Mitogen-activated protein kinase (MAPK/ERK) regulates adenomatous polyposis coli during growth-factor-induced cell extension

Coordinated organization of microtubule and actin networks in response to extracellular signals is crucial for cell polarization, the formation of cell extensions and initiation of cell migration. Adenomatous polyposis coli (APC) plays an important role in regulating the cytoskeleton during cell migration (Aoki and Taketo, 2007). APC accumulates at the plus end of microtubule bundles in actively extending membranes (Mimori-Kiyosue et al., 2000; Näthke et al., 1996; Votin et al., 2005; Zhou et al., 2004), stimulates microtubule assembly and bundling in vitro (Zumbrunn et al., 2001) and promotes microtubule growth and rescue in vivo (Kita et al., 2006; Reilein and Nelson, 2005). APC also associates with actin structures (Harris and Nelson, 2010; Rosin-Arbesfeld et al., 2001) and decorates the tip of microtubules that track along actin bundles in neuronal growth cones (Zhou et al., 2004).

Regulation of these functions of APC has focused on APC phosphorylation by glycogen synthase kinase-3β (GSK-3β), which reduces the interaction between APC and microtubules in vitro (Zumbrunn et al., 2001). Local inhibition of GSK-3β downstream of nerve growth factor (NGF) activation of neurons and rat adrenal medulla pheochromocytoma (PC12) cells is essential for APC localization, microtubule stabilization and further cell extension (Mills et al., 2003; Zhou et al., 2004). It is generally assumed, therefore, that APC phosphorylation decreases upon NGF-mediated cell extension, and that this regulates the function of APC in stabilizing microtubules during cell extension (Barth et al., 2008). Surprisingly, however, the phosphorylation state of APC has not been examined directly. Here, we show that early growth factor signaling in PC12 cells results in APC phosphorylation by the MAPK/ERK pathway, which in vitro decreases the interaction of APC with actin. Moreover, we show that ERK activity and the ERK-activating protein IQGAP3 are required for efficient localization of APC at the tip of cell extensions. We suggest that this pathway acts in parallel with inhibition of GSK-3β to coordinate APC-mediated re-organization of the actin and microtubule cytoskeletons at cell extensions.

NGF-induced cell extension in PC12 cells requires APC (Dobashi et al., 2000) and involves inhibition of GSK-3β (Jin et al., 2007), similarly to axonal outgrowth in ex vivo neuronal cultures and directional cell migration in other cell types (Etienne-Manneville and Hall, 2003; Zhou et al., 2004). We have shown previously that APC clusters localize to the tip of neurite extensions after prolonged (1–2 day) NGF signaling (Votin et al., 2005). Here, for the first time, we examined the regulation of APC during the first few hours of NGF and epidermal growth factor (EGF) signaling and formation of initial cell extensions, and characterize the immediate effects of these growth factors on APC protein modification and localization.

In untreated cells, small puncta of APC localized diffusely throughout the cell body. However, within 4–6 hours of addition of NGF APC rapidly localized into clusters at the tip of very short initial cell extensions [Fig. 1A; see also quantification in Fig. 2B: 30.21±3.06% of cells with APC clusters at 4 hours NGF, compared with 1.23±0.04% in untreated (Ctrl) cells; P=0.0007]. APC also localized in clusters at the tip of EGF-induced cell extensions (supplementary material Fig. S1A,B; 35.80±1.52% cells with APC clusters at 4 hours EGF compared with 1.23±0.04% in Ctrl cells; P=0.00002), indicating that APC relocalization is regulated by a pathway common to different growth factors.

Western blots of solubilized APC revealed two closely migrating bands (Fig. 1B). In the absence of NGF (time 0, Fig. 1B) the faster migrating APC band was enriched relative to the slower migrating band, but upon NGF stimulation the amount of the slower migrating APC increased at the apparent expense of the faster migrating band (Fig. 1B,C; 47.77±0.34% of slower migrating APC band out of total APC at 4 hours of NGF treatment, compared with 37.15±2.69% at time 0; P=0.008). Treatment of cells with EGF induced a similar change in the electrophoretic mobility of APC (supplementary material Fig. S1C,D; EGF, 40.92±1.53% slower migrating band compared with 26.5±1.0% slower migrating band for untreated cells; P=0.0002).

Comparison of NGF-induced activation of ERK (shown by ERK phosphorylation) with NGF-mediated APC mobility shift showed that maximal ERK activation preceded maximal APC modification (Fig. 1B–D). Maximal ERK activation occurred within 15 minutes of NGF treatment, after which it decreased and stayed slightly elevated for the duration of the time course (Fig. 1C,D, see also Fig. 3A,C). Similar time courses of NGF-induced ERK activation in PC12 cells have been published previously (Traverse et al., 1992; Santos et al., 2007). NGF-mediated APC mobility shift was strongest 1 hour after treatment with growth factor and persisted for the duration of the time course (Fig. 1B,C). Note that APC clustering increased significantly after 4–6 hours (Fig 1A), indicating that APC modification preceded APC tip clustering.

