1847.Full-Text.Pdf

Published OnlineFirst January 29, 2016; DOI: 10.1158/0008-5472.CAN-15-1752

Cancer Molecular and Cellular Pathobiology Research

RASSF1A Directly Antagonizes RhoA Activity through the Assembly of a Smurf1-Mediated Destruction Complex to Suppress Tumorigenesis Min-Goo Lee, Seong-In Jeong, Kyung-Phil Ko, Soon-Ki Park, Byung-Kyu Ryu, Ick-Young Kim, Jeong-Kook Kim, and Sung-Gil Chi

Abstract

RASSF1A is a tumor suppressor implicated in many tumorigenic as Rhotekin. As predicted on this basis, RASSF1A competed with processes; however, the basis for its tumor suppressor functions are Rhotekin to bind RhoA and to block its activation. RASSF1A not fully understood. Here we show that RASSF1A is a novel mutants unable to bind RhoA or Smurf1 failed to suppress antagonist of protumorigenic RhoA activity. Direct interaction RhoA-induced tumor cell proliferation, drug resistance, epitheli- between the C-terminal amino acids (256–277) of RASSF1A and al–mesenchymal transition, migration, invasion, and metastasis. active GTP-RhoA was critical for this antagonism. In addition, Clinically, expression levels of RASSF1A and RhoA were inversely interaction between the N-terminal amino acids (69-82) of correlated in many types of primary and metastatic tumors and RASSF1A and the ubiquitin E3 ligase Smad ubiquitination regu- tumor cell lines. Collectively, our ﬁndings showed how RASSF1A latory factor 1 (Smurf1) disrupted GTPase activity by facilitating may suppress tumorigenesis by intrinsically inhibiting the tumor- Smurf1-mediated ubiquitination of GTP-RhoA. We noted that the promoting activity of RhoA, thereby illuminating the potential RhoA-binding domain of RASSF1A displayed high sequence mechanistic consequences of RASSF1A inactivation in many can- homology with Rho-binding motifs in other RhoA effectors, such cers. Cancer Res; 76(7); 1847–59. 2016 AACR.

Introduction RASSF1A is one of the most heavily methylated genes in human cancers and restoration of its expression decreases in The Ras-association domain family (RASSF) of proteins vitro colony formation, suppresses anchorage-independent comprises 10 members encoded by different genes growth, and reduces in vivo tumorigenicity (6, 7). RASSF1A (RASSF1–RASSF10), which share the presence of the RA interacts with proapoptotic MST1 and MST2, known to activate domain (1). The RASSF1 gene located at 3p21.3 is the best the c-Jun N-terminal kinase pathway and also enhances death characterized RASSF member that is epigenetically silenc- receptor–evoked apoptosis by binding to modulator of apo- ed in a wide variety of adult and childhood cancers (2–4). ptosis 1 (MOAP1), a Bax-binding protein and stimulating a The RASSF1 gene generates seven tissue-speciﬁctranscripts complex formation with TNFa receptor-1 (4, 8–10). RASSF1A (RASSF1A-G) and two major variants, RASSF1A and C,are inhibits cyclin D1 accumulation and regulates G –S cell-cycle ubiquitously expressed in normal tissues and have four com- 1 progression in a p53-dependent manner by promoting MDM2 mon exons (exons 3–6), which encode a RA domain (1). self-ubiquitination through disruption of the MDM2–DAXX– RASSF1A encodes a 39 kDa peptide containing several HAUSP complex (11, 12). RASSF1A also induces prometaphase domains that are important for its role as a tumor suppressor, arrest through interaction with Cdc20, an activator of the including the C1 zinc ﬁnger domain for death receptor asso- anaphase-promoting complex (APC) and consequent blockade ciation, the ataxia telangiectasia mutated phosphorylation site of the APC–Cdc20 interaction (13). RASSF1A also binds to and for DNA damage repair and the SAV/RASSF/HIPPO (SARAH) stabilizes microtubules and controls the tubulin dynamics, domain for association with mammalian sterile 20-like which is intimately related with its capacity to promote cell- kinases, MST1 and MST2 (4, 5). cycle arrest and suppress cell motility (14–16). Although RASSF family members contain a RA domain that potentially associates with the Ras family of GTPase, the ability Department of Life Sciences, Korea University, Seoul, Korea. of most RASSF proteins to associate with Ras has yet to be Note: Supplementary data for this article are available at Cancer Research clearly established. To date, only RASSF2, RASSF4, and RASSF5 Online (http://cancerres.aacrjournals.org/). have been observed to associate directly with K-Ras (17–19). Corresponding Author: Sung-Gil Chi, Department of Life Sciences, School of Contradictory reports of Ras association with RASSF1A lead to Life Sciences and Biotechnology, Korea University, 5-1 Anam-dong, Sungbuk- the conjecture that Ras association with RASSF1A might be gu, Seoul 136-701, Republic of Korea (South). Phone: 822-3290-3443; Fax: 822- indirect and most likely mediated through heterodimerization 927-5458; E-mail: [email protected] with RASSF5A (20). Moreover, it is still subject to debate that doi: 10.1158/0008-5472.CAN-15-1752 activeRasisrequiredforRASSF1Atofunctionasatumor 2016 American Association for Cancer Research. suppressor and RASSF1A modulates truly the growth inhibitory

www.aacrjournals.org 1847

Lee et al.

