Human p21-activated kinase (Pak1) regulates actin organization in mammalian cells

202

Research Paper

Human p21-activated kinase (Pak1) regulates actin organization in mammalian cells Mary Ann Sells*, Ulla G. Knaus†, Shubha Bagrodia‡, Diane M. Ambrose*, Gary M. Bokoch† and Jonathan Chernoff* Background: The Rho family GTPases Cdc42, Rac1 and RhoA regulate the reorganization of the actin cytoskeleton induced by extracellular signals such as growth factors. In mammalian cells, Cdc42 regulates the formation of filopodia, whereas Rac regulates lamellipodia formation and membrane ruffling, and RhoA regulates the formation of stress fibers. Recently, the serine/threonine protein kinase p65pak was isolated by virtue of its ability to bind to activated Cdc42 and Rac1. Upon binding, p65pak autophosphorylates, thereby increasing its catalytic activity towards exogenous substrates. This kinase is therefore a candidate effector for the changes in cell shape induced by growth factors. Results: Here, we report that the microinjection of activated Pak1 protein into quiescent Swiss 3T3 cells induces the rapid formation of polarized filopodia and membrane ruffles. The prolonged overexpression of Pak1 amino-terminal mutants that are unable to bind Cdc42 or Rac1 results in the accumulation of filamentous actin in large, polarized membrane ruffles and the formation of vinculin-containing focal complexes within these structures. This phenotype resembles that seen in motile fibroblasts. The amino-terminal Pak1 mutant displays enhanced binding to the adaptor protein Nck, which contains three Src-homology 3 (SH3) domains. Mutation of a proline residue within a conserved SH3-binding region at the amino terminus of Pak1 interferes with SH3-protein binding and alters the effects of Pak1 on the cytoskeleton.

Addresses: *Fox Chase Cancer Centre, 7701 Burholme Avenue, Philadelphia, Pennsylvania 1911, USA. †Departments of Immunology and Cell Biology, Research Institute of Scripps Clinic, La Jolla, California 92037, USA. ‡Department of Pharmacology, Cornell University, Ithaca, New York 14853, USA. Correspondence: Jonathan Chernoff E-mail: [email protected] Received: 15 November 1996 Revised: 23 January 1997 Accepted: 7 February 1997 Published: 18 February 1997 Electronic identifier: 0960-9822-007-00202 Current Biology 1997, 7:202–210 © Current Biology Ltd ISSN 0960-9822

Conclusions: These results indicate that Pak1, acting through a protein that contains an SH3 domain, regulates the structure of the actin cytoskeleton in mammalian cells, and may serve as an effector for Cdc42 and/or Rac1 in promoting cell motility.

Background Many growth factors, in addition to stimulating cell division, also cause rapid changes in cell shape. The Rho family of small GTPases, which includes Cdc42, Rac1 and RhoA, seems to be essential in directing actin-reorganization pathways that result in these changes in cell morphology, characterized by the formation of filopodia, lamellipodia, membrane ruffles and stress fibers [1– 4]. Similar proteins are required for the establishment and maintenance of cell polarity in budding and fission yeast [5,6]. The molecular details of how these Rho family GTPases are regulated are not well understood, although several candidate regulatory proteins have been identified [7,8]. Even less is known about the mechanism by which the activated GTPases transmit signals to the actin cytoskeleton. Recently, two protein kinases were isolated by virtue of their ability to bind to activated Rac1. Both human p120ack [9] and rat p65pak [10] bind preferentially to activated (GTP-bound) Rac1 and Cdc42, inhibit subsequent GTPase-activating protein (GAP) activity on these proteins, and, at least in the case

of p65pak, autophosphorylate upon binding — thereby increasing their catalytic activity towards exogenous substrates [10]. These kinases are therefore candidate effector proteins that cause cell shape changes in response to growth factors.

Results In order to test whether p65pak or related proteins are part of the signaling apparatus for cell morphogenesis, we examined the effects of a human p65pak protein on actin reorganization. We used the polymerase chain reaction to identify Pak-like gene fragments from a human lymphocyte cDNA library, and subsequently cloned the fulllength cDNAs from the same library. One of the clones encoded a 544 amino-acid protein, Pak1, which has 98% identity to rat p65pak and which is expressed primarily in brain, skeletal muscle and spleen. A second clone encoded a ubiquitously expressed, 525 amino-acid protein, termed Pak2, with 78% identity to rat p65pak (data not shown). Pak1 has been well-characterized biochemically [10,11], and so we chose to examine in detail the function of this

Research Paper Pak1 regulates actin organization Sells et al.

Figure 1

(a) 0 min

(b) 10 min

(c) 20 min

(d) 25 min

time-lapse microscopy. GST–Pak1 is constitutively active in vitro [11], possibly because of conformational changes induced at the amino terminus of Pak1 by fusion with GST (see below). Cells injected with GST–Pak1 rapidly developed filopodia and lamellipodia at a leading edge and retraction fibers at a trailing edge, suggesting that GST–Pak1 stimulates cell movement (Fig. 1). These effects were observed in more than 80% of injected cells. The simultaneous development of filopodia and retraction fibers has also been reported in Cdc42-injected cells [3]. The injection of GST or buffer alone had no discernible effect on cell morphology. These effects were not inhibited by co-injection of GST–Pak1 with dominant-negative (N17) forms of Cdc42 or Rac1 (data not shown), consistent with the idea that Pak1 functions downstream of these GTPases in an actin-reorganization signaling pathway. This notion is supported by the observation that endogenous Pak1 colocalizes with filamentous (F) actin to filopodia, lamellipodia and membrane ruffles in Swiss 3T3 cells treated with platelet-derived growth factor (S. Dharmawardhane, R.H. Daniels and G.M.B., unpublished observations); increased physical association of Pak1 with F-actin is also observed under these conditions.

