Activating ROCK1 Somatic Mutations in Human Cancer

PA Lochhead1,2, G Wickman1, M Mezna and MF Olson

The Beatson Institute for Cancer Research, Glasgow, Scotland, UK

Cancer cells acquire characteristics of deregulated rous cancers (http://www.sanger.ac.uk/genetics/CGP/ growth, survival and increased metastatic potential. and http://cancergenome.nih.gov/). Statistical analysis Genetic mutations that provide a selective advantage of mutation frequencies can be used to classify them as by promoting these characteristics have been termed active ‘driver’ mutations that provide a selective ‘drivers,’ whereas mutations that do not contribute to advantage during cancer initiation and progression, or disease initiation/progression are termed ‘passengers.’ silent ‘passengers,’ which are expanded from progenitor The advent of high-throughput methodologies has facili- cells, yet provide no overt selective advantage. Although tated large-scale screening of cancer genomes and the this approach has tremendous value in identifying subsequent identification of novel somatic mutations. important genes in cancer, it remains incomplete until Although this approach has generated valuable results, the biological consequences of detected mutations are the data remain incomplete until the functional conse- determined to differentiate drivers from passengers. quences of these mutations are determined to differentiate Cancer genome sequencing has revealed three novel potential drivers from passengers. ROCK1 is an essential ROCK1 mutations in human malignant disease (http:// effector kinase downstream of Rho GTPases, an im- www.sanger.ac.uk/cgi-bin/genetics/CGP/cosmicsearch?_ portant pathway involved in cell migration. The Cancer q ¼ rock1). The three somatic mutations in ROCK1 are: Genome Project identified three nonsynonymous mutations AA insertion between base pair (bp) 1214–1215, in the ROCK1 gene. We now show that these somatic resulting in frameshift and premature termination at ROCK1 mutations lead to elevated kinase activity and Y405*; C to G transition at bp 3377, resulting in drive actin cytoskeleton rearrangements that promote premature termination at S1126*; and C to T trans- increased motility and decreased adhesion, characteristics version at bp 3577, resulting in the amino-acid of cancer progression. Mapping of the kinase-interacting substitution P1193S (Figure 1a). The Y405* and regions of the carboxy terminus combined with structural S1126* mutants were both identified in primary human modeling provides an insight into how these mutations breast cancers, whereas the third mutation, P1193S, was likely affect the regulation of ROCK1. Consistent with the identified from the established human non-small-cell frequency of ROCK1 mutations in human cancer, these lung carcinoma line NCI-H1770 (Greenman et al., results support the conclusion that there is selective 2007). pressure for the ROCK1 gene to acquire ‘driver’ mutations ROCK1 and ROCK2 (Rho-associated kinases 1 and 2) that result in kinase activation. are significant effector kinases of the small GTPase Oncogene (2010) 29, 2591–2598; doi:10.1038/onc.2010.3; RhoA (Olson and Sahai, 2009). The primary function of published online 8 February 2010 this pathway is the generation of acto-myosin contrac- tile force, which influences behaviors including motility, Keywords: ROCK; kinase; Rho; actin; cytoskeleton; survival and proliferation. There is a wealth of data cancer implicating Rho GTPases in human cancer (Sahai and Marshall, 2002; Benitah et al., 2004) further suggesting that ROCK1 and ROCK2, as key mediators of Rho signaling, may have important roles (Narumiya et al., The identification of mutated genes that drive human 2009). In support of this, preclinical studies have shown oncogenesis is an ongoing major objective worldwide. beneficial effects of ROCK inhibition on tumor inci- The development of high-throughput detection dence rates, volume, invasiveness and metastasis (Itoh methodologies has facilitated the identification of et al., 1999; Somlyo et al., 2000; Nakajima et al., 2003; novel somatic mutations associated with nume- Ying et al., 2006; Ogawa et al., 2007; Xue et al., 2008). In addition, elevated ROCK1 and/or ROCK2 expres- Correspondence: Professor MF Olson, Molecular Cell Biology, The sion have been detected in several human cancers, which Beatson Institute for Cancer Research, Garscube Estate, Switchback Road, Glasgow, Scotland G61 1BD, UK. correlated with poor outcome (Kamai et al., 2002, 2003, E-mail: [email protected] 2004; Abe et al., 2008). Although alterations in the 1These authors contributed equally to this work. regulation of ROCK1 signaling as an important 2Current address: The Laboratory of Molecular Signalling, The contributory factor to human cancer may be