ABSTRACT
The microtubule motor cytoplasmic dynein contributes to radial migration of newborn pyramidal neurons in the developing neocortex. Here, we show that AMP-activated protein kinase (AMPK) mediates the nucleus-centrosome coupling, a key process for radial neuronal migration that relies on dynein. Depletion of the catalytic subunit of AMPK in migrating neurons impairs this coupling as well as neuronal migration. AMPK shows overlapping subcellular distribution with cytoplasmic dynein and the two proteins interact with each other. Pharmacological inhibition or activation of AMPK modifies the phosphorylation states of dynein intermediate chain (DIC) and dynein functions. Furthermore, AMPK phosphorylates DIC at Ser81. Expression of a phospho-resistant mutant of DIC retards neuronal migration in a similar way to AMPK depletion. Conversely, expression of the phospho-mimetic mutant of DIC alleviates impaired neuronal migration caused by AMPK depletion. Thus, AMPK-regulated dynein function via Ser81 DIC phosphorylation is crucial for radial neuronal migration.
INTRODUCTION
Neuronal migration is a fundamental process for proper brain development. Defects in this process cause severe neurodevelopmental disorders (Feng and Walsh, 2001; Guerrini and Parrini, 2010; Kato and Dobyns, 2003; Reiner et al., 2016). In the developing cerebral cortex, postmitotic neurons migrate long distances along fibers of radial glial cells from their birthplace in the ventricular zone (VZ) toward the pial surface (Hatten, 1999; Kriegstein and Noctor, 2004; Marín et al., 2010). As cortical neurons migrate, they show bipolar morphology with a long, thick leading process oriented toward the pia and a thin tailing process directing to the ventricle. During neuronal migration, cortical neurons undergo a series of highly organized subcellular events, including the extension of the leading process and the forward movement of the centrosome ahead of the nucleus. These changes are followed by translocation of the nucleus toward the centrosome (Bertipaglia et al., 2018; Cooper, 2013; Tsai and Gleeson, 2005). The nuclear translocation following the centrosomal forward movement is key for neuronal migration, and these events rely on proper modulation of the microtubule/actin cytoskeletons and their motor proteins.
In migrating neurons, the microtubule network emanating from the centrosome extends out in the leading process and rearward to surround the nucleus (Bertipaglia et al., 2018; Cooper, 2013; Tsai and Gleeson, 2005). Cytoplasmic dynein, a minus end-directed motor of microtubules, is enriched at the nuclear envelope and in the dilated region (known as cytoplasmic swelling) near the base of the leading process (Tsai et al., 2007; Zhang et al., 2009). It has been proposed that this regional (compartmentalized) localization of dynein is required for pulling the centrosome/nucleus forward along the microtubules. In fact, knockdown of dynein heavy chain or dynein-associated proteins such as Lis1 (also known as Pafah1b1) and Ndel1/Nde1 weakens the coordination between the centrosome and the nucleus, and inhibits the centrosomal/nuclear forward movement (Shu et al., 2004; Tanaka et al., 2004). The actin cytoskeleton and myosin also contribute to nuclear translocation in migrating neurons. When positioned ahead of the nucleus, actomyosin appears to pull the structure (Solecki et al., 2009); when behind the nucleus, actomyosin squeezes the rear side of the cells to push the nucleus forward (Schaar and McConnell, 2005). In brief, motor proteins play a key role in nuclear migration but the signals that regulate their behaviors in migrating neurons remain poorly understood.
AMP-activated protein kinase (AMPK) is an evolutionarily conserved Ser/Thr kinase with multiple functions including energy homeostasis, cell cycle, cell polarity and cell motility (Dasgupta and Chhipa, 2016; Hardie et al., 2012; Mihaylova and Shaw, 2011). AMPK consists of three subunits: a catalytic α-subunit, and regulatory β- and γ-subunits. Full activation of AMPK requires its phosphorylation at Thr172 in the kinase domain of the α-subunit. In addition, AMP binding to the γ-subunit promotes net phosphorylation at Thr172 by protecting Thr172 from its dephosphorylation and enhancing the association of AMPK with the upstream kinase (Jeon, 2016; Stein et al., 2000). To date, two protein kinases, liver kinase B1 (LKB1) and calcium/calmodulin-dependent protein kinase β (CaMKKβ), have been identified as major upstream kinases responsible for Thr172 phosphorylation and activation of AMPK in mammalian cells (Dasgupta and Chhipa, 2016; Jeon, 2016). Noticeably, LKB1 directs centrosomal/nuclear forward movement in migratory neurons in the developing neocortex (Asada and Sanada, 2010; Asada et al., 2007).
In the present study, we report that AMPK associates with dynein, phosphorylates dynein intermediate chain at Ser81, regulates cytoplasmic dynein behavior and drives nuclear migration in migratory neurons of the developing neocortex. Knockdown of AMPKα or expression of the phospho-resistant Ser81A mutant of dynein intermediate chain disrupts the nucleus-centrosome coupling and impairs neuronal migration. The impaired neuronal migration observed upon AMPK depletion is alleviated by co-expression of the phospho-mimetic Ser81D mutant of dynein intermediate chain. Thus, AMPK modulates dynein behavior and nuclear migration in migratory neurons.
RESULTS
Expression of AMPKα and Thr172-phosphorylated AMPKα in the developing neocortex
In mammals, two isoforms of the α catalytic subunit of AMPK (AMPKα) exist: AMPKα1 and AMPKα2. We first examined the protein expression of the isoforms in the developing neocortex. AMPKα1 and AMPKα2 were both expressed in the developing mouse neocortex from E13 to P0, as assessed by western blotting using antibodies specific to AMPKα1 and AMPKα2 (Fig. 1A). Using an antibody that recognizes both AMPKα1 and AMPKα2, we also found that total AMPKα levels were relatively constant during the examined period (E13 to P0) (Fig. 1A). Moreover, Thr172-phosphorylated/activated AMPKα (pAMPKα), which was detected with an antibody recognizing the phosphorylated forms of both AMPKα1 and AMPKα2, was also present in the developing neocortex throughout this period, with relatively low levels at E13 and higher levels at E15-P0.
Expression of AMPKα in the developing mouse neocortex. (A) Lysates prepared from E13, E15, E17 and P0 neocortices were immunoblotted with the antibodies indicated. (B) E14 neocortical sections immunostained using antibodies against AMPKα, pAMPKα and pericentrin. Images of the entire cerebral wall are shown (left panels). Magnified images of the boxed areas are shown in the panels on the right. Nuclei were stained with TO-PRO-3 iodide. Two examples of pAMPKα-positive puncta are indicated by arrowheads in each panel. Scale bars: 50 µm (left panels); 10 µm (right panels). (C) Plasmids expressing GFP and DsRed-Centrin2 were together electroporated into E14 mouse neocortex, and brains were fixed at E17. The brain sections were immunostained with antibodies against GFP and pAMPKα. High-magnification images of a GFP-labeled migratory neuron located at CP are shown. The GFP-labeled neuron is outlined with a dashed line. Arrowheads indicate the position of the centrosome. Magnified images around the centrosome are shown in the insets. Scale bars: 5 µm; 2 µm (insets).
Expression of AMPKα in the developing mouse neocortex. (A) Lysates prepared from E13, E15, E17 and P0 neocortices were immunoblotted with the antibodies indicated. (B) E14 neocortical sections immunostained using antibodies against AMPKα, pAMPKα and pericentrin. Images of the entire cerebral wall are shown (left panels). Magnified images of the boxed areas are shown in the panels on the right. Nuclei were stained with TO-PRO-3 iodide. Two examples of pAMPKα-positive puncta are indicated by arrowheads in each panel. Scale bars: 50 µm (left panels); 10 µm (right panels). (C) Plasmids expressing GFP and DsRed-Centrin2 were together electroporated into E14 mouse neocortex, and brains were fixed at E17. The brain sections were immunostained with antibodies against GFP and pAMPKα. High-magnification images of a GFP-labeled migratory neuron located at CP are shown. The GFP-labeled neuron is outlined with a dashed line. Arrowheads indicate the position of the centrosome. Magnified images around the centrosome are shown in the insets. Scale bars: 5 µm; 2 µm (insets).
Immunohistochemical analysis revealed that total AMPKα is expressed in the VZ as well as in the cortical plate (CP) and intermediate zone (IZ), where migrating neurons are abundant. Immunofluorescent signals of pAMPKα were observed in a punctate pattern in the VZ/IZ/CP and showed remarkable overlapping localization with the centrosomal marker pericentrin in these regions (Fig. 1B). Detailed analysis of migrating neurons transfected with GFP and the centrosomal marker DsRed-Centrin2 confirmed that pAMPKα is enriched around the centrosome (Fig. 1C).
We further examined the subcellular distribution of pAMPKα in primary cortical neurons and Cos-7 cells. In cultured cortical neurons, pAMPKα signals were prominent around the centrosome but weak/diffuse in the cytoplasm (Fig. 2A). In addition, pAMPKα accumulation around the centrosome was remarkably diminished by the treatment of neurons with the microtubule-destabilizing reagent nocodazole, suggesting a microtubule-dependent accumulation of the activated kinase at the centrosome (Fig. S1). In Cos-7 cells, pAMPKα and AMPKα signals were punctate and highly enriched in the proximal part of the interphase microtubule arrays emanating from the centrosome. Around the centrosome and on the microtubule arrays, pAMPKα and AMPKα showed substantial overlapping distribution with dynein intermediate chain, a subunit of cytoplasmic dynein, and LIS1, a dynein-associated protein (Fig. 2B). Similar overlapped distribution was also observed in neurons in which pAMPKα and dynein intermediate chain were colocalized around the centrosome (Fig. S2). In this context, AMPKα may accumulate at this site via specific centrosomal and/or microtubule-binding component. Together, these observations suggest that AMPKα likely decorates microtubules in a similar manner to cytoplasmic dynein.
