Potential conflict of interest: Nothing to report.
This work was supported by grants from La Ligue Contre le Cancer (comité de la Dordogne), Association pour la Recherche sur le Cancer (grant 5077), Association Française pour l'Etude du Foie and Schering-Plough laboratory (to V.M.), La Ligue Nationale Contre le Cancer “Equipe labellisée” (to J.R.), INCa-ARC-ANRS (PAIR-CHC “No-FLiC”), and BioIntelligence (OSEO) (to J.Z-R.). S.S. was supported by a postdoctoral fellowship from the Fondation pour la Recherche Médicale.
We performed a review of public microarray data that revealed a significant down-regulation of Rnd3 expression in hepatocellular carcinoma (HCC), as compared to nontumor liver. Rnd3/RhoE is an atypical RhoGTPase family member because it is always under its active GTP-bound conformation and not sensitive to classical regulators. Rnd3 down-regulation was validated by quantitative real-time polymerase chain reaction in 120 independent tumors. Moreover, Rnd3 down-expression was confirmed using immunohistochemistry on tumor sections and western blotting on human tumor and cell-line extracts. Rnd3 expression was significantly lower in invasive tumors with satellite nodules. Overexpression and silencing of Rnd3 in Hep3B cells led to decreased and increased three-dimensional cell motility, respectively. The short interfering RNA-mediated down-regulation of Rnd3 expression induced a loss of E-cadherin at cell-cell junctions that was linked to epithelial-mesenchymal transition through the up-regulation of the zinc finger E-box binding homeobox protein, ZEB2, and the down-regulation of miR-200b and miR-200c. Rnd3 knockdown mediated tumor hepatocyte invasion in a matrix-metalloproteinase–independent, and Rac1-dependent manner. Conclusion: Rnd3 down-regulation provides an invasive advantage to tumor hepatocytes, suggesting that RND3 might represent a metastasis suppressor gene in HCC. (HEPATOLOGY 2012;55:1766–1775)
Hepatocellular carcinoma (HCC) is the main primary malignancy of the liver worldwide.1 Its overall poor prognosis is the result of high rates of postoperative recurrence and metastasis incidence. Intrahepatic metastases, especially venous metastases, are hallmark features of HCC progression. The escape of carcinoma cells from the tumor may be influenced by a permissive liver microenvironment and a gain of invasive abilities of tumor cells. Many data suggest that the latter involves dedifferentiation of epithelial cells, which occurs by loss of cell polarity and cell-cell contacts and the concomitant acquisition of migratory and invasive features, referred to as the epithelial-mesenchymal transition (EMT).2 Although recent evidence suggest that hepatocellular EMT plays a pivotal role in the dissemination of malignant hepatocytes during HCC progression, the underlying molecular mechanisms remain to be characterized.3, 4
Ras homolog (Rho) GTPases, including RhoA, Rac1, and cell division cycle 42 (Cdc42), are the main regulators of the actin cytoskeleton and therefore general modulators of cellular processes important for tumor biology. Moreover, deregulated Rho GTPase signaling was reported to play an important role in the initiation and the progression of HCC.5, 6 Rnd3/RhoE is an atypical member of the Rho GTPase family because it is devoid of GTPase activity. The best-characterized function of Rnd3 is the inhibition of RhoA activity and the subsequent down-regulation of ROCK-mediated actomyosin contractility.7, 8 Through its role as a negative regulator of the Rho/ROCK pathway, Rnd3 was involved in tumor cell migration and invasion9-11 and myoblast fusion.12 More recently, Rnd3 was shown to inhibit cell-cycle progression, apparently independently of cytoskeleton remodeling.8 Thus, Rnd3 has been implicated in different steps of cancer development, such as regulation of cell proliferation and apoptosis,13-15 cell transformation,13 or cell migration and invasion.
Our reanalysis of five transcriptomic studies revealed an alteration of Rnd3 messenger RNA (mRNA) expression in HCC, compared to nontumor liver tissues,5 with four of five showing a down-expression16-19 and a single one, based on only four cases, an overexpression.20 Here, we confirm that Rnd3 is down-regulated in most human HCC and HCC-related cell lines, and we provide evidence that Rnd3 down-regulation increases HCC invasion and thus may favor HCC progression.
