p21-activated kinase 1 is required for efficient tumor formation and progression in a Ras-mediated skin cancer model

Abstract

The RAS genes are the most commonly mutated oncogenes in human cancer and present a particular therapeutic dilemma, as direct targeting of Ras proteins by small molecules has proved difficult. Signaling pathways downstream of Ras, in particular Raf/Mek/Erk and PI3K/Akt/mTOR, are dominated by lipid and protein kinases that provide attractive alternate targets in Ras-driven tumors. As p21-activated kinase 1 (Pak1) has been shown to regulate both these signaling pathways and is itself upregulated in many human cancers, we assessed the role of Pak1 in Ras-driven skin cancer. In human squamous cell carcinoma (SCC), we found a strong positive correlation between advanced stage and grade and PAK1 expression. Using a mouse model of Kras-driven SCC, we showed that deletion of the mouse Pak1 gene led to markedly decreased tumorigenesis and progression, accompanied by near total loss of Erk and Akt activity. Treatment of Kras^G12D mice with either of two distinct small molecule Pak inhibitors (PF03758309 and FRAX597) caused tumor regression and loss of Erk and Akt activity. Tumor regression was also seen in mice treated with a specific Mek inhibitor, but not with an Akt inhibitor. These findings establish Pak1 as a new target in KRAS-driven tumors and suggest a mechanism of action through the Erk, but not the Akt, signaling pathway.

Introduction

The Ras genes - HRAS, KRAS, and NRAS - represent one of the most important oncogene families in human cancer, with activating mutations seen in approximately 30% of solid tumors (1). Ras proteins act as switch molecules by transmitting mitogenic signals in response to variety of extracellular stimuli by binding and hydrolyzing GTP, as well as regulating diverse cellular processes such as proliferation, migration, senescence, differentiation, and survival. In human cancer, activating mutations in RAS promote cell proliferation and result in tumorigenesis that generally correlates with poor prognosis and poor therapeutic response (2). Since the oncogenic role of the Ras protein is well-established, numerous attempts have been made to target this GTPase for the treatment of human cancers. Strategies for blocking activated Ras have included attempts to reduce its expression, interfere with its subcellular localization, and inhibit its downstream effectors (3, 4). With regard to the latter, more than twenty proteins have been reported as effectors of Ras, and many of these provide potentially suitable drug targets (5, 6).

The phosphoinositol-3 kinase (PI3K)/Akt/mTOR and Raf/Erk signaling modules are among the best-studied Ras effector pathways. A growing body of evidence indicates that members of the p21-activated kinase (Pak) family, in particular Pak1, are required for the activation of both these pathways. Paks are serine-threonine-specific protein kinases that act downstream of the small GTPases Cdc42 and Rac in a variety of signaling pathways (7–9). Mammalian cells encode six Pak isoforms - group A (Pak1, -2, and -3) and group B (Pak4, -5, and -6) – with partly overlapping but also clearly distinct signaling properties (10). In Erk signaling, Pak1 phosphorylates c-Raf at S338 and Mek1 at S298, sites that are required for full activation of these proteins in some cell types (10, 11). In the Akt pathway, Pak1 is thought to act in a non-catalytic fashion, acting as a scaffold to bridge PDK1 to Akt (12, 13). Inhibition or loss of Pak1 might therefore be expected to interfere with the oncogenic potential of proteins such as Ras that induce transformation at least in part by activation of these pathways. A wealth of in vitro data support this view, as expression of dominant negative alleles of Pak1, reduction of Pak1 expression by RNAi, and small molecule inhibitors of Pak1 have all been shown to interfere with Ras-driven transformation (11, 14, 15). However, the role of Pak1 in Ras-driven tumorigenesis in vivo, and the particular signaling pathways affected, are not defined.

In this work, we asked if Pak1 plays a role in Ras-induced skin cancer. Using human skin cancer tissue microarrays, we found that PAK1 expression levels are associated with more aggressive grades and poorer differentiation of squamous cell carcinoma (SCC). Functional data were then obtained by crossing an inducible Kras^G12D driven mouse model of skin cancer to Pak1 wild-type, heterozygous, or knock-out mice. In such mice, we found that Pak1 gene dosage was positively correlated with tumor initiation and progression. Kras^G12D mice lacking Pak1 showed marked reduction in both Erk and Akt activation, indicating that Pak1 function is required for activation of these signaling pathways by Kras in vivo. Tumor regression was also noted when Kras^G12D mice were treated with either of two distinct small molecule Pak inhibitors (PF03758309 or FRAX597) or a Mek inhibitor, but not with an Akt inhibitor. These findings establish Pak1 as a new target in Kras-driven tumors and define a mechanism of action primarily through the Erk, but not the Akt, signaling pathway.

