Sequential phosphorylation of NDEL1 by the DYRK2-GSK3β complex is critical for neuronal morphogenesis

Neuronal morphogenesis requires multiple regulatory pathways to appropriately determine axonal and dendritic structures, thereby to enable the functional neural connectivity. Yet, however, the precise mechanisms and components that regulate neuronal morphogenesis are still largely unknown. Here, we newly identified the sequential phosphorylation of NDEL1 critical for neuronal morphogenesis through the human kinome screening and phospho-proteomics analysis of NDEL1 from mouse brain lysate. DYRK2 phosphorylates NDEL1 S336 to prime the phosphorylation of NDEL1 S332 by GSK3β. TARA, an interaction partner of NDEL1, scaffolds DYRK2 and GSK3β to form a tripartite complex and enhances NDEL1 S336/S332 phosphorylation. This dual phosphorylation increases the filamentous actin dynamics. Ultimately, the phosphorylation enhances both axonal and dendritic outgrowth and promotes their arborization. Together, our findings suggest the NDEL1 phosphorylation at S336/S332 by the TARA-DYRK2-GSK3β complex as a novel regulatory mechanism underlying neuronal morphogenesis.


Introduction
Establishment of neuronal morphology is a key process during neurodevelopment. The neuronal morphogenesis process involves the extension and the branching of axons and dendrites in order to allow each neuron to determine functional connections with other neurons. Indeed, perturbations of this process can cause severe deficits in brain functions in various neurodevelopmental disorders such as autism spectrum disorder, attention deficit hyperactive disorder, and schizophrenia (Birnbaum and Weinberger, 2017;Forrest et al., 2018;Schubert et al., 2015). Although its complex regulatory pathways composed of various players are still largely unknown, the orchestrated remodeling of the cytoskeleton is a crucial step for neuronal morphogenesis (Coles and Bradke, 2015;Dent and Gertler, 2003;Rodriguez et al., 2003).
NDEL1 directly binds to Trio-associated repeat on actin (TARA, also known as TRIOBP isoform 1) , a short isoform of Trio-binding protein (TRIOBP) generated by alternative splicing (Riazuddin et al., 2006;Seipel et al., 2001). TARA associates with filamentous actin (F-actin) and has functions in cell mitosis and cell migration Seipel et al., 2001;Zhu et al., 2012). Although its abnormal aggregation has also been observed in the postmortem brains of patients with schizophrenia (Bradshaw et al., 2014;, the role of the TARA in neurodevelopment remains largely unknown. Furthermore, the molecular mechanisms underlying functions of NDEL1-TARA complex have yet to be unraveled.
Here, we introduced the large-scale human kinome library screening and the unbiased LC-MS/MS analysis of NDEL1 in order to systematically search regulatory mechanisms for its functions in brain development. We identified the novel sequential phosphorylation at S336 and S332 by DYRK2 and GSK3b and its function in neuronal morphogenesis, particularly in axon/dendrite outgrowth and neuronal arborization, through modulation of F-actin dynamics. We propose a new signaling mechanism that TARA scaffolds DYRK2 and GSK3b and recruits them to NDEL1, thereby inducing sequential phosphorylation of NDEL1 S336/S332 that is crucial for establishing the neuronal morphology. Taking together, our results provide a new biological insight to understand underlying mechanism for neuronal morphogenesis thereby for relevant neurodevelopmental disorders.
We then assessed the PTMs of endogenous NDEL1 proteins isolated from postnatal day 7 (P7) developing mouse brain by LC-MS/MS analysis to see whether the S332 and 336 can be phosphorylated in vivo ( Figure 1C, Figure 1-figure supplement 3). Interestingly, among multiple PTMs    Figure 1 continued on next page identified, we observed masses indicating phosphorylation at NDEL1 S332 or S336 ( Figure 1D), suggesting the potential roles of NDEL1 phosphorylation in neurodevelopmental processes.
To monitor the phosphorylation state of NDEL1 S336 and S332, we raised an NDEL1 S336/S332 phosphorylation-specific antibody (anti-pNDEL1). As detected by anti-pNDEL1 antibody, the phosphorylation was decreased in NDEL1 S332A and virtually absent in NDEL1 S336A (Figure 1-figure supplement 2B). Anti-pNDEL1 antibody also detected bands corresponding to endogenous NDEL1 proteins immunoprecipitated (IPed) from lysates of developing mouse brain (Figure 1-figure supplement 2C). In the brain lysates from various developmental stages, pNDEL1 signal peaked at embryonic day 18 (E18) and P7 ( Figure 1E), the stage where residual neuronal migration and intense neurite outgrowth and maturation occur. Additionally, the increment of the endogenous NDEL1 phosphorylation by DYRK2-GSK3b S9A expression was detected by anti-pNDEL1 antibody ( Figure 1F).
The pNDEL1 signal in mouse brain slices was significantly diminished upon NDEL1 knockdown by in utero electroporation of an shRNA construct, further validating the immunostaining specificity of the antibody (Figure 1-figure supplement 2D). In cultured primary hippocampal neurons, the significant pNDEL1 signal was detected in the growth cone, a region of dynamic crosstalk between actin and microtubules ( Figure 1G). This crosstalk is required for correct neurite extension and branching (Coles and Bradke, 2015;Pacheco and Gallo, 2016;Rodriguez et al., 2003). Moreover, the pNDEL1 signal overlapped with that of both phalloidin, a marker for F-actin, and a-tubulin. Notably, phosphorylated NDEL1 was prominently present in filopodia-like structures where both F-actin and microtubules are colocalized ( Figure 1G, arrowheads).