Samples came from resected or explanted livers with HCC of patients treated in Bordeaux from 1992 to 2005. Fragments of fresh tumor and nontumor liver tissues (taken at a distance of at least 2 cm from the tumor) were immediately snap-frozen in liquid nitrogen and stored at −80°C. RNA or proteins were extracted as previously described.21 HCCs used as the Affymetrix hybridization set (57 HCCs) and the quantitative reverse-transcription polymerase chain reaction (qRT-PCR) validation set (63 HCCs) were described.21 Characteristics of HCCs used for the immunoblotting analysis (27 HCCs) are indicated in Supporting Table 1.
BioCoat Matrigel invasion chambers (BD Biosciences, Franklin Lakes, NJ) were used according to the manufacturer's protocol. Briefly, cells were trypsinized, washed, resuspended in serum-free medium (Dulbecco's modified Eagle's medium [DMEM]; Glutamax; Invitrogen, Carlsbad, CA), supplemented with 0.1% bovine serum albumin, and 5 × 104 cells were placed in the top portion of the invasion chamber. The lower portion of the chamber contained 5% fetal bovine serum as a chemoattractant. After 20 hours, cells that migrated to the bottom chamber were fixed in 3% paraformaldehyde, stained with phalloidin/Alexa 546 and Hoechst, photographed, and counted. For assays in which cells were exposed to drugs, both the top and bottom chambers contained either 10 μM of GM6001 or 5 μM of EHT1864 or EHT4063 throughout the assay. To analyze the morphology of invading cells, cells were included in a type I collagen gel (BD Biosciences) added to the upper chamber of a Transwell plate, as described previously.22
Statistical analysis was performed with GraphPad Prism software (GraphPad Software, Inc., La Jolla, CA). Differences between means were assessed with Mann-Whitney's test or the Student's t test. When comparing multiple means, we used an analysis of variance (ANOVA). Correlations between the mRNA level of expression and qualitative variables were calculated with Kruskal-Wallis' nonparametric test. Pearson's test was used to compare quantitative values of expression. P values less than 0.05 were considered significant.
Materials and Methods.
See the Supporting Materials and Methods for details regarding antibodies and reagents, short interfering RNA (siRNA) and microRNA (miRNA) transfection, stable cell-line construction, cell-growth assay and culture, immunohistochemistry (IHC), immunofluorescence (IF), and reverse-transcription polymerase chain reaction (RT-PCR) procedures.
RND3 Is Down-regulated in HCC and HCC Cell Lines.
To investigate the expression levels of RND3 in HCC, we reanalyzed the Affymetrix GeneChip arrays of our own series of 57 HCCs and five samples of pooled nontumor tissues.21 A highly significant down-regulation of RND3 mRNA was observed when HCCs were compared to nontumor tissues (Supporting Fig. 1A). Quantitative RT-PCR (qRT-PCR) results on the same sample set correlated very well with the array data (Supporting Fig. 1B; Pearson's r = 0.7915; P < 0.0001). These data, in addition to qRT-PCR analysis on a second independent set of 63 tumors, demonstrated that RND3 mRNA expression was significantly lower in HCC than in cirrhotic livers, benign hepatocellular adenomas, and nontumor livers (Fig. 1A,B). The mean level of RND3 mRNA expression in malignant specimens was approximately 2-fold lower than that in benign tissue. Rnd3 expression level was not correlated to HCC etiology (i.e., virus- or alcohol-related HCC) (Supporting Fig. 1C-E). However, RND3 mRNA expression was significantly lower in tumors with satellite nodules, which is indicative of local invasion of HCC (P = 0.0313; Fig. 1C).
The expression of Rnd3 protein was analyzed by western blotting on a subset of 27 HCC samples (Fig. 2A; Supporting Fig. 1F). Twenty-three of twenty-seven cases of HCCs (85%) showed a decreased Rnd3 expression, when compared to peritumor tissue. Mean tumor/nontumor ratio was 0.68 ± 0.08 (P = 0.0005). Using IHC, Rnd3 expression in peritumor tissue varied from faint to intense and was predominantly localized to the cytoplasm of hepatocytes (Fig. 2B). In contrast, low or no expression was observed in tumor samples (Fig. 2B). Rnd3 protein expression was also determined in healthy primary human hepatocytes as well as in the tumor cell lines, Huh6, Huh7, SNU398, SNU475, Hep3B, and HepG2. Results showed that Rnd3 expression was reduced in all tumor cell lines tested, as compared to primary hepatocytes (Fig. 2C).