well Babraham Research Institute, Babraham Research Campus, Cam- bridge CB22 3AT, UK accepted, this is the first report of ROCK1 activation Received 10 July 2009; revised 2 December 2009; accepted 30 December as a direct consequence of somatic mutations. In this 2009; published online 8 February 2010 study, we describe the biochemical and biological Activating ROCK1 somatic mutations in cancer PA Lochhead et al 2592 activities of these mutants and conclude that these of the MYC-tagged ROCK1 proteins in NIH 3T3 ROCK1 mutations are activating. These observations fibroblasts, assessed by flow cytometry, was comparable are consistent with the statistical analysis of nonsynony- for WT, S1126* and P1193S, whereas expression of mous versus synonymous mutation rates, which indi- K105G and Y405* was lower (Supplemental Figures 1a cates that the ROCK1 gene is under positive selection and b). Within 6 h of transfection, the active ROCK1 pressure as a cancer driver (Greenman et al., 2007) mutants Y405*, S1126* and P1193S induced cellular To determine how these mutations affected ROCK1 contraction, significantly reducing the mean area per cell activity, MYC-tagged wild type (WT), kinase-dead (Figure 2a). No differences in subcellular localization of K105G, and Y405*, S1126* and P1193S proteins were the ROCK1 mutants compared with WT were observed expressed in NIH 3T3 mouse fibroblast cells. WT, (data not shown). When the effects of the active ROCK1 K105G and P1193S migrated at 158 kDa, whereas mutants on cell morphology were examined, Y405*, Y405* and S1126* migrated as expected at 47 and S1126* and P1193S expression markedly modified cell 113 kDa, respectively, as determined by western blotting shape, shifting the distribution of morphologies from (Figure 1b). In vitro kinase assays of the MYC- predominant spread to contracted and rounded immunoprecipitated proteins revealed that the activity (Figure 2b). Consistent with the effects on cell contrac- of all three ROCK1 mutants was increased significantly tion, analysis of individual random cell motility by time- (2.5- to 3.5-fold) relative to WT. The kinase-dead lapse microscopy revealed that S1126*- and P1193S- K105G mutant had no detectable activity, indicating expressing NIH 3T3 cells were more motile than GFP- that the active forms possessed intrinsic kinase activity expressing control cells, with significantly longer cumu- that did not result from contaminating MYC-tag lative track lengths (Figure 2c). Although track lengths coimmunoprecipitating kinases (Figure 1b). We next of WT ROCK1-expressing cells were apparently longer determined whether ROCK1 mutants that retained the than GFP control cells, they were not statistically Rho-binding domain could be further activated by significant (Figure 2c), whereas Y405* expression was GTP-loaded RhoA. Consistent with previous results, insufficient to reliably assess its effects. Expression of WT ROCK1 was significantly activated fourfold by S1126* also increased the Euclidean distance traveled RhoA-GTP (Matsui et al., 1996), S1126* activity was and cell velocity (Supplemental Figure 2a and b). We not increased and P1193S was marginally elevated but next examined the effects of ROCK1 mutants on failed to achieve statistical significance. Nonetheless, adhesion using a Transwell dissociation assay (Doki neither mutant was activated to a greater degree than et al., 1993). All three ROCK1 mutants significantly WT, indicating that the ROCK1 mutations do not increased the number of cells that had crossed and enhance RhoA-GTP-induced activation. In addition, dissociated from the Transwell membrane (Figure 2d), expression of Y405*, S1126* and P1193S mutants in indicating reduced cell adhesiveness, a property asso- NIH 3T3 fibroblasts stimulated actin stress fiber ciated with increased metastatic potential. Expression of formation and increased myosin light chain phospho- S1126* was also sufficient to further increase dissocia- rylation on serine19 (Figure 1d), two hallmarks of tion of the highly metastatic MDA MB 231 breast ROCK activity in cells. These results show that the cancer cell line in this assay (Supplemental Figure 2d). novel ROCK1 somatic mutations identified from human Interestingly, the S1126* mutant had the greatest effects cancers have enhanced activity in vitro and in cells. on cell contraction, morphological changes, motility and Given the prominent role of ROCK1 in tumor cell dissociation (Figure 2) as well as inducing more invasion and metastasis (Olson and Sahai, 2009), prominent actin stress fibers and MLC phosphorylation we assessed the effects of the somatic ROCK1 mutations (Figure 1c) relative to the other mutants. These on cell morphology, motility and adhesion. Expression observations indicate that somatic ROCK1 cancer