Phosphorylation of NDEL1 S336/S332 regulates neuronal morphogenesis
The phosphorylated NDEL1 was colocalized with both actin and microtubules at the growth cone, thus we next examined the roles for NDEL1 S336/S332 phosphorylation in the neuronal development. In cultured hippocampal neurons, NDEL1 knockdown significantly reduced both the total neurite length and the longest neurite length by about 31% and 33%, respectively (Figure 2A-C). Coexpression of an shRNA-resistant form of NDEL1 WT , but not NDEL1 S332/336A , effectively reversed these phenotypes providing specificity to the NDEL1 phosphorylation. These NDEL1 phosphorylation-specific effects were found for both axonal and dendritic neurites (Figure 2-figure supplement 1A-B). In addition, co-expression of DYRK2 and GSK3b S9A with NDEL1 WT , but not with NDEL1 S332/336A , significantly increased the total neurite length and the longest neurite length by about 22% and 19%, respectively ( Figure 2D-F, Figure 2-figure supplement 1C-D). Furthermore, to avoid potential complications caused by an overexpression of NDEL1, we applied CRISPR/Cas9 Figure 1 continued brain lysates. Peptide containing S336 and S332 phosphorylation is indicated by bold and underlined letters. (D) MS/MS spectrum for the phosphorylated fragments of NDEL1 peptide including S332 and S336 residues. The sequence of the peptide (aa 315-345) and all detected fragment ions are shown above. The b-and y-ions annotated in the spectrum include the sizes of y 14 and b 22 ions indicative of phosphorylation at S332 or S336. (E) Phosphorylation levels of endogenous NDEL1 S336/S332 in the developing mouse brain. Amount of lysates subjected to IP was normalized by NDEL1 protein levels. N = 7 for E15, P7, P14, P28, and adult brain lysates and N = 6 for E18. All results are presented as means ± SEM. (F) Increased endogenous NDEL1 phosphorylation by DYRK2 and GSK3b. Transfected HEK293 cell lysates were IPed with pan-NDEL1 antibody followed by western blot with anti-pNDEL1 antibody. Over-expression of DYRK2 and GSK3b S9A increased the endogenous NDEL1 S336/S332 phosphorylation. The number of samples is shown at the bottom of the bar of the graph. All results are presented as means ± SEM. *p<0.05 from Student's t-test. (G) Endogenous NDEL1 S336/S332 phosphorylation detected at the growth cone of the primary cultured mouse hippocampal neuron. Anti-pNDEL1 antibody signal was enriched at the growth cone and colocalized with both phalloidin and a-tubulin (indicated by arrowheads). Magnified confocal images show the strong overlap between pNDEL1, phalloidin, and a-tubulin. The scale bar represents 10 mm. See also Figure 1-figure supplements 1, 2 and 3 and Figure 1-source data 1 and 2. The online version of this article includes the following source data and figure supplement(s) for figure 1: Source data 1. Source data for quantitation of endogenous pNDEL1 in developing mouse brain lysates. Source data 2. Source data for quantitation of endogenous pNDEL1 with kinases over-expression.     Taken together, these results show that NDEL1 phosphorylation at S336 and S332 up-regulates axon/dendrite outgrowth.
Suppression of NDEL1 expression in mouse brain decreases the number of dendritic branches (Saito et al., 2017). In Drosophila, the loss of NudE, a homolog of both NDE1 and NDEL1, results in abnormal dendritic arborization (Arthur et al., 2015). Thus, we next examined the roles of NDEL1 phosphorylation in the arborization of dendrites in the developing mouse brain. NDEL1 S336/S332 phosphorylation was prominently detected in neurons of the cortical layers II/III in P14 mouse brain ( Although in utero electroporation is intrinsically cell-type nonspecific and thus we cannot fully exclude the contribution of neuron-nonautonomous effects from neural or glial progenitors, these results collectively indicate that NDEL1 phosphorylation at S336 and S332 plays critical roles for neuronal arborization.
TARA recruits DYRK2 and GSK3b to induce sequential phosphorylation of NDEL1 S336/S332 The interaction between NDEL1 and TARA has been identified previously . Interestingly, TARA co-expression introduced at least two additional forms of NDEL1 bands distinct in size ( Figure 4A). Moreover, deletion of the NDEL1-interacting domain, hTARA D413-499  or mTARA D401-487 , effectively blocked the band shift ( Figure 4-figure supplement 1A-B). To test whether the band mobility shifts are caused by PTMs, we treated protein phosphatase in vitro and it significantly diminished the mobility shifts of NDEL1 bands ( Figure 4A). In addition, TARA enhanced the signals detected by phosphoserine-specific antibody (anti-PhosphoSerine) (Figure 4figure supplement 1C). These results indicate that the NDEL1-TARA interaction promotes NDEL1 phosphorylation.
To characterize the TARA-dependent NDEL1 phosphorylation, we utilized LC-MS/MS analysis of FLAG-NDEL1 proteins upon co-expression of NDEL1 and TARA in HEK293 cells (Figure 4-figure supplement 2). Among the multiple PTMs detected, a phospho-peptide containing S336 residue was present ( Figure 4B). The phosphorylations of both S336 and S332 in a single peptide were not recovered, in part due to technical limitations of LC-MS/MS. When we tested each single alanine Source data 1. Source data for axon/dendrite outgrowth of NDEL1 knockdown and rescue groups. Source data 2. Source data for axon/dendrite outgrowth of NDEL1 and kinases over-expression groups. Source data 3. Source data for axon/dendrite outgrowth assay of NDEL1 S332/336A KI.   To assess how TARA increases NDEL1 phosphorylation by DYRK2-GSK3b kinases, we tested protein-protein interactions among TARA and the kinases. Endogenous TARA was co-IPed with both DYRK2 and GSK3b ( Figure 4D-E). Furthermore, over-expression of TARA enhanced the interaction between DYRK2 and GSK3b, suggesting that TARA acts as a scaffold ( Figure 4F). Immunocytochemistry analysis confirmed their colocalization, further supporting their functional association ( Figure 4G).
In cultured neurons, the co-expression of NDEL1 WT , but not NDEL1 S332/336A , with TARA significantly increased both the total neurite length and the longest neurite length ( Figure 5A Thus, these results indicate that TARA recruits DYRK2 and GSK3b to induce NDEL1 S336/S332 phosphorylation, thereby enhancing neuronal morphogenesis.