Materials and Methods

Generation of transgenic mice and tumor measurement

K5-rTA and tet-Kras^G12D transgenic mice (FVB-N) (16) were crossed with Pak1 knockout mice (FVB-N) (17) separately to generate Pak1^+/−::K5-tTR and Pak1^+/−::tet-Kras^G12D colonies. Progeny from these colonies were subsequently bred to generate K5-tTR::tet-Kras^G12D mice that were wild-type, heterozygous, or knockout for Pak1. Genotyping was performed by PCR analysis of tail biopsy DNA. A doxycycline diet was given to a similar number of mice from each cohort as well as littermates from all three colonies not receiving doxycycline treatment as controls. All mice were examined daily to assess tumor onset. Tumor measurements were taken by calipers thrice weekly and tumor volume was calculated as width² × length × 0.5. Mice were euthanized when the tumor size reached >1 cm or if mice exhibited signs of illness.

Doxycycline administration

Doxycycline-containing food was administered to mice at 3 months of age in a pelleted form with complete grain-based rodent diet and the control littermates were fed an identical diet, lacking doxycycline.

Cancer therapy using inhibitors

Mek inhibitor PD0325901(18) was reconstituted and administered to mice receiving either a single dose of 20 mg/kg/day or an equivalent volume of vehicle (0.5% methylcellulose, 0.2% Tween-80) via oral gavage. Pak inhibitor PF3758309 was synthesized according to published protocols (19) and was formulated in saline and administered by intraperitoneal injection at 25 mg/kg/day or an equivalent volume of vehicle. Akt inhibitor GSK690693 (20) was dissolved in water and administered by intraperitoneal injection at dose of 30 mg/kg/day or equivalent volume of vehicle. FRAX-597 (6-(2-chloro-4-(thiazol-5-yl)phenyl)-8-ethyl-2-(4-(4-methylpiperazin-1-yl)phenylamino)pyrido[2,3-d]pyrimidin-7(8H)-one, Chemical Formula: C29H28ClN7OS, Exact Mass: 557.18, Molecular Weight: 558.10 (Fig. 4B) (Afraxis, La Jolla, CA) was formulated in 10% (PEG400:Tween 80:PVP-K30 – 90:5:5) 15% Vitamin E-TPGS and 75% of HPC (0.5%) in 50 mM citrate buffer (pH 3) and administered by oral gavage at 90 mg/kg/day. All treatments were begun when tumors became visible (~20 mm³, ~5 day post doxycycline feedings) and were continued for 7 or 10 days, at which time the animals were sacrificed and SCC tissues taken for analysis.

Figure 4.

Erk and Akt-mTOR signaling in Pak1^+/+, Pak1^+/−, and Pak1^−/− tumors. (A–C) Immunoblot analyses of Erk and Akt-mTOR signaling pathways from tumor lysates. Tumors were excised from newly sacrificed animals and protein lysates obtained and probed with the indicated antibodies. As tumors in Pak1^−/− mice were usually small (50 mm³ or less), equal sized tumors were also used for analysis of signaling from Pak1^+/+, and Pak1^+/− mice. (D) Papillomas (P) or large (>75 mm³) carcinomas (C) from Pak1^+/+::K5-tTR::tet-Kras^G12D, Pak1^+/−::K5-tTR::tet-Kras^G12D, or Pak1^−/−::K5-tTR::tet-Kras^G12D mice were excised and analyzed by immunoblot with the indicated antibodies. c-Raf was immunoprecipitated prior to analysis by immunoblot. Numbers at the top of each column indicate specific individual mice used for these experiments.

