PAK1 Negatively Regulates the Activity of the Rho Exchange Factor NET1*

Rho family small G-protein activity is controlled by guanine nucleotide exchange factors that stimulate the release of GDP, thus allowing GTP binding. Once activated, Rho proteins control cell signaling through interactions with downstream effector proteins, leading to changes in cytoskeletal organization and gene expression. The ability of Rho family members to modulate the activity of other Rho proteins is also intrinsic to these processes. In this work we show that the Rac/Cdc42hs-regulated protein kinase PAK1 down-regulates the activity of the RhoA-specific guanine nucleotide exchange factor NET1. Specifically, PAK1 phosphorylates NET1 on three sites in vitro: serines 152, 153, and 538. Replacement of serines 152 and 153 with glutamate residues down-regulates the activity of NET1 as an exchange factor in vitro and its ability to stimulate actin stress fiber formation in cells. Using a phospho-specific antibody that recognizes NET1 phosphorylated on serine 152, we show that PAK1 phosphorylates NET1 on this site in cells and that Rac1 stimulates serine 152 phosphorylation in a PAK1-dependent manner. Furthermore, coexpression of constitutively active PAK1 inhibits the ability of NET1 to stimulate actin polymerization only when serines 152 and 153 are present. These data provide a novel mechanism for the control of RhoA activity by Rac1 through the PAK-dependent phosphorylation of NET1 to reduce its activity as a guanine nucleotide exchange factor.

three classes of proteins known as guanine nucleotide exchange factors (GEFs), 1 GTPase activating proteins, and guanine nucleotide dissociation inhibitors (6). GEFs regulate Rho GTPase activation by catalyzing the release of GDP, thus allowing the binding of GTP. GTPase activating proteins inactivate Rho proteins by stimulating their intrinsic GTPase activities. Guanine nucleotide dissociation inhibitors preferentially bind to the inactive, GDP-bound forms of Rho proteins, thereby sequestering them from their sites of action. In this regulatory scheme it is the GEFs that control small G-protein activation in response to growth factor and mitogen stimulation.
The Rho GEF family currently consists of 25 cloned genes and is predicted to contain upwards of 50 family members when fully characterized (7)(8)(9)(10). Each Rho GEF displays a unique specificity for different Rho family small G-proteins. For example, the Rho GEF Dbl efficiently catalyzes GDP release for RhoA, Rac1, and Cdc42hs (11). On the other hand, the Rho GEF Tiam1 specifically catalyzes nucleotide exchange only for Rac1 (12). The regulatory mechanisms controlling the enzymatic activities of different Rho GEFs are equally diverse. These include binding to phosphoinositides, altered proteinprotein interactions, changes in subcellular localization, and site-specific phosphorylation (9,13). The Rho GEF Vav, for example, is activated by the combined effects of phosphorylation on specific tyrosine residues and the binding of phosphoinositide 3,4,5-phosphate. This relieves the actions of an autoinhibitory domain and stimulates its GDP exchange activity toward Rho proteins (14 -17). Similarly, the activity of the Rho GEFs Tiam1, Dbl, and FGR are positively regulated by phosphorylation (18 -20).
The neuroepithelioma transforming gene 1 (NET1) is a Rho GEF that was first identified in a screen for transforming genes in NIH 3T3 cell focus formation assays (21). It contains a negative regulatory domain in its amino terminus, and deletion of this domain creates an oncogenic form of the protein. In addition to transforming NIH 3T3 cells, expression of oncogenic NET1 stimulates actin stress fiber formation, c-Jun NH 2 -terminal kinase/mitogen-activated protein kinase activation, and serum response factor activation (21,22). Recently it was demonstrated that wild-type NET1 is localized in the nucleus, and that truncation of the amino terminus results in relocalization of a fraction of the NET1 to the cytoplasm. This was at least partially due to the elimination of two putative nuclear localization signals within the amino terminus (23). Thus, NET1 activity is regulated at least in part through subcellular localization. No other mechanisms controlling NET1 activity have been described.
In this work we show that NET1 is negatively regulated through phosphorylation by the serine/threonine protein kinase PAK1. Specifically, PAK1 phosphorylates the oncogenic form of NET1 (NET1⌬N) on serines 152, 153, and 538 in vitro. Replacement of serines 152 and 153 with glutamate residues, which mimics their phosphorylation, inhibits the GDP exchange activity of NET1 toward RhoA in vitro, as well as the ability of NET1 to stimulate actin stress fiber formation in cells. Using an antibody that specifically recognizes NET1 phosphorylated on serine 152, we show that PAK1 phosphorylates this site in cells and that expression of constitutively active Rac1 stimulates the phosphorylation of NET1 on serine 152 in a PAK-dependent manner. Furthermore, we show that PAK specifically down-regulates NET1 activity in cells only when serine 152 is intact. These data demonstrate a novel mechanism for controlling NET1 activity and provide an additional means by which Rac1 controls RhoA activity.