response mediated by Ras and that RASSF1A utilizes an as was performed using tissue arrays (SuperBioChips Laboratory) yet unidentified GTPase to provoke its tumor suppression and Vectastain ABC (avidin-biotin-peroxidase) kit (Vector Labo- effects (4, 10). ratories) as described previously (22). To understand the molecular basis of RASSF1A-mediated tumor suppression, we explored the possible involvement of Fluorescence resonance energy transfer analysis RASSF1A in the regulation of Rho family of small GTPase. Here To determine RASSF1A effect on RhoA activation, a fluo- we present evidence that RASSF1A is a novel antagonist of RhoA rescence resonance energy transfer (FRET) assay was per- tumor-promoting activity, which binds directly to active GTP- formed using Raichu-Rhotekin-RBD as a biosensor of active RhoA and facilitates its proteosomal degradation via interaction RhoA (23). The cells were cotransfected with pRaichu-Rho- with Smad ubiquitination regulatory factor 1 (Smurf1) ubiquitin tekin-RBD and either control-RFP or RASSF1A-RFP and trea- E3 ligase. Thus, our study demonstrates a new mechanism by ted with EGF (50 ng/mL) for 3 hours. The FRET measure- which RASSF1A controls multiple processes in RhoA-driven ments were performed on a Zeiss LSM 700 using emission tumor progression. filters for CFP (433 nm excitation, 450/88 nm emission), YFP (433 nm excitation, 585/42 nm emission), and FRET channel Materials and Methods (433 nm excitation, 530/30 nm emission). Quantification of the average FRET/CFP ratio obtained from image data was Human cell lines and tissue specimens analyzed using MetaMorph software (Version 7.5, Molecular Human cancer cell lines (DU145, HCT116, HeLa, A549, and Devices). MDA-MB-231) were purchased from ATCC. IMR-90 cells (human fetal lung fibroblast) were obtained from Korea Cell In vitro translation, binding, and GST pull-down assays Line Bank. All these cell lines were authenticated by short In vitro transcription and translation were carried out using tandem repeat profiling at Korea Cell Line Bank before use. TNT Quick Coupled reticulocyte lysate system (Promega) and Allelic score data revealed a pattern related to the scores biotinylated tRNA molecule (Transcend tRNA, Promega). For reported by the ATCC, and consistent with their presumptive preparation of nucleotide-free RhoA-G14V and RhoA-T19N, identity. A549 (Tet-RASSF1A) cells were generated by cotrans- Flag-RhoA proteins were incubated in nucleotide binding buffer fection of RASSF1A (pcDNA4/TO) and tetracycline repressor and coupled to Flag agarose beads. The complexes were purified vector (pcDNA6/TR; Invitrogen) and selected under blasticidin with 0.1 mol/L glycine buffer. To generate GTPgS or GDP bound (5 mg/mL) and zeocin (100 mg/mL). A total of 60 primary form, nucleotide-free Flag-RhoA proteins were incubated in tumor specimens and their adjacent normal tissues were nucleotide binding buffer containing GTPgS or GDP and 10 obtained by surgical resection in the Kyung Hee University mmol/L MgCl .Purified RASSF1A-V5 was incubated with either Medical Center (Sungbuk-gu, Seoul, Korea). 2 GTPgS-loaded Flag-RhoA-G14V or GDP-loaded Flag-RhoA- T19N and immunoprecipitated with anti-V5 or anti-Flag anti- Expression plasmids and siRNA body. For GST pull-down assay, 1 mgofGTPgS-loaded GST- Expression vectors for wild-type (WT) and deletion mutants RhoA (G14V) was immobilized on GSH beads and incubated of RASSF1A, RhoA, and Smurf1 were constructed using a PCR- with 1 mg of purified Rhotekin-V5 and increasing amounts of based approach. Single amino acid substitution mutants of RASSF1A-V5. RASSF1A and dominant negative (DN) and constitutively active forms of RhoA, Rac1, Cdc42, H-Ras, K-Ras, and N-Ras were generated using the QuickChange site-directed mutagenesis kit Tumor invasion assay (Stratagene). siRNA duplexes against RASSF1A, RhoA, Smurf1, and Cells were plated onto Biocoat Matrigel invasion membranes S100A4 and short hairpin RNA (shRNA) constructs against (BD Biosciences). After 25-hour incubation, the remaining tumor fi RASSF1A and Smurf1 were synthesized by Dharmacon Research cells on the top surface of the lters were removed by wiping with or purchased from Invitrogen. cotton swabs, and the invading cells on the bottom surface were stained with May–Grunwald€ –Giemsa staining. The number of Small GTPase pull-down assay cells on the bottom surface was counted under a microscope at a fi GTP-bound RhoA, Rac1, and Cdc42 levels were measured using magni cation of 200 . Upstate active Rho assay kit (Millipore Corporation). Briefly, cell lysates were incubated with Rhotekin RGD-agarose (for GTP- Animal studies RhoA) and PAK1 RBD-agarose (for GTP-Rac1 and GTP-Cdc42) Mouse tumor xenograft assay was carried out as described beads for 45 minutes at 4C. The protein/beads complexes were previously (21). For tumor cell colonization assay, A549 cells washed with Upstate lysis buffer and the bound proteins were were injected intravenously and numbers of nodules in the lungs eluted in freshly prepared 2X SDS sample buffer. were counted after 20 days of injection. All animal studies were performed with the approval of Korea University Institutional Immunoprecipitation and IHC Animal Care and Use Committee and Korea Animal Protection Immunoprecipitation (IP) assays were carried out as described Law. previously (21). Antibodies specific for RASSF1A (eB114), Smurf1 (sc-25510), phospho-MLCl (Ser19, #3671), RhoA (sc-418), Rac1 Statistical analysis (sc-95), CDC42 (sc-6083), RhoGAP (sc-30206), RhoGEF Cell proliferation, apoptosis, migration, and invasion assays (sc-20804), b-tubulin (T0198), and Actin (sc-1616) were pur- were performed in triplicate and data were presented as a mean chased from eBioscience, Santa Cruz Biotechnology, Cell Signal- SD. Student t test was used to determine the statistical significance. ing Technology, and Sigma-Aldrich. Immunohistochemical assay Pearson correlation coefficient (r) was used to measure the

1848 Cancer Res; 76(7) April 1, 2016 Cancer Research

RASSF1A Is an Antagonist of RhoA

strength of the association between RASSF1A and RhoA expres- acids 1-97 of RhoA are essential for the interaction (Fig. 2D sion levels in human tissues. A P value of less than 0.05 was and E and Supplementary Fig. S2B). RASSF1A mutants lacking considered significant. this region showed no inhibitory effect on RhoA expression. As predicted, RASSF1A-D256-277 failed to block RhoA-induced Results MLC and IkB phosphorylation (Fig. 2F). Interestingly, the DN4 (lacking amino acids 1-119) but not DN2 (lacking RASSF1A suppresses RhoA activation by destabilizing GTP- amino acids 1-68) mutant also failed to downregulate RhoA, RhoA suggesting that the N-terminal 69-119 residues might be To delineate the RASSF1A regulation of Rho family GTPases, we involved in RhoA degradation (Fig. 2E). Furthermore, all of initially examined its effect on Rho GTPase-induced actin reor- the single amino acid substitution mutants (L266G, R267F, ganization using DU145 prostate cancer cells. In pcDNA control L268G, R269F, L270G, and G272L) we tested failed to bind subline cells, transfection of constitutively active RhoA (G14V), and downregulate RhoA. Likewise, RASSF1A-R269F showed Rac1 (T115I), and Cdc42 (G12V) led to the formation of stress no inhibitory effect on EGF activation of RhoA signaling (Sup- fiber, lamellipodia, and filopodia, respectively (Fig. 1A). Howev- plementary Fig. S2C). Interestingly, RASSF1C, which shares er, RhoA-induced stress fiber and membrane ruffling were sub- amino acids 129-340 with RASSF1A but has unique exon 2g- stantially attenuated in DU145-RASSF1A subline cells whereas encoding N-terminal 49 amino acids, did not interact with Rac1-induced lamellipodia and Cdc42-induced filopodia were RhoA, but a exon 2g-deleted RASSF1C mutant showed RhoA- not affected. RASSF1A inhibition of RhoA-induced membrane binding ability, suggesting that the exon 2g may block the protrusion was also observed in J82 (bladder), A549 (lung), and RhoA-binding activity provoked by the 256-277 residues HCT116 (colon) cells (Supplementary Fig. S1A). Moreover, a (Fig. 2B and Supplementary Fig. S2D and S2E). FRET assay using Raichu-Rhotekin-RBD as a biosensor for active RhoA revealed that EGF-mediated RhoA activation is blocked by RASSF1A (Fig. 1B). Consistently, RASSF1A inhibited RhoA acti- RASSF1A stimulates Smurf1-mediated RhoA ubiquitination via vation of FAK, MLC, LIMK1/2, and Myc (Supplementary Fig. S1B). direct binding to Smurf1 Ras-induced Raf–Mek–Erk1/2 signaling was not affected by We asked whether Smurf1, a previously reported RhoA-target- RASSF1A (Supplementary Fig. S1C). RASSF1A transfection ing ubiquitin E3 ligase, is involved in RASSF1A-mediated RhoA resulted in a dose-associated decrease in RhoA protein level but degradation (25). A CHX chase assay showed that RASSF1A- did not affect two closely related isoforms, RhoB and RhoC (Fig. induced RhoA degradation is impaired by Smurf1 depletion 1C and D). Unlikely RASSF1A, RASSF1C, an isoform with a single (Supplementary Fig. S3A). Moreover, RASSF1A expression strong- N-terminal exon (exon 2g), showed no RhoA-reducing activity. ly promoted RhoA ubiquitination and this effect was blocked by RhoA cycles between an active GTP-RhoA state and an inactive Smurf1 depletion (Fig. 3A and Supplementary Fig. S3C). The off- target effect of siSmurf1 was excluded by a rescue assay using GDP-RhoA state, and GTP-RhoA level is upregulated and down- 0 regulated by guanine nucleotide exchange factors (GEF) and siSmurf1-3UTR designed to target the 3 -untranslated region and GTPase–activating proteins (GAP), respectively (24). Using small nontargetable Smurf1 expression (Supplementary Fig. S3C). GTPase pull-down assay, we found that GTP-RhoA level is pro- RhoA ubiquitination by RASSF1A was upregulated and down- foundly decreased by RASSF1A (Fig. 1E). Moreover, GTP-RhoA regulated by WT- and DN-Smurf1, respectively (Supplementary In vitro induction by serum, lysophosphatidic acid, or EGF was strongly Fig. S3D). binding and sequential IP assays revealed that upregulated and downregulated by RASSF1A depletion and over- RASSF1A binds directly to Smurf1 and form a protein complex expression, respectively (Fig. 1F and Supplementary Fig. S1D with RhoA and Smurf1 (Fig. 3B and C). It was also shown that the and S1E). RASSF1A exerted no detectable effect on Rho-GAP and N-terminal amino acids 69-82 of RASSF1A and the HECT domain Rho-GEF level, suggesting that RASSF1A may target directly of Smurf1 are responsible for the interaction (Fig. 3D and Sup- GTP-RhoA (Supplementary Fig. S1F). To address this, we com- plementary Fig. S3E). RASSF1A mutants lacking the Smurf1- pared RASSF1A effect on RhoA-G14V (GTP-binding) and RhoA- binding region showed no activity to reduce GTP-RhoA level, – T19N (GDP-binding) and found that RhoA-G14V but not RhoA- and the Smurf1 RhoA interaction was enhanced by WT-RASSF1A T19N level was downregulated by RASSF1A (Fig. 1G). It was but not by RASSF1A-R269F (Fig. 3E and F). These data indicate also shown that the half-life of RhoA protein is reduced from that RASSF1A binds to Smurf1 and promotes Smurf1-mediated – approximately 8.1 hour to 3.9 hour by RASSF1A overexpression RhoA ubiquitination by facilitating Smurf1 RhoA interaction and this effect of RASSF1A is blocked by proteosomal inhibitors (Supplementary Fig. S3F). MG132, LLnL, or Lactacystin (Fig. 1H and I and Supplementary Fig. S1G). RASSF1A suppresses Rhotekin interaction with and activation of RhoA RASSF1A binds to GTP-RhoA via the C-terminal 256-277 Multiple effector molecules interact with RhoA through a residues within the RA domain conserved leucine zipper-like motif (26). Interestingly, we not- Next we examined whether RASSF1A interacts with RhoA. IP ed that the RhoA-binding region (amino acids 256-277) of assay of endogenous and transfected proteins revealed that RASSF1A has a high degree of similarity to the Rho-binding RASSF1A interacts with RhoA-G14V but not with RhoA-T19N motifs of other Rho effectors, including Rhotekin, PKN, PRK2, and active Ras proteins (Fig. 2A and B and Supplementary and Rhophilin, raising the possibility that RASSF1A may com- Fig. S2A). Moreover, in vitro translation-binding assay showed pete with these effectors in binding RhoA (Fig. 4A). As pre- that RASSF1A binds directly to RhoA-G14V (Fig. 2C). Using a dicted, RhoA–Rhotekin and RhoA–RASSF1A interactions were series of deletion mutants, we detected that the C-terminal strongly impeded by RASSF1A and Rhotekin, respectively amino acids 256-277 of RASSF1A and the N-terminal amino (Fig. 4B and C and Supplementary Fig. S4A). Furthermore, a

www.aacrjournals.org Cancer Res; 76(7) April 1, 2016 1849

Lee et al.

A B DU145-pcDNA DU145-RASSF1A A549 Phalloidin GFP DIC Phalloidin GFP DIC (Raichu-Rhotekin-RBD) - ++ EGF (3 h) ++ - Control-RFP --+ RASSF1A-RFP

FRET

1.05 2.24 1.13 Emission ratio

RFP (T115I) Control (G14V)

DIC Cdc42 Rac1 RhoA (G12V)

CEHCT116 D m 0 0.5 1 2 RASSF1A ( g) HCT116 A549 MDA-MB231 RhoA 0 0.5 1 2 RASSF1A (mg) 1.0 0.4 0.3 0.1 - + --+ - RASSF1A-V5 --+ --+ RASSF1C-V5 GTP-RhoA Cdc42 kDa RhoA Commasi Rac1 RhoB RASSF1A GTP-Cdc42 RhoC Tubulin Commasi 40 RASSF1(V5) RhoA 25 GTP-Rac1 Cdc42 Tubulin Commasi Rac1 RASSF1A RT-PCR IB GAPDH

G HCT116 F siControl siRASSF1A G14V T19N Flag-RhoA 0 2 4 8 0 2 4 8 Serum (h) 0 1 2 0 1 2 RASSF1A (mg) GTP-RhoA 1.0 2.3 10.7 0.5 1.0 4.9 22.8 9.2 RhoA (Flag) RASSF1A RASSF1A IMR90 Tubulin Tubulin

H HCT116 (RhoA-G14V) I pcDNA RASSF1A A549 0 2 4 6 8 10 12 0 2 4 6 8 10 12 CHX (h) ------++Lactacystin ---- ++ --LLnL RhoA-G14V --+ + ----MG132 RASSF1A - + - + - + - + RASSF1A-V5 Tubulin RhoA 1 0.9 0.8 0.7 0.5 0.4 0.2 1 0.9 0.5 0.1 0.02 0 0 RASSF1A (V5) Relative RhoA levels (RhoA/Tubulin) Tubulin

Figure 1. RASSF1A suppresses RhoA activation through GTP-RhoA degradation. A, RASSF1A inhibition of RhoA-induced stress fiber and membrane ruffleformation. DU145 sublines were transfected with GFP-tagged RhoA, Rac1, or Cdc42. Microscopic examination of membrane protrusions (GFP) and stress fiber formation (phalloidin) was performed at 9 hours after transfection. DIC, differential interface contrast. B, a FRET assay showing RASSF1A inhibition of RhoA activity. A549 cells were cotransfected with Raichu-Rhotekin-RBD sensor and either control-RFP or RASSF1A-RFP. The cells were treated with EGF (10 ng/mL, 3 hours) and emission ratio (YFP/CFP) was measured. C, RASSF1A-mediated RhoA reduction. Immunoblot (IB) and RT-PCR were performed at 48 hours after transfection. D, RASSF1A specificity of RhoA inhibition. E, small GTPase pull-down assay showing RASSF1A downregulation of GTP-RhoA. Commasi staining of the gels was used to adjust input amounts. F, effect of RASSF1A depletion on serum (10%)-induced GTP-RhoA in IMR-90 cells. G, RASSF1A inhibition of GTP-RhoA (G14V) but not of GDP-RhoA (T19N). H, a cycloheximide (CHX) chase assay showing RASSF1A stimulation of RhoA degradation. I, blockade of RASSF1A-induced RhoA degradation by proteosomal inhibitors. Cells were treated with MG132 (5 mmol/L), LLnL (10 mmol/L), or Lactacystin (5 mmol/L) for 6 hours.

1850 Cancer Res; 76(7) April 1, 2016 Cancer Research

RASSF1A Is an Antagonist of RhoA

A B A549 - + -- + -- Flag-RhoA-WT -- + -- + - Flag-RhoA-G14V --- + -- + Flag-RhoA-T19N - + + + --- RASSF1A-V5 Rac1 IP : IgG RASSF1A RhoA Cdc42 ---- + + + RASSF1C-V5 IB: RASSF1A kDa IB: Flag (RhoA) IB: RhoA 40 RASSF1A

HeLa IB: Rac1 IB: V5 IB: Cdc42 25 RASSF1C IP: V5 35 25 RhoA (Flag) 40 RASSF1A (V5) C 25 RASSF1C (V5) WCL Tubulin

––+ ––+ + + + + + + RASSF1A-V5 – + – + – + – –+ Flag-RhoA-G14V kDa – + – + – + – + Flag-RhoA-T19N E 40 RASSF1A 35 RASSF1A-V5 RhoA-T17N RhoA-G14V 25 IB: biotin Input IP: V5 Flag IgG

kDa N1 (45-340) N2 (69-340) N4 (120-340) C2 (1-150) C3 (1-255) C4 (1-264) C5 (1-277) C6 (1-292) D D D D D D D D WT (1-340) pcDNA D 55 256 277 IP: RhoA LRKLLDDEQPLRLRLLAGPSDK IB: V5 35 1 52 101 194 289 340 RhoA IB: RhoA WT (1-340) C1 RA SARAH binding DN1 (45-340) + 55 DN2 (69-340) + 35 RASSF1A DN4 (120-340) + (V5) D C2 (1-150) - 25 WCL D C3 (1-255) - Tubulin DC4 (1-264) - DC5 (1-277) + DC6 (1-292) +

F A549 G Control RhoA-G14V RASSF1A-V5 - + -- + - RASSF1A-WT -- + -- + RASSF1A-D256-277 R269F R269F L270G WT WT L266G L268G G272L R267F R267F kDa P-MLC pcDNA IB: RhoA P-IkB IB: V5 RhoA-G14V IP IB: V5 40 RASSF1A (V5) 25 IB: RhoA Tubulin RhoA V5

Figure 2. RASSF1A binds directly to GTP-RhoA. A, IP assay showing the RASSF1A–RhoA interaction in HeLa cells. B, RASSF1A interaction with RhoA-G14V but not with RhoA-T19N. Cotransfection and IP assays were performed as indicated. C, in vitro translation-binding assay showing a direct interaction between RASSF1A and RhoA. D, RASSF1A deletion mutants and their RhoA-binding status. C1, protein kinase C1 binding domain. RA, Ras association domain; SARAH, SAV/RASSF/HIPPO domain. E, IP assay for determination of RhoA-binding region of RASSF1A. F, loss of RhoA-inhibiting activity of RASSF1A by deletion of the RhoA-binding region. A549 subline cells were transfected with WT or D256-277 RASSF1A and its effect on P-MLC and P-IkB levels was examined at 48 hours after transfection. G, no RhoA-binding and destabilizing activity of mutant RASSF1A carrying single amino acid substitution within the RhoA-interacting region.

www.aacrjournals.org Cancer Res; 76(7) April 1, 2016 1851

Lee et al.

A A549 B ---- siSmurf1 0 1 2 3 3 3 3 RASSF1A-V5 + --- + - + - RASSF1A-V5 kDa + + + + + + + His-Ub - --- - 130 + + + RASSF1C-V5 -- + - + + + + Flag-RhoA-G14V IP: RhoA --- 100 kDa + + + + + Myc-Smurf1 100 75 IB: His Smurf1 40 55 RASSF1A IB: RhoA RASSF1C 35 RhoA-G14V IB: biotin RASSF1A (V5) 25 Smurf1 Input IP: V5 IgG

WCL RhoA Tubulin

D RASSF1A-V5

C N1 (45-340) N2 (69-340) N4 (120-340) C6 (1-292) C2 (1-150) C1 (1-119) N3 (83-340) N3 (83-340) C3 (1-255) D D D D D D D D pcDNA3.1 WT (1-340) kDa Myc-Smurf1 Smurf1 + + + + + + + + + + RASSF1A 40 st nd IgG 2 1 Smurf1 IgG IP : RASSF1A IB: RASSF1A IB: V5 25 IB: Smurf1 IB: RhoA IB: Myc RASSF1A IP : Myc IgG Smurf1 40 RASSF1A

WCL RhoA 25 (V5) Tubulin Smurf1 (Myc) Tubulin WCL

E F A549 RASSF1A-V5 RASSF1A-WT RASSF1A-R269F N3 (83-340) N2 (69-340) N1 (45-340) N4 (120-340) C3 (1-255) D D D D D WT (1-340) WT (1-340) IB: Myc (Smurf1) kDa GTP-RhoA IB: RASSF1A 55 IP:RhoA RhoA (Input) 35 RASSF1A (V5) 25 RASSF1A (V5) Tubulin Smurf1 + + + +++ - RhoA binding WCL RhoA + - + ++ - + Smurf1 binding + - + + + -- RhoA inhibition Tubulin

Figure 3. RASSF1A promotes Smurf1-mediated RhoA ubiquitination. A, Smurf1 dependency of RASSF1A-mediated RhoA ubiquitination. B, in vitro translation-binding assay showing a direct interaction between RASSF1A and Smurf1. C, sequential IP assay showing the RASSF1A–Smurf1–RhoA complex formation in HeLa cells. Cell lysates were precipitated with anti-RASSF1A antibody and the resulting complexes were precipitated with anti-Smurf1 antibody. D, characterization of Smurf1-binding region of RASSF1A. E, loss of RhoA-inhibiting activity of RASSF1A by deletion of either RhoA- or Smurf1-binding region. F, stimulation of the Smurf1–RhoA interaction by WT but not by mutant (R269F) RASSF1A. IP was conducted using equal amounts of RhoA protein (input).

1852 Cancer Res; 76(7) April 1, 2016 Cancer Research

RASSF1A Is an Antagonist of RhoA

A RASSF1A (256-277) L RKL L D D E Q P L R L R L LA G P SD K Rhotekin (21-59) L QRK L D H E - - I R M R D - - G A CK L RTKN (28-45) L QRK L D H E - - I R M R D - - G A CK L Rhophilin (47-64) L HQQ I SK E -- L R M R T- - G A EN L PKN (42-59) L RRE I R K E - - L K L K E - - G A EN L PRK2 (141-158) L QKQ L D I E - - L K V K Q - - G A EN M MEKK1 (437-456) L L GM L D E E - S L T V C E - D G C RN K

B A549 C IMR90 A549 0 0.5 1 2 2 2 2 2 RASSF1A-V5 siRhotekin 2 2 2 2 0 0.5 1 2 Myc-Rhotekin siRASSF1A - + + + + - + + + + EGF (12 h) IB: V5 IB: Rhotekin IB: Myc IB: RASSF1A IB: RhoA

IP: RhoA IB: RhoA RASSF1A (V5) Rhotekin Rhotekin (Myc) RASSF1A WCL WCL WCL WCL IP: RhoA RhoA Tubulin Tubulin

D E MDA-MB-231 RASSF1A-WT Pull-down RASSF1A-R269F - + + + + - + + + + LPA GST GST-RhoA IB: S100A4 m 1 1 1 1 1 1 1 1 Rhotekin-V5 ( g) IB: Rhotekin 0 0.5 1 2 0 0.5 1 2 RASSF1A-V5 (mg) IB: RASSF1A

Rhotekin (V5) IP: RhoA RhoA (Input) RASSF1A (V5) S100A4 Rhotekin GST-RhoA RASSF1A (V5)

GST WCL RhoA Tubulin

F HeLa G MDA-MB-231 siRASSF1A GFP-Control GFP-RASSF1A - + + + + EGF siS100A4 WT R269F 69-82 IB: S100A4 IB: Rhotekin IB: RASSF1A IP: RhoA RhoA (Input) S100A4 Rhotekin RASSF1A WCL RhoA DIC GFP DIC Phalloidin Tubulin RhoA-G14V

Figure 4. RASSF1A hinders RhoA interaction with effector molecules. A, amino acid sequence similarity of RhoA-binding regions of RASSF1A and Rho effectors. B, reciprocal inhibition of RASSF1A and Rhotekin in binding RhoA. C, knockdown assay showing RASSF1A competition with Rhotekin in binding RhoA. D, GST pull-down assay for the competitive relationship between RASSF1A and Rhotekin. Puriﬁed GTPgS-loaded GST-RhoA (G14V) was immobilized on GSH beads and incubated with puriﬁed Rhotekin-V5 and RASSF1A-V5 as indicated. E, RASSF1A suppression of S100A4–RhoA interaction through its RhoA-binding property. F, increased Rhotekin–RhoA interaction by RASSF1A depletion. G, RASSF1A suppression of RhoA-induced, S100A4-mediated lamelliopoia formation.

www.aacrjournals.org Cancer Res; 76(7) April 1, 2016 1853

Lee et al.