Tissue preparation, histology, immunohistochemistry, and immunoblotting

All tumor lesions, control tissues and internal organs were fixed overnight in 4% paraformaldehyde, dehydrated and embedded in paraffin. Hematoxylin and eosin (H&E) stained sections were used for diagnostic purposes and unstained sections for immunohistochemical (IHC) studies. IHC was performed with the following antibodies: rabbit polyclonal antibody for Pak1 (1:50), Pak2 (1:50), phospho-Erk1/2 (pThr202/pTyr204) (1:100), phospho-Akt (pThr308) (1:30), phospho-S6 (pSer235/p236) (1:5000), anti-cleaved caspase 3 (Cell Signaling Technology, Danvers, MA), and rat monoclonal antibody for Ki67 (Dako). The evaluation of the IHC was conducted blindly, without knowledge of the origin or genotype. The percentage of Ki67 positive cells was determined by scanning the slides using an Aperio CS Scanscope scanner and a nuclear detection software from the same manufacturer. (2000–5000 cells were counted per mouse, 3–5 mice/group). Apoptosis was evaluated counting CC3 positively stained in five high magnification fields per mouse (X400) (3–5 mice per group, minimal number of cells counted/mouse was 500 per mouse).

Immunoblot analyses were performed on lysates extracted from tumors. Protein concentration was determined, and equal amounts of total proteins were separated on SDS-PAGE. Antibodies used included Pak1, Mek, Erk, phospho-Erk1/2 (pThr202/pTyr204), Akt, phospho-Akt (pThr308), GSK3β, phospho-GSK3β (pSer9), mTOR, phospho-mTOR (pSer2448), p70S6K, phospho-p70S6K (pThr389), S6, phospho-S6 (pSer235/p236), and cyclin D1 from Cell Signaling Technology; K-ras, phospho-Pak (pSer141) and phospho-Mek (pSer298) were from Invitrogen. GAPDH was used as loading control.

Statistical analysis

Kaplan-Meier survival curves and statistical analysis were performed using GraphPrism program. Tumor volumes of mice treated with different inhibitors were compared with control group using a two-tailed Student’s t-test with P < 0.05 considered statistically significant

Results

Expression of Pak1 correlates with advanced grade of human squamous cell skin cancer

In normal human skin, Pak1 expression was largely restricted to the basal layer and scattered inflammatory cells (Figure 1A and Supplemental Results) by IHC. Consistent with this, diffuse cytoplasmic Pak1 expression was seen in 5/5 (100%) of basal cell carcinomas (Figure 1B). In squamous cell carcinomas (SCC), expression of Pak1 was associated with a more aggressive grade; 31/45 (69%) of moderate/poorly differentiated cases showed diffuse cytoplasmic Pak1 expression (Figure 1C, and Supplemental Results; Table S1), but only 27/69 (39%) of well-differentiated cases (Figure 1D) expressed cytoplasmic Pak1 (p=0.0027). In SCCs, Pak1 expression was also associated with activation of Erk (p=0.0621) and Akt (p=0.05), though the former association did not reach statistical significance (Supplemental Results, Table S2A). It should be noted that we assessed total Pak1 levels but not phospho-Pak1 levels because, in our experience, antibodies to activated forms of Pak1 did not perform well in IHC settings.

Figure 1.

Pak1 immunohistochemistry in human skin cancer. (A) Pak1 immunohistochemistry showing basal expression in keratinocytes and Langerhans cells in normal human skin. (B) In a basal cell carcinoma and a (C) poorly differentiated SCC, diffuse cytoplasmic Pak1 expression is observed. Nuclear Pak1 expression is also seen. (D) In well-differentiated SCC, Pak1 expression is either negative or restricted to keratinocytes adjacent to the stroma.

Expression of the proliferation markers cyclin D1 and Ki67 was also investigated in the same human tissue microarrays. Cyclin D1 and Ki67 immunohistochemistry was performed as previously described using antibodies SP4 (DAKO, Carpinteria, CA) and SP6 (Neomarkers, Thermo Fisher Scientific, Kalamazoo, MI, USA) (21). The percentage of tumor cell nuclei that were positive for Cyclin D1 or Ki67 was scored in each core. The mean percentage of nuclei was compared in Pak1 negative (score = 0) and Pak1 (score ≥ 1) positive cases. Both cyclin D1 and Ki67 were significantly positively associated with Pak1 expression. Pak1 positive cases (n = 58) had a mean 13.9% of cyclin D1 positive cells, compared to Pak1 negative cases (n = 56), which had a mean 5.7% (p=0.0092, unpaired two-tailed t-test). Pak1 positive cases (n = 58) had a mean 35.0% of Ki67 positive cells, compared to Pak1 negative cases (n = 56), which had a mean 15.2% (p < 0.0001, unpaired two-tailed t-test). Supplemental Results; Table S2B