For the production of GST fusion proteins, all NET1 cDNAs, as well as RhoA, were subcloned into pGEXKG (Amersham Biosciences). NET1 proteins and RhoA were produced in BL21 DE3 Escherichia coli (Stratagene) after transformation with the appropriate construct. One-liter cultures were grown at 37°C to an A 600 ϭ 0.8, and expression of fusion proteins was stimulated by addition of isopropyl 1-thio-␤-D-galactopyranoside to 400 M, followed by incubation at 37°C for 4 h. Cell pellets were then collected by centrifugation and frozen. Cells were lysed and GST-NET1 fusion proteins or GST-RhoA were purified using a glutathione-agarose affinity resin (Sigma) essentially as described (24). GST-RhoA was purified in the presence of 10 M GDP. GST-PAK1 was purified from E. coli as described (25). The concentration of each GST fusion protein was determined by bicinchoninic acid assay (Pierce) and confirmed by SDS-PAGE followed by Western blotting. Proteins were stored at Ϫ80°C for later use.
In Vitro Kinase Assays, Antibodies, Immunoprecipitations, and Western Blotting-GST-NET1 fusion proteins were phosphorylated in vitro with GST-PAK1 (26) at 30°C for 30 min in kinase buffer (20 mM Tris-HCl, pH 8.0, 10 mM MgCl 2 , 1 mM dithiothreitol, 100 M ATP, 2 Ci of [␥-32 P]ATP (PerkinElmer Life Sciences)). Proteins were then resolved by 12% SDS-PAGE and visualized by Coomassie staining. After drying the gel, phosphorylated proteins were visualized by autoradiography. Radioactive phosphate incorporated into NET1 was determined by scintillation counting of the excised protein bands. Phosphoamino acid analysis of phosphorylated NET1 proteins was performed as described (26).
For production of the anti-phosphoserine 152 (anti-pSer-152) antibody, the peptide NH 2 -TKRR(pS)SALWSEC-COOH, corresponding to mouse NET1 amino acids 148 -159 phosphorylated on serine 152, was conjugated to keyhole limpet hemocyanin. New Zealand White rabbits were immunized with this conjugate and the blood was recovered as described (Sigma-Genosys). Mouse and rabbit anti-hemagglutinin antibodies and mouse anti-GST were from Santa Cruz Biotechnology. For analysis of serine 152 phosphorylation by Western blotting, cells were lysed in RIPA plus phosphatase and protease inhibitors (50 mM Tris-HCl (pH 8.0), 100 mM NaCl, 1% Triton X-100, 0.5% sodium deoxy-cholate, 0.1% SDS, 80 mM ␤-glycerophosphate, 0.5 mM sodium orthovanadate, 10 g/ml pepstatin A, 10 g/ml leupeptin, 2 g/ml aprotinin, 1 mM phenylmethylsulfonyl fluoride); the DNA was sheared by 10 passages through a 23-gauge needle; and insoluble material was pelleted by centrifugation (16,000 ϫ g for 10 min at 4°C). Hemagglutinin (HA)-tagged NET1 proteins were then immunoprecipitated from soluble lysates using a mouse anti-HA antibody, and the immunoprecipitates were washed 3 ϫ 1 ml with buffer A (500 mM NaCl, 20 mM Tris-HCl (pH 8.0)), followed by washing 1 ϫ 1 ml with 20 mM Tris-HCl (pH 8.0). Immunoprecipitated proteins were resuspended in 1 ϫ Laemmli sample buffer, resolved by SDS-PAGE, and transferred to polyvinylidene difluoride membranes (Amersham Biosciences). The anti-pSer-152 antibody was used for Western blotting by diluting in Tris-buffered saline plus 0.05% Tween 20 (TBST) ϩ 5% nonfat milk and incubating overnight at room temperature. All other antibodies were diluted in TBST ϩ 0.025% nonfat milk and incubated with membranes for 1 h at 37°C or overnight at 4°C. After washing with TBST, membranes were incubated with horseradish peroxidase-conjugated goat anti-rabbit or anti-mouse antibodies (KPL) for 30 min at room temperature. After washing with TBST, the membranes were developed by enhanced chemiluminesence. Quantification of Western blots was performed by densitometry followed by analysis using NIH Image.