GST pull-down assay using purified GST-RhoA (G14V), Rhote- regulated and upregulated by RASSF1A expression and deple- kin-V5, and RASSF1A-V5 proteins revealed that RASSF1A–RhoA tion, respectively (Fig. 5I). interaction interferes with Rhotekin–RhoA interaction, supporting that RASSF1A competes with Rhotekin in binding RASSF1A suppresses RhoA-driven tumor growth and RhoA (Fig. 4D). Consistently, RASSF1A and Rhotekin regula- colonization tion of RhoA ubiquitination was blocked by cotransfection of To delineate the RASSF1A regulation of RhoA in vivo,we Rhotekin and RASSF1A, respectively (Supplementary Fig. S4B carried out mouse tumor xenograft assays using shRNA-medi- and S4C). In contrast, RASSF1A-R269F exerted no effect on ated knockdown of RASSF1A or Smurf1. Compared with shCon- the Rhotekin–RhoA interaction and RhoA ubiquitination trol tumors, shRASSF1A tumors exhibited substantially higher (Supplementary Fig. S4D). Recent studies showed that Rho- growth rate and elevated total and GTP-bound RhoA protein tekin interacts with S100A4, a myosin IIA heavy chain–binding levels (Fig. 6A and B). Moreover, RhoA-driven tumor growth protein, thereby permitting S100A4 to complex with RhoA and was significantly suppressed by WT-RASSF1A but not by D256- switch RhoA function from stress fiber formation to mem- 277 and D69-82 mutants (Fig. 6C and D). Likewise, RASSF1A- brane ruffling to confer an invasive phenotype (27). On the mediated growth suppression and RhoA downregulation were basis of this, we further tested whether RASSF1A antagonizes markedly debilitated in Smurf1-depleted tumors (Fig. 6E and F). RhoA function by blocking RhoA–Rhotekin interaction. Tumor cell colonization assay using tail vein injection of A549 RASSF1A did not affect S100A4 protein level, but its over- cells showed that RhoA-driven tumor cell colonization in the expression and depletion profoundly downregulated and lungs is blocked by WT-RASSF1A but not by D256-277 and D69- upregulated S100A4 interaction with RhoA, whereas 82 mutants in a highly Smurf1-dependent manner (Fig. 6G–I RASSF1A-R269F exerted no effect on S100A4–RhoA interac- and Supplementary Fig. S6A). tion (Fig. 4E and F and Supplementary Figs. S4E and S4F). S100A4 stimulation of RhoA-mediated P-MLC, MMP-9, and P- GTP-RhoA level is inversely correlated with RASSF1A p65 induction was abrogated by WT-RASSF1A but not by expression in human cancers RASSF1A-R269F (Supplementary Fig. S4G). Likewise, RhoA- To ascertain the relationship between RASSF1A and RhoA in induced, S100A4-mediated lamellipodia and membrane ruf- human tumor tissues, we characterized their expression status in fling was attenuated by WT-RASSF1A or RASSF1A-D69-82 but 47 cancer cell lines and 60 primary tumor specimens derived not by RASSF1A-R269F (Fig. 4G). These data indicate that from the colon, stomach, and bladder. Compared with normal RASSF1A antagonizes RhoA function by impeding effector tissues, a substantial fraction of primary tumors displayed interaction with and activation of RhoA. higher GTP-RhoA and lower RASSF1A expression, and an inverse correlation of RASSF1A and GTP-RhoA levels was identified in RASSF1A-mediated tumor suppression is linked to its both cancer cell lines and primary tumors (Fig. 7A and B). RhoA-antagonizing activity Immunohistochemical study also revealed a strong inverse cor- Next we asked whether the RhoA-antagonizing activity of relation between RASSF1A and RhoA in lung, breast, and colo- RASSF1A is linked to its tumor suppression function. RhoA rectal tissues (Fig. 7C and Supplementary Fig. S7A). Although all or EGF activation of cyclins (E and D1) and suppression of of 27 normal tissues we tested exhibited high ( level 2.5) CDK inhibitors (p21WAF1 and p27KIP1) were blocked by RASSF1A and low (< level 2.5) RhoA levels, high RhoA immu- WT-RASSF1A but not by RASSF1A-R269F, and this effect of noreactivity was detected in 82 of 117 (70.1%) low RASSF1A RASSF1A was not detected in RhoA-depleted cells (Fig. 5A and tumors but only in 7 of 33 (21.2%) high RASSF1A tumors WAF1 Supplementary Fig. S5A). RhoA repression of p21 mRNA (Fig. 7D). Furthermore, 26 of 30 (86.7%) metastatic tumors was also impeded by WT-RASSF1A but not by D256-277 or displayed higher RhoA and lower RASSF1A levels compared D69-82 mutant (Supplementary Fig. S5B). Cell growth and with their matched primary tumors. Collectively, our study DNA synthesis were inhibited by WT-RASSF1A but not by identifies that RASSF1A is a novel antagonist of RhoA, whose D256-277 or D69-82 mutant, and RhoA-induced tumor cell loss of expression contributes to RhoA-driven tumor progres- resistance to 5-FU- or etoposide-induced apoptosis was abol- sion (Fig. 7E). ished by WT-RASSF1A but not by R269F or D256-277 mutant (Fig. 5B–D and Supplementary Fig. S5C and S5D). Wound healing and Matrigel assays revealed that RhoA-driven tumor Discussion cell migration and invasion are suppressed by WT-RASSF1A Loss of expression of RASSF1A is implicated in many aspects but not by R269F or L266G mutant, and that increased of tumorigenesis, including tumor growth, invasion, and metas- cell migration triggered by RASSF1A depletion is attenuated tasis. However, important questions that remain unanswered are by codepletion of RhoA (Fig. 5E and F and Supplementary that RASSF1A associates with the Ras family of GTPase and Fig. S5E and S5F). Likewise, RhoA-induced MMP-9 expres- RASSF1A utilizes an as yet unidentified GTPase to provoke its sion was blocked by WT-RASSF1A but not by R269F mut- tumor suppression effects. In the current study, we demonstrated ant (Fig. 5G). Moreover, in DU145 cells, TGFb1-induced, first that RASSF1A binds directly to RhoA GTPase, promotes RhoA-mediated epithelial–mesenchymal transition (EMT) Smurf1-mediated RhoA ubiquitination, and blocks RhoA-driven was blocked by WT-RASSF1A but not by R269F mutant tumorigenic processes. Our study thus establishes that RASSF1A whereas it is further promoted by depletion of RASSF1A or represents one critical RhoA-interacting protein that antag- Smurf1 (Fig. 5H and Supplementary Figs. S5G–S5I). Consis- onizes RhoA's tumor-promoting activity, adding a new mecha- tently, TGFb1 repression of epithelial markers (E-cadherin, nism by which RASSF1A functions as a tumor suppressor. ZO-1, and cytokeratin 18) and induction of mesenchymal The presence of the RA domain initially led to the conjecture markers (fibronectin, N-cadherin, and Vimentin) were down- that the RASSF family proteins could associate with Ras GTPases

1854 Cancer Res; 76(7) April 1, 2016 Cancer Research

RASSF1A Is an Antagonist of RhoA

A HCT116 B HCT116 sublines - + + + Flag-RhoA-G14V 10 pcDNA ** ) -- + - RASSF1A-WT-V5 5 RASSF1A-WT --- + RASSF1A-R269F-V5 8 RASSF1A- D256-277 D Cyclin D1 RASSF1A- 69-82 * Cyclin E 6 p21WAF1 4 * p27KIP1 RhoA-G14V (Flag) 2 Number of cell (× 10 RASSF1A (V5) 0 Tubulin 0 1 2 3 4 Days after seeding C HCT116 D - + + + RhoA-G14V HCT116 -- WT R269F RASSF1A RASSF1A-WT RASSF1A-∆256-277 - - +++ ++++++ Etoposide (48 h) --- + ++++++ RhoA-G14V Cleaved PARP mol/L, 48 h) m mol/L, 48 RhoA (Flag) + RASSF1A (V5) 5-FU (50 Tubulin E - + + + RhoA-G14V -- WT R269F RASSF1A HCT116 pcDNA F RhoA-G14V 0h 30 ** ** 25 ** HCT116 18 h 20 15 HCT116 G 10

RASSF1A-WT-V5 (%) Invasion RASSF1A-R269F-V5 5 - + + + + - + + + + Flag-RhoA-G14V 0 MMP-9 - WT R269FL266G RhoA-G14V (Flag) RASSF1A RASSF1A (V5)

