PF-3758309

P21-Activated Kinase 1 (PAK1) as a Therapeutic Target in BRAF Wild-Type Melanoma
Christy C. Ong, Adrian M. Jubb, Diana Jakubiak, Wei Zhou, Joachim Rudolph, Peter M. Haverty, Marcin Kowanetz, Yibing Yan, Jarrod Tremayne, Richard Lisle, Adrian L. Harris, Lori S. Friedman, Marcia Belvin, Mark R. Middleton, Elizabeth M. Blackwood, Hartmut Koeppen, Klaus P. Hoeflich
Manuscript received July 19, 2012; revised February 15, 2013; accepted February 20, 2013.
Correspondence to: Klaus P. Hoeflich PhD, Department of Translational Oncology Genentech, MS 50, 1 DNA Way, South San Francisco, CA 94080 ([email protected]).

Background Although remarkable clinical response rates in melanoma have been observed using vemurafenib or dabrafenib in patients with tumors carrying oncogenic mutations in BRAF, a substantial unmet medical need remains for the subset of patients with wild-type BRAF tumors.
Methods To investigate the role of p21-activated kinases (PAKs) in melanoma, we determined PAK1 genomic copy number and protein expression for a panel of human melanoma tissues. PAK1 was inhibited in vitro and in vivo using RNA interference or PF-3758309 inhibitor treatment in a panel of melanoma cell lines with known BRAF and RAS (rat sarcoma) genotype to better understand its role in melanoma cell proliferation and migration. Tumorigenesis was assessed in vivo in female NCR nude mice and analyzed with cubic spline regression and area under the curve analyses. All statistical tests were two-sided.
Results Strong cytoplasmic PAK1 protein expression was prevalent in melanomas (27%) and negatively associated with activating mutation of the BRAF oncogene (P < .001). Focal copy number gain of PAK1 at 11q13 was also observed in 9% of melanomas (n = 87; copy number ≥ 2.5) and was mutually exclusive with BRAF mutation (P < .005). Selective PAK1 inhibition attenuated signaling through mitogen-activated protein kinase (MAPK) as well as cytoskeleton- regulating pathways to modulate the proliferation and migration of BRAF wild-type melanoma cells. Treatment of BRAF wild-type melanomas with PF-3758309 PAK inhibitor decreased tumor growth for SK-MEL23 and 537MEL xenografts (91% and 63% inhibition, respectively; P < .001) and MAPK pathway activation in vivo. Conclusions Taken together, our results provide evidence for a functional role of PAK1 in BRAF wild-type melanoma and thera- peutic use of PAK inhibitors in this indication. J Natl Cancer Inst;2013;105:606–615 Malignant melanoma accounts for approximately 80% of deaths from skin cancer. Although melanoma is surgically curable when discovered at early stages, regional and systemic spread of the disease considerably worsens the prognosis, with only 14% of metastatic melanoma patients surviving for 5 years (1). The mitogen-activated protein kinase (MAPK) pathway has recently been elucidated as a critical growth pathway in several melanoma subtypes (2). For instance, from a pooled analysis of 4493 patients, the occurrence of BRAF (v-Raf murine sarcoma viral oncogene homolog B1) mutation was 41% in cutaneous melanomas (3). The most frequent BRAF somatic mutation in malignant melanoma is substitution of valine at residue 600 to confer constitutive catalytic activity and signaling (4). Genetic studies have confirmed that BRAF is required for initiation and maintenance of melanoma in preclinical model systems (4–6). These discoveries prompted a flurry of drug discovery activity to develop small molecule inhibitors of BRAF, including GDC-0879, PLX-4720, PLX-4032/ vemurafenib, and dabrafenib (7–10). These inhibitors selectively decrease the growth of BRAF oncogene addicted tumor cells and provide hope for patients with the subset of melanoma that has activating mutations in the BRAF oncogene (11). However, much less antitumor efficacy with current BRAF small molecule inhibitors is observed for wild-type BRAF melanoma cells (7,8), raising the need to identify additional melanoma-associated driver genes to provide new insights into the biology, oncogenic signaling, and possible therapeutic targets for disease management of melanoma patients of all classifications. The RAF kinase family is composed of three members, ARAF, BRAF, and CRAF, which play a pivotal role in transducing signals in the canonical MAPK signaling pathway from RAS (rat sarcoma gene) to downstream kinases MEK1/2 (MAPK/ERK kinase 1/2) and ERK1/2 (extracellular signal-regulated kinase 1/2). However, additional kinases have been reported to also play a role in ERK activation. In particular, several groups have reported that group 1 p21-activated kinases (PAKs), serine/threonine protein kinases that serve as important mediators of Rac (RAS-related C3 botulinum toxin substrate) and Cdc42 (cell division control protein 42 homolog) GTPase function, as well as pathways required for RAS-driven tum- origenesis, contribute to MAPK pathway activation. PAK1 phospho- rylates CRAF at Ser338, a critical residue for activation, and MEK1 at Ser298, a site that is proximal to the activation loop residues Ser217/Ser221 that are substrates for the RAF kinases (12–14). The pathway cross-talk between PAKs and the MAPK pathway signaling in epithelial cells can be induced by a variety of conditions, includ- ing growth factor stimulation and cell adhesion to the extracellular matrix (15–17). As a major downstream effector of the Rho family small GTPases Cdc42 and Rac1, PAK1 also plays a fundamental role in linking extracellular signals to changes in actin cytoskeleton organization, cell shape, and adhesion dynamics (18–20). PAK1 is widely expressed in a variety of normal tissues, and expression is sub- stantially increased in breast and lung cancers (21–23). Functional studies have also implicated PAK1 in cell transformation (24) and tumor growth (23,25,26). These findings indicate that PAK1 may contribute to tumorigenesis in some disease contexts. Herein, we show that