Subcellular localization of AMPKα. (A) Neocortical neurons at days in vitro 2 (DIV 2) were immunostained with antibodies against pAMPKα, Tuj1 (a neuronal marker), pericentrin and LIS1. Magnified images around the centrosome are shown in the insets. Scale bars: 5 µm; 1 µm (insets). (B) Cos-7 cells were immunostained using the antibodies indicated. Scale bars: 10 µm; 5 µm (insets).
Subcellular localization of AMPKα. (A) Neocortical neurons at days in vitro 2 (DIV 2) were immunostained with antibodies against pAMPKα, Tuj1 (a neuronal marker), pericentrin and LIS1. Magnified images around the centrosome are shown in the insets. Scale bars: 5 µm; 1 µm (insets). (B) Cos-7 cells were immunostained using the antibodies indicated. Scale bars: 10 µm; 5 µm (insets).
AMPKα depletion causes defects in neuronal migration
To investigate the putative role of AMPK in neuronal migration in the developing neocortex, we generated plasmids expressing short hairpin RNAs against AMPKα1 (AMPKα1 shRNA and shRNA#2) and AMPKα2 (AMPKα2 shRNA and shRNA#2). The shRNA constructs efficiently silenced the expression of the respective kinases that were transiently expressed in Cos-7 cells (Fig. S3A). On the other hand, expression of shRNA-resistant mutants of AMPKα1 (AMPKα1res) and AMPKα2 (AMPKα2res), which contain two silent mutations within the target sequence of the AMPKα1 shRNA and AMPKα2 shRNA, were not affected by their respective shRNA constructs (Fig. S3A). In addition, endogenous AMPKα expression was remarkably diminished in AMPKα shRNA-introduced primary cortical neurons (Fig. S3B). We then electroporated the AMPKα shRNA constructs together with the GFP-expressing plasmid into E14 mouse neocortices. In neocortices electroporated at E14 with the control shRNA, GFP-labeled control cells were mainly located in the IZ (28.3±1.4%) and CP (67.0±2.0%) at E17, and they migrate further into the CP at E18 (IZ, 5.6±2.4%; CP, 92.3±3.5%). In contrast, AMPKα2 shRNA-introduced cells were mainly located at the IZ at E17 (86.7±0.7%) and E18 (67.8±1.0%) (Fig. 3A,B). Similar abnormal distribution pattern of GFP-labeled neurons was observed in neocortices electroporated with the AMPKα1 shRNA construct as well as with a mixture of the AMPKα1 shRNA plasmid and the AMPKα2 shRNA plasmid (hereafter referred to as AMPKα1/2 shRNA). Furthermore, introduction of second shRNA plasmids for AMPKα1 (AMPKα1 shRNA#2) and for AMPKα2 (AMPKα2 shRNA#2) also significantly retarded neuronal migration (Fig. S4A). Of note, about 75% of AMPKα1/2 shRNA-introduced cells in the IZ/CP display a thick leading process, while the remaining 25% show multipolar morphology in the E17 neocortices electroporated at E14. This proportion is not significantly different from that of control cells with a leading process (79.1±4.5% in control versus 75.3±6.5% in AMPKα1/2 shRNA, n=3 embryos; Fig. S5). This observation suggests that the multipolar-to-bipolar transition of newborn neurons is not affected by AMPKα depletion and that early stages of radial migration are severely affected in these cells, thereby causing their accumulation in the IZ. In addition, mispositioning of GFP-labeled neurons in the neocortices electroporated with the AMPKα1 and α2 shRNA constructs was not caused by enhanced neuronal apoptosis: cleaved caspase 3 signal was not enhanced upon electroporation of the AMPKα1/2 shRNA construct when compared with that generated with the control shRNA (Fig. S6). Furthermore, neuronal differentiation of progenitor cells was also unaffected by AMPKα1/2 shRNA electroporation: the populations of GFP-labeled cells positive for Pax6 (a neural progenitor marker) were unchanged in E15 neocortices electroporated with AMPKα1/2 shRNA at E14 (Fig. S7). Importantly, the defect in neuronal positioning in AMPKα1 shRNA- and AMPKα2 shRNA-introduced neocortices was almost completely rescued by co-expression of respective AMPKα1res and AMPKα2res that are resistant to RNAi (Fig. 3C,D, Fig. S4B). These results not only confirmed the specificity of AMPKα shRNAs but also indicated that impaired neuronal migration of AMPKα shRNA-introduced neurons is indeed caused by AMPKα depletion. Similarly, the positioning defect of AMPKα1 shRNA-introduced neocortices was significantly rescued by co-expression of AMPKα2 (Fig. S4B), suggesting that AMPKα1 and AMPKα2 redundantly affects neuronal migration, a process sensitive to the dosage of total AMPKα.
Knockdown of AMPKα impairs neuronal migration. (A,B) Plasmids expressing control shRNA and AMPKα shRNAs, as indicated, were electroporated into E14 neocortices, together with the GFP-expressing plasmid. Brains were fixed at E17 and E18 as indicated. Brain sections were immunostained using anti-GFP antibody. Nuclei were stained with TO-PRO-3 iodide. Representative images are shown in A. The percentage of GFP-labeled cells in the VZ, IZ and CP was calculated and is plotted in B as the mean±s.e.m. (n=3 embryos for E17 AMPKα2 shRNA; n=4 embryos for E17 control shRNA, E17 and E18 AMPKα1 shRNA, E18 AMPKα2 shRNA, and E17 AMPKα1/2 shRNA; n=5 embryos for E18 control shRNA and E18 AMPKα1/2 shRNA). ***P<0.001 versus control (two-tailed Welch's t-test). In B, a total of 1662 (E17) and 1805 (E18) cells were counted for control shRNA, 1423 (E17) and 1238 (E18) cells for AMPKα1 shRNA, 938 (E17) and 1467 (E18) cells for AMPKα2 shRNA, and 801 (E17) and 1944 (E18) cells for AMPKα1/2 shRNA. (C,D) Plasmids expressing control shRNA, AMPKα2 shRNA, AMPKα2res (AMPKα2 resistant to AMPKα2 shRNA) and a kinase-dead form (T172A) of AMPKα2res (AMPKα2 resistant to AMPKα2 shRNA), as indicated, were electroporated into E14 neocortices, together with the GFP-expressing plasmid. E18 brain sections were immunostained with anti-GFP antibody. Nuclei were stained with TO-PRO-3 iodide. Representative images are shown in C. The percentage of GFP-labeled cells in the VZ, IZ and CP was calculated, and is plotted in D as the mean±s.e.m. (n=4 embryos for control shRNA, AMPKα2 shRNA, AMPKα2 shRNA+AMPKα2res, AMPKα2 shRNA+T172A AMPKα2res and T172A AMPKα2res). **P<0.01 (two-tailed Welch's t-test). NS, not significant. In D, a total 815 cells was counted for control shRNA, 518 cells for AMPKα2 shRNA, 857 cells for AMPKα2 shRNA+AMPKα2res, 851 cells for AMPKα2 shRNA+T172A AMPKα2res and 675 cells for T172A AMPKα2res. Scale bars: 100 µm.
Knockdown of AMPKα impairs neuronal migration. (A,B) Plasmids expressing control shRNA and AMPKα shRNAs, as indicated, were electroporated into E14 neocortices, together with the GFP-expressing plasmid. Brains were fixed at E17 and E18 as indicated. Brain sections were immunostained using anti-GFP antibody. Nuclei were stained with TO-PRO-3 iodide. Representative images are shown in A. The percentage of GFP-labeled cells in the VZ, IZ and CP was calculated and is plotted in B as the mean±s.e.m. (n=3 embryos for E17 AMPKα2 shRNA; n=4 embryos for E17 control shRNA, E17 and E18 AMPKα1 shRNA, E18 AMPKα2 shRNA, and E17 AMPKα1/2 shRNA; n=5 embryos for E18 control shRNA and E18 AMPKα1/2 shRNA). ***P<0.001 versus control (two-tailed Welch's t-test). In B, a total of 1662 (E17) and 1805 (E18) cells were counted for control shRNA, 1423 (E17) and 1238 (E18) cells for AMPKα1 shRNA, 938 (E17) and 1467 (E18) cells for AMPKα2 shRNA, and 801 (E17) and 1944 (E18) cells for AMPKα1/2 shRNA. (C,D) Plasmids expressing control shRNA, AMPKα2 shRNA, AMPKα2res (AMPKα2 resistant to AMPKα2 shRNA) and a kinase-dead form (T172A) of AMPKα2res (AMPKα2 resistant to AMPKα2 shRNA), as indicated, were electroporated into E14 neocortices, together with the GFP-expressing plasmid. E18 brain sections were immunostained with anti-GFP antibody. Nuclei were stained with TO-PRO-3 iodide. Representative images are shown in C. The percentage of GFP-labeled cells in the VZ, IZ and CP was calculated, and is plotted in D as the mean±s.e.m. (n=4 embryos for control shRNA, AMPKα2 shRNA, AMPKα2 shRNA+AMPKα2res, AMPKα2 shRNA+T172A AMPKα2res and T172A AMPKα2res). **P<0.01 (two-tailed Welch's t-test). NS, not significant. In D, a total 815 cells was counted for control shRNA, 518 cells for AMPKα2 shRNA, 857 cells for AMPKα2 shRNA+AMPKα2res, 851 cells for AMPKα2 shRNA+T172A AMPKα2res and 675 cells for T172A AMPKα2res. Scale bars: 100 µm.