Tubulin DU145 HeLa I H TGFb1 (2 ng/mL, 48 h) WT-RASSF1A - - siRASSF1A + + - + ++ + - + + ++ TGFb1 (48 h) E-cadherin ZO-1 Cytokeratin 18 Fibronectin N-cadherin Vimentin

RASSF1A GTP-RhoA

R269F R269F WT Control RASSF1A Tubulin DU145-GFP A549

Figure 5. Tumor suppression function of RASSF1A stems from its RhoA-inhibiting property. A, RASSF1A inhibition of RhoA regulation of cyclins and CDK inhibitors. B, loss of growth-inhibitory activity of RASSF1A by deletion of RhoA- or Smurf1-binding domain. Data represent the mean SD (n ¼ 3 experimental replicates; , P < 0.05; , P < 0.01, Student t test). C and D, Annexin V and cleaved PARP assay showing no apoptotic activity of RASSF1A-R269F and RASSF1A-D256-277. E, RASSF1A inhibition of RhoA-induced cell migration. F, Matrigel assay showing blockade of RhoA-driven tumor cell invasion by WT-RASSF1A but not by RhoA binding-deﬁcient mutants (R269F and L266G). G, inhibition of RhoA-mediated MMP-9 induction by WT-RASSF1A but not by RASSF1A-R269F. H and I, microscopic and immunoblot assays of RASSF1A suppression of TGFb1-induced EMT.

www.aacrjournals.org Cancer Res; 76(7) April 1, 2016 1855

Lee et al.

A B C HeLa HCT116 (RhoA-G14V)

) shControl shControl shRASSF1A 3 HeLa pcDNA RASSF1A-WT 12 shRASSF1A 10

8 ** 6 RASSF1A -D256-277 -D69-82 GTP-RhoA 4 RhoA 2

Smurf1 Tumor volume (×100 mm 0 4 8 12 16 20 24 (d) Tubulin

D HCT116 (RhoA-G14V) E DU145 pcDNA shControl shSmurf1

) WT 3 12 D256-277 RASSF1A pcDNA RASSF1A pcDNA RASSF1A 10 D69-82 8

6 ** 4 2 RASSF1A Smurf1 Tumor volume (×100 mm 0 4 8 12 16 20 24 (d) GTP-RhoA RhoA shControl-pcDNA Tubulin

) shSmurf1-pcDNA

F 3 12 shControl-RASSF1A shSmurf1-RASSF1A 10 G A549 8 RASSF1A D D 6 WT 256-277 69-82 4 ** 2

Tumor volume (×100 mm 0 0 5 10 15 20 25 30 35 40 (d) RhoA -G12V -G12V pcDNA H 20 pcDNA RhoA-G14V I 15 15 ** ** 10 10 5 5 Lung nodules Lung nodules 0 0 - + + - + + RhoA-G14V pcDNA WT D256-277 69-82 -- + -- + RASSF1A RASSF1A shControl shSmurf1

Figure 6. RASSF1A suppresses RhoA-driven tumor growth and colonization in vivo. A and B, mouse tumor xenograft showing RASSF1A depletion effect on tumor growth. GTP-RhoA levels in shRASSF1A and shControl tumors were compared by immunoblot assay. Data represent the mean SD (n ¼ 5pergroup; , P < 0.01). C and D, comparison of suppression effect of WT, D256-277 and D69-82 RASSF1A on RhoA-driven tumor growth. E and F, a Smurf1 dependency of RASSF1A effect on tumor growth. G and H, RASSF1A suppression of RhoA-driven tumor cell colonization in the lungs. The NOD/SCID mice were injected intravenously with A549 subline cells expressing RhoA-G14V and RASSF1A (WT, D256-277 or D69-82) and metastatic nodules in the lungs were counted after 20 days. Data represent the mean SD (n ¼ 5 per group; , P < 0.01). I, a Smurf1 dependency of RASSF1A inhibition of RhoA-driven tumor cell colonization in the lungs.

1856 Cancer Res; 76(7) April 1, 2016 Cancer Research

RASSF1A Is an Antagonist of RhoA

ABN1 T1 T2 T3 T4 T5 T6 T7 Cell lines (N = 47) Colon CA (N = 20) GTP-RhoA 2.0 r 2 = 0.745 r 2 = 0.904 P < 0.01 P < 0.01 RASSF1A 1.5 Colon Tubulin 1.0 RhoA level

GTP-RhoA -

P 0.5 RASSF1A GT 0

Stomach Tubulin Stomach CA (N = 20) Bladder CA (N = 20) GTP-RhoA 2.0 r 2 = 0.736 r 2 = 0.711 P P RASSF1A < 0.01 < 0.01 level 1.5

Bladder Tubulin 1.0 RhoA -

C Squamous Lymph node P 0.5

Normal lung cell carcinoma metastasis GT 0 0 0.5 1.0 0 0.5 1.0 RASSF1A protein level

Normal Primary tumor Metastatic tumor Colon (N = 59) Lung (N = 59) Breast (N = 59) r2 r2 r 2 5 = 0.574 = 0.619 = 0.409 RASSF1A RASSF1A RhoA P < 0.01 P < 0.01 P < 0.01

Lung 4 Normal rectum Adenocarcinoma metastasis 3 2 RhoA level 1

0 0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5 RASSF1A level (Immunostaining) RASSF1A RASSF1A RhoA

E Growth factors (EGF, TGFb1 etc) GDP RhoA Cell proliferation Antiapoptosis Rho-GAP Rho-GEF Actin reorganization Rhotekin S100A EMT GTP RhoA GTP RhoA Invasion Metastasis

RASSF1A RASSF1A RhoA-driven GTP RhoA Smurf1 Smurf1 tumor progression Ub Ub

Figure 7. Inverse correlation of RASSF1A and RhoA expression in human tumor tissues. A, expression of RASSF1A and GTP-RhoA in tumor tissues. N, normal tissue; T, tumors. B, inverse correlation of RASSF1A and GTP-RhoA expression in cancer cell lines and primary tumor tissues. CA, cancer. r, Pearson correlation coefﬁcient. C, immunohistochemical analysis of RASSF1A and RhoA in tumor and matched normal tissues. D, inverse correlation of RASSF1A and RhoA immunoreactivity in colon, lung, and breast tissues. E, schematic representation of RASSF1A suppression of RhoA and its implication in RhoA-driven tumorigenesis. RASSF1A binds directly to RhoA and Smurf1 and stimulates Smurf1 interaction with and ubiquitination of GTP-RhoA. RASSF1A suppresses RhoA-driven tumor cell growth, EMT, invasion, and metastasis, identifying RASSF1A as a novel antagonist of RhoA.

www.aacrjournals.org Cancer Res; 76(7) April 1, 2016 1857

Lee et al.

