RASSF1A Suppresses the Invasion and Metastatic Potential of Human Non–Small Cell Lung Cancer Cells by Inhibiting YAP Activation through the GEF-H1/RhoB Pathway

One of the key events in the natural history of lung carcinomas is the early inactivation of the tumor suppressor gene RASSF1 [Ras association (RalGDS/AF-6) domain family member 1], in 30% of non–small cell lung cancer (NSCLC), by promoter hypermethylation (1). The RASSF1A protein, encoded by 1 of 8 splicing isoforms, termed 1A to 1H (2), is expressed in all normal human tissues, and carries several domains mediating protein–protein interactions with multiple partners, making RASSF1A a key regulator of apoptosis, cell-cycle progression, and genome integrity (2). It was also suggested that RASSF1A further plays a role in tumor cell adhesion and motility (3), though the molecular mechanisms are poorly understood. In Drosophila melanogaster, the dRASSF protein is thought to interfere with the Hippo pathway, composed of the Ste20-like Hippo kinase, (fly ortholog of human MST1/2), the AGC kinase Warts (Drosophila ortholog of human Lats1/2), and the terminal target Yorkie, a transcriptional cofactor (Drosophila ortholog of Yes-associated protein, YAP; ref. 4). In the nucleus of mammal cells, active YAP interacts with TEAD transcription factors to regulate transcription of multiple genes, some of which have been shown to mediate epithelial-to-mesenchymal transition (EMT; ref. 5). YAP was recently reported to transduce mechanical signals from extracellular matrix (ECM), which are mediated by a Rho GTPase activity, and tension of actin–myosin cytoskeleton, independently of Hippo signaling (6). The Rho (A, B, and C) small GTP/GDP–binding proteins, are indeed well-known master regulators of cell adhesion and motility (7). Their activation state is highly regulated by guanine nucleotide exchange factors (GEF), which trigger the exchange of GDP for GTP (8). Once activated, RhoA, RhoC proteins mediate cell motility and tumor invasion, whereas RhoB is rather a migration-inhibiting protein, counteracting the action other Rho proteins (9). Rho proteins were reported to influence YAP subcellular localization, but the links connecting RASSF1A, YAP, and Rho signaling network remain elusive in human cells (5).

We previously reported that patients with stage I–II NSCLC and RASSF1A promoter gene methylation displayed a 3-fold decrease in 5-year overall survival, suggesting that RASSF1A inactivation could not only be involved in initial bronchial carcinogenesis, but also in the acquisition of a metastatic phenotype, as patients ultimately die from metastasis spread (10). Here, we sought to decipher this putative RASSF1A specific role in the acquisition of invasiveness, by using RNAi to mimic RASSF1A loss of function in lung cancer cell lines, as in nontumorigenic, untransformed HBEC-3 bronchial cell lines (11). We found that RASSF1A knockdown, by inducing EMT and increasing cell motility and invasion capacities, favors in vivo lung metastasis process. Mechanistically, we reveal a connection between RASSF1A depletion, RhoB downregulation, and YAP activation, leading to the acquisition of tumorigenic and invasive hallmarks.

Isogenic HBEC-3 and HBEC-3-KRasV12 bronchial cells (Dr. White, UT Southwestern Medical Center, Dallas, TX; ref. 12), tumorigenic epithelial cell lines BEAS-2B, cancer-derived cell lines A549 and H1975 from ATCC, and BEAS-2B-KRasV12 from Dr. Pradines (9, 13) were grown in supplemented appropriate media (Supplementary Table S1), transfected at 30% confluence, using Lipofectamine RNAiMAX (Invitrogen) with siRNA, plasmid DNA, or control mimics (Dharmacon; Supplementary Tables S2 and S3). HBEC-3 and HBEC-3-KRasG12V were authenticated using standard karyotyping techniques as previously described (12). A549, H1975, BEAS-2B, were passaged for fewer than 6 months after receipt from ATCC. BEAS-2B cells transformed by KRasV12 by Dr. Pradines (UMR INSERM 1037-CRCT, Toulouse, France) were obtained by transfection with the pZip–KRasV12 plasmid (9). KRas–mutant allele in HBEC-3-KRasV12 or BEAS2B-KRasV12 cells was checked by using a RT-PCR/RFLP method as described previously (12).

After extraction, RT-PCR was done with each primer sets (Supplementary Table S4) as described previously (12). RT-PCR data were normalized to the human S16. Relative quantification was calculated using the ΔΔC_t method.