GST–Pak1 induces the formation of filopodia and lamellipodia in Swiss 3T3 cells. Quiescent Swiss 3T3 cells were injected with 200 mg ml–1 GST–Pak1. At the indicated intervals, phase-contrast micrographs were obtained. Arrows indicate lamellipodia and arrowheads indicate filopodia. Scale bar equals 10 mm.

kinase by testing its ability to carry out some of the known effector functions of Cdc42 and Rac1. We first expressed Pak1 in Swiss 3T3 cells, and assessed the effects on cell morphology. Quiescent cells were microinjected with a fusion protein of Pak1 and glutathione-S-transferase (GST–Pak-1), and studied by

As GST–Pak1 is susceptible to proteolytic degradation during production and storage [11], we asked whether the sustained expression of Pak1 from a plasmid vector induced similar effects on cell morphology, and which features of the protein were required for these events. We made Pak1 with point mutations that abolish binding to Cdc42 and Rac1 and/or alter kinase activity (Fig. 2). Pak1L83,L86, with leucine residues were substituted for two highly conserved histidine residues within the p21-binding domain (PBD), was unable to bind to Cdc42 or Rac1 and had enhanced kinase activity towards itself (Fig. 2) and exogenous substrates (data not shown). A catalytically inactive form of this PBD mutant (Pak1L83,L86,R299) was constructed by replacing an essential lysine in the protein kinase subdomain II with arginine. A second activated form of Pak1 (Pak1E423) was generated by replacing a threonine

An ti-M Cd yc c blo Ra 42 o t c1 ve Au ove rlay to rla kin y as e

Figure 2 Activity and interactions between Pak1 and Rho family GTPases. COS1 cells were transfected with pCMV6M bearing sequences encoding the indicated forms of Pak. Anti-Myc immunoprecipitates were assayed for the presence of Pak1 by immunoblot, for binding to Rac1 and Cdc42 by an overlay assay, and for autokinase activity.

203

PBD

Kinase domain

– Wild-type

L83, L86

L83, L86 L83, L86

R299

L83, L86, R299 E423

E423

204

Current Biology, Vol 7 No 3

residue in kinase subdomain VIII with glutamic acid. Similar activating mutations have been made in other serine/threonine kinases [12], and imply that this residue may serve as a regulatory phosphorylation site. The ability of Pak1E423 to bind to Cdc42 and Rac1 was unimpaired.

Figure 3

We compared the morphological effects of expressing Rho family GTPases or Pak1 in quiescent Swiss 3T3 cells. In agreement with previous reports [1–4], cells expressing constitutively active mutants of Cdc42, Rac1 and RhoA predominantly formed filopodia, circumferential lamellipodia and increased stress fibers, respectively (Fig. 3f–h). Expression of wild-type Pak1 (tagged with a Myc epitope) had little effect on actin distribution (Fig. 3b), whereas expression of Myc-tagged Pak1L83,L86 induced a distinct phenotype, characterized by increased stress fibers and large, protruding, ruffle-like structures positioned at one side of the cell (Fig. 3c). This distinctive morphology, similar to that of motile fibroblasts, was observed in more than 70% of the transfected cells, and identical results were obtained using a different cell type (REF52) and a different transfection method (plasmid microinjection). Omission of the small, amino-terminal epitope tag gave equivalent results (data not shown). Confocal imaging revealed that a substantial fraction of the Pak1L83,L86 mutant protein was concentrated within these ruffle-like structures, where it colocalized with F-actin (Fig. 4) and vinculin in focal complexes (data not shown). The remainder of the Pak1L83,L86 appeared to be distributed throughout the cytoplasm, most prominently in the perinuclear region. As Pak1L83,L86 does not bind to Rac1 or Cdc42, the observed effects on actin distribution are unlikely to be caused by increased p21–GTP levels resulting from the protection of these GTPases from GAPs [10]. The subcellular distributions of wild-type Pak1 and the PBD mutant were grossly similar (data not shown), and so the effects of these proteins on cytoskeletal architecture may instead be due to differences in kinase activity, or to differences in the conformation of the amino terminus, as reflected by the inability of Pak1L83,L86 to bind p21 GTPases (Fig. 2). In order to distinguish between these possibilities, we tested the effects of expressing the activated Pak1E423 mutant, which can bind p21. Cells expressing Pak1E423 were flattened and occasionally showed filopodia (1–2 per cell) and complete loss of stress fibers (Fig. 3d), but were otherwise similar to untransfected quiescent cells (Fig. 3a). In contrast, the kinase-dead Pak1L83,L86,R299 mutant, which cannot bind p21, induced the formation of unipolar ruffles (Fig. 3e), similar to those observed in cells expressing Pak1L83,L86 (Fig. 3c), although the cells expressing Pak1L83,L86,R299 were more elongated and crescent-shaped. In Swiss 3T3 cells, therefore, Pak1-mediated actin reorganization appears to have two components: a generalized effect

