Oncogene (2012) 31, 3397–3408 & 2012 Macmillan Publishers Limited All rights reserved 0950-9232/12 www.nature.com/onc ORIGINAL ARTICLE PAK1 is a breast cancer oncogene that coordinately activates MAPK and MET signaling

Y Shrestha1,2, EJ Schafer1,2, JS Boehm2, SR Thomas1,2,FHe2,3,JDu2,3, S Wang4, J Barretina1,2,5, BA Weir1,2, JJ Zhao6, K Polyak1,5,7, TR Golub2,3,5, R Beroukhim1,2,6,7 and WC Hahn1,2,5,7

1Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, MA, USA; 2Broad Institute of Harvard and MIT, Cambridge, MA, USA; 3Department of Pediatric Oncology, Dana-Farber Cancer Institute, Boston, MA, USA; 4Cytogenetics Core Facility, Dana-Farber/Harvard Cancer Center, Boston, MA, USA; 5Center for Cancer Genome Discovery, Dana-Farber Cancer Institute, Boston, MA, USA; 6Department of Cancer Biology, Dana-Farber Cancer Institute, Boston, MA, USA and 7Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, MA, USA

Activating mutations in the RAS family or BRAF Lau and Haigis, 2009). Oncogenic mutations in the RAS frequently occur in many types of human cancers but family genes are common in selected cancer types, are rarely detected in breast tumors. However, activation including pancreatic, colon and non-small-cell lung of the RAS–RAF–MEK–ERK MAPK pathway is com- cancers. RAS activation may occur directly through monly observed in human breast cancers, suggesting that these oncogenic mutations or indirectly due to activa- other genetic alterations lead to activation of this tion of RAS regulators or effectors (reviewed in Down- signaling pathway. To identify breast cancer oncogenes ward, 2003). Activation of growth factor signaling is a that activate the MAPK pathway, we screened a library of predominant mechanism upstream of RAS that leads human kinases for their ability to induce anchorage- to its activation. In epithelial cancers, EGFR and independent growth in a derivative of immortalized human ERBB2—two ErbB family tyrosine kinase (TK) recep- mammary epithelial cells (HMLE). We identified p21- tors—commonly activate RAS oncogenic function activated kinase 1 (PAK1) as a kinase that permitted (Mendelsohn and Baselga, 2000). Similarly, loss of HMLE cells to form anchorage-independent colonies. function of a RAS-GAP, neurofibromatosis type 1 PAK1 is amplified in several human cancer types, (NF1), also drives RAS downstream signaling (Bollag including 30–33% of breast tumor samples and cancer et al., 1996). cell lines. The kinase activity of PAK1 is necessary for Several alternate mechanisms that activate RAS PAK1-induced transformation. Moreover, we show that signaling have also been previously described. The PAK1 simultaneously activates MAPK and MET signal- RAS effector pathway, PI3K, is activated by mutations ing; the latter via inhibition of merlin. Disruption of these of its catalytic subunit PIK3CA, amplification of its activities inhibits PAK1-driven anchorage-independent downstream target AKT or loss of function of PTEN ,a growth. These observations establish PAK1 amplification negative regulator (Scheid and Woodgett, 2001). Simi- as an alternative mechanism for MAPK activation in larly, activating mutation of BRAF occurs in 50% of human breast cancer and credential PAK1 as a breast melanomas, leading to constitutive activation of MAPK cancer oncogene that coordinately regulates multiple signaling (Davies et al., 2002). These two effector signaling pathways, the cooperation of which leads to pathways have important roles in RAS-mediated cell malignant transformation. transformation, as co-inhibition of PI3K and MAPK Oncogene (2012) 31, 3397–3408; doi:10.1038/onc.2011.515; efficiently suppresses RAS-driven tumor growth (Engel- published online 21 November 2011 man et al., 2008; Sos et al., 2009). Less than 5% of human breast tumors exhibit Keywords: PAK1; transformation; MAPK; MET; oncogenic mutations in the RAS family of genes (Miyakis merlin et al., 1998; Lau and Haigis, 2009). RAS signaling in breast cancer is more commonly activated by alterations upstream or downstream of RAS itself. For example, the RAS pathway is activated through ERBB2 amplification Introduction in about 20% of breast tumors (Hynes and MacDonald, 2009). Such growth factor activation in breast cancer The RAS family of small GTPases, HRAS, KRAS and results in a co-activation of RAS effectors PI3K and NRAS, are often mutated in human cancers, rendering MAPK (Neve et al., 2002). Co-inhibition of these them constitutively active and oncogenic (reviewed in synergistic pathways has proven more effective than single-pathway inhibition for suppression of breast tumor- Correspondence: Dr WC Hahn, Dana-Farber Cancer Institute, 450 igenesis (Hoeflich et al., 2009; Mirzoeva et al., 2009). Brookline Avenue, Dana 1538, Boston, MA 02215, USA. E-mail: [email protected] Moreover, co-activation of these RAS effectors is common Received 3 April 2011; revised 17 August 2011; accepted 10 September in breast cancer. Activating mutations in PIK3CA are 2011; published online 21 November 2011 found in 25–30% of breast tumors, whereas loss of PAK1 drives breast transformation via MAPK and MET Y Shrestha et al 3398 function of PTEN by mutation or loss of protein activates cancer cell lines, where PAK1 enhanced anchorage- the PI3K pathway in 5–30% of human breast tumors independent colony formation (Vadlamudi et al., 2000). (Freihoff et al., 1999; Bachman et al., 2004; Hennessy Our findings in HMLEA cells indicate that PAK1 also et al., 2005; Miron et al., 2010). In contrast, oncogenic cooperates with other oncogenic alterations to induce mutations in the MAPK pathway components have not cell transformation. been reported in breast tumors. Therefore, alternative mechanisms that activate MAPK signaling to promote PAK1 amplifications in human breast cancer breast oncogenesis need to be defined. The observation that PAK1 induces cell transformation To identify genes that activate or substitute for suggested that PAK1 could function as an oncogene. To MAPK activation in breast cancer, we performed a assess the frequency of PAK1 amplification in human kinase-focused screen in a human mammary epithelial cancer, we analyzed a large set of human cancer samples cell transformation model (HMLE) driven by oncogenic in Tumorscape, a dataset that includes whole-genome RAS. Expression of oncogenic HRAS (HRASV12)in analyses of somatic copy number alterations in a HMLE cells, which are immortalized with the catalytic collection of 3131 human samples, including 243 breast subunit of telomerase and SV40 early region, promotes tumor samples and cancer cell lines (Beroukhim et al., anchorage-independent growth and tumorigenesis 2010). We found that PAK1 is amplified in 33% of (Elenbaas et al., 2001). In this transformation model, breast samples (Figure 2a), as well as in a smaller simultaneous expression of myristoylated (myr)-AKT1 fraction of non-small-cell lung, ovarian, small-cell lung, and an activated allele of MEK1 (MEK1S218D/S222D, melanoma and esophageal squamous cancers. These MEKDD) can substitute for oncogenic HRAS. Using findings confirm and extend smaller scale studies that HMLE cells expressing MEKDD, we previously found focused on chromosome 11q13, mainly, but not IKBKE as a kinase oncogene amplified in 30% of exclusively, in ovarian cancer (Bekri et al., 1997; human breast cancers, and CSNK1E as a transforming Schraml et al., 2003; Lambros et al., 2005; Kim et al., gene that also functions as a synthetic lethal partner 2006; Bostner et al., 2007; Brown et al., 2008; with activated b-catenin (Boehm et al., 2007; Kim et al., Chattopadhyay et al., 2010; Tu et al., 2011). Overall, 2010). To identify genes that activate MAPK signaling we found that 16% of all cancers in the Tumorscape and cooperate with active PI3K pathway, we screened exhibited amplification of the locus that contains PAK1. for oncogenes in HMLE cells expressing myr-AKT1 To confirm that PAK1 amplifications occur in breast (HMLEA). We identified p21-activated kinase 1 (PAK1) cancer, we performed fluorescence in situ hybridization as an amplified kinase capable of inducing potent (FISH) and found normal PAK1 copy number in breast mammary cell transformation through the coordinate cancer cell lines that did not show evidence of PAK1 regulation of MAPK and MET signaling. amplification in Tumorscape (EFM19 and SKBR3). In contrast, when we examined metaphase spreads derived from breast cancer cell lines that exhibited amplification Results of PAK1 in Tumorscape (SUM52 and SUM190), we found clear evidence of PAK1 copy number gain Kinome focused gain-of-function screen for kinases that (Figure 2b). The average number of PAK1 signals per promote anchorage-independent growth of immortalized nucleus was greater than the chromosome 11 centromere HMLE cells reference signals in SUM52 and SUM190 lines To identify oncogenes that activate the MAPK pathway (Figure 2c). The PAK1 FISH signal in the four cell lines in mammary epithelial cells, we expressed an open correlated with the corresponding PAK1 copy number in reading frame (ORF) library composed of kinases in Tumorscape. Furthermore, PAK1 copy number directly HMLEA cells. Specifically, we pooled 597 myristoyla- correlated with its expression both at the RNA and tion- and Flag-epitope-tagged kinases, introduced them protein levels in the four cell lines (Figure 2d). into HMLEA cells and measured anchorage-indepen- Using the genomic identification of significant targets dent growth (Figure 1a). Macroscopic colonies formed in cancer (GISTIC) algorithm (Beroukhim et al., 2007) by HMLEA cells expressing kinase pools or controls in Tumorscape samples, we found that PAK1 lies on an were counted at 2 weeks. We found that pool no. 9 amplification peak on chromosome 11q distinct from a formed numerous colonies (Figure 1b). When the 27 neighboring peak containing CCND1 (Figure 2e). kinases in pool no. 9 were individually tested for CCND1 is amplified in 42% of all samples in anchorage-independent growth of HMLEA cells, we Tumorscape, whereas 43% of those that exhibit PAK1 found that PAK1 induced robust colony formation amplification also carry CCND1 amplification. These (Figure 1c). We also observed that myristoylation of observations strongly suggest that amplifications of PAK1 was unnecessary to induce anchorage-indepen- PAK1 and CCND1 are independent events. dent growth (Supplementary Figure 1A), and we We also analyzed The Cancer Genome Atlas (TCGA) used non-myristoylated PAK1 for subsequent validation for PAK1 copy number gain in human cancer samples. and characterization studies. Moreover, expression of We found that 30% of breast invasive adenocarcinoma PAK1 sufficed to induce anchorage-independent growth samples showed significant PAK1 amplification (Q- of HMLE cells (Figure 1d). Thus, our observations value 8.04E-39). Together, our observations establish establish that PAK1 can transform immortalized that PAK1 is amplified in a significant fraction of human HMLE cells, unlike prior studies carried out in breast breast tumors and tumor-derived cell lines.