PAK1 protein is genomically amplified and highly expressed in primary melanomas and that PAK1 dysreg- ulation is negatively associated with mutational activation of BRAF. Based on these data, we hypothesized that PAK1 may be required for MAPK pathway activation and, accordingly, also proliferation of BRAF wild-type melanoma cells. Both genetic and pharmaco- logical approaches were used to demonstrate the dependence of BRAF wild-type melanoma cells on PAK1 activity. Our results sup- port further characterization of PAK1 as a novel therapeutic target in this subset of melanoma that has substantial unmet medical need. Methods Tissue Analyses Immunohistochemistry (IHC) was performed as described previ- ously (23). Antigen retrieval was performed by incubating slides in tris(hydroxymethyl)aminomethane-ethylenediaminetetraacetic acid (Tris-EDTA), pH 9.0 using a Decloaking Chamber (Biocare Medical, Concord, CA). Sections were incubated with a primary anti–p21-activated kinase 1 (PAK1) rabbit polyclonal antibody at 1:100 dilution (catalog number 2602; Cell Signaling Technology, Danvers, MA). Bound antibody was labeled with anti-rabbit/mouse Envision (DAKO, Carpinteria, CA). Intensity of PAK1 expression was scored separately in the cytoplasm and nuclei of neoplastic cells on a scale of zeor to three. The highest intensity score among repli- cate cores was used as the score for each patient. The same patholo- gist (A.M. Jubb) scored all cases, blind to the clinical data. The χ2 test was used to evaluate associations between categorical variables. Mutation status was determined for BRAF codon 600 and NRAS codons 12, 13, 61, and 146 by KASPar (KBioscience, Herts, UK) and conventional Sanger DNA sequencing methods. Cell Viability Assays Melanoma cell lines that are BRAF wild-type and NRAS wild-type include SK-MEL23, 537MEL, MeWo, and SK-MEL30. NRAS mutant melanoma lines included HMV-II (Q61K), IPC-298 (Q61L), MEL-JUSO (Q61L), and SK-MEL-31 (Q61K). C32 cells have both NRAS(Q61R) and BRAF(V600E) mutations. SK-MEL-2 cells are NRAS(Q61R) mutant but are PAK1-deficient because of chromo- some loss at 11q13. Cell line identity was verified by high-throughput single nucleotide polymorphism genotyping using Illumina Golden gate multiplexed assays (Illumina, Hayward, CA). Cell transfections and treatments were performed using short interfering RNA (siRNA) oligonucleotides for PAK1 and PAK2 from Dharmacon RNAi Technologies (Chicago, IL); GDC-0879, PLX-4720, G-945, and PF-3758309 were described previously (5,8,27). These BRAF inhibi- tors have similar properties with respect to biochemical and cellular potency. For instance, BRAF(V600E) median inhibitory concentra- tion (IC50) values are 0.18, 13, and 0.13 nM, and p-ERK1/2 cellular half maximal effective concentration (EC50) values are 4.6, 46, and 63 nM for G-945, PLX-4720, and GDC-0879, respectively. Cellular viability was assessed by ATP content using the CellTiter-Glo Luminescent Assay (Promega, Madison, WI), and results represent mean 95% con- fidence intervals from three independent experiments. Immunoprecipitation and Enzymatic Assays Immunoprecipitations and kinase assays for MEK1, MEK2, and CRAF were performed as described previously (28). In brief, cell lysates were incubated with rabbit anti-CRAF antibody and 50 μL of Protein A agarose beads (EMD Millipore, Billerica, MA) for 2 hours at 4°C. After washing with lysis buffer plus protease and phosphatase inhibitor mixtures, protein A beads were incubated with 0.4 μg of inactive MEK1 (EMD Millipore, Billerica, MA) in 40 μL of kinase buffer (20 mM MOPS, pH 7.2, 25 mM β-glycerol phosphate, 5 mM ethylene glycol tetraacetic acid (EGTA), 1 mM sodium orthovanadate, 1 mM DTT, 120 μM ATP, 18 mM magne- sium chloride) for 30 minutes at 30°C. Samples were analyzed by gel electrophoresis and immunoblotting. Tumor Xenograft Models Cultured SK-MEL-23, A2058.X1, and A375.X1 cells were removed from culture, suspended in Hank’s buffered saline solution, mixed 1:1 with Matrigel (BD Biosciences, San Jose, CA), and implanted subcutaneously into the right flank of naive female NCR nude (Taconic Farms, Hudson, NY) or Beige Nude XID (Harlan Laboratories, Livermore, CA) mice. Mice with tumors of a mean volume of approximately 250 mm3 were randomized into treatment cohorts (n = 10 mice per group). Tumor volumes were calculated by the following formula: tumor volume = 0.5 × (a × b2), where “a” is the largest tumor diameter and “b” is the perpendicular tumor diameter. Tumor volume results are presented as mean tumor volumes ± 95% confidence intervals. Percent growth inhibition at the end of study was calculated as: percent growth inhibition = 100 [(end of study vehicle − end of study treatment)/(end of study vehicle)]. Data analysis and generation of P values using the Dunnett t test was done using JMP software (SAS Institute, Cary, NC). All experimental procedures conformed to the guiding principles of the American Physiology Society and were approved by Genentech’s Institutional Animal Care and Use Committee. For pharmacodynamics studies, the xenograft tumors were harvested 1 hour after dosing. Statistical Analysis To analyze the repeated measurement of tumor volumes from the same animals over time, a mixed modeling approach was used (29). This approach addresses both repeated measurements and modest dropouts due to any nontreatment-related death of animals before study end. Cubic regression splines were used to fit a nonlinear pro- file to the time courses of log2 tumor volume at each dose level. These nonlinear profiles were then related to dose within the mixed model. Tumor growth inhibition as a percentage of vehicle (%TGI) was calculated as the percentage of the area under the fitted curve (AUC) for the respective dose group per day in relation to the vehi- cle, using the formula %TGI = 100 × (1  AUCdose/AUCveh). To get uncertainty intervals for %TGI, the fitted