment. Because Rac1 is mainly activated on the plasma membrane (19), and both FE65 and ELMO1 stimulate Rac1, we determined whether FE65 and ELMO1 are expressed on the plasma membrane. Plasma membrane was prepared from both FE65 ϩ ELMO1 co-transfected cells and primary cortical neurons by using a Qproteome plasma membrane protein kit (Qiagen). Approximately 30% of plasma membrane elution and 1% total lysate (as size control) were subjected to immunoblotting for FE65, ELMO1, and various subcellular compartment markers. We found that both FE65 and ELMO1 were detected in the plasma membrane elutions from FE65 ϩ ELMO1 co-transfected cells (Fig. 5A) and primary cortical neurons (Fig. 5B).
In rat cortical neurons, both distinct and overlapping staining were detected (Fig. 5D), in particular at the cell bodies. In fact, ICA at the cell bodies indicated a moderate covariance (ICQ ϭ 0.054 Ϯ 0.02; p Ͻ 0.001; n ϭ 40). Because Rac1 is a regulator of actin cytoskeletal dynamics of growth cones at the tip of growing neurites (18,22,23), and FE65 potentiates the effect of ELMO1 on Rac1 activation, we also enquired whether FE65 and ELMO1 co-localize in the growth cones of neurons. Akin to previous reports (6, 24), FE65 was observed at the growth cones. Importantly, ELMO1 immunostain signal overlapped significantly with FE65 in the growth cones (ICQ ϭ 0.26 Ϯ 0.01; p Ͻ 0.001; n ϭ 40) (Fig. 5D).

FE65 releases ELMO1 autoinhibitory conformation and promotes plasma membrane targeting of ELMO1
ELMO1 is shown to adopt a closed autoinhibitory conformation via intramolecular interaction between its N-terminal EID and C-terminal EAD that plays a role in regulating Rac1 signaling (15). It is proposed that the relief of ELMO1 autoinhibition is a prerequisite for targeting ELMO1-DOCK180 complex to the plasma membrane. Moreover, although yet to be identified, ELMO1 C-terminal interactors are suggested to play a role in Rac1 signaling by releasing ELMO1 autoinhibition constraints (7,13). Because our data reveal that FE65 interacts with the C terminus of ELMO1, we tested whether FE65 influences ELMO1 EID-EAD interaction. Mammalian expression constructs of Myc-ELMO1(1-315) and GST-ELMO1(315-727) were transfected to the cells with empty vector (EV), FE65, or FE65m. GST-ELMO1(315-727) was captured from the cell lysates by GSH-Sepharose 4B resin. Similar to the previous report, GST-ELMO1(1-315) bound to ELMO1(315-727) in the pulldown assay. Intriguingly, GST-ELMO1(315-727) could not pull down ELMO1(1-315) when co-transfected with FE65 but not FE65m mutant (Fig. 6A). Because FE65m interacts less with ELMO1 (Figs. 1H and 6A (bottom)), our data may suggest that FE65 interferes with ELMO1 EID-EAD interaction by competing with ELMO1 N-terminal EID for binding to the C-terminal EAD.