Oncogene PAK1 drives breast transformation via MAPK and MET Y Shrestha et al 3399 597 pBP-MF-Kinase ORFs

1-25, 26-52...... 597 10.0 Kinase Pools 9.0 Controls 8.0 Scoring Pools Pool #1 Pool #2.....Pool #22 7.0 + 6.0 #9 HMLEA 5.0

4.0 1:25

1.5 SD D 3.0 D

2.0 MEK Median Vector Macroscopic colony number 1.0 0.0 Pools 1 to 22

250 HMLE ** 60 PAK1 200 50 Individual Kinases Controls Scoring Kinases 40 150

30 100

Colony number 20 1 SD 50 DD 10

Macroscopic colony number

Median Vector MEK 0 0 Vector PAK1 Pool #9 - Kinases 1 to 27 Figure 1 Human kinase expression screen in HMLEA cells. (a) An ORF library containing 597 Myr-Flag-epitope-tagged kinase ORFs were organized into 22 pools and screened for their ability to promote anchorage-independent growth of HMLEA cells. (b) The average number of macroscopic colonies formed by HMLEA cells expressing the pools or controls at 2 weeks. Vector, pBP; MEKDD 1:25, positive control. (c) Anchorage-independent growth of HMLEA cells expressing controls or each of the 27 kinases in pool no. 9. (d) Anchorage-independent growth of HMLE cells expressing vector control or PAK1. **P-value o0.001.

We then investigated whether cell lines that harbor Prior work has shown that the kinase activity of PAK1 amplifications were dependent on PAK1 for PAK1 can be assessed by the status of phosphorylation proliferation or anchorage-independent colony formation. at several internal sites. PAK1 phosphorylation at T423 We failed to find changes in population-doubling times of relieves autoinhibitory conformation, allowing autopho- SUM52 cells expressing PAK1- or lacZ-specific short sphorylation at S144 to maintain kinase activity hairpin RNAs (shRNAs) or in HMLE cells expressing (reviewed in Bokoch, 2003). We evaluated the effect of control vector or PAK1, indicating that PAK1 expression overexpressing PAK1 on its kinase activity by assessing level does not affect proliferation (Figure 3a). In contrast, the phosphorylation status of PAK1 itself as well as its suppression of PAK1 by two different PAK1-specific reported phosphorylation targets. When we expressed a shRNAs in SUM52 and SUM190 lines compromised control vector, PAK1 or PAK1K299R in HMLE cells, we anchorage-independent growth (Figure 3b). Importantly, found that wild-type PAK1 but not PAK1K299R was in cell lines (SKBR3 and EFM19) that do not harbor phosphorylated at both T423 and S144 sites (Figure 4b). PAK1 amplification, suppression of PAK1 failed to affect However, comparison of total and phosphorylated anchorage-independent growth (Figure 3c). These obser- PAK1 levels indicates that a fraction of PAK1 in vations demonstrate that cell lines that exhibit amplifica- HMLE–PAK1 cells may remain in an autoinhibitory tion and overexpression of PAK1 depend on it for conformation. Nevertheless, we found that the phos- anchorage-independent growth, strongly indicating that phorylation levels of three PAK1-specific substrates; PAK1 is a targeted gene in the more-telomeric 11q RAF1S338, MEK1S298 and merlinS518, were enhanced in GISTIC amplification peak in Tumorscape. HMLE–PAK1 cells (Figure 4c; Coles and Shaw, 2002; Xiao et al., 2002; Zang et al., 2002; Zang et al., 2008). PAK1 kinase activity in anchorage-independent growth The elevated phosphorylation levels of PAK1 itself as PAK1 regulates signaling pathways through both well as its substrates exclusively in HMLE–PAK1 cells kinase-dependent and kinase-independent mechanisms confirm that a greater portion of the wild-type PAK1 is (reviewed in Dummler et al., 2009). We evaluated the in its active conformation, and hence catalytically active requirement for PAK1 kinase activity in cell transfor- compared to PAK1K299R. These findings establish that mation by expressing wild-type PAK1 and its previously the kinase activity of PAK1 is necessary for the characterized kinase-deficient mutant, PAK1K299R (Tang transformation of mammary epithelial cells. Moreover, et al., 1997). We found that wild-type PAK1, but not PAK1-induced phosphorylation of RAF1, MEK1 and PAK1K299R, formed robust anchorage-independent co- merlin suggests that these pathways may contribute to lonies in HMLE cells (Figure 4a). PAK1-driven transformation.

Oncogene PAK1 drives breast transformation via MAPK and MET Y Shrestha et al

3400

0 0

20 20 40 40 60 CCND1 60 CCND1 80 PAK1 PAK1 80 PAK1 100 100 Megabases along chromosome 11q 120 100 10-5 10-10 10-15 10-20 10-25 Megabases along chromosome 11q 120 Significance (as q-value)

Telomere Breast NSC Lung Ov SCL MelEsq Telomere Deleted Neutral Amplified

Average signal/nucleus (100 nuclei per cell line) Copy Number (CEP11) (PAK1) (Tumorscape) SUM52 6.37 3 >10 SUM190 4.41 2 6 SKBR3 2.07 4 4 EFM19 1.54 55

SUM52 SUM190 6 5 4 3 2

1 Normalized PAK1 mRNA PAK1 0 HMLE EFM19 SKBR3 SUM52 SUM190 PAK1

EFM19 SKBR3 GAPDH

Figure 2 PAK1 is amplified in human cancers. (a) Copy number profile along chromosome 11q of human tumor samples and cancer cell lines in Tumorscape that exhibit highest level of PAK1 amplification divided according to cancer type: breast, non-small cell (NSC) lung, ovarian (Ov), small cell lung (SCL), melanoma (Mel) and esophageal squamous (Esq). PAK1 and CCND1 loci are marked. (b) Fluorescence in situ hybridization of breast cancer cell lines with (SUM52 and SUM190) and without (EFM19 and SKBR3) PAK1 amplification. Orange fluorescence indicates chromosome 11 centromere probe. Green fluorescence indicates PAK1 probe. White bar ¼ 100 m.(c) Tumorscape copy number data are tabulated along with corresponding average number of PAK1 and reference signals from 100 nuclei per cell line. (d) mRNA and protein levels of PAK1 in the four breast cancer cell lines and HMLE cells. (e) GISTIC analysis of 3131 Tumorscape human cancers along chromosome 11q showing two distinct amplification peaks containing PAK1 or CCND1. The arrow indicates a higher magnification representation of the dotted area to show PAK1 locus (red).

MAPK signaling and PAK1-driven anchorage- dramatically fewer colonies, whereas EFM19 cells, independent growth which exhibit no PAK1 copy number gain, were largely Phosphorylation of RAF1S338 and MEK1S298 prime the unaffected (Figure 5c). We also found that U0126 activation of the MAPK pathway by facilitating phos- inhibited ERK1/2T202/Y204 phosphorylation as well as the phorylation of MEK1 at its activation loop, and ability of HMLE cells expressing PAK1 or HRASV12 subsequent phosphorylation of ERK1/2. PAK1 expres- (HMLER) to form colonies in a dose-dependent manner sion in HMLE cells not only increased phosphorylation of (Figure 5c). Together, these genetic and pharmacologi- RAF1S338 and MEK1S298, two critical upstream members cal studies indicate that the activation of RAF–MEK– of the MAPK pathway, but also increased phosphoryla- ERK MAPK signaling is essential for PAK1-induced tion of the downstream effectors, ERK1/2T202/Y204 (Figures anchorage-independent growth. 4c and 5a). Conversely, suppression of PAK1 in Prior work has implicated PAK1 in the regulation SUM52 repressed phosphorylation of RAF1S338 and of RAS signaling. Even in the presence of oncogenic ERK1/2T202/Y204, thus indicating regulation of the MAPK RAS, loss of PAK1 function or inhibition of PAK1 pathway by PAK1 (Supplementary Figure 2A). activators inhibited colony formation (Tang et al., 1997; In consonance with this observation, suppressing Appledorn et al., 2010). Similarly, we found that PAK1 RAF1 in two PAK1-dependent cell lines, HMLE– suppression by shRNAs partially inhibited anchorage- PAK1 and SUM52, suppressed phospho-ERK1/2 levels independent growth of HMLER cells, reinforcing the and compromised anchorage-independent growth (Sup- role of PAK1 in RAS-driven transformation (Figure plementary Figure 2B; Figure 5b). Furthermore, when 5d). We, however, did not observe any difference in total we treated SUM52 and EFM19 with U0126, a MEK1/2- or phosphorylated levels of PAK1 between HMLE and specific inhibitor (Favata et al., 1998), SUM52 formed HMLER cells (Supplementary Figure 3A). Hence, we

Oncogene PAK1 drives breast transformation via MAPK and MET Y Shrestha et al 3401 16 SUM52 25 HMLE 14 20 12

10 15 8 6 10 4 shPAK1 #1 Vector Population Doublings Population Doublings shPAK1 #2 5 2 shLACZ PAK1 0 0 0 3 5 7 10 12 14 17 19 21 0 2479111416 18 21 23 Days Days

300 SUM52 3500 EFM19 250 3000 2500 200 2000 150

shPAK1 #1 shPAK1 #2 shLACZ 1500 shPAK1 #1 shPAK1 #2 shLACZ 100 PAK1 1000 PAK1

Colony number Colony number 50 *** *** GAPDH 500 GAPDH 0 0 shPAK1 #1shPAK1 #2 shLACZ shPAK1 #1shPAK1 #2 shLACZ

1600 SUM190300 SKBR3 1400 250 1200 1000 200 800 150 shPAK1 #1 shPAK1 #2 shLACZ shPAK1 #1 shPAK1 #2 shLACZ 600 100 *** PAK1 PAK1 Colony number 400 *** Colony number 50 200 GAPDH GAPDH 0 0 shPAK1 #1shPAK1 #2 shLACZ shPAK1 #1shPAK1 #2 shLACZ Figure 3 PAK1-amplified cell lines are dependent on PAK1 for transformation. (a) Proliferation of SUM52 cells expressing PAK1- or lacZ-specific shRNAs or HMLE cells expressing vector control or PAK1 calculated over 21 or 23 days. Effects of suppressing PAK1 on anchorage-independent growth of breast cancer cell lines with (b: SUM52 and SUM190) or without (c: EFM19 and SKBR3) PAK1 amplification. Error bars ¼ 1 s.d. Insets indicate PAK1 protein levels. ***P-value o0.0001. A full colour version of this figure is available at the Oncogene journal online. conclude that PAK1 gain of function activates MAPK inhibited anchorage-independent growth but also sup- pathway, a RAS effector, which is necessary for PAK1- pressed merlinS518 phosphorylation (Figure 3b and 6a). driven anchorage-independent growth. Moreover, am- Furthermore, shRNA-mediated suppression of NF2 plification resulting in overexpression of PAK1 is an induced anchorage-independent growth of HMLE cells, alternate mechanism by which MAPK signaling may be indicating a tumor suppressive function of merlin in activated in human breast tumors that do not harbor breast cancer (Figure 6b). These observations demon- activating mutations in components of the RAS-RAF- strate that PAK1 regulates merlin at an inhibitory MEK-ERK MAPK pathway. phosphorylation site, and that suppression of merlin promotes transformation of immortalized mammary Merlin function in MET signaling and PAK1-mediated epithelial cells. anchorage-independent growth Merlin has recently been shown to negatively regulate Merlin or NF2 is a tumor suppressor gene deleted in a EGFR internalization and signaling through membrane subset of patients with neurofibromatosis and is a sequestration (Curto et al., 2007). In particular, merlin known substrate of PAK1 (reviewed in McClatchey, activity at the membrane of confluent mouse embryonic 2007). Merlin and PAK1 are involved in an inhibitory fibroblasts resulted in a substantial downregulation of loop where merlin binds to the p21-binding domain of overall phosphotyrosine levels, whereas NF2À/À mouse PAK1, inhibiting its activation and suppressing RAS- embryonic fibroblasts exhibited high phosphotyrosine dependent transformation (Kissil et al., 2003; Hirokawa levels, indicating that merlin regulates phosphotyrosine et al., 2004). PAK1, on the other hand, phosphorylates signaling (Curto et al., 2007). To investigate the role of merlin at S518, which induces translocation of merlin merlin in membrane receptor activity, we compared away from the membrane and inhibits its function phosphotyrosine levels in HMLE cells expressing wild- (Shaw et al., 2001; Kissil et al., 2002; Xiao et al., 2002). type PAK1 or the kinase-inactive mutant PAK1K299R. To investigate whether PAK1 regulation of merlin We found that wild-type but not the kinase-inactive contributes to mammary cell transformation, we exam- mutant of PAK1 induced an increase in total phospho- ined the consequences of altering PAK1 or NF2 tyrosine levels (Figure 6c). On the basis of this expression in HMLE cells and breast cancer cell lines. observation, we investigated EGFR phosphorylation PAK1 expression in HMLE cells, under conditions that status in HMLE–PAK1 cells compared with HMLE led to anchorage-independent growth, resulted in cells expressing either a control vector or PAK1K299R and increased phosphorylation of merlinS518 (Figure 4a and c). found only a small increase, if any, in phosphorylation Conversely, suppression of PAK1 in SUM52 not only of EGFR at sites that indicate its activation status,

Oncogene PAK1 drives breast transformation via MAPK and MET Y Shrestha et al 3402 HMLE associated proteins exhibited at least a significant 1.5- 250 ** fold increase in phosphotyrosine levels in HMLE–PAK1 cells when compared to HMLE–PAK1K299R cells 200 (Figure 6e; Du et al., 2009). Confirming our prior findings, ERK2, but not EGFR, scored in this assay. 150 Interestingly, we found that three different MET-specific antibodies showed that MET phosphotyrosine levels 100 were significantly elevated in HMLE cells expressing K299R

Colony number PAK1 compared to PAK1 . Together, these obser- 50 vations strongly suggested that PAK1-mediated inhibi- tion of merlin function led to activation of MET 0 signaling rather than EGFR activation. Vector PAK1 PAK1K299R In HMLE–PAK1 cells, we found that MET was phosphorylated at Y1234/1235 and Y1003—sites asso- ciated with MET activation and trafficking (Figure 7a; reviewed in Trusolino et al., 2010). We also found that a K299R K299R MET adaptor protein, GAB1 (Weidner et al., 1996), and

Vector PAK1 PAK1 Vector PAK1 PAK1 a MET downstream effector molecule, STAT3 (Boccac- cio et al., 1998), were phosphorylated in HMLE–PAK1, pPAK1T423 S338 pRAF1 providing strong evidence that MET signaling was activated by PAK1 overexpression (Figure 7a). More- S144 pPAK1 RAF1 over, when NF2 was overexpressed in HMLE–PAK1 cells, phosphorylation of PAK1T423 decreased slightly, PAK1 pMEK1S298 whereas that of METY1234/1235 was eliminated (Figure 7b). Conversely, when NF2 was suppressed in HMLE cells, GAPDH MEK1 phosphorylation of MET was enhanced (Supplementary Figure 5A). pMerlinS518 To determine the contribution of MET signaling to PAK1-induced anchorage-independent growth, we trea- Merlin ted HMLE–PAK1 and SUM52 with the MET inhibitor PHA-665752, which inhibits MET signaling through GAB1 (Christensen et al., 2003). We found that PHA- GAPDH 665752 inhibited anchorage-independent growth of both Figure 4 PAK1 kinase activity is required for anchorage- HMLE–PAK1 and SUM52 in a dose-dependent manner independent growth. (a) Anchorage-independent growth of HMLE (Figure 7c; Supplementary Figure 6A). cells expressing control vector, wild-type or kinase-dead PAK1 When we introduced MET- or GAB1-specific (PAK1K299R). Error bars ¼ 1 s.d. **P-value o0.003 (compared with shRNAs in HMLE–PAK1 and SUM52, we found that either vector or kinase-dead control). (b) Expression level of PAK1 and its phosphorylation status at T423 and S144 in HMLE cells suppression of MET or GAB1 compromised anchorage- expressing vector control, PAK1 or PAK1K299R.(c) Phosphoryla- independent growth of both cell lines that are dependent tion levels of three PAK1 substrates, RAF1S338, MEK1S298 and on PAK1 for transformation (Figure 7d; Supplementary merlinS518. Corresponding total protein levels are shown. Figures 6B and 6C). These observations demonstrate that PAK1 regulates MET activity by modulating merlin, and that MET signaling through GAB1 is required for Y1173 and Y1068 (Figure 6d; Batzer et al., 1994; PAK1-driven anchorage-independent growth. Yamauchi et al., 1997). Additionally, suppression of Inhibition or activation of MET signaling in HMLE EGFR in HMLE–PAK1 cells did not alter colony derivatives by pharmacological or genetic methods did formation, indicating EGFR may not be the appro- not alter phospho-ERK1/2 levels (Supplementary Fig- priate target of merlin inhibition in this context ures 5A and 5B), indicating that MAPK activation and (Supplementary Figure 4). MET regulation through merlin are independent events Prior work has indicated that NHERF1, the adaptor regulated by PAK1 (Figure 8). protein that interacts with merlin for EGFR inhibition, is not specific to EGFR and may associate with other growth factor receptors (reviewed in Weinman et al., 2006). We thus postulated that receptor TKs other than Discussion EGFR may be affected by merlin inhibition due to PAK1 amplification and overexpression. To identify MAPK pathway activation is common in breast cancers receptor TKs regulated by PAK1, we performed a (reviewed in Santen et al.,2002).Insomecases,thisactiv- multiplex antibody-based assay that measures activated ation is driven by amplification of ERBB2.However,a TKs and associated proteins in HMLE–PAK1 and substantial proportion of breast cancers with normal HMLE–PAK1K299R cells regardless of their mode of ERBB2 copy number also show activation of MAPK activation (Du et al., 2009). We found that six TKs or signaling. Here, we used an experimental model of

Oncogene PAK1 drives breast transformation via MAPK and MET Y Shrestha et al 3403

HMLE

HMLER

shLACZ shRAF1 #1 shRAF1 shRAF1 #2 shRAF1 K299R 200

180 PAK1

PAK1 PAK1 Vector 160 GAPDH 140 pERK1/2T202/Y204 120 *** 100 *** ERK1/2 80 60 Colony number 40 GAPDH 20 0 shRAF1 #1 shRAF1 #2 shLACZ

HMLE-PAK1 SUM52 500 2500

450

shLACZ shRAF1 #1 shRAF1 2000 400 #2 shRAF1 350 RAF1

1500 300 GAPDH

shLACZ shRAF1 #1 shRAF1 shRAF1 #2 shRAF1 250 1000 RAF1 200 *** *** GAPDH 150

Colony number Colony number 500 100 *** *** 50 0 0 shRAF1 #1 shRAF1 #2 shLACZ shRAF1 #1 shRAF1 #2 shLACZ

1200 HMLE-PAK12500 SUM52 1000 2000 800 U0126 U0126 DMSO 1uM 10uM 25uM DMSO 10uM 1uM 25uM 1500 600 pERK1/2 ** pERK1/2 1000 400 ERK1/2 ***

Colony number 200 Colony number 500 *** ** *** 0 0 DMSO 1uM 10uM 25uM DMSO 10uM 25uM U0126 U0126 U0126 U0126 U0126

1400 HMLER 8000 EFM19 1200 7000 U0126 1000 6000 DMSO 1uM 10uM 25uM U0126 pERK1/2 800 DMSO 1uM 10uM 25uM 5000 pERK1/2 4000 600 3000 400

Colony number

Colony number 2000 200 1000 0 0 DMSO 1uM 10uM 25uM DMSO 10uM 25uM U0126 U0126 U0126 U0126 U0126 Figure 5 MAPK activation by PAK1 expression is required for anchorage-independent growth. (a) Phosphorylation status of the MAPK pathway downstream effectors ERK1/2T202/Y204 in HMLE cells expressing a control vector, PAK1 or PAK1K299R. Total ERK1/2 levels are also shown. (b) Anchorage-independent colony formation by HMLE–PAK1 and SUM52 expressing RAF1- or lacZ-specific shRNAs. Corresponding RAF1 protein levels are shown. (c) Anchorage-independent growth of HMLE–PAK1, HMLER, SUM52 and EFM19 treated with U0126 at the indicated doses. Representative phospho-ERK1/2T202/Y204 levels are shown. (d) Anchorage- independent colony formation by HMLER cells expressing PAK1- or lacZ-specific shRNAs. PAK1 protein levels are shown. Error bars ¼ 1 s.d. ***P-value o0.0001 and **P-value o0.001. mammary epithelial cell transformation to search for a significant proportion of other epithelial cancers. We kinases that transform these cells in a MAPK-dependent further showed that breast cancer cell lines harboring manner. PAK1 amplification are dependent on its expression for We screened 597 kinase-related ORFs and found transformation but not proliferation. These observa- PAK1 as a kinase whose expression induced anchorage- tions substantiate PAK1 as the target gene in one of the independent colony formation in HMLE cells. On the two 11q13-14 amplicons and further define PAK1 as a basis of these observations, we concluded that PAK1 is bona fide breast cancer oncogene. a transforming kinase, which may also cooperate with Through similar transformation and copy number other oncogenic alterations, such as PI3K activation, to analyses of group I PAKs, we found that PAK1 and enhance transformation. PAK3, but not PAK2, promoted colony formation of In parallel, although prior studies have reported HMLE cells. However, we failed to find significant amplifications of 11q13-q14, we used two largest amplification of PAK3 in Tumorscape (Supplementary collections of copy number data in human tumors Figure 1B, data not shown). Hence, among its closest currently available (Tumorscape and TCGA) to confirm family members, PAK1 shows strongest oncogenic that PAK1 is amplified in 30–33% of breast cancer and characteristics.

Oncogene PAK1 drives breast transformation via MAPK and MET Y Shrestha et al 3404 HMLE HMLE

SUM52 K299R 300 ***

shNF2 #1 shNF2 #2 shLACZ

PAK1 Vector 250 PAK1 Merlin Y1173 200 ** pEGFR shPAK1 #2 shLACZ shPAK1 #1 GAPDH 150 pMerlinS518 pEGFRY1068 100

Merlin Colony number 50 EGFR 0 shNF2 #1 shNF2 #2 shLACZ GAPDH

HMLE

9R 8.0

K29 or

7.0 PTK6

PAK1

Vect PAK1

150 K299R 6.0

100 5.0

75 4.0 AXL MET 3.0 MET 50

Fold signal change HER2 MET 2.0 ERK2 GRB2 HMLE-PAK1 / 1.0

0.0 0 500 1000 1500 2000 2500 3000 3500 4000 phosphotyrosine Normalized HMLE-PAK1 signal Figure 6 PAK1 expression inhibits merlin function, subsequently upregulating phosphotyrosine levels. (a) Phosphorylation status of merlinS518 in SUM52 cells expressing PAK1- or lacZ-specific shRNAs. (b) Anchorage-independent colonies formed by HMLE cells expressing NF2- or lacZ-specific shRNAs. Merlin protein levels are shown. ***P-value ¼ 0.0002 and **P-value ¼ 0.0059. Immunoblots showing overall phosphotyrosine levels (c) or phosphotyrosine levels of EGFR at Y1173 and Y1068 (d) in HMLE cells expressing a control vector, PAK1 or PAK1K299R.(e) Luminex assay showing normalized phosphotyrosine signals of TKs and associated proteins in HMLE–PAK1 (x axis) and the fold phosphotyrosine signal increase in HMLE–PAK1 when compared to HMLE–PAK1K299R (y axis). Each diamond (gray and black) represents individually distinct antibodies. Analytes considered positive (410) and showing significant (1.5) fold increase in signal are labeled and shown as black diamonds.

PAK1 is known to regulate several signaling path- observations also suggest that amplification of PAK1 is ways. We found that PAK1-induced activation of both an alternative mechanism of activating MAPK signaling MAPK and MET signaling are essential for PAK1- in human breast cancers. driven cell transformation. Although PAK1 also Prior work showed that the tumor suppressor regulates LIMK1/2, the suppression of LIMK1/2 failed merlin regulates EGFR signaling. However, we to affect anchorage-independent growth (data not failed to find any evidence using both focused and shown). Furthermore, PAK1, as an effector of the small unbiased assays that PAK1-mediated regulation of GTPases, RAC1 and CDC42, may regulate invasion or merlin induced alterations in EGFR function. Instead, metastasis in some cellular contexts. Our observations, we found that PAK1 expression leads to activation of however, implicate PAK1 as an oncogene that induces MET signaling, which was required for anchorage- transformation in breast and possibly other cancers. independent growth. These findings established the In addition, we found that two signaling pathways— contribution of MET signaling to PAK1-dependent MAPK and merlin–MET—were coordinately regulated transformation and demonstrated that merlin regulates by PAK1 to drive anchorage-independent growth. multiple receptor TK signaling cascades in a similar Using both genetic and pharmacological manipulations, manner. we showed that anchorage-independent growth induced The identification of PAK1 as a breast cancer by PAK1 requires independent activation of MAPK oncogene provides one mechanism by which the as well as MET signaling. We have also observed MAPK pathway is activated in breast cancers that that activation of the MAPK pathway by PAK1 is do not harbor amplifications of ERRB2 or oncogenic not specific to HMLE or SUM52 cells but also occurs in mutations in the RAS–RAF–MEK–ERK pathway. human mammary epithelial cells only expressing Although ERBB2 expression has been shown to hTERT (HME), in which case the expression of correlate with PAK1 activation, (Arias-Romero et al., PAK1 also enhances phosphorylation of ERK1/2 and 2010), indicating that they function in the same transforms them (Supplementary Figure 7). As we pathway, co-amplification of ERBB2 and PAK1 may and others have found that PAK1 expression is not entirely be redundant. ERBB2 amplification can necessary for RAS-induced cell transformation, our lead to activation of other RAS effector pathways in

Oncogene PAK1 drives breast transformation via MAPK and MET Y Shrestha et al 3405 HMLE 800 HMLE-PAK1 700 K299R HMLE-PAK1 600 500 Vector PAK1 PAK1 400 pMETY1234/1235

Vector NF2 300 200 ** Merlin pMETY1003 100 ** 0 T423 DMSO 1uM 2.5uM 5uM MET pPAK1 PHA-665752

PAK1 pGAB1Y307 350 300 Y1234/1235 SUM52 GAB1 pMET 250 200 MET pSTAT3Y705 150 100 GAPDH Colony number Colony number STAT3 50 * * 0 DMSO 1uM 2.5uM 5uM GAPDH PHA-665752

HMLE-PAK1 HMLE-PAK1 1800 2500 1600 1400 2000 1200 ** 1000 1500 800 *** 1000 *** 600 Colony number Colony number 400 500 200 *** 0 0 shMET #1 shMET #2 shLACZ shGAB1 #1 shGAB1 #2 shLACZ

450 SUM52 450 SUM52 400 400 350 350 300 300 250 *** 250 200 200 150 150 ***

Colony number *** 100 Colony number 100 50 50 *** 0 0 shMET #1 shMET #2 shLACZ shGAB1 #1 shGAB1 #2 shLACZ Figure 7 MET signaling activation due to PAK1 expression promotes anchorage-independent growth. (a) Phosphotyrosine levels of MET at Y1234/1235 and Y1003, its adaptor protein GAB1Y307 and downstream effector STAT3Y705 in HMLE cells expressing a control vector, PAK1 or PAK1K299R. Corresponding total protein levels are shown. (b) Immunoblots of HMLE–PAK1 cells expressing vector control or NF2 showing levels of merlin, phosphorylated MET at Y1234/1235 and phosphorylated PAK1 at T423. Total protein levels are shown. (c) Anchorage-independent growth of HMLE–PAK1 and SUM52 after treatment with PHA-665752, a MET inhibitor, at indicated doses. (d) Anchorage-independent colonies formed by HMLE–PAK1 or SUM52 expressing MET-, GAB1- or lacZ-specific shRNAs. ***P-value o0.0001, **P-value o0.002 and *P-value o0.01. addition to MAPK, whereas PAK1 amplification also functions as a driving oncogene, a cooperating onco- activates MET signaling pathway through merlin gene, as well as an essential non-oncogene that mediates inhibition. transformation of mammary cells through regulation of Our studies using previously characterized kinase- multiple signaling pathways. dead PAK1-mutant allele (K299R) showed that PAK1 kinase activity is required for transformation, and establish PAK1 as a target for small-molecule inhibition Materials and methods in cancers that harbor PAK1 amplification. Our observations further indicate that breast cancer cell Cell lines and reagents lines, such as SUM52 and SUM190, that harbor PAK1 HMLE derivatives were cultured in Clonetics MEGM (Lonza, amplification and exhibit correspondingly elevated Basel, Switzerland). SUM52 and SUM190 were obtained from expression levels are best examples of breast tumor Stephen Ethier and cultured in a 50:50 mixture of mammary derived cells that show oncogenic addiction to PAK1. epithelial cell growth medium (MEGM) and DMEM/F12 (Invitrogen Carlsbad, CA, USA) with 10% inactivated fetal Conversely, although HMLER cells did not show calf serum (Sigma-Aldrich, St Louis, MO, USA). SKBR3 and overexpression or elevated activity of PAK1, they were EFM19 were cultured in RPMI or RPMI (À) phenol red dependent on PAK1 for transformation, indicating that (Invitrogen), respectively, with 10% inactivated fetal calf PAK1 can also behave as an essential gene in RAS- serum. All media contained penicillin/streptomycin (Cellgro, driven cancers. In sum, we present evidence that PAK1 Manassas, VA, USA).

Oncogene PAK1 drives breast transformation via MAPK and MET Y Shrestha et al 3406 negative controls, respectively. HMLEA cells expressing the pools or controls were seeded in 0.3% Noble agar (Sigma- Aldrich) in six-well plates, six replicate wells/pool, and assayed for colony formation. Bottom agar consisted of Dulbecco0s Modified Eagle’s Medium (DMEM) with 0.6% Noble agar, 8% inactivated fetal calf serum and Antibiotic-Antimycotic (Invitrogen). Colony formation was assessed after 2 weeks when images of each well were taken at  6.25 magnification on Olympus SZX9 microscope with Olympus (Center Valley, PA, USA) Qcolor 3 camera using Magnafire or Q Capture and analyzed with ImageJ (Bethesda, MA, USA) software. Macro- scopic colonies (4100 square pixels or 0.004 m) with the circularity of 0.5–1 were counted. Median colony number of the pools was calculated and those that scored 1.5 s.d.s above median were selected for further study. Colonies 450 square pixels (0.002 m) with the circularity of 0.8–1 were counted in all colony formation assays post screening.

Immunoblotting For immunoblot analyses, samples were harvested in RIPA (Cell Signaling Technology (CST), Boston, MA, USA) or CLB (CST) with 1 mM phenylmethanesulfonyl fluoride (Sigma- Aldrich), protease inhibitor (Roche, Manheim, Germany) and phosphatase inhibitor (EMD Biosciences). Samples were Figure 8 PAK1 simultaneously activates two distinct signaling run in 4–12% bis-tris gels (Invitrogen) in MOPS buffer pathways to promote transformation. The model depicts PAK1 (Invitrogen). Gels were transferred using the iBlot system mechanism of action to activate the RAS-MAPK (orange) pathway (Invitrogen). For drug inhibition analysis by immunoblot, cells through direct activation of RAF1 and MEK1. The second arm were treated overnight with dimethyl sulfoxide, U0126 or (purple) shows that PAK1 and merlin are in an inhibitory loop, PHA-665752 at the indicated concentrations with constant which can be pushed by PAK1 amplification and overexpression to total volume of dimethyl sulfoxide with or without drug. inhibit merlin that leads to MET signaling activation through Antibodies used in the study were obtained from CST except GAB1. A full colour version of this figure is available at the for anti-EGFR, which was a gift from Jeonghee Cho, anti- Oncogene journal online. merlin (Santa Cruz Biotechnology, Santa Cruz, CA, USA), and 4G10 (Millipore, Temecula, CA, USA), pBabe-Puro-Flag-PAK1K299R was derived by site-directed mutagenesis of pBabe-Puro-Flag-PAK1 using QuikChange- Quantitative reverse transcriptase PCR IIXL (Stratagene, Santa Clara, CA, USA). Primer sequences RNA was harvested using QiaShredder and RNeasy (Qiagen, are shown in Supplementary Table 1. Retroviral and lentiviral Valencia, CA, USA). Complementary DNA was prepared production was carried out as previously described (Boehm using Advantage RT-for-PCR (Clontech, Mountain View, et al., 2005; Moffat et al., 2006). Infections were modified to CA, USA). Quantitative PCR was carried out using SYBR incorporate spins at 1178 Â g (2250 r.p.m.) for 30 min at 25 1C. (Applied Biosystems, Beverly, MA, USA). The primers used U0126 (EMD Biosciences, Gibbstown, NJ, USA) and PHA- are listed in Supplementary Table 1. 665752 (Tocris, Ellisville, MO, USA) were reconstituted in dimethyl sulfoxide and added to cells while seeding in soft Proliferation assay agar. shRNAs were obtained from the RNAi Consortium SUM52 expressing shRNAs against PAK1 or lacZ or HMLE (TRC) at the Broad Institute; the corresponding reference expressing vector control or PAK1 were plated in triplicate in a numbers are as follows: shPAK1 no. 1-TRCN0000195500, six-well plates. Cells were counted the next day (day 1) as shPAK1 no. 2-TRCN0000025258, shNF2 no. 1- baseline and then counted and replated for 21 or 23 days at TRCN0000039974, shNF2 no. 2-TRCN0000010397, shRAF1 indicated time points to calculate population-doubling times. no. 1-TRCN0000195502, shRAF1 no. 2-TRCN0000196969, shMET no. 1-TRCN0000199327, shMET no. 2-TRCN00- 00009850, shGAB1 no. 1-TRCN0000074283 and shGAB1 no. Fluorescence in situ hybridization 2-TRCN0000074285. Fluorescence in situ hybridization was carried out as pre- viously described (Knoll and Lichter, 2005). Human chromo- some 11 cep probe and BAC probe 11-381E5 (PAK1) were Pooled human kinase ORF screen used. Hybridization signals were viewed on an Olympus BX-51 The second-generation pBabe-Puro-Myr-Flag kinase ORF fluorescent microscope system. library was developed by the Center for Cancer Systems Biology at the Dana-Farber Cancer Institute and TRC Luminex assay (Johannessen et al., 2010). This ORF library consists of 597 Lysates were harvested in normal growth conditions in kinase-related ORFs that were pooled at equal DNA amounts MEGM. The Luminex assay was carried out for TKs and in random order into 22 pools, each containing 25–27 ORFs. associated proteins as previously described (Du et al., 2009). HMLEA was infected with virus produced from the DNA Raw data were normalized by subtracting antibody and pools. Cells were then selected for puromycin resistance at sample backgrounds. One or more antibodies (analytes) 1 mg/ml. pBabe-Puro-MEKDD, diluted 1:25 with pBabe-Puro, against a specific TK or associated protein may be present. and the parental vector pBabe-Puro were used as positive and Average normalized signal of triplicates for each of the 244

Oncogene PAK1 drives breast transformation via MAPK and MET Y Shrestha et al 3407 analytes was calculated. Normalized signals of 10 or higher in Conflicts of interest HMLE–PAK1 and HMLE–PAK1K299R were considered posi- tive. Normalized signals for analytes in HMLE–PAK1 were WCH, KP, JJZ and RB are consultants for Novartis divided by corresponding signals in HMLE–PAK1K299R, and Pharmaceuticals. those with signal fold difference of 1.5 or higher were considered significantly increased. Acknowledgements Copy number analysis Tumorscape was analyzed as described (Beroukhim et al., We thank Nicole A Spardy, Christine L Nguyen and Yaara C 2010). GISTIC profile for Tumorscape was created with all Zwang for critical reading of the manuscript. We also thank 3131 tumor samples and cancer cell lines including 243 breast the Dana-Farber/Harvard Cancer Center Cytogenetics Core samples. Copy number profiles were created for each cancer for their assistance with fluorescence in situ hybridization type with up to 80 samples that show PAK1 amplification. studies. This work was supported in part by a Department of Breast samples with a score of 0.8 (log2 copy number ratio) or Defense Breast Cancer Research Program Pre-doctoral higher in Tumorscape were taken into consideration for the Fellowship (W81XWH-08-1-0767, YS), grants from the US PAK1-CCND1 co-amplification study. GISTIC profile for National Cancer Institute (R33 CA128625 and R01 TCGA breast invasive adenocarcinoma was created with 507 CA130988, WCH) and a Cancer Center Support Grant no. cancer samples. NIH 5 P30 CA06516 (Cytogenetics Core).

References

Appledorn DM, Dao KH, O’Reilly S, Maher VM, McCormick JJ. Brown LA, Kalloger SE, Miller MA, Shih Ie M, McKinney SE, Santos (2010). Rac1 and Cdc42 are regulators of HRasV12-transformation JL et al. (2008). Amplification of 11q13 in ovarian carcinoma. Genes and angiogenic factors in human fibroblasts. BMC Cancer Chromosomes Cancer 47: 481–489. 10: 13. Chattopadhyay I, Singh A, Phukan R, Purkayastha J, Kataki A, Arias-Romero LE, Villamar-Cruz O, Pacheco A, Kosoff R, Huang M, Mahanta J et al. (2010). Genome-wide analysis of chromosomal Muthuswamy SK et al. (2010). A Rac-Pak signaling pathway is alterations in patients with esophageal squamous cell carcinoma essential for ErbB2-mediated transformation of human breast exposed to tobacco and betel quid from high-risk area in India. epithelial cancer cells. Oncogene 29: 5839–5849. Mutat Res 696: 130–138. Bachman KE, Argani P, Samuels Y, Silliman N, Ptak J, Szabo S et al. Christensen JG, Schreck R, Burrows J, Kuruganti P, Chan E, Le P (2004). The PIK3CA gene is mutated with high frequency in human et al. (2003). A selective small molecule inhibitor of c-Met kinase breast cancers. Cancer Biol Ther 3: 772–775. inhibits c-Met-dependent phenotypes in vitro and exhibits cytor- Batzer AG, Rotin D, Urena JM, Skolnik EY, Schlessinger J. (1994). eductive antitumor activity in vivo. Cancer Res 63: 7345–7355. Hierarchy of binding sites for Grb2 and Shc on the epidermal Coles LC, Shaw PE. (2002). PAK1 primes MEK1 for phosphorylation growth factor receptor. Mol Cell Biol 14: 5192–5201. by Raf-1 kinase during cross-cascade activation of the ERK Bekri S, Adelaide J, Merscher S, Grosgeorge J, Caroli-Bosc F, pathway. Oncogene 21: 2236–2244. Perucca-Lostanlen D et al. (1997). Detailed map of a region Curto M, Cole BK, Lallemand D, Liu CH, McClatchey AI. (2007). commonly amplified at 11q13–4q14 in human breast carcinoma. Contact-dependent inhibition of EGFR signaling by Nf2/Merlin. Cytogenet Cell Genet 79: 125–131. J Cell Biol 177: 893–903. Beroukhim R, Getz G, Nghiemphu L, Barretina J, Hsueh T, Linhart D Davies H, Bignell GR, Cox C, Stephens P, Edkins S, Clegg S et al. et al. (2007). Assessing the significance of chromosomal aberrations (2002). Mutations of the BRAF gene in human cancer. Nature 417: in cancer: methodology and application to glioma. Proc Natl Acad 949–954. Sci USA 104: 20007–20012. Downward J. (2003). Targeting RAS signalling pathways in cancer Beroukhim R, Mermel CH, Porter D, Wei G, Raychaudhuri S, therapy. Nat Rev Cancer 3: 11–22. Donovan J et al. (2010). The landscape of somatic copy-number Du J, Bernasconi P, Clauser KR, Mani DR, Finn SP, Beroukhim R alteration across human cancers. Nature 463: 899–905. et al. (2009). Bead-based profiling of tyrosine kinase phosphoryla- Boccaccio C, Ando M, Tamagnone L, Bardelli A, Michieli P, tion identifies SRC as a potential target for glioblastoma therapy. Battistini C et al. (1998). Induction of epithelial tubules by Nat Biotechnol 27: 77–83. growth factor HGF depends on the STAT pathway. Nature 391: Dummler B, Ohshiro K, Kumar R, Field J. (2009). Pak protein kinases 285–288. and their role in cancer. Cancer Metastasis Rev 28: 51–63. Boehm JS, Hession MT, Bulmer SE, Hahn WC. (2005). Transforma- Elenbaas B, Spirio L, Koerner F, Fleming MD, Zimonjic DB, tion of human and murine fibroblasts without viral oncoproteins. Donaher JL et al. (2001). Human breast cancer cells generated by Mol Cell Biol 25: 6464–6474. oncogenic transformation of primary mammary epithelial cells. Boehm JS, Zhao JJ, Yao J, Kim SY, Firestein R, Dunn IF et al. Genes Dev 15: 50–65. (2007). Integrative genomic approaches identify IKBKE as a breast Engelman JA, Chen L, Tan X, Crosby K, Guimaraes AR, Upadhyay cancer oncogene. Cell 129: 1065–1079. R et al. (2008). Effective use of PI3K and MEK inhibitors to treat Bokoch GM. (2003). Biology of the p21-activated kinases. Annu Rev mutant Kras G12D and PIK3CA H1047R murine lung cancers. Nat Biochem 72: 743–781. Med 14: 1351–1356. Bollag G, Clapp DW, Shih S, Adler F, Zhang YY, Thompson P et al. Favata MF, Horiuchi KY, Manos EJ, Daulerio AJ, Stradley DA, (1996). Loss of NF1 results in activation of the Ras signaling Feeser WS et al. (1998). Identification of a novel inhibitor pathway and leads to aberrant growth in haematopoietic cells. Nat of mitogen-activated protein kinase kinase. J Biol Chem 273: Genet 12: 144–148. 18623–18632. Bostner J, Ahnstrom Waltersson M, Fornander T, Skoog L, Freihoff D, Kempe A, Beste B, Wappenschmidt B, Kreyer E, Nordenskjold B, Stal O. (2007). Amplification of CCND1 and Hayashi Y et al. (1999). Exclusion of a major role for the PAK1 as predictors of recurrence and tamoxifen resistance in PTEN tumour-suppressor gene in breast carcinomas. Br J Cancer postmenopausal breast cancer. Oncogene 26: 6997–7005. 79: 754–758.