curve and the fitted covariance matrix were used to generate a random sample as an approximation to the distribution of %TGI. The random sam- ple is composed of 1000 simulated realizations of the fitted-mixed model, where the %TGI has been recalculated for each realization. Our reported uncertainty interval gives the values for which 95% of the time the recalculated values of %TGI will fall in this region given the fitted model. The 2.5 and 97.5 percentiles of the simu- lated distribution were used as the upper and lower uncertainty intervals. Plotting was performed and generated using R version 2.8.1 and Excel version 12.0.1 (Microsoft, Redmond, WA). Data were analyzed using R version 2.8.1 (R Development Core Team 2008; R Foundation for Statistical Computing, Vienna, Austria), and the mixed models were fit within R using the nlme package, version 3.1–89. All statistical tests were two-sided. Results Elevated PAK1 Protein Expression and Genomic Amplification in Melanoma To determine the possible extent of PAK1 dysregulation in human melanoma, we assayed primary tumor tissue from 87 melanoma patients for DNA copy number changes using high-resolution sin- gle nucleotide polymorphism arrays (Figure 1A). The frequency of PAK1 amplification was eight of 87 specimens with copy num- ber greater than or equal to 2.5 in this tumor panel. RNA was purified from 42 melanoma tumor and cell lines specimens, and increased PAK1 copy number was correlated with mRNA expres- sion (Spearman’s Rho = 0.59; P = .005) (Figure 1B). In compari- son, PAK2 and PAK3 genes were not amplified or highly expressed in melanoma tissues (Supplementary Figure 1, A and B, available online). Dysregulated PAK1 expression was more frequent than would be predicted by genomic amplification alone, thereby sug- gesting that additional transcriptional or regulatory mechanisms increase PAK1 expression in this indication (30,31). Elevated expression of PAK1 in melanoma compared with normal skin tis- sues was also demonstrated using gene expression data deposited in the Gene Expression Omnibus database (GSE4587). Interestingly, PAK1 gene amplification was preferentially observed in tumors that lacked activating mutations in the BRAF oncogene (22% vs 0% for BRAF wild-type or mutant, respectively; P = .005, two- sided t test) (Figure 1B). The levels of PAK1 mRNA expression differed between wild-type and BRAF(V600E) or BRAF(V600M) genotypes (P = .006 and P = .13, respectively). Taken together, this suggests that PAK1 could be a tumor-promoting “driver” gene in a subset of BRAF wild-type melanomas. To further evaluate the extent of PAK1 dysregulation in human melanomas, PAK1 protein expression level and subcellular localization were ascertained by IHC staining of a distinct set of tissue microarrays. Robust and selective IHC reactivity of PAK1 antibody was previously demonstrated (23). In malignant mela- noma, 46 of 92 (50%) primary tumor samples were positive for PAK1 expression, and 27% of all cases showed staining of moder- ate (≥2) or strong (≥3) intensity in the malignant cells (Figure 1C, panels c and d; Table 1). Nuclear localization of PAK1 was only evident in a very small proportion of melanomas. Identical results were seen with an alkaline phosphatase label and fast red chromo- gen in place of a horseradish peroxidase label and brown diamin- obenzidine (data not shown). PAK1 was weakly expressed in basal keratinocytes in normal skin, whereas lymphocytes and presumed Langerhans cells were positive for PAK1 expression (Figure 1C, panel d). Together, these data show that PAK1 DNA copy number, mRNA, and protein expression are broadly upregulated in human melanoma. Negative Association Between PAK1 Overexpression and BRAF Mutation in Primary Melanomas Given the prevalence of oncogenic mutation of BRAF and NRAS in melanoma (3), melanoma tissues were genotyped for known hotspot mutations in BRAF (codon 600) and NRAS (codons 12, 13, 61, and 146) genes. Genotype data for BRAF (39 Val600Glu, 1 Val600Lys, and 46 wild-type) and NRAS (1 Gln61His, 7 Gln61Lys, 1 Gln61Lys + Gln61Arg + Leu59Ala, 1 Gln61Leu, 19 Gln61Arg, 2 Gln61Arg + Gln61Lys, and 53 wild-type) were available for 86 and 84 tumors, respectively, and are consistent with the ranges of mutation frequencies that have been observed for similarly sized sample sets in the Catalogue of Somatic Mutations in Cancer (COSMIC) database. PAK1 IHC staining was scored blind to clinicopathological details, and mutation status and results are summarized in Table 1 and Supplementary Tables 1 and 2 (available online). Notably, PAK1 protein expression was dysregulated selectively in BRAF wild-type tumors (n = 19 of 46 were positive for strong IHC staining of PAK1) compared with melanomas expressing oncogenic V600E or V600K mutants (n = 4 of 40 tumors with high IHC staining). This negative correlation between PAK1 expression and BRAF mutation was statistically significant (P < .001, Fisher exact test). A similar trend, albeit not statistically significant, was observed when dichotomizing samples into only NRAS mutant and nonmutant status (Table 1). Ten melanoma tumors were identified with both BRAF and NRAS mutations, and each of these tumors was negative for PAK1 overexpression (n = 7 with IHC score of 0; n = 3 with score of 1). There was no statistically significant association between PAK1 protein expression and mitotic count (P = .61, Student t test), pathological tumor stage (P = .14, χ2 test), Breslow thickness (P = .85, Student t test), or ulceration (P = .91, χ2 test). Taken together, these results provide evidence that PAK1 dysregulation is strongly associated with cutaneous melanomas that lack oncogenic mutation of BRAF and define a subset of human melanoma for which there is no effective targeted therapy. PAK1 Expression and Proliferation of BRAF Wild-Type Melanoma Cells Given the genomic and histologic data for elevated PAK1 