Extraction of APC from cells in the presence of the phosphatase inhibitor sodium vanadate (NaVan) had little or no effect on the electrophoretic mobility of APC, but treatment of NaVan-free cell extracts with λ-phosphatase abolished the appearance of the NGF-induced slower migrating APC band (Fig. 1E). Thus, the mobility shift of APC after growth factor addition is due to phosphorylation.

NGF and EGF activate several signaling pathways (supplementary material Fig. S2A) including the PI3K–AKT pathway which inhibits GSK-3β (Kleijn et al., 1998). Inhibition of GSK-3β with SB216763 (Coghlan et al., 2000) resulted in APC clustering in the tip of cell extensions, which was similar to that induced by NGF or EGF (Fig. 2A,B; 26.25±3.07% of SB-treated cells with APC clusters compared with 1.23±0.04% for Ctrl cells, P=0.001). Incubation with both a GSK-3β inhibitor and NGF had an additive effect on APC clustering at the tips of cell extensions, resulting in almost double the number of cells with clusters compared with those treated with either NGF or SB216763 alone [Fig. 2A,B; 51.76±2.41% of SB + NGF cells with APC clusters compared with 30.22±3.06% for NGF alone (P=0.005) and 26.25±3.06% for SB alone (P=0.003)].

Surprisingly, analysis of APC by western blotting showed that inhibition of GSK-3β with SB216763 in the absence of NGF did not cause the appearance of the slower migrating (phosphorylated) APC band (Fig. 2C,D; SB, 38.17±1.03% slower migrating band; control, 40.33±1.15% slower migrating band); similar results were obtained using GSK-3β inhibitor I (GSKI, Fig. 2C). That GSK-3β was inhibited by SB216763 was shown by a corresponding increase in β-catenin level (supplementary material Fig. S2D) (Hart et al., 1998). Addition of NGF in the presence of SB216763, however, induced the electrophoretic mobility shift of APC to the slower migrating, phosphorylated species (Fig. 2C,D; NGF+SB, 54.10±1.93% slower migrating band, compared with 40.33±1.15% for control cells; P=0.001).

In summary, these results show that: (1) APC is phosphorylated downstream of NGF and EGF; (2) GSK-3β inhibition can mediate some APC clustering independently of growth-factor-mediated APC phosphorylation; and (3) GSK-3β inhibitor and NGF have additive effects on APC clustering. Based on these new findings, we investigated the pathway involved in APC phosphorylation and whether APC phosphorylation has a role in promoting APC clustering.

To test signaling pathways downstream of NGF involved in APC phosphorylation, we analyzed the effect of different inhibitors on NGF-induced electrophoretic mobility shift of APC by SDS-PAGE (supplementary material Fig. S2). Inhibition of the NGF-receptor TrkA with K252a (Berg et al., 1992; Hashimoto, 1988) or addition of a blocking antibody against NGF inhibited NGF-mediated changes in APC protein electrophoretic mobility (supplementary material Fig. S2A–C). However, inhibitors of phosphatidylinositol 3-kinase (PI3K) (LY294002), protein kinase C (PKC) (Bisindolylmaleimide I), Jun N-terminal kinase (JNK) (SP600125), cyclin-dependant kinases (CDKs) (Roscovitine) and p38 MAPK (SB203580) did not block NGF-induced phosphorylation of APC (supplementary material Fig S2E–G).

U0126, a specific inhibitor of MEK (Duncia et al., 1998; Favata et al., 1998), which blocks MEK-mediated ERK phosphorylation (Fig. 3A,C), blocked the NGF-induced shift in APC electrophoretic mobility (Fig. 3A,B; 35.90±1.09% slower migrating band compared with 48.11±1.62% slower migrating band for NGF-treated cells; P=0.003). This result indicates that the ERK pathway mediates the APC phosphorylation induced by NGF and EGF.

To test whether ERK-mediated APC phosphorylation is required for efficient APC clustering at the tip of cell extensions, we analyzed APC clustering in cells treated with NGF or EGF in the presence of the MEK inhibitor. Inhibition of MEK activity reduced the percentage of cells that accumulated APC at the tip of NGF-induced cell extensions (Fig. 3D,E; 14.97±1.80% cells with APC clusters compared with 30.22±3.06% for NGF alone; P=0.01) and EGF-induced cell extensions (supplementary material Fig. S1A,B: 11.69±2.56% cells with APC clusters compared with 35.80±1.52% for EGF alone; P=0.001). Residual APC clustering in response to NGF or EGF when MEK was inhibited might be due to MEK-independent GSK-3β inhibition downstream of these growth factors (Fig. 3D,E and supplementary material S1A,B). However, combined inhibition of MEK and GSK-3β did not reduce the number of cells with APC clusters at the tip of cell extensions compared with that induced by inhibition of GSK-3β alone (34.7±15.28% compared with 26.26±3.07% for GSK inhibition alone; Fig. 3F), indicating that MEK activation and GSK-3β inhibition induce APC clustering independently.