Antibodies against vimentin; ZO-1; β-catenin; Twist; Lats1; phospho-cofilin(Ser3); cofilin; phospho-LIMK(Thr508)/LIMK2(Thr505); LIMK1; PDXP; ROCK1; Yap; RhoA; RhoB; GEFH1; phospho-GEFH1 (Ser885); calpain; PARP; phospho-PP1α(Thr320; Cell Signaling Technology); PP2ACalpha (Millipore); and actin (Sigma-Aldrich), were used at 1:1,000 dilution. Whole cell protein extracts were prepared as described previously (14), and proteins detected by immunoblotting by enhanced chemiluminescence with ECL kit (Promega).

Immunofluorescence studies used E-cadherin and Yap primary antibodies from Cell Signaling Technology, syndecan-1 and syndecan-4 from R&D, CD44v6 from Santa Cruz Biotechnology, N-cadherin, and RASSF1A from eBioscience. The AlexaFluo555 and 486-labeled (Invitrogen) secondary antibodies were added for 1 hour. Coverslips were mounted with DAPI (Santa Cruz Biotechnology), and image captured with high-throughput confocal microscopy (FluoView FV1000, Olympus).

Transfected cells grown onto 24-well collagen IV or fibronectin-coated plates (BD Biocoat Matrigel Invasion Chamber) were pretreated with mitomycin C (1 μg/mL) 12 hours before an artificial “wound” created at 0 hour. Photographs were taken (×10) at 0 hour and 6 hours. The distances subtracted across the wound at 0 hour and 6 hours were expressed as μm/hour.

A total of 25 × 10³ cells in serum-free medium were added to the top invasion chambers of 24-well transwell plates containing cell culture insert with 8 μm pore size (BD BioCoat Matrigel Invasion Chamber, BD Biosciences). Complete media were added to the bottom chambers. Cells were incubated 24 hours (migration assay) to 48 hours (invasion assay). Top (nonmigrating) cells were removed, bottom (migrating) cells were stained with crystal violet. Assays were performed in triplicate, and data presented as the average number of migrating/invading cells.

A total of 2 × 10⁴ transfected cells were seeded in 96-well plates coated with collagen IV or fibronectin (BD Biosciences). After 1 hour incubation at 37°C, cells were washed and stained with crystal violet (30 minutes), washed again, and OD read at 590 nm.

Base agar matrix (50 μL, Cell Biolabs) was seeded in 96-well plate and 3,000 cells/well layered on agar followed by 50 μL of 2× complete medium. Colonies were counted for each well 21 days later.

Cells were lysed in chilled immunoprecipitation buffer and the cleared lysate (500 μg) incubated with 3 μg of the indicated antibody and 30 μL of protein-A agarose beads (Repligen) in 1 mL of IP buffer (13). Beads were resuspended in 30 μL of 2× Laemmli buffer and subjected to Western blotting. For GTP-Rho pulldown assays, cell lysates were incubated with beads glutathione-S-transferase (GST)-Rhotekin Rho binding domain (RBD) 45 minutes at 4°C and precipitates analyzed by Western blotting using RhoB or RhoA antibodies.

The experiments were performed according to the European Convention for the Protection of Vertebrates Used for Scientific Purposes. Groups of 7, strain 250, 6-week-old, female Fox Chase SCID^−/− Beige mice from Charles River Laboratory were injected subcutaneously with ShRASSF1A-777- or ShRASSF1A-779–infected HBEC-3 or H1975 cell suspension (1 × 10⁷ cells/0.1 mL) in the right flank of each animal (see Supplementary Table S3 for sh sequences). Mice were monitored for tumor growth twice a week and tumors were allowed to grow to 1,000 mm³ before euthanasia of the mice. The postmortem examination included macroscopic description of lungs and liver. Tumor xenografts and lungs were rapidly removed and dissected into two parts: one fragment was snap-frozen in nitrogen for RNA extraction and qRT-PCR, the other fragment fixed in PFA 4% for histologic analysis. Lung-specific tetracycline-inducible human EGFRL858R bitransgenic mice (CCSP-rtTA; tetOEGFRL858R) crossed into rhob null (rhob^−/−), and wild-type (rhob^+/+) strains, have been described previously (14). Five- to 6-week-old mice were fed ad libitum with food pellets that contained doxycycline (1 g/kg) for 8 weeks, and then euthanized, lung excised, and paraffin-embedded. Paraffin sections (4 μm) served for hematoxylin and eosin staining and YAP IHC, using standard procedures. Ten tumor lesions from each of the 5 mice with each genetic status were examined for YAP nuclear staining.