Rnd3 Regulates Invasion and Growth of HCC Cells.
Because RND3 expression showed a strong correlation with the presence of satellite nodules in HCC, we analyzed the effect of changes in RND3 expression level on cell invasion in the Hep3B HCC cell line. Lentiviral transduction led to a 6-fold overexpression of Rnd3, which was associated with a 4.5-fold reduction in cell ability to invade Matrigel (Fig. 3A). On the other hand, transient Rnd3 knockdown using two different siRNAs led to decreased expression of Rnd3 protein by 95% (S1) or 75% (S3) (Fig. 3B) and resulted in a significant increase in invasion (Fig. 3B). The increase was more drastic with S1 than S3, which is in agreement with their silencing efficacy. While performing invasion assays, we also analyzed cell growth and found that Rnd3 knockdown led to an inhibition of HCC cell growth (Fig. 3C). Thus, our results demonstrate that Rnd3 expression levels inversely regulate HCC cell-invasion and growth properties. Because of our initial observation that down-regulation of Rnd3 was associated with evidence of an invasive phenotype in patients, and to better characterize Rnd3 involvement in HCC progression, we focused our study on the invasion mechanism.
Rnd3 Depletion Induces Down-regulation of E-Cadherin.
Because loss of the cell-junction protein, E-cadherin, is associated with HCC cell invasiveness,23, 24 we evaluated E-cadherin expression in Rnd3-depleted cells. Rnd3 silencing in Hep3B (Fig. 4A,B) using both siRNAs led to a significant decrease in E-cadherin mRNA expression, whereas a significant down-regulation of E-cadherin protein was only observed with S1. However, decrease of E-cadherin protein expression was significantly observed with both siRNAs in Huh7 cells (Supporting Fig. 2A). IF analyses confirmed that E-cadherin expression was strongly reduced at cell-cell contacts in Rnd3-silenced cells (Supporting Fig. 2B). These results suggest that Rnd3 depletion affected the integrity of adherens junctions. We then sought to analyze E-cadherin protein levels in the 27 HCCs previously used for measuring Rnd3 expression. E-cadherin expression was down-regulated in 16 of 27 HCCs, as compared to peritumor tissue (Supporting Fig. 3). Interestingly, Rnd3 expression was significantly correlated with E-cadherin expression in patient samples (Fig. 4C; Pearson's r = 0.6190; P = 0.0006).
Rnd3 Knockdown Leads to a Partial EMT.
Because loss of E-cadherin expression and increased invasiveness are hallmarks of EMT, we further analyzed the expression of mesenchymal markers, such as vimentin, and the transcription factors, Snail1, Slug, zinc finger E-box binding homeobox 1 (ZEB1), and Smad-interacting protein 1 (SIP1)/ZEB2, which are described as transcriptional repressors of E-cadherin.25 qRT-PCR analysis revealed that Rnd3 silencing induced the mRNA expression of ZEB2, but not of ZEB1 or other EMT markers (Fig. 5A; Supporting Fig. 4). Because ZEB1/2 expression is under the control of the miR-200 family that targets their 3′ untranslated regions (UTRs),26 we monitored miR-200b and miR-200c expression in Rnd3-silenced Hep3B cells. The expression of both miRNAs was significantly decreased upon Rnd3 silencing (Fig. 5B). Moreover, forced overexpression of miR-200b and/or miR-200c in hepatoma cells down-regulated ZEB1 and ZEB2 expression, leading to E-cadherin up-regulation and increased cell-cell contacts (Supporting Fig. 5). Thus, Rnd3 knockdown induced a decrease in expression of the guardians of the epithelial phenotype, miR-200, and an increase in that of the EMT promoter, ZEB2, leading to E-cadherin repression.
Rnd3 Depletion Favors HCC Cell Invasion Through an Amoeboid (Pseudopodal)-Like Mechanism.