FE65-ELMO1 complex stimulates neurite outgrowth
It is proposed that ELMO1 activation is a prerequisite for its plasma membrane targeting and Rac1 activation there (7,8). Thus, we tested whether FE65 alters plasma membrane recruitment of ELMO1. Intriguingly, knockdown of FE65 reduced the amount of both transfected and endogenous ELMO1 in the plasma membrane elutions (Fig. 7, A and B). Similarly, knockdown of FE65 decreased the number of cells with ELMO1 plasma membrane localization significantly (Fig. 7F). Moreover, FE65m decreased the amount of ELMO1 in the plasma membrane elution as compared with the WT counterpart . DOCK180 is required for FE65-and ELMO1-mediated Rac1 activation and neurite outgrowth. A, DOCK180, ELMO1, and FE65 interact in immunoprecipitation assays from transfected cells. Immunoprecipitations were performed from CHO cells transfected with DOCK180 ϩ ELMO1 and DOCK180 ϩ ELMO1 ϩ FE65. DOCK180 was immunoprecipitated using anti-FLAG against the FLAG tag on DOCK180. DOCK180 in the immunoprecipitates was detected by anti-DOCK180 C-19 (1:2500), and ELMO1, FE65, and ␣-tubulin were detected by anti-ELMO1 PA5-28406, anti-FE65 E20, and anti-␣-tubulin DM1A, respectively. IgG and M2 refer to the presence of mouse control IgG and anti-FLAG M2 in the immunoprecipitations, respectively. ELMO1 level in IPs was measured and analyzed as described in the legend to Fig. 1. The relative ELMO1 level in IP was expressed as a densitometric ratio of ELMO1in IP/DOCK180 in IP. Data for graphs were obtained from three independent experiments. ns, p Ͼ 0.05. Results are means Ϯ S.D. (error bars). B, DOCK180, ELMO1, and FE65 form a complex at the endogenous level. DOCK180 was immunoprecipitated from rat brain lysate using goat anti-DOCK180 C-19. The immunoprecipitates were analyzed for the presence of DOCK180, ELMO1, FE65, and ␣-tubulin by using anti-DOCK180 C-19 (1:1000), anti-ELMO1 PA5-28406 (1:500), anti-FE65 E20 (1:1000), and anti-␣-tubulin DM1A (1:5000), respectively. IgG and C-19 refer to the presence of goat control IgG and anti-DOCK180 C-19 in the immunoprecipitations, respectively. C, CHO cells were transfected with the constructs EV control plasmid, FE65, and ELMO1 and either control or DOCK180 siRNAs, as indicated. Rac1 activation assays were performed as described in the legend to Fig. 3. Knockdown of DOCK180 reduces the effects of FE65, ELMO1, and FE65 ϩ ELMO1 on Rac1 activation. In C, the relative Rac1-GTP level was expressed as a densitometric ratio of Rac1-GTP/total Rac1. Data for graphs were obtained from three independent experiments. *, p Ͻ 0.001. Results are means Ϯ S.D. (error bars). D, rat primary cortical neurons were transfected with the constructs EV control plasmid and different combinations of FE65, ELMO1, control siRNA, and DOCK180 siRNA, as indicated, for the neurite outgrowth assay. Knockdown of DOCK180 suppresses the stimulatory effects of FE65, ELMO1, and FE65 ϩ ELMO1 on neurite outgrowth. The expressions of EGFP, FE65-Myc/ Myc-ELMO1, and DOCK180 were determined by Western blot analysis using anti-GFP JL-8 (1:5000), anti-Myc 9B11 (1:2500), and anti-DOCK180 C-19 (1:2500), respectively. ␣-Tubulin in the lysates was determined by ␣-tubulin DM1A antibody (1:5000). Data were obtained from at least 40 cells/transfection, and the experiments were repeated three times. *, p Ͻ 0.001. Error bars, S.E. (Fig. 7C). These observations may indicate that FE65-ELMO1 interaction is required, at least in part, for targeting of ELMO1 to the plasma membrane. On the other hand, knockdown of ELMO1 did not have a significant effect on the plasma membrane localization of FE65 in both biochemical plasma membrane isolation (Fig. 7, D and E) and immunofluorescence analyses (Fig. 7G). Collectively, our findings suggest that FE65 interacts with ELMO1 to release ELMO1 autoinhibition constraints and facilitates the targeting of ELMO1 to the plasma membrane to stimulate Rac1 signaling.

Discussion
In the present study, ELMO1 is identified as a novel interactor for the neuronal adaptor protein FE65. A number of inter-actors are reported for the conserved domains of FE65. To our knowledge, ELMO1 is the first reported FE65 NTR interactor. In addition to FE65, two closely related isoforms, FE65L1 and FE65L2, are identified. Although functional redundancy between FE65 proteins is suggested, there is evidence that FE65 may have some specific biological roles. For instance, only FE65 is capable to mediate FE65-AICD transactivation but not the other two isoforms (reviewed in Ref. 25). Moreover, learning and memory deficits are observed in FE65 single knockout mice, which suggests that FE65L1 and FE65L2 may not be able to fully compensate for some unique FE65 brain functions (3). It is noteworthy that all three FE65 proteins contain a WW and two PTB domains, and their structural divergence occurs pri-   In A-E, pan-cadherin and GAPDH serve as plasma membrane and cytosol markers, respectively. 30% of plasma membrane elution of each treatment was analyzed as described in the legend to Fig. 5. ELMO1 level in the PM elution and total ELMO1 level in the lysate were measured and analyzed as described in the legend to Fig. 5. The relative ELMO1 level in PM elution was expressed as a densitometric ratio of ELMO1 in PM/total ELMO1 in the lysate. Data for graphs were obtained from three independent experiments. *, p Ͻ 0.001; ns, p Ͼ 0.05. Results are means Ϯ S.D. (error bars). COS7 cells were transfected with ELMO1 ϩ control siRNA or ELMO1 ϩ FE65 siRNA (F) or with FE65 ϩ control siRNA or FE65 ϩ ELMO1 siRNA (G). Cells in F and G were then immunostained for ELMO1 and FE65 as described in the legend to