We tested whether MEK-induced ERK activity is required for NGF-induced APC clustering by co-depleting ERK1 (MAPK3) and ERK2 (MAPK1) (Fig. 4). Immunofluorescence showed that NGF-induced phosphorylated ERK levels were reduced by 64% in cells co-depleted for ERK1 and ERK2 (ERK1/2) compared with control cells (Fig. 4A,B; ERK1/2-depleted cells 24±1.2 arbitrary fluorescence units, control cells 67±4.3 units). NGF-induced APC clustering at the tip of extensions was significantly reduced in those cells compared with that in cells transfected with control shRNA (24±3% ERK1/2-depleted cells with clusters and 54±1% control cells with clusters; P<0.008; Fig. 4C,D).

IQGAP3 mediates neurite extension in PC12 cells (Wang et al., 2007a) and promotes ERK activation (Nojima et al., 2008). Therefore, cells were depleted of IQGAP3 to test whether IQGAP3 is required for APC clustering during initial NGF-induced cell extension. In depleted cells, the level of IQGAP3 was reduced to 25±5.6% of the level in control cells (Fig. 5A,B). Depletion of IQGAP3 resulted in a significant decrease of APC clustering in cells treated for 4 hours with NGF (33±2% IQGAP3-depleted cells with clusters and 58±3% control cells with clusters; P<0.008; Fig. 5C,D).

The MEK inhibition and ERK depletion experiments indicate that APC localization and electrophoretic mobility shift are modified by MEK-mediated activation of ERK. To test directly whether APC is directly phosphorylated by ERK, we focused on the C-terminal 130 kDa of APC (GST–SAMP3end, Fig. 6A). SAMP3end has binding sites for microtubules and actin, and serine and threonine residues that are consensus sites for phosphorylation by ERK (S/TP) (Davis, 1993), some of which are phosphorylated during cell cycle progression and nuclear export (Dephoure et al., 2008; Olsen et al., 2006; Trzepacz et al., 1997) and in several mouse tissues (Huttlin et al. 2010).

In vitro phosphorylation of GST–SAMP3end with MEK1-activated ERK1 and ERK2 (Fig. 6B,C) revealed that ³²P was incorporated into GST–SAMP3end by activated ERK1 and ERK2, whereas ³²P was not incorporated into GST alone (Fig. 6D). MEK1 alone was unable to phosphorylate GST–SAMP3end and the MEK inhibitor PD0325901 had no effect on ERK2 phosphorylation of GST–SAMP3end (Fig. 6C), indicating that the phosphorylation is not caused by active MEK1 in the ERK2 preparation. Inhibition of MEK1 by PD0325901 completely blocked ERK2 phosphorylation, indicating that the inhibitor was specific and effective (Fig. 6E).

To identify sites in SAMP3end that are phosphorylated by ERK2, we performed LC-MS/MS. We identified ten ERK2 phosphorylation sites in the C-terminal region of APC (Fig. 6A; LC-MS/MS results in supplementary material Fig. S3), most of which localized in domains that are proposed to mediate cytoskeletal interactions (Deka et al., 1998; Moseley et al., 2007; Munemitsu et al., 1994; Zumbrunn et al., 2001).

Because the SAMP3end of APC contains microtubule and actin-binding sites, we tested whether phosphorylation of ERK affected binding of GST–SAMP3end to microtubules in a high-speed microtubule-pelleting assay. Both unphosphorylated and ERK-phosphorylated GST–SAMP3end co-pelleted with microtubules in a concentration-dependant manner (Fig. 7A,B). A small difference in microtubule binding of phosphorylated GST–SAMP3end was not an artifact of measuring levels of phosphorylated versus non-phosphorylated GST–SAMP 3end (supplementary material Fig S4A,B), but was also not statistically significant (Fig. 7A,B; 22.69±4.78% GST–SAMP 3end bound to 2 nM polymerized tubulin compared with 14.63±0.78% of ERK-phosphorylated GST–pSAMP3end; four experiments; P=0.55). Unphosphorylated and ERK2-phosphorylated GST–SAMP3end also co-pelleted with bundled microtubules and promoted microtubule bundling in a concentration-dependent manner, reaching a plateau at 25 nM GST–SAMP3end (Fig. 7C,D), further demonstrating that ERK2 phosphorylation does not significantly alter APC binding to or bundling of microtubules.

We analyzed whether ERK2 phosphorylation reduced APC binding to F-actin in a high-speed pelleting assay. GST–SAMP3end co-pelleted with pre-polymerized F-actin in a F-actin concentration-dependant manner (Fig. 8A,B). ERK phosphorylation, however, inhibited GST–SAMP3end binding to actin filaments (46.39±1.41% unphosphorylated GST-SAMP3end bound to 4 μM polymerized actin compared with 14.88±4.77% of phosphorylated GST–pSAMP3end; four experiments; P=0.003, Fig. 8A,B). In a low-speed pelleting assay, GST–SAMP3end bundled F-actin in a GST–SAMP3end concentration-dependent manner, reaching saturation at 50 nM GST–SAMP3end (Fig. 8C,D). However, ERK2-phosphorylated GST–SAMP3end protein did not promote F-actin bundling (Fig. 8C,D). Thus, ERK2 phosphorylation significantly decreases APC binding to and bundling of F-actin.