AMPKα kinase activity is also required for proper migration of newborn cortical neurons as expression of kinase-dead AMPKα (T172A AMPKα and D157A AMPKα) (Stein et al., 2000) did not rescue migration defects of AMPKα-depleted neurons (Fig. 3C,D, Fig. S4C). In addition, at P4, almost all AMPKα-depleted neurons were found within the CP (Fig. S8). Thus, AMPKα depletion strongly delays neuronal migration, rather than simply arresting it. Of note, AMPKα-depleted neurons at P4 resided in more superficial layers when compared with control neurons (Fig. S8). Considering that neuronal differentiation of neural progenitor cells was not altered upon AMPKα depletion (Fig. S7), we reason that this superficial positioning phenotype is not caused by alterations in the birthdate but rather by delayed migration.
AMPKα-depletion disrupts nuclear translocation
The directed centrosomal movement and proper centrosome-nucleus coupling are essential for nuclear migration and consequently for neuronal migration. To investigate the effect of AMPKα knockdown on the positioning of the centrosome and nucleus in migratory neurons, we introduced into E14 embryos the DsRed-Centrin2-expressing plasmid (a maker of the centrosome) and the GFP-expressing plasmid together with either the control shRNA plasmid or AMPKα1/2 shRNA plasmid. We then analyzed migratory neurons located at upper part of the IZ and lower part of the CP at E17. We found that, in AMPKα-depleted neurons, the centrosome was far away from the nucleus (6.13±0.63 µm) when compared with the centrosome-nucleus distance in control neurons (1.53±0.33 µm) (Fig. 4A,B). In normal migrating neurons, the nucleus is elongated toward the direction of migration due to the application of strong traction forces on the nucleus by dynein (Tsai et al., 2007). The ratio of nuclear length over its width (L/W) of control neurons was 2.42±0.06, while the L/W value of dynein heavy chain-depleted neurons was significantly lower (1.98±0.06) than that of control (Fig. 4A,B). These results support the idea that the distorted nuclear shape is caused by the traction force generated by dynein. Notably, the L/W values of AMPKα-depleted neurons were similar to those of dynein heavy chain-depleted neurons and remarkably lower than those of control cells (AMPKα1/2 shRNA, 1.88±0.05; AMPKα1 shRNA, 2.18±0.07; AMPKα2 shRNA, 1.85±0.06) (Fig. 4A,B). Thus, like dynein, AMPK is required for the spatial positioning of the centrosome/nucleus and coordination of the nucleus-centrosome coupling.
Knockdown of AMPKα impairs centrosomal and nuclear movement. (A) Plasmids expressing control shRNA, AMPKα1/2 shRNA and dynein heavy chain shRNA (dynein HC shRNA) were electroporated into E14 neocortices together with plasmids expressing GFP and DsRed-Centrin2. E17 brain sections were immunostained using anti-GFP antibody. Nuclei were stained with TO-PRO-3 iodide. GFP-positive migrating neurons in the upper IZ and lower CP were analyzed. Arrowheads indicate the position of the centrosome labeled by DsRed-Centrin2. GFP-labeled cells and their nuclei are outlined with dashed lines. Scale bar: 5 µm. (B) The distance between the nucleus and the centrosome was measured and is plotted as the mean±s.e.m. (n=27 for control shRNA and n=24 for AMPKα1/2 shRNA) (left). In the right graph, the ratios of nuclear lengths over widths (L/W) were measured and are plotted as the mean±s.e.m. (n=71 for control shRNA, n=94 for AMPKα1/2 shRNA, n=67 for dynein HC shRNA). ***P<0.001 versus control (two-tailed Welch's t-test). Scale bar: 5 µm. (C) Either the control shRNA plasmid or the AMPKα1/2 shRNA plasmid was electroporated into E14 cortices together with the CFP-expressing plasmid, DsRed-Centrin2-expressing plasmid and Histone H2B-GFP-expressing plasmid. Coronal brain slices were prepared at E17 and subjected to time-lapse imaging with 20 min intervals over a period of 420-700 min. Time is indicated in h:min in the top of each panel. Brackets indicate the position of the swelling. Scale bars: 10 µm. (D) Representative tracings of the centrosomal movement and the nuclear movement in control cells and AMPKα1/2-depleted cells are presented in the graphs.
Knockdown of AMPKα impairs centrosomal and nuclear movement. (A) Plasmids expressing control shRNA, AMPKα1/2 shRNA and dynein heavy chain shRNA (dynein HC shRNA) were electroporated into E14 neocortices together with plasmids expressing GFP and DsRed-Centrin2. E17 brain sections were immunostained using anti-GFP antibody. Nuclei were stained with TO-PRO-3 iodide. GFP-positive migrating neurons in the upper IZ and lower CP were analyzed. Arrowheads indicate the position of the centrosome labeled by DsRed-Centrin2. GFP-labeled cells and their nuclei are outlined with dashed lines. Scale bar: 5 µm. (B) The distance between the nucleus and the centrosome was measured and is plotted as the mean±s.e.m. (n=27 for control shRNA and n=24 for AMPKα1/2 shRNA) (left). In the right graph, the ratios of nuclear lengths over widths (L/W) were measured and are plotted as the mean±s.e.m. (n=71 for control shRNA, n=94 for AMPKα1/2 shRNA, n=67 for dynein HC shRNA). ***P<0.001 versus control (two-tailed Welch's t-test). Scale bar: 5 µm. (C) Either the control shRNA plasmid or the AMPKα1/2 shRNA plasmid was electroporated into E14 cortices together with the CFP-expressing plasmid, DsRed-Centrin2-expressing plasmid and Histone H2B-GFP-expressing plasmid. Coronal brain slices were prepared at E17 and subjected to time-lapse imaging with 20 min intervals over a period of 420-700 min. Time is indicated in h:min in the top of each panel. Brackets indicate the position of the swelling. Scale bars: 10 µm. (D) Representative tracings of the centrosomal movement and the nuclear movement in control cells and AMPKα1/2-depleted cells are presented in the graphs.
Of note, the perinuclear cage-like microtubules were clearly detected in AMPKα-depleted neurons in culture (Fig. S9), implicating that disruption of the microtubule array is not responsible for the phenotype. Based on a previous study indicating that swelling formation and nuclear elongation are coupled (Nishimura et al., 2014), we next inspected swelling formation in these neurons. We found that formation of the swelling is not disrupted but rather promoted in AMPKα-depleted neurons in E17 neocortices electroporated at E14. Indeed, ∼65% of AMPKα-depleted neurons have a swelling near the base of the leading process (63.5±5.6% in AMPKα1/2, n=3), whereas only 40% of control neurons display the swelling (37.0±2.5% n=3 embryos, P=0.023). We hypothesize that the increased proportion of AMPKα-depleted neurons with a swelling is related to their arrested migration at stages with swelling. Alternatively, this phenotype may reflect a problem in the coupling between nuclear elongation (Fig. 4) and swelling formation.
To further define the role for AMPKα in neuronal migration, we next analyzed the centrosomal/nuclear movements in migrating neurons in brain slices. For this, we introduced plasmids expressing DsRed-Centrin2, Histone H2B-GFP (nuclear marker) and CFP into neurons, together with either the control shRNA plasmid or the AMPKα1/2 shRNA plasmid. We imaged AMPKα-depleted neurons that accumulated in the upper IZ (Fig. 3). Consistent with previous reports (Tsai et al., 2007; Zhang et al., 2009), control neurons migrated radially in a typical saltatory manner: a cytoplasmic swelling is formed near the base of the leading process, the centrosome moves forward with relatively constant motion until it reaches the swelling, and the nucleus moves to the centrosome in a saltatory manner (Fig. 4C,D). In addition, the nucleus appeared distorted toward the direction of migration. On the other hand, AMPKα1/2-depleted neurons showed impaired centrosomal and nuclear movements with the centrosome stuck just ahead of the nucleus or arrested at the swelling formed in the leading process (Fig. 4C,D). In addition, the nucleus of AMPKα1/2-depleted neurons mostly adopted a rounded-shape rather than a distorted shape when compared with that of control cells (Fig. 4C). Altogether, these results indicate that AMPK is required for proper nuclear and centrosomal movements; by coordinating these motions, the kinase regulates neuronal migration.
AMPK interacts with cytoplasmic dynein in vivo
A proteomics screen performed for proteins phosphorylated by AMPK in the mouse brain suggested that DIC is a candidate substrate for the kinase (Tuerk et al., 2007). In this study, we found the overlapping distribution of AMPKα with cytoplasmic dynein in cells (Fig. 2, Fig. S2) and the remarkable phenocopy between AMPKα-depleted and dynein heavy chain-depleted neurons in terms of L/W nuclear ratio and neuronal migration defects (Fig. 4) (Tsai et al., 2007). Therefore, we sought to investigate the potential physical interaction of AMPK with cytoplasmic dynein. We first performed co-immunoprecipitation assays using lysates from Cos-7 cells and E17 telencephalon. As shown in Fig. 5A, the dynein intermediate chain (DIC) antibody pulled down DIC as well as the dynein accessory proteins p150glued and Lis1. DIC antibody also co-immunoprecipitated AMPKα and pAMPKα (Fig. 5A). On the other hand, these proteins were hardly detectable in immunoprecipitates assayed with a control antibody (GFP antibody). Furthermore, Flag-tagged AMPKα and HA-tagged DIC transfected into HEK 293 cells were reciprocally co-immunoprecipitated (Fig. 5B). In brief, AMPK physically associates with the cytoplasmic dynein in vitro and in vivo.