and play a role as effector molecules in the Ras signaling path- andIIthatarealsoknowntobinddirectlytoRhotekinand ways (1). Studies showed that several RASSF family members, Rhophilin (26, 34). This suggests that in addition to promoting including RASSF2, 4, 5, and 6, interact with the active forms Smurf1-mediated RhoA ubiquitination, RASSF1A could sup- of H- or K-Ras (28, 29). Although RASSF1A was reported to form press RhoA signaling by hindering effector binding and acti- a complex with activated K-Ras, mediate oncogenic K-Ras– vation of RhoA. Indeed, our results showed that RASSF1A dependent apoptosis, and block Ras-induced genomic inst- blocks RhoA signaling by disrupting Rhotekin interaction with ability, it is still controversial that RASSF1A associates directly RhoA. Therefore, RASSF1A suppresses RhoA through Smurf1- withRasandtheRAdomainisrequiredforitsRaseffector dependent and Smurf1-independent mechanisms. function (15, 30). A structural study indicated that RASSF1A RhoA plays essential roles in the regulation of various cellu- does not bind to classical Ras GTPases and the Ras-RASSF1A lar processes, in particular, actin cytoskeleton dynamics, cell- associationismuchweakerthantheRas-RASSF5Aassociation, extracellular matrix adhesion, and cell migration, which are con- raising the possibility that Ras association with RASSF1A may nected directly to invasion and metastasis of tumor cells (35). be weak and/or mediated indirectly through heterodimeriza- RASSF1A binds to tubulins and controls the microtubule dynam- tion with RASSF5A (20,31). On the basis of these observations, ics and loss of this property is associated with reduced ability to it was proposed that the RA domain of RASSF1A may associate promote cell-cycle arrest and suppress cell motility (15, 36). with a yet unknown small GTPases. In the current study, we Although accumulating evidences support that RASSF1A associa- found that RASSF1A binds to RhoA GTPase through the amino tion with microtubules is intimately related with its tumor sup- acid 256-277 region within the RA domain. Moreover, RASSF1A pression function, the molecular basis for this activity remains was identified to bind specifically to a GTP-bound active largely undefined (14, 16). In the current study, we observed that RhoA but not interact with other Rho isoforms RhoB and RhoC RASSF1A is colocalized with RhoA and Smurf1 at the protruding and other Rho GTPase members Rac1 and Cdc42. RASSF1A was edge of cells and blocks RhoA-induced actin rearrangement and also shown to bind to Smurf1 ubquitin E3 ligase, promotes lamellipodia and protrusion formation. It was also shown that Smurf1 interaction with RhoA, and thus stimulates Smurf1- RASSF1A suppresses RhoA-induced tumor cell migration, inva- mediated RhoA ubiquitination. An important role for RASSF1A sion, and in vivo growth and colonization and this effect is highly in the formation of the Smurf1–RhoA complex in RhoA deg- relied on its property to bind RhoA and Smurf1. Considering that radation was supported by the observation that mutant Smurf1 regulates cell polarity and protrusion formation by target- RASSF1A lacking either Smurf1- or RhoA-binding motif fails ing local RhoA for degradation, our data support that RASSF1A to destabilize RhoA. Furthermore, our study showed that high controls directly RhoA-mediated membrane movements and its GTP-RhoA level is closely associated with low RASSF1A level microtubule stabilizing is provoked, at least in part, via a complex in multiple human cancer cells and tumor tissues, supporting formation with RhoA and Smurf1. It is also noticeable that a that RhoA-driven tumor progression is attributed partially to RASSF1A mutant lacking the RhoA-binding motif loses the micro- RASSF1A inactivation. tubule localization and is redistributed within the nucleus, sup- RASSF1C shares amino acids 129-340 with RASSF1A. How- porting that RASSF1A microtubule localization is linked to its ever, we observed that RASSF1C does not bind to RhoA despite tumor suppression function. RhoA activation is critical for onco- the presence of the RA domain. RASSF1C has no SH3 binding genic EMT, particularly during early steps of TGFb1-induced EMT and C1 domains encoded by exons 1a and 2ab in RASSF1A, but (37). In agreement with these, we found that RASSF1A suppresses instead carries a unique N-terminal domain (amino acids 1-49) TGFb1-induced EMT via its RhoA-antagonizing activity. Although encoded by alternative exon 2g (1). A NMR study revealed that the physiologic consequence of the RASSF1A–RhoA–Smurf1 inter- RASSF5 has an intramolecular association between the C1 action is only beginning to be determined, the evidence we domain and the RA domain, leading to the conjecture that the obtained here demonstrates that RASSF1A has the prominent C1 domain-dependent conformation of RASSF1A, which has roles to modulate and integrate a series of critical cellular processes 60% amino acid identity to RASSF5, may play a role in its RA involved in tumor invasion and metastasis progression, including domain-mediated interaction with RhoA (32). However, our EMT. study showed that a mutant RASSF1A lacking the C1 domain In conclusion, our study establishes that RASSF1A antagonizes normally binds to RhoA, excluding the role of the C1 domain in RhoA oncogenic activity by GTP-RhoA destabilization. RASSF1A the interaction with RhoA. It is thus conceivable that the unique appears to evoke this key function by interacting directly with exon 2g-encoding N-terminal region of RASSF1C may have an RhoA and Smurf1, leading to the formation of a RhoA-destruct- inhibitory role for the RA domain association with RhoA. ing protein complex. RASSF1A thus represents one critical neg- Indeed, we found that a RASSF1C mutant lacking exon 2g ative regulator of RhoA signaling, adding a new mechanism by interacts with RhoA. Although further studies are required, our which RASSF1A functions as a tumor suppressor. data supports that the N-terminal variability of RASSF1 proteins may play a key role in determining isoform-specificfunctions, Disclosure of Potential Conflicts of Interest including the specificity of interacting proteins. No potential conflicts of interest were disclosed. RhoA signaling is controlled by multiple effector molecules that interact directly with RhoA through a conserved leucine Authors' Contributions zipper-like motif (26, 33). In this study, we identified that the Conception and design: M.-G. Lee, I.-Y. Kim, S.-G. Chi RhoA-binding region (amino acids 256-277) of RASSF1A has a Development of methodology: M.-G. Lee, K.-P. Ko high degree of amino acid similarity to the RhoA-binding motif Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.-G. Lee, S.-G. Chi of several Rho effectors, including Rhotekin, RTKN, Rhophilin, Analysis and interpretation of data (e.g., statistical analysis, biostatistics, and PKN. Moreover, RASSF1A interacts with RhoA N-terminal computational analysis): M.-G. Lee, S.-G. Chi region (amino acids 1-97), which comprises the switch region I Writing, review, and/or revision of the manuscript: M.-G. Lee, S.-G. Chi

1858 Cancer Res; 76(7) April 1, 2016 Cancer Research

RASSF1A Is an Antagonist of RhoA

Administrative, technical, or material support (i.e., reporting or orga- Grant Support nizing data, constructing databases): M.-G. Lee, S.-I. Jeong, S.-K. Park, This work was supported by the Korean Health Technology R&D Project B.-K.Ryo,S.-G.Chi (HI12C1277) and National Research Foundation of Korea (NRF- Study supervision: M.-G. Lee, J.-K. Kim, S.-G. Chi 2015R1A2A1A01005389). Acknowledgments The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked The authors thank Dr. Kozo Kaibuchi (Nagoya University, Nagoya, Japan) advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate for Myc-ROCK, Dr. Aafke Arianens (The Netherlands Cancer Institute, this fact. Amsterdam, the Netherlands) for Myc-p190RhoGEF, Dr. Takeshi Imamura (The JFCR Cancer Institute) for Smurf1 plasmids, Dr. Koh-ichi Nagata (Aichi Human Service Center, Japan) for Rhotekin plasmid, and Dr. Michiyuki Received June 30, 2015; revised December 19, 2015; accepted January 16, Matsuda (Tokyo University, Japan) for Raichu-Rhotekin-RBD sensor. 2016; published OnlineFirst January 29, 2016.