The prevailing model of APC regulation during the formation of cell extensions posits that decreased APC phosphorylation upon inhibition of GSK-3β results in increased localization of APC to the tip of cell extensions. Increased microtubule bundling by unphosphorylated APC then promotes the formation and stabilization of membrane extensions (Etienne-Manneville and Hall, 2003; Purro et al., 2008; Zhou et al., 2004).

We confirmed that inhibition of GSK-3β is part of the regulation of APC during growth-factor-induced APC clustering, and that it is sufficient to induce some APC clustering at the tip of cell extensions in the absence of growth factors. Although GSK-3β phosphorylation of APC decreased the electrophoretic mobility of APC (Harris and Nelson, 2010; Ikeda et al., 2000), we did not detect a change in electrophoretic mobility of APC after inhibition of GSK-3β. This indicates that in untreated cells, GSK-3β phosphorylation of APC might be transient and therefore difficult to detect as a change in the electrophoretic mobility of APC (Ikeda et al., 2000). Alternatively, APC might not be phosphorylated by GSK-3β, in which case, GSK-3β inhibition must induce APC clustering by a different mechanism from that involved in reducing APC phosphorylation at GSK-3β sites. Importantly, we showed that co-treatment of cells with the GSK-3β inhibitor and NGF had an additive effect on APC clustering, indicating that a GSK-3β-independent, NGF-responsive pathway is involved in APC cluster localization.

In the first direct analysis of APC phosphorylation during cell extension, we found increased APC phosphorylation upon addition of growth factors. The fact that either NGF or EGF caused this increase shows that the pathway involved is not specific for neuronal differentiation or NGF activation, but is a more general mechanism of cytoskeletal regulation during cell extension. Of several kinases downstream of NGF and EGF signaling, we found that ERK activation by MEK was responsible for NGF-induced APC phosphorylation and increased APC clusters in cell extensions. MEK inhibition reduced APC phosphorylation in response to NGF; and MEK inhibition or ERK1/2 co-depletion reduced APC clustering at the tips of extensions.

In addition, ERK1/2, but not MEK1, phosphorylated the APC C-terminus in vitro, indicating that ERK1/2 rather than MEK are the kinases that can directly phosphorylate APC after NGF-induced activation of this pathway (see model in Fig. 8E). This specific role for ERK brings together previous observations that activated ERK localizes along microtubules (Reszka et al., 1995) and at sites of cell extension (Wang et al., 2007b), is required for the formation of cell extensions (Atwal et al., 2000; Brahmbhatt and Klemke, 2003; Reszka et al., 1997; Webb et al., 2004) and regulates several microtubule-associated proteins (MAPs) and cell-substrate adhesion turnover at the leading edge (Drewes et al., 1992; Gotoh et al., 1990; Hoshi et al., 1992; Webb et al., 2004).

However, because ERK phosphorylation of APC precedes APC clustering, it might not be restricted to the cell edge, but could occur in other parts of the cell, with enough time for phosphorylated APC to accumulate at the cell tip.

It has been reported that inhibition of MEK and ERK does not affect APC localization to the tip of established axonal growth cones in fully polarized DRG neurons (Zhou et al., 2004). This inconsistency with our results might be due to differences between long-term axonal outgrowth and the formation of initial cell extension; for example, ERK phosphorylation of APC might be required for the initiation, but not the long-term maintenance of cell extensions.

IQGAP3 is required for neurite extension in PC12 cells (Wang et al., 2007a) and promotes ERK activation in proliferating mammary epithelial cells (Nojima et al., 2008). We showed that depletion of IQGAP3 reduced NGF-induced formation of APC clusters. Therefore, IQGAP3 might promote APC clustering at cell tips by facilitating phosphorylation of APC by the MEK–ERK pathway (see model in Fig. 8E), although a direct link between these pathways remains to be shown.

Mass spectrometry analysis identified ten ERK phosphorylation sites in the C-terminus of APC. These sites are located in or near functional domains of APC: (1) S2093 is at a putative secondary microtubule-binding site (Zumbrunn et al., 2001); (2) S2244, S2449, S2473, T2481, S2485 and S2535 are located in a region that mediates binding of microtubules and actin (Deka et al., 1998; Moseley et al., 2007; Munemitsu et al., 1994); (3) S2535 is part of a small domain ANS2 that is required for actin nucleation (Okada et al., 2010); (4) S2830 localizes very close to the end of the EB1-binding domain (Honnappa et al., 2005); (5) S2145 and S2774 are not located in a known functional domain of APC, but their proximity to the microtubule-, actin- and EB1-binding domains could affect interactions of APC with the cytoskeleton. Note that a high-throughput analysis of phospho-peptides generated during the cell cycle and upon EGF signaling identified S2093, S2473 and S2535 as phosphorylated sites in APC, although the kinase(s) involved was unknown (Dephoure et al., 2008; Olsen et al., 2006), and that S2093, S2449, S2473 and S2535 were also identified as phosphorylated sites in several mouse tissues including brain and kidney (Huttlin et al. 2010).