In a three-dimensional (3D) environment, individual cancer cells use a broad spectrum of migration and invasion mechanisms, which are dictated by the extracellular matrix (ECM) together with specific cell determinants. These include amoeboid and mesenchymal modes of movement, which are distinguished by their different usage of Rho GTPase-signaling pathways and distinct requirements for extracellular proteolysis.27 Amoeboid cells show high levels of actomyosin contractility involving signaling through RhoA/ROCK, and their movement is associated with deformation of the cell body through the ECM without proteolysis. In the mesenchymal-type movement, cells have an elongated morphology with Rac/Cdc42-induced protrusions at the leading edge, and this movement requires ECM proteolysis. We first attempted to discriminate between the two modes of invasion through the inhibition of matrix metalloproteinases (MMPs), whose activity is only required for the mesenchymal movement. The broad-spectrum MMP inhibitor, GM6001, did not decrease the invasion induced by Rnd3 depletion, suggesting that Rnd3-silenced cells invade the ECM without degrading it (Fig. 6A; Supporting Fig. 6A). Second, we analyzed the morphology of cells invading a thick type I collagen matrix.22 Although both control and Rnd3-silenced cells showed a rounded morphology, Rnd3-silenced cells were observed as isolated cells in the matrix and developed long actin-based protrusions, such as pseudopodia (Fig. 6B; Supporting Fig. 6B,C). Finally, to analyze the involvement of RhoGTPases in invasion induced by Rnd3 loss, Hep3B cells were transfected with GTPase-specific siRNAs, in addition to Rnd3 siRNAs, and invasion assays were performed in Matrigel-coated Transwells. The invasion of Rnd3-silenced cells was strongly inhibited by Rac1 or Cdc42, but not by RhoA knockdown (Fig. 6C). Rac1 requirement was also demonstrated using the pharmacological Rac1 inhibitor, EHT186428 (Fig. 6D; Supporting Fig. 6D). Collectively, our data show that Rnd3 knockdown induces HCC cell invasion through a Rac1-dependent and MMP-independent mechanism, thus suggesting an amoeboid pseudopodal-like mechanism.29
We report that RND3 is down-regulated in a majority of HCC cell lines and tissues. Previously, Rnd3 expression was also reported as low in biopsies from prostate and gastric cancers15, 30 and was suggested to act as a tumor suppressor in these cancers. However, RND3 expression was not systematically decreased in tumors because it was found high in non-small-cell lung31, 32 and pancreatic cancers.33 Thus, despite its ubiquitous expression in healthy tissues,34 Rnd3 regulation and biological effect may be significantly different in various tumors.
Rnd3 belongs to the Rnd subfamily of the Rho GTPase family. Because Rnd proteins are not regulated by the typical GTP/GDP cycle, they are thought to be regulated primarily at their transcriptional level. Here, using qRT-PCR, immunoblotting analysis, and IHC, we showed a down-regulation of RND3 mRNA and protein in HCC. The mechanism for Rnd3 down-regulation is still unclear. Although RND3 was reported to be a direct transcriptional target of p53,35 no correlation with p53 mutations could be established in our HCC samples (data not shown). Recently, it was reported that the miRNA miR-200b, directly reduced the expression of RND3 in HeLa cells.36 However, the relevance of this regulation in HCC remains to be evaluated.37 In addition, it was described that Rnd3 is regulated by histone deacetylation in gastric cancer cells,30 raising the hypothesis that it may also be regulated at the epigenetic level in hepatic tumors.
Rnd3 has been involved in diverse cellular functions, including actin cytoskeleton remodeling and cell-cycle progression.8 Because of the striking down-regulation of RND3 in HCC and of its biological functions, we hypothesized that the low expression of RND3 would give an advantage to liver tumor cells and contribute to the development of HCC. Surprisingly, we found that Rnd3 knockdown inhibited HCC cell growth. However, cells with reduced Rnd3 expression also acquired invasive capacity. This suggests that Rnd3 regulates a switch to attenuate cell growth and favor cell invasion. This is consistent with the concept that profound morphological changes are incompatible with high proliferation, and that attenuation of cell proliferation favors invasion versus tumor growth.2 In this respect, the EMT-promoting factor, Snail, was described to induce partial G1/S cell-cycle arrest that is, at least in part, a result of the repression of CCND2-encoding cyclin D2.38 Further experiments are required to decipher the mechanistic insights responsible for HCC growth arrest observed upon Rnd3 silencing.