Role of Pak1 in a conditional model of Kras-driven squamous cell skin cancer

To investigate the effect of Pak1 signaling on Ras-mediated oncogenesis in mice, we employed a well-characterized mouse model in which transgenic animals express a tetracycline-inducible oncogenic Kras^G12D mutant under the control of the K5 promoter, which is active in keratinocytes (16). Upon doxycycline induction, these mice develop a broad range of proliferative lesions and carcinomas in squamous epithelial cells such as skin and salivary gland (16, 22). Mice carrying K5-rTA or tet-Kras^G12D transgenes were bred with Pak1 knockout mice (17) separately to generate Pak1^+/−::K5-rTA and Pak1^+/−::tet-Kras^G12D animals. Intercrosses of these mice generated Pak1 wild-type, heterozygous, and knockout mice bearing the K5-rTA and tet-Kras^G12D transgenes, in which Kras^G12D activation is restricted to the basal stratified epithelial cells upon doxycycline induction.

In non-transgenic mice, loss of Pak1 did not affect skin development or cellular architecture (Figure S1). In wild-type and heterozygous animals, Pak1 immunohistochemistry showed staining in the epidermis, hair follicles and sebaceous glands. Pak1^−/− mice exhibited an almost total absence of Pak1 immunostain in the epidermis with mild immunostain still remaining in some sebaceous glands. Interestingly, Pak2 immunohistochemistry showed that all cutaneous structures expressed Pak2 in the three mouse groups including Pak1^−/− mice.

Mice at three months of age were fed either a control diet or a diet containing doxycycline. In mice containing both the K5-rTA and tet-Kras transgenes, the presence of doxycycline in the diet resulted in two to three-fold elevated ras expression whereas doxycycline did not affect Kras levels in animals that contained only a single transgene (i.e., K5-rTA or tet-Kras^G12D alone) (Figure 2A). Interestingly, both Pak1 and Pak2 protein levels were two- to four-fold elevated in K5-rTA::tet-Kras^G12D mice after doxycycline induction of Kras^G12D expression.

Figure 2.

Effect of Pak1 on Kras-driven tumorigenesis. (A) The effect of doxycycline administration on Kras, Pak1, and Pak2 protein levels in single transgenic (K5-tTR or tet-Kras) and doubly transgenic mice (K5-tTR::tet-Kras), fed a control or a doxycycline-containing diet. Kras, Pak1, and Pak2 expression in skin was analyzed by immunoblot. Numerals represent relative expression to GAPDH. (B) Kaplan-Meier tumor-free survival curves showing effects on latency of tumor formation in Pak1^−/− mice versus Pak1^+/+ and Pak1^+/− mice. (C) Median days until onset of tumor initiation, (D) total number of tumors and (E) tumor volume. Abbreviations: SCC, squamous cell carcinoma; ND, not detected. * Median treatment days to detect tumor.

As early as five days after doxycycline administration, K5-rTA::tet-Kras^G12D mice developed multiple dome-shaped lesions on the surface of the skin. Almost all such lesions rapidly transformed to frank squamous cell carcinoma (SCC). Thus, activation of the K-ras oncogene is tightly regulated in vivo and results in rapid formation of SCC. A control group of mice bearing only single transgenes (K5-rTA or tet-Kras^G12D) that were treated with doxycycline did not develop any lesions or exhibit any signs of illness, irrespective of Pak1 genotype. No control mice died or developed skin lesions during this time period. (data not shown).

We compared the incidence and latency of tumor appearance in Pak1^+/+, Pak1^+/− and Pak1^−/− mice in the presence of the Kras^G12D transgene. After induction of Kras, Pak1^+/+ mice rapidly developed skin tumors, with 50% of mice bearing visible lesions by 8 days, and all mice bearing visible lesions by twenty days (Figure 2B, 2C). In contrast, the Pak1 heterozygotes took twice as long to develop visible lesions and a few mice remained tumor free by forty days. Pak1-null mice had an even greater latency period and about a third of the mice remained free of visible tumors at 40 days. Pak1^+/+ mice showed much earlier tumor development, with a median of eight days until tumors detection versus seventeen days for Pak1^+/− and 25 days for Pak1^−/− mice (Figure 2C). The significant delay in tumor initiation and progression in Pak1^−/− mice was associated with increased survival, as all Kras^G12D expressing Pak1^+/+ and Pak1^+/− mice had to be euthanized by one month, whereas 86% (6/7) of matched Pak1^−/− mice were alive at one month. While most of the K5-rTA::tet-Kras^G12D mice fed with doxycycline-containing food eventually developed papillomas and SCC, irrespective of the Pak1 genotype, the total number of tumors (Figure 2D) and the average total tumor volume was dramatically reduced in Pak1^−/− mice (40 mm³) compared with those in Pak1^+/− (90 mm³) and Pak1^+/+ (140 mm³) mice (Figure 2E).