Phosphorylation of the APC C-terminus regulates binding to microtubules and actin filaments (Moseley et al., 2007; Zumbrunn et al., 2001), and to EB1 (Askham et al., 2000). We found that ERK phosphorylation of the APC C-terminus did not affect microtubule binding or bundling by APC, but inhibited APC binding and bundling of F-actin. Significantly, ERK phosphorylation sites are not present in the microtubule-binding domain of APC, but in one of the two characterized actin-binding domains (Okada et al., 2010). This might explain why ERK phosphorylation primarily affected F-actin, but not microtubule binding and bundling.

We have focused on ERK phosphorylation of the C-terminus of APC because this domain interacts directly with the actin and microtubule cytoskeletons. There are other domains in and outside the C-terminal region of APC that can mediate interactions with the cytoskeleton. For example, the armadillo repeat domain of APC interacts with the actin-binding protein IQGAP1 and kinesin superfamily-associated protein Kap3 (Watanabe et al., 2004; Jimbo et al., 2002), and it will be important in future studies to determine whether ERK phosphorylation regulates these interactions or the interaction of the APC C-terminus with the microtubule plus-end binding protein EB1 (Su et al., 1995).

We suggest that growth-factor-induced ERK phosphorylation of APC releases a pool of APC bound to F-actin and, thereby, increases the amount of APC that would be available to bind microtubules and accumulate at microtubule plus-ends at the tip of cell extensions (see schematic representation in Fig. 8E). Parallel inhibition of GSK-3β downstream of NGF or EGF signaling could increase binding of APC to microtubules and localization at the tip of extensions (Zhou et al., 2004; Zumbrunn et al., 2001). Reduced interactions between microtubules and actin through ERK phosphorylation of APC might enable microtubule bundles to penetrate the actin network at the cell periphery, and thereby stabilize cell extensions (Burnette et al., 2007; Sabry et al., 1991; Schaefer et al., 2002).

We emphasize that, at present, the specific sites in APC that are phosphorylated in vivo in response to growth factor signaling remain to be defined. Although a correlation between APC phosphorylation and localization is clear, the functional significance of phosphorylation sites in the C-terminus and in other, still untested, domains of APC remains to be determined. Site-directed mutagenesis of the in vitro defined phosphorylation sites, and ERK phosphorylation of additional APC domains, will be important areas for future investigations to identify whether all or only some of these sites are required for regulation of EGF-and NGF-induced formation of cell extensions.

PC12 cells were cultured as described previously (Votin et al., 2005). For all except the depletion experiments, cells were plated on plates or glass coverslips coated with rat-tail collagen type I for 1 day, which was changed to serum-free medium overnight. For inhibitor experiments, cells were treated for 4 hours with 50 ng/ml NGF (Upstate, Lake Placid, NY) or 100 ng/ml EGF (Sigma) unless otherwise stated. Inhibitors were added 15 minutes before addition of NGF or EGF, and incubated for the indicated times.

The following inhibitors were used: U0126 (10 μM; Promega, Madison, WI), SB216763 (20 μM; Sigma), LY294002 (40 μM; Biomol, Plymouth Meeting, PA); K252a (400 nM), Bisindolylmaleimide I (20 μM), Roscovitine (25 μM), SB203580 (25 μM), SP600125 (10 μM) and the GSK-3β inhibitor I (TDZD-8, 10 μM) were from Calbiochem (Gibbstown, NJ). Sheep anti-NGF blocking antibody was purchased from Chemicon (Temecula, CA).

For depletion of ERK1, ERK2 and IQGAP, PC12 cells were transfected with Qiagen SureSilencing shRNA/GFP plasmids (SABiosciences Corp., Frederick, MD) containing shRNA-encoding inserts for rat ERK1 (MAPK3, CAACCACATTCTAGGTATACT) and ERK2 (MAPK1, CCATGATCATACAGGGTTCTT), IQGAP3 (AAAGCAAACCTGAATGCTATT) or control (GGAATCTCATTCGATGCATAC). Cells were transfected using Lipofectamine LTX and Plus Reagent (Invitrogen, Carlsbad, CA) as described by the manufacturer and plated 6 hours after transfection onto collagen-I-coated plastic dishes for FACS sorting and western blotting of FACS-sorted GFP-positive cells (IQGAP3 depletion) or onto collagen-I-coated Rinzl plastic coverslips (Electron Microscopy Sciences, Hatfield, PA) for NGF treatment and immunostaining (IQGAP3 and ERK1/2 depletion). Rabbit antiserum against IQGAP3 was provided by Kozo Kaibuchi and Takashi Watanabe, Nagoya University, Japan.

Transfected cells were cultured for 3 days (ERK1/2 co-depletion) or 5 days (IQGAP3 depletion), serum-starved overnight and treated as indicated on day 4 or 6 after transfection with 50 ng/ml NGF (Upstate, Lake Placid, NY).

