|Home | About | Journals | Submit | Contact Us | Français|
In migrating cells, external signals polarize the microtubule (MT) cytoskeleton by stimulating the formation of oriented, stabilized MTs and inducing the reorientation of the MT organizing center (MTOC). Glycogen synthase kinase 3β (GSK3β) has been implicated in each of these processes, although whether it regulates both processes in a single system and how its activity is regulated are unclear. We examined these issues in wound-edge, serum-starved NIH 3T3 fibroblasts where MT stabilization and MTOC reorientation are triggered by lysophosphatidic acid (LPA), but are regulated independently by distinct Rho GTPase-signaling pathways. In the absence of other treatments, the GSK3β inhibitors, LiCl or SB216763, induced the formation of stable MTs, but not MTOC reorientation, in starved fibroblasts. Overexpression of GSK3β in starved fibroblasts inhibited LPA-induced stable MTs without inhibiting MTOC reorientation. Analysis of factors involved in stable MT formation (Rho, mDia, and EB1) showed that GSK3β functioned upstream of EB1, but downstream of Rho-mDia. mDia was both necessary and sufficient for inducing stable MTs and for up-regulating GSK3β phosphorylation on Ser9, an inhibitory site. mDia appears to regulate GSK3β through novel class PKCs because PKC inhibitors and dominant negative constructs of novel PKC isoforms prevented phosphorylation of GSK3β Ser9 and stable MT formation. Novel PKCs also interacted with mDia in vivo and in vitro. These results identify a new activity for the formin mDia in regulating GSK3β through novel PKCs and implicate novel PKCs as new factors in the MT stabilization pathway.
Microtubule (MT) arrays become polarized to perform their functions in development, differentiation, and cell migration. One of the major ways MT arrays become polarized is through the generation of a subset of stabilized MTs (Gundersen and Bulinski, 1988 ; Bulinski and Gundersen, 1991 ; Gundersen et al., 2004 ). In contrast to dynamic MTs, selectively stabilized MTs display longer half-lives and increased resistance to MT-depolymerizing drugs (Webster et al., 1987 ; Khawaja et al., 1988 ). Selectively stabilized MTs are abundant in differentiated cells, but are not restricted to differentiated cells: in wounded monolayers of fibroblasts, stable MTs form selectively so that they orient toward the leading edge (Gundersen and Bulinski, 1986 ; Gundersen and Bulinski, 1988 ; Palazzo et al., 2004 ). Selective MT stabilization is thought to occur by the formation of a plus-end cap that prevents the MT from growing or shrinking (Infante et al., 2000 ), and factors implicated in selective MT stabilization are localized at the ends of these MTs (Wen et al., 2004 ). Stabilized MTs accumulate posttranslationally modified tubulin, including detyrosinated and acetylated tubulin, and this may contribute to the recognition of stabilized MTs by factors that selectively interact with modified tubulin, such as kinesin motors (Liao and Gundersen, 1998 ). The functions of these stabilized, detyrosinated MTs in polarized cells are beginning to be understood, and studies have highlighted their importance in vimentin intermediate filament extension (Gurland and Gundersen, 1995 ; Kreitzer et al., 1999 ), endosomal recycling (Lin et al., 2002 ), migration (Wen et al., 2004 ), and virus infection (Naranatt et al., 2005 ). In addition, mice lacking the detyrosination/tyrosination cycle die shortly after birth and have neuronal defects (Erck et al., 2005 ).
In mammalian fibroblasts, the formation of long-lived stable MTs is regulated by the serum factor lysophosphatidic acid (LPA), which stimulates a signaling pathway that involves the heterotrimeric G protein, Gα12/13, the small GTPase Rho, its effector, the formin mDia, and the MT plus-end binding proteins EB1 and APC (Cook et al., 1998 ; Palazzo et al., 2001a ; Wen et al., 2004 ; Goulimari et al., 2005 ). Rho and mDia also appear to be involved in MT stabilization in embryonic endodermal cells, where they function along with the plakin, ACF-7 (Kodama et al., 2003 ). Binding of activated Rho to mDia is thought to relieve mDia's autoinhibited state and lead to its activation (Alberts, 2001 ). Activated forms of mDia bind both EB1 and APC, and all three proteins can be found on the ends of stable MTs, consistent with a capping mechanism (Wen et al., 2004 ). MTs stabilized by this Rho-mDia signaling pathway do not grow or shrink (Cook et al., 1998 ; Palazzo et al., 2001a ; Wen et al., 2004 ), making this form of stabilization different from that of MAPs, which stabilize MTs by binding along their length, but do not block tubulin exchange at the ends.
The MT cytoskeleton is also polarized by positioning the MT organizing center (MTOC) at a specific place in the cell. In migrating fibroblasts, astrocytes, macrophages, and endothelial cells, the MTOC is reoriented so that it lies between the nucleus and the leading edge (Gotlieb et al., 1981 ; Kupfer et al., 1982 , 1983 ; Gundersen and Bulinski, 1988 ; Etienne-Manneville and Hall, 2001 ). Cdc42, Par6, atypical PKC (aPKC), and dynein comprise a common “MTOC reorientation” pathway in fibroblasts, astrocytes, and endothelial cells (Etienne-Manneville and Hall, 2001 , 2003 ; Palazzo et al., 2001b ; Tzima et al., 2003 ; Gomes et al., 2005 ). In fibroblasts, this pathway, like that regulating MT stabilization, is triggered by LPA (Palazzo et al., 2001b ). In fibroblasts, it is known that LPA triggers a second Cdc42-regulated pathway that functions to move the nucleus rearward, whereas the Par6-aPKC-dynein pathway maintains the MTOC at the cell centroid. The combined effects of the two pathways are necessary for MTOC reorientation (Gomes et al., 2005 ). The second Cdc42-regulated pathway involves the Cdc42 effector, MRCK (myotonic dystrophy kinase-related Cdc42-binding kinase), myosin and actin retrograde flow, which is thought to drive the nucleus rearward (Gomes et al., 2005 ).
