Hepatocellular carcinoma (HCC) accounts for>80% of primary liver cancer and is the fifth most prevalent cancer in the world, with approximately 500 000 new cases per year (1, 2). HCC is the third leading cause of mortality among cancer patients. The high mortality rate of HCC is mainly related to HCC metastasis, which is also one of the major causes of tumour recurrence in HCC patients after tumour resection. Cancer metastasis is the most devastating but least well-understood aspect of HCC. Mounting evidence has shown that RhoGTPases and Rho-effectors play important roles in HCC movement and metastasis. This review will first focus on the cellular and molecular mechanisms involved in HCC metastasis. Following the signalling axis of the RhoGTPases/Rho-effector pathway, we will sequentially discuss the findings on the upstream RhoGTPase regulators, RhoGTPases and Rho-effectors and their downstream targets in HCC metastasis. We will also discuss the potential of this pathway as a therapeutic target for HCC treatment. A better understanding of the RhoGTPases/Rho-effector pathway in the metastatic process is awaited so that more effective therapeutic strategies can be devised to treat metastatic and recurrent HCC.
Hepatocellular carcinoma (HCC) is an intractable disease with an extremely high mortality rate. Metastasis is the major factor of liver failure, tumour recurrence and death in HCC patients. Unfortunately, no promising curative therapy for HCC metastasis is available as yet; therefore, treatment for advanced HCC still remains a formidable challenge. A large body of evidence has demonstrated that the RhoGTPases/Rho-effector pathway plays important roles in mediating HCC metastasis based on their foremost functions in orchestrating the cell cytoskeletal reorganization. This review will first discuss the general principles of cancer metastasis and cancer cell movement with a particular focus on HCC. We will then summarize the implications of various members in the RhoGTPases/Rho-effectors signalling cascade including the upstream RhoGTPase regulators RhoGTPases and Rho-effectors and their downstream targets in HCC metastasis. Finally, we will discuss the therapeutic insight of targeting the RhoGTPases/Rho-effector pathway in HCC. Taken together, the literature demonstrates the importance of the RhoGTPases/Rho-effector signalling pathway in HCC metastasis and marks the necessity to have a more thorough knowledge of this complicated signalling network in order to develop novel therapeutic strategies for HCC patients.
Fundamental steps of cancer cell movement and metastasis
Metastasis is a multistep process, involving cancer cell invasion, intravasation, extravasation and colonization (Fig. 1). First, cancer cells detach from the primary tumour mass and invade into the extracellular matrix (ECM). This step requires degradation of the ECM with matrix metalloproteinases (MMPs) and involves substantial cell cytoskeleton reorganization to drive the cell movement. Cancer cells then enter the circulatory system and survive in the adverse environment inside the vasculature and the lymphatic system. Eventually, cancer cells exit the circulatory system and colonize into new favourable sites in the same organ or in distant organs. The microenvironments of the secondary sites are essential in determining the efficiency of colonization. The liver itself is an ideal site for HCC metastasis as it has a microenvironment identical to that of the primary tumour. For intrahepatic metastasis, HCC cells travel through the portal veins to finally colonize in other parts of the liver. For extrahepatic metastasis, HCC cells enter the blood stream through the hepatic vein, then the inferior vena cava to the lungs and eventually to other organs including the bones, adrenal glands and brain.
Cell motility is vital for many steps involved in cancer metastasis including invasion, intravasation and extravasation. External signals such as chemoattractants initiate cell movement by triggering a series of signalling events when they bind to their receptors on the cell membrane. Upon stimulation, a cell undergoes a cyclical mechanism of three steps of coordinated actomyosin contraction involving the assembling and disassembling of actin filaments (3, 4). The first step requires the extension of the cell body by generating protrusions towards the direction of cell migration at the anterior end (Fig. 2). This protruding step includes the formation of new adhesions that anchor the protrusions to the substratum. The second step requires actomyosin contraction that drives the cell body to move forward towards the protrusive end. Actomyosin contraction refers to the force generated by actin filaments, known as stress fibres, which are composed of polymerized actin and myosin (Fig. 2). Stress fibres stretch across the cells and attach to a group of macromolecules called focal adhesion molecules. The focal adhesion molecules anchor the cells to the substratum and act as mechanical linkages to the ECM (Fig. 2). Following actomyosin contraction, the third step emerges concomitantly with the release of the attached adhesions at the posterior end. This de-adhering step facilitates the actomyosin contraction in dragging the cell to move forward.
RhoGTPases and Rho-effectors
There are 20 members of RhoGTPases family. They can be divided into eight subfamilies, namely, Rho (RhoA, RhoB and RhoC), Rac (Rac1, Rac2, Rac3 and RhoG), Cdc42 (Cdc42, RhoJ and RhoQ), Rnd (Rnd1, Rnd2 and Rnd3), RhoBTB (RhoBTB1 and RhoBTB2), RhoV and RhoU, RhoD and RhoF and RhoH subfamilies [see (5) for a review]. RhoGTPases play important roles in actin reorganization and cytoskeletal arrangement that are essential to cell movement. Among these, Rho, Rac and Cdc42 are the three most well-characterized subfamilies of RhoGTPases. RhoGTPases act as molecular switches and function through activation of their specific downstream effectors to modulate the cytoskeletal rearrangement. Wiskott–Aldrich syndrome protein (WASP) family and P21-activated protein kinase (PAK) are the downstream effectors of Cdc42 and Rac, while Rho-kinase (ROCK) is the most well-characterized downstream effector of Rho. Aberrant regulations of RhoGTPases and their effectors drastically alter the cytoskeleton arrangement and contribute to cancer metastasis and we will next discuss the regulators of this pathway in HCC.