expression in the subset of human melanoma that is wild-type for BRAF, we examined PAK1 expression and the effect of RNA Figure 1. P21-activated kinase 1 (PAK1) expression in human melanoma. A) PAK1, at 11q13, is focally amplified in human melanoma tissues. B) PAK1 DNA copy and mRNA expression (226507_at Affymetrix MAS 5.0 signal) are correlated for melanoma tumor samples. C) Representative images of PAK1 immunohistochemistry in primary human malignant melanomas. Cytoplasmic expression score 0 (a), 1 (b), 2 (c), and 3 (d). Brown chromogen deposition indicates immunoreactivity against a blue hematoxylin counterstain. Scale bar = 200 μM. PAK1 expression is also seen in stromal cells (c) and cells intercalating within the epidermis that may represent Langerhan’s cells (d). The stratum basalis (arrow) and stratum spinosum (arrowhead) of the normal epidermis, as well as melanoma (asterisk), are indicated. Mb = megabases. Table 1. P21 activated kinase (PAK1) protein is overexpressed in v-Raf murine sarcoma viral oncogene homolog B1 (BRAF) wild-type melanoma* PAK1 IHC† MAPK activator Genotype 0, 1‡ 2, 3‡ P BRAF WT 27 19 .001 Mutant 36 4 NRAS WT 37 16 .61 Mutant 24 7 * All statistical tests were two-sided. P values were calculated using the Fisher exact test. IHC = immunohistochemistry; MAPK = mitogen-activated protein kinase; WT = wild-type. † Total number of patient tumors that were analyzed. ‡ PAK1 IHC scores are 0, 1, 2, or 3. interference–mediated knockdown of PAK1 in a panel of melanoma cell lines to clarify the contribution of PAK1 toward tumor cell proliferation. Increased PAK1 protein expression in melanomas expressing wild-type vs mutant BRAF was also observed for melanoma cell lines (Supplementary Figure 2, A and B, available online). Commercially available siRNA oligonucleotide duplexes were previously characterized for efficiency and selectivity of PAK1 and PAK2 knockdown (23). Melanoma cell lines 537MEL, MeWo, SK-MEL23, and SK-MEL30 express high levels of PAK1 protein. Transient knockdown of PAK1 by a pool of multiple PAK1-selective siRNA oligonucleotides resulted in a 1.8- to 4.3-fold reduction in cell viability when compared with cells transfected with a nontargeting, negative control siRNA oligonucleotide (P < .0001) (Figure 2A). Furthermore, inhibition of PAK1 generally reduced proliferation of BRAF wild-type melanoma cells relative to BRAF(V600E) cells (P < .07; n = 14), further supporting a role for PAK1 as a driver of proliferation in this melanoma subtype (Figure 2B). Melanoma cells with activating mutations of NRAS displayed variable sensitivity to PAK inhibition (Figure 2C). As an additional control, PAK1- selective siRNA oligonucleotides had no effect on SK-MEL2 cells, a cell line in which the chromosomal region encompassing the PAK1 gene is deleted (Figure 2C). To better assess the mechanism by which PAK1 contributes to proliferation, PAK-dependent cellular signaling was assessed in 537MEL and SK-MEL23 cells. MAPK pathway activation, as determined by phosphorylation of ERK and MEK, was dramatically inhibited by PAK knockdown (Figure 2D). In agreement with this result, cyclin D1 levels (which are essential for regulating cyclin-dependent kinases and G1/S progression) were also diminished as a consequence of PAK1 ablation. Certain NRAS mutant cell lines also displayed diminished MAPK pathway modulation, albeit to a lesser extent than BRAF wild-type cells, after perturbation of PAK1 expression (Supplementary Figure 2C, available online). PAK1 signaling in BRAF wild-type melanoma cells was further investigated using a reverse-phase protein array phosphoproteomics platform. Decreased signaling to MAPK, proliferation, nuclear factor-κB, cap-dependent protein translation, and cytoskeletal pathways was observed after PAK1 inhibition (Figure 2E). PAK1 has been shown to phosphorylate both CRAF(Ser338) and MEK1(Ser298) (17,32–34). Hence, we next sought to determine Figure 2. Contribution of p21-activated kinase 1 (PAK1) to proliferation of v-Raf murine sarcoma viral oncogene homolog B1 (BRAF) wild-type (WT) melanoma cells. A) Proliferation of melanoma cells after transfec- tion with small interfering RNA (siRNA) oligonucleotides was measured by Cell TiterGlo ATP consumption assay. PAK1 was required for cell growth, and the data are normalized to control and shown as the mean ± 95% confidence intervals. B) In a panel of melanoma cell lines, PAK1 inhibition selectively impaired growth of cells without BRAF(V600E) mutation (n = 4; 537MEL, MeWO, SK-MEL23, SK-MEL30) compared with those with BRAF(V600E) mutation (n = 9; P = .07; 624MEL, 888MEL, 928MEL, RPMI-7951, A375, Colo829, LOX-IMVI, Malme-3M, A375). C) Proliferation of neuroblastoma RAS viral (v-ras) oncogene homolog (NRAS) mutant melanoma cells was measured after transfection with siRNA oligonucleotides for nontargeting control (NTC), NRAS, or PAK1/2. SK-MEL2 cells contain a chromosomal deletion of the region containing the PAK1 gene (ΔPAK1). Data are normalized to control and shown as the mean ± 95% confidence intervals. D) Inhibition of PAK1/2 decreases extracellular signal-regulated kinase 1/2 (ERK1/2) and mitogen-activated protein kinase/extracellular signal-regulated kinase 1/2 (MEK1/2) phosphorylation and cyclin D1 levels in SK-MEL23 and 537MEL cells. E) PAK1/2 inhibition in SK-MEL23 BRAF WT mela- noma cells decreases signaling to the cytoskeletal, mitogen-activated protein kinase (MAPK), proliferation, and nuclear factor κB (NF-κB) pathways as determined by reverse phase protein array (RPPA) analy- sis. Normalized RPPA results are presented as mean ± 95% confidence intervals. Statistical tests were two-sided. . siNTC = nontargeting control siRNA; siNRAS = NRAS-specific siRNA; siPAK1 = PAK1-specific siRNA; ΔPAK1 = chromosomal deletion of PAK1 gene. Figure 3. p21-activated kinase 1 (PAK1)–mediated v-Raf murine sar- coma viral oncogene (CRAF) activation in v-Raf