FE65-ELMO1 complex stimulates neurite outgrowth
marily in their NTR. Thus, FE65 NTR may be required for its unique functions. In fact, emerging evidence reveals that FE65 NTR has distinct cellular roles. For instance, an N-terminal truncated FE65 product, named p65FE65, has been identified, and the truncated form has a stronger interaction with several FE65 interactors than the full-length counterpart. Moreover, the truncated FE65 has a more potent effect on suppression of sAPP␣ secretion and reduced transactivation capability (26). The identification of ELMO1 as an FE65 NTR interactor provides evidence that the FE65 NTR could also function in recruiting proteins for modulating FE65 functions.
Previously, we have shown that FE65 stimulates Rac1 activation (6). Because it does not possess any GEF activities, the exact mechanism(s) by which FE65 promotes Rac1 signaling remains elusive. Intriguingly, we show here that FE65 interacts directly with ELMO1, a regulatory subunit of the ELMO1-DOCK180 Rac1 GEF complex, and knockdown of either ELMO1 or DOCK180 decreases the stimulatory effect of FE65 on Rac1 activation. These findings suggest that FE65 activates Rac1, at least in part, via ELMO1-DOCK180 Rac1 GEF.
Here, we also show that Met-692 and Glu-693 within ELMO1 EAD are essential for FE65-ELMO1 interaction in both Y2H and co-immunoprecipitation assays. The two residues are also crucial for mediating ELMO1 EID-EAD intramolecular interaction (15). These suggest that both FE65 and ELMO1 EID bind to a similar region on the ELMO1 EAD. It is noteworthy that ELMO1 adopts an inactive closed conformation by its EID-EAD interaction at basal state, and release of such constraints is required for proper functioning of DOCK180 in Rac1 activation (15). Although it has long been proposed that ELMO1 interactors play a part in relieving ELMO1 autoinhibition (reviewed in Refs. 7 and 13), such a molecule was yet to be identified. Our findings reveal that FE65 is an interactor of ELMO1 EAD to diminish ELMO1 EID-EAD intramolecular inhibitory interaction. Of note, overexpression of either FE65 or ELMO1 also stimulates Rac1 activation and neurite outgrowth. This may be due to the exogenous proteins interacting with the endogenous partners, as both FE65 and ELMO1 are expressed in the cells used in this study ( Fig. 2A). As mentioned above, once the two proteins interact, FE65 releases ELMO1 autoinhibition conformation. The FE65-ELMO1 complex becomes "active" for the downstream activities, such as Rac1 activation. Hence, it is essential to identify the upstream mechanism(s) that regulates FE65-ELMO1 interaction. It is reported that phosphorylation modulates the interactions between FE65 and its interactors (27)(28)(29). Therefore, it is possible that phosphorylation of FE65 and/or ELMO1 regulates their interaction and thereby FE65-ELMO1 complex activation.
In addition to ELMO1, FE65 has been shown to bind to ARF6, a small GTPase implicated in various neuronal processes, including axonal elongation (6,30,31). The effects of ARF6 are attributed, at least in part, by activation of Rac1. However, the exact mechanism by which ARF6 stimulates Rac1 remains unclear. Of note, ARF family members have been shown to form multimeric complexes to activate various biological events. For example, a multimeric complex that contains clathrin, ARF1, AP1, and CYFIP is reported to play a significant regulatory role in clathrin-AP1-coated carrier biogenesis (32). Additionally, several GTPases and their GTPase-activating proteins complex with ARF4 for the ciliary trafficking of polycystin-1 (33). Although ARF6 activation is reported to stimulate Rac1 signaling through the ELMO1-DOCK180 complex (34), the linkage between ARF6 and the ELMO1-DOCK180 complex was unknown. Therefore, it is worth investigating whether FE65 functions as a link to orchestrate ARF6 and ELMO1-DOCK180 for stimulating Rac1 signaling.
It is noteworthy that deregulation of Rac1, a well-known Rho family small GTPase, has been implicated in a variety of diseases, including, but not limited to, cancer, cardiovascular diseases, and neurodegenerative disorders (35). For instance, defective Rac1 signaling is associated with Alzheimer's disease and motor neuron disease (36 -38). These diseases are characterized by synaptic loss and dysfunction. Of note, FE65 has been implicated in synaptic functions (5,6,39,40), and reduced FE65 expression is observed in both Alzheimer's disease and motor neuron disease (41,42). Signaling pathways that regulate embryonic neurite development are conserved in adults for controlling synaptic plasticity, including Rac1 signaling (43). Because we show here that FE65 regulates Rac1 activity by recruiting ELMO1, dysfunction of FE65 may play a role in the pathogenesis of Alzheimer's disease and motor neuron disease.
Increasing evidence suggests roles of FE65 in cancer. For example, FE65 expression is reported to be increased in thyroid carcinomas, which could stimulate cancer cell growth by promoting sAPP generation (44,45). Moreover, FE65 has been shown to enhance the transcriptional activity of estrogen receptor ␣ and to stimulate breast cancer cell growth (46). Additionally, it has been reported that FE65 suppresses the migration and invasion of estrogen receptor ␣-negative breast cancer cells (47). In addition to neuronal functions, our finding opens an avenue for studying the association between FE65, Rac1 signaling, and carcinogenesis, which may lead to novel therapeutic targets for intervention. Altogether, we show that FE65 interacts with ELMO1-DOCK180 to stimulate Rac1-mediated neurite outgrowth by activating ELMO1 (Fig. 8).