Data are means ± SEM of experiments realized three times independently. Statistical differences were determined either by the Student paired t test for single comparison or by one-way ANOVA followed by Dunnett multiple comparison test to compare each conditions of an experiment with a single control (siNeg; GraphPad Software, Inc.). Statistical significance was set at P ≤ 0.05.

RASSF1A was targeted in human bronchial epithelial cell (HBEC) lines using two different siRNAs (siRASSF1A and siRASSF1AC; Fig. 1A and B, siRNA-targeting regions on RASSF1 gene shown in Supplementary Fig. S1). Noteworthy, siRASSF1AC also inhibited RASSF1C mRNA expression (Fig. 1C). First, we surprisingly observed 5-fold more giant cells (P < 0.05) in RASSF1A-depleted cells than in cells treated with a control siRNA (siNeg), and many cells presented long cytoplasmic extensions (Fig. 1D, RASSF1A and AC), as observed in motile cells (15). In an air-liquid interface (ALI; Fig. 1E, a and b), HBEC-3 cells established a single layer of columnar cells and exhibited both intercellular tight and dense adherens junctions (Fig. 1E, c, siNeg). Conversely, in this condition, the majority of RASSF1A-knockdown cells (RASSF1A depletion efficacy shown by IHC in Fig. 1E, d) formed a monolayer of flat cells with small, scarce, and disorganized adherens junctions (Fig. 1E, c), resembling squamous metaplasia cells, encountered in preneoplastic bronchial lesions (16). A fraction of RASSF1A-deficient cells also exhibited focal features of hyperplasia (Fig. 1E, a and b; siRASSF1A and AC) with several layers of enlarged elongated cells, and focal features of dysplasia, such as a significant increase of binucleated cells (6-fold vs. normal HBEC-3 cells, P < 0.01, Fig. 1D). Electronic microscopy, with quantification performed in three independent ALI cultures, with 50 to 70 cells for each condition (Supplementary Fig. S1), confirmed that RASSF1A-depleted cells exhibited a complete loss of tight junctions, a significant decrease in dense adherens junctions, replaced by more diffuse adherens structures (Fig. 1E, c, SiRASSF1A and AC).

We examined the expression of several EMT markers in bronchial cell lines, after RASSF1A depletion (Fig. 2). In RASSF1A-depleted cells, several epithelial markers such as E-cadherin (Fig. 2A and E), miR-200b/miR-200c (Fig. 2B and C), syndecan-1 (Fig. 2F), and ZO-1 (Fig. 2J), decreased, whereas various mesenchymal markers, such as miR-21 (Fig. 2D), N-cadherin (Fig. 2G), syndecan-4 (Fig. 2H), CD44v6 (Fig. 2I), Twist1 (Fig. 2K), β-catenin (Fig. 2L), and vimentin (Fig. 2M) increased.

To ensure that the EMT induced by RASSF1A knockdown was not the result of a RNAi off-target effect, HBEC-3 cells were cotransfected with si-RASSF1A and a siRNA-resistant RASSF1A-encoding expression plasmid. Significantly, restoration of RASSF1A expression (Supplementary Fig. S3A) in RASSF1A-knockdown HBEC-3 cells decreased vimentin and Twist1 reexpression (Supplementary Fig. S3B).