murine sarcoma viral oncogene homolog B1 (BRAF) wild-type melanoma cells. PAK1- and PAK2-selective or nontargeting control (NTC) small interfering RNA (siRNA) oligonucleotides were transfected into SK-MEL23 and 537MEL melanoma cells. After 48 hours, endogenous mitogen-activated protein kinase/extracellular signal-regulated kinase 1 (MEK1) (A), mitogen- activated protein kinase/extracellular signal-regulated kinase 2 (MEK2) (B), or CRAF (C) proteins were immunoprecipitated, and the complexes were immunoblotted to detect phosphorylation of residues critical for catalytic activation. Total protein levels in the immunocomplexes were also determined as loading controls. D) Cells were treated with dime- thyl sulfoxide (DMSO) or 5 μM PF-3758309 for 4 hours, and endog- enous CRAF was immunoprecipitated and immunoblotted for Ser338 phosphorylation. Total CRAF levels in the immunocomplexes are also shown. E) SK-MEL23 cells were treated with DMSO, 5 μM PF-3758309, or 20 μM IPA-3 for 4 hours. CRAF immunocomplexes were incubated with inactive MEK1 protein in kinase buffer for 30 minutes. Levels of phospho-MEK1 (Ser217/Ser221) were determined, and CRAF catalytic activity is reported as the levels of MEK1 phosphorylation normalized to total CRAF protein. IgG = immunoglobulin G; IP = immunoprecipitation; siPAK1 = PAK1-specific siRNA. the molecular mechanism by which PAK1 triggers activation of the MAPK pathway in BRAF wild-type melanoma cells. Because phospho-specific antibodies that are raised to the Ser217/Ser221 activation loop sites on MEK proteins do not distinguish between MEK1 and MEK2, the MEK isoforms were immunoprecipitated from cells transfected with either control or PAK-selective siRNA oligonucleotides, and MEK activation was detected by immu- noblotting with phospho-MEK1/2(Ser217/Ser221) antibodies. PAK knockdown diminished both MEK1 (Figure 3A) and MEK2 (Figure 3B) phosphorylation in 537MEL and SK-MEL23 cells. Because the Ser298 phosphorylation site on MEK1 is not con- served in MEK2, PAK-dependent activation of both MEK iso- forms would suggest that upstream signaling to CRAF might be a driver of MAPK pathway regulation in BRAF wild-type mela- noma cells. Accordingly, PAK ablation reduced phosphorylation of CRAF on Ser338, a residue critical for full activation of this kinase (Figure 3C). The dependence of CRAF(Ser338) phospho- rylation (Figure 3D) and CRAF effector signaling (Figure 3E) on PAK catalytic activity was also confirmed using PF-3758309, an inhibitor of PAKs that was in clinical development (27), and IPA-3, an allosteric inhibitor that binds PAK1–3 and prevents activation by Rho family GTPases (35). Taken together, the functional con- sequences of PAK1 blockade in BRAF wild-type melanoma cells encompasses pronounced cytostatic effects by reduced CRAF acti- vation and subsequent MAPK pathway signaling. Differential Sensitivity of BRAF Wild-type Melanoma Cells to PAK and BRAF Inhibition To more closely investigate the activity and cellular mechanism of action of PAK signaling within sensitive and insensitive tumor types, small molecule inhibition of PAK and BRAF were compared using SK-MEL23 BRAF wild-type and A375 BRAF(V600E) cells. Administration of PF-3758309 resulted in profound MAPK pathway modulation in SK-MEL23 cells (Figure 4A, lane 2), but not A375 cells (Figure 4A, lane 5), as determined by measure- ment of ERK1/2 and MEK1/2 phosphorylation on kinase loop residues that are critical for catalytic activity. In comparison, analysis of PLX-4720–mediated signaling changes revealed only modest inhibition of ERK1/2 and MEK1/2 phosphorylation in SK-MEL23 cells (Figure 4A, lane 3), whereas the same treatment conditions potently inhibited MAPK activation in BRAF(V600E) cells (Figure 4A, lane 6). As a control, no differences were noted for total ERK1/2 or MEK1/2 protein levels in this experiment. Consistent with previous reports, MEK1-Ser298 was confirmed as a PAK-specific phosphorylation site, but Ser298 phosphoryla- tion was unlinked to MEK activation loop phosphorylation in Figure 4. Differential sensitivity of mitogen-activated protein kinase (MAPK) signaling in v-Raf murine sarcoma viral oncogene homolog B1 (BRAF) wild-type (WT) and BRAF(V600E) melanoma cells to PAK inhibi- tion. A) SK-MEL23 and A375 cells were treated with dimethyl sulfoxide (DMSO), 5 µM PF-3758309, or 0.2 µM PLX-4720 for 4 hours, and lysates were analyzed for phosphorylation of MAPK pathway components. The normalized ratios for phosphorylated vs total extracellular signal- regulated kinase 1/2 (ERK1/2) levels (quantified by densitometry anal- ysis) as well as lighter and darker exposures of p–mitogen-activated protein kinase/extracellular signal-regulated kinase 1 (p-MEK1/2)(S217/ S221) immunoblots are shown. B) Ectopic expression of Flag epitope- tagged PAK1 drives MAPK pathway activation in A375 cells. Specificity was demonstrated using PF-3758309 inhibitor treatment as a control. C) Pharmacodynamic response of BRAF WT tumors measured by pho- sophorylation of v-Raf murine sarcoma viral oncogene (CRAF)(Ser338) after either vehicle or 25 mg/kg PF-3758309 administration. D) Antitumor efficacy of 10, 15, and 25 mg/kg PF-3758309 intraperitoneal daily dos- ing in the SK-MEL23 preclinical tumor model of BRAF WT melanoma. E) Antitumor efficacy of 15 and 25 mg/kg PF-3758309 intraperitoneal daily dosing in the 537MEL preclinical tumor model of BRAF WT mela- noma. Treatment PF-3758309 impaired tumor growth of both SK-MEL23 and 537MEL tumors relative to the vehicle cohort at the final day of dosing (P < .001). Results are presented as mean tumor volumes ± 95% confidence intervals. Ten mice were used per efficacy cohort. BRAF(V600E) melanoma cells (Figure 4A, lane 5). The biological consequence of PAK1 phosphorylation of MEK1-Ser298 is pres- ently not well understood; however it has been shown that PAK1- MEK1 signaling can be mediated by cell–cell contact