Effect of Pak1 on filamentous actin distribution. Swiss 3T3 cells were transfected with (a) pCMV6 alone, or pCMV6 bearing sequences encoding (b) Myc–Pak1, (c) Myc–Pak1L83,L86, (d) Myc–Pak1E423, (e) Myc–Pak1L83,L86,R299, (f) HA–Cdc42L61, (g) HA–RacL61 or (h) HA–RhoL63. 24 h post transfection, cells expressing recombinant proteins, which were tagged with either the hemagglutinin (HA) or Myc peptide epitope at the amino terminus, were identified by indirect immunofluorescence. For each transfection, at least 100 antigen-positive cells were examined. Each panel shows the actin distribution of a typical cell expressing recombinant protein. Scale bar equals 10mm.

on cell shape that is dependent on the protein kinase activity of Pak1; and an effect on F-actin accumulation in localized, ruffle-like structures which requires conformational changes at the amino terminus of Pak1, but which is largely

Research Paper Pak1 regulates actin organization Sells et al.

205

Figure 4 Co-localization of Pak1 and F-actin. Swiss 3T3 cells were transfected with pCMV6 encoding Myc–Pak1L83,L86 and processed as described in the legend to Figure 3. The distribution of (a) F-actin and (b) Myc epitope was examined by indirect immunofluorescence, using confocal microscopy. A combined image, shown in color (c), indicates areas of colocalization in yellow. Scale bar equals 10 mm.

independent of its protein kinase activity. For wild-type Pak1, such conformational changes presumably occur in vivo upon binding to activated Cdc42 or Rac1. Microinjected GST–Pak1 protein, which has no mutation in the PBD, does induce actin remodeling (Fig. 1); this effect may be related to unfolding of the amino terminus resulting from fusion to the bulky GST moiety, a possibility supported by the finding that GST–Pak1 is constitutively active in vitro [11]. Which non-kinase domain on Pak1 might be affected by binding to Rac1 or Cdc42? Amino-terminal to the PBD, all known mammalian Paks, as well as Drosophila and Caenorhabditis elegans Pak1, have a highly conserved, proline-rich motif which directs interactions with proteins that contain a Src-homology 3 (SH3) domain, such as Nck [13–15]. We compared the ability of wild-type Pak1 and the Pak1L83,L86 PBD mutant to bind to Nck in vitro. Lysates from cells expressing wild-type and mutant Pak1 were overlayed on nitrocellulose filters containing defined amounts of GST–Nck, and then analyzed for bound Pak1 by western immunoblotting. Pak1L83,L86 bound much more avidly than wild-type Pak1 to GST–Nck, as judged by the intensity of the signal on the immunoblots (Fig. 5a). No binding of Pak1 was detected when GST alone was spotted on the filter in place of GST–Nck. When the lysates were preincubated with activated recombinant Rac1, the binding of wild-type Pak1 to GST–Nck increased about four-fold, whereas the binding of Pak1L83,L86 was unaffected (data not shown). Similar results were obtained in vivo (Fig. 5b). Endogenous Nck was immunoprecipitated from cells transfected with either wild-type Pak1 or Pak1L83,L86 (both were tagged with a hemagglutinin (HA)-epitope), and the immunoprecipitates were examined by anti-HA immunoblotting. Although the wild-type and mutant Pak1 were expressed at approximately equal levels, approximately four-fold more of the mutant Pak1 co-immunoprecipitated with Nck. These data suggest that the proline-rich domain at the amino terminus of Pak1 becomes more accessible to SH3 proteins following the activation of Pak1 by Rac1, and that this phenomenon can be elicited in the absence of Rac1 by mutations within the nearby PBD.

In order to determine whether the amino-terminal, proline-rich region of Pak1 mediates any of its effects on the cytoskeleton, we compared actin organization and focal complex formation in cells transfected with the PBD mutant Pak1L83,L86 or with Pak1A13,L83,L86; this latter form of Pak1 has an additional mutation within the amino-terminal, proline-rich region that completely abolishes binding to Nck [14,15] and to several other proteins that contain SH3 domains (J.C., unpublished observations). When expressed in Swiss 3T3 cells, Pak1A13,L83,L86 caused an increase in stress fibers, but failed to induce polarized membrane ruffles in the majority of cells (Fig. 6k). These data indicate that the formation of polarized membrane ruffles requires the amino-terminal, proline-rich region of Pak1. As the subcellular distribution of both Pak1 mutants was similar (data not shown), the lack of membrane ruffling in cells expressing Pak1A13,L83,L86 does not appear to be due to a gross mislocalization of this protein. It is possible, however, that a subtle mislocalization of Pak1 would be missed in these analyses, as the high level of expression used might obscure such differences. The expression of either Pak1L83,L86 or Pak1A13,L83,L86 induced the formation of vinculin-containing focal complexes, but these complexes were much more prominent in cells expressing Pak1L83,L86, and were concentrated within the ruffles of such cells (Fig. 6j,l). The vinculincontaining focal complexes in cells expressing Pak1L83,L86 were comparable in size and shape to those seen in cells expressing Cdc42 or Rac1 (Fig. 6b,d), but were distinct from the larger, arrow-head shaped adhesion complexes induced by RhoA (Fig. 6f). Interestingly, the increase in stress fibers could be blocked by coexpression of dominant-negative (N19) RhoA (data not shown). These results suggest that the increase in stress fibers seen with these Pak1 mutants is mediated by the activation of endogenous Rho.