EMT enhances tumor cell migration and invasion (16). To test whether RASSF1A deficiency induced an invasive phenotype in HBEC-3 cells, we measured the expression of 84 genes involved in invasiveness (Supplementary Table S5). There was a significant reduction (>2-fold) in the expression of MMP9, mDia1, and RhoB, whereas MMP2, MMP14, RhoA, RhoC, RhoE, cofilin, and integrinβ2 mRNAs increased (>2-fold) in RASSF1A-depleted HBEC-3 cells. RASSF1A depletion could thus promote the acquisition of motile and invasive properties, by altering the Rho-family proteins expression, actin dynamic regulators, a cell-adhesion protein (ITGβ2), or matrix-metalloproteases. Accordingly, wound-healing assays revealed that RASSF1A-depleted HBEC-3 cells migrated faster than controls (Fig. 3A). Similar results were observed when RASSF1A was reduced in several other untransformed and transformed bronchial cell lines (Supplementary Fig. S4), namely, HBEC-3-Rasv12 (12), BEAS-2B-RasV12, or H1975 (Supplementary Table S1). This effect was independent of the adhesion substrate tested such as collagen I (Fig. 3A), collagen IV (Fig. 3A, Supplementary Fig. S4B), fibronectin (Supplementary Fig. S4B). HBEC-3 two-dimensional migration induced by RASS1A knockdown showed features evoking both collective migration and multicellular streaming (Supplementary Movie S1; ref. 16). Cell migration involves rearrangements of adhesion structures, and upon RASSF1A knockdown, talin, paxillin, moesin, and vinculin mRNAs expression levels actually decreased (Supplementary Fig. S5A). Consistently, RASSF1A depletion in HBEC-3 cells reduced adhesion to laminin- (Supplementary Fig. S5B) and fibronectin-coated plates (Supplementary Fig. S5C). Thus, RASSF1A knockdown affects cell adhesion to substrate, cell–cell contacts (Fig. 1), EMT markers (Fig. 2), and cell motility (Fig. 3A). The siRNA impact on cell migration was specific as expression of a siRNA-resistant RASSF1A actually reversed the cell motility to normal levels (Supplementary Fig. S3C). The increased cell motility of RASSF1A-silenced cells translates into increased cell invasion capacity, as ascertained in matrigel invasion (Fig. 3B) and in transendothelial migration assays (Supplementary Fig. S4A). Similar results were observed upon RASSF1A depletion, in other bronchial cell lines (Supplementary Fig. S4B and S4C). The stimulating effect on migration of RASSF1A depletion was not rescued by cotransfection of a RASSF1C-encoding expression plasmid in HBEC-3 cells, which conversely reversed the inhibiting effect on migration obtained by depleting RASSF1C by siRASSF1AC (Supplementary Fig. S3C). Besides, codepletion of both RASSF1A and RASSF1C using siRASSF1AC (Fig. 1C), had variable effects on cell migration, depending on substratum and cell types (Fig. 3, Supplementary Fig. S4). To further assess the differing requirements for RASSF1A and RASSF1C in migration, we used A549 cells with hypermethylated RASSF1A gene and no RASSF1A expression. In A549 cells, restoring RASSF1A expression, or depletion of RASSF1C by siRASSF1AC, decreased cell invasion, whereas overexpression of RASSF1C increased cell migration through Matrigel (Fig. 3D).

We next evaluated anchorage-independent growth capacities of HBEC in soft-agar colony formation assays (Fig. 3C). RASSF1A depletion provided HBEC-3 and HBEC-3 RasV12 cells anchorage-independent growth, whereas control HBEC-3 and HBEC3 RasV12 cells failed to form colonies in soft agar (Fig. 3C and Supplementary Fig. S4D). The anchorage-independent growth of BEAS-2B was also stimulated by RASSF1A depletion (Supplementary Fig. S4D).

We infected tumorigenic EGFR-mutated lung cancer H1975 cells expressing wild-type RASSF1A, and untransformed HBEC-3 cells, with lentivirus expressing shRNAs targeting RASSF1A (17). qRT-PCR (not shown) or Immunofluorescence analysis (Supplementary Fig. S6A) confirmed that RASSF1A expression decreased by 30% to 50% with either shRNA construct. Despite this modest silencing, shRASSF1A cells exhibited significantly increased migration, invasion, and growth in soft agar when compared with control cells (Supplementary Fig. S6B and S6C). When injected in SCID mice, normal and RASSF1A-depleted HBEC-3 cells failed to form tumors after 6 months of observation (not shown), suggesting that RASSF1A depletion alone does not support xenograft growth. Conversely, parental tumorigenic H1975 showed the first signs of tumors at subcutaneous injection sites 12 to 21 days after injection, whereas tumors formed by RASSF1A-depleted H1975 cells emerged as soon as 9 days after injection (Fig. 4A). Tumor growth was significantly accelerated (2.8-fold) upon RASSF1A knockdown compared with controls (Fig. 4A). Using qRT-PCR (not shown) and Western blotting, we confirmed that shRASS1A-infected primary tumors still exhibited a significant decrease of RASSF1A (Fig. 4B). Finally, RASSF1A-depleted H1975 cells formed significantly much more lung metastatic tumor foci than native cells (Fig. 4C and D), confirming that increased cell migration and cell invasion induced by RASSF1A loss, led in vivo to increased metastatic potential.