GSK3β was named for its role in regulating glycogen synthase, although it is now known to participate in many signaling pathways and cellular activities. GSK3β is usually constitutively active and most signaling pathways result in the inhibition of GSK3β. For example, insulin stimulates protein kinase B (PKB) to phosphorylate Ser9 (pSer9) of GSK3β and this inhibits its activity (Cross et al., 1995 ). Insulin treatment of starved 3T3 fibroblasts does promote a small increase in the levels of stable MTs (Gundersen et al., 1994 ), suggesting that regulation of GSK3β activity may influence stable MT formation. In the Cdc42 pathway involved in MTOC reorientation in astrocytes, activation of atypical PKC also increases GSK3β pSer9 levels (Etienne-Manneville and Hall, 2003 ), but whether GSK3β participates in MTOC reorientation in fibroblasts or other systems that orient their MTOC has not been tested.
LPA is another cytokine that causes the inhibition of GSK3β by phosphorylation on Ser9, and although PKCs seem to be involved, the specific isoform(s) has not been identified (Fang et al., 2002 ). Because LPA stimulates both MT stabilization and MTOC reorientation in starved NIH 3T3 fibroblasts, we were interested in whether the inhibition of GSK3β was important for the effects of LPA on either or both of these MT-polarizing events. Given the evidence that GSK3β is involved in MTOC reorientation in astrocytes, we predicted that it would also be involved in both processes in fibroblasts. However, we find that GSK3β is not involved in MTOC reorientation in fibroblasts, but instead is involved in MT stabilization. Because LPA-induced MT stabilization in fibroblasts is regulated by a Rho-mDia pathway, this suggested that there might be a novel pathway involved in regulating GSK3β in response to LPA stimulation. Indeed, we find that GSK3β pSer9 is stimulated by both LPA and active forms of mDia1 or mDia2 and that mDia1 was both necessary and sufficient for enhancing GSK3β pSer9. mDia1 appears to regulate GSK3β phosphorylation through a pathway involving PKCε, a novel PKC isoform, because mDia1 directly binds PKCε, and PKCε was necessary for both the LPA- and mDia-stimulated phosphorylation of GSK3β and for formation of stable MTs. These results describe a novel activity for a formin in regulating kinase activity and show that in fibroblasts, GSK3β contributes to the Rho-mDia-EB1-APC pathway that regulates MT stabilization.
All materials were purchased from Sigma (St. Louis, MO) unless otherwise indicated. NIH 3T3 cells were maintained as previously described (Cook et al., 1998 ; Kreitzer et al., 1999 ; Palazzo et al., 2001a ). For microinjection, confluent NIH 3T3 cells were serum starved for 2 d, wounded, and microinjected as previously described (Cook et al., 1998 ; Kreitzer et al., 1999 ; Palazzo et al., 2001a ). Plasmid DNA was microinjected in H-KCl (10 mM HEPES, pH 7.4, 140 mM KCl) at 100–200 μg/ml and allowed to express 2–4 h after microinjection. To induce stable MTs and MTOC reorientation, 5 μM LPA was added for 90–120 min. GSK3β inhibitors LiCl (20 mM) or SB216763 (Tocris, Ballwin, MO) were added to starved cells and incubated for 2 h. SB216763 was used at 2–10 μM with similar results.
All HA-tagged PKC constructs were a gift from J. W. Soh (Soh et al., 1999 ), and flag-tagged dominant negative PKCλ was from C. Carpenter (Coghlan et al., 2000 ). Full-length HA-Rokα was from L. Lim (Leung et al., 1996 ). mDia1 and mDia2 constructs were from A. Alberts (Alberts, 2001 ), except for pEGFP-C1-521-1040-mDia2, which was made by subcloning 521-1040-mDia2 from pEFm-EGFP (Wen et al., 2004 ) into pEGFP-C1 (Clontech, Palo Alto, CA). HA-tagged GSK3β was constructed by PCR amplification of pcDNA3-GSK3β (a gift from J. Kitajewski) with primers 5′-GGCGGCGTCGACAGAAGAGCCATCATG-3′ and 3′-GCCTCCAACTCTACCTCTAGAGCCCCC-5′ and cloned into pKHA3 (a gift from D. Brautigen). To prepare GFP-GSK3β, GSK3β was subcloned from pKHA3 into pEGFP-C1 (Clontech). The GFP-GSK3β fusion was then subcloned into pCAGGS/ES (a gift from P. Scheiffle) downstream of the chicken β-actin promoter. All constructs were verified by sequencing, and DNA was purified with Qiagen midiprep kits (Chatsworth, CA).