Figure 1 ROCK1 somatic cancer mutants are active. (a) ROCK1 domain structure for wild-type (WT) and mutants. RBD, Rho-binding domain. Split pleckstrin homology (PH) domains (yellow) have an inserted C1 domain (dark blue). Asterisk indicates the position of amino-acid substitution. Drawn to scale. Plasmids expressing WT and K105G ROCK1 (pCAG MYC ROCK1) were previously described (Coleman et al., 2001). Y405*, S1126* and P1193S were generated using QuikChange (Stratagene, Stockport, UK) site-directed mutagenesis according to the manufacturer’s protocol. (b) MYC-tagged WT, K105G, Y405*, S1126* and P1193S were immunoprecipitated from transfected NIH3T3 cells as previously described (Coleman et al., 2001). Upper panel immunoprecipitates were assayed for ROCK1 kinase activity at 30 1C for 10 min in a total volume of 50 ml containing 50 mM Tris/ 32 6 HCl, pH 7.5, 0.1 mM EGTA, 0.1% (v/v) 2-mercaptoethanol, 10 mM MgCl2, 0.1 mM [g- P]ATP (2 Â 10 c.p.m./nmol) and 50 mM LIMKtide (lkkpdrkkrytvvgnpywma). The reaction was stopped by spotting onto p81 paper and submerging in 1% (v/v) orthophosphoric acid. The amount of phosphate incorporation was determined using Cerenkov counting. Results are mean±s.e.m. activity relative to WT (n ¼ 3, *Po0.05 by t-test). Lower panel, lysates were separated by SDS–polyacrylamide gel electrophoresis and immunoblotted with anti-MYC antibody. Arrows indicate ROCK1 proteins. ~, nonspecific bands. (c) MYC-tagged WT, S1126* and P1193S were immunoprecipitated from transfected HEK293 cells and assayed as described above either without or with the addition of GTP-loaded RhoA. Results are mean±s.e.m. activity relative to activity in the absence of RhoA-GTP (n ¼ 3, *Po0.05 relative to absence of RhoA-GTP by t-test). (d) Representative immunofluorescence confocal images of serum-starved NIH-3T3 fibroblasts stained for actin (top panels) and phosphorylated myosin light chain (pMLC; bottom panels). Cellular nuclei, stained with DAPI, are shown in blue. Arrows indicate cells positive for MYC-tag ROCK1. Scale bars represent 50 mm. After transfection, cells were serum- starved overnight and then fixed and stained as previously described (Coleman et al., 2001). Confocal images were obtained using an Olympus FV1000 (Olympus, Southend, UK) using a 60 Â oil-immersion objective.

Oncogene Activating ROCK1 somatic mutations in cancer PA Lochhead et al 2593 mutations that result in increased kinase activity also active conformation without activation loop phospho- promoted changes in cell morphology, increased moti- rylation (Jacobs et al., 2006). Instead, ROCK1 activity is lity and decreased adhesion. restrained by its inhibitory C terminus; relief from Next, we addressed how these mutations lead to autoinhibition results from Rho-GTP binding or by increased ROCK1 activity. ROCK1 is an unusual AGC caspase cleavage at a C-terminal site (Ishizaki et al., kinase family member as its kinase domain assumes an 1997; Coleman et al., 2001). All three ROCK1

Oncogene Activating ROCK1 somatic mutations in cancer PA Lochhead et al 2594

Figure 2 ROCK1 somatic cancer mutants promote actin rearrangements and cell migration. (a) Mean cell area of ROCK1- transfected NIH 3T3 cells. For cell size determinations, six random fields of MYC-tag-positive cells were captured per treatment. Images were analyzed for percentage MYC-tag-positive area divided by the cell number. Data are mean±s.e.m. (n ¼ 3), statistical comparisons determined by analysis of variance followed by Dunnett’s multiple comparison test. (*Po0.01 vs GFP). (b) Representative images of cell morphologies (spread, contracted and rounded) stained for F-actin (red) and ROCK1 expression (MYC; green). For cell morphology, images were captured and scored for spread, contracted or rounded. Average distribution of morphologies from three experiments for each ROCK1 variant in bottom panel. (c) Representative cell tracks of ROCK1-expressing cells (left panel). Cells were co-transfected in the ratio of 1:2 of GFP and ROCK1 constructs and plated into a glass bottom six-well plate in complete medium. Three hours after nucleofection, the medium was replaced with reduced serum (3.3% fetal bovine serum, FBS) medium. Three bright field and GFP time lapse images were captured for each treatment at every 2 min for 20 h. Individual cell motility was analyzed using Metamorph (Wokingham, UK) and Image J. Mean cumulative track lengths±s.e.m. (n ¼ 4, right panel). Statistical comparison performed by analysis of variance (ANOVA) followed by Dunnett’s multiple comparison test (*Po0.05 vs GFP; **Po0.01 vs GFP). (d) Mean number of dissociated cells following chemotaxis of ROCK1-transfected NIH 3T3 cells through Transwell membranes. Cells were co-transfected in the ratio of 1:2 of GFP and ROCK1 constructs and incubated within an 8 mm pore Transwell insert for 3 h. Growth medium was then aspirated from the Transwell insert and replaced with serum-free medium, and the insert was placed in a fresh well containing medium containing 10% FBS and incubated for 18 h. The number of GFP-positive cells dissociated from the membrane were counted from three independent 20 Â fields. Results are mean±s.e.m. (n ¼ 4), statistical analysis performed by ANOVA followed by Dunnett’s multiple comparison test (**Po0.01 vs GFP).

mutations affect the C terminus, suggesting that they C terminus in ROCK1 autoinhibition, the kinase- activate by disrupting inhibitory interactions between interacting regions had not been previously mapped. N- and C-terminal domains. Despite the role of the We probed a 21-mer peptide array (Figure 3a) compris-

Oncogene Activating ROCK1 somatic mutations in cancer PA Lochhead et al 2595

Figure 3 Characterization of ROCK1 C-terminal kinase-interacting regions. (a) Schematic representation of ROCK1 domain structure, regions analyzed on a peptide array, and the probe used. The probe consisted of MYC-tagged ROCK1 kinase domain (amino acids 1–404). 21-mer peptides offset by two amino acids covered region 1 (red) amino acids 1–75 and region 2 (black) amino acids 853–1354. (b) Autoradiogram of the peptide array probed with 35S labeled protein MYC-tagged ROCK1 kinase domain (upper panel). The five binding regions are outlined with pink, green, brown, orange, blue and red. The locations of these regions in ROCK1 are indicated in the schematic diagram (lower panel). (c) Representative dot blots of protein fragments binding to immobilized ROCK1 kinase domain. Recombinant GST-ROCK1 kinase domain (amino acids 1–535; 200 ng) was spotted directly on nitrocellulose and blocked in 5% milk/TBS, and then incubated with equivalent amounts of indicated recombinant protein fragments in 5% milk/TBS for 18 h at 4 1C. Membranes were washed and then probed with rabbit anti-GST and mouse anti-FLAG antibodies to detect ROCK1 and FLAG-labeled binding proteins, respectively. Infrared fluorophore-conjugated secondary antibodies were detected at 700 and 800 nm using a Li-Cor Odyssey. Surface plot indicates signal intensity for each condition. Data represent mean±s.e.m. of fold increase in protein binding over blank from three to nine replicates. Statistical analysis calculated by analysis of variance followed by Bonferroni’s multiple comparison test (**Po0.01 vs blank or indicated group; ***Po0.001 vs blank). ing amino acids 1–75 containing the N-terminal the Rho-binding domain. The first segment of the split dimerization region and amino acids 853–1354 of the pleckstrin homology (PH) domain contained regions 2f, C terminus with [35S]-labeled ROCK1 kinase domain whereas the second segment overlapped with 2i. Binding (amino acids 1–404). As expected, the probe bound to region 2h was contained within the cysteine-rich the ROCK1 N-terminal residues 27–75 that mediate C1 domain. Kinase binding to the region containing kinase dimerization (Figure 3b, Supplementary Figure P1193 (2g) was weak though detectable. To further 3a) (Jacobs et al., 2006). In addition, the kinase domain characterize these kinase-binding regions, we expressed strongly bound four C-terminal regions (Figures 3a FLAG-tagged protein fragments in Escherichia coli and and b; regions 2d, 2f, 2h and 2i), indicating that tested their abilities to bind recombinant GST-ROCK1 potentially there are multiple sites engaged in kinase kinase domain (200 ng) spotted onto nitrotrocelluose regulation. Region 2d overlapped with the latter half of membranes. Regions 2g, 2h and 2i significantly bound

Oncogene Activating ROCK1 somatic mutations in cancer PA Lochhead et al 2596 ROCK1 kinase domain, whereas 2d binding was cancer drivers. These kinase mutations associated with detectable but did not achieve significance (Figure 3c). cancer have revealed two ROCK1 mutants similar to Region 2f could not be tested because of protein previously characterized activating deletions (caspase- insolubility. The P1193S mutation significantly reduced cleaved active form (Coleman et al., 2001) and an kinase binding by over 30% (Figure 3c). To test the experimentally derived form (Itoh et al., 1999)). In role of these domains in kinase regulation, representa- addition, a novel activating single amino-acid substitu- tive 21-mer peptides from each binding region were tion was identified revealing new information about synthesized and tested in an IMAP fluorescence- ROCK1 regulation. Mapping of the C-terminal kinase polarization ROCK1 kinase assay. Kinase activity interacting regions and structural modeling has revealed was significantly inhibited by 2f and 2g peptides that ROCK1 likely keeps kinase activity in check (Figure 4a), whereas the P1193S mutation significantly through interactions with multiple surfaces. reduced kinase inhibition relative to WT 2g (Figure 4b). Three somatic mutations have also been described in These results indicate that multiple surfaces in the the ROCK2 gene in human cancers; a T to C ROCK1 C terminus likely contribute to kinase-domain transversion at position 3508 in the LB2518 malignant binding and regulation. These interactions range from melanoma cell line that results in an S1194P substitution relatively weak and non-regulatory (2d) to strong (equivalent to S1162 in ROCK1) in the first PH domain, and inhibitory (2g). In addition to the significant insertion of A at 3518–3519 in the CP50-MEL-B contribution of region 2g that contains P1193, region malignant melanoma cell line resulting in a frameshift 2h appears to be an important kinase-binding region, and termination at Y1174* (equivalent to Y1142 in possibly contributing to stabilizing N- and C-terminal ROCK1) in the first PH domain, and deletion of T at interactions. 412 resulting in a frameshift and termination at W138* The data presented above make it easy to conceive in a primary stomach carcinoma (http://www.sanger. how the Y405* and S1126* mutants, which respectively ac.uk/cgi-bin/genetics/CGP/cosmicsearch?_q ¼ rock2). lack all or most of the kinase-interacting and auto- Given the results of this study, the clustering of the inhibitory regions 2f, 2g and 2h, would be constitutively Y1174* and S1194P mutations in the first PH domain activated. Given the high degree of conservation suggests that they would be activating mutations, between the ROCK1 and ROCK2 PH-C1-PH domains although the W138* mutation retains little of the kinase in general, and the 2f, 2g and 2h regions in particular region. (Supplemental Figure 3b), the kinase-interacting and There are numerous examples of kinase activation by inhibitory regions were modeled onto the NMR solution cellular proteolysis (Bokoch, 1998), as exemplified by structures of the PH-C1-PH domains (Wen et al., 2008) caspase cleavage of ROCK1 (Coleman et al., 2001) and (Figures 4c and d). In addition to directly affecting granzyme B cleavage of ROCK2 (Sebbagh et al., 2005). kinase binding and inhibition, as can be seen in The organization of ROCK1 and ROCK2 domains Figure 4c, P1193 breaks a short a-helix before a b-sheet containing N-terminal kinase domains and C-terminal that leads on toward the C1 domain. We predict that the autoinhibitory domains lends them to activation by P1193S mutation does not provide the appropriate proteolysis or by mutations that result in premature a-helix break, and as a consequence will disorder the termination. Therefore, it could be predicted that other split PH domain and likely disrupt the C1 domain. In kinases organized in a similar manner would also be addition, the structural rigidity provided by a proline subjected to activating mutations that affected C- residue would be reduced when substituted with a serine terminal regulatory regions. Indeed, the Cdc42-regu- residue, which might result in less ordered folding. lated MRCKa (aka CDC42BPA), which has a similar These alterations would result in inefficacious kinase role in actin cytoskeleton regulation as ROCK1 and inhibition. In addition, modeling of the 2f, 2g and 2h ROCK2, has an example of a frameshift and termina- regions reveals that they form multiple surfaces located tion at K697* in the HCC1395 breast ductal carcinoma directly opposite to a series of charged lysine and cell line that removes autoinhibitory C1 and PH-like arginine residues that cooperate to mediate lipid domains (Tan et al., 2001). Interestingly, four non- interaction and to orientate ROCK PH-C1-PH domains synonymous mutations were identified for MRCKb in membrane bilayers BPA (Wen et al., 2008) (aka CDC42BPB), all of which occur C terminal to the (Figure 4d). As a result, the critical kinase interaction kinase domain and therefore are potentially activating, and inhibitory domains are exposed to the cytosol and with the mutation most likely to be activating consisting are thus accessible for interaction and inhibition of the of a frameshift and termination at R1092* found in an ROCK catalytic domain. These results show that intestinal adenocarcinoma tumor. although the ROCK1 somatic mutations affect the gene The increased activity of the ROCK1 mutations product differently, they all have a common mechanism suggests independence from the upstream regulator of kinase activation: attenuation of the C-terminal Rho, similar to Ras-independent activating B-Raf inhibitory function. mutations. Although the rate of ROCK1 mutations In summary, the three ROCK1 mutations, Y405*, was low, these results justify screening a larger set of S1126* and P1193S have enhanced kinase activity, cancers to determine a more representative rate and promote contraction, increase motility and decrease identify additional mutations. Furthermore, these ob- adhesion; indicating that these mutations are all servations additionally validate the value of high- constitutively active, consistent with them being active throughput sequencing for identification of new cancer

Oncogene Activating ROCK1 somatic mutations in cancer PA Lochhead et al 2597

Figure 4 ROCK1 kinase-binding regions inhibit kinase activity (a) Representative peptides corresponding to the following spots on the peptide array 2d (d17), 2f (f16), 2g (g15), 2h (h12) and 2i (i12) were synthesized and tested for ROCK1 inhibition. Buffer containing 10 mM Tris/HCl pH 7.2, 10 mM MgCl2, 0.05% NaNO3 and 0.1% phospho-free bovine serum albumin and supplemented with 100 nM fluorescein-tagged peptide substrate derived from S6 Kinase (5FAM-AKRRRLSSLRA), 100 mM ATP and 100 mM of the indicated peptide in 50 ml was incubated for 30 min at room temperature. IMAP binding reagent diluted 1:400 in the buffer described above was added to each well (65 ml), incubated for 30 min at room temperature and then fluorescence polarization read with a Tecan Safire2 multimode microplate reader by exciting the sample at 470 nm and measuring emission at 525 nM.(b) Fluoresence polarization assay of ROCK1 inhibition by 100 mM of 2g, 2g P1193S and Y-27632. Data represent mean kinase activity±s.e.m. from two to five replicates expressed as a percent of control. Statistical analysis calculated by analysis of variance followed by Dunnett’s multiple comparison test (*Po0.05; **Po0.01; ***Po0.001 vs control). (c) Ribbon diagram showing that P1193 (green) is located between an a-helix and a b-sheet in the PH domain (yellow). A part of the kinase-interacting region 2g is highlighted in orange. In full-length ROCK, the interrupting C1 domain would be expected to continue from the orange ribbon (2g) and rejoin the PH domain at the indicated residue in blue. Residues and domains modeled and rendered onto the ROCK2 PH structure (2rov) from PDB coordinates using Pymol. (d) Surface modeling of PH and C1 domains with predicted membrane orientation. ROCK1 binding and inhibitory domains 2 g (orange), 2f (brown) and 2h (light blue) modeled onto separately solved PH (yellow) and C1 (dark blue) structures of ROCK2 (PDB file 2rov and 2row, respectively). Location of P1193 is indicated in green. Residues highlighted in red mediate lipid binding (Wen et al., 2008). Charged head groups of the inner leaflet of a phospholipid bilayer are drawn in gray. Dashed lines indicate the expected structural attachments between the PH and C1 domains. Structures were modeled and rendered in Pymol.

Oncogene Activating ROCK1 somatic mutations in cancer PA Lochhead et al 2598 genes, and for providing valuable unbiased insight into Acknowledgements protein regulation and function. This study was supported by Cancer Research UK and by a Conﬂict of interest grant to MFO from the NIH (CA030721). Assistance with protein modeling was provided by Felix Krueger (Babraham The authors declare no conﬂict of interest. Institute).

References

Abe H, Kamai T, Tsujii T, Nakamura F, Mashidori T, Mizuno T et al. Matsui T, Amano M, Yamamoto T, Chihara K, Nakafuku M, Ito M (2008). Possible role of the RhoC/ROCK pathway in progression of et al. (1996). Rho-associated kinase, a novel serine/threonine kinase, clear cell renal cell carcinoma. Biomed Res 29: 155–161. as a putative target for small GTP binding protein Rho. EMBO J Benitah SA, Valeron PF, van Aelst L, Marshall CJ, Lacal JC. (2004). 15: 2208–2216. Rho GTPases in human cancer: an unresolved link to upstream and Nakajima M, Hayashi K, Egi Y, Katayama K, Amano Y, Uehata M downstream transcriptional regulation. Biochim Biophys Acta 1705: et al. (2003). Effect of Wf-536, a novel ROCK inhibitor, 121–132. against metastasis of B16 melanoma. Cancer Chemother Pharmacol Bokoch GM.. (1998). Caspase-mediated activation of PAK2 during 52: 319–324. apoptosis: proteolytic kinase activation as a general mechanism of Narumiya S, Tanji M, Ishizaki T. (2009). Rho signaling, ROCK and apoptotic signal transduction? Cell Death Differ 5: 637–645. mDia1, in transformation, metastasis and invasion. Cancer Coleman ML, Sahai EA, Yeo M, Bosch M, Dewar A, Olson MF. Metastasis Rev 28: 65–76. (2001). Membrane blebbing during apoptosis results from caspase- Ogawa T, Tashiro H, Miyata Y, Ushitora Y, Fudaba Y, Kobayashi T mediated activation of ROCK I. Nat Cell Biol 3: 339–345. et al. (2007). Rho-associated kinase inhibitor reduces tumor Doki Y, Shiozaki H, Tahara H, Inoue M, Oka H, Iihara K et al. recurrence after liver transplantation in a rat hepatoma model. (1993). Correlation between E-cadherin expression and invasiveness Am J Transplant 7: 347–355. in vitro in a human esophageal cancer cell line. Cancer Res 53: Olson MF, Sahai E. (2009). The actin cytoskeleton in cancer cell 3421–3426. motility. Clin Exp Metastasis 26: 273–287. Greenman C, Stephens P, Smith R, Dalgliesh GL, Hunter C, Bignell G Sahai E, Marshall CJ. (2002). RHO-GTPases and cancer. Nat Rev et al. (2007). Patterns of somatic mutation in human cancer Cancer 2: 133–142. genomes. Nature 446: 153–158. Sebbagh M, Hamelin J, Bertoglio J, Solary E, Breard J. (2005). Direct Ishizaki T, Naito M, Fujisawa K, Maekawa M, Watanabe N, Saito Y cleavage of ROCK II by granzyme B induces target cell membrane et al. (1997). p160ROCK, a Rho-associated coiled-coil forming blebbing in a caspase-independent manner. JExpMed201: 465–471. protein kinase, works downstream of Rho and induces focal Somlyo AV, Bradshaw D, Ramos S, Murphy C, Myers CE, Somlyo adhesions. FEBS Lett 404: 118–124. AP. (2000). Rho-kinase inhibitor retards migration and in vivo Itoh K, Yoshioka K, Akedo H, Uehata M, Ishizaki T, Narumiya S. dissemination of human prostate cancer cells. Biochem Biophys Res (1999). An essential part for Rho-associated kinase in the Commun 269: 652–659. transcellular invasion of tumor cells. Nat Med 5: 221–225. Tan I, Seow KT, Lim L, Leung T. (2001). Intermolecular and Jacobs M, Hayakawa K, Swenson L, Bellon S, Fleming M, Taslimi P intramolecular interactions regulate catalytic activity of myotonic et al. (2006). The structure of dimeric rock i reveals the mechanism dystrophy kinase-related Cdc42-binding kinase {alpha}. Mol Cell for ligand selectivity. J Biol Chem 281: 260–268. Biol 21: 2767–2778. Kamai T, Arai K, Sumi S, Tsujii T, Honda M, Yamanishi T et al. Wen W, Liu W, Yan J, Zhang M. (2008). Structure basis and (2002). The rho/rho-kinase pathway is involved in the progression unconventional lipid membrane binding properties of the PH-C1 of testicular germ cell tumour. BJU Int 89: 449–453. tandem of Rho kinases. J Biol Chem 283: 26263–26273. Kamai T, Tsujii T, Arai K, Takagi K, Asami H, Ito Y et al. (2003). Xue F, Takahara T, Yata Y, Xia Q, Nonome K, Shinno E et al. (2008). Signiﬁcant association of Rho/ROCK pathway with invasion and Blockade of Rho/Rho-associated coiled coil-forming kinase signal- metastasis of bladder cancer. Clin Cancer Res 9: 2632–2641. ing can prevent progression of hepatocellular carcinoma in matrix Kamai T, Yamanishi T, Shirataki H, Takagi K, Asami H, Ito Y et al. metalloproteinase-dependent manner. Hepatol Res 38: 810–817. (2004). Overexpression of RhoA, Rac1, and Cdc42 GTPases is Ying H, Biroc SL, Li WW, Alicke B, Xuan JA, Pagila R et al. (2006). associated with progression in testicular cancer. Clin Cancer Res 10: The Rho kinase inhibitor fasudil inhibits tumor progression in 4799–4805. human and rat tumor models. Mol Cancer Ther 5: 2158–2164.

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

Oncogene