Histological analysis indicated that the tumors from Pak1^+/+::K5-rTA::tet-Kras^G12D mice were hyperproliferative, as shown by an increased number of Ki67-positive cells (Figure 3A and 3B). There was no statistically significant difference between Ki67 staining of tumors from these mice and their counterpart heterozygous Pak1^+/− mice (23% vs. 21%). On the other hand, Ki67 staining was markedly reduced in Pak1^−/−::K5-rTA::tet-Kras^G12D mice (Figure 3A and 3B). In addition, skin lesions were detected in 100%, 87% and 71% of Pak1^+/+ (n=15), Pak1^+/− (n=15), and Pak1^−/− (n=7) mice, respectively (Figure 3C). In Pak1^+/+ mice, only 20% of tumors were papillomas while 80% were SCC, with some presenting a predominantly well-differentiated pattern of squamous differentiation but most featuring a poorly differentiated squamous histotype. The latter tumors frequently invaded the subjacent dermal connective tissue (Figure 3A). However, this advanced grade of carcinoma progression was much less common in Pak1^−/− mice. In Pak1^−/− mice, about 60% had benign tumors (papillomas) and just 14% SCCs (Figure 3C), and these few tumors tended to be smaller than those seen on wild-type mice (Figure 2E). Taken together, the rapid tumor onset and progression of SCC in Pak1^+/+ but not Pak1^−/− mice suggests that Pak1 is a key component in Kras-mediated oncogenesis. In this aggressive disease model system, deletion of the Pak1 gene is sufficient to slow tumor initiation and progression, alter the tumor spectrum, and prolong survival.

Figure 3.

Effect of Pak1 gene dosage on histology and proliferative capacity of Ras-driven SCC. (A) Tumors in Pak1^+/+, Pak1^+/− and Pak1^−/− mice stained with H&E and Ki-67. Note reduced Ki67-positive staining cells in tumors of Pak1^−/− mice. Magnifications ×10 and ×40 for H&E staining and Ki-67 respectively. Scale bar =100 µm. (B) Histogram showing Ki-67 labeling index (i.e. percentage of positively stained cells) of tumors from Pak1^+/+, Pak1^+/− and Pak1^−/− mice. (C) Percentage of mice with papillomas or SCC in K5-tTR::tet-Kras animals.

Down-regulation of multiple signaling pathways in Pak1^−/− tumors

Pak1 has been implicated in regulating Erk signaling downstream of Ras, via phosphorylation of Mek1 and c-Raf (14, 23, 24) and also in activating Akt via a scaffolding interaction with PDK1 (12, 15). To investigate the molecular mechanisms underlying reduced cancerous growth in Pak1^−/− mice, we investigated the status of the Raf-Mek-Erk and PI3K-Akt pathways in the Kras^G12D model. We found that the Raf-Mek-Erk pathway is persistently activated in the epithelial lesions after onset of Kras mutant expression in transgenic mice. As shown in Figures 4A and S2, activation of Erk signaling cascade was observed in tumors of Pak1^+/+ and Pak1^+/− mice. Strikingly, we found a strong reduction in phosphorylation of Mek and Erk in tumors isolated in Pak1^−/− mice (Figure 4A and Figure S2).

We next determined the status of the Akt pathway. Activation of Akt and GSK3β were markedly attentuated in Pak1^−/− mice, as assessed by immunoblot (Figure 4B) and IHC (Figure S2). In addition, there was a decrease in the phosphorylation of Akt targets mTOR, p70 S6K, and S6 in Pak1^−/− tumors (Figure 4C and Figure S2). Suppression of cyclin D1 expression was also noted in Pak1^−/− tumor tissues (Figure 4C). Interestingly, in the few cases of large SCC observed in Pak1^−/− animals, c-Raf, Erk, and Akt activity was restored, as was P-Pak (Figure 4D). As these tissues lack Pak1, the ~67 kD P-Pak signal most likely represents Pak3, which has an identical migration as Pak1 on SDS/PAGE.