Oncogene PAK1 drives breast transformation via MAPK and MET Y Shrestha et al 3408 Hennessy BT, Smith DL, Ram PT, Lu Y, Mills GB. (2005). Exploiting Neve RM, Holbro T, Hynes NE. (2002). Distinct roles for the PI3K/AKT pathway for cancer drug discovery. Nat Rev Drug phosphoinositide 3-kinase, mitogen-activated protein kinase and Discov 4: 988–1004. p38 MAPK in mediating cell cycle progression of breast cancer cells. Hirokawa Y, Tikoo A, Huynh J, Utermark T, Hanemann CO, Oncogene 21: 4567–4576. Giovannini M et al. (2004). A clue to the therapy of neurofibro- Santen RJ, Song RX, McPherson R, Kumar R, Adam L, Jeng MH matosis type 2: NF2/merlin is a PAK1 inhibitor. Cancer J 10: 20–26. et al. (2002). The role of mitogen-activated protein (MAP) kinase in Hoeflich KP, O’Brien C, Boyd Z, Cavet G, Guerrero S, Jung K et al. breast cancer. J Steroid Biochem Mol Biol 80: 239–256. (2009). In vivo antitumor activity of MEK and phosphatidylinositol Scheid MP, Woodgett JR. (2001). Phosphatidylinositol 30 kinase 3-kinase inhibitors in basal-like breast cancer models. Clin Cancer signaling in mammary tumorigenesis. J Mammary Gland Biol Res 15: 4649–4664. Neoplasia 6: 83–99. Hynes NE, MacDonald G. (2009). ErbB receptors and signaling Schraml P, Schwerdtfeger G, Burkhalter F, Raggi A, Schmidt D, pathways in cancer. Curr Opin Cell Biol 21: 177–184. Ruffalo T et al. (2003). Combined array comparative genomic Johannessen CM, Boehm JS, Kim SY, Thomas SR, Wardwell L, hybridization and tissue microarray analysis suggest PAK1 at Johnson LA et al. (2010). COT drives resistance to RAF inhibition 11q13.5-q14 as a critical oncogene target in ovarian carcinoma. Am through MAP kinase pathway reactivation. Nature 468: 968–972. J Pathol 163: 985–992. Kim DH, Muto M, Kuwahara Y, Nakanishi Y, Watanabe H, Aoyagi Shaw RJ, Paez JG, Curto M, Yaktine A, Pruitt WM, Saotome I et al. K et al. (2006). Array-based comparative genomic hybridization of (2001). The Nf2 tumor suppressor, merlin, functions in Rac- circulating esophageal tumor cells. Oncol Rep 16: 1053–1059. dependent signaling. Dev Cell 1: 63–72. Kim SY, Dunn IF, Firestein R, Gupta P, Wardwell L, Repich K et al. Sos ML, Fischer S, Ullrich R, Peifer M, Heuckmann JM, Koker M (2010). CK1epsilon is required for breast cancers dependent on et al. (2009). Identifying genotype-dependent efficacy of single and beta-catenin activity. PLoS One 5: e8979. combined PI3K- and MAPK-pathway inhibition in cancer. Proc Kissil JL, Johnson KC, Eckman MS, Jacks T. (2002). Merlin Natl Acad Sci USA 106: 18351–18356. phosphorylation by p21-activated kinase 2 and effects of phosphor- Tang Y, Chen Z, Ambrose D, Liu J, Gibbs JB, Chernoff J et al. (1997). ylation on merlin localization. J Biol Chem 277: 10394–10399. Kinase-deficient Pak1 mutants inhibit Ras transformation of Rat-1 Kissil JL, Wilker EW, Johnson KC, Eckman MS, Yaffe MB, Jacks T. fibroblasts. Mol Cell Biol 17: 4454–4464. (2003). Merlin, the product of the Nf2 tumor suppressor gene, is an Trusolino L, Bertotti A, Comoglio PM. (2010). MET signalling: inhibitor of the p21-activated kinase, Pak1. Mol Cell 12: 841–849. principles and functions in development, organ regeneration and Knoll JH, Lichter P. (2005). In situ hybridization to metaphase cancer. Nat Rev Mol Cell Biol 11: 834–848. chromosomes and interphase nuclei. Curr Protoc Hum Genet Tu HF, Chang KW, Chiang WF, Liu CJ, Yu EH, Liu ST et al. (2011). Chapter 4: Unit 4 3. The frequent co-expression of the oncogenes PIK3CA and PAK1 in Lambros MB, Fiegler H, Jones A, Gorman P, Roylance RR, Carter oral carcinomas. Oral Oncol 47: 211–216. NP et al. (2005). Analysis of ovarian cancer cell lines using array- Vadlamudi RK, Adam L, Wang RA, Mandal M, Nguyen D, Sahin A based comparative genomic hybridization. J Pathol 205: 29–40. et al. (2000). Regulatable expression of p21-activated kinase-1 Lau KS, Haigis KM. (2009). Non-redundancy within the RAS promotes anchorage-independent growth and abnormal organiza- oncogene family: insights into mutational disparities in cancer. tion of mitotic spindles in human epithelial breast cancer cells. J Biol Mol Cells 28: 315–320. Chem 275: 36238–36244. McClatchey AI. (2007). Neurofibromatosis. Annu Rev Pathol 2: Weidner KM, Di Cesare S, Sachs M, Brinkmann V, Behrens J, 191–216. Birchmeier W. (1996). Interaction between Gab1 and the c-Met Mendelsohn J, Baselga J. (2000). The EGF receptor family as targets receptor tyrosine kinase is responsible for epithelial morphogenesis. for cancer therapy. Oncogene 19: 6550–6565. Nature 384: 173–176. Miron A, Varadi M, Carrasco D, Li H, Luongo L, Kim HJ et al. Weinman EJ, Hall RA, Friedman PA, Liu-Chen LY, Shenolikar S. (2010). PIK3CA mutations in in situ and invasive breast carcino- (2006). The association of NHERF adaptor proteins with g protein- mas. Cancer Res 70: 5674–5678. coupled receptors and receptor tyrosine kinases. Annu Rev Physiol Mirzoeva OK, Das D, Heiser LM, Bhattacharya S, Siwak D, 68: 491–505. Gendelman R et al. (2009). Basal subtype and MAPK/ERK kinase Xiao GH, Beeser A, Chernoff J, Testa JR. (2002). p21-activated kinase (MEK)-phosphoinositide 3-kinase feedback signaling determine links Rac/Cdc42 signaling to merlin. J Biol Chem 277: 883–886. susceptibility of breast cancer cells to MEK inhibition. Cancer Res Yamauchi T, Ueki K, Tobe K, Tamemoto H, Sekine N, Wada M et al. 69: 565–572. (1997). Tyrosine phosphorylation of the EGF receptor by the kinase Miyakis S, Sourvinos G, Spandidos DA. (1998). Differential expres- Jak2 is induced by growth hormone. Nature 390: 91–96. sion and mutation of the ras family genes in human breast cancer. Zang M, Gong J, Luo L, Zhou J, Xiang X, Huang W et al. (2008). Biochem Biophys Res Commun 251: 609–612. Characterization of Ser338 phosphorylation for Raf-1 activation. Moffat J, Grueneberg DA, Yang X, Kim SY, Kloepfer AM, Hinkle G J Biol Chem 283: 31429–31437. et al. (2006). A lentiviral RNAi library for human and mouse Zang M, Hayne C, Luo Z. (2002). Interaction between active Pak1 and genes applied to an arrayed viral high-content screen. Cell 124: Raf-1 is necessary for phosphorylation and activation of Raf-1. 1283–1298. J Biol Chem 277: 4395–4405.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

Oncogene