COS-7 cells were transfected with mDia1 siRNA in a six-well dish with 4 μl of 20 μM siRNA and 10 μl of Lipofectamine 2000 (Invitrogen, Carlsbad, CA) in 1 ml of growth media (DMEM with 10% FBS). The mDia1 siRNA sequence was 5′-GCUGGUCAGAGCCAUGGAU-3′ (Arakawa et al., 2003 ) and the control GFP sequence was 5′-GGCUACGUCCAGGAGCGCACC-3′. Both siRNAs were from Dharmacon (Boulder, CO). After 24 h, the medium was changed to serum-free DMEM for another 24 h. Cells were lysed and analyzed for GSK3β phosphorylation. Similar knockdown results in COS-7 cells were obtained with a second mDia1 siRNA constructed using the Ambion Silencer siRNA construction kit (Austin, TX). This mDia1 sequence was 5′-AAGGUGAAGGAGGACCGCUUU-3′. The GAPDH siRNA used was the control provided with the kit. For microinjection of siRNA into NIH 3T3 fibroblasts, confluent monolayers were serum-starved and wounded. A few hours later, Ambion kit-synthesized siRNAs were diluted in H-KCl and microinjected at 10 μM with 1 mg/ml 500 kDa FITC-dextran (Molecular Probes, Eugene, OR) as an injection marker. After microinjection (48 h), cells were fixed for 5–10 min in methanol at −20°C and processed for immunofluorescence. To visualize individual siRNA-treated cells, the Ambion kit-synthesized siRNA was labeled with CX-rhodamine using the Mirrus Label IT siRNA Tracker kit (Boston, MA) according to the manufacturer's instructions. NIH 3T3 fibroblasts passaged at least three times in FBS were transfected with 2 μl Mirrus TKO Transfection reagent and 200 nM siRNA in 300 μl of growth media in a 24-well plate. After transfection (24 h), cells were replated onto coverslips and allowed to adhere another 24 h. Cells were then starved for 24 h, treated with LPA, and fixed and processed for immunofluorescence as above. To examine GSK3β pSer9 levels (see below) in siRNA treated cells, NIH 3T3 cells were transfected with siRNA to mDia1 or GAPDH (Invitrogen) with Lipofectamine Plus. Cells were split 24 h after transfection and transfected a second time 48 h after transfection. After the second transfection, cells were washed, serum-starved overnight, treated with LPA or insulin, and processed for immunofluorescence (see below).
Staining of fixed cells was performed as previously described (Gundersen et al., 1984 ). Glu MTs were detected with a polyclonal rabbit antibody (Ab) SG that specifically recognizes detyrosinated (Glu) tubulin (Gundersen et al., 1984 ; 1:400), and tyrosinated (Tyr) MTs were detected with the rat mAb YL1/2 (1:10, European Collection of Animal Cell Cultures, Salisbury, United Kingdom). The following Abs were used for detection of overexpressed or endogenous proteins; mouse mAb to HA (12CA5, 1:100, Roche, Indianapolis, IN), mouse mAb to GSK3β (0011-A, 1:100, Santa Cruz Biotechnology, Santa Cruz, CA), mouse mAb to mDia1 (clone 51, 1:100, BD Transduction, Lexington, KY), and mouse mAb to GAPDH (6C5, 1:1000, Ambion). For certain experiments, cells were also stained with mouse mAb to actin (C4D6, 1:200, a gift from J. Lessard, University of Cincinnati), β-tubulin (3F3, 1:200), or γ-tubulin (GTU-88, 1:200, Sigma). For localizing F-actin, cells were fixed in 4% paraformaldehyde for 10 min, permeabilized with 0.5% Triton X-100 for 5 min, and stained with rhodamine-phalloidin (1:500, Molecular Probes). Digital images were acquired as previously described (Palazzo et al., 2001a ).
For staining with GSK3β pSer9 Ab (1:100, Cell Signaling Technology, Beverly, MA), cells were fixed with 2% paraformaldehyde + 0.2% glutaraldehyde in PBS for 20 min and then permeabilized with 0.2% Triton X-100 in PBS for 10 min. Cells were then blocked in 2% BSA in TBS and then stained as previously described, except primary antibody incubations were for 1 h. To quantify GSK3β pSer9 in individual cells, digital images were acquired with equivalent exposure times as previously described (Palazzo et al., 2001a ). The average fluorescence intensity of the entire cell was measured using MetaMorph software (Universal Imaging, West Chester, PA) on a per cell basis after subtracting the background intensity (measured in an adjacent cell-free area of the coverslip). We attempted to correct the fluorescence measurements for nonspecific staining of the secondary antibody; however, these readings were too variable to give consistent results between the experiments and so were not corrected for.
COS-7 cells were transfected in growth medium (DMEM with 10% CS) with Lipofectamine 2000 according to the manufacturer's specifications. After transfection (6–24 h), cells were washed twice with serum-free medium (SFM; DMEM + 10 mM HEPES) and incubated in SFM for an addition 18–24 h. Cells were treated with 10 μM LPA or 10–100 μg/ml insulin for 5 min. To inhibit PKC, 10 μM bisindolylmaleimide I (Calbiochem, LaJolla, CA) was added 1 h before treatment with LPA or insulin. Cells were washed once with cold PBS and lysed in 1% Triton X-100, 150 mM NaCl, 20 mM Tris, pH 7.4 or 8.0, phosphatase inhibitors (5 mM β-glycerophosphate, 1 mM sodium orthovanadate, 2.5 mM sodium pyrophosphate), and Protease Inhibitor Cocktail (Sigma). After 10 min extraction on ice, lysates were clarified by centrifugation (13,000 × g, 10 min, 4°C). Protein concentration was determined with a BCA protein assay and equal amounts of protein were run on SDS-PAGE, transferred to nitrocellulose and analyzed by Western blotting with rabbit Ab to GSK3β pSer9 (1:5000, Cell Signaling Technology), mouse mAb GSK3β (0011-A, 1:5000, Santa Cruz), mouse mAb to mDia1 (clone 51, 1:1000, BD Transduction), rat mAb to HA (3F10, 1:5000, Roche), and mouse mAb to vinculin (VIN-11–5, 1:5000, Sigma).