Cells on dishes were collected 6 days after transfection for sorting of GFP-positive cells by fluorescence-activated cell sorting using a Vantoo DIVa BD Digital Vantage Sorter (Stanford shared FACS Facility, Stanford University, CA). Equal numbers of GFP-positive cells transfected with control shRNA or IQGAP3 shRNA were analyzed by SDS-PAGE and western blotting for IQGAP3 protein levels.

Control, growth factor and inhibitor treatments were performed in parallel. Controls on each blot are DMSO at the same concentration as in the inhibitor treatments. Cells were washed in ice-cold Tris buffer (50 mM Tris-HCl, pH 7.4, 100 mM NaCl, 1 mM MgCl₂) and scraped in MECB (50 mM Tris-HCl, pH 7.4, 100 mM NaCl, 1 mM MgCl₂, 0.5% Nonidet P-40, 1 mM Pefabloc, 5 μg/ml leupeptin, 5 μg/ml pepstatin, 5 μg/ml aprotinin, 0.1 mM sodium orthovanadate, 20 μg/ml TPCK, 1 unit α₂-macroglobulin, 1 mM DTT). The MECB buffer was supplemented with Complete Protease Inhibitor Cocktail from Roche (Indianapolis, IN) and Phosphatase Inhibitor Cocktail 1 and 2 (Sigma). Cell extracts were prepared after a 10 minute incubation on ice by 10 minute centrifugation at 14,000 r.p.m. at 4°C. For λ-phosphatase treatment, cell extracts were prepared in the absence of sodium orthovanadate and phosphatase inhibitors and incubated with 400 U λ-protein phosphatase (New England Biolabs, Ipswich, MA) supplemented with 1 mM MnCl₂ and supplied 10× buffer at 30°C for 30 minutes. Protein concentrations in cell extracts were determined using the BCA Protein Assay Kit (Thermo Scientific, Rockford, IL) and equal amounts of protein were loaded in each lane for SDS-PAGE.

For APC western blots, cell extracts were subjected to SDS-PAGE in pre-cast Tris-acetate 3–8% polyacrylamide gradient gels (Bio-Rad, Hercules, CA) and transferred to PVDF membranes. Rabbit antibodies against APC (C-20, 1:500) and ERK (K-23, 1:1000) were from Santa Cruz Biotechnology (Santa Cruz, CA); mouse anti-pERK (3A7, 1:500), rabbit anti-pAKT (Ser473, 1:500) and rabbit anti-AKT (1:1000) were from Cell Signaling Technology (Danvers, MA). Mouse anti-β-catenin was from BD Transduction Laboratories (San Diego, CA). Rabbit antiserum against IQGAP3 was provided by Kozo Kaibuchi and Takashi Watanabe, Nagoya University, Japan. Secondary goat anti-mouse antibody IRDye 800 (1:15,000) was from Li-COR Biotechnology (Lincoln, NE), anti-rabbit Alexa Fluor 650 (1:30,000) was from Molecular Probes (Eugene, OR). Immunoblots were scanned at 680 nm and 800 nm using an Odyssey infrared imaging system (Li-COR Biotechnology, Lincoln, NE) and quantified with ImageJ. To quantify APC phosphorylation, the integrated intensity of the slower migrating band (pAPC, closed arrowhead, Figs 1, 2, 3 and supplementary material Figs S1,S2) was normalized to the total APC integrated intensity (closed and open arrowheads, Figs 1, 2, 3 and supplementary material Figs S1,S2).

For phosphorylated AKT to AKT ratios, total AKT levels were confirmed with rabbit anti-AKT in a separate immunoblot that was loaded in parallel with equal amounts of the same samples as used for the immunoblot incubated with rabbit pAKT primary antibody. For phosphorylated ERK to ERK ratios, equal amounts of protein were loaded and the membrane was co-stained with mouse anti-pERK, and rabbit anti-ERK primary antibodies and secondary anti-mouse IRDye 800 and anti-rabbit Alexa Fluor 680 antibodies. Phosphorylated ERK levels in relation to ERK were directly quantified by scanning the immunoblots at 800 and 700 nm, respectively, using the LI-COR Infrared scanner system (LI-COR Biosciences, Lincoln, NB); this obviates the need for stripping the blots. ERK phosphorylation is expressed as pERK integrated intensity normalized by ERK integrated intensity. Significance was calculated using unpaired, two-tailed Student’s t-test.