Guanine Nucleotide Exchange Assays-Exchange factor assays measuring NET1⌬N activity toward GST-RhoA in vitro were performed essentially as described (27). Briefly, GST-RhoA was loaded with GDP for 5 min at room temperature in GDP loading buffer (20 mM Tris-HCl Microinjection of NET1 Expression Plasmids-NIH 3T3 cells plated on glass coverslips were incubated in Dulbecco's modified Eagle's medium with 0.05% fetal bovine serum for 24 h prior to injection. All plasmids were injected into the nuclei of cells at a concentration of 0.05 mg/ml as previously described (22). Briefly, 4 h after injection, the cells were fixed in 3.7% formaldehyde in phosphate-buffered saline at 37°C for 5 min followed by permeabilization with 0.3% Triton X-100 in phosphate-buffered saline for 5 min at room temperature, and the cells were then incubated with mouse anti-HA antibody (monoclonal antibody 12CA5) to detect expression of HA-tagged NET1 proteins. Coexpressed, Myc-tagged PAK L107F was detected using rabbit anti-Myc tag antibody (Invitrogen). Following extensive washing in phosphatebuffered saline, binding of primary antibodies was detected by incubating with donkey anti-rabbit or anti-mouse antibodies coupled to either fluorescein isothiocyanate or 7-amino-4-methylcoumarin-3-acetic acid, as indicated. F-actin morphology and accumulation were monitored by co-staining with tetramethylrhodamine-conjugated phalloidin (Molecular Probes). Coverslips were mounted on slides with Gelvatol and imaged with a Nikon E600 epifluorescence microscope equipped with a Spot CCD camera using fixed exposure times. Images were imported into the OpenLab software package. Relative fluorescence intensities of TRITC-phalloidin-labeled actin in injected and adjacent uninjected cells were recorded on a 0-to 256-bit scale. The fluorescence intensities from 5 to 10 injected, HA-tag expressing cells per experiment are reported as the mean intensity.

Characterization of NET1 Phosphorylation by PAK1 in
Vitro-To date there have been no descriptions of the biochemical mechanisms controlling the activity of the Rho GEF NET1. Because Rac1 has been reported to down-regulate the activity of RhoA (28 -30), we examined whether NET1 activity is regulated through phosphorylation by the Rac1 effector PAK1. We first examined whether PAK1 could phosphorylate NET1 in vitro. For these assays, wild-type NET1 and an amino-terminal truncated, transforming version of NET1 (NET1⌬N, amino acids 122-595) were expressed in bacteria as GST fusion pro-teins and affinity purified. These proteins were then tested as PAK1 substrates in vitro using recombinant, constitutively active GST-PAK1. As shown in Fig. 1A, GST-PAK1 efficiently phosphorylated both full-length NET1 and NET1⌬N. Phosphoamino acid analysis revealed that both NET1 proteins were exclusively phosphorylated on serine residues (data not shown). We then constructed a series of deletion mutants to identify the subdomains of NET1 that were phosphorylated by PAK1. Thus, GST fusion proteins corresponding to the Dbl and pleckstrin homology domains (amino acids 156 -361 and 362-501, respectively), as well as the amino-and carboxyl-terminal flanking regions (amino acids 122-155 and 502-592, respectively) were produced in bacteria and tested as PAK1 substrates. As shown in Fig. 1B, recombinant PAK1 efficiently phosphorylated flanking regions 122-155 and 502-595, but not the isolated Dbl or pleckstrin homology domains. Thus, in vitro, PAK1 phosphorylates multiple residues of NET1⌬N that are located in two distinct domains.
Alignment of the amino acid sequences surrounding a number of published PAK1 substrates demonstrates that PAK1 phosphorylation sites often contain the consensus sequence (R/X)-(R/X)-X-S*-S/⌽, where ⌽ is a hydrophobic amino acid. 2 Furthermore, in proteins containing two adjacent serine residues at the P 0 and P ϩ1 sites, the first serine is often preferentially phosphorylated. Thus, using this consensus sequence as a guide, we mutated potential PAK1 phosphorylation sites in both NET1-(122-155) and NET1-(156 -595) to alanine, expressed these polypeptides as GST fusion proteins in E. coli, and tested them for phosphorylation by GST-PAK1 in vitro.
The amino-terminal portion of NET1 contained two potential PAK1 phosphorylation sites, located at serines 152 and 153, with serine 152 predicted to be the preferred site. As shown in Fig. 1C, replacement of serine 152 with alanine reduced the phosphorylation of NET1-(122-155) to 72% of wild-type NET1-(122-155), whereas replacement of serine 153 had no effect. On the other hand, replacement of both serines with alanine (S152A/S153A) completely eliminated phosphorylation by PAK1. Thus, these data indicate that PAK1 phosphorylates NET1-(122-155) mainly on serine 152 in vitro, and that if this site is mutated to alanine, serine 153 will be used as a substrate instead. Phosphorylation of the serine at the P ϩ1 position after mutation of the P 0 serine to a non-phosphorylatable residue has been observed previously for the PAK1 substrates MEK1 and Raf-1 (26,31,32).
We then identified the site phosphorylated by PAK1 in the C terminus of NET1. This sequence also contained two potential PAK1 phosphorylation sites, located at serines 538 and 557. However, substitution of these sites with alanines revealed that only serine 538 was phosphorylated by PAK1 in vitro (Fig.  1D). To confirm that we had identified all of the PAK1 phosphorylation sites in NET1⌬N, we then created a triple alanine mutant of NET1⌬N (S152A/S153A/S538A) and tested it as a substrate for PAK1. As shown in Fig. 1E, NET1⌬N containing the triple serine to alanine alteration was no longer phosphorylated by PAK1. Thus, PAK1 phosphorylates three sites on NET1⌬N in vitro: serines 152, 153, and 538 (Fig. 1F).