Phosphorylation of NDEL1 S336/S332 enhances F-actin dynamics
We then looked for the underlying mechanism by which NDEL1 S336/S332 phosphorylation increases axon/dendrite outgrowth and neuronal arborization. Since we observed anti-pNDEL1 antibody signal overlapping with both F-actin and microtubule at the growth cone ( Figure 1G), we tested if NDEL1 S336/S332 phosphorylation affects cytoskeletal dynamics.
TARA directly binds to and stabilizes the F-actin structure and it recruits NDEL1 toward F-actin Seipel et al., 2001). When we isolated F-actin from soluble G-actin via F-actin fractionation protocol, NDEL1 S336/S332 phosphorylation induced by TARA was detected in both insoluble F-actin fraction and soluble G-actin fraction, verifying that it can associate with F-actin structure ( Figure 6A) comparable to its colocalization at growth cone ( Figure 1G). Furthermore, we employed fluorescence recovery after photobleaching (FRAP) in combination with transfection of RFP-LifeAct to measure F-actin dynamics in the growth cone-like structure of differentiating SH-SY5Y cells (Belin et al., 2014;Riedl et al., 2008). NDEL1 knockdown suppressed the fluorescence recovery without changing the half-maximum recovery time ( Figure 6B-F, Figure 6-figure supplement 1A-D, and Figure 6-video 1). Co-expression of an shRNA-resistant form of NDEL1 WT , but not plots (B), the total length of dendrites (C), the total number of branches (D), and the number of primary/secondary/tertiary dendrites (E) were analyzed by using Simple neurite tracer plug-in of ImageJ software. White arrowheads in (A) indicate neurons with migration defect. (F-J) NDEL1 S336/S332 phosphorylation induced by DYRK2-GSK3b kinases increased dendritic arborization of layer II/III pyramidal neurons. All constructs were electroporated in utero to E15 mouse brain and P14 brain was subjected for analysis. (F) Representative images of the brain slices with the tracked neuron (above) and the overlapped dendritic structures of five independent neurons (bottom). Sholl analysis plots (G), the total length of dendrites (H), the total number of branches (I), and the number of primary/secondary/tertiary dendrites (J) were analyzed by using Simple neurite tracer plug-in of ImageJ software. Each n number is shown at the bottom of the bar of the graph. Scale bars represent 100 mm. All results are presented as means ± SEM. *p<0.05, **p<0.01, and ***p<0.001 from one-way ANOVA for (C), (D), (H), and (I) and two-way ANOVA for (B), (E), (G), and (J). All brain samples for each group were collected from offspring of at least three independent in utero electroporation surgeries. See also Figure 3-figure supplement 1, Figure 3-videos 1 and 2, and Figure 3-source data 1 and 2. The online version of this article includes the following video, source data, and figure supplement(s) for figure 3: Source data 1. Source data for dendritic arborization of NDEL1 knockdown and rescue groups. Source data 2. Source data for dendritic arborization of NDEL1 and kinases over-expression groups.  - Figure 4. TARA recruits DYRK2 and GSK3b to induce sequential phosphorylation of NDEL1 S336/S332. (A) In vitro phosphatase assay. Calf intestinal alkaline phosphatase (CIP) was treated in vitro after IP of FLAG-tagged NDEL1. Additional bands of NDEL1 disappeared upon CIP treatment indicating that these additional bands are caused by the multiple phosphorylation. (B) MS/MS spectrum for the phosphorylated fragments of NDEL1 peptide including S332 and S336 residues. NDEL1 proteins from HEK293 cells over-expressing NDEL1 and TARA were subjected to LC-MS/MS analysis. The sequence of the peptide (aa 306-345) and all detected fragment ions are shown above. The b-and y-ions annotated in the spectrum include the sizes of y 12 and b 33 ions indicative of a single phosphorylation at either S335 or S336. (C) TARA-induced NDEL1 phosphorylation S336 and S332. TARA increased sequential phosphorylation of NDEL1, first at S336 followed by S332. (D) The protein-protein interaction between DYRK2 and TARA. Endogenous DYRK2 was co-precipitated by IP of endogenous TARA from HEK293 cell lysates. Rabbit IgG was used as a negative control. At input lanes, 1% and 10% of lysates were loaded for anti-DYRK2 and anti-TARA blots, respectively. (E) The protein-protein interaction between GSK3b and TARA. Endogenous TARA was co-precipitated by IP of endogenous GSK3b proteins from HEK293 cell lysates. Mouse IgG was used as a negative control. (F) Co-IP among DYRK2, GSK3b, and TARA. Over-expression of TARA increased an amount of GSK3b proteins co-precipitated by IP of DYRK2, implying that TARA scaffolds these kinases to form a DYRK2-GSK3b-TARA tripartite complex. (G) Colocalization of DYRK2, GSK3b, and TARA in mouse hippocampal neurons. GFP-hDYRK2, FLAG-hGSK3b, and MYC-hTARA colocalized at the soma (above) and the growth cone (bottom) regions. See also  Figure 4 continued on next page NDEL1 S332/336A , significantly rescued the impaired F-actin dynamics indicating its phosphorylation dependency. Reduced amount of fluorescence recovery can be interpreted as either by the more stable F-actin structures or by the less newly formed F-actin. Since NDEL1 S336/S332 phosphorylation had a minimal effect on an insoluble fraction of F-actin fraction, it is more likely that suppression of NDEL1 phosphorylation reduced F-actin formation.
Taken together, our results indicate that NDEL1 phosphorylation at S336 and S332 up-regulates F-actin dynamics, at the growth cone of extending neurites, which is likely to underlie neuronal morphogenesis.