AMPK interacts with cytoplasmic dynein. (A) Lysates from E17 telencephalons (right) and Cos-7 cells (left), as well as control buffer (lysis buffer), were subjected to immunoprecipitation with an anti-DIC antibody (DIC mAb), an anti-GFP antibody (GFP mAb) and control IgG (control mAb). Immunoprecipitates were then immunoblotted with antibodies against pAMPKα, AMPKα, DIC, p150glued and Lis1. Inputs are 1% starting material. (B) HEK293 cells were transfected with Flag-tagged AMPKα (Flag-AMPKα) and HA-tagged DIC (dynein IC-HA) as indicated. Cell lysates were subjected to immunoprecipitation with an anti-Flag tag antibody (upper panels) and an anti-HA tag antibody (lower panels). Immunoprecipitates were immunoblotted with antibodies against Flag-tag and HA-tag. Inputs are 5% starting material.
AMPK interacts with cytoplasmic dynein. (A) Lysates from E17 telencephalons (right) and Cos-7 cells (left), as well as control buffer (lysis buffer), were subjected to immunoprecipitation with an anti-DIC antibody (DIC mAb), an anti-GFP antibody (GFP mAb) and control IgG (control mAb). Immunoprecipitates were then immunoblotted with antibodies against pAMPKα, AMPKα, DIC, p150glued and Lis1. Inputs are 1% starting material. (B) HEK293 cells were transfected with Flag-tagged AMPKα (Flag-AMPKα) and HA-tagged DIC (dynein IC-HA) as indicated. Cell lysates were subjected to immunoprecipitation with an anti-Flag tag antibody (upper panels) and an anti-HA tag antibody (lower panels). Immunoprecipitates were immunoblotted with antibodies against Flag-tag and HA-tag. Inputs are 5% starting material.
AMPK phosphorylates DIC at Ser81
We next tested whether AMPKα phosphorylates DIC. For this purpose, phosphate-affinity gel electrophoresis using the acrylamide-pendant Mn2+-Phos-tag (Phos-tag SDS-PAGE) (Kinoshita et al., 2005) was used to evaluate the phosphorylation states of DIC. DIC in E13 to P0 neocortical lysates ran as a single band on the normal SDS-PAGE, whereas it ran as two bands on the Phos-tag SDS-PAGE (Fig. 6A). Noticeably, alkaline phosphatase treatment abolished the slower-migrating DIC band with a concomitant increase in the faster-migrating species (Fig. 6B). These observations suggest that DIC is a phosphoprotein in the developing neocortex and that the slower-migrating species corresponds to the phosphorylated protein. Similar patterns of phosphorylation state of DIC were also observed in Cos-7 cells and primary cortical neurons (Fig. 6C,D).
DIC is a phosphoprotein in the developing neocortex. (A) Lysates from neocortices (E13, E15, E17 and P0) were subjected to Phos-tag SDS-PAGE [Phos-tag(+), upper panel] and normal SDS-PAGE [Phos-tag(−), lower panel], followed by immunoblotting with a DIC antibody. (B) Lysates from E15 neocortices were incubated with or without alkaline phosphatase and subjected to Phos-tag SDS-PAGE, followed by immunoblotting with a DIC antibody. (C,D) Lysates from Cos-7 cells (C) and 4 DIV primary neurons (D) were incubated with or without alkaline phosphatase and subjected to Phos-tag SDS-PAGE, followed by immunoblotting with a DIC antibody. (E,F) Cos-7 cells (E) and cultured neurons (F) were treated with vehicle (0.2% DMSO), 100 µM A769662 (an AMPK activator), 2 mM AICAR (an AMPK activator) or 20 µM compound C (an AMPK inhibitor) as indicated for 1 h. The cell lysates were subjected to Phos-tag SDS-PAGE, followed by immunoblotting with a DIC antibody (upper panels). The lysates subjected to normal SDS-PAGE were also immunoblotted with antibodies indicated (lower panels). Representative results are shown. Similar results were obtained from three or four independent experiments.
DIC is a phosphoprotein in the developing neocortex. (A) Lysates from neocortices (E13, E15, E17 and P0) were subjected to Phos-tag SDS-PAGE [Phos-tag(+), upper panel] and normal SDS-PAGE [Phos-tag(−), lower panel], followed by immunoblotting with a DIC antibody. (B) Lysates from E15 neocortices were incubated with or without alkaline phosphatase and subjected to Phos-tag SDS-PAGE, followed by immunoblotting with a DIC antibody. (C,D) Lysates from Cos-7 cells (C) and 4 DIV primary neurons (D) were incubated with or without alkaline phosphatase and subjected to Phos-tag SDS-PAGE, followed by immunoblotting with a DIC antibody. (E,F) Cos-7 cells (E) and cultured neurons (F) were treated with vehicle (0.2% DMSO), 100 µM A769662 (an AMPK activator), 2 mM AICAR (an AMPK activator) or 20 µM compound C (an AMPK inhibitor) as indicated for 1 h. The cell lysates were subjected to Phos-tag SDS-PAGE, followed by immunoblotting with a DIC antibody (upper panels). The lysates subjected to normal SDS-PAGE were also immunoblotted with antibodies indicated (lower panels). Representative results are shown. Similar results were obtained from three or four independent experiments.
When Cos-7 cells were treated with an AMPK activator (A769662), the slower-migrating DIC band intensity increased with a concomitant decrease in the faster-migrating species (Fig. 6E). Conversely, treatment with the AMPK inhibitor compound C substantially decreased the upper band intensity, and this coincided with an increase in the lower band intensity (Fig. 6E). We also confirmed that AMPK activity increases following treatment with A769662, as the drug upregulated the levels of pAMPKα and phosphorylated form of acetyl-coenzyme A-carboxylase (ACC), a major AMPK substrate (Fig. 6E). Conversely, compound C treatment decreased phosphorylated ACC levels (Fig. 6E). Similar results were obtained in primary neurons treated with AMPK activators (A769662 and AICAR) and an AMPK inhibitor (compound C) (Fig. 6F). Since a substantial decrease in the slower-migrating species was observed following treatment with compound C, we suspected that the slower-migrating phospho-species were generated predominantly by endogenous AMPK activity. Consistent with this notion, the levels of the slower-migrating DIC species correlated with the levels of pAMPKα in the developing neocortex: they are both expressed at low levels at E13 and are upregulated at E15-P0 (Figs 1A and 6A). Altogether, these observations indicate that AMPK contributes to the phosphorylation of DIC.
To determine whether AMPKα directly phosphorylates DIC, we performed in vitro phosphorylation assay using recombinant proteins. In mammals, there are two different dynein intermediate chain genes, Ic1 (Dync1i1) and Ic2 (Dync1i2) and each gene has several splice variants. In our experiments, we used the IC-2B isoform, as it is known to be expressed abundantly in neurons and developing brain (Kuta et al., 2010; Pfister et al., 1996a,b). We incubated bacterially expressed recombinant GST-tagged DIC with recombinant active AMPKα1/β1/γ2 and thereafter subjected them to normal SDS-PAGE. As shown in Fig. 7A, GST-DIC incubated with active AMPK showed a significant mobility shift. The upshifted band was detected with an antibody recognizing phospho-Ser (Fig. 7A), suggesting that AMPK directly phosphorylates DIC at Ser residue(s). To identify the site(s) of phosphorylation, the upshifted band of GST-DIC was excised from the gels and subjected to in-gel tryptic digestion followed by liquid chromatography-tandem mass spectrometry (LC-MS/MS). The mass spectrometry analysis identified two tryptic peptides (42-EAAVSVQEESDLEK-55 and 59-EAEALLQSMGLTTDSPIVPPPMSPSSK-85), each with a single phosphorylation (Fig. 7B). Fragmentation analysis of these peptides revealed the Ser51-phosphorylation in the first peptide and a single phosphorylation between Ser81 and Lys85 of the second peptide. As three Ser residues exist in between, Ser81, Ser83 and Ser84, in conjunction with Ser51, were considered as the potential phosphorylation sites of DIC by AMPK. Of note, these sites identified are highly conserved among IC-1 and IC-2 splice variants in mammals.
AMPK phosphorylates DIC at Ser81. (A) Recombinant GST-DIC was incubated with or without recombinant active AMPKα1/β1/γ2 for 180 min and subjected to SDS-PAGE, followed by Coomassie Brilliant Blue staining (left panel). The reaction mixtures (incubated for 30 min/60 min) were also subjected to western blotting with anti-phospho-Ser antibody (pSer) and anti-DIC antibody (right panels). (B) Schematic representation of two tryptic fragments that were identified to be phosphorylated in DIC. Domains present in DIC are also shown. LC-MS/MS fragmentation analysis of GST-DIC incubated with active AMPK identified Ser51-phosphorylation in the first peptide (red) and a single phosphorylation between Ser81 and Lys85 of the second peptide (red). (C) Lysates from Cos-7 cells transfected with either Flag-tagged wild-type DIC (WT) or Flag-tagged DIC mutants indicated were subjected to Phos-tag SDS-PAGE, followed by immunoblotting with an antibody against Flag-tag. (D) Cos-7 cells were transfected with plasmids expressing constitutively active AMPK (AMPK CA), HA-tagged wild-type DIC (WT) and HA-tagged DIC S81A, as indicated. Cell lysates were subjected to Phos-tag SDS-PAGE, followed by immunoblotting with an antibody against HA-tag. (E) Lysates from cultured cortical neurons transfected with either HA-tagged wild-type DIC (WT) or HA-tagged DIC S81A were subjected to Phos-tag SDS-PAGE, followed by immunoblotting with anti-HA antibody.