Replacement of Serines 152 and 153 with Acidic Residues Inhibits the GDP Exchange Activity of NET1⌬N in Vitro-We next tested whether replacement of the PAK1 phosphorylation sites in NET1⌬N with alanine or glutamate residues affected the exchange factor activity of NET1⌬N toward RhoA in vitro.
In these experiments, replacement with alanine was expected to be neutral in effect, whereas replacement with negatively charged glutamic acid was expected to mimic phosphorylation by PAK1. We first tested the effect of replacing serines 152 and 153 with alanine or glutamate. As shown in Fig. 2A, wild-type NET1⌬N efficiently catalyzed GTP␥S binding by RhoA. In addition, replacing these serines with alanines did not significantly affect NET1⌬N enzymatic activity. However, substitution with glutamates at these sites significantly reduced the ability of NET1⌬N to stimulate GTP␥S binding by RhoA. These data suggest that phosphorylation of serines 152 and 153 by PAK1 negatively regulates the ability of NET1⌬N to stimulate GDP exchange for RhoA.
We then tested the effect of replacing serine 538 with alanine or glutamate. As shown in Fig. 2B, neither change significantly affected the exchange activity of NET1⌬N. This suggests that phosphorylation of serine 538 by PAK1 does not affect NET1⌬N activity toward RhoA. Lastly, we tested the effects of replacing all three PAK1 phosphorylation sites with alanine or glutamate residues. As shown in Fig. 2C, the triple alanine mutant exhibited the same GEF activity toward RhoA as wildtype NET1⌬N, but the triple glutamate mutant was much less effective at catalyzing GTP␥S binding by RhoA. Similar results were observed with recombinant, full-length GST-NET1 (data not shown). Thus, we conclude from these experiments that phosphorylation of either NET1⌬N or full-length NET1 by PAK1 on serines 152 and 153 may negatively regulate NET1 activity toward RhoA, and that phosphorylation of serine 538 is likely to be without effect.
Acidic Substitutions at the Amino-terminal PAK1 Phosphorylation Sites Block the Ability of NET1⌬N to Stimulate Actin Stress Fiber Formation-It has been shown that expression of NET1⌬N in cells stimulates F-actin accumulation and stress fiber formation, similar to the effect of expressing constitutively active RhoA (22). In NIH 3T3 cells, the two primary RhoA effectors are thought to be the formin mDia1 and Rho kinase, which nucleate and bundle actin filaments into stress fibers, respectively (33). To test whether alteration of the amino-terminal PAK1 phosphorylation sites affected the ability of NET1⌬N to regulate cytoskeletal organization, serum-starved NIH 3T3 cells were microinjected with HA epitope-tagged NET1⌬N, or NET1⌬N containing alanine or glutamate substitutions at these sites. Four hours after injection, the cells were fixed and stained by indirect immunofluorescence for expression of NET1⌬N proteins (anti-HA) and filamentous actin (Factin) (Fig. 3A, right and left columns, respectively). The presence of F-actin was detected by incubating the cells with TRITC-labeled phalloidin, which binds specifically to polymerized actin. The relative amounts of F-actin in each cell were then quantified by monitoring fluorescence because of the TRITC-phalloidin. In these assays, expression of NET1⌬N induced the formation of numerous actin stress fibers, consistent with previously published results (22) (Fig. 3A). Quantification showed that NET1⌬N stimulated an average 3-fold increase in mean TRITC fluorescence relative to neighboring uninjected cells, or to cells expressing a catalytically inactive variant of NET1⌬N (L321E) (22) (Fig. 3B). Substitution of serines 152 and 153 with alanines did not significantly affect the ability of NET1⌬N to stimulate stress fiber formation or F-actin accumulation (Fig. 3A, bottom panels). In contrast, glutamate substitution at the PAK1 phosphorylation sites significantly disrupted the ability of NET1⌬N to stimulate stress fiber formation and F-actin assembly (Fig. 3, middle panels). Quantification of these effects showed that NET1⌬N S152E/S153E stimulated only a slight increase in mean F-actin staining relative to the surrounding, non-injected cells (Fig. 3B). These results strongly indicate that phosphorylation of NET1⌬N on serines 152 and 153 negatively regulates its activity in cells, and are consistent with the observed effects of glutamate substitution of these sites on the enzymatic activity of NET1⌬N in vitro (Fig. 2).
NET1 Is Phosphorylated on Serine 152 in Cells Following the Expression of Constitutively Active PAK1 or Rac1-To determine whether NET1 was phosphorylated by PAK1 in cells, an antibody was produced that specifically recognized NET1 only when phosphorylated on serine 152 (anti-pSer-152). This residue was chosen rather than serine 153 because PAK1 primarily phosphorylated NET1 on serine 152 in vitro (Fig. 1C).