COS-7 cells were transfected with myc-GFP-ΔGBD-mDia1 or myc-GFP as a control along with pHACE-PKCε, pHACE-PKCη, pHACE-PKCζ, or HA-ROKα plasmids using Lipofectamine 2000 according to the manufacturer's protocol. After 24 h transfection, cells were washed once with cold PBS and lysed in RIPA buffer (1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, 50 mM Tris, pH 7.4, 150 mM NaCl, and phosphatase and protease inhibitors). After 10-min extraction on ice, lysates were clarified by centrifugation (13,000 × g, 10 min, 4°C). To each extract, rabbit Ab to full-length GFP (1:500, BD Transduction) and protein A-Sepharose beads were added, and lysates were mixed overnight at 4°C. Beads were washed in RIPA buffer, eluted with SDS sample buffer and loaded on gels. Expressed proteins were detected by Western blotting with mouse mAb to myc (9E10, 1:5000, Santa Cruz) and rat mAb to HA (3F10, 1:5000, Roche). For endogenous binding, 5 mg of precleared mouse brain extract was immunoprecipitated with 20 μg rabbit anti-PKCε (Santa Cruz) or control rabbit IgG overnight at 4°C in lysis buffer containing 1% TX-100, 150 mM NaCl, 20 mM Tris, pH 7.4, and phosphatase and protease inhibitors. Protein G-Sepharose beads were then added and mixed for an additional 3 h. Beads were washed with lysis buffer, eluted with SDS sample buffer, and analyzed by SDS-PAGE followed by Western blotting. Antibodies used for blotting were mouse mAb to mDia1 (1:1000, BD Transduction) and rabbit Ab to PKCε (1:1000, Santa Cruz). For direct in vitro binding, 20 μg GST-521-1040-mDia2 or an equal molar amount of GST was bound to glutathione-agarose beads for 1–2 h at 4°C. Then, 6 pmol recombinant purified human PKCε or PKCζ (Calbiochem) was added in lysis buffer with or without lipids (100 μg/ml phosphatidylserine and 10 μg/ml diacylglycerol; Avanti, Aalabaster, AL) and mixed overnight at 4°C. Beads were washed with lysis buffer, and bound proteins were eluted with SDS sample buffer and run on gels. Bound proteins were detected by Western blotting with rabbit Ab to PKCε (1:2000, Santa Cruz), rabbit Ab to PKCζ (1:2000, Upstate Biotechnology, Lake Placid, NY), or mouse mAb to GST (1:20,000, Cell Signaling Technology). For blot overlay analysis, equimolar amounts of PKCε or PKCζ were subject to SDS-PAGE and transferred to nitrocellulose. The membrane was incubated overnight at 4°C with 10 μg/ml GST-521-1040-mDia2 or an equimolar amount of GST in TBST containing 5% BSA, 100 μg/ml phosphatidylserine and 10 μg/ml diacylglycerol. The bound GST-mDia2 or GST was detected by mouse Ab to GST, and quantified using the LI-COR Odyssey Infrared Imaging System (Cincinnati, OH).
We first tested whether GSK3β was involved in the LPA-stimulation of MT stabilization and/or MTOC reorientation in wound-edge serum-starved NIH 3T3 fibroblasts. For MT stabilization, we assayed the formation of stabilized MTs by immunofluorescently staining cells for post-translationally detyrosinated tubulin (or Glu tubulin after its C-terminus), which accumulates in MTs after they are stabilized (Webster et al., 1987 ; Khawaja et al., 1988 ). We then scored cells for their content of Glu MTs. Cells containing five or more brightly stained, extended Glu MTs were counted as positive for Glu MTs. For MTOC reorientation, we assessed the position of the MTOC (detected with antibodies to γ-tubulin) relative to the leading edge and nucleus, scoring MTOCs as reoriented if they were in a pie-shaped segment facing the wound; random orientation of the MTOC gives a value 33% “reoriented” by this method (see Palazzo et al., 2001b ).
Because LPA stimulates inhibition of GSK3β (Fang et al., 2002 ; and see below), we first tested whether known inhibitors of GSK3β, LiCl, or SB216763 mimicked the effect of LPA treatment of starved fibroblasts. When starved fibroblasts were treated with either GSK3β inhibitor, the level of Glu MTs increased without significantly affecting the levels of the bulk, dynamic MTs detected with Tyr tubulin antibody (Figure 1, A and B). As with LPA treatment, Glu MTs induced by the GSK3β inhibitors constituted a subset of the total MTs. However, unlike those triggered by LPA, they were not preferentially oriented toward the wound edge (Figure 1A). Glu MTs induced by GSK3β inhibitors were not noticeably bundled, indicating that MAP-like proteins were not involved. SB216763 treatment of starved NIH 3T3 fibroblasts also promoted an increase in nocodazole-resistant MTs (Figure 1C), indicating that inhibition of GSK3β affects MT stability and not the enzymes responsible for tubulin post-translational modifications. These results show that inhibition of GSK3β is sufficient to trigger the formation of stabilized MTs in fibroblasts.
To test whether inhibition of GSK3β was necessary for stable MT formation, serum-starved fibroblasts were microinjected with DNA encoding full-length GSK3β and after expression (1–2 h), treated with LPA to induce stable MTs. Overexpression of wild-type GSK3β inhibited the formation of Glu MTs induced by LPA (Figure 1, A and B). These results demonstrate that GSK3β is a negative regulator of LPA-stimulated MT stability.
To test whether GSK3β was involved in MTOC reorientation in fibroblasts, we treated wounded, starved NIH 3T3 fibroblast monolayers with SB216763 in the absence or presence of LPA to promote MTOC reorientation. SB216763 did not induce MTOC reorientation in starved cells and did not inhibit MTOC reorientation induced by LPA (Figure 2, A and B). To extend this finding, starved NIH 3T3 fibroblasts were microinjected with wild-type GSK3β in the presence or absence of LPA to examine whether overexpression of GSK3β inhibited MTOC reorientation. Expression of wild-type GSK3β, which inhibited MT stabilization induced by LPA (Figures 1, A and B, and and2A),2A), did not inhibit MTOC reorientation induced by LPA or promote MTOC reorientation on its own (Figure 2, A and B). Thus, in starved NIH 3T3 fibroblasts, GSK3β is involved in regulating MT stabilization but not MTOC reorientation. This is consistent with our previous studies showing that factors working in one of the MT regulatory pathways in fibroblasts do not function in the other pathway (Palazzo et al., 2001b ; Wen et al., 2004 ).