Inhibition of tumor growth and Ras pathway by small molecule signaling inhibitors

The long latency and slow growth of tumors in Pak1^−/− mice could be related to defective activation of the Erk and/or Akt-mTOR signaling pathways that require Pak1 for full activity. To determine the relative contributions of these pathways, we used small-molecule inhibitors that target Pak directly, or that selectively target Erk or Akt activation. First, two Pak inhibitors - PF03758309, which potently suppresses both group A (Pak1, -2, -3) and B (Pak4, -5, -6) Paks (19), and FRAX-597, a new group A-specific Pak inhibitor (Figure 5B) – were assessed for effects on tumor growth and signaling. It should be noted that both Pak inhibitors have certain off-target effects on other kinases, but that these off-target effects are largely non-overlapping (Figure 5A, Figure S3, and Table S3).

Figure 5.

Tumor regression in mice treated with Pak inhibitors. (A) Specificity of Pak inhibitors. 1 µM FRAX597 or PF3758309 was tested for inhibitory activity against a panel of >300 protein kinases (Invitrogen). Inhibition of >50% is represented by yellow spheres, and >75% by red spheres, mapped onto a kinome tree diagram. Inset shows protein kinase family groups plotted on tree diagram. Unique and common targets of the two kinase inhibitors (>75% inhibition) are shown on a Venn diagram. (B) Structure of Pak inhibitor FRAX-597. (C–F) Volumetric changes in tumor size between untreated mice (vehicle) and mice treated with inhibitors for seven (FRAX597) or ten (PF03758309, PD0325901, GSK690693) days. (G–J) Immunoblot analysis of tumor lysates from animals treated with inhibitors or vehicle.

To test the effects of these compounds, a cohort of K5-rTA::tet-Kras^G12D mice were fed a doxycycline diet and, when tumors became visible (~5 days post onset of doxycyline), the mice were separated into two groups and treated daily with either inhibitor or vehicle, while remaining on a doxycycline diet. Mice were weighed weekly and tumor volume measured every two days for two weeks. The inhibitory effects of these compounds were evaluated at the end of the treatment course. Mice treated with either Pak inhibitor showed marked tumor regression. In the case of the pan-Pak inhibitor PF03758309, average tumor volume was reduced by 92% (from 64.80 ± 6.52 mm³ to 4.88 ± 0.89 mm³) (Figure 5C). Similar effects were noted in mice treated with FRAX-597, with reductions on average tumor volumes of 89% (Figure 5D). In treated mice, tumor tissue, if any remained, was characterized by an increase in apoptosis without a notable change in proliferation, as assessed by staining with cleaved caspase-3 and Ki-67, respectively (Figure S4B).

For both cohorts, tumor tissues were examined by immunoblot analysis to determine the effects of the Pak inhibitors on signal transduction in vivo. As expected, both PF03758309 and FRAX-597 induced near complete abolition of Pak activation (Figure 5G and 5H). Interestingly, FRAX-597, but not PF03758309 treatment was associated with marked loss of total Pak1 expression (Figure 5G and 5H). Levels of Pak2, but not Pak4, were also reduced in tumor tissues from these animals (Figure 5H). For either compound, Mek and Erk activity was reduced, comparable to levels seen in Pak1 knockout mice (Figure 5G and 5H). Interestingly, Akt activity, as assessed by anti-phospho-Thr308 antisera, was also severely reduced in mice treated with either anti-Pak agent. Analysis of tumor tissue from PF03758309-treated mice showed a marked increase in apoptosis, consistent with Akt inhibition (Figure S4B). It should be noted that SCC was not present in FRAX-597 treated mice, precluding evaluation of apoptosis in tumor tissue.