For ERK1/2 and IQGAP3 depletion experiments (Figs 4,5), PC12 cells were fixed with 4% formaldehyde EM grade (Electron Microscopy Sciences, Hatfield, PA) in Cytoskeleton Buffer (10 mM MES, pH 6.1, 138 mM KCl, 3 mM MgCl, 2 mM EGTA, 0.32 M sucrose) to preserve GFP fluorescence, and immunostained with affinity-purified APC-2 rabbit antiserum 1:500 or with phospho-ERK1/2 mouse monoclonal antibody 3A7 1:200 (Cell Signaling Technology, Danvers, MA). To determine the percentage of cells with APC clusters, at least 100 GFP-expressing depleted or control cells were analyzed for each experiment in five independent experiments. For all other experiments, PC12 cells were fixed in 100% methanol at –20°C for 15 minutes and stained with anti-APC antibody H290 (1:200) (Santa Cruz Biotechnology Santa Cruz, CA). This APC antibody was validated for its use in immunofluorescence by comparing immunostaining results with results obtained with APC-2 and APC-3 antisera that we have characterized previously (Näthke et al., 1996; Votin et al., 2005). Images were obtained using an Axiovert 200M microscope with a 100× Plan-APOCHROMAT 1.40 oil differential interference contrast objective (Carl Zeiss MicroImaging, Thornwood, NY) equipped with a Photometrics CoolSNAP camera (Roper Scientific, Tucson, AZ), Images were analyzed using ImageJ. Controls, growth factor and inhibitor treatments were performed in parallel in three independent experiments. DMSO had no effect on NGF-induced clustering of APC at the tip of cell extensions at the concentrations used (supplementary material Fig. S5). A total of 35–100 cells were counted for each treatment in each experiment. We defined an APC cluster as a local accumulation of APC fluorescence at the extension tip that was above APC fluorescence along the cell extension. APC cluster formation is expressed in percentage of cells with APC clusters ± s.e.m. with P values from unpaired, two-tailed Student’s t-tests.

GST–SAMP3end was produced by cloning DNA encoding amino acid residues 2038–2843 of human APC into BamHI and SalI sites of pGEX-6P-3 vector (Amersham Biosciences, Piscataway, NJ) and transformed into BL21 cells (Invitrogen, Carlsbad, CA). Protein expression and purification was performed as previously described (Barth et al., 2002) with some modifications. Briefly, protein expression was induced with 0.3 mM IPTG at 30°C for 4 hours. Proteins were purified over glutathione–agarose resin (Sigma) and eluted in 20 mM Tris-HCl, pH 8.7, 150 mM KCl, 1 mM DTT and 15 mM reduced glutathione, flash frozen and stored at –80°C. Protein concentration was determined by densitometric analysis of Coomassie-Blue-stained gels.

For ³²P labeling, 260 nM of purified GST–SAMP3end were incubated in the presence or absence of 2.5 ng/μl MEK-activated recombinant human ERK1 or ERK2 (SignalChem, Richmond, BC, Canada) or 2.5 ng/μl constitutively active recombinant human MEK1 (Calbiochem, Gibbstown, NJ). To determine whether the phosphorylation was due to residual MEK1 in the ERK2 preparation, 1 μM and 10 μM of the MEK inhibitor PD0325901 (Stemgent, Cambridge, MA) was included in the assay. A tenfold molar amount of GST was used as a negative control for ERK2 phosphorylation to ensure detection by Coomassie Brilliant Blue staining and to rule out the possibility of incorporation of ³²P in the GST domain. Inactivated ERK2 (Calbiochem, Gibbstown, NJ) was used as substrate to test for MEK activity and proper action of PD0325901, and detected by immunoblotting with antibodies against ERK and phosphorylated ERK. Reactions were performed in kinase buffer (25 mM Tris-HCl, pH 7.5, 10 mM MgCl₂, 0.05 mM EGTA, 5 mM β-glycerophosphate, 0.1 mM sodium orthovanadate, 2 mM DTT) and 87.5 nCi/nmol [γ-³²P]ATP in a total concentration of 200 μM ATP and a final volume of 50 μl. Reactions were incubated at 30°C for 30 minutes and stopped by addition of Laemmli SDS sample buffer, followed by SDS-PAGE and autoradiography or western blot in the case of MEK1 phosphorylation of inactive ERK2.

The preparation of protein for analysis of phosphorylation sites by LC-MS/MS, and cytoskeleton binding and bundling experiments was as described above with 3.85 ng/μl of activated recombinant human ERK2 (SignalChem, Richmond, BC, Canada) and a total of 1 mM ATP at 30°C for 1 hour to promote complete phosphorylation of the substrate. Unphosphorylated controls were incubated in the same conditions without kinase. After phosphorylation, samples were cleared by centrifugation at 4°C and 200,000 g in a TL100 Ultracentrifuge (Beckman) and used immediately.

Samples for LC-MS/MS analysis were prepared by a previously described method (Shevchenko et al., 2006). Polyacrylamide gel slices were sectioned into ∼1 mm strips, destained with 50% acetonitrile, 50% 100 mM ammonium bicarbonate buffer, dehydrated with acetonitrile and then rehydrated in 10 mM ammonium bicarbonate containing 13 ng/μl modified sequencing grade trypsin (Promega). After covering with 100 mM ammonium bicarbonate, the gel pieces were incubated for 12 hours at 37°C, the peptides were recovered, desalted and the mixture analyzed by capillary LC-MS/MS. The peptide mixtures were separated on a 0.32 mm×10 cm C18 capillary reversed-phase column with buffers containing 0.1% formic acid using a linear gradient of 0–55% acetonitrile delivered by a capillary HPLC pump (Agilent Model 1100). The outlet of the column was connected directly to the electrospray source of a LTQ Orbitrap XL model hybrid mass spectrometer system (Thermo Fisher Scientific). The data were analyzed by generating chromatograms using a 3 mTh (milli-Thomson) window around the calculated theoretical mass of the tryptic phosphopeptides for what were deemed the most likely phosphorylation sites and manually interpreting the MS/MS scans. Other phosphopeptides were found from neutral-fragment mass chromatograms to identify those peptides that included a possible loss of phosphoric acid in the MS/MS scan. Finally, the entire set of MS/MS scans was searched against a protein sequence database to which the sequence of the protein construct had been added, to see whether additional phosphopeptides not identified by the other procedures could be identified.