Discussion The results reported here indicate that Pak1 modulates the structure of the actin cytoskeleton and may serve as a morphogenic effector for Cdc42 and/or Rac1 in vivo. This

206

Current Biology, Vol 7 No 3

Figure 5 (a)

Pak1 (Wild-type)

2

1

0.5

0.25

0.125

0.05

0.025

0.0125

0.005

Blank

2

1

0.5

0.25

0.125

0.05

0.025

0.0125

0.005

Blank

Effect of Pak1 PBD mutations on binding to Nck. (a) The indicated amounts (in mg) of GST–Nck SH3 domain 2 were spotted onto nitrocellulose. Extracts from COS1 cells transfected with wild-type Pak1 or Pak1L83,L86 were then overlayed as described in Materials and methods, and immunoblotted with monoclonal anti-Myc. To the left, an anti-Myc immunoblot is shown of the Pak1-containing lysates used for the overlay. (b) Lysates from NIH-3T3 cells stably transfected with HAtagged wild-type Pak1 or Pak1L83,L86 were immunoprecipitated with anti-Nck antibodies and immunoblotted with monoclonal anti-HA.

Pak1 (L83, L86)

(b)

Wild-type

L83, L86

Wild-type

L83, L86

Anti-HA immunoblot

Anti-Nck IP

Lysate

conclusion is based on the ability of Pak1 mutants, when introduced into cells, to stimulate the formation of polarized lamellipodia, membrane ruffles and filopodia. These effects are independent of the action of Rac1 and/or Cdc42, as dominant-negative (N17) forms of these GTPases were unable to block Pak1-induced cytoskeletal responses. Also, like Rac and Cdc42, Pak1 induces the formation of vinculin-containing focal complexes. Although we and others have shown that mammalian Paks mediate activation of stress-activated protein kinase cascades by Rac1/Cdc42 [16–19], probably via a MEKK, the relevance of this activation to cytoskeletal rearrangements or cell motility is not clear. In this study, we have shown that the ability of Pak1 to regulate the actin cytoskeleton is in part independent of its intrinsic serine/threonine kinase activity. Pak1 constructs with a modified amino terminus have dramatic effects on the polarity of the cytoskeleton in Swiss 3T3 cells, whether or not they have an active catalytic domain. Conversely, Pak1 that is not modified at the amino terminus, but that has high levels of constitutive kinase activity, lacks these effects. The ability of Pak1 to regulate polarized actin assembly appears to depend on an aminoterminal, proline-rich domain in Pak1 which has a predicted

SH3-binding motif, PXXP (single-letter amino-acid code, X indicating any amino acid); mutation of this SH3-binding domain inhibits the ability of Pak1 to induce the formation of polarized membrane ruffles and focal complexes. The importance of the amino-terminal PXXP domain for cytoskeletal regulation by Pak1 is shown by the enhanced binding of the highly active Pak1L83,L86 mutant and Cdc42activated, wild-type Pak1 to the SH3-containing Nck. Furthermore, it has been established that this domain serves as a functional SH3-binding site in vivo, interacting with the second SH3 domain of the Nck adaptor protein [14,15]. Our data suggest that Pak1 binds a protein which contains an SH3 domain and which regulates its localization and/or affects actin dynamics. Interestingly, an analogous situation has been described in Saccharomyces cerevisiae [20]. In this species, a protein known as Bem1p binds to the yeast Pak homolog, Ste20p, through an SH3 domain present on Bem1p ([20] and E. Leberer, personal communication). This interaction with Bem1p mediates a physical association of Ste20p with F-actin and contributes to the regulation of cell morphology in yeast. Although no mammalian Bem1p homologs are known, a fission yeast homolog, Scd2p/Ral3p [21], binds to fission yeast Pak1p (J.C.,

Research Paper Pak1 regulates actin organization Sells et al.

207

Figure 6

Focal complex formation in cells expressing wild-type or mutant forms of Pak1. Actin (a,c,e,g,i,k) and vinculin (b,d,f,h,j,l) distribution in transgeneexpressing cells are shown 24 h after transfection of quiescent Swiss 3T3 with pCMV6 bearing sequences encoding (a,b) HA–Cdc42L61, (c,d) HA–RacL61, (e,f) HA–RhoAL63, (g,h) Myc–Pak1, (i,j) Myc–-Pak1L83,L86 or (k,l) Myc–Pak1A13,L83,L86. Scale bar equals 10 mm.

unpublished observations). Therefore, it is plausible that mammalian Paks are linked to the cytoskeleton via a Bem1p-like protein. This physical association with actinbinding proteins might give Pak access to key substrates, perhaps within focal adhesions. Alternatively, as suggested by the cytoskeletal activity of the kinase-dead Pak-1 protein (Fig. 3e), this enzyme might serve as a scaffold to assemble additional molecules that affect actin structure. From these other observations and our current data, we hypothesize that the binding of activated Rac1–GTP or Cdc42–GTP to Pak1 changes the conformation of its amino terminus such that the SH3-binding domain is exposed and/or has higher affinity for SH3-containing target proteins. These putative target proteins, perhaps serving as the functional equivalent of Bem1p in budding yeast, would enable Pak1 to associate with and/or regulate F-actin. In this

model, Pak1 acts as a bifunctional regulatory molecule, with the amino-terminal PXXP motifs primarily regulating polarization of the actin cytoskeleton through interactions with SH3-containing proteins, and the catalytic carboxyl terminus primarily mediating activation of the stress-activated protein kinase cascades by Rac1/Cdc42 (Fig. 7). However, as the phenotypes elicited by expression of kinase-active and kinase-dead Pak1L83,L86 are not completely equivalent (compare Fig. 3c,e), the kinase activity of this protein may also influence some aspect of F-actin assembly, perhaps indirectly via the phosphorylation of myosin. The established morphogenic roles of Pak homologs in budding and fission yeast [20,22–26], as well as the effects of Pak expression in mammalian cells ([27] and this study), suggest that Paks regulate cytoskeletal structure. Are Paks, then, cytoskeletal effectors for Cdc42 and/or

208

Current Biology, Vol 7 No 3

Figure 7

PXXP NH2 Inactive COOH

Kinase T423

Cdc42/Rac1–GTP

Active

NH2 PXXP SH3 protein

Polarization, actin remodeling

Myosin ?