The transcriptional coactivator YAP is involved in EMT regulation and connected with RASSF1A (18, 19). Indeed, upon RASSF1A depletion, we observed an increased expression of ITGB2, a YAP-transcriptional target (Supplementary Table S5; ref. 20). We further showed that active nuclear YAP was significantly increased upon RASSF1A depletion (>2-fold; Fig. 5A and B) as shown by nuclear/cytoplasmic fractionation experiments (Fig. 5A), immunofluorescence studies (Fig. 5B), and the concomitant upregulation of two other YAP target genes (6), besides ITBG2 (Supplementary Table S5): ANKRD1 and CTGF (Fig. 5C). RASSF1A siRNA transfection also induced YAP nuclear accumulation in ALI HBEC-3 cultures (Fig. 5D). A time-course evaluation of YAP nuclear accumulation upon RASSF1A depletion further showed an increase of nuclear YAP, as soon as 12 hours posttransfection (Supplementary Fig. S7), when 30% decrease in RASSF1A protein content was already detectable (not shown), whereas both nuclear and cytoplasmic YAP were detected from 24 hours to 36 hours posttransfection. Cotransfection of siRNA-resistant RASSF1A (Supplementary Fig. S3D) or Yap-RNAi (Fig. 5B) reverted this phenotype.

The simultaneous silencing of RASSF1A/YAP reverted (i) the EMT-like phenotype due to RASSF1A deficiency [epithelial marker syndecan-1 (Fig. 5E) and mesenchymal vimentin (Fig. 5F) were reverted to normal levels as observed in controls], (ii) the increased cell invasion (Fig. 5G), but failed to prevent colony formation of RASSF1A-depleted HBEC-3 (Fig. 5H). YAP seems thus to be crucial for the EMT-like and cell migration/invasion phenotypes in RASSF1A-depleted cells, but not for anchorage-independent growth of RASSF1A-depleted cells.

The cofilin pathway is a key regulator of actin dynamics (21). Cofilin mRNA expression increased upon RASSF1A depletion (Supplementary Table S5) and cofilin phosphorylation significantly decreased, suggesting RASSF1A depletion increased cofilin activity (Fig. 6A). In support of this finding, active phospho-LIMK, which suppresses cofilin activity, was reduced, contrasting with the increase in global LIMK (Fig. 6A), whereas expression of the cofilin-specific phosphatases PDXP/chronophin (22) increased (Fig. 6A). In addition, expression of Rock1, the kinase activating LIMK (23) was also reduced (Fig. 6A), whereas activation of Lats1 kinase, which is able to inhibit LIMK was increased as mentioned above. To our knowledge, these findings provide the first in vitro evidence that RASSF1A regulates the cofilin pathway.

LIMK controls cytoskeleton dynamics by regulating Rho family GTPases (24). We described (Supplementary Table S5) that mRNA expression of some Rho proteins was either upregulated (RhoA, Rho C), or downregulated (RhoB), upon RASSF1A depletion. We focused on RhoB because of its antimigratory role (9). We first confirmed the decrease of RhoB mRNA upon RASSF1A depletion using different RhoB mRNA primer pairs (Fig. 6B). We then checked whether reduced RhoB mRNA expression upon RASSF1A depletion led to a decrease of active GTP-bound RhoB protein. We actually found a significant decrease of endogenous active RhoB-GTP in RASSF1A-depleted HBEC-3 cells (Fig. 6C), but not of endogenous RhoA-GTP levels (Supplementary Fig. S8A). Expression of siRNA-resistant RASSF1A or constitutively active RhoBV14 restored normal RhoB-GTP levels (Fig. 6C). Significantly, coexpression of siRASSF1A with wild-type RhoB, or constitutively active RhoBV14, strongly reduced HBEC-3 cells invading capacity, whereas the dominant-negative RhoBN19 had no effect (Fig. 6D). Concurrent RASSF1A and RhoB depletion, as RhoB depletion alone, also increased HBEC-3 cells invasion (Supplementary Fig. S9A). Conversely, RhoA wild-type or RhoAV14 failed to inhibit invasion of RASSF1A-depleted HBEC-3 (Supplementary Fig. S8B) and HBEC-3-RasV12 cells (not shown). Interestingly, expression of the RhoAN19 mutant, which can trap Rho-GEFs (including RhoA-specific GEFs), strongly inhibited invasion of RASSF1A-depleted HBEC-3 and HBC3-RasV12 cells (Supplementary Fig. S8B). By using Narciclasine, a Rho activator (25), in RASSF1A-depleted HEBC-3 cells, we revealed that Rho proteins sustained activation impaired cell migration and invasion. As we had showed that active RhoAV14 had no effects on cell migration upon RASSF1A depletion, the Narciclasine effect could then mainly due to RhoB reactivation (Supplementary Fig. S8C). Collectively, our results suggest that RASSF1A depletion specifically alters RhoB-inhibiting functions on migration.