AMPK phosphorylates DIC at Ser81. (A) Recombinant GST-DIC was incubated with or without recombinant active AMPKα1/β1/γ2 for 180 min and subjected to SDS-PAGE, followed by Coomassie Brilliant Blue staining (left panel). The reaction mixtures (incubated for 30 min/60 min) were also subjected to western blotting with anti-phospho-Ser antibody (pSer) and anti-DIC antibody (right panels). (B) Schematic representation of two tryptic fragments that were identified to be phosphorylated in DIC. Domains present in DIC are also shown. LC-MS/MS fragmentation analysis of GST-DIC incubated with active AMPK identified Ser51-phosphorylation in the first peptide (red) and a single phosphorylation between Ser81 and Lys85 of the second peptide (red). (C) Lysates from Cos-7 cells transfected with either Flag-tagged wild-type DIC (WT) or Flag-tagged DIC mutants indicated were subjected to Phos-tag SDS-PAGE, followed by immunoblotting with an antibody against Flag-tag. (D) Cos-7 cells were transfected with plasmids expressing constitutively active AMPK (AMPK CA), HA-tagged wild-type DIC (WT) and HA-tagged DIC S81A, as indicated. Cell lysates were subjected to Phos-tag SDS-PAGE, followed by immunoblotting with an antibody against HA-tag. (E) Lysates from cultured cortical neurons transfected with either HA-tagged wild-type DIC (WT) or HA-tagged DIC S81A were subjected to Phos-tag SDS-PAGE, followed by immunoblotting with anti-HA antibody.
To determine which of these sites is (are) phosphorylated by AMPK, we examined the phosphorylation status of mutant DIC in which each of the four sites was substituted with alanine. We then transfected these mutants (DIC S51A, S81A, S83A and S84A) with the Flag-tag into Cos-7 cells, and cell lysates were subjected to Phos-tag SDS-PAGE followed by western blotting with the Flag antibody. DIC S51A, S83A and S84A mutants retained the slower-migrating species and thus displayed two bands similar to wild-type DIC (Fig. 7C). On the other hand, the DIC S81A mutant did not exhibit the slower-migrating phospho-species (Fig. 7C). Even in the presence of constitutively active AMPK transfected, the slower-migrating species were not detected (Fig. 7D). These observations indicate that substitution of Ser81 is sufficient to block DIC phosphorylation by AMPK. Similarly, when the DIC S81A mutant was introduced into primary cortical neurons, the slower-migrating species was not observed, when compared to neurons transfected with wild-type DIC (Fig. 7E). Thus, Ser81 is the main phosphorylation site for AMPK in neurons.
AMPK inhibition and forced expression of the DIC S81A mutant negatively regulates dynein functions
To determine whether Ser81 phosphorylation of DIC is important for dynein function, we examined the effect of DIC S81A mutant overexpression on the dynein-dependent organelle motility. Distribution of the Golgi apparatus is restricted to around the centrosome, and this is maintained by dynein-dependent transport of vesicles towards the minus-ends of microtubules. Indeed, disruption of cytoplasmic dynein activity by knockout of dynein heavy chain or microinjection of a function-blocking DIC antibody causes dispersion of the Golgi apparatus (Burkhardt et al., 1997; Harada et al., 1998). We first analyzed the distribution of the Golgi apparatus in Cos-7 cells transfected with GFP, C-terminally GFP-fused wild-type DIC or C-terminally GFP-fused DIC S81A. In GFP-introduced control cells, the Golgi apparatus, visualized using an antibody against the cis-Golgi marker GM130, was clustered at perinuclear region (Fig. 8A). The distribution of the Golgi apparatus in wild-type DIC-introduced cells was indistinguishable from that in control cells (Fig. 8A). In contrast, DIC S81A expression caused significant dispersion of the organelle, thereby increasing the area marked by GM130 signals (Fig. 8A). Furthermore, similar dispersion was observed in Cos-7 cells treated with the AMPK inhibitor compound C (Fig. 8B), suggesting that AMPK activity and Ser81 phosphorylation of DIC positively regulate dynein functions. Finally, we examined the distribution of the Golgi apparatus in migrating neurons in vivo. In control GFP-labeled neurons, the organelle was distributed in cluster within the proximal part of the leading process (Fig. S10). Remarkably, in DIC-S81A-introduced neocortices, the Golgi apparatus became significantly elongated along the leading process (Fig. S10). These observations indicate that the AMPKα contributes to the modulation of dynein behavior in vivo.
DIC S81A mutant expression impairs dynein functions. (A) Plasmids expressing GFP, C-terminally GFP-tagged wild-type DIC (DIC WT-GFP) and C-terminally GFP-tagged DIC S81A (DIC S81A-GFP) were transfected into Cos-7 cells as indicated. The cells were fixed 24 h after transfection and immunostained with anti-GM130 antibody (a cis-Golgi marker). Nuclei were stained with TO-PRO-3 iodide. The Golgi area and cell area of individual transfected cells were measured and are plotted as the mean± s.e.m. (n=52 for GFP, n=51 for WT-GFP and n=54 for S81A-GFP). ***P<0.001 versus control GFP-transfected cells (two-tailed Welch's t-test). NS, not significant. Scale bar: 10 µm. (B) Cos-7 cells were treated with either vehicle (0.2% DMSO) or compound C (20 µM) for 1 h and fixed, followed by immunostaining with anti-GM130 antibody. Nuclei were stained with TO-PRO-3 iodide. Scale bar: 10 µm.
DIC S81A mutant expression impairs dynein functions. (A) Plasmids expressing GFP, C-terminally GFP-tagged wild-type DIC (DIC WT-GFP) and C-terminally GFP-tagged DIC S81A (DIC S81A-GFP) were transfected into Cos-7 cells as indicated. The cells were fixed 24 h after transfection and immunostained with anti-GM130 antibody (a cis-Golgi marker). Nuclei were stained with TO-PRO-3 iodide. The Golgi area and cell area of individual transfected cells were measured and are plotted as the mean± s.e.m. (n=52 for GFP, n=51 for WT-GFP and n=54 for S81A-GFP). ***P<0.001 versus control GFP-transfected cells (two-tailed Welch's t-test). NS, not significant. Scale bar: 10 µm. (B) Cos-7 cells were treated with either vehicle (0.2% DMSO) or compound C (20 µM) for 1 h and fixed, followed by immunostaining with anti-GM130 antibody. Nuclei were stained with TO-PRO-3 iodide. Scale bar: 10 µm.
Defects in neuronal migration upon AMPK depletion are rescued by expression of a phospho-mimetic mutant of DIC
We next examined whether Ser81 phosphorylation of DIC contributes to radial migration of cortical neurons. We expressed either wild-type DIC or DIC S81A together with the GFP-expressing plasmid into E14 mouse neocortices and analyzed the distribution of the GFP-labeled cells at E18. In E18 neocortices electroporated with wild-type DIC, the majority of the GFP-labeled neurons were located within the CP (66.3±3.7%), whereas a smaller population of cells was detected in the IZ and VZ (Fig. 9A,B). On the other hand, in DIC S81A-introduced neocortices, most of the GFP-labeled cells were located at the IZ (58.6±3.9%), and a small population of cells migrated into the CP (37.8±4.3%) (Fig. 9A,B). Detailed immunohistochemical analysis of migratory neurons electroporated with GFP and DsRed-Centrin2 showed that the nucleus-centrosome distance was remarkably increased in DIC S81A-introduced neurons when compared with that of control neurons and wild-type DIC-electroporated neurons (control, 2.14±0.35 µm; wild-type DIC, 2.75±0.37 µm; DIC S81A, 5.82±0.64 µm). In addition, the L/W ratio of the nucleus in DIC S81A-introduced neurons was significantly lower than those of control and wild-type DIC-introduced neurons (control, 2.51±0.11; wild-type DIC, 2.49±0.08; DIC S81A, 1.87±0.08), but similar to the ratio found in AMPKα-depleted neurons (Fig. 9C,D). These results indicate that overexpression of DIC S81A in neurons disrupts the nucleus-centrosome coupling and impairs neuronal migration in the same ways as AMPKα depletion does.