Kinase T423 MEKK? P

COOH

SAPK cascade

that the activation of JNK induced by Rac1A37,L61 or Cdc42A37,L61 in COS cells is mediated through a kinase other than Pak1. Indeed, the relative levels of Paks versus mixed-lineage kinases have not been described in COS cells or in murine fibroblasts. Thus, activated Pak1 could mediate some of the cytoskeletal effects of Rac1 or Cdc42 in specific cell types. A similar argument has been proposed by Lamarche et al. [22] to account for the actin-reorganizing activity of the Wiskott–Aldrich syndrome protein (WASP), which, like Pak1, binds to Cdc42 via a conserved PBD [34]. It is also possible that, in some circumstances, Pak1 can regulate the actin cytoskeleton independently of activation by GTPases — for example, by recruitment to the plasma membrane in response to Nck translocation stimulated by growth factors [14,35].

© 1997 Current Biology

A model for Pak1 activity. Upon binding to Cdc42 or Rac1, a conformational change in Pak1 may unmask its amino-terminal prolinerich domain as well as cause increased kinase activity. The exposed proline-rich domain may then engage in productive interactions with SH3-containing proteins which dictate localization and/or mediate some of the cytoskeletal effects of Pak1. The kinase domain of Pak1 may also affect cytoskeletal structure through phosphorylation of targets such as myosin, as well as mediate the activation of stressactivated protein kinase (SAPK) cascades, probably via a MEKK.

Rac? Expression of Pak1 in Swiss 3T3 cells induces a phenotype that differs from that elicited by Rac1 and Cdc42. Pak1-induced filipodia and lamellipodia are polarized to the leading edge of the cell, rather than symmetrically distributed, as is the case for Cdc42 and Rac1. The resulting appearance of such cells is similar to that seen in motile fibroblasts. In addition, recent mutational analyses of Cdc42 and Rac1 have cast doubt on the roles of Pak or other PBD-containing proteins in mediating the cytoskeletal effects of these GTPases. Mouse fibroblasts expressing Rac1 or Cdc42 mutants that cannot activate Pak nevertheless form filopodia and circumferential membrane ruffling, respectively [28,29]. Furthermore, a second set of Rac1 or Cdc42 mutants that are expected to retain the ability to activate Pak fail to induce formation of these structures. However, neither of these experiments preclude the possibility that Paks regulate the organization of cortical actin and influence focal complex assembly. For example, while Rac1A37,L61 and Cdc42A37,L61 (which fail to induce ruffles or filopodia, respectively) were shown to be capable of binding Pak1 in vitro, and of activating the Jun N-terminal kinase (JNK) pathway in transfected COS cells, there is as yet no evidence that endogenous Pak1 is actually activated by these mutant GTPases in murine fibroblasts. In addition to the three known mammalian Paks, Rac1 and Cdc42 bind several other PBD-containing protein kinases, such as ‘mixed lineage’ kinases, many of which also can activate the JNK pathway [30–33]. Therefore, it is possible

The phenotype of Swiss 3T3 cells expressing Pak1L83,L86 most closely resembles that seen during cell locomotion, and preliminary data indicate that such cells in fact display altered motility (M.A.S. and J.C., unpublished observations). The mechanism for some of these effects appears to involve the interaction of Pak1 with an as yet uncharacterized SH3-containing actin regulatory protein, perhaps related in structure or function to budding yeast Bem1p. It is also possible that some of the effects of Pak1 on the actin cytoskeleton are mediated indirectly through myosin. In Acanthamoeba and Dictyostelium, Pak homologues have recently been shown to phosphorylate myosin heavy chain, stimulating its Mg⋅ATPase activity [36,37]. Myosin I contains an SH3 domain, which might provide a binding site for Pak1 [38]. Although vertebrate myosin I proteins lack an equivalent activating phosphorylation site [39], a related isoform, myosin VI, does contain this site [38]. It is also interesting to note that certain forms of myosin II [40] and Pak1 (S. Dharmawardhane, R.H. Daniels and G.M.B., unpublished observations) have been localized to the leading edge of polarized cells. Finally, a known Pak1-binding SH3 protein, such as Nck, may be involved in the accumulation of F-actin in mammalian cells (J.Y. Bu, R. Pappu, B. Mayer, J.C., D. Straus, and A.C. Chan, unpublished observations). The Cdc42binding protein WASP also binds to Nck, and thus, with Pak1, potentially forms part of a morphogenic signaling apparatus in some cell types [34,41]. The identification of proteins that interact with Pak1 promises to extend our understanding of the role of these enzymes in eukaryotic cell morphogenesis.