References 1. Dammann R, Li C, Yoon JH, Chin PL, Bates S, Pfeifer GP. Epigenetic 20. Gordon M, Baksh S. RASSF1A: not a prototypical Ras effector. Small inactivation of a RAS association domain family protein from the lung GTPases 2011;2:148–57. tumour suppressor locus 3p21.3. Nat Genet 2000;25:315–9. 21. Lee MG, Han J, Jeong SI, Her NG, Lee JH, Ha TK, et al. XAF1 directs 2. van der Weyden L, Adams DJ. The Ras-association domain family (RASSF) apoptotic switch of p53 signaling through activation of HIPK2 and members and their role in human tumourigenesis. Biochem Biophys Acta ZNF313. Proc Natl Acad Sci U S A 2014;111:15532–7. 2007;1776:58–85. 22. Han J, Kim YL, Lee KW, Her NG, Ha TK, Yoon S, et al. ZNF313 is a novel cell 3. Agathanggelou A, Cooper WN, Latif F. Role of the Ras-association domain cycle activator with an E3 ligase activity inhibiting cellular senescence by family 1 tumor suppressor gene in human cancers. Cancer Res 2005;65: destabilizing p21WAF1. Cell Death Differ 2013;20:1055–67. 3497–508. 23. Yoshizaki H, Ohba Y, Kurokawa K, Itoh RE, Nakamura T, Mochizuki N, 4. Foley CJ, Freedman H, Choo SL, Onyskiw C, Fu NY, Yu VC, et al. Dynamics et al. Activity of Rho-family GTPases during cell division as visualized with of RASSF1A/MOAP-1 association with death receptors. Mol Cell Biol FRET-based probes. J Cell Biol 2003;162:223–32. 2008;28:4520–35. 24. Rossman KL, Der CJ, Sondek J. GEF means go: turning on RHO GTPases 5. Hamilton G, Yee KS, Scrace S, O'Neill E. ATM regulates a RASSF1A- with guanine nucleotide-exchange factors. Nat Rev Mol Cell Biol 2005; dependent DNA damage response. Curr Biol 2009;19:2020–5. 6:167–80. 6. Burbee DG, Forgacs E, Zochbauer-Muller S, Shivakumar L, Fong K, Gao B, 25. Sahai E, Garcia-Medina R, Pouyssegur J, Vial E. Smurf1 regulates tumor cell et al. Epigenetic inactivation of RASSF1A in lung and breast cancers and plasticity and motility through degradation of RhoA leading to localized malignant phenotype suppression. J Natl Cancer Inst 2001;93:691–9. inhibition of contractility. J Cell Biol 2007;176:35–42. 7. Byun DS, Lee MG, Chae KS, Ryu BG, Chi SG. Frequent epigenetic inacti- 26. Blumenstein L, Ahmadian MR. Models of the cooperative mechanism for vation of RASSF1A by aberrant promoter hypermethylation in human Rho effector recognition: implications for RhoA-mediated effector activa- gastric adenocarcinoma. Cancer Res 2001;61:7034–8. tion. J Biol Chem 2004;279:53419–26. 8. Praskova M, Khoklatchev A, Ortiz-Vega S, Avruch J. Regulation of the MST1 27. Chen M, Bresnick AR, O'Connor KL. Coupling S100A4 to Rhotekin kinase by autophosphorylation, by the growth inhibitory proteins, RASSF1 alters Rho signaling output in breast cancer cells. Oncogene 2013;32: and NORE1, and by Ras. Biochem J 2004;381:453–62. 3754–64. 9. Baksh S, Tommasi S, Fenton S, Yu VC, Martins LM, Pfeifer GP, et al. The 28. Park J, Kang SI, Lee SY, Zhang XF, Kim MS, Beers LF, et al. Tumor suppressor tumor suppressor RASSF1A and MAP-1 link death receptor signaling to bax ras association domain family 5 (RASSF5/NORE1) mediates death recep- conformational change and cell death. Mol Cell 2005;18:637–50. tor ligand-induced apoptosis. J Biol Chem 2010;285:35029–38 10. Vos MD, Dallol A, Eckfeld K, Allen NP, Donninger H, Hesson LB, et al. The 29. Ikeda M, Hirabayashi S, Fujiwara N, Mori H, Kawata A, Iida J, et al. Ras- RASSF1A tumor suppressor activates Bax via MOAP-1. J Biol Chem association domain family protein 6 induces apoptosis via both caspase- 2006;281:4557–63. dependent and caspase-independent pathways. Exp Cell Res 2007;313: 11. Shivakumar L, Minna J, Sakamaki T, Pestell R, White MA. The RASSF1A 1484–95. tumor suppressor blocks cell cycle progression and inhibits cyclin D1 30. Donninger H, Vos MD, Clark GJ. The RASSF1A tumor suppressor. J Cell Sci accumulation. Mol Cell Biol 2002;22:4309–18. 2007;120:3163–72. 12. Song MS, Song SJ, Kim SY, Oh HJ, Lim DS. The tumour suppressor 31. Stieglitz B, Bee C, Schwarz D, Yildiz O, Moshnikova A, Khokhlatchev A, RASSF1A promotes MDM2 self-ubiquitination by disrupting the et al. Novel type of Ras effector interaction established between tumour MDM2-DAXX-HAUSP complex. EMBO J 2008;27:1863–74. suppressor NORE1A and Ras switch II. EMBO J 2008;27:1995–2005. 13. Song MS, Song SJ, Ayad NG, Chang JS, Lee JH, Hong HK, et al. The tumour 32. Harjes E, Harjes S, Wohlgemuth S, Muller€ KH, Krieger E, Herrmann C, et al. suppressor RASSF1A regulates mitosis by inhibiting the APC-Cdc20 com- GTP-Ras disrupts the intramolecular complex of C1 and RA domains of plex. Nat Cell Biol 2004;6:129–37. Nore1. Structure 2006;14:881–8. 14. Liu LM, Tommasi S, Lee DH, Dammann R, Pfeifer GP. Control of micro- 33. Shibata H, Mukai H, Inagaki Y, Homma Y, Kimura K, Kaibuchi K, et al. tubule stability by the RASSF1A tumor suppressor. Oncogene 2003;22: Characterization of the interaction between RhoA and the amino-terminal 8125–36. region of PKN. FEBS Lett 1996;385:221–4. 15. Vos MD, Martinez A, Elam C, Dallol A, Taylor BJ, Latif F, et al. A role for the 34. Ihara K, Muraguchi S, Kato M, Shimizu T, Shirakawa M, Kuroda S, et al. RASSF1A tumor suppressor in the regulation of tubulin polymerization Crystal structure of human RhoA in a dominantly active form complexed and genomic stability. Cancer Res 2004;64:4244–50. with a GTP analogue. J Biol Chem 1998;273:9656–66. 16. El-Kalla M, Onyskiw C, Baksh S. Functional importance of RASSF1A micro- 35. Timpson P, McGhee EJ, Morton JP, von Kriegsheim A, Schwarz JP, Karim tubule localization and polymorphisms. Oncogene 2010;29:5729–40. SA, et al. Spatial regulation of RhoA activity during pancreatic cancer cell 17. Ortiz-Vega S, Khokhlatchev A, Nedwidek M, Zhang XF, Dammann R, invasion driven by mutant p53. Cancer Res 2011;71:747–57. Pfeifer GP, et al. The putative tumor suppressor RASSF1A homodimerizes 36. Dallol A, Agathanggelou A, Tommasi S, Pfeifer GP, Maher ER, Latif F. and heterodimerizes with the Ras-GTP binding protein Nore1. Oncogene Involvement of the RASSF1A tumor suppressor gene in controlling cell 2002;21:1381–90. migration. Cancer Res 2005;65:7653–9. 18. Avruch J, Xavier R, Bardeesy N, Zhang XF, Praskova M, Zhou D, et al. Rassf 37. Bhowmick NA, Ghiassi M, Bakin A, Aakre M, Lundquist CA, Engel ME, et al. family of tumor suppressor polypeptides. J Biol Chem 2009;284:11001–5. Transforming growth factor-beta1 mediates epithelial to mesenchymal 19. Vos MD, Clark GJ. RASSF family proteins and Ras transformation. Methods transdifferentiation through a RhoA-dependent mechanism. Mol Biol Cell Enzymol 2006;407:311–22. 2001;12:27–36.

www.aacrjournals.org Cancer Res; 76(7) April 1, 2016 1859

RASSF1A Directly Antagonizes RhoA Activity through the Assembly of a Smurf1-Mediated Destruction Complex to Suppress Tumorigenesis

Min-Goo Lee, Seong-In Jeong, Kyung-Phil Ko, et al.

Cancer Res 2016;76:1847-1859. Published OnlineFirst January 29, 2016.

Updated version Access the most recent version of this article at: doi:10.1158/0008-5472.CAN-15-1752

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2016/01/29/0008-5472.CAN-15-1752.DC1

Cited articles This article cites 37 articles, 20 of which you can access for free at: http://cancerres.aacrjournals.org/content/76/7/1847.full#ref-list-1

Citing articles This article has been cited by 7 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/76/7/1847.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at Subscriptions [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/76/7/1847. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.