Bovine brain tubulin (Cytoskeleton, Denver, CO) was polymerized according to the manufacturer’s specifications. Purified chicken G-actin was a gift from Daniel Dickinson (Stanford University, Palo Alto, CA). Actin was polymerized in 20 mM imidazole, pH 7.0, 100 mM KCl, 2 mM MgCl₂, 500 μM ATP, 1 mM EGTA for 1 hour at room temperature (RT) and stabilized with an equimolar amount of phalloidin for 30 minutes at RT. G-actin was removed by centrifuging the reaction at 417,200 g for 20 minutes and resuspending the pellet in 20 mM imidazole, pH 7.0, 150 mM NaCl, 2 mM MgCl₂, 500 μM ATP, 1 mM EGTA by passing the solution through a 26G needle. After polymerization, all filaments were handled using wide-mouthed pipetteman tips to minimize microtubule shearing. For all pull-down experiments GST-SAMP3end and controls were centrifuged at 100,000 g for 40 minutes at 4°C after the phosphorylation reaction and incubation with the filaments in order to avoid unspecific spin down of the fragments.

For microtubule-binding assays, we followed the protocol provided by vendor (Cytoskeleton) with modifications. Briefly, 150 nM of unphosphorylated or phosphorylated GST–SAMP3end was incubated without or with increasing concentrations of polymerized microtubules in General Tubulin Buffer (80 mM PIPES, pH 7, 2 mM MgCl₂, 0.5 mM EGTA, 20 μM taxol) in a total volume of 50 μl for 30 minutes at RT. Samples were loaded onto 100 μl of cushion buffer (50% glycerol, 80 mM PIPES, pH 7, 2 mM MgCl₂, 0.5 mM EGTA, 20 μM taxol) and centrifuged at 100,000 r.p.m. for 40 minutes at RT in a TLA-100.1 rotor (Beckman). Pellets were analyzed by SDS-PAGE, Coomassie Brilliant Blue staining and quantified with ImageJ. Signal intensity values for GST-SAMP3end pelleted with microtubules were corrected for the respective signal intensities pelleted without filaments and expressed as percentage of total input.

For microtubule bundling assays, GST and unphosphorylated and phosphorylated GST–SAMP3end were equilibrated at RT, and the indicated concentrations were mixed with 2 μM polymerized microtubules in kinase buffer with 1 mM ATP in a final volume of 60 μl in 0.5 ml Eppendorf tubes. The reactions were incubated for 5 minutes at RT and centrifuged in a microcentrifuge at 6500 g for 5 minutes at RT. Pellets were washed with General Tubulin Buffer and centrifuged again. Equivalent amounts of supernatant and pellets were separated by SDS-PAGE and stained with Coomassie Brilliant Blue. Quantification of microtubule bundling was performed with ImageJ and expressed as the integrated intensity of the pellet normalized by the total amount of tubulin in the sample.

Actin-binding and -bundling experiments were performed as previously described (Harris et al., 2006) with modifications. For actin-binding assays, 100 nM of unphosphorylated or phosphorylated GST–SAMP3end was incubated without or with increasing concentrations of F-actin in actin reaction buffer (25 mM Tris-HCl, pH 7.5, 1 mM EGTA, 500 μM ATP, 2 mM MgCl₂, 1 mM DTT and 0.01% Nonidet P-40) for 10 minutes at room temperature. Samples were centrifuged at 100,000 r.p.m. for 20 minutes at 4°C in a TLA-100.1 rotor (Beckman). Pellets were washed in 60 μl actin reaction buffer and re-centrifuged. Pellets and supernatants were analyzed by SDS-PAGE, and Coomassie Brilliant Blue staining. The amount of sedimented GST–SAMP3end was quantified as described above for microtubules.

For F-actin-bundling assays, unphosphorylated or phosphorylated GST–SAMP3end or GST at the indicated concentrations were mixed with 2 μM F-actin in actin reaction buffer (25 mM Tris-HCl, pH 7.5, 1 mM EGTA, 500 μM ATP, 2 mM MgCl₂, 1 mM DTT and 0.01% Nonidet P-40) in a final volume of 60 μl in 0.5 ml Eppendorf tubes. After incubation for 10 minutes at room temperature, samples were centrifuged in a microcentrifuge at 16,000 g for 5 minutes at RT. Pellets were washed briefly with 60 μl of actin reaction buffer and centrifuged again. Equivalent amounts of supernatants and pellets were analyzed by SDS-PAGE and staining with Coomassie Brilliant Blue, and pelleted actin was quantified as described above for tubulin.

We thank Kozo Kaibuchi and Takashi Watanabe for antiserum to IQGAP3.