To examine whether GSK3β functions with the known components of the MT stabilization pathway or alternatively acts through a parallel pathway, we tested whether inhibition of GSK3β was able to promote MT stabilization in the presence of known inhibitors of the MT stabilization pathway. Botulinum C3 toxin, which specifically inhibits Rho (Aktories et al., 1989 ), blocks LPA-stimulated formation of stable MTs (Cook et al., 1998 ). Microinjection of C3 toxin did not prevent the induction of stable MTs by LiCl (Figure 3, A and B), but did reduce the levels of actin fibers, demonstrating that the injected protein was functional. The MT tip protein EB1 acts downstream of both Rho and mDia in the stabilization pathway, and the C-terminal fragment of EB1 (EB1-C) inhibits stable MTs induced by LPA or mDia (Wen et al., 2004 ). Expression of EB1-C inhibited the induction of stable MTs by LiCl (Figure 3, A and B).
These results suggest that GSK3β functions in the MT stabilization pathway and acts downstream of Rho but upstream of EB1. As a further test of this, we explored whether GSK3β inhibition in starved cells stimulated the formation of actin filaments. If GSK3β acts downstream of Rho and mDia, it should not affect the actin cytoskeleton. Indeed, addition of LiCl to serum-starved cells did not promote formation of actin stress fibers, compared with LPA (Figure 3C), which activates Rho and mDia and induces stress fiber formation (Ridley and Hall, 1992 ; Watanabe et al., 1997 ). This result confirms that GSK3β functions downstream of Rho but upstream of EB1 in the MT stabilization pathway and also suggests that stable MTs formed by the inhibition of GSK3β do not require actin stress fibers, consistent with earlier studies (Palazzo et al., 2001a ; Wen et al., 2004 ).
LPA treatment of starved Swiss 3T3 fibroblasts was shown earlier to increase GSK3β Ser9 phosphorylation, although the mechanism for this was not fully delineated (Fang et al., 2002 ). We confirmed that LPA treatment of starved COS-7 cells or NIH 3T3 fibroblasts increased GSK3β pSer9 levels (Figure 4, A and B). In both cases, the levels of GSK3β pSer9 triggered by LPA was similar to that induced by insulin, which stimulates GSK3β pSer9 through a pathway involving PKB/Akt (Cross et al., 1995 ).
The cellular factors involved in LPA-induced GSK3β pSer9 are unknown. Given that GSK3β appears to function downstream of Rho-mDia (but upstream of EB1) in the MT stabilization pathway, upstream factors such as mDia are potential candidates for mediating the effect of LPA on GSK3β. The GSK3β pSer9 levels were assessed in COS-7 cells transiently transfected with HA-tagged GSK3β with or without constitutively active versions of mDia (ΔGBD-mDia1 or ΔGBD-mDia2). By examining the pSer9 levels of HA-tagged GSK3β, which was distinguishable from the endogenous GSK3β by a difference in migration on SDS gels (see Figure 4A), we could specifically determine the effect of mDia in the cotransfected cells. Expression of either active mDia1 or mDia2 increased HA-GSK3β pSer9 levels (Figure 4A). The levels of HA-GSK3β pSer9 stimulated by active mDia1 or mDia2 were about half of that stimulated by LPA. Because ~50% of the cells were cotransfected (unpublished data), these results suggest that activation of mDia alone may be sufficient to account for the LPA-stimulated increase in GSK3β pSer9 levels.
Next, we determined whether activation of endogenous mDia1 was sufficient to stimulate endogenous GSK3β phosphorylation on Ser9. NIH 3T3 cells were transfected with GFP-tagged Dia autoregulatory domain (GFP-DAD), a fragment of mDia1 that is able to activate the endogenous protein by relieving autoinhibition (Alberts, 2001 ). Transfection with DAD was sufficient to promote phosphorylation of endogenous GSK3β in NIH 3T3 cells, whereas the control transfection with GFP alone had no effect (Figure 4B). In addition, expression of GFP-521-1040-mDia2, which contains the FH1 and FH2 domains and is the minimal domain of mDia that is sufficient to stimulate stable MT formation (Wen et al., 2004 ), was also sufficient to promote GSK3β pSer9 in NIH3T3 cells (Figure 4B).
To test whether endogenous mDia1 was required for the LPA-stimulated increase in GSK3β pSer9 levels, we knocked down mDia1 with siRNA. COS-7 cells were transfected with HA-GSK3β as before, in combination with siRNA specific for mDia1 or control siRNA specific for GFP. mDia1 siRNA decreased endogenous mDia1 by ~75%, whereas the control siRNA against GFP did not affect mDia1 levels (Figure 4C). mDia1 siRNA, but not GFP siRNA, reduced LPA-stimulated HA-GSK3β pSer9 by ~50%, but did not affect basal HA-GSK3β pSer9 levels in unstimulated cells or the induction of HA-GSK3β pSer9 in cells stimulated with insulin (Figure 4C). These data demonstrate that mDia1 is responsible for at least part of the enhanced GSK3β pSer9 levels induced by LPA and that mDia1 does not participate in insulin-stimulated regulation of GSK3β, which occurs through the PKB/Akt pathway (Cross et al., 1995 ).