Yeast two-hybrid system
A yeast two-hybrid screen of the mouse brain Mate and Plate cDNA library (Clontech) was performed essentially as described previously, using the cDNA that encodes for the human FE65 amino acids 1-217 as bait (6). ␤-gal colony-lift filter assays were performed essentially as described previously (54).

Cell culture and transfection
CHO, HEK293, and E18 primary rat cortical neurons were prepared and cultured as described previously (6,54).
For plasmid transfection, CHO and HEK293 cells were transfected with X-tremeGENE TM 9 (Roche Applied Science), and rat cortical neurons were transfected with Lipofectamine 2000 (Thermo Fisher Scientific) according to the manufacturers' instructions and as described previously (6). Accell siRNAs and regular siRNAs were obtained from GE Dharmacon. siRNAs were transfected to CHO and HEK293 by using Lipofectamine RNAiMAX (Thermo Fisher Scientific).

Protein-binding assays
The GST and GST fusion proteins were expressed in E. coli BL21 and captured by GSH-Sepharose 4B resin according to the manufacturer's instructions (GE Healthcare). The expressed proteins were used as "baits" in GST pulldown assays as described (6,27,54). Co-immunoprecipitation assays were performed essentially as described previously from transfected cell and rat brain lysates (6,27,54).
Direct binding assays were performed as described previously (6,54). In brief, GST and GST-ELMO1(626 -727) were expressed and purified from E. coli. The GST baits were used to pull down purified His-FE65(1-217). The protein complexes were captured by GSH-Sepharose 4B resin and then analyzed by SDS-PAGE and Western blotting.

Rac1 activation assays
Rac1-GTP levels in cells were determined by using a Rac1 activation assay kit (Thermo Scientific) as described previously (6). The relative amount of Rac1-GTP in PAK-PBD pulldowns was expressed as a densitometric ratio of Rac1-GTP/total Rac1.

Densitometric analysis of Western blots
Densitometric analysis of Western blots was performed by C-DiGit blot scanner (LI-COR) and analyzed by LI-COR Image Studio TM software. Data were obtained from at least three independent experiments.