Treatment of mice with the Mek inhibitor PD0325901 (25) had similar beneficial effects on tumor regression (Figure 5E), in the absence of any notable toxicity such as anemia. With this compound, activation of Mek and Erk, but not Pak or Akt, was reduced (Figure 5I). In contrast, the Akt inhibitor GSK690693, which has activity in leukemia cell lines (26), in xenografts models (20), and in pre-clinical settings (27), had only a small effect on SCC tumor regression, even at a dose that markedly reduced Akt signaling activity (Figure 5F and 5J) and induced apoptosis (Figure S4). As expected, GSK690693 did not notably affect the activity of Pak or Erk (Figure S4). These data show that Akt inhibition alone does not have a robust anti-tumor effect in this animal model.

Discussion

This study represents the first use of Pak-deficient animals and small molecule Pak inhibitors to study the role of these kinases in a genetically engineered mouse model of cancer. We report that loss of Pak1 gene function caused a marked reduction in the number, latency, and progression of Kras-driven squamous cell skin cancer. In cancer cells from such animals, Erk and Akt signaling activity were severely reduced. Moreover, loss of Pak function induced by either of two distinct small molecule Pak inhibitors was associated with regression of Ras-driven tumors of the skin, as well as reduction in Erk and Akt activity. These results could be replicated treating the mouse model with a small-molecule inhibitor of Mek, but not with an inhibitor of Akt signaling. These data suggest that Pak1, via activation of the Erk cascade, is required for efficient tumorigenesis and tumor maintenance in this aggressive and highly penetrant Kras cancer model.

Of the six Pak isoforms, Pak1 in particular has garnered much attention with respect to tumorigenesis, as amplification of the PAK1 gene, with concomitant overexpression of the Pak1 protein, is commonly observed in human cancers of the breast, ovary, and bladder (11). Unlike other oncogenic serine/threonine protein kinases such as B-Raf, activating point mutations or deletions of Pak1 have not been found in human cancers, despite the ability of activated Pak1 mutants to transform cells in vitro and in vivo. PAK1 gene amplification appears to be particularly relevant in human breast cancer, as it has been shown that such amplification is associated with resistance to tamoxifen treatment (28), but may well also play a role in Ras-driven tumors. For example, a recent study by Ong et al. reported that elevated Pak1 expression is prevalent in 61% of head and neck tumors and 64% of squamous non-small cell lung cancer (NSCLC) (21). In both human head and neck and in human NSCLC cell lines with PAK1 gene amplification, reduction of Pak1 expression by shRNA induced loss of Erk activity and gain of caspase activity, accompanied by increased cellular apoptosis. As both these cancers are associated with frequent KRAS mutation, these data suggest a general role for Pak1 in Ras-driven tumorigenesis. Such studies are also consistent with earlier work showing that active Pak1 is required for transformation by Kras in vitro (14, 15, 29).

Wang et al. recently reported that Rac1, a direct upstream activator of Pak1, is essential for DMBA/TPA-induced skin tumor formation in mice (30). In this model of skin cancer, Hras (and occasionally Kras) is frequently mutated and represents the essential initiating oncogenic event (31, 32). The effects of Rac1 gene deletion in this model were likely mediated at least in part by group A Paks, as loss of the Rac1 gene was accompanied by decreased keratinocyte hyperproliferation and diminished activation of Mek and Akt, two pathways known to be linked to Pak. Interestingly, these signaling pathways were not altered in untreated Rac1-deficient skin, indicating a hyperproliferation-specific function of Rac1 in vivo. Our data are consistent with these findings, as Pak1-deficient mice do not have any notable defects in skin or hair development (Figure S1), yet resist Kras-driven carcinogenesis and show loss of Erk activation in this setting. We have also previously shown that Pak1 regulates Erk activation downstream of Ras in an NF1 mast cell model, in which Ras is activated (33). These findings support a model in which Pak1 is required for efficient activation of the Erk pathway by Ras in various cell types. Interestingly, our data also suggest that in some conditions, other group A Paks such as Pak3 may substitute for Pak1, as the rare large tumors that arose in Pak1^−/−::K5-rTA::tet-Kras^G12D mice showed restoration of an active Pak species of 67 kD, accompanied by reactivation of c-Raf, Erk and Akt (Figure 4D).

The cell of origin in Ras-driven SCC is thought to reside in the hair follicle stem cell niche or from immediate progenitors (34, 35). We found that Pak1 is expressed in this compartment in both human and murine skin (Figures 1A and S1). Both the Erk and Akt pathways are known to be activated during Ras-induced tumorigenesis from hair follicle stem cells; thus the effect of Pak1 loss or inhibition might be mediated by reduced signaling through Erk and Akt in these cells.