Defects in neuronal migration upon AMPK depletion are rescued by expression of phospho-mimetic mutant of DIC. (A-D) Plasmids expressing wild-type DIC (DIC WT) and DIC S81A were electroporated into E14 neocortices, together with GFP and DsRed-Centrin2, as indicated. E18 brain sections were immunostained with anti-GFP antibody. Nuclei were stained with TO-PRO-3 iodide. (A) Images of the entire cerebral wall are shown. Scale bar: 100 µm. (B) The percentage of GFP-labeled cells in the VZ, IZ and CP was measured and is plotted as the mean±s.e.m. (n=4 embryos; 1271 neurons were analyzed for DIC WT and 948 neurons for DIC S81A). **P<0.01 versus wild-type DIC (two-tailed Welch's t-test). (C) High-magnification images of GFP-labeled migratory neurons are shown. Arrowheads indicate the position of the centrosome labeled by DsRed-Centrin2. The nuclei in GFP-labeled neurons are outlined with dashed lines. Scale bar: 5 µm. (D) In GFP-labeled neurons at the upper IZ and lower CP, the distance between the nucleus and the centrosome (left graph), and the ratio of the nuclear length over its width (L/W, right graph) were measured and are plotted as mean±s.e.m. [n=44 (control), 42 (DIC WT) and 40 (DIC S81A) neurons for the nucleus-centrosome distance; n=41 (control), 39 (DIC WT) and 39 (DIC S81A) neurons for the nuclear shape]. ***P<0.001 (two-tailed Welch's t-test). NS, not significant. (E,F) Plasmids expressing control shRNA, AMPKα2 shRNA, DIC WT and DIC S81D, as indicated, were electroporated into E14 neocortices, together with the GFP-expressing plasmid. E18 brain sections were immunostained with anti-GFP antibody. Nuclei were stained with TO-PRO-3 iodide. Representative images are shown in E. Scale bar: 100 µm. (F) The percentage of GFP-labeled cells in the VZ, IZ and CP was measured and is plotted as the mean± s.e.m. (n=4 embryos for control shRNA, control shRNA+DIC WT, control shRNA+DIC S81D and AMPKα2 shRNA; n=5 embryos for AMPKα2 shRNA+DIC WT and AMPKα2 shRNA+DIC S81D; a total of 1069 neurons were analyzed for control shRNA, 975 neurons for control shRNA+ DIC WT, 1119 neurons for control shRNA+DIC S81D, 1014 neurons for AMPKα2 shRNA, 1208 neurons for AMPKα2 shRNA+DIC WT and 1373 neurons for AMPKα2 shRNA+DIC S81D). **P<0.01, ***P<0.001 (two-tailed Welch's t-test).
Defects in neuronal migration upon AMPK depletion are rescued by expression of phospho-mimetic mutant of DIC. (A-D) Plasmids expressing wild-type DIC (DIC WT) and DIC S81A were electroporated into E14 neocortices, together with GFP and DsRed-Centrin2, as indicated. E18 brain sections were immunostained with anti-GFP antibody. Nuclei were stained with TO-PRO-3 iodide. (A) Images of the entire cerebral wall are shown. Scale bar: 100 µm. (B) The percentage of GFP-labeled cells in the VZ, IZ and CP was measured and is plotted as the mean±s.e.m. (n=4 embryos; 1271 neurons were analyzed for DIC WT and 948 neurons for DIC S81A). **P<0.01 versus wild-type DIC (two-tailed Welch's t-test). (C) High-magnification images of GFP-labeled migratory neurons are shown. Arrowheads indicate the position of the centrosome labeled by DsRed-Centrin2. The nuclei in GFP-labeled neurons are outlined with dashed lines. Scale bar: 5 µm. (D) In GFP-labeled neurons at the upper IZ and lower CP, the distance between the nucleus and the centrosome (left graph), and the ratio of the nuclear length over its width (L/W, right graph) were measured and are plotted as mean±s.e.m. [n=44 (control), 42 (DIC WT) and 40 (DIC S81A) neurons for the nucleus-centrosome distance; n=41 (control), 39 (DIC WT) and 39 (DIC S81A) neurons for the nuclear shape]. ***P<0.001 (two-tailed Welch's t-test). NS, not significant. (E,F) Plasmids expressing control shRNA, AMPKα2 shRNA, DIC WT and DIC S81D, as indicated, were electroporated into E14 neocortices, together with the GFP-expressing plasmid. E18 brain sections were immunostained with anti-GFP antibody. Nuclei were stained with TO-PRO-3 iodide. Representative images are shown in E. Scale bar: 100 µm. (F) The percentage of GFP-labeled cells in the VZ, IZ and CP was measured and is plotted as the mean± s.e.m. (n=4 embryos for control shRNA, control shRNA+DIC WT, control shRNA+DIC S81D and AMPKα2 shRNA; n=5 embryos for AMPKα2 shRNA+DIC WT and AMPKα2 shRNA+DIC S81D; a total of 1069 neurons were analyzed for control shRNA, 975 neurons for control shRNA+ DIC WT, 1119 neurons for control shRNA+DIC S81D, 1014 neurons for AMPKα2 shRNA, 1208 neurons for AMPKα2 shRNA+DIC WT and 1373 neurons for AMPKα2 shRNA+DIC S81D). **P<0.01, ***P<0.001 (two-tailed Welch's t-test).
Finally, we investigated the functional relationship between AMPK and DIC in neuronal migration. We used a phospho-mimetic mutant of DIC (DIC S81D) that has Ser81 replaced with Asp to mimic the negative charge generated by phosphorylation. We electroporated the plasmid expressing DIC S81D and control shRNA or AMPKα2 shRNA into E14 neocortices. The distribution of the GFP-labeled cells in E18 neocortices electroporated with both DIC S81D and control shRNA (Fig. 9E,F) was not significantly different from that of neocortices electroporated solely with control shRNA or with both DIC wild-type and control shRNA. Thus, DIC S81D expression does not affect migration of neurons that have normal levels of AMPKα. Noticeably, mispositioning of neurons in neocortices electroporated with AMPKα2 shRNA was significantly rescued by overexpression of DIC S81D but not wild-type DIC (Fig. 9E,F). The rescue was partial, probably because both endogenous DIC competes with exogenous DIC S81D in the dynein protein complex.
DISCUSSION
AMPK regulates nuclear movement and neuronal migration
Motor proteins, particularly cytoplasmic dynein, are known to direct nuclear movement for neuronal migration in the developing neocortex. However, the signals that contribute to the regulation and function of dynein during both nuclear and neuronal migration remain unclear. The present study demonstrates that AMPKα is required for centrosomal and nuclear motility, nucleus-centrosome coupling as well as for neuronal migration in the developing neocortex. Specifically, we showed that AMPKα depletion causes remarkable retardation of neuronal migration in the developing neocortex (Fig. 3). This defect is associated with discoordination between the centrosome and nucleus, as revealed by the longer centrosome-nucleus distance in the AMPKα-depleted neurons (Fig. 4A,B), and with profound retardation of the centrosomal/nuclear movement, as demonstrated by time-lapse imaging of migratory neurons in brain slices (Fig. 4C,D).
A previous study reported that AMPKα is not required for radial migration in the developing neocortex (Williams et al., 2011). The authors electroporated Cre recombinase with GFP into E15.5 neocortices of AMPKα1−/−;AMPKα2F/F mice to knock out both AMPKα1 and AMPKα2, dissected brain slices immediately, and analyzed the distribution and morphology of the AMPKα1/α2-deficient cortical neurons after 5 days in culture. Because ∼40% of the GFP-labeled neurons were observed into the CP, the authors suggested that AMPK is not required for neuronal migration. There was no statistical analysis for the neuronal distribution but it seems that less GFP-labeled AMPKα1/α2-deficient neurons were present in the CP, and concomitantly more neurons were in the IZ/VZ when compared with the distribution of control GFP-neurons. Our study shows that AMPKα depletion slows down but does not arrest neuronal migration and that AMPKα-depleted neurons reach the CP by early postnatal stage when AMPKα is knocked down at E14 (6.2% at E17, 28.7% at E18 and >90% at P4) (Fig. 3, Fig. S8). Hence, Williams et al. may have not detected migration defects because they analyzed the samples too late, i.e. after 5 days in culture, at a time point when a specific population of delayed migrating neurons have already caught up and reached the CP.
AMPK-mediated Ser81 phosphorylation of DIC contributes to dynein functions
Cytoplasmic dynein is a large protein complex with dynein heavy chain forming the motor domain and DIC acting as a scaffolding protein. Using lysates from embryonic brain, Cos-7 cells and primary neurons, endogenous and exogenous DIC, phosphatase and AMPK inhibitor treatment, we provide much evidence that DIC associates with and is phosphorylated by AMPK (Fig. 6). Furthermore, we also demonstrate that AMPK directly phosphorylates DIC in vitro (Fig. 7). Using recombinant phospho-dead mutants, we have also identified Ser81 as the main site of phosphorylation by AMPK in vivo (Fig. 8).
Importantly, we demonstrate that Ser81 phosphorylation of DIC by AMPK positively regulates dynein functions. Transfection of phospho-resistant DIC S81A mutant into Cos-7 cells disrupts the normal perinuclear enrichment of the Golgi apparatus, which is a marker for dynein function (Fig. 8, Fig. S10). The N-terminal region of DIC contains a coiled-coil domain involved in binding to accessary proteins such as p150glued and Ndel1/Nde1, whereas the C-terminal region is composed of binding sites for other dynein subunits. Ser81 resides in a serine-rich region that is adjacent to the N-terminal coiled-coil domain (see Fig. 7B). Several lines of evidence suggest that dynein functions are modulated by the phosphorylation of DIC at Ser/Thr residues in the serine-rich region. In fact, Ser81 phosphorylation of DIC promotes dynein binding to and dynein-mediated transport of signaling endosomes such as Trk-containing signaling endosomes (Ginty and Segal, 2002; Mitchell et al., 2012). In addition, phosphorylation of Ser88/Thr89 has been shown to cause dissociation of Ndel1, a potent dynein force generator, thereby reducing dynein motility (Gao et al., 2015). Thr89 phosphorylation can also promote dynein binding to the kinetochore protein ZW-10, for dynein localization on kinetochore (Whyte et al., 2008), rather than to p150glued, a mediator for interaction of dynein with various organelles/cargos (Schroer, 2004). In addition, live-cell imaging of dynein composed of Ser84 phospho-mimetic/phospho-resistant mutants of DIC suggests that phosphorylation at DIC Ser84 modulates dynein motor activity (Blasier et al., 2014). Thus, phosphorylation within the serine-rich domain positively or negatively regulates dynein behaviors through stimulating/inhibiting dynein motor activity and/or the association of dynein with organelles/cargos.