and adhesion (15). PAK signaling was also induced by ectopic expression of Flag-PAK1 in BRAF(V600E) cells with only moderate endog- enous expression of PAK1 to confirm these findings,. Elevated PAK1 signaling in A375 cells resulted in a substantial increase in CRAF and MEK phosphorylation that was reversible by addition of PF-3758309 (Figure 4B), suggesting that acquisition of PAK1 overexpression could be another mechanism to overcome depend- ence on oncogenic BRAF in melanoma (36). To extend our in vitro observations, pharmacodynamic modu- lation by PAK small molecule inhibitors was evaluated using tumor xenograft models (Figure 4C). After tumor establishment, mice were either administered saline or PF-3758309 (25 mg/kg), and tumors were harvested 1 hour after dosing. Treatment with PF-3758309 resulted in a substantial decrease in CRAF, MEK1/2, and ERK1/2 phosphorylation in SK-MEL23 tumors (Figure 4C). In A2058.X1 BRAF(V600E) tumors, decreased phosphorylation of CRAF(Ser338) was not observed after PF-3758309 dosing (data not shown). The effect of PF-3758309 on growth and maintenance of SK-MEL23 and 537MEL BRAF wild-type tumors was also evalu- ated in efficacy experiments (Figure 4, D and E; Supplementary Figure 3, available online). Treatment with 15 and 25 mg/kg PF-3758309 impaired tumor growth of both SK-MEL23 (76% and 91% inhibition, respectively) and 537MEL (33% and 63%, respec- tively) relative to the vehicle cohort as measured on the final day of dosing (P < .001) (Supplementary Table 3, available online). In comparison, only minimal inhibition of either CRAF phosphoryla- tion or tumor growth were observed for SK-MEL23 tumors treated with a potent RAF inhibitor in vitro and in vivo (Supplementary Figure 4, A and B, available online) (7). Together, the magnitude of MAPK pathway inactivation by PF-3758309 was associated with antitumor efficacy in a BRAF wild-type melanoma xenograft model, and these data support the conclusion that interfering with PAK signaling could have therapeutic efficacy in this subset of melanoma. Discussion PAK family kinases have been implicated as central players in growth factor signaling networks (19,22). Hence, we initiated this study to investigate the contribution of PAK1 signaling to mela- noma. PAK1 genomic amplification and elevated cytoplasmic pro- tein expression were observed in 9% and 26% of primary human melanomas, respectively (Figure 1, A and B). Recently, recurrent activating mutation of PAK1 upstream activators—namely, Rac1 (37,38) and PREX2 (phosphatidylinositol-3,4,5-trisphosphate- dependent Rac exchange factor 2) (39)—have also been identi- fied in melanoma. These somatic mutations increase the activity and tumorigenic properties of Rac1 and PREX2 in melanocytes and promote Rac1 binding to PAK1 (38,39). Interestingly, we observed that PAK1 dysregulation was prevalent in tumors that lack activating, oncogenic mutations in BRAF (Table 1; Supplementary Table 1, available online). Loss-of-function stud- ies to analyze the role of PAK1 in BRAF wild-type melanoma cells revealed a requirement for PAK1 in CRAF phosphoryla- tion, MAPK pathway signaling, proliferation (Figures 2 and 3), and melanoma cell migration (Supplementary Figure 5, available online). Consistent with these cell culture experiments, admin- istration of PF-3758309 in vivo also resulted in reduced MAPK signaling (Figure 4C) and stasis in two xenograft models of BRAF wild-type melanoma (Figure 4, D and E). Together, these findings demonstrate that PAK is critically important for BRAF wild-type melanoma (model depicted in Figure 5), and additional predictors of response or resistance to PAK inhibition remain to be identi- fied for this indication. PAK1-dependent activation of MAPK signaling in tumor cells has been demonstrated for other indications (22,40,41), and anti- tumor efficacy by targeting the MAPK pathway has been validated in the clinic (9). However, attenuation of PAK1 signaling may also result in additional phenotypes relevant for tumor inhibition, such as decreased invasion and metastasis (18,19). Moreover, recent evi- dence in model systems points to mechanisms that promote sign- aling through CRAF by overexpression or mutational activation of upstream regulators to mediate resistance to BRAF inhibitors (42–44). It is therefore possible that elevated PAK1 signaling con- tributes to BRAF inhibitor resistance by switching the dependency for MAPK pathway activation to the CRAF isoform. In prelimi- nary support of this hypothesis, group 1 PAKs were identified in a functional screen to characterize mechanisms of de novo and Figure 5. Diagram depicting the mechanism of action for p21-activated kinase 1 (PAK1) in v-Raf murine sarcoma viral oncogene homolog B1 (BRAF) wild-type melanoma. A) In the context of oncogenic mutation, BRAF strongly drives activation of the mitogen-activated protein kinase (MAPK) signaling pathway, and these tumor cells are sensitive to inhi- bition of this kinase. B) In melanomas in which BRAF is not mutated, elevated expression and genomic amplification of PAK1 is frequent and results in increased signaling to v-Raf murine sarcoma viral oncogene (CRAF)– mitogen-activated protein kinase/extracellular signal-regulated kinase (MEK)– extracellular signal-regulated kinase (ERK) and poten- tially additional effector pathways.This subset of melanoma is relatively insensitive to BRAF inhibition and proliferative capacity is dependent on PAK1. SMI = small molecule inhibitor. acquired resistance to BRAF inhibition (36). Additionally, given the contribution of PAK1 to proliferation of both NRAS mutant tumor cells (Figure 2C) (45) and to proliferation of squamous cell carcinomas (23), it is plausible that PAK1 inhibition may allevi- ate the cutaneous squamous cell carcinomas and keratoacanthomas that are driven by RAS and CRAF activation that occur in patients treated with BRAF inhibitor as a