On the other hand, the alteration in the 3D cell motility observed when Rnd3 expression was modulated is consistent with findings in healthy9 and transformed fibroblasts,39 showing a reduced invasion subsequent to overexpression of Rnd3. These results are, however, in sharp contrast with the reported implication of Rnd3 in the acquisition of an invasive phenotype of melanoma cells. Indeed, Rnd3 is overexpressed in melanoma cell lines and its down-regulation reduced cell-invasion ability.11 This could reflect the plasticity of cancer cells and the different implication of Rnd3 in various tumors.
Characterization of invasion of HCC cells induced by Rnd3 knockdown revealed the absence of MMP activity requirement, suggesting an amoeboid-like movement. However, we demonstrated that this movement occurs in a RhoA-independent manner. Because Rnd3 was mainly described as a RhoA pathway antagonist, this may represent a novel RhoA-independent role of Rnd3. We further characterized cell invasion induced by Rnd3 silencing as a Rac1-dependent movement, with a round morphology and the presence of actin-rich pseudopodia. Thus, according to the multiscale tuning model from Friedl and Wolf,29 we assume that the loss of Rnd3 induced an amoeboid pseudopodal-like mode of movement facilitated by the loss of strong adhesive cell-cell interactions, which is itself linked to the repression of E-cadherin expression. Remarkably, Rnd3 down-regulation strongly correlated with E-cadherin down-expression in HCC samples, and low levels of Rnd3 also correlated with the presence of satellite nodules, suggesting that our observation may be relevant for HCC progression. Although no publication has reported on an effect of Rnd3 on E-cadherin expression as yet, our data agree with others showing a role of Rnd3 on the expression of M-cadherin12 and, more generally, on the assembly of adherens and tight junctions.40 Consistent with this, depletion of Rnd3 in A431 squamous-cell carcinoma cells led to loss of cell-cell cohesion and defective collective cell invasion.41 We found that the repression of E-cadherin occurs at the mRNA level through the up-regulation of the EMT transcription repressor, ZEB2. We demonstrate, for the first time, that Rnd3 regulates the miR-200/ZEB/E-cadherin pathway. ZEB1 and ZEB2 are master regulators of the mesenchymal phenotype that repress the transcription of genes containing E-box elements in their promoters, including E-cadherin.42 The miR-200 family has been shown to target ZEB1 and ZEB2 through their 3′ UTRs. In addition, ZEB1 and ZEB2 directly repress miR-200 miRNA expression, demonstrating a double-negative feedback loop between ZEB1/ZEB2 and the miR-200 family during EMT and tumorigenesis.26 Here, we demonstrated that Rnd3 knockdown induces a decrease of miR-200b and miR-200c and an increase in ZEB2 expression, resulting in decreased E-cadherin expression and the acquisition of mesenchymal features (Fig. 7). These results are also in agreement with the previously reported role of ZEB2 in HCC cell invasion.43 The mechanisms by which Rnd3 silencing alters the miR-200/ZEB balance remain to be characterized (Fig. 7). However, because ZEB2, but not ZEB1, expression was altered in response to Rnd3 silencing, we postulate that Rnd3 silencing may probably first act on ZEB2 expression, which, in consequence, alters miR-200 transcriptional levels. In addition, Rnd3 silencing induced only a partial EMT, because we did not find an up-regulation of vimentin and MMP members (data not shown) shown to be under the control of Snail and ZEB2 in liver tumor cells.43 Because E-cadherin loss and the dissolution of the E-cadherin-mediated adherens junction represent key preliminary steps in EMT, Rnd3 may participate in the establishment of an invasive phenotype of liver tumor cells. In conclusion, our results suggest that RND3 is a potential metastasis suppressor gene in HCC. The targeting of its regulatory pathway with specific inhibitors may consequently offer a new therapeutic avenue in the management of cancer progression.
The authors thank Dr. C. Perret (Cochin Institute, Paris, France) for the Huh6 cell line, Dr. L. Désiré (Exonhit Therapeutics, Paris, France) for EHT1864 and EHT4063 components, and C. Gauthier-Rouvière for discussions. The authors also acknowledge V. Guyonnet-Duperat and V. Pitard from the vectorology and flow-cytometry core facilities, respectively (SFR TransBioMed, Bordeaux, France). The authors thank S. Loriot, C. Péanne and Dr. F. Sagliocco for their help with, respectively, IHC, cell-growth analyses and tumor protein extract preparations. The authors are grateful to Drs. E. Chevet and F. Saltel (INSERM U1053, Bordeaux, France) for their critical reading of the manuscript for this article.