To understand mechanistically how RASSF1A knockdown affected RhoB-GTP levels, we examined three well-defined Rho-GEFs, GEFH1 (ARHGEF2), Vav2, and Ect2, reported to regulate RhoB and cell migration (26–28) negatively (29) or positively (30), depending on cellular context. Cotransfected with siRASSF1A, GEFH1, Vav2, or Ect2 inhibited RASSF1A depletion induced cell invasiveness (Fig. 6E). Similarly, overexpression of mDia-1, an effector of activated RhoB-GTP (31), of which mRNA was found to be decreased upon RASSF1A depletion (Supplementary Table S5), impaired cell invasion induced by RASSF1A depletion (Fig. 6F), further supporting a role for RhoB inactivation in RASSF1A depletion-promoted cell invasion. Given that (i) murine GEFH1 was reported as binding partner of RhoBV14 in a Yeast two-hybrid screen (32), (ii) GEFH1 exhibits a Rho-specific activity in vitro, but is unable to catalyze GDP exchange on Rac (33), and (iii) GEFH1 was identified in dendritic cells as the major protein associated and colocalized with the upregulated RhoB, upon lipopolysaccharide stimulation (29), we focused on GEF-H1.

We found that RASSF1A knockdown increases Ser885-phosphorylation and inactivation of GEFH1 (Fig. 6G). Moreover, similar to RASSF1A depletion (Fig. 6H), GEFH1 knockdown in HBEC-3 cells decreased the activated RhoB-GTP level (Fig. 6H). Conversely, GEFH1 overexpression in RASSF1A-depleted cells increased RhoB-GTP levels (Fig. 6H). Significantly, GEFH1 depletion alone increased the HBEC-3 cells invasion (Fig. 6I), and codepletion of RASSF1A and GEFH1 additively elevated cell invasion (Fig. 6I). Thus, RASSF1A loss abolishes the antiinvasive role of active RhoB-GTP, presumably by inactivating the GEF-H1 activity.

RASSF1A depletion also decreased PP2A activity in HBEC-3 cells (Fig. 6J), whereas PP1A phosphatase amounts remained unchanged (Supplementary Fig. S8E), and we hypothesized that such phosphatase regulation could influence GEFH1 phosphorylation state. Recently, siRNA-mediated depletion of RhoB was reported to downregulate PP2A activity (13). By immunoprecipitating PP2ACa from cells lysates derived from BEAS2B cells expressing RASSF1A, we detected both GEFH1 and RhoB with specific antibodies (Supplementary Fig. S8G), suggesting PP2A forms a complex with both proteins. Depletion of PP2A, RhoB, or RASSF1A in bronchial cells with specific siRNAs, induced an increase of phospho-GEFH1, as detected by immunofluorescence with a specific anti-phospho-Ser-885 GEFH1 antibody, supporting a role for PP2A in controlling GEFH1 activation state upon RASSF1A depletion (Supplementary Fig. S8F). Moreover, by immunoprecipitating GEFH1 in HBEC-3 cells, again we found that PP2ACa and RhoB associated; this association decreased upon RASSF1A depletion (Fig. 6L), supporting a role for RASSF1A in regulating the stochiometry of the complex RhoB/GEFH1/PP2A.

Hence, RASSF1A could exert a basal migration inhibitory effect via PP2A/GEFH1/RhoB regulation of cofilin-regulated F-actin polymerization. Indeed, we checked that overexpression of the constitutively active RhoBV14 increased inactive phospho-cofilin, whereas RASSF1A depletion decreased phospho-cofilin, and DN RhoBN19 had no effect (Fig. 6K). In addition, cotransfection of a specific siRNA targeting cofilin together with siRASSF1A was able to revert YAP cytoplasmic-nuclear shuttling induced by RASSF1A depletion and decreased transcription of YAP transcriptional targets (Supplementary Fig. S8H).