single agent (11). This study also had some limitations. Our in vivo results were obtained from BRAF wild-type cell lines grown as xenograft tumors in nude mice and might not fully represent the heterogeneity observed for melanomas in the clinic. Additionally, we believe that there are therapeutic strategies for employing PAK inhibitors in melanomas with activating mutation of PAK1 upstream activators, Rac1, and PREX2; however, this study did not evaluate these mech- anisms, and additional work is needed to evaluate these and other biomarkers for sensitivity of melanoma cells to PAK inhibition. Currently, only ipilimumab treatment, which potentiates T-cell responses (45), is available for melanoma patients with BRAF wild- type tumors, and there is a need to identify pharmacologically trac- table driver genes associated with BRAF wild-type melanoma. The dependence of BRAF wild-type melanomas on PAK1 has not been reported previously, and our studies refine our understanding of this therapeutic target. Given that this melanoma subtype com- prises 40% to 50% of new cases in the United States, our findings have important implications for the development of new strategies and agents for the treatment of this disease. References 1. American Cancer Society. Cancer Facts & Figures 2011 http://www. cancer.org/research/cancerfactsfigures/cancerfactsfigures/cancer- facts-figures-2011. 2. Lopez-Bergami P. The role of mitogen- and stress-activated protein kinase pathways in melanoma. Pigment Cell Melanoma Res. 2011;24(5):902–921. 3. Lee JH, Choi JW, Kim YS. Frequencies of BRAF and NRAS mutations are different in histological types and sites of origin of cutaneous melanoma: a meta-analysis. Br J Dermatol. 2011;164(4):776–784. 4. Davies H, Bignell GR, Cox C, et al. Mutations of the BRAF gene in human cancer. Nature. 2002;417(6892):949–954. 5. Hoeflich KP, Gray DC, Eby MT, et al. Oncogenic BRAF is required for tumor growth and maintenance in melanoma models. Cancer Res. 2006;66 (2):999–1006. 6. Dankort D, Filenova E, Collado M, Serrano M, Jones K, McMahon M. A new mouse model to explore the initiation, progression, and therapy of BRAFV600E-induced lung tumors. Genes Dev. 2007;21(4):379–384. 7. Hoeflich KP, Herter S, Tien J, et al. Antitumor efficacy of the novel RAF inhibitor GDC-0879 is predicted by BRAFV600E mutational status and sustained extracellular signal-regulated kinase/mitogen-activated protein kinase pathway suppression. Cancer Res. 2009;69(7):3042–3051. 8. Tsai J, Lee JT, Wang W, et al. Discovery of a selective inhibitor of onco- genic B-Raf kinase with potent antimelanoma activity. Proc Natl Acad Sci U S A. 2008;105(8):3041–3046. 9. Bollag G, Hirth P, Tsai J, et al. Clinical efficacy of a RAF inhibi- tor needs broad target blockade in BRAF-mutant melanoma. Nature. 2011;467(7315):596–599. 10. Ribas A, Flaherty KT. BRAF targeted therapy changes the treatment para- digm in melanoma. Nat Rev Clin Oncol. 2011;8(7):426–433. 11. Chapman PB, Hauschild A, Robert C, et al. Improved survival with vemurafenib in melanoma with BRAF V600E mutation. N Engl J Med. 2011;364(26):2507–2516. 12. King AJ, Sun H, Diaz B, et al. The protein kinase Pak3 positively reg- ulates Raf-1 activity through phosphorylation of serine 338. Nature. 1998;396(6707):180–183. 13. Tang Y, Yu J, Field J. Signals from the Ras, Rac, and Rho GTPases con- verge on the Pak protein kinase in Rat-1 fibroblasts. Mol Cell Biol. 1999;19(3):1881–1891. 14. Frost JA, Steen H, Shapiro P, et al. Cross-cascade activation of ERKs and ternary complex factors by Rho family proteins. EMBO J. 1997;16(21):6426–6438. 15. Slack-Davis JK, Eblen ST, Zecevic M, et al. PAK1 phosphorylation of MEK1 regulates fibronectin-stimulated MAPK activation. J Cell Biol. 2003;162(2):281–291. 16. Zang M, Waelde CA, Xiang X, Rana A, Wen R, Luo Z. Microtubule integrity regulates Pak leading to Ras-independent activation of Raf-1: insights into mechanisms of Raf-1 activation. J Biol Chem. 2001;276(27):25157–25165. 17. Beeser A, Jaffer ZM, Hofmann C, Chernoff J. Role of group A p21-acti- vated kinases in activation of extracellular-regulated kinase by growth fac- tors. J Biol Chem. 2005;280(44):36609–36615. 18. Arias-Romero LE, Chernoff J. A tale of two Paks. Biol Cell. 2008;100(2):97–108. 19. Kumar R, Gururaj AE, Barnes CJ. p21-activated kinases in cancer. Nat Rev Cancer. 2006;6(6):459–471. 20. Ong CC, Jubb AM, Zhou W, et al. p21-activated kinase 1: PAK’ed with potential. Oncotarget. 2011;2(6):491–496. 21. Holm C, Rayala S, Jirstrom K, Stal O, Kumar R, Landberg G. Association between Pak1 expression and subcellular localization and tamoxifen resist- ance in breast cancer patients. J Natl Cancer Inst. 2006;98(10):671–680. 22. Arias-Romero LE, Villamar-Cruz O, Pacheco A, et al. A Rac-Pak signaling pathway is essential for ErbB2-mediated transformation of human breast epithelial cancer cells. Oncogene. 2010;29(43):5839–5849. 23. Ong CC, Jubb AM, Haverty PM, et al. Targeting p21-activated kinase 1 (PAK1) to induce apoptosis of tumor cells. Proc Natl Acad Sci U S A. 2011;108(17):7177–7182. 24. Vadlamudi RK, Adam L, Wang RA, et al. Regulatable expression of p21-activated kinase-1 promotes anchorage-independent growth and abnormal organization of mitotic spindles in human epithelial breast can- cer cells. J Biol Chem. 2000;275(46):36238–36244. 25. Yi C, Wilker EW, Yaffe MB, Stemmer-Rachamimov A, Kissil JL. Validation of the p21-activated kinases as targets for inhibition in neurofibromatosis type 2. Cancer Res. 2008;68(19):7932–7937. 26. Chow HY, Stepanova D, Koch J, Chernoff J. p21-Activated kinases are required for transformation in a cell-based model of neurofibromatosis type 2. PLoS One. 2010;5(11):e13791. 27. Murray BW, Guo C, Piraino J, et al. Small-molecule p21-activated kinase inhibitor PF-3758309 is a potent inhibitor of oncogenic sign- aling and tumor growth. Proc Natl Acad Sci U S A. 2010;107(20): 9446–9451.