We first tested the anti-pSer-152 antibody for its ability to recognize NET1 when phosphorylated on serine 152 in vitro. Recombinant, GST-NET1⌬N or GST-NET⌬N fusion proteins containing alanine substitutions at serines 152 and 153, or at serine 538, were incubated in kinase buffer, with or without recombinant GST-PAK1. Following separation by SDS-PAGE, proteins were transferred to polyvinylidene difluoride membrane and immunoblotted with the anti-pSer-152 antibody. As shown in Fig. 4A, the anti-pSer-152 antibody recognized NET1⌬N only when it was phosphorylated by PAK1 ( compare  lanes 1 and 5). In addition, substitution of serines 152 and 153 To determine whether the antibody was specific for phosphorylation on serine 152, we tested its ability to recognize NET1 with single alanine substitutions at serines 152 and 153. Thus, recombinant GST-NET1-(122-155) or GST-NET1-(122-155) containing alanine substitutions at serines 152, 153, or both residues were used as substrates for recombinant PAK1 in vitro, separated by SDS-PAGE, and tested for recognition by  (Fig. 1C), we interpret these data to mean that the anti-pSer-152 antibody recognizes NET1 only when phosphorylated on serine 152, and that the unphosphorylated form of serine 153 may contribute to the epitope recognized by this antibody.
To determine whether PAK1 phosphorylated NET1 on serine 152 in cells, HEK 293 cells were transfected with HA epitopetagged NET1⌬N, with or without constitutively active PAK1. The cells were lysed in RIPA buffer, the NET1⌬N was immunoprecipitated, and the precipitates were immunoblotted with the anti-pSer-152 antibody. As shown in Fig. 5A, NET1⌬N exhibited a low level of phosphorylation on serine 152 in serumstarved cells (lane 2). Co-expression of constitutively active PAK1 (PAK1*), on the other hand, strongly stimulated the phosphorylation of this site (lane 3). As expected, NET1⌬N containing alanine substitutions at serines 152 and 153 was not recognized by this antibody (lanes 4 and 5). We also tested whether full-length NET1 was phosphorylated by PAK1 in cells. As shown in Fig. 5B, PAK1 effectively stimulated the phosphorylation of full-length NET1 on serine 152 in cells (compare lanes 2 and 3), although the degree of phosphorylation was less than that observed for NET1⌬N. It is not clear why full-length NET1 was phosphorylated less efficiently than NET1⌬N. One possible explanation is that a substantial portion of full-length NET1 is localized to the nucleus and therefore may be inaccessible to PAK1. Alternatively, the aminoterminal truncation that makes NET1⌬N oncogenic may expose serine 152 for phosphorylation by PAK1 (see "Discus-sion"). Nevertheless, these data clearly demonstrate that PAK1 phosphorylates both NET1⌬N and full-length NET1 in cells.
We then tested whether Rac1 could stimulate the phosphorylation of NET1 on serine 152, and whether this depended on endogenous PAK activity. Thus, HEK 293 cells were transfected with NET1⌬N, minus or plus constitutively active Rac1 (V 12 Rac1). The NET1⌬N was then immunoprecipitated and tested for phosphorylation on serine 152 by Western blotting. As shown in Fig. 6A, expression of V 12 Rac1 strongly stimulated the phosphorylation of NET1⌬N on serine 152 (compare lanes  2 and 3). Furthermore, the ability of Rac1 to stimulate the phosphorylation of NET1⌬N on serine 152 was dependent on endogenous PAK activity, because co-expression of the autoinhibitory domain from PAK1 blocked Rac1-dependent phosphorylation of NET1⌬N (compare lanes 3 and 4). Given that expression of this autoinhibitory domain will block the activation of endogenous PAKs 1-3 (24,34,35), these data indicate that Rac1 requires the activity of one or more endogenous PAKs to stimulate the phosphorylation of NET1⌬N on serine 152. We also tested whether Rac1 would stimulate the phosphorylation of full-length NET1 on serine 152. As shown in Fig. 6B, expression of constitutively active Rac1 also stimulated full-length NET1 phosphorylation at this site, and this was dependent on the activation of endogenous PAKs. As was the case for phosphorylation by constitutively active PAK1, the phosphorylation of wild-type NET1 stimulated by V 12 Rac1 was less robust than for NET1⌬N. However, these data plainly demonstrate that both full-length NET1 and NET1⌬N are phosphorylated on serine 152 by endogenous PAKs in response to Rac1 activation.
The Cellular Activity of Full-length NET1 Is Regulated through Phosphorylation of Serines 152 and 153-Overexpression of full-length NET1 also stimulates actin reorganization, albeit to a much lower extent than NET1⌬N (23). To test if

FIG. 3. Substitution of the amino-terminal PAK1 phosphorylation sites with glutamates inhibits actin stress fiber formation stimulated by expression of NET1⌬N.
A, NIH 3T3 cells were maintained in 0.05% fetal calf serum/Dulbecco's modified Eagle's medium for 24 h and then microinjected with plasmids for HA epitope-tagged NET1⌬N (top panels), NET1⌬N containing alanine substitutions at serines 152 and 153 (NET1⌬N S152A/S153A, middle panels), or NET1⌬N containing glutamate substitutions at these sites (NET1⌬N S152E/S153E, bottom panels). Four hours later the cells were fixed and stained for F-actin (left column) and for NET1 expression (anti-HA(12CA5), right column). Shown is a representative experiment. B, quantification of the effect of NET1⌬N proteins on F-actin accumulation in microinjected NIH 3T3 cells. Cells were stained with TRITC-labeled phalloidin, which binds to polymerized F-actin. The amount of F-actin staining was quantified as mean fluorescence per cell from 5-10 cells per experiment. Results are the average from three separate experiments. Error bars are the S.E. serine 152 was a critical site for regulation of the full-length NET1 protein in cells, serum-starved NIH 3T3 cells were microinjected with expression vectors for HA epitope-tagged, fulllength NET1, or NET1 variants bearing alanine or glutamate substitutions at the amino-terminal PAK1 phosphorylation sites. We also tested whether co-expression of constitutively active PAK1 would affect the ability of NET1 or NET1 S152A/ S153A to elevate F-actin levels. As for experiments with NET1⌬N, the cells were fixed 4 h after injection and stained for NET1 expression (anti-HA) and polymerized actin (F-actin), and the levels of F-actin were quantified by TRITC-phalloidin staining. As shown in Fig. 7A, overexpression of full-length NET1 stimulated a moderate degree of actin polymerization, with a nearly 2-fold increase in F-actin levels compared with the neighboring uninjected cells (Fig. 7C). In these experiments, the NET1 was localized predominantly to the nucleus, although a low level of cytoplasmic staining was observed. Because RhoA must be activated at the plasma membrane to stimulate actin polymerization, this suggests that only very low levels of cytoplasmic NET1 are required for effects on the actin cytoskeleton. When we tested the phosphorylation site mutants, we observed that glutamate substitution at the PAK1 phosphorylation sites (NET1 S152E/S153E) only modestly inhibited the ability of NET1 to stimulate stress fiber formation and F-actin accumulation (Fig. 7, A, middle panels, and C). On the other hand, alanine substitution at these sites significantly enhanced the ability of full-length NET1 to induce stress fiber formation and F-actin accumulation (Fig. 7A, bottom panels). In fact, the ability of NET1 S152A/S153A to stimulate actin polymerization was nearly as great as NET1⌬N (compare Figs. 3B and 7C). Because alanine substitution of these sites did not affect the enzymatic activity of NET1 in vitro (see text for Fig.  2), these data suggest that a fraction of the full-length NET1 that localizes to the cytoplasm in these cells is negatively regulated by phosphorylation of serines 152 and 153. This would not be unexpected, because NIH 3T3 cells exhibit a high level of basal PAK activity even after serum starvation. 2 Thus, the activity of full-length NET1 may be reduced in these cells compared with what would be possible in the absence of endog-  5. PAK1 phosphorylates both NET1⌬N and full-length NET1 on serine 152 in cells. A, phosphorylation of NET1⌬N by PAK1 in HEK 293 cells. Cells were transfected with HA epitope-tagged NET1⌬N or NET1⌬N S152A/S153A, minus or plus constitutively active Myc epitope-tagged constitutively active PAK1 (PAK1*). After serum starvation, the cells were lysed in RIPA buffer and NET1⌬N proteins were immunoprecipitated with an antibody against the HA-epitope and tested for phosphorylation on serine 152 by Western blotting. The panels show, from top to bottom, a membrane probed with anti-pSer-152, the same membrane reprobed with an anti-HA antibody, and a Western blot of the cell lysate with anti-Myc to detect PAK1 expression. B, phosphorylation of full-length NET1 by PAK1 in HEK 293 cells. Cells were transfected with HA epitope-tagged full-length NET1 or NET1 S152A/S153A, minus or plus Myc epitope-tagged, constitutively active PAK1, and then were tested for phosphorylation on serine 152, as described in A. enous PAK activity. Even so, these data still demonstrate that serines 152 and 153 are critical to the ability of full-length NET1 to stimulate actin polymerization, and suggest that phosphorylation of these sites down-regulates NET1 activity in the cell.
To determine whether PAK1 directly regulated the activity of full-length NET1, we co-injected constitutively active PAK1 with full-length NET1 or NET1 containing alanine substitutions at the NH 2 -terminal phosphorylation sites (NET S152A/ S153A). As shown in Fig. 7B, coexpression of active PAK1 effectively blocked the ability of NET1 to stimulate actin stress fiber formation. Quantification of this effect showed that in these cells the level of F-actin staining was comparable with the surrounding, noninjected cells (Fig. 7C). On the other hand, NET1 S152A/S153A was resistant to down-regulation by PAK1 (Fig. 7B), although the formation of stress fibers (which typically traverse most of the cell) was somewhat impaired (Fig. 7B, bottom panels). Because constitutively active PAK1 stimulates the dissolution of actin stress fibers (24), this suggests that active PAK1 expression affected the assembly of F-actin into stress fibers, but did not affect the ability of NET1 to stimulate the generation of new actin filaments per se. Thus, these results demonstrate that PAK1 negatively regulates the activity of fulllength NET1 only when serines 152 and 153 are intact. DISCUSSION In this study we have identified NET1 as a novel PAK1 substrate. Specifically, NET1 is phosphorylated by PAK1 on serines 152, 153, and 538 in vitro, and on serine 152 in cells. Furthermore, expression of constitutively active Rac1 stimulates the phosphorylation of NET1 on serine 152 in a PAK-dependent manner. Three separate findings indicate that phosphorylation of this site negatively regulates the nucleotide exchange activity of NET1 toward RhoA. First, substitution of serines 152 and 153 with glutamate residues, which act as phospho-mimetics for many proteins, negatively regulates the enzymatic activity of NET1 toward RhoA in vitro. Second, NET1 S152E/S153E is impaired in its ability to stimulate Factin accumulation. Lastly, coexpression of constitutively active PAK1 with NET1 blocks the ability of full-length NET1 to stimulate actin stress fiber formation only when serines 152 and 153 are intact, indicating that PAK1 down-regulates NET1 activity in cells through the phosphorylation of these sites.
The biochemical mechanism by which phosphorylation of serines 152 and 153 inhibits the GDP exchange activity of NET1 is not clear, but may be because of disruption of the hydrophobic core within a novel NH 2 -terminal extension of the NET1 Dbl domain. This extension was first identified as a region of homology between NET1 and a subset of Rho GEFs (22), and its role was recently characterized in the crystal structure of RhoA with the related Rho GEF LARG (36). In LARG this NH 2 -terminal extension consisted of two small ␣-helices (␣N1 and ␣N2) that directly interacted with the switch 1 domain of RhoA (36). Furthermore, deletion of this extension, or mutation of a key tryptophan residue within the extension inhibited the enzymatic activity of LARG by nearly 80% in vitro. Thus, phosphorylation of this segment in NET1 may down-regulate its enzymatic activity by blocking interaction with the switch 1 domain of RhoA. In support of this hypothesis, we have observed that the region between residues 122 and 155 of NET1 is of critical importance to the catalytic activity of NET1. For example, although NET1⌬N (amino acids 122-595) is fully active as an exchange factor in vitro, a smaller protein lacking amino acids 122-155 (NET1-(156 -595)) is significantly impaired in its GDP exchange activity toward RhoA. 3 By producing an antibody that specifically recognizes NET1 phosphorylated on serine 152, we have successfully demonstrated that PAK1 stimulates its phosphorylation in cells. We have also shown that expression of constitutively active Rac1 led to serine 152 phosphorylation, and that this depended on the activation of endogenous PAKs. A requirement for endogenous PAK activity was proven by co-expressing the autoinhibitory domain from PAK1, which blocks the activation of group I PAKs (isoforms 1-3) (24,34,35). Because HEK 293 cells do not express PAK3, 2 the isoforms that phosphorylate NET1 in these cells must be PAKs 1 and/or 2. Both of these kinases play important roles in growth factor-stimulated, Rac1-and Cdc42-dependent signaling to the actin cytoskeleton. In addition, activation of PAK2 by caspases is important for some of the morphological changes associated with apoptosis (reviewed by Bokoch (37)). Thus, down-regulation of NET1 activity by PAKs may be critical for ligand-stimulated morphological changes associated with a number of extracellular stimuli.
Previous work has shown that the ability of Rac1 to negatively regulate RhoA activity is important for controlling Rac1-FIG. 6. Constitutively active Rac1 stimulates the phosphorylation of NET1⌬N and full-length NET1 on serine 152 in a PAKdependent manner. A, HEK 293 cells were transfected with HA epitope-tagged NET1⌬N, minus or plus Myc epitope-tagged, constitutively active Rac1 (V 12 Rac1) and the PAK1 autoinhibitory domain (PAK1 AID). After serum starvation, the cells were lysed in RIPA buffer and the NET1⌬N proteins were immunoprecipitated with an antibody against the HA epitope and tested for phosphorylation on serine 152 by Western blotting. The panels, from top to bottom, show a membrane probed with the anti-pSer-152 antibody, the same membrane reprobed with an anti-HA antibody, and a Western blot of the cell lysate to detect Myc epitope-tagged Rac1 expression. B, HEK 293 cells were transfected with HA epitope-tagged full-length NET1, minus or plus Myc epitopetagged V 12 Rac1 and the PAK1 AID, and tested for phosphorylation on serine 152 as described in A. dependent cellular differentiation and transformation (28 -30), and recently a biochemical mechanism that contributes to this activity has been described (38). Specifically, Rac1 was shown to stimulate the formation of reactive oxygen species which, in turn, caused the inactivation of the low molecular weight protein tyrosine phosphatase LMW-PTP. This led to an increase in the tyrosine phosphorylation and the activity of p190 Rho-GTPase activating protein, which then down-regulated RhoA activity. In addition, Dan et al. (39) have shown that expression of PAK5 also down-regulates RhoA activity, although a mechanism accounting for this regulation was not identified (39). We have not been able to demonstrate a global down-regulation of RhoA activity following the expression of constitutively active PAK1 (data not shown), making it unlikely that phosphorylation of NET1 by PAK1 accounts for the general decrease in total GTP-RhoA levels observed by others. However, the downregulation of NET1 activity that we observe may provide for a localized decrease in RhoA activation under conditions in which NET1 normally regulates RhoA.