Finally, we wondered whether the effects on YAP and RhoB signaling, upon RASSF1A depletion, were connected. We tested whether manipulation of RhoB levels/activity would alter YAP cytosol to nucleus translocation. Expression of dominant-negative RhoBN19 increased active YAP nuclear accumulation (Fig. 7A). Similar results were obtained with siRNA-RhoB (Supplementary Fig. S9B). Expression of RhoB-WT, RhoBV14 or individual expression of the three RhoB-GEFs (Fig. 7C) suppressed nuclear accumulation of YAP upon RASSF1A knockdown. Expression of constitutively active RhoBV14 mutant, or RhoB wild-type alone, decreased YAP transcriptional activity, and was able to counteract YAP activity increase in RASSF1A-depleted cells, on the transcription of both ANKRD1 and CTGF genes (Fig. 7B). The RhoB influence on YAP nuclear accumulation and activation was the opposite of RhoA effect, as overexpression of RhoAwt or RhoAV14 significantly increased nuclear YAP, whereas DNRhoAN19 plasmid transfection did not significantly alter YAP localization, but drastically reduced nuclear YAP in RASSF1A-depleted cells (Supplementary Fig. S8D). Conversely, YAP depletion could increase RhoB-GTP active form (Fig. 7E), suggesting a specific reciprocal interaction between RhoB and YAP. Significantly, RhoB depletion alone as codepletion of RhoB and RASSF1A also increased cell invasion (Supplementary Fig. S9A). This final set of data suggests that the RASSF1A–GEF-H1–RhoB axis not only controls cell motility and shape, but also regulates the cytoplasmic nuclear shuttling of YAP. We raised the question whether such YAP/RhoB regulation loop can occur in vivo by analyzing YAP nuclear expression in a previously described transgenic mouse model. This model was obtained by crossing rhob wild-type or null mice, with mice prone to develop lung neoplastic changes through an inducible lung-specific EGFRL858R mutation (14). These RhoB-deficient mice develop focal or diffuse epithelial hyperplasia, atypical adenomatous hyperplasia, multifocal lung adenomas, and adenocarcinomas. Actually, we found by analyzing 10 tumors from each of 5 rhoB−/rhoB− mice, and 10 tumors from each of rhoB+/rhoB+ mice, that the RhoB-deficient mice exhibited 2.3-fold more tumors with strong nuclear YAP staining (54.2% vs. 23.4%, P < 0.05; Fig. 7D). Collectively, these findings support a novel function for RASSF1A in controlling the level of active RhoB through regulating the activity of its exchange factor GEF-H1, and further identify RhoB as a main determinant for regulation of YAP cytoplasmic nuclear shuttling.

Here, we demonstrate a critical role for RASSF1A in the bronchial epithelial phenotype maintenance, reporting that RASSF1A is required to restrict EMT, cell migration, cell invasion, and metastasis of HBEC lines. To our knowledge, this work is the first to describe RASSF1A as a metastasis suppressor in lung carcinogenesis. Such promigratory effect could be attributed in part to the increase of YAP transcriptional activity in RASSF1A-depleted cells. YAP promigratory effects have been already reported in ovarian cancer cells treated by LPA (34) or in Drosophila border cells migration (35). In line with such data, we report here that RASSF1A knockdown promotes a nuclear accumulation of active YAP. A regulation of YAP has been previously reported through a Trio-Rho/Rac signaling circuitry, downstream Gαq activation (36). ECM rigidity and mechanical stress were also shown to regulate YAP transcriptional activity, via the tension the actomyosin cytoskeleton (37), and a Rho GTPase activity (6). However, the specific Rho protein involved in such process remained unknown (38).

We previously reported that RASSF1A depletion dramatically altered tubulin cytoskeleton (10) as previously reported in A549 cells (39). We also found that paclitaxel, a tubulin-interacting drug, actually impaired cell invasion induced by RASSF1A depletion (data not shown). We show here that RASSF1A could alter cell migration by specifically activating actin cytoskeleton-regulating cofilin. Besides, in our hands, cofilin depletion reduced the RASSF1A-induced YAP nuclear localization, whereas Aragona and colleagues previously reported that YAP/TAZ nuclear localization and activity were upregulated upon cofilin depletion in dense cells, grown on stiff substrate ECM (37). This discrepancy could result from cofilin different functions according to different contexts: mechanotransduction and RASSF1A-dependent transformation. Aragona and colleagues showed cofilin was required to limit YAP/TAZ activity in cells experiencing low mechanical stress (low density or soft substrate), while we show here that cofilin is required to increase nuclear YAP in cells lacking RASSF1A expression.