Conclusions In Swiss 3T3 cells, the expression of Pak1 proteins bearing mutations at the amino terminus induces the polarized redistribution of F-actin and the formation of focal complexes — a phenotype similar to that seen in motile fibroblasts. Many of these effects appear to be mediated by an SH3-binding motif located at the amino terminus of Pak1. We propose that these mutant forms of

Research Paper Pak1 regulates actin organization Sells et al.

Pak1 mimic the effects of the endogenous, wild-type protein, which, upon binding to activated Cdc42 or Rac1, undergoes a conformational change that increases exposure of the SH3-binding motif, allowing productive interactions with downstream cytoskeletal effectors.

Materials and methods Cell Culture COS1 and Swiss 3T3 cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% FBS. COS1 cells were transfected using Lipofectamine and Swiss 3T3 cells were transfected by a calcium phosphate method [42].

Protein preparation and microinjections Human Pak1 and Pak2 were cloned by standard methods from a human lymphoblastoid cDNA library [43] (GenBank accession numbers: Pak1, U24152; Pak2, U24153). Bacterial expression vectors were created by subcloning cDNAs into pGEX-2T [44], and GST fusion proteins were purified by standard techniques, and dialyzed against phosphate-buffered saline. Cells were injected with 100 mg ml–1 GST–Pak1, with or without 500 mg ml–1 N17-Rac1, N17Cdc42, or N19-RhoA, and 500 mg ml–1 rat immunoglobulin G (IgG). For time-lapse experiments, the cells were photographed at the indicated intervals. In co-injection experiments, the cells were fixed 15 min post-injection, and stained for the presence of IgG and for F-actin.

Eukaryotic expression vectors cDNAs encoding Pak1 and p21 GTPases were subcloned into the epitope tag expression vectors pJ3M and pJ3H, respectively [45], then subcloned as SalI–EcoRI fragments into pCMV6 (a modified version of pCMV5) [46]. For some experiments, HA-tagged Pak1 was also used.

Immunofluorescence Cells were fixed in 3.5% paraformaldehyde for 15 min, permeabilized in 0.1% Tween-20, and blocked with 3% BSA. In Figure 1, cells were stained for the presence of IgG (Lissamine-Rhodamine B-conjugated goat anti-rat IgG) and for filamentous actin (0.1 mM FITC-phalloidin) for 45 min. In Figure 3, cells were stained with mouse anti-epitope tag antibodies, followed by Lissamine-Rhodamine-conjugated goat anti-mouse antibodies and FITC-conjugated phalloidin. In Figure 6, cells were stained with with rabbit anti-epitope tag and mouse antivinculin antibodies, followed by Lissamine Rhodamine-conjugated anti-rabbit antibodies (to localize cells expressing recombinant protein), 0.1 mM FITC-conjugated phalloidin, and AMCA-conjugated anti-mouse antibodies (to localize vinculin). Confocal microscopy was performed using a Bio-Rad MRC 600 laser scanning confocal microscope.

Binding studies COS1 cells (maintained in DMEM) supplemented with 10% FBS) were transfected with control or protein-encoding plasmid DNA. 48 h posttransfection, the cells were lysed in a buffer containing 50 mM Tris-HCl pH 8.0, 137 mM NaCl, 1% NP40, 10% glycerol, 10 mM b-glycerol phosphate, 1 mM Naávanadate, 1 mM PMSF, 10 mg ml–1 aprotinin. A portion of the lysate was boiled in an equal volume of 2× SDS/PAGE sample buffer, and the remainder was used either for in vitro binding studies, by incubation with GST–p21 beads for 2 h at 4°C, or immunoprecipitated with either anti-HA monoclonal antibody 12CA5 (BabCO) or anti-Nck polyclonal sera (a gift from Ed Skolnik). Lysates and immunoprecipitates were analyzed by immunoblot following SDS/10% PAGE and transferred to nitrocellulose or polyvinylidine flouride (Immobilon) membranes. Signals were detected using enhanced chemiluminescence (Pierce). Rac1 and Cdc42 overlay assays (Fig. 2) were performed essentially as described in [47] using 2 mg ml–1 g[35S]GTP-loaded Cdc42 or Rac1 for 20 min at 4°C, then 10 min at room temperature. After washing to remove non-specific binding, the blots were exposed to Kodak XAR film at –70°C for 24 h.

209

Pak1 overlay binding assays (Fig. 5a) were carried out by spotting the indicated amount of GST–Nck-SH3 domain 2 (amino acids 101–166) onto a nitrocellulose membrane, blocking with 5% non-fat milk, then incubating the membrane at 4°C for 30 min with lysates from COS1 cells transfected with Myc-tagged Pak1 or Pak1L83,L86. The membrane was then washed extensively and immunoblotted with anti-Myc antibodies. Signals were detected using enhanced chemiluminescence.

Kinase assays Kinase activity was assessed by incubating anti-Myc immunoprecipitates in a buffer containing 25 mM Tris-HCl pH 7.5, 10 mM MgCl2, 20 mM g-[32P]ATP (5 000 cpm pmol–1) for 20 min at 30°C. The reaction was terminated with SDS/PAGE sample buffer, separated by 10% SDS/PAGE, blotted to a PVDF membrane and autoradiographed at –70°C for 16 h.

Acknowledgements We thank Erica Golemis for comments on the manuscript, Steve Elledge for the gift of the human lYES cDNA library, and Ed Skolnik for anti-Nck sera. This work was supported by grants from the ACS (J.C.), NCI (D.M.A.), and NIH (G.M.B., U.K., and S.B.), and an appropriation from the Commonwealth of Pennsylvania.

References 1. Ridley AJ, Hall A: The small GTP-binding protein rho regulates the assembly of focal adhesions and actin stress fibers in response to growth factors. Cell 1992, 70:389–399. 2. Ridley AJ, Paterson HF, Johnston CL, Diekmann D, Hall A: The small GTP-binding protein rac regulates growth factors-induced membrane ruffling. Cell 1992, 70:401–410. 3. Kozma R, Ahmed S, Best A, Lim L: The Ras-related protein Cdc42Hs and bradykinin promote formation of peripheral actin microspikes and filopodia in Swiss 3T3 cells. Mol Cell Biol 1995, 15:1942–1952. 4. Nobes CD, Hall A: Rho, Rac, and Cdc42 GTPases regulate the assembly of multi-molecular focal complexes associated with actin stress fibers, lamellipodia and filopodia. Cell 1995, 81:53–62. 5. Drubin DG: Development of cell polarity in budding yeast. Cell 1991, 65:1093–1096. 6. Chant J: Cell polarity in yeast. Trends Genet 1994, 10:328–333. 7. Hall A: Small GTP-binding proteins and the regulation of the actin cytoskeleton. Annu Rev Cell Biol 1994, 10:31–54. 8. Ridley AJ: Rho-related proteins: actin cytoskeleton and cell cycle. Curr Opin Genet Dev 1995, 24:24-30. 9. Manser E, Leung T, Salihuddin H, Tan L, Lim L: A non-receptor tyrosine kinase that inhibits the GTPase activity of p21cd42. Nature 1993, 363:364–367. 10. Manser E, Leung T, Salihuddin H, Zhao Z-s, Lim L: A brain serine/threonine protein kinase activated by Cdc42 and Rac1. Nature 1994, 367:40–46. 11. Knaus UG, Morris S, Dong H-J, Chernoff J, Bokoch GM: Regulation of p21-activated kinases by human leukocyte G protein coupled receptors. Science 1995, 269:221–223. 12. Johnson LN, Noble MEM, Owen DJ: Active and inactive protein kinases — structural basis for regulation. Cell 1996, 85:149–158. 13. Bagrodia S, Taylor S, Creasy CL, Chernoff J, Cerione RA: Identification of a murine p21Cdc42/Rac activated protein kinase (PAK). J Biol Chem 1995, 270:22731–22738. 14. Galisteo ML, Chernoff J, Su Y-C, Skolnik EY, Schlessinger J: The adaptor protein Nck links tyrosine kinases with the serinethreonine kinase Pak1. J Biol Chem 1996, 271:20997–21000. 15. Bokoch GM, Wang Y, Bohl BP, Sells MA, Quilliam LA, Knaus UG: Interaction of Nck adapter protein with p21-activated kinase (PAK1). J Biol Chem 1996, 271:25746–25749. 16. Zhang S, Han J, Sells MA, Chernoff J, Knaus UG, Ulevitch RJ, Bokoch GM: Rho family GTPases regulate p38 mitogen-activated protein kinase through the downstream mediator Pak1. J Biol Chem 1995, 270:23934–23936. 17. Bagrodia S, Derijard B, Davis RJ, Cerione RA: Cdc42 and Pakmediated signaling leads to Jun kinase and p38 mitogenactivated protein kinase activation. J Biol Chem 1995, 270:27995–27998.

210

Current Biology, Vol 7 No 3

18. Polverino A, Frost JP, Yang P, Hutchinson M, Neiman AM, Cobb MH, Marcus S: Activation of mitogen-activated protein kinase cascades by p21-activated protein kinases in cell-free extracts of Xenopus oocytes. J Biol Chem 1995, 270:26067–26070. 19. Brown JL, Stowers L, Maer M, Trejo J, Coughlin S, Chant J: Human Ste20 homologue hPak1 links GTPases to the JNK MAP kinase pathway. Curr Biol 1996, 6:598–605. 20. Leeuw T, Fourest-Lieuvin A, Wu C, Chenevert J, Clark K, Whiteway M, Thomas DY, Leberer E: Pheromone response in yeast: association of Bem1p with proteins of the MAP kinase cascade and actin. Science 1995, 270:1210–1213. 21. Chang EC, Barr M, Wang Y, Jung V, Xu H-P, Wigler MH: Cooperative interaction of S. pombe proteins required for mating and morphogenesis. Cell 1994, 79:131–141. 22. Liu H, Styles CA, Fink GR: Elements of the yeast pheromone response pathway required for filamentous growth of diploids. Science 1993, 262:1741–1744. 23. Leberer E, Wu C, Leeuw T, Forest-Lieuvin A, Segall JE, Thomas DY: Functional characterization of the Cdc42p binding domain of yeast Ste20p protein kinase. EMBO J 1997, 16:83–97. 24. Marcus S, Polverino A, Chang E, Robbins D, Cobb MH, Wigler MH: Shk1, a homolog of the Saccharomyces cerevisiae Ste20 and mammalian p65PAK protein kinases, is a component of a Ras/Cdc42 signaling module in the fission yeast Schizosaccharomyces pombe. Proc Natl Acad Sci USA 1995, 92:6180–6184. 25. Ottilie S, Miller PJ, Johnson DI, Creasy C, Sells MA, Bagrodia S, Forsburg SL, Chernoff J: Fission yeast pak1+ encodes a protein kinase that interacts with Cdc42p and is involved in the control of cell polarity and mating. EMBO J 1995, 14:5908–5919. 26. Mosch H-U, Radclyffe RL, Fink GR: Ras2 signals via the Cdc42/Ste20/mitogen-activated protein kinase module to induce filamentous growth in Saccharomyces cerevisiae. Proc Natl Acad Sci USA 1996, 93:5352–5356. 27. Manser E, Huang H-Y, Loo T-H, Chen XQ, Leung T, Lim L: Expression of constitutively active a-Pak reveals effects of the kinase on actin and focal complexes. Mol Cell Biol, in press. 28. Lamarche N, Tapon N, Stowers L, Burbelo PD, Aspenström P, Bridges T, Chant J, Hall A: Rac and Cdc42 induce actin polymerization and G1 cell cycle progression independently of p65Pak and the JNK/SAPK MAP kinase cascade. Cell 1996, 87:519–529. 29. Joneson T, McDonough M, Bar-Sagi D, Van Aelst L: RAC regulation of actin polymerization and proliferation by a pathway distinct from Jun kinase. Science 1996, 274:1374–1376. 30. Fan G, Merritt SE, Kortenjann M, Shaw PE, Holzman LB: Dual leucine zipper-bearing kinase (DLK) activates p46SAPK and p38mapk but not ERK2. J Biol Chem 1996, 271:24788–24793. 31. Burbelo PD, Dreschel D, Hall A: A conserved binding motif defines numerous candidate target proteins for both Cdc42 and Rac GTPases. J Biol Chem 1995, 270:29071–29074. 32. Rana A, Gallo K, Godowski P, Hirai S, Ohno S, Zon L, Kyriakis JM, Avruch J: The mixed lineage kinase SPRK phosphorylates and activates the stress-activated protein kinase activator, SEK-1. J Biol Chem 1996, 271:19025–19028. 33. Teramoto H, Coso OA, Miyata H, Igishi T, Miki T, Gutkind JS: Signaling from the small GTP-binding proteins Rac1 and Cdc42 to the c-Jun N-terminal kinase/stress-activated protein kinase pathway: a role for mixed lineage kinase 3/protein-tyrosine kinase 1, a novel member of the mixed lineage kinase family. J Biol Chem 1996, 271:27225–27228. 34. Symons M, Derry JMJ, Kariak B, Jiang S, Lemahieu V, McCormick F, Francke U, Abo A: Wiskott–Aldrich syndrome protein, a novel effector for the GTPase CDC42Hs, is implicated in actin polymerization. Cell 1996, 84:723–734. 35. Lu W, Katz S, Gupta R, Mayer BJ: Activation of Pak by membrane localization mediated by Nck SH3 domain. Curr Biol 1997, 7:85–94. 36. Brzeska H, Szczepanowska J, Hoey J, Korn ED: The catalytic domain of Acanthamoeba myosin I heavy chain kinase. II. Expression of active catalytic domain and sequence homology to p21-activated kinase (PAK). J Biol Chem 1996, 271:27056–27062. 37. Lee SF, Egelhoff TT, Mahasneh A, Cote GP: Cloning and characterization of a Dictyostelium myosin I heavy chain kinase activated by Cdc42 and Rac. J Biol Chem 1996, 271:27044–27048. 38. Bement WM, Mooseker MS: TEDS rule: a molecular rationale for differential regulation of myosins by phosphorylation of the heavy chain head. Cell Motil Cytoskeleton 1995, 31:87–92.

39. Brzeska H, Korn ED: Regulation of class I and class II myosins by heavy chain phosphorylation. J Biol Chem 1996, 271:16983–16986. 40. Kelley CA, Seller JR, Gard DL, Bui D, Adelstein RS, Baines IC: Xenopus nonmuscle myosine heavy chain isoforms have different subcellular localizations and enzymatic activities. J Cell Biol 1996, 134:675–687. 41. Rivero-Lezcano OM, Marcilla A, Sameshima JH, Robbins KC: Wiskott–Aldrich syndrome protein physically associates with Nck through Src homology 3 domains. Mol Cell Biol 1995, 15:5725–5731. 42. Chen C, Okayama H: High-efficiency transformation of mammalian cell by plasmid DNA. Mol Cell Biol 1987, 7:2745–2752. 43. Elledge SJ, Mulligan JT, Ramer SW, Spottswood M, Davis RW: lYES: A multifunctional cDNA expression vector for the isolation of genes by complementation of yeast and Escherichia coli mutations. Proc Natl Acad Sci USA 1991, 88:1731–1735. 44. Smith DB, Johnson KS: Single-step purification of polypeptides expressed in Escherichia coli as fusions with glutathione Stransferase. Gene 1988, 67:31–40. 45. Sells MA, Chernoff J: Epitope tagging vectors for eukaryotic protein production. Gene 1995, 152:187–189. 46. Andersson S, Davis DL, Dahlbäck H, Jörnvall H, Russell DW: Expression cloning and regulation of steroid 5 alpha-reductase, an enzyme essential for male sexual differentiation. J Biol Chem 1989, 264:8222–8229. 47. Manser E, Leung T, Monfries C, Teo M, Hall C, Lim L: Diversity and versatility of GTPase activating proteins for the p21rho subfamily of ras G proteins detected by anovel overlay assay. J Biol Chem 1992, 267:16025–16028.

Because Current Biology operates a ‘Continuous Publication System’ for Research Papers, this paper has been published via the internet before being printed. The paper can be accessed from http://biomednet.com/cbiology/cub.htm — for further information, see the explanation on the contents page.