In this context, we hypothesize that Ser81 phosphorylation of DIC promotes dynein motility and/or the association of dynein with its targets to promote neuronal migration. In support of this view, we found the association between DIC and p150glued to be altered by Ser81 mutations (S81A and S81D) in DIC, as assessed by a co-immunoprecipitation assay using cell lysates derived from COS-7 cells transfected with DIC wild type, DIC S81A and DIC S81D (Fig. S11). When compared with DIC wild type, more p150glued is co-immunoprecipitated with DIC S81A and, conversely, less p150glued is found with DIC S81D in the immunoprecipitates. These differential associations between proteins of the motor complex may represent a potential mechanism to explain the AMPK-dependent regulation of dynein behavior.
AMPK regulates neuronal migration through Ser81 phosphorylation of DIC
We found that depletion of AMPKα or overexpression of phospho-resistant DIC S81A significantly retarded neuronal migration. Conversely, overexpression of the phospho-mimetic DIC S81D mutant ameliorated migration defects caused by AMPK depletion. Together with our findings of direct phosphorylation of DIC by AMPK at Ser81 and modulation of dynein behaviors by the Ser81 phosphorylation (see above), we conclude that AMPK contributes to neuronal migration via Ser81-phosphorylation of DIC and positive regulation of dynein functions.
In this context, we propose a model in which AMPK colocalizes with dynein, associates with and phosphorylates DIC, potentiates dynein functions, and boosts the pulling forces on the nucleus and centrosome to direct proper nuclear and centrosomal forward movement. Along this line, we have noticed that nuclei of AMPKα-depleted neurons and DIC S81A-introduced neurons show a rather relatively rounded (rather than distorted) shape when compared with that of control cells (Figs 3 and 9). This finding is reminiscent of the shape of nuclei in dynein heavy chain-depleted neurons and, therefore, supports the idea that disruption of the AMPKα-dynein axis results in less pulling forces on the nucleus. In a non-mutually exclusive scenario, mispositioning of organelles (e.g. Golgi apparatus, see Fig. S8) caused by dysregulation of the AMPKα-dynein pathway may be an alternative mechanism to explain the delayed migration of AMPKα-depleted neurons (Bellion et al., 2005; Nishimura et al., 2014).
In conclusion, our study provides novel mechanistic insights into the regulation of dynein function during neuronal migration. We identified AMPK as a regulator for DIC through Ser81 phosphorylation during centrosomal and nuclear movements, both crucial for proper neuronal migration. Our findings may help understand the molecular basis of neuronal migration disorders.
MATERIALS AND METHODS
Animals
ICR mice were purchased from SLC (Hamamatsu, Japan) and were housed under 12 h light-12 h dark cycle. All animal experiments were conducted in accordance with guidelines set by The University of Tokyo and approved (permit number 21-01) by the Committee on Animal Care and Use of the Graduate School of Science in The University of Tokyo.
Plasmids
The pCAGEN, pCAG-IRES-GFP (pCAGIG) and pCAG-CFP plasmids were kindly provided by Dr Takahiko Matsuda (University of Hyogo, Japan). The pBS/U6 plasmid and an RNAi plasmid for dynein heavy chain were kind gifts from Dr Yang Shi (Harvard Medical School, MA, USA) and Dr Li-Huei Tsai (Massachusetts Institute of Technology, MA, USA), respectively. The full-length open reading frame (ORF) of Ampkα1, Ampkα2 and Dync1I2 were amplified by PCR from cDNA derived from the E16 mouse cortex and cloned into the pCAGEN vector that directs subcloned gene expression from the CAG promoter. The AMPKα1 and AMPKα2 shRNA plasmids were generated by inserting the annealed oligonucleotides into the pBS/U6 plasmid (Sui et al., 2002), as described previously (Sanada and Tsai, 2005). Target sequence for RNAi were as follows: AMPKα1 shRNA, 5′-GGCACACCCTGGATGAATTAA-3′; AMPKα1 shRNAi#2, 5′-GGGACTGCTACTCCACAGAGA-3′; AMPKα2 shRNA, 5′-GGTAGACAGTCGGAGCTATCT-3′; and AMPKα2 shRNA#2, 5′-GGATGACAGCGCCATGCATAT-3′. Plasmids encoding silent mutants of AMPKα, kinase-dead mutants of AMPKα, as well as phospho-resistant and phospho-mimetic mutants of DIC were generated using the QuickChange mutagenesis technique. Constitutively active mutants of AMPKα2 [C-terminal truncated form of AMPKα2 (amino acids, 1-312), with Thr172 changed to Glu] were generated using QuickChange mutagenesis followed by subcloning of the 1-312 amino acids region into the pCAGEN vector. pCAG-Histone H2B-GFP plasmids were generated by subcloning histone H2B-GFP from the pBOS-H2BGFP vector (BD Biosciences) into the pCAGEN vector. For bacterial expression, Dync1i2 was subcloned into the pGEX-4T2 vector that expresses N-terminal GST-tagged dynein IC2B.
Cell culture and transfection
Cos-7 cells and HEK293 cells (purchased from the European Collection of Authenticated Cell Cultures, ECACC) were maintained in DMEM with 10% fetal bovine serum (FBS) and antibiotics. Plasmids were transfected into cells using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Transfected cells were then cultured in 10% FBS/DMEM for 24-48 h prior to being subjected to immunoblotting, immunocytochemistry and immunoprecipitation.
Primary cortical neurons were prepared as described previously (Asada et al., 2007). Plasmids were transfected into primary neurons using a Nucleofector device (Amaxa), according to the manufacturer's instructions. In some experiments, Cos-7 cells and primary cortical neurons were treated with A769662 (final concentration 100 µM; AdooQ Bioscience), AICAR (2 mM; Wako) and compound C (20 µM; Calbiochem) 1 h before being subjected to immunoblotting and immunocytochemistry.
For immunoblotting and co-immunoprecipitation assay, Cos-7 cells and cultured cortical neurons were lysed in lysis buffer [20 mM HEPES, 100 mM NaCl, 1.0% TritonX-100, 10 mM NaF, 2 mM Na3VO4, and protease inhibitor cocktail (Complete, EDTA-free; Roche Molecular Biochemicals); pH 7.4]. After a 30 min incubation on ice, the cell lysates were centrifuged for 10 min at 20,000 g, and the resultant supernatant was collected as cell extracts.
Immunoblotting
Proteins separated by SDS-PAGE and Phos-tag SDS-PAGE were transferred to PVDF membranes (Merck Millipore). Phos-tag SDS-PAGE was performed in typical SDS-PAGE gels (6%) supplemented with 20 µM Phos-tag Acrylamide (Wako; AAL-107) and 80 µM MnCl2. The blots were blocked with 5% (w/v) skim milk or 3% (w/v) BSA in TBS (10 mM Tris-HCl, 140 mM NaCl; pH 7.6) at room temperature for 1 h, and then incubated with primary antibodies in 3% BSA in TBS overnight at 4°C, followed by incubation with secondary antibodies overnight at 4°C.
Primary antibodies used were rabbit anti-AMPKα (1:1000; 2532, Cell Signaling Technology), rabbit anti-pT172-AMPKα [40H9] (1:1000; 2535, Cell Signaling Technology), rabbit anti-AMPKα1 [Y365] (1:1000; ab32047, Abcam), rabbit anti-AMPKα2 (1:1000; ab3760, Abcam), rabbit anti-phospho-ACC (Ser79) [D7D11] (1:1000; 4150, Cell Signaling Technology), rabbit anti-ACC [C83B10] (1:1000; 3676, Cell Signaling Technology), mouse anti-dynein IC [74.1] (1:3000; sc-13524, Santa Cruz Biotechnology), rabbit anti-dynein HC (1:1000; R-325, sc-9115, Santa Cruz Biotechnology), mouse anti-p150glued (1:2000; 610473, BD Biosciences), goat anti-LIS1 (1:200; N-19, sc-7577, Santa Cruz Biotechnology), rabbit anti-phosphoserine (1:1000; 61-8100, Invitrogen), mouse anti-Flag (1:1000; F1804, Sigma), mouse anti-DYKDDDDK (1:10,000; 014-22383, Wako), rabbit anti-HA (1:2000; 561, MBL), mouse anti-β-actin (1:20,000; A1978, Sigma) and rabbit anti-GFP (1:1000; A11122, Invitrogen). The secondary antibodies used were HRP (horseradish peroxidase)-conjugated anti-rabbit IgG antibody (1:10,000; NA934, GE Healthcare), HRP-conjugated anti-mouse IgG antibody (1:10,000; NA931, GE Healthcare) and HRP-conjugated anti-goat IgG antibody (1:5000; 14-13-06, Kirkegaard and Perry Laboratories).