Our work also provides an added level of control to cellular mechanisms that generate new actin filaments and regulate their assembly into stress fibers. Previous work suggests that in NIH 3T3 cells RhoA utilizes two key effectors in the production of stress fibers, Rho kinase (ROCK) and the mammalian Diaphanous-related formin mDia1 (33). ROCK is thought to contribute to stress fiber assembly by phosphorylating proteins, such as the myosin-binding subunit of protein phosphatase type 1, that regulate the bundling of actin filaments into stress fibers, and its activation results in the formation of stellate actin bundles contracted at the cell center (40). The formin mDia1 is thought to contribute to stress fiber formation by nucleating non-branched actin filaments, and the activities of ROCK and mDia1 have been proposed to cooperate to produce fully functional stress fibers (41). In support of this, coexpression of the so-called "activated" versions of ROCK and mDia1 together can lead to a profound induction of stress fiber FIG. 7. Coexpression of constitutively active PAK1 blocks F-actin accumulation stimulated by expression of full-length NET1, but not by NET1 S152A/S153A. A, NIH 3T3 cells maintained in 0.05% fetal bovine serum were microinjected with expression plasmids for HA epitope-tagged, full-length NET1 (top panels), NET1 S152E/S153E (middle panels), or NET1 S152A/S153A (bottom panels). Four hours later the cells were fixed and stained for F-actin (left column) and for NET1 expression (anti-HA(12CA5), right column). Shown is a representative experiment. B, serum-starved NIH 3T3 cells were microinjected with expression plasmids for HA epitope-tagged, full-length NET1 (top panels), or NET1 S152A/S153A (bottom panels), with or without Myc epitope-tagged, constitutively active PAK1 (PAK1*; staining shown in insets). C, quantification of F-actin accumulation stimulated by NET1 expression, as described in A and B. Error bars are the S.E. Ϯ mean. Asterisks over the NET1 and NET1 ϩ PAK1* bars show that the difference in mean fluorescence intensity between these samples is significant, as determined by Student's t test (p Ͻ 0.01).
formation that mimics the effects of expression of either activated RhoA or NET1⌬N. In the present study, we observed that NET1 S152A/S153A co-expression with active PAK1 could "rescue" F-actin accumulation (quantified as TRITC-phalloidin staining), but not stress fiber assembly. This result suggests that PAK1 does not affect the ability of the formin to nucleate and elongate actin filaments but does interfere with or counteract signals that drive filament assembly into stress fibers. Thus, the ability of PAK1 to interfere with stress fiber formation (stimulated by NET1) is likely because of disruption of ROCK-dependent effects on actin stress fiber formation.
Although NET1 potently stimulates actin polymerization in the cell, a role for NET1 in normal cell physiology has not yet been determined. Recent studies indicate that the cellular activity of overexpressed NET1 is controlled through nuclear localization, such that the full-length protein exists in the nucleus, sequestered from cytoplasmic and plasma membranebound pools of RhoA (23). Although the authors of this study were unable to demonstrate ligand-dependent export of fulllength NET1 from the nucleus, they did show that the cytoplasmic, oncogenic form of NET1 (NET1⌬N) accumulates in the nucleus when nuclear export is inhibited. This suggests that full-length NET1 may shuttle between the cytoplasm and nucleus. Our results suggest that NET1 may be negatively regulated by cytoplasmic PAK proteins when nuclear export occurs. Alternatively, PAK1 may negatively regulate nuclear pools of NET1 at latter stages in the cell cycle. For example, PAK1 is also active during mitosis, at which time it translocates to the nucleus (42)(43)(44). Thus, it is possible that PAK1 negatively regulates NET1 activity at this point in the cell cycle. There is precedence for the activity of a Rho GEF being crucial for cell division, because the Rho GEF ECT2 is required for cytokinesis in HeLa cells (45,46). Thus, although a mitotic role for NET1 has yet to be described, it is clear that precise regulation of RhoA activity is critical during mitosis and cytokinesis (46 -49), and multiple Rho GEFs may contribute to this regulation. Future experiments will explore the possible role of NET1 and its control by PAK1 at latter stages in the cell cycle.