Regulation of RhoGTPases and Rho-effectors
The activity of RhoGTPases is determined by their binding status to GTP or GDP molecules. GTP-bound RhoGTPases are functionally active, whereas GDP-bound RhoGTPases are inactive (Fig. 3A). Although most members of the RhoGTPase family themselves possess intrinsic GTPase activities hydrolysing GTP to GDP, their activities are tightly controlled by three upstream regulators. The Rho guanine nucleotide exchange factor (RhoGEF) activates RhoGTPases by promoting the exchange of GDP to GTP, while RhoGTPase-activating protein (RhoGAP) inactivates RhoGTPases by accelerating its intrinsic GTPase activity (Fig. 3A). Rho GDP-dissociation inhibitor (RhoGDI) acts as an additional layer of regulation that suppresses the activity of RhoGTPases by inhibiting the release of the GDP molecule from the RhoGTPases, hence inhibiting the association of GTP molecule with the RhoGTPases (Fig. 3A).
Once RhoGTPases are activated, they in turn activate the Rho-effectors to signal further downstream targets (Fig. 3B). Many Rho-effectors such as WASP, PAK1 and ROCK are activated by RhoGTPases through a common mechanism. These Rho-effectors are in autoinhibitory conformations and their functional domains are inaccessible to their substrates. Active RhoGTPases disrupt the intra-molecular autoinhibitory binding of the Rho-effectors so that the effectors' functional domains are exposed to their substrates (Fig. 4) (6).
RhoGTPase-activating protein in hepatocellular carcinoma
Among the three upstream regulators of RhoGTPases, the functions of RhoGAP are most extensively studied in HCC and will thus be further discussed in this review. Different microarray studies have revealed that many RhoGAP proteins such as p190RhoGAP and ArhGAP are downregulated in HCC [see (7) for a review]. However, the precise functions of individual RhoGAP members in HCC have not been clearly revealed, except for one RhoGAP family, deleted in liver cancer (DLC). There are three identified members in the DLC family: DLC1, DLC2 and DLC3. They are functionally similar, with reported tumour-suppressive properties in various human cancers [see (8) for a review]. Somatic mutation in DLC1 appears to be an uncommon event in HCC (9); nevertheless, the loss of DLC1 and DLC2 is frequent in HCC. In primary HCC, it was reported that 67.5% of the cases showed at least a two-fold reduced DLC1 mRNA expression level relative to the adjacent non-tumorous livers (9), and 18% of the cases showed reduced DLC2 mRNA expression relative to the adjacent non-tumorous livers (10, 11). Underexpression of DLC1 was mainly contributed by genetic deletion and promoter hypermethylation (9, 12, 13), while underexpression of DLC2 was mainly contributed by genetic deletion in HCC (10, 11).
Tumour- and metastasis-suppressive roles of deleted in liver cancer in hepatocellular carcinoma
Deleted in liver cancer 1 and DLC2 share many common characteristics including their suppressive effects on tumour growth, cell migration and cytoskeleton reorganization in HCC. Independent research groups have demonstrated the tumour-suppressive functions of DLC in various experimental models in HCC. Restoration of DLC1 expression in HCC cell lines, 7730K, Focus and SMMC-7721, significantly suppressed in vitro cell proliferation and in vivo subcutaneous tumour growth (10, 11). Overexpression of DLC1 in a mouse hepatoma cell line expressing oncogenic Ras suppressed tumour formation in mouse livers (14). A reverse model has demonstrated consistent tumour-suppressive effects of DLC1 with the use of p53-deficient and myc-overexpressing liver progenitor cells, hepatoblasts. DLC1 knockdown in the genetically modified hepatoblasts significantly accelerated HCC onset and enhanced HCC aggressiveness in mice (14). Overexpression of DLC2 in a hepatoblastoma cell line, HepG2, suppressed cell proliferation and anchorage-independent growth (15). However, our group has recently demonstrated that DLC2 knockout mice do not show a higher incidence of tumour formation in a diethylnitrosamine -induced HCC model, indicating that loss of DLC2 alone may not be sufficient to drive HCC because of a possible compensation from DLC1 (15).
Deleted in liver cancer also suppresses cancer migration, invasion and metastasis (16, 17). Ectopic expression of DLC1 can suppress cell migration and invasion in HCC, non-small-cell lung cancer, breast cancer, lung cancer and ovarian cancer cell line models (16, 18–23). Overexpression of DLC1 suppresses cell migration and invasion in 7703K, Focus and SMMC-7721, while overexpression of DLC2 suppