The p21 activated kinases (Paks), an evolutionarily conserved family of serine/threonine kinases, are becoming increasingly important for a variety of cellular functions in mammalian cells. Paks act as downstream effectors for the small GTPases, Cdc42 and Rac1. The kinase domain of mammalian Paks is homologous to that of the yeast protein STE20, which has been implicated in pheromone-response pathways (Leberer et al., 1997). Paks regulate cell morphology and polarity in most, if not all, eukaryotic cells (Sells et al., 1999). The Pak family currently consists of six members, Pak1 through Pak6 (Manser et al., 1994; Abo et al., 1998; Sells et al., 1999; Dan et al., 2002; Lee et al., 2002). All the Paks are similar in structure, containing an N-terminal regulatory domain and a C-terminal kinase domain.
Based on the structural and architectural similarities, the Pak family can be grouped into two sub-groups: group I, consisting of Pak 1 (also referred to as αPak), Pak2 (also referred to as γPak), and Pak3 (also referred to as βPak) (> 90% homology in kinase domain and 73% overall homology compared with Pak1); and group 2, consisting of Pak4, Pak5, and Pak6, (∼50% homology with Pak1 kinase domain). Structural differences among different groups have been recently reviewed (Jaffer and Chernoff, 2002). The C-terminal kinase domain represents the region of highest homology among the Paks, and the N-terminal region represents the region of least homology. Differences in the N-terminal region are thought to contribute activation of distinct signaling pathways by recruitment of unique downstream effector molecules.
Expression of Pak isoforms varies among tissues. Of all the members of the Pak family, Pak2 is ubiquitously expressed in a number of tissues, whereas Pak3 and Pak5 are specifically expressed in brain (Dan et al., 2002). Pak1 is highly expressed in the brain, muscle, and spleen (Manser et al., 1994). Pak4 appears to be highly expressed in prostate, testis, and colon (Abo et al., 1998). Pak6 is also expressed in a number of tissues, with highest expression in testis and prostrate (Lee et al., 2002).
Because of the nature and depth of the information available on Paks, only a few representative studies are included in this review. Extensive information on the biochemical characteristics of Paks, and their regulation by Rho GTPases is available in the recent reviews (Knaus and Bokoch, 1998; Bagrodia and Cerione, 1999; Sells et al., 1999; Jaffer and Chernoff, 2002).
Paks—An effector of the small GTPases
One of the earliest responses of most cells to extracellular signals is the rapid reorganization of the actin cytoskeleton leading to the formation of motile structures, alteration in cell shape, and hence, to alterations in cell adhesiveness and locomotion. In mammalian cells, generation of actin-based dynamic motile structures is regulated by the small GTPases of the Rho family (Ridley et al., 1992; Kozma et al., 1996; Hall, 1998). Members of the Rho family of the small GTPases, Rac1, Cdc42, and RhoA, have been implicated in the regulation of cytoskeletal rearrangements (Aspenstorm, 1999). RhoA is involved in the maintenance of actin stress-fibers and focal adhesion points, Rac1 in the formation of lamellipodia and membrane ruffles (Ridley et al., 1992), and Cdc42 in the formation of peripheral actin microspikes and filopodia (Kozma et al., 1996). In mammalian cells, Paks are identified as downstream effector target of Cdc42 and Rac1, and binding of GTP-bound GTPases to Pak1 stimulates its kinase activity via autophosphorylation (Manser et al., 1994). In addition to Rac1 and Cdc42, newly identified homologs of the Rho family of GTPases such as Wrch-1 and Chp can also activate Paks and induce filopodium formation and stress fiber dissolution (Aronheim et al., 1998; Tao et al., 2001).
The Paks form complexes specifically with activated (GTP-bound) p21, inhibiting p21 GTPase activity and leading to kinase autophosphorylation and activation. Autophosphorylated kinase has a decreased affinity for Cdc42/Rac1, freeing the p21 for further stimulatory activities or downregulation by GTPase-activating proteins (Manser et al., 1994). Guanine nucleotide exchange factors (GEFs) and GTPase-activating proteins (GAPs), which regulate the GTP-GDP bound states of the Rho family of GTPases, are important determinants of downstream signaling activated by Pak1 kinases (Zhou et al., 1998). Even though, all Paks can interact with GTPases, not all Paks are activated by such interaction. The GTPase interaction of Pak1, Pak2, and Pak3 by itself leads to activation of Paks 1–3, whereas binding of GTPases to Pak4, Pak5, and Pak6 does not lead to their activation (Jaffer and Chernoff, 2002). However, Pak4 has been shown to mediate the induction of filopodia in response to the Rho GTPase Cdc42 (Abo et al., 1998), suggesting GTPases regulate group II Paks distinctly from that of group I Paks.
Recently the crystal structure of a complex formed by the N-terminal autoregulatory fragment and the C-terminal kinase domain of Pak1 has shown that GTPase binding triggers a series of conformational changes, involving disruption of a Pak1 dimer and rearrangement of the kinase active site into a catalytically competent state (Lei et al., 2000; Parrimi et al., 2002). Inhibition of the Rho-family GTPases also blocks the activation of Pak by a membrane-targeted Nck SH3 domain, demonstrating the dependence of in vivo activation of Pak1 on Rho-family GTPases (Lu and Mayer, 1999). Use of the mammalian two-hybrid screen has identified Chp, a homolog of the GTPase Cdc42Hs, as a Pak-binding protein that is implicated in reorganizing the actin cytoskeleton (Aronheim et al., 1998). Shk1, the Ste20/p21-activated kinase homolog, forms a complex with the SH3 domain protein Scd2 (also called Ral3). Overexpression of Scd2 stimulates autophosphorylation of wild-type Shk1 in fission yeast. In addition, Scd2 may stimulate Shk1 catalytic function, at least in part, by positively modulating protein–protein interaction between Cdc42 and Shk1 (Chang et al., 1999).
GTPase-independent regulation of Paks
In addition to kinase activity, Src homology-3 (SH3)-binding PXXP motifs (where X is any amino acid) in Pak1 have been implicated in cytoskeletal reorganization, and Pak1 localization to the membrane in the absence of other signals is sufficient for its activation (Lu and Mayer, 1999), thus, emphasizing the existence of GTPase-independent mechanisms of regulation (Fig. 1). Pak1 localization, probably, via its interaction with the adapter molecules such as Nck, guanine nucleotide factor Pix, and paxillin may also modulate Pak-signaling (Bokoch et al., 1996; Bagrodia and Cerione, 1999; Zhao et al., 2000).
Lipids activate Pak1 to the same degree as GTPases. Pak1 activity could be rapidly induced by sphingosine and several related long-chain sphingoid bases (Bokoch et al., 1998). Lipid stimulation of Pak1 activity depends upon intact Pak1 kinase functions. Using Pak1 mutants, it was shown that lipids act at a site over-lapping or identical to the GTPase-binding domain on Pak (Bokoch et al., 1998). Lipids also activate Pak1 activity via phosphorylated 3-phosphoinositide-dependent kinase 1 (PDK1) (King et al., 2000a). The stimulatory effect of sphingosine results in exposure of the activation loop, making it accessible for intermolecular phosphorylation (Zenke et al., 1999). These studies suggest that Pak(s) may be an important mediator of the biological effects of sphingolipids.
Certain in vitro protein substrates of Pak1, such as histone 2B and histone 4, stimulate both autophosphorylation and Pak kinase activity to levels similar to those observed with Cdc42. This substrate-level activation does not require autophosphorylation of Thr402 in the activation loop (Jakobi et al., 2000). Pak1 also interact with filamin A, a cytoskeletal protein important for the stability of lamellopodia, and binding sites in Pak1 were localized between amino acids 52 and 132 in the conserved Cdc42/Rac1 interacting binding domain (CRIB domain). Interestingly, the binding of filamin A to the Pak1 CRIB domain stimulates Pak1 activity, and Pak1 and filamin A interaction is required for growth factor-stimulation of cofilin phosphorylation via LIM kinase. In addition, filamin A is essential for Pak1-induced cytoskeletal reorganization, and two-way regulatory interaction between Pak1 and filamin A may contribute to the local amplification and local stimulation of Pak1 activity and its targets in cytoskeleton structures (Vadlamudi et al., 2002).
Activation of heterotrimeric G proteins by chemoattractant neutrophils exhibit rapid activation of Pak1 and Pak2, and interruption of the calcium/calmodulin (CaM) signaling blocks their activation, suggesting that the Ca++/CaM complex also plays a major role in the activation of Paks (Lian et al., 2001). Pak1 also phosphorylate Galpha(z), a member of the Galpha(i) family that is found in the brain, platelets, and adrenal medulla. Pak1 regulates Galpha(z) function by attenuating the inhibitory effects of both GAPs and Gbeta gamma via phosphorylation of Ser16 of Galpha(z) (Wang et al., 1999).
Recently, Pak1 and Pak3 have been shown to form complexes with the protein-serine/threonine phosphatase 2A (PP2A) and p70 S6kinase in brain extracts, suggesting a role for PP2A in the regulation of Pak activation (Westphal et al., 1999). Similarly, Pak1's interaction with POPX 1/2 also leads to inhibition of Pak1 activity (Koh et al., 2002). A new Pak1 interacting protein, hPIP1, is shown to inhibit Cdc42/Rac-stimulated Pak1 activity, and it is proposed to act as a negative regulator of Pak signaling pathways (Xia et al., 2001).
Emerging functions of Pak
Accumulating evidence indicates Paks phosphorylate a number of substrates on serine/threonine residues preferably in the context of basic residues such as (K/R) (R/X) (X) (S/T) (Table 1). Pak family members are involved in the regulation of phosphorylation of regulatory non-muscle and smooth muscle myosin II light chain (MLC). Pak2 catalyzes monophosphorylation of MLC at Ser19, and such phosphorylation is implicated in cell retraction (Chew et al., 1998; Zeng et al., 2000). Expression of catalytically active Pak1 increased myosin MLC phosphorylation in fibroblasts and in endothelial cells (Kiosses et al., 1999).
Paks also phosphorylate MLCK, an enzyme that catalyzes phosphorylation of MLC and such phosphorylation inhibits further MLCK activation, thus reducing further phosphorylation of MLC (Sanders et al., 1999). In contrast to Paks, MLCK phosphorylates MLC on two sites (Ser18, 19). Since Paks regulate both mono-and diphosphorylation of MLC, it has been proposed that differential activation of Paks and MLCK may have a role in specific cellular cytoskeletal reorganizations (Goeckeler et al., 2000). In addition to MLC phosphorylation, activated GTPases regulate actin depolymerization through Pak1 via LIM-kinase (Lawler, 1999). Pak1 phosphorylates LIM-kinase at threonine residue 508 within LIM-kinase's activation loop and increases LIM-kinase-mediated phosphorylation of the actin-regulatory protein cofilin leading to polymerization of actin (Edwards et al., 1999).
The mechanism by which Pak1 regulates cytoskeletal reorganization remains elusive. Kinase-dead Pak1 induces lamellipodia formation and accumulation of focal points, thus these Pak1 effects are independent of its kinase activity (Manser et al., 1997). Studies also demonstrated a kinase independent effect of Pak1 on the ruffling and these studies utilized Cd42 or Rac mutant H83, 86L Pak1 (Frost et al., 1998). However, H83, 86L Pak1 mutant has a partial constitutive kinase activity, which may account for excessive ruffling. Breast cancer cells show excessive ruffling by H83, 86L Pak1 (Vadlamudi et al., 2000), while no ruffling was observed by kinase dead Pak1-K299R (Adam et al., 2000). However, K299R expression in MDA-MB435 cells resulted in or was associated with excessive cell spreading and accumulation of mature focal points (Adam et al., 2000). Recent data suggest that Pak1 interacts with and phosphorylates filamin A. In filamin A-negative cells, constitutively active Pak1 failed to induce ruffling. Furthermore, filamin A mutant which cannot be phosphorylated by Pak1, failed to induce ruffling (Vadlamudi et al., 2002). These results suggest that Pak1-mediated ruffling requires both kinase-dependent and -independent functions of Pak1 and may depend on the cellular context.
Evidence suggests that Paks may have a role in intermediate filament reorganization. Pak1 phosphorylates desmin and inhibits its ability to bind filament (Ohtakara et al., 2000). Pak3 has also been shown to regulate Ca(2+)-independent contraction of smooth muscle via phosphorylation of caldesmon (Foster et al., 2000). Pak1 regulates the reorganization of vimentin filaments through direct vimentin phosphorylation (Goto et al., 2002).
Recent evidence also implicates Paks in microtubule reorganization (Zhang et al., 2001). Pak1 phosphorylates stathmin on Ser16 and is believed to play a role in stabilization of microtubules (Daub et al., 2001). A recent study showed a Pak5 homolog from Xenopus stabilized microtubules and may have a role in linking actin signaling with the microtubule network (Cau et al., 2001). Incubation of cells with nocodazole leads to activation of Pak1/2 kinases, and such activation leads to ras-independent phosphorylation of Raf-1 on Ser338 (Zhang et al., 2001).
In addition to cytoskeletal changes, Paks also stimulate a number of signaling pathways, including mitogen-activated protein kinases (p42/44 MAPK, p38 MAPK), Jun N-terminal kinase activities (JNK), and nuclear factor kappa B (NF-κB). Rac1 and Cdc42 appear to regulate a protein kinase cascade initiated at the level of Pak and leading to activation of p38MAPK and JNK (Bagrodia et al., 1995; Zhang et al., 1995). Overexpression of Pak1 and Pak2 in 293 cells is sufficient to activate JNK and, to a lesser extent p38 MAPK (Frost et al., 1996). Breast cancer cells expressing inducible T423E Pak1 exhibited a regulatable stimulation of P42/44 MAPK and Jun N-terminal kinase activities (Vadlamudi et al., 2000). Pak1 also stimulates nuclear translocation of the p65 subunit of NFκB, but it does not activate the inhibitor of kappaB kinases alpha or beta, and thus it could influence the extent of NF-kappaB activation by multiple stimuli (Frost et al., 2000).
Pak1 phosphorylates MAPK/ERK kinase 1 (MEK1) on Ser298, a site important for binding of Raf-1 to MEK1 in vivo. Expression of interfering Pak1 also reduces stimulation of Ternary Complex Factors (TCF) function by serum growth factors, while expression of active Pak1 enhances epidermal growth factor (EGF)-stimulated MEK1 activity (Frost et al., 1997). Phosphorylation of Ser338 or Ser339 in the catalytic domain of Raf-1 regulates Raf-1 activation in response to Ras, Src, and EGF signaling. Pak3 phosphorylates Raf-1 on Ser338 in vitro and in vivo, suggesting that signaling via Raf-1 depends on the activation of the Ras and Pak pathways as well as on activation of Ras-dependent phosphatidylinositol-3-OH kinase (PI-3 kinase). PI 3-kinase regulates phosphorylation of Raf1 Ser338 through the serine/threonine kinase Paks. Phosphorylation of Raf1 Ser338 through PI 3-kinase and Pak provides a co-stimulatory signal, which together with Ras leads to strong activation of Raf-1 kinase activity by integrins (Chaudhary et al., 2000).
Pak plays a critical role in natural killer (NK) cell lysis of tumor cells. Dominant-negative N17Rac1 and Pak1 effect cytotoxic function; whereas, constitutively active V12Rac1 has the opposite effect. Results of this study indicated a specific pathway PI3K → Rac1 → PAK1 → MEK → ERK in NK cells that effects lysis (Jiang et al., 2000b). The Rac1/Cdc42-Pak1-MEKK1 coupled pathway is shown to have a role in signal transduction and facilitates cross-communication between the CD28 costimulatory signal and the T cell antigen receptor (TCR) signal (Kaga et al., 1998). Pak can phosphorylate tumor suppressor gene neurofibromatosis type 2 tumor suppressor gene (NF2, Merlin) on Ser518 (Xiao et al., 2002).
Paks may play an important role in exocytosis, endocytosis, and vesicle trafficking pathways. Macropinocytosis is an essential aspect of normal cell function, contributing to both the growth and motile processes of cells. Pak1 localizes to areas of membrane ruffling, as well as to amiloride-sensitive pinocytic vesicles. Pak activity is required for normal growth factor-induced macropinocytosis, and such regulation may have a role for Pak in directed cell motility. Constitutively activated Pak1 enhances both growth factor-stimulated 70-kDa dextran uptake and efflux, suggesting that Pak1 activity modulates pinocytic vesicle cycling (Dharmawardhane et al., 1999, 2000). Pak phosphorylates a conserved serine or threonine residue in the myosin heavy chain (Brzeska et al., 1997). Mutation of serine to glutamic acid, which mimics phosphorylation and therefore activation of myosin, results in an accumulation of internalized plasma membranes/vesicles results from an activation of endocytosis, thus suggests an in vivo significance of regulatory phosphorylation on class I myosin by Pak1 (Yamashita and May, 1998). Cyclin-dependent kinase Cdk5-p35 localizes with Pak1 in the Golgi apparatus where it phosphorylates Pak1. Inhibition of Cdk5-p35 resulted in lack of transported exocytotic vesicles (Paglini et al., 2001).
In an effort to better understand the functions of Pak, in recent years researchers have identified a number of Pak1 binding proteins (Table 2). The first type of interacting proteins identified belongs to the Cool/Pix family of proteins. There are three proteins in this group. Pak-interacting exchange factors (alpha Pix and beta Pix) were identified by Manser et al. (1998). Bagrodia et al. also identified three proteins, and referred them as the Cool family (cloned out of library) proteins (Bagrodia et al., 1998). Pix and Cool are identical proteins and appear to be generated by alternative splicing, and they contain a common Src homology-3 (SH3) domain, DBL-homology (DH) domain, and pleckstrin-homology (PH) domain but they differ in the lengths of the N- or C-terminal region. These three proteins are now commonly referred to as p50Cool-1, P85Cool-1/beta Pix, Cool-2/alpha Pix. Cool/Pix family proteins activate Paks by cooperating with Cdc42/Rac1 and also by direct interaction with Paks. The binding of p50/Cool-1 inhibits autophosphorylation and the ability to phosphorylate exogenous substrates of Pak3 (Bagrodia et al., 1999), whereas Cool-2/alpha Pix is capable of stimulating Pak activity. Pak-interacting exchange factor (alpha Pix) is activated by PI3-kinase and colocalizes with Pak in a focal complex (Yoshii et al., 1999). P85Cool-1/beta Pix interacts with Pak through its SH3 domain and may have a role in the nuclear signaling and actin cytoskeleton regulation by Pak. The effect of beta-Pix on p38MAPK is believed to be exerted through the Cdc42/Rac1-Pak pathway and requires Pak activity (Lee et al., 2001).
G-protein-coupled receptor kinase-interacting protein known as GIT1 is a multidomain protein. It has binding domains for paxillin as well as for pix (Manabe et al., 2002). GIT1 may have a role in drawing activated Pak1 to adhesions and the leading edge of cell via its interaction with paxillin. Overexpression of GIT1 in fibroblasts or epithelial cells causes a loss of paxillin from focal-adhesion complexes and stimulates cell motility. Since GIT1 directly associates with focal-adhesion kinase (FAK), it may link FAK to Pak signaling and thus, playing an important role in orchestrating focal complex dynamics, which underlie cell motility (Zhao et al., 2000).
The adaptor protein Nck can interact and mediate the relocalization and subsequent activation of Pak kinases. Nck associates in vivo with Pak, using the second of its three SH3 domains, which results in activation of Pak and stimulation of downstream MAPK cascades. The Nck adaptor protein could function to link changes in tyrosine phosphorylation of membrane receptors to the Pak signaling pathway. The human immunodeficiency virus (HIV) protein Nef associates with and activates a Pak-related kinase in lymphocytes infected in vitro. Nef-mediated activation of a Pak-related kinase correlates with the induction of high virus loads and the development of AIDS in the infected host (Sawai et al., 1996; Barber et al., 1998; Renkema et al., 1999).
Another Pak1-interacting protein, identified by yeast two-hybrid screening, is filamin A. Pak1 interacts with filamin A in vivo and is physiologically stimulated to be phosphorylated by, interact with and to colocalize with endogenous Pak1 in membrane ruffles (Vadlamudi et al., 2002). Pak1's ruffle-forming activity was functional in filamin A-expressing cells but not in filamin A-deficient cells (Fig. 2). Furthermore, Pak ruffle-forming activity was regained in filamin A-deficient cells upon restoration of filamin A expression, and it was impaired by a Pak1 inhibitory fragment in filamin A-expressing, growth factor-stimulated cells (Vadlamudi et al., 2002).
Paxillin can bind directly to Pak3, and paxillin can compete with Nck and beta Pix in the binding of Pak3. Moreover, paxillin can be phosphorylated by Pak3. Therefore, paxillin may link Pak3 to integrins independent of Nck and beta Pix, as Nck links Pak1 to growth factor receptors (Hashimoto et al., 2000). Paxillin is a focal adhesion adaptor protein involved in the integration of growth factor- and adhesion-mediated signal transduction pathways. Paxillin LD4 is implicated in the binding of a complex of proteins containing active p21 GTPase-activated kinase (Pak), Nck, and the guanine nucleotide exchange factor, and Pix. The association of this complex with paxillin is mediated by a new 95-kD protein, p95PKL (paxillin-kinase linker), which binds directly to paxillin LD4 and Pix. Overexpression of the paxillin LD4 deletion mutant in neuroblastoma cells inhibits lamellipodia formation in response to insulin-like growth factor-1 suggesting that paxillin is a mediator of p21 GTPase-regulated actin cytoskeletal reorganization through the recruitment to nascent focal adhesion structures of an active Pak/Pix complex, potentially by means of interactions with p95PKL (Turner et al., 1999).
Regulation of Paks by tyrosine phosphorylation via cytoplasmic tyrosine kinases is an emerging theme in the area of Pak research. Etk/Bmx, a member of the Tec family of non-receptor protein-tyrosine kinases, directly associates with Pak1 by means of its N-terminal pleckstrin homology domain and it also phosphorylates Pak1 on tyrosine residues (Bagheri-Yarmand et al., 2001). Pak2 interacts and phosphorylates with c-Abl on sites located in the kinase domain, in a region that is implicated in protein–protein interactions and in subcellular localization, and results in activation of c-Abl. Pak2 is also phosphorylated by activated c-Abl and results in downregulation of Pak2 activity by ubiquitination, suggesting a negative feedback loop between c-Abl and Pak2 (Roig et al., 2000). Constitutively active mutant EGF receptor induces tyrosine phosphorylated forms of Pak1. The catalytic activity of the Pak family of serine/threonine kinase is stimulated by tyrosine phosphorylation (McManus et al., 2000).
Pak and growth factor signaling
Growth factors and their receptors play an essential role in the regulation of epithelial cell proliferation, metastasis, angiogenesis, and tumorigenesis. Stimulation of growth factor signaling has been implicated in the development of an invasive phenotype and Pak1 activation in human breast cancer cells. Heregulin, a combinatorial ligand for human EGF receptor 2 (HER2) and HER3 receptors, triggers a rapid stimulation of Pak1 activity and its redistribution into the leading edges of motile cells. HRG stimulation promotes physical interactions between Pak1, actin, and HER2; and these interactions are dependent on the activation of PI-3 kinase (Adam et al., 1998). The blockade of HER2 receptor by an anti-HER2 monoclonal antibody results in the inhibition of HRG-mediated stimulation of the PI-3 kinase/Pak pathway and also the formation of motile actin cytoskeleton structures (Adam et al., 1998). Platelet-derived growth factor (PDGF)-induced p38MAPK activation is also blocked by expression of a mutant of Pak1 binding protein beta-Pix SH3m (W43K) and beta Pix DHm (L238R, L239R) suggesting a role for the Pak1-beta Pix pathway in PDGF-mediated cytoskeletal reorganization (Lee et al., 2001).
Hepatocyte growth factor (HGF), the ligand for the Met receptor tyrosine kinase, is a potent modulator of epithelial-mesenchymal transition and dispersal of epithelial cells, processes that play crucial roles in tumor development, invasion, and metastasis. HGF induces activation of the Cdc42/Rac-regulated Pak and its translocation to membrane ruffles and is shown to play a role in HGF-induced epithelial cell spreading (Royal et al., 2000). PDGF mediated activation of Pak1 family kinases and cell migration of fibroblasts requires transactivation of EGFR (He et al., 2001). Overexpression of mutant Pak1 blocked PDGF-induced chemotactic cell migration, suggesting that PDGF may require Pak1-mediated signaling to p38MAPK among other pathways (Dechert et al., 2001).
Paks and growth regulation
In recent years, Pak signaling has been implicated in cell growth regulation. The expression of a catalytically active Pak1 mutant stimulates anchorage-independent growth of breast cancer cells in soft agar in a preferential MAPK-sensitive manner (Vadlamudi et al., 2000). In another study, researchers observed that inhibition of PKA allows anchorage independent stimulation of the MAPK cascade by growth factors via Pak (Howe and Juliano, 2000). PKA negatively regulates Pak in vivo, and though induction of anchorage-independent signaling resulting from PKA suppression is blocked by dominant negative Pak, it is markedly prolonged by constitutively active Pak1 (Howe and Juliano, 2000). Pak4 was also shown to have oncogenic potential, and Pak4 activity is required for Ras-driven, anchorage-independent growth (Callow et al., 2002).
The Ras pathway leading to Pak1 activation has been shown to have a role in cellular transformation. Existing evidence suggests Raf-1 activation by ras is achieved through a combination of both physical interaction and indirect mechanisms involving the activation of a second ras effector, PI 3-kinase, which directs Pak-mediated regulatory phosphorylation of Raf-1 (Sun et al., 2000). Ras effectors also cooperate with the PI 3-kinase to activate Pak activity. However, results from studies with rat fibroblasts suggest that Pak activation may be necessary, but not sufficient, for cooperative transformation of Rat-1 fibroblasts by ras, rac, and rho (Tang et al., 1999). Rat-1 cell lines expressing Pak1 (K299R) were highly resistant to ras transformation, while cells expressing wild-type Pak1 were efficiently transformed by Ras. Pak1 (L83, L86, R299), a mutant that fails to bind either Rac or Cdc42, also inhibited Ras transformation (Tang et al., 1997).
Pak1 may affect the cell survival pathway by directly phosphorylating and inactivating the pro-apoptotic functions of Bad (Schurmann et al., 2000). Pak1 phosphorylated Bad in vitro and in vivo at the Ser112 and Ser136 sites, resulting in a markedly reduced interaction between Bad and Bcl-2 or Bcl-x(L) and an increased association of Bad with 14-3-3 tau. Overexpression of a constitutively-active Pak1 T423E mutant promotes the survival of NIH 3T3 murine fibroblasts, while overexpression of the autoinhibitory domain of Pak1 (amino acids 89–143) enhances apoptosis (Schurmann et al., 2000).
Pak2 has been shown to act as both a pro- and anti-apoptotic factor, depending on the mode of activation. Caspases activated during apoptosis cleave Pak2, and such cleavage is shown to activate Pak2 and JNK, thereby, modulating cell death (Widmann et al., 1998). Proteolytic activation of Pak2 induces morphological changes and elicits apoptosis. This C-terminal Pak2 fragment also activates the c-Jun N-terminal kinase pathway in vivo (Lee et al., 1997). However, activation of full-length Pak2 promotes cell survival and suppresses stress-induced cell death of BALB3T3 fibroblasts in response to TNF alpha, growth factor withdrawal, and ultraviolet light. Expression Pak2-T402E increases phosphorylation of the pro-apoptotic Bcl-2 family protein Bad and prevents cell death induced by ectopic expression of Bad (Jakobi et al., 2000; Jakobi et al., 2001).
Pak in invasion and metastasis
Exposure of cells to growth factors induces cytoskeleton reorganization, lamellipodia formation, and membrane ruffling; such changes contribute to increased cell migration and invasion (Nobes and Hall, 1995). A close correlation exists between the status of Pak1 activity and the baseline invasiveness of human breast cancer cells and breast tumor grades (Vadlamudi et al., 2000). Expression of Pak4 has been shown to be elevated in a variety of tumor cell lines (Callow et al., 2002). The expression of Pak1-interacting protein Etk in non-invasive MCF-7 human breast cancer cells significantly increased proliferation and anchorage-independent growth of epithelial cancer cells. Conversely, expression of the kinase-inactive mutant Etk-KQ suppressed the proliferation, anchorage-independent growth, and tumorigenicity of human breast cancer MDA-MB435 cells (Bagheri-Yarmand et al., 2001). These observations suggest that Pak1 is a target of Etk, and that Etk controls the proliferation, as well as the anchorage-independent and tumorigenic growth of mammary epithelial cancer cells (Bagheri-Yarmand et al., 2001). Further, we observed that constitutive expression of active Pak1 (T423EPak1) promotes hyperplasia in mouse mammary gland (Wang et al., 2002). In addition, hyperactive Rac3 is present in highly proliferative human breast cancer-derived cell lines and tumor tissues. Experiments with dominant negative Pak mutants revealed that the Rac3-Pak pathway was critical for DNA synthesis, independent of JNK activation, suggesting an important role for Rac3 and Paks in tumor growth (Mira et al., 2000). HRG-mediated invasion of the MCF-7 breast cancer cell line is mediated by the Pak1 pathway (Adam et al., 1998). Recently, the PI-3 kinase/Pak1/p38MAPK pathway has been also shown to be hyper-activated in breast tumors, while no activation of AKT or p70S6K was observed (Salh et al., 2002). Furthermore, PI-3 kinase and Pak1 are co-overexpressed in breast tumors, and Pak1 has higher activity in grade II breast tumors. Together, these observations suggest that Pak1 may participate in the diversification of signals downstream PI-3 kinase (Salh et al., 2002).
Constitutively activated mutant EGF receptor induces the formation of a transformation-specific signaling module that complexes with myosin II containing signal adapter proteins Shc, Grb2, and Nck, and with tyrosine-phosphorylated forms of Pak, caldesmon, and MLC kinase. Transformation-specific, tyrosine-induced phosphorylation of Pak1 enhances the catalytic activity of Pak1. The tyrosine phosphorylation of Pak is dependent on Rho but not on ras, rac-, or cdc42. Activation of Pak1 by a ligand-independent EGF receptor mutant may constitute a novel pathway for the coupling of oncogenic receptor tyrosine kinases with the actomyosin molecular motor contributing to the regulation of the mechanical forces governing cellular adhesion and cytoskeletal tension, and hence on anchorage-independent cell growth (McManus et al., 2000).
Pak2 is activated between two- and fivefold in response to ionizing radiation (IR) in 3T3-L1 fibroblasts and U937 leukemia cells. Pak2 is activated in a dose- and time-dependent manner. UV (80 J/m2) and the DNA-damaging drugs cytosine beta-d-arabinofuranoside (AraC) and cis-platinum (II) diamine dichloride (cisplatin) also induce Pak2 activation. In contrast to Pak2, Pak1 and JNK are activated only by cisplatin and UV in 3T3-L1 cells, suggesting differential regulation of Pak1. Since Pak1 expression is deregulated or overexpresed in tumors, activation of Paks by therapeutic agents may lead to resistance because of Pak's ability to promote cell survival and anchorage-independent growth (Roig and Traugh, 1999).
Wnt-1 transformation induces expression of a novel Rho GTPase Wrch-1. Overexpression of Wrch-1 mimics the Wnt-1 phenotypes in morphological transformation of mouse mammary epithelial cells. Wrch-1 also mediates the effects of Wnt-1 signaling in the regulation of cell morphology, cytoskeletal organization, and cell proliferation. Thus, Pak1 signaling may be important in Wnt-1-mediated transformation (Tao et al., 2001). Pak1-mediated phosphorylation and inactivation of neurofibromatosis type 2, a tumor suppressor gene may play a role in tumor cell spreading and metastasis (Xiao et al., 2002).
Amplification of loci present on band q13 of human chromosome 11 is a feature of a subset of estrogen receptor (ER)-positive breast carcinomas prone to metastasis. Four new genes were placed on the regional map, namely CBP2, CLNS1A, UVRAG, and Pak1 (Bekri et al., 1997). To our knowledge, no studies were published looking the amplification of the Pak1 gene or mutation of Paks in tumors. However, based on the existing evidence of the importance of this molecule in motility, cell survival signaling increased protein expression/activity in cancer cells/tumors and amplification of its chromosomal loci in a subset of ER-positive tumors, we speculate that Pak1 expression may be deregulated in tumors.
Pak and angiogenesis
Several recent studies suggest that tumor growth and progression are intimately linked with the process of angiogenesis. The onset of tumor angiogenesis depends on the production of angiogenic factors by the tumor cells or by the tumor microenvironment that stimulates host-organ vascular endothelial cell growth and chemotaxis (Fidler, 1995). Vascular endothelial growth factor (VEGF), also known as vascular permeability factor (Fidler, 1995; Hanahan and Folkman, 1996), is a potent, selective endothelial-cell mitogen and chemotactic protein that is regulated by a number of cytokines, growth factors, and oncogenes (Rak et al., 1995; Shweiki et al., 1992), and by phosphatidylinositol (PI)-3 kinase signaling (Jiang et al., 2000). Pak1 is a potential modulator of endothelial cell migration (Kiosses et al., 1999). Heregulin-β1 regulates angiogenesis by means of upregulation of VEGF expression. Pak1 plays an important role in controlling VEGF expression and, consequently, in controlling VEGF secretion and functions (Bagheri-Yarmand et al., 2000). Pak1 activation stimulates JNK in vascular smooth muscle cells (VSMCs) and Angiotensin II (Ang II) uses Pak1 as an upstream mediator of JNK in Ang II signaling, which suggests there is role for Paks in proinflammatory signaling in VSMCs (Schmitz et al., 1998). Treatment of cells with a Pak peptide specifically inhibits endothelial cell migration and contractility, similar to treatment with a full-length dominant-negative Pak. These results further suggest a role for Paks in angiogenesis (Kiosses et al., 1999).
Pak and mitosis
Earlier studies with Pak homologues in Saccharomyces cervisiae (Ste20) showed that Pak homologs may have a role in cytokinesis and mitosis. In Xenopus, X-Paks are shown to be involved in negative control of the process of oocyte maturation (Faure et al., 1999). Using a microinjection approach, Faure et al. (1999) demonstrated that G2/M progression is prevented in Xenopus cycling extracts in the presence of active X-Pak1 (Faure et al., 1999). In breast cancer cells, regulated expression of kinase-active Pak1 resulted in abnormal organization of mitotic spindles, characterized by the appearance of multiple spindle orientations (Vadlamudi et al., 2000). Ste20p, a member of the Ste20/p21-activated family of protein kinases, is post-translationally modified by phosphorylation in a cell cycle-dependent manner. Ste20p is a substrate for the Cdc28p kinase, and undergoes a Cln2p-Cdc28p-mediated mobility shift as cells initiate budding and DNA replication. In cells that express only Cln2p G1 cyclin, minor overexpression of Ste20p causes aberrant morphology, suggesting proper coordination of Ste20p and Cln-Cdc28p activity may be required for control of cell shape (Wu et al., 1998). Paka, a Dictyostelium discoideum homolog of Pak, colocalizes with myosin II in the cleavage furrow of dividing cells. Paka-null cells are defective in completing cytokinesis in suspension (Chung and Firtel, 1999). Recent studies have shown a putative chromosomal function of Pak1 in mammalian cells (Li et al., 2002). At the onset of mitosis, Pak1 becomes activated and translocates to the nucleus. From this point on, Pak1 behaves like a chromosomal passenger protein. Screening of G2/M expression library with glutathione transferase-Pak1 solid-phase-based-kinase reaction identified histone H3 as a specific interacting substrate of Pak1, which phosphorylates histone H3 on Ser10 both in vitro and in vivo. Histone H3 phosphorylation has been shown to be required for the initiation of chromosome condensation and cell–cycle progression (Cheung et al., 2000). Pak1 is also phosphorylated during mitosis in mammalian fibroblasts on Thr 212 (Thiel et al., 2002). Deregulation of Pak1 signaling thus may influence DNA ploidy and may contribute to anchorage-independent growth and other chromosomal abnormalities observed in tumor cells (Fig. 3) (Vadlamudi et al., 2000; Li et al., 2002).
Pak as therapeutic target
Emerging evidence strongly implicates Pak1 signaling in the process of tumorigenesis thus interference with Pak1 signaling may be a useful therapeutic tool (Fig. 4). Results of studies using dominant negative expression plasmids clearly showed the promise of this approach in controlling metastasis of a number of tumor cells (Adam et al., 1998; Bagheri-Yarmand et al., 2000). Since Pak1 also supports a number of cell survival signals, interfering with such signaling may promote apoptosis in tumors, thus may lead to tumor regression. However, no known chemical inhibitor of Pak1 is currently available. The structure of Pak1 has been recently solved (Lei et al., 2000) and is expected to faciliate the development of Pak1 inhibitors.
Growth factor-induced activation of Pak1 can be inhibited by the PDGF-R-specific inhibitor (AG1478) or the EGF receptor specific inhibitors (AG12995), and therefore these compounds may be clinically useful in regulating Pak activity growth factor-regulated tumors (He et al., 2001). Bufalin, a component of the Chinese medicine chan′su, induces apoptosis in various lines of human tumor cells by means of the Tiam1-Rac1-Pak1-JNK pathway (Kawazoe et al., 1999). A plethora of proteins that bind Pak1 are being identified, and such protein–protein interactions further provide novel targets for interference with Pak signaling. A recent study identified a small peptide that can interfere with NCK binding to Pak1 can interfere with angiogenesis (Kiosses et al., 1999), suggesting that small molecule interference of Pak interactions with its binding protein may have therapeutic value.
Many questions need to be addressed to fully appreciate the cellular functions of the Pak family of proteins and to effectively target this fascinating protein kinase for drug development (Fig. 5). Some of the key questions include the following. (a) What are the physiological substrates and interacting proteins of Paks? It is expected that identification of these substrates and proteins will be aided by emerging new proteomic technologies. Emerging data also raise important questions about the functions of Pak1 during mitosis, including its potential involvement in the condensation, capture, and/or movement of chromosomes during mitosis; (b) Does the Pak family have a role in cell cycle progression? (c) Does Paks control cell cycle checkpoints? (d) Are the mitotic functions of Paks are dependent on GTPase signaling? It will also be interesting to determine when and how Pak activation occurs during the process of tumorigenesis. The availability of specific antibodies against phosphorylated Paks is likely to provide insight into the role of Paks in human cancers. Future studies are clearly required to determine whether dominant-active mutations in Paks occur during the process of tumorigenesis. Paks are abundantly expressed in neuronal tissues and future studies are warranted for studying the role of Paks and their binding proteins in neurodegenerative diseases. The ability of Pak6 to interact with the androgen receptor brings a new dimension to the area of Pak1 research. Do other members of the Pak family also interact with the nuclear receptors and modulate their functions? Does upregulation of Pak activity in tumors alter the responsiveness of tumors to hormones and/or anti-hormones? Does this process involve regulatory interactions between Paks and nuclear receptor cofactors? A better understanding of the functions of Paks and their binding proteins is expected to result in development of a comprehensive and integrated model for dealing with Paks in management of human cancer cells.