Co-immunoprecipitation assays
E17 telencephalons (four hemispheres) were isolated and incubated in HBSS containing trypsin and DNase I for 20 min at 37°C, and then triturated with a fire-polished Pasteur pipet, followed by rinsing three times with HBSS. The telencephalic cells were lysed with the lysis buffer as above.
Cell extracts from E17 telencephalon and Cos-7 cells were precleared by incubation with 20 µl of Protein G-Sepharose for 30 min, and thereafter incubated with anti-dynein IC antibody [74.1] (5 µg) or control IgG (5 µg; Sigma) for 24 h at 4°C, followed by incubation with 20 µl of Protein G-Sepharose for 1 h. The beads were washed three times with lysis buffer and subjected to immunoblotting.
Lysates of HEK293 cells transfected with or without plasmids expressing Flag-AMPKα2 and dynein IC-HA were prepared 48 h after transfection and subjected to a co-immunoprecipitation assay with anti-Flag antibody (2 µg) or anti-HA antibody (2 µg).
Immunocytochemistry
For immunostaining, cells were fixed with 4% paraformaldehyde in PBS for 15 min at 37°C, permeabilized with 0.1% Triton X-100 in PBS for 5 min and blocked with PBS containing 3% BSA and 0.2% Triton X-100, followed by incubation with primary antibodies diluted in the blocking solution overnight at 4°C. The cells were then incubated with Alexa 488/Cy3/Cy5-conjugated secondary antibodies. For immunostaining with anti-dynein IC [74.1] antibody, cells were fixed with methanol for 10 min at −20°C and were subjected to immunostaining as above. For measurements of the size of the Golgi apparatus, the outer edge of the region where GM130 signals are clustered was outlined. In addition, the outer edge of cells was outlined based on the GFP image. Then the outlined area size was calculated using ZEN software (Carl Zeiss Microimaging).
Primary antibodies used were rabbit anti-AMPKα (1:100; 2532, Cell Signaling Technology), rabbit anti-pT172-AMPKα [40H9] (1:100; 2535, Cell Signaling Technology), mouse anti-dynein IC [74.1] (1:50; sc-13524, Santa Cruz Biotechnology), goat anti-LIS1 (1:200; N-19, sc-7577, Santa Cruz Biotechnology), rat anti-GFP (1:1000; 04434-34, Nacalai Tesque), mouse anti-pericentrin (1:200; 611814, BD Biosciences), mouse anti γ-tubulin [GTU-88] (1:200; T6557, Sigma), mouse anti-Tuj1 antibody (1:3000; MMS-435P, Covance), mouse anti-GM130 (1:100; 610822, BD Biosciences) and mouse anti-α-tubulin (1:1000; T5168, Sigma).
In utero electroporation
DNA solution (∼1 µl) in PBS containing 0.01% Fast Green was injected into the lateral ventricle of mouse embryos, followed by electroporation. The electric pulses were generated by CUY21-EDIT (Nepa gene) and applied to the cerebral wall at five repeats of 42 V for 50 ms with an interval of 950 ms with forceps-type electrodes (CUY650P5, Nepa gene). The GFP-expressing plasmid was co-injected with the shRNA constructs. Final concentrations of the plasmids used are as follows: pCAGIG (2 µg/µl), AMPKα shRNAs (1 µg/µl for Fig. 9E, 2 µg/µl for Fig. 3C and Fig. S4C, and 5 µg/µl for the others), DsRed-Centrin2 (5 µg/µl), AMPKα1/AMPKα2 with a silent mutation (5 µg/µl), kinase-dead AMPKα (5 µg/µl) and wild-type/mutant DIC (2 µg/µl for Fig. 9E, 10 µg/µl for Fig. 9A,C, Fig. S10).
For time-lapse imaging of migratory neurons, the CFP-expressing plasmid (2.5 µg/µl, pCAG-CFP), the Histone H2B-GFP expressing plasmid (1 µg/µl) and the DsRed-Centrin2-expressing plasmid (5 µg/µl) were co-injected with the shRNA constructs (5 µg/µl).
Immunohistochemistry
Brains were fixed with 4% paraformaldehyde in PBS overnight at 4°C. Brain sections (30 µm for immunohistochemistry with anti-pAMPKα antibody, 60 µm for the others) were prepared with a vibratome (Leica), blocked with PBS containing 3% BSA and 0.2% Triton X-100, and then subjected to immunohistochemistry as described previously (Asada et al., 2007).
Primary antibodies used were rabbit anti-AMPKα (1:50; 2532, Cell Signaling Technology), rabbit anti-pT172-AMPKα [40H9] (1:50; 2535, Cell Signaling Technology), rabbit anti-GFP (1:1000; A11122, Invitrogen), rat anti-GFP (1:1000; 04434-34, Nacalai Tesque), rabbit anti-HA (1:1000; 3724, Cell Signaling Technology), and mouse anti-Pericentrin (1:200; 611814, BD Biosciences) and rabbit anti-cleaved caspase 3 (1:500, 9661, Cell Signaling Technology).
Preparation of acute brain slices and time-lapse imaging
Acute brain slices were prepared as described previously (Sanada and Tsai, 2005; Asada and Sanada, 2010) with slight modifications. Electroporated brains were dissected from E17 embryos in ice-cold DMEM/F12 supplemented with D-glucose (final concentration, 6.05 g/l) (dissection media) that was oxygenized and conditioned with 95% O2/5% CO2 for 20 min on ice. The slices (300 µm) were prepared with a vibratome in the ice-cold and preconditioned dissection media. The slices were then transferred onto 35 mm glass-bottom dishes (Matsunami Glass), and were overlaid with collagen (Cellmatrix IA, type I Collagen; Nitta Gelatin) for immobilization. The dishes were then placed in a CO2 incubator for 10-20 min for gel solidification, and 1 ml of prewarmed culture media (Neurobasal medium supplemented with B27, 5% horse serum and 5% FBS) was added into the dish, followed by incubation for 1 h in a CO2 incubator.
Fluorescently labeled neurons were viewed under 5% CO2/95% air at 37°C through a 40× oil-immersion objective (NA 1.0) of the Axio observer inverted microscope with a PM S1 incubator (Carl Zeiss Microimaging). Time-lapse images were collected with an AxioCam cooled CCD camera (Carl Zeiss Microimaging) every 20 min for 7-12 h.
Protein kinase assay
Bacterially expressed GST-Dynein IC2B was purified by using a glutathione-Sepharose column (GE Healthcare) according to the manufacturer's protocol. Recombinant GST-Dynein IC2B was subjected to SDS-PAGE analysis with Coomassie Brilliant Blue staining, and the protein content was estimated by densitometry of band intensities in the SDS-PAGE gel in comparison with those of serial dilutions of BSA standards in the same gel. Active AMPK complex (AMPK α1/β1/γ2) were purchased from Signalchem (P55-10H). Recombinant bacterially expressed GST-Dynein IC2B (300 ng) was incubated with or without 100 ng of the recombinant AMPK in 20 µl of kinase reaction buffer (20 mM HEPES, 10 mM MgCl2, 1 mM DTT, 0.2 mM AMP, 1 mM ATP; pH 7.4) at 30°C for the indicated time. The reaction was started by addition of ATP solution and stopped by SDS-PAGE loading buffer.
Mass spectrometry analysis
Recombinant bacterially expressed GST-DIC (3 µg) was phosphorylated with 1 µg of recombinant active AMPK in vitro (see above) followed by subjecting to SDS-PAGE. The GST-DIC band was excised out from the gel and was subjected to in-gel digestion with modified trypsin (sequencing grade, Promega). Mass spectrometry analysis of digested peptides was carried out by Genomine. In brief, the tryptic peptides extracted from the gel were separated and analyzed using reverse-phase capillary HPLC directly coupled to a Finnigan LCQ ion trap mass spectrometer. The individual spectra from MS/MS were processed using the TurboSEQUEST software (Thermo Quest). The generated peak list files were used to query either MSDB database or NCBI using the MASCOT program (www.matrixscience.com). Only significant hits, as defined by MASCOT probability analysis, were considered.
Statistical analysis
All bar graphs in migration assays were plotted as mean±s.e.m. (n=3-5 embryos from two or three pregnant dams). Direct comparisons were made using two-tailed Welch's t-test using Microsoft Excel. The significance level was set at P<0.05. Sample sizes used in the present study are similar to those generally employed in the field. The experiments were randomized, and data collection and analyses were performed blind to the experimental condition.
Acknowledgements
We thank Dr Takahiko Matsuda, Dr Yang Shi and Dr Li-Huei Tsai for plasmids.
Footnotes
Author contributions
Conceptualization: Y.N., N.A., M.D.N., K.S.; Methodology: Y.N., K.S.; Validation: Y.N.; Formal analysis: Y.N.; Investigation: Y.N., N.A., K.S.; Writing - original draft: Y.N.; Writing - review & editing: M.D.N., K.S.; Visualization: Y.N.; Funding acquisition: Y.N., M.D.N., K.S.
Funding
This work was supported in part by Grants-in-Aid for Scientific Research (C) (17K07045 and 20K06868 to K.S.) from the Japan Society for the Promotion of Science, by Scientific Research on Innovative Areas (19H04769 to K.S.) from the Ministry of Education, Culture, Sports, Science and Technology of Japan, and by the Canadian Institutes of Health Research (M.D.N.). Y.N. was supported by Japan Society for the Promotion of Science Research Fellowships for Young Scientists.
Peer review history
The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.187310.reviewer-comments.pdf
References
Competing interests
The authors declare no competing or financial interests.