To quantify the effect of mDia1 knockdown on endogenous Ser9 phosphorylation of GSK3β in NIH 3T3 cells, a quantitative immunofluorescence approach was taken because the transfection efficiency was not sufficient to permit a Western blot assay. NIH 3T3 were transfected with siRNA to mDia1, or siRNA to GAPDH as a control, serum-starved, and then treated with LPA or insulin. Cells were fixed and stained for GSK3β pSer9, and mDia1 or GAPDH. The average fluorescence intensity per cell was measured in cells stained for GSK3β pSer9 that had visible mDia1 (or GAPDH) knockdown (for examples, see Figure 4E). This assay showed a consistent 30–50% increase in GSK3β pSer9 fluorescence in LPA- and insulin-treated cells, demonstrating the validity of this assay (Figure 4D). Knockdown of mDia1, but not GAPDH, resulted in a significant decrease in LPA-stimulated GSK3β pSer9 (Figure 4D). There was also a decrease in insulin-stimulated GSK3β pSer9 in mDia1 siRNA-treated cells, but this decrease was less significant compared with LPA (Figure 4D). These data demonstrate that mDia1 is required for at least some of the enhanced GSK3β-pSer9 levels induced by LPA.
Previous studies identified mDia as the likely Rho effector for LPA-mediated MT stabilization (Palazzo et al., 2001a ). Given that mDia1 is required for LPA-induced increase in GSK3β pSer9 and that GSK3β appeared to be involved in the MT stabilization pathway, it was important to confirm that mDia was necessary for MT stabilization induced by LPA as previously reported (Goulimari et al., 2005 ). Wound-edge starved NIH 3T3 fibroblasts, which express mDia1 but not mDia2 (Tominaga et al., 2000 ; and unpublished data), were microinjected with siRNA against mDia1 that we had validated in COS-7 cells (Figure 4C). Two days after mDia1 siRNA treatment or GAPDH siRNA as a control, the cells were treated with LPA to stimulate formation of stable MTs. The majority of cells injected with siRNA had reduced levels of mDia1 or GAPDH as detected by immunofluorescence, confirming the efficacy of the knockdown in NIH 3T3 fibroblasts (Figure 4E). In cells with visible knockdown of mDia1, there was a decrease in the level of Glu MTs in response to LPA compared with uninjected cells or GAPDH siRNA injected cells (Figure 4E). Quantification of this effect in cells transfected with labeled siRNA (to identify the transfected cells) showed that mDia1 knockdown significantly reduced the induction of Glu MTs by LPA compared with untransfected or siRNA GAPDH controls (Figure 4F).
LPA-stimulated GSK3β pSer9 is blocked by PKC inhibitors (Fang et al., 2002 ; Yang et al., 2005 ), and we confirmed this result in NIH 3T3 and COS-7 cells, where treatment with the general PKC inhibitor, bisindolylmaleimide I, strongly inhibited LPA-stimulated, but not insulin-stimulated, GSK3β pSer9 (Figure 5, A and B). Next, we tested whether mDia-stimulated GSK3β pSer9 also required PKC. COS-7 cells transfected with HA-GSK3β were pretreated with bisindolylmaleimide I and then treated with LPA or insulin or cotransfected with constitutively active ΔGBD-mDia2, and their levels of GSK3β pSer9 were assayed. GSK3β pSer9 induced by LPA or active mDia2 was inhibited by bisindolylmaleimide I, indicating that LPA/mDia-induced GSK3β pSer9 required PKC(s) (Figure 5B). Insulin-stimulated GSK3β pSer9 was less affected, as expected (Figure 5B).
There are three classes of PKC isoforms: conventional, novel, and atypical. We tested whether a specific class of PKC was involved in the increase in GSK3β pSer9 and stable MTs triggered by mDia1. Initially, we screened a set of dominant negative PKC isoforms for effects on LPA-induced stable MT formation. Expression of either dominant negative PKCε or PKCη, both novel isoforms of PKC, prevented LPA-induced stable MTs in starved NIH 3T3 fibroblasts (Figure 5, C and D). In contrast, expression of dominant negative conventional (α, β1, γ) or atypical (ζ, λ) PKC isoforms had no effect on LPA-induced stable MTs (Figure 5, C and D). Cells expressing dominant negative PKCε or PKCη still retained their ability to reorient their MTOC in response to LPA (Figure 5E), again confirming that proteins involved in MT stabilization do not affect MTOC reorientation. Expression of atypical PKCζ or λ isoforms interfered with LPA-induced MTOC reorientation (Figure 5C and unpublished data), as previously reported (Etienne-Manneville and Hall, 2001 ; Gomes et al., 2005 ), confirming the activity of these proteins.
To test whether novel PKCs were sufficient to promote stable MTs, plasmids encoding the isolated catalytic (cat) domain were expressed in starved cells. Expression of PKCε-cat or PKCη-cat did not induce stable MTs in the absence of LPA treatment (unpublished data). Thus, although novel PKCs are required for the formation of stable MTs induced by LPA, their catalytic activity alone is not sufficient.
To check whether the novel PKCs affected GSK3β pSer9, we analyzed the ability of the dominant negative constructs to inhibit LPA- and mDia-induced GSK3β pSer9. Coexpression of dominant negative PKCε in COS-7 cells with GFP-GSK3β as a reporter construct inhibited the increase in GSK3β pSer9 levels in response to LPA and ΔGBD-mDia1, but not insulin (Figure 5F). Thus, dominant negative PKCε that inhibits MT stabilization also inhibits GSK3β phosphorylation.