Immunofluorescence studies and neurite length measurements
Immunofluorescence staining and neurite length measurement were performed as described previously (6,54). Three independent experiments, with at least 40 neurons each, were In the growth cone, FE65 recruits ELMO1-DOCK180 bipartite Rac1 GEF complex and activates ELMO1 by disrupting its autoinhibitory closed conformation from between EID and EAD. Once activated, the FE65-ELMO1-DOCK180 complex is targeted to the plasma membrane to promote Rac1 activation and thereby neurite outgrowth.

FE65-ELMO1 complex stimulates neurite outgrowth
performed in a blind manner. The lengths were determined as the distance from the growth cone tip to the periphery of the cell body and quantified using ImageJ (National Institutes of Health) with the NeuronJ plugin (55). For the co-localization study, intensity correction analyses were performed as described using ImageJ with the JAcoP plugin (6,20,21). Three independent experiments with at least 40 cells were analyzed. Statistical analyses were performed using a one-way analysis of variance test with the Bonferroni post hoc test. Differences were considered significant at p Ͻ 0.05.

Plasma membrane isolation
Plasma membrane was isolated from cells by using a Qproteome plasma membrane protein kit (Qiagen). Briefly, cells were harvested in ice-cold PBS and then washed with 1ϫ lysis buffer PM two times. Cell pellets were lysed in 1ϫ lysis buffer PM with protease inhibitors and lysis solution PL at 4°C for 15 min. The lysates were then cleared by centrifugation at 4°C, 450 ϫ g for 5 min. Supernatants were incubated with binding ligand PBL and Strep-Tactin magnetic beads at 4°C for 1 h. The magnetic beads were washed with lysis buffer PM and followed by wash buffer twice. The isolated plasma membrane proteins were eluted by boiling in 1ϫ SDS sample buffer, and 30% of plasma membrane elutions were subjected to Western blot analysis for FE65 and ELMO1. The purity of the plasma membrane preparations was validated by probing the samples with various subcellular compartment marker antibodies. Approximately 1% of total lysates were loaded for size comparison.

Proximity ligation assays
A PLA was performed by using a Duolink In Situ fluorescence kit (Sigma). In brief, HEK293 cells were seeded on coverslips 24 h before transfection. Mammalian expression constructs Myc-ELMO1(1-315) and FLAG-ELMO1(315-727) were transfected into the cells with EV, FE65, or FE65m. The cells were fixed with 4% paraformaldehyde and permeabilized with 0.1% Triton 24 h post-transfection. After blocking with 5% FBS in PBS at 37°C for 1 h, the cells were then incubated with goat anti-ELMO1 N-20 (1:100) and mouse anti-FLAG M2 (1:200) for 1 h at room temperature to probe for Myc-ELMO1(1-315) and FLAG-ELMO1(315-727), respectively. The cells were then washed three times with 1ϫ Wash Buffer A, followed by incubation with Duolink In Situ PLA probe antimouse PLUS and anti-goat MINUS at 37°C for 1 h in a humid incubator. After incubation, the cells were washed three times with 1ϫ Wash Buffer A. Ligation was performed by adding 1ϫ ligation stock and diluted ligase at 37°C for 30 min, followed by two washes with 1ϫ Wash Buffer A. Amplification was carried out in a darkened humid incubator by incubating the cells with 1ϫ amplification stock and diluted polymerase at 37°C for 100 min. Then the cells were washed two times with 1ϫ Wash Buffer B and then one time with 0.01ϫ Wash Buffer B. The coverslips were mounted with Duolink In Situ mounting medium with 4Ј,6-diamidino-2-phenylindole (DAPI). Images were captured by using an Olympus IX71 fluorescence microscope with an UPlanSAPo ϫ60 water immersion objective. Fluorescence images were captured by a Nikon DS-Qi2 camera, and the fluorescence signals were quantified by the Object Count tool in Nikon NIS Elements. Cells were also stained with anti-␤-tubulin as a morphology marker.

Statistical analyses
All experiments were repeated at least three times. Statistical analyses were performed using one-way analysis of variance tests with the Bonferroni post hoc test. Significance between different treatments is indicated as follows: *, p Ͻ 0.001; **, p Ͻ 0.01; ***, p Ͻ 0.05; ns, not significant (p Ͼ 0.05). Error bars show either S.D. or S.E.