Discussion
In this study, we have identified a novel mechanism underlying neuronal morphogenesis that sequential phosphorylation of NDEL1 at S336 and S332 mediated by the TARA-DYRK2-GSK3b complex promotes the modulation of F-actin dynamics to impact this process (Figure 7).
The identification of the TARA-DYRK2-GSK3b signaling module may allow the discovery of novel mechanisms and players in the neuronal morphogenesis. The concerted action of DYRK2 and GSK3 on substrates, such as eIF2Be, tau, CRMP4, DCX, c-Jun, and c-Myc (Cole et al., 2006;Nishi and Lin, 2005;Slepak et al., 2012;Taira et al., 2012;Tanaka et al., 2012;Weiss et al., 2013;Woods et al., 2001), has implicated these kinases in cell cycle, neurite outgrowth, neuronal migration, microtubule regulation, and apoptosis. In these processes, DYRK2 primes the substrate followed by the preferential action of GSK3 at a nearby residue. Likewise, TARA also functions in cell mitosis, migration, and neurite outgrowth Hong et al., 2016;Zhu et al., 2012), suggesting that it may co-operate with these kinases to modulate these cellular functions. Here, we newly identified TARA directly binds to DYRK2 and GSK3b and recruits them to NDEL1 for phosphorylation at S336 and S332, respectively. NDEL1 S336 is the priming site; once mutated, phosphorylation at S332 no longer occurs. Furthermore, enhancing TARA expression augments NDEL1 phosphorylation by DYRK2 and GSK3b. Taken together, our results indicate that TARA acts as a molecular scaffold to functionally link these two kinases to NDEL1. Of particular note, TARA and CRMP1, another substrate of GSK3, have been identified in insoluble aggregates present in brain samples of schizophrenia patients (Bader et al., 2012), hinting at the potential involvement of TARA-DYRK2-GSK3b signaling module in the associated disease pathogenesis.
NDEL1 S336/S332 phosphorylation may highlight the functional difference between NDEL1 and its paralog NDE1. The C-terminal of NDEL1 (aa 191-345) contains multiple phosphorylation sites that are targeted by Aurora A and CDK1/CDK5 (Mori et al., 2007;Niethammer et al., 2000). Here, we identified and characterized two novel sites, S336 and S332, that are also located in the C-terminus. Interestingly, these TARA-DYRK2-GSK3b responsible sites, but not other phosphorylation sites, are absent in NDE1, a paralog of NDEL1 (Bradshaw et al., 2013;Mori et al., 2007;Niethammer et al., 2000;Shmueli et al., 2010). Indeed, the function of NDEL1 S336/S332 phosphorylation in axon/dendrite outgrowth is not shared by NDE1 (Figure 2-figure supplement 1L-O). Thus, we expect that other NDEL1-specific functions are also regulated by S336/S332     (Okamoto et al., 2015). This dual phosphorylation inhibits LIS1-DISC1 transport to the neurite tips and thus microtubule elongation, thereby inhibiting radial migration of neurons. Interestingly, phosphorylation by CDK5 alone enhances neuronal migration via increasing NDEL1 binding to katanin p60 . Similarly, phosphorylation at S251 by Aurora A alone increases radial migration (Takitoh et al., 2012). These results indicate that dual phosphorylation of NDEL1 can have distinct effects to those elicited by single phosphorylation, and similar regulation may also exist regarding NDEL1 S336/S332 phosphorylation and TARA-DYRK2-GSK3b signaling.
The actin-microtubule crosstalk is critical for neuronal morphogenesis (Coles and Bradke, 2015;Dong et al., 2015;Pacheco and Gallo, 2016;Rodriguez et al., 2003). TARA directly associates to F-actin (Seipel et al., 2001) while NDEL1 regulates actin, microtubules, and intermediate filaments Nguyen et al., 2004;Niethammer et al., 2000;Shim et al., 2008;Toth et al., 2014). We previously showed that TARA recruits NDEL1 toward the cell periphery and together they up-regulate local F-actin levels . Here, we showed that TARA-DYRK2-GSK3b axis promotes NDEL1 S336/S332 phosphorylation to modulate F-actin dynamics ( Figure 6). We also observed this phosphorylation to be prominent in the neuronal growth cone where they colocalize with both F-actin and microtubules ( Figure 1G). Some proteins involved in the actin-microtubule crosstalk, such as EB3, Drebrin, and CLIP-170, modulate F-actin dynamics while bound to microtubules (Jaworski et al., 2009;Lewkowicz et al., 2008;Mikati et al., 2013). Likewise, we postulate that NDEL1 may be associated with microtubule and simultaneously up-regulate F-actin dynamics, enhancing actin-microtubule crosstalk at growth cones. Also, we cannot exclude the possibility that TARA-mediated phosphorylation modulates the association of NDEL1 with microtubule. Additional investigations in this regard are required for further understanding how NDEL1 modulate the actinmicrotubule crosstalk.
In summary, we have identified the sequential phosphorylation of NDEL1 at S336 and S332 mediated by the TARA-DYRK2-GSK3b complex as a novel regulatory mechanism for neuronal morphogenesis through modulating F-actin dynamics. Harnessing the biology of NDEL1 S336/S332 phosphorylation or the TARA-DYRK2-GSK3b signaling module will provide new insights toward the discovery of novel components and pathways that are pertinent to brain development and neurodevelopmental disorders. The total neurite length (E) and the longest neurite length (F) were measured by using ImageJ software. (G-K) Up-regulated NDEL1 S336/S332 phosphorylation increased dendritic arborization of layer II/ III pyramidal neurons. All constructs were electroporated in utero to E15 mouse brain and P14 brain was subjected for analysis. (G) Representative images of the brain slices with the tracked neuron (above) and the overlapped dendritic structures of five independent neurons (bottom). Sholl analysis plots (H), the total length of dendrites (I), the total number of branches (J), and the number of primary/secondary/tertiary dendrites (K) were analyzed by using Simple neurite tracer plug-in of ImageJ software. Each n number is shown at the bottom of the bar of the graph. Scale bars represent 100 mm. All results are presented as means ± SEM. **p<0.01 and ***p<0.001 from one-way ANOVA for (B), (C), (E), (F), (I), and (J) and two-way ANOVA for (H) and (K). All neurite outgrowth experiments for (A-C) and (D-F) were independently repeated for at least three times. All brain samples for each group of (G-K) were collected from offspring of at least three independent in utero electroporation surgeries. See also Figure 5-figure supplement 1, Figure 5-video 1, and Figure 5-source data 1 and 2. The online version of this article includes the following video, source data, and figure supplement(s) for figure 5: Source data 1. Source data for axon/dendrite outgrowth of NDEL1 and TARA over-expression groups. Source data 2. Source data for dendritic arborization of NDEL1 and TARA over-expression groups.  Representative time-lapse images of FRAP assay to measure F-actin dynamics at differentiating SH-SY5Y cells expressing RFP-LifeAct. All of NDEL1 over-expressing constructs here contain an shRNA-resistant mutation. A yellowdashed circle indicates region-of-interest used for bleaching. Bleaching was given by stimulating with 10% 568 nm laser for 10 s. (C) Time-dependent fluorescence recovery graph. Comparisons of the area under FRAP curves (D), the percentage of mobile F-actin fraction calculated by the amount of eventual fluorescence recovery (E), and the average half-max (t 1/2 ) of RFP-LifeAct fluorescence recovery (F). NDEL1 knockdown cells had decreased fluorescence recovery meaning more immobile fraction of F-actin and could not be rescued by NDEL1 S332/336A implying its phosphorylation dependency. Each n number is shown at the bottom of the bar of the graph. The scale bar at (B) represents 10 mm. All results are presented as means ± SEM. *p<0.05, **p<0.01, and ***p<0.001 from two-way ANOVA for (C) and one-way ANOVA for (D-F). See also Figure 6-figure supplements 1, 2 and 3, Figure 6-video 1, and Figure 6-source data 1. The online version of this article includes the following video, source data, and figure supplement(s) for figure 6:

Antibodies and plasmids
Anti-NDEL1 rabbit polyclonal antibody (Cat# 17262-1-AP, RRID:AB_2235821) was purchased from Proteintech Group (Rosemont, IL, USA). Anti-TARA rabbit polyclonal antibody (Cat# PA5-29092, Figure 6 continued Source data 1. Source data for F-actin FRAP assay of NDEL1 knockdown and rescue groups.     Figure 7. A model for the mechanism by which NDEL1 S336/S332 phosphorylation regulates neuronal morphogenesis. The phosphorylation of NDEL1 at S336 by DYRK2 primes S332 phosphorylation by GSK3b. TARA mediates this process by recruiting DYRK2 and GSK3b to NDEL1 and forming a tripartite complex in association with F-actin. The phosphorylated NDEL1 enhances F-actin dynamics at the interface with microtubule cytoskeleton in growth cones, thereby facilitating axon/dendrite length and neuronal arborization. Primary cultures of hippocampal neurons were established by isolating E18 SD rat embryo or E15 ICR mouse embryo hippocampal tissues in HBSS (Gibco) and dissociating tissues in 0.25% trypsin (Sigma-Aldrich) and 0.1% DNase I (Sigma-Aldrich) for 10 min at 37˚C. Cells were resuspended in Neurobasal medium (Gibco) supplemented with 10 mM HEPES pH 7.4% and 10% (v/v) horse serum for final cell concentration being 4.0 Â 10 5 cells/mL and plated on glass coverslips pre-coated with poly-D-lysine and laminin. After 2 hr of plating, cell medium was replaced to Neurobasal medium containing 2 mM glutamine, 2% (v/v) B27 supplement (Gibco), and 1% (v/v) penicillin/streptomycin. The neurons were transfected 9 hr or 48 hr after plating with Lipofectamine 2000 and medium was replaced to the culture medium 2 hr after transfection.