To quantify the effect of dominant negative PKCε on endogenous GSK3β pSer9, we again performed quantitative immunofluorescence to measure GSK3β pSer9. COS-7 cells were transfected with dominant negative PKCε, GFP-ΔGBD-mDia1, or both, serum-starved, and then treated with LPA or insulin as indicated. Cells were fixed and stained, and individual cells expressing one or both proteins were measured for their levels of GSK3β pSer9. Again, this assay showed a consistent increase in GSK3β pSer9 fluorescence in LPA- and insulin-treated cells, similar to what was observed in NIH 3T3 cells (Figure 5G). In addition, transfection with active mDia1 significantly increased endogenous GSK3β pSer9 over levels in untreated cells, as had been observed for transfected GSK3β. Expression of PKCε-KR significantly reduced the ΔGBD-mDia1 and LPA stimulation of GSK3β pSer9 to background levels (Figure 5G). PKCε-KR also decreased GSK3β pSer9 stimulation due to insulin, but this decrease was not significant and did not reduce the levels to those of untreated cells (Figure 5G).
The above results raise the possibility that mDia acts in conjunction with novel PKC isoforms to phosphorylate and inhibit GSK3β. To explore whether this might be a direct effect of mDia on novel PKCs, we tested whether the two proteins interacted in vivo. Mouse brain extract was immunoprecipitated with antibody to PKCε. A small amount of mDia1 was visualized in the immunoprecipitate, demonstrating that PKCε and mDia1 interact at physiological levels (Figure 6A). Because we found that only novel PKC isoforms specifically affected stable microtubules, we examined whether other isoforms of PKC were capable of binding to mDia. Tagged versions of active mDia1 and full-length PKC isoforms were transfected into COS-7 cells and immunoprecipitated. Immunoprecipitation of myc-GFP-ΔGBD-mDia1, but not myc-GFP alone, coimmunoprecipitated the novel PKC isoforms, PKCη and PKCε, and also the atypical PKC isoform, PKCζ (Figure 6B). Another HA-tagged kinase, ROKα, did not coimmunoprecipitate with mDia1. Thus, active mDia1 is capable of binding multiple PKC isoforms in vivo when overexpressed.
To examine whether the interaction between mDia and PKCs is direct, the binding of purified recombinant PKCε or PKCζ with a GST-tagged active fragment of mDia2 was examined by a blot overlay assay. Increasing equimolar amounts of PKCε or PKCζ immobilized on nitrocellulose were incubated with 10 μg/ml GST-521-1040-mDia2 (Figure 6C). Before the overlay, the membrane was Ponceau-stained to demonstrate equal loading. Both PKCε and PKCζ bound to GST-521-1040-mDia2, although PKCε consistently bound more GST-mDia2 than did PKCζ (Figure 6C). At the highest concentration, PKCε bound 1.8-fold more than PKCζ (±0.14) to GST-mDia2, with similar differences in binding at lower concentrations. Binding of both PKC isoforms to GST was minimal compared with GST-521-1040-mDia2, and correcting for the GST binding did not alter the higher level of binding observed with PKCε compared with PKCζ. Novel PKCs are characterized by the ability to be activated by lipids, but not calcium; thus we wanted to analyze the lipid dependence of mDia binding to PKCε. The interaction of PKCε with GST-521-1040-mDia2 was increased in the presence of the PKC activating lipids phosphatidylserine and diacylglycerol (Figure 6D). Binding of PKCζ, which does not require diacylglycerol for activation, was not affected (Figure 6D). Thus, both PKCε and PKCζ bind mDia2 directly, but PKCε has a higher affinity for mDia2 than PKCζ, and the binding of PKCε is also enhanced by the presence of lipids.
We demonstrate that GSK3β functions and is phosphorylated in the Rho-mDia mediated MT stabilization pathway, acting downstream of Rho and upstream of EB1. GSK3β is negatively regulated by signals that are sufficient to promote MT stabilization, because knockdown of both mDia1 and dominant negative novel PKCs block MT stabilization and GSK3β phosphorylation. These data provide a novel role for a formin protein in regulating the action of a kinase, although earlier work showed mDia could interact with Src kinase (Tominaga et al., 2000 ). Recent studies of formins in mammalian cells have focused on their effects on the actin and MT cytoskeletons (reviewed in Wallar and Alberts, 2003 ; Zigmond, 2004 ). Here, we demonstrate that LPA, through mDia1, can regulate the activity of GSK3β by controlling its phosphorylation.
Our results point to novel PKCs as the likely candidates for mediating the effects of mDia on GSK3β. We found that dominant negative novel PKCs blocked active mDia induced GSK3β pSer9 and that mDia interacted directly with PKCε. Although mDia also bound to another PKC isoform, PKCζ, the affinity of PKCε was higher for mDia and could account for the specific activity of novel PKCs toward stable MTs. Furthermore, although the binding of PKCε to mDia was enhanced in the presence of lipids, the binding of PKCζ was not. These results suggest that binding of active mDia to novel PKCs may regulate their activity toward substrates such as GSK3β. Previous work showed that novel PKCs can phosphorylate GSK3β on Ser9 (Fang et al., 2002 ). We suggest that mDia may regulate the activity or localization of novel PKCs to enhance their activity toward GSK3β. This LPA-stimulated pathway involving mDia and novel PKC regulation of GSK3β functions independently of insulin-PKB regulation of GSK3β.
This work identifies two new components, novel PKCs and GSK3β that function in the Rho-mDia-EB1-APC pathway, which regulates plus-end mediated stabilization of MTs. Our work points to an new role for novel PKCs in governing MT stabilization. Novel PKCs are characterized by their dependence on lipids for activation, but they do not require calcium. Novel PKCs (δ, θ, ε, and η) are widely and differentially expressed in tissues and each isoform has unique functions. We do not know how mDia binding to PKC may regulate its activity. Perhaps the binding of PKCε to mDia brings it into the vicinity of MTs, where it is sometimes localized in vivo (Kato et al., 2001 ; Palazzo et al., 2001a ; Wen et al., 2004 ). PKCε is also localized to the actin cytoskeleton in some cells (Prekeris et al., 1996 ; Hernandez et al., 2001 ), and it is also possible that mDia regulates its activity toward the actin cytoskeleton.