Our data strongly suggest that among Rho actin–regulating proteins, RhoB could be targeted by RASSF1A and lead to the activation of the actin-depolymerizing factor cofilin. Three main results led us to this hypothesis. First, in RASSF1A-depleted cells, RhoB mRNA expression is decreased, which paralleled a decrease of active RhoB-GTP protein. Second, overexpressed RhoB increased inactive phospho-cofilin, whereas a dominant negative, nucleotide-free, RhoBN19 mutant, did not impact cofilin regulation. Third, we showed that cell motility induced by RASSF1A depletion could be efficiently inhibited by GEFs activating RhoB, while dominant negative RhoBN19 mutant or siRhoB were able to increase cell motility. Noteworthy, in our settings, overexpressing RhoB-GEF, GEF-H1, restrained RASSF1A-induced cell migration, whereas previous studies had assigned a promigratory role to GEF-H1 (29, 40). Conversely, GEF-H1 depletion (Fig. 6I), as RhoB depletion or RhoBN19 dominant-negative mutant transfection, increased HBEC-3 cells migration, in line with the cell spreading phenotype observed in PC3 cells depleted of GEF-H1 or RhoB (41). We confirmed that GEF-H1 was phosphorylated upon RASSF1A depletion, thus presumably inactivated (42). Mechanistically, RASSF1A depletion, by inducing a substantial decrease in PP2A activity, could downregulate GEF-H1 and subsequently RhoB activity (Fig. 7F). This model is consistent with a previous report showing PP2A and RhoB physical interaction (13), and we show here, that actually, RASSF1A levels can alter the composition of a multiprotein complex composed of GEFH1, RhoB, and PP2ACA. We report that overexpression of RhoB (wild-type or constitutively GTP-bound), inhibited the YAP nuclear accumulation upon RASSF1A depletion. Conversely, dominant-negative RhoBN19 or RhoB depletion led to YAP nuclear accumulation. Finally, in vivo, lung tumors observed in transgenic RhoB-deficient mice exhibited increased nuclear YAP. An opposite effect for RhoA and Rac was recently suggested, favoring YAP activation (36, 43). Thus, if control of YAP nuclear entrance was first linked to the promigratory RhoA/Rac proteins, we unravel here how RASSF1A and RhoB could act as cell migration and metastasis suppressors, by opposing to YAP nuclear entrance.

Our results enlightened that RASSF1A is a master-organizer of epithelial cell differentiation maintenance and migration-inhibiting genetic program. In lung cancer, RASSF1A inactivation by promoter methylation, could therefore contribute cancer invasiveness. The prognostic effect of RASSF1A methylation in early lung cancer could result from the removal of a physiologic program, involving RhoB and its regulator GEF-H1, restricting YAP activity and the ability to form distant metastases (Fig. 7F). Thus, YAP pharmacologic targeting in RASSF1A-methylated tumors, could be efficient in preventing metastasis spread, and improve lung cancer overall survival by delaying tumor progression.

No potential conflicts of interest were disclosed.

Conception and design: G. Levallet, G. Zalcman

Development of methodology: F. Dubois, M. Keller, F. Soncin, M. Vaisse-Lesteven, G. Levallet, G. Zalcman

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): F. Dubois, M. Keller, O. Calvayrac, F. Soncin, L. Hoa, J. Mazières, G. Levallet, G. Zalcman

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): F. Dubois, M. Keller, O. Calvayrac, A. Hergovich, J. Mazières, G. Levallet, G. Zalcman

Writing, review, and/or revision of the manuscript: F. Dubois, O. Calvayrac, F. Soncin, A. Hergovich, M.-C. Parrini, J. Mazières, J. Camonis, G. Levallet, G. Zalcman

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): F. Dubois, L. Hoa, A. Hergovich, G. Zalcman

Study supervision: A. Hergovich, G. Levallet, G. Zalcman

Other (many discussions, interpretations of data, modelling): J. Camonis

The authors thank Drs. Goux (CMAbio) and Elie (HiQ) for their assistance in electronic and confocal microscopy.

This work was supported by Cancéropole Nord-Ouest (G. Zalcman), the Ligue contre le Cancer de Normandie (G. Zalcman, G. Levallet), and the CNMR (G. Levallet). A. Hergovich is a Wellcome Trust Career Development Fellow (090090/Z/09/Z).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.