28. Hatzivassiliou G, Song K, Yen I, et al. RAF inhibitors prime wild- type RAF to activate the MAPK pathway and enhance growth. Nature. 2010;464(7287):431–435.
29. Pinheiro J, Bornkamp B, Bretz F. Design and analysis of dose-finding stud- ies combining multiple comparisons and modeling procedures. J Biopharm Stat. 2006;16(5):639–656.
30. Reddy SD, Ohshiro K, Rayala SK, Kumar R. MicroRNA-7, a homeobox D10 target, inhibits p21-activated kinase 1 and regulates its functions. Cancer Res. 2008;68:(20)8195–8200.
31. de la Torre-Ubieta L, Gaudilliere B, Yang Y, et al. A FOXO-Pak1 tran-
scriptional pathway controls neuronal polarity. Genes Dev. 2010;24(8): 799–813.
32. Mason CS, Springer CJ, Cooper RG, Superti-Furga G, Marshall CJ, Marais R. Serine and tyrosine phosphorylations cooperate in Raf-1, but not B-Raf activation. EMBO J. 1999;18(8):2137–2148.
33. Zang M, Hayne C, Luo Z. Interaction between active Pak1 and Raf-1
is necessary for phosphorylation and activation of Raf-1. J Biol Chem. 2002;277(6):4395–4405.
34. Eblen ST, Slack JK, Weber MJ, Catling AD. Rac-PAK signaling stimulates extracellular signal-regulated kinase (ERK) activation by regulating formation of MEK1-ERK complexes. Mol Cell Biol. 2002;22(17):6023–6033.
35. Deacon SW, Beeser A, Fukui JA, et al. An isoform-selective, small-mol- ecule inhibitor targets the autoregulatory mechanism of p21-activated kinase. Chem Biol. 2008;15(4):322–331.
36. Johannessen CM, Boehm JS, Kim SY, et al. COT drives resistance to
RAF inhibition through MAP kinase pathway reactivation. Nature. 2011;468(7326):968–972.
37. Hodis E, Watson IR, Kryukov GV, et al. A landscape of driver mutations in melanoma. Cell. 2012;150(2):251–263.
38. Krauthammer M, Kong Y, Ha BH, et al. Exome sequencing identi-
fies recurrent somatic RAC1 mutations in melanoma. Nat Genet. 2012;44(9):1006–1014.
39. Berger MF, Hodis E, Heffernan TP, et al. Melanoma genome sequenc- ing reveals frequent PREX2 mutations. Nature. 2012;485(7399): 502–506.
40. McDaniel AS, Allen JD, Park SJ, et al. Pak1 regulates multiple c-Kit medi- ated Ras-MAPK gain-in-function phenotypes in Nf1+/- mast cells. Blood. 2008;112(12):4646–4654.
41. Rouquette-Jazdanian AK, Sommers CL, Kortum RL, Morrison DK, Samelson LE. LAT-independent Erk activation via Bam32-PLC- gamma1-Pak1 complexes: GTPase-independent Pak1 activation. Mol Cell. 2012;48(2):298–312.
42. Montagut C, Sharma SV, Shioda T, et al. Elevated CRAF as a potential mechanism of acquired resistance to BRAF inhibition in melanoma. Cancer Res. 2008;68(12):4853–4861.
43. Heidorn SJ, Milagre C, Whittaker S, et al. Kinase-dead BRAF and onco- genic RAS cooperate to drive tumor progression through CRAF. Cell. 2010;140(2):209–221.
44. Poulikakos PI, Rosen N. Mutant BRAF melanomas—dependence and resistance. Cancer Cell. 2011;19(1):11–15.
45. Singhal R, Kandel ES. The response to PAK1 inhibitor IPA3 distin-
guishes between cancer cells with mutations in BRAF and Ras oncogenes.
Oncotarget. 2012;3(7):700–708.

Funding
AMJ is supported by a career development fellowship from the Pathological Society of Great Britain and Ireland. ALH and MRM are supported by Cancer Research U.K. and the Oxford NIHR Biomedical Research Centre.

Notes
C.C. Ong and A.M. Jubb contributed equally to this work.
The authors are solely responsible for the study design, data collection, analy- sis and interpretation of the data, writing the manuscript, and decision to submit the manuscript for publication.
We thank Fred de Sauvage, Amy Young, Kyung Song, Helen Turley, Brian Buckwalter, Bruno Alicke, Marie-Claire Wagle and our immunohisto- chemistry facility for providing insightful discussions, suggestions and techni- cal assistance.

Affiliations of authors: Department of Translational Oncology (CC), DJ, WZ, JT, LSF, MB, EMB, KPH), Department of Pathology (AMJ, HK), Department of Discovery Chemistry (JR), Department of Bioinformatics (PMH), and Department of Oncology Biomarker Development (MK, YY), Genentech, Inc, San Francisco, CA; Weatherall Institute of Molecular Medicine, University of Oxford, Headington, Oxford, UK (AMJ, RL, ALH, MRM).