Interestingly, we found that Akt activity was reduced in mice treated with either anti-Pak agent, PF03758309 or FRAX-597. While these results are expected for FRAX-597, which induces loss of Pak1 protein and thus essentially phenocopies Pak1^−/− mice, it is less clear why the pan-Pak inhibitor PF03758309 also inhibited Akt signaling, as the scaffold functions of Pak1 that are thought to link it to PDK and Akt would presumably be unaffected by this ATP-competitive compound. It is possible that Akt inhibition in this setting represents an off-target effect of PF03758309. Indeed, we found that PF03758309 has substantial inhibitory activity against the Akt activator PDK1 (Fig. 4A and Table S3), and Murray et al. reported that one Akt isoform, Akt3, is strongly inhibited by this compound (19). Alternatively, PF03758309 might stabilize a particular conformation of Pak1 that shields its scaffolding elements.

Regarding the new group A Pak inhibitor, FRAX-597, this molecule has the interesting and unexpected property of reducing Pak1 and Pak2 expression levels in treated animals (Figure 5H). While the basis for these effects are not yet understood, we have found that similar effects are also seen in vitro when various growth-factor-stimulated cell lines are exposed to FRAX-597, and that loss of Pak expression, but not of Pak enzymatic inhibition, can be prevented by proteosome inhibitors such as MG-132 (data not shown). Combined with the animal data depicted in Figure 5H and the biochemical data in Table S3, these data suggest that FRAX-597 has a dual inhibitory effect on group A Paks: it acts as a competitive inhibitor and also as a destabilizing agent, perhaps by binding to an ‘open’ form of Pak.

In human skin cancer, we found that the expression of PAK1 is highly associated with advanced grade of SCC as illustrated in poorly differentiated carcinomas or undifferentiated tumors (Figure 1C). In our mouse model of SCC, expression of Kras was associated with increased Pak1 expression (Figure 2A). Whether Ras signaling augments Pak1 gene transcription or translation, or affects the stability of the Pak1 protein, is not known, though it is of interest that Reddy have reported that levels of the micro-RNA miR-7, which down-regulates Pak1 expression, are reduced in highly invasive breast cancer cells compared to their non-invasive counterpart (36). In our Kras-driven mouse model of SCC, Pak1 levels correlated with histologic grade: 80% of tumors isolated in Pak1^+/+ mice were histologically SCC whereas only 14% of epidermal lesions in Pak1^−/− mice displayed a malignant phenotype (Figure 3C). Also of note, there was 25% reduction of SCC found between Pak1^+/− and Pak1^+/+ mice (60% vs. 80%). However, loss of one Pak1 allele did not affect the activity of the Erk or Akt-mTOR pathways (Figure 4), suggesting that signaling pathways in addition to Erk and Akt-mTOR are also regulated by Pak1.

Disruption of Ras-signaling pathways has been a major goal of anti-cancer drug development, with a particular emphasis on identifying key, targetable signaling proteins downstream of Ras. There is ample evidence supporting a key role for PI3K-Akt pathway in oncogenic Ras signaling, but our results suggest that, in the aggressive K5-tTR::Tet-Kras SCC model, Ras oncogenic signals are transmitted predominantly through a Pak1-Mek-Erk pathway (Figure S5). An inhibitor targeting Mek signaling (PD0325901) showed impressive tumor regression, consistent with other recent reports in Ras models (37–40). In particular, Scholl et al. have shown that genetic deletion of Mek1/2 in mouse epidermis abolishes transformation by Ras (41), and Ehrenreiter et al. have shown that c-Raf is also required for Ras-driven tumorigenesis in this setting (42). In contrast, we found only a small effect on tumor size with an Akt inhibitor (GSK690693), despite appropriate target pathway inhibition. This finding is consistent with the idea that inhibition of PI3K-Akt signaling alone is not adequate to diminish tumors driven by mutant Kras, once established, though such inhibition did add to the anti-tumor effects of a Mek inhibitor (43). Anti-Pak agents may offer a dual benefit by simultaneously inhibiting both Erk and Akt signaling, thus impeding proliferation and promoting apoptosis in cancer cells. Given that a Pak inhibitor (PF3758309) recently entered clinical trials (http://clinicaltrials.gov/show/NCT00932126), it should soon be possible to determine the potential of Pak proteins as anti-cancer targets in human malignancies.