Human kinome library screening
The Center for Cancer Systems Biology (Dana Farber Cancer Institute)-Broad Human Kinase ORF collection plasmid kit (Johannessen et al., 2010;Yang et al., 2011)[2, 5] was a gift from William Hahn and David Root (Addgene kit # 1000000014). Each kinase ORF in pDONR-223 vector was cloned into pEZYmyc-His or pEZYflag Gateway destination vectors via LR Clonase II Plus enzyme (Thermo Fisher Scientific) reaction for 4 hr at room temperature followed by transformation into DH5a competent cells and selection from ampicillin-containing LB agar plate. Each kinase ORF expressing clone was confirmed by sequencing analysis.
For screening responsible kinases for NDEL1 S336/S332 phosphorylation, FLAG-NDEL1 plasmid and the expressing clone plasmid were transfected into HEK293 cells and incubated for 48 hr. pEGFP-C3 plasmid and MYC-hTARA construct were transfected with FLAG-NDEL1 to be used as a negative control and a positive control, respectively. Cells were lysed into 1X ELB lysis buffer (50 mM Tris pH 8.0, 250 mM NaCl, 5 mM EDTA, 0.1% NP-40) supplemented with 2 mM NaPPi, 10 mM NaF, 2 mM Na 3 VO 4 , 1 mM DTT, and protease inhibitor cocktail (Roche, Mannheim, Germany). The lysates were subjected to immunoblotting with FLAG antibody. The candidate kinases were selected by the increment of NDEL1 phosphorylation evidenced by the band shift.

Immunoblot assay and immunoprecipitation
Transfected HEK293 cells were lysed in 1X ELB lysis buffer supplemented with 2 mM NaPPi, 10 mM NaF, 2 mM Na 3 VO 4 , 1 mM DTT, and protease inhibitor cocktail (Roche). Mouse brain tissues were isolated from anesthetized and perfused mice followed by homogenization and lysis in 1X modified RIPA lysis buffer (50 mM Tris pH 7.5, 150 mM NaCl, 1% NP-40, 5 mM EDTA, 1% Triton X-100, 0.5% sodium deoxycholate) supplemented with 2 mM NaPPi, 10 mM NaF, 2 mM Na 3 VO 4 , 1 mM DTT, protease inhibitor cocktail (Roche) and 10 U/mL Benzonase nuclease (Sigma-Aldrich). For immunoblotting analysis, proteins were denatured by mixing lysates with 5X SDS sample buffer (2% SDS, 60 mM Tris pH 6.8, 24% glycerol, and 0.1% bromophenol blue with 5% b-mercaptoethanol) and incubating at 95˚C for 10 min. Proteins were separated by SDS-PAGE with 9% polyacrylamide gel and transferred to PVDF membrane (Millipore, Billerica, MA, USA). Membranes were blocked with 5% skim milk or 4% bovine serum albumin (BSA) in Tris-buffered saline (20 mM Tris pH 8.0, and 137.5 mM NaCl) with 0.25% Tween20 (TBST) for 30 min and incubated with primary antibodies at 4˚C for more than 6 hr and HRP-conjugated secondary antibodies at room temperature for more than 2 hr. Protein signals were detected by ECL solutions (BioRad, Hercules, CA, USA). For IP, lysates were incubated with 1-5 mg of antibody at 4˚C for more than 6 hr with constant rotation. Protein-A agarose beads (Roche) washed three times with lysis buffer were mixed with IPed lysates and incubated at 4C for 2 hr or overnight with constant rotation. Beads were collected by centrifugation, washed three times, and mixed with SDS sample buffer for immunoblotting analysis.