GSK3β is the first negative regulator to be identified in the Rho-mDia-EB1-APC pathway for MT stabilization. A number of earlier studies have used inhibitors to show that GSK3β regulates MT stability (Goold et al., 1999 ; Krylova et al., 2000 ; Akhmanova et al., 2001 ). Our work extends these studies by showing how a known MT regulatory pathway controls GSK3β phosphorylation in response to a physiological stimulus. Most previous studies have implicated GSK3β in regulating MT stability through its affects on MAPs or CLASPs binding laterally to MTs (Lovestone et al., 1996 ; Wagner et al., 1996 ; Goold et al., 1999 ; Krylova et al., 2000 ; Akhmanova et al., 2001 ). Our results indicate that GSK3β also regulates MT stability through its action on factors involved in MT capture at the cell cortex. Glu MTs are a stabilized subset of MTs that are capped at their distal ends, preventing subunit addition or loss and their ends are found in close proximity to the cell cortex (Infante et al., 2000 ; Wen et al., 2004 ). Direct interaction of mDia with the MT tip proteins EB1 and APC contributes to this end-mediated MT stabilization (Wen et al., 2004 ). The interaction of these proteins with each other or with MTs are potential targets for regulation by GSK3β. The only known substrate of GSK3β among these proteins is APC, and GSK3β phosphorylation of APC decreases its interaction with MTs in vitro (Zumbrunn et al., 2001 ). APC interaction with EB1 is also regulated by phosphorylation, although this is by other kinases (Askham et al., 2000 ). It will be interesting to test whether other interactions among these proteins and with MTs are regulated by GSK3β phosphorylation. GSK3β also appears to negatively regulate the interaction of CLASPs with microtubules (Akhmanova et al., 2001 ; Wittmann and Waterman-Storer, 2005 ), so this is another candidate target for GSK3β.
The observation that inhibition of GSK3β induces stable MTs that, in contrast to LPA-stimulated stable MTs, are not polarized toward the leading edge, is likely due to the global inactivation caused by GSK3β inhibitors and the fact that GSK3β functions downstream of Rho. Selective stabilization of MTs oriented toward the wound edge occurs because of integrin engagement in the front of the cell, leading to activation of FAK (Palazzo et al., 2004 ). This adhesion signal is thought to restrict the ability of active Rho to stimulate mDia to the region near the leading edge. Consistent with this idea, overexpression of active Rho promotes the formation of a polarized array of stable MTs, but overexpression of constitutively active mDia does not (Cook et al., 1998 ; Palazzo et al., 2001a ). GSK3β bypasses the adhesion signal that polarizes the newly formed stable MTs by acting downstream of mDia. Indeed, we have found that GSK3β inhibition by LiCl is able to promote MT stabilization in nonadherent fibroblasts that lack integrin signals (Palazzo and Gundersen, unpublished observations).
The MT array is also polarized in migrating cells by reorientation of the MTOC to a position between the nucleus and the leading edge of the cell (Gotlieb et al., 1981 ; Kupfer et al., 1982 ; Gundersen and Bulinski, 1988 ). We previously found that inhibition of Cdc42, dynein, or dynactin, proteins involved in MTOC reorientation, does not block MT stabilization and that inhibition of Rho or EB1, proteins involved in MT stabilization, does not block MTOC reorientation (Palazzo et al., 2001b ; Wen et al., 2004 ). We find that GSK3β and novel PKCs function in the MT stabilization pathway, which is consistent with these results, but not in the MTOC reorientation pathway in NIH 3T3 fibroblasts and that atypical PKCs function in MTOC reorientation, as reported previously (Etienne-Manneville and Hall, 2003 ; Gomes et al., 2005 ), but not in MT stabilization. That inhibition of GSK3β is both necessary and sufficient for MT stabilization, but not for MTOC reorientation, is in contrast to results in wounded astrocytes, where GSK3β overexpression or inhibition has been found to inhibit MTOC reorientation (Etienne-Manneville and Hall, 2003 ). A role for GSK3β in MT stabilization in astrocytes has not yet been tested. We do not know the reason for the different effects of GSK3β in the two systems, but they may be related to differences in cell types, conditions used to stimulate MTOC reorientation, or to the much longer time to reorient MTOCs in astrocytes (6 h) compared with fibroblasts (1 h).
Despite the differences between the responses of different cell types to GSK3β, it is clear that GSK3β is an important regulator of microtubule polarity, and thus a key determinant of polarity during development and differentiation. GSK3β has a crucial role in Wnt signaling involving APC, Dvl, Axin, and β-catenin, which mediates embryonic polarity. Recent studies in neurons have also pointed to the significance of GSK3β in the establishment of axons and maintenance of neuronal polarity via APC, PI3K, and CRMP-2 (Shi et al., 2004 ; Jiang et al., 2005 ; Yoshimura et al., 2005 ). Whether the establishment of polarity by GSK3β in these situations requires the formation of stable microtubules will be an interesting question to pursue.
We thank J. W. Soh, A. Alberts, D. Brautigan, J. Kitwajewski, P. Scheiffle, C. Carpenter, and F. Bartolini for DNA constructs; Y. Wen for preparation of GST protein; and members of the Gundersen lab for comments on the manuscript. C.H.E. was supported by a Howard Hughes Medical Institute Predoctoral Fellowship. This work was supported by National Institutes of Health Grant GM 62939 (G.G.G.).
This article was published online ahead of print in MBC in Press (http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E05-10-0914) on September 20, 2006.