Immunocytochemistry and immunohistochemistry
For immunocytochemistry, cells were fixed with 4% paraformaldehyde in PBS or 4% paraformaldehyde and 4% sucrose in PBS for 20 min and washed with PBS for three times. Cells were permeabilized with 0.5% Triton X-100 in PBS for 5 min and blocked with 5% goat serum in PBS or 4% BSA in PBS for more than 30 min. For staining proteins, cells were incubated with primary antibodies diluted in the blocking solution for 1 hr at room temperature or overnight at 4˚C, washed with PBS for three times, and treated with secondary antibodies diluted in the blocking solution for 1 hr at room temperature.
For sequential immunostaining, cells were incubated with the first primary antibody diluted in the blocking solution for 2 hr followed by two rounds of incubation with Alexa Fluor 488-conjugated secondary antibody in the blocking solution for 1 hr each at room temperature. Cells were washed with PBS for more than three times, incubated with the second primary antibody diluted in the blocking solution for 2 hr at room temperature, and treated with Alexa Fluor 647-conjugated secondary antibody in the blocking solution for 1 hr at room temperature.
For immunohistochemistry, mouse brain slices on slide glass were washed with PBS and additionally fixed with 4% paraformaldehyde in PBS for 20 min. After three times of PBS washing, 0.5% Triton X-100 in PBS was treated for permeabilization for 10 min and 5% goat serum or 5% BSA blocking solution was treated for 1 hr. Primary antibody diluted in blocking solution was treated for overnight at 4˚C. After three times of PBS washing, fluorescent-conjugated secondary antibody diluted in blocking solution was treated for 2 hr at room temperature. For sequential staining, it was done as same as immunocytochemistry. Tissue slides were washed with PBS and mounted by using UltraCruz Aqueous Mounting Medium with DAPI (Cat# sc-24941, RRID:AB_10189288, Santa Cruz Biotechnology).
Cell images were acquired by using FV3000 confocal laser scanning microscope (Olympus, Tokyo, Japan) and processed by using ImageJ (Fiji) software (RRID:SCR_002285, National Institute of Health, Bethesda, MD, USA) (Schindelin et al., 2012). For quantitation of colocalization between NDEL1, TARA, pNDEL1, and IgG control staining patterns, all images were deconvolved using advanced constrained iterative (CI) algorithm-based deconvolution program of cellSens software (Olympus) and were subjected for Pearson's colocalization coefficient analysis through cellSens.

In utero electroporation
Pregnant ICR mice at E15 were anesthetized with an intraperitoneal injection of ketamine (75 mg/ Kg) (Yuhan Corporation, Seoul, South Korea) and xylazine (11.65 mg/Kg) (Bayer AG, Leverkusen, Germany) in PBS. Coding sequences of target genes in pCIG2 vectors or shRNA sequences in pLL3.7-EGFP and pLL3.7-mRFP vectors were purified by using EndoFree plasmid maxi kit (Qiagen, Germantown, MD, USA). Each DNA solution (2.0 mg/mL) mixed with Fast Green solution (0.001%) was injected into bilateral ventricles of the embryo through pulled microcapillary tube (Drummond Scientific, Broomall, PA, USA). Tweezer-type electrode containing two disc-type electrodes was located with appropriate angle and electric pulses were given as 35 V, 50 ms, five times with 950 ms intervals by using an electroporator (Harvard Apparatus, Holliston, MA, USA). After electroporation, embryos were put back into the mother's abdomen, the incision was sutured, and mice were turned back to their home cage. The mice were sacrificed at E18 or P14 and brains were fixed with 4% paraformaldehyde in PBS for 24 hr, dehydrated with 10% and 30% sucrose in PBS for more than 24 hr, and soaked and frozen in Surgipath FSC22 Clear OCT solution (Leica Biosystems, Richmond, IL, USA). Brain tissue was sectioned by using cryostats (Leica Biosystems) with 100 mm thickness and each section was immediately bound to Superfrost Plus microscope slides (Fisher Scientific). Brain slice images were acquired by using 10x, 20x, and 40x objective lenses of FV3000 confocal laser scanning microscope (Olympus) with Z-stacks of 1 mm intervals.
In vitro axon/dendrite outgrowth assay Primary cultured rat hippocampal neurons were subjected to transfection at either 9 or 48 hr after plating. The neurons were fixed by 4% (w/v) paraformaldehyde in PBS for 20 min after 72 or 48 hr after transfection for the knockdown group or the over-expression groups, respectively. Cell images were acquired by using a 40x objective lens of fluorescent microscopy and analyzed by using ImageJ software.

In vitro phosphatase assay
Transfected HEK293 cells were lysed in 1X ELB lysis buffer and were IPed with FLAG antibody and protein-A agarose beads to enrich FLAG-tagged NDEL1 proteins. After washing and removal of PBS, beads were incubated with 1X in vitro kinase assay buffer (30 mM HEPES pH 7.2, 10 mM MgCl 2 , and 0.2 mM DTT) and 10 units of calf intestinal alkaline phosphatase (CIP, New England Biolabs, Beverley, MA, USA) at 37˚C for 30 min. 5X SDS sample buffer was mixed to stop dephosphorylation activity of CIP and to subject for western blot analysis.

Liquid Chromatography (LC)-Mass Spectrometry (MS)/MS Analysis
To enrich endogenous NDEL1 proteins from P7 mouse brains, brain lysates with total of 20 mg proteins were IPed with 2 mg of anti-NDEL1 antibody. IPed proteins were reduced with 5 mM dithiothreitol for 0.5 hr at 56˚C and alkylated with 20 mM iodoacetamide at room temperature in the dark for 20 min followed by in-bead protein digestion with 1 mg trypsin (Promega) at 37˚C overnight. The beads were removed by a short spin and digested peptides were desalted with a C18 spin column (#89870, Thermo Fisher Scientific). The peptides extracted from the IP were analyzed with a nano liquid chromatography (LC) system (Dionex) coupled to a Q-Exactive Plus Orbitrap (Thermo Fisher Scientific). A binary solvent system composed of 0.1% formic acid in water and 0.1% formic acid in acetonitrile was used for all analysis. Peptide fractions were separated on an Ultimate 3000 RSLCnano System with a PepMap 100 C18 LC column (#164535) serving as a loading column followed by a PepMap RSLC C18 (#ES803) analytical column with a flow rate of 0.3 mL/min for 135 min. Full scan mass spectrometry (MS) with a data-dependent MS/MS acquisition was performed in a range from 350 to 2000 m/z. All raw LC-MS/MS data were processed with Proteome Discoverer 2.2 (Thermo Fisher Scientific). Data were filtered to a 1% false discovery rate and searched for dynamic phosphorylation modifications (79.966 Da) at a fragment mass tolerance setting of 0.02 Da.

Lysosomal trafficking assay
Primary cultured mouse hippocampal neurons were subjected to transfection of GFP-LAMP1 with indicated constructs at DIV 5-7 and imaged at 37˚C with supplying 5% CO 2 gas by using FV3000 confocal laser scanning microscope. Through FV31S-DT software (Olympus), a 50 mm-length region of interest (ROI) at the axon was determined and recorded for total 3 min with 1 s interval. Generation of kymographs and data analysis were performed by using CellSens and ImageJ using the KymoAnalyzer v1.01 plug-in (Neumann et al., 2017) combined with manual analysis.

Neuronal dendritic morphology analysis
For neuronal dendritic morphology analysis, mouse embryos were electroporated in utero at E15 and sacrificed at P14. From sectioned brain slices, the somatosensory cortex was located based on mouse brain atlas with hippocampal structure and cortical layer II-IV was distinguished by DAPI staining pattern. Each layer II/III pyramidal neuron in which cell soma is located in the middle of Z-stacks with clearly observable apical dendritic structure was subjected for analysis. Apical and basal dendritic morphologies of each selected neuron were traced out by using Simple Neurite plug-in of ImageJ or Imaris software (Bitplane, Zurich, Switzerland) and total length, the longest length, number of branches, and number of primary/secondary dendrites were measured. For Sholl analysis, the tracing data were subjected to Sholl Analysis plug-in of ImageJ by 10 mm radius step size.

Time-lapse live imaging with FRAP assay
To test F-actin or microtubule dynamics, differentiated SH-SY5Y cells expressing either RFP-LifeAct or mCherry-a-tubulin were imaged at 37˚C with supplying 5% CO 2 gas by using FV3000 confocal laser scanning microscope. Through FV31S-DT software, either a 10 mm-radius circular region of interest (ROI) around cellular process tips for F-actin dynamics or 6 mm x 3 mm rectangular ROI at the middle of the cellular process for microtubule dynamics was determined. The ROI was photobleached by scanning with 10% power 568 nm laser and 20 ms/pixel scan speed for total 10 s. For FRAP analysis, five frames were acquired as pre-bleach images followed by bleaching and 150 frames were acquired as post-bleach with each 2 s interval. FRAP results were analyzed by automatically with the easyFRAP-web application (Koulouras et al., 2018) combined with additional manual analysis.

Statistical analysis
All graphs were presented as the mean ± SEM. Statistical significance of the data was analyzed by two-tailed Student's t-test for comparisons between two groups and one-way or two-way ANOVA followed by Bonferroni's post-hoc test for comparisons among multiple groups. . Transparent reporting form Data availability All data generated or analysed during this study are included in the manuscript and supporting files. Source data files have been provided for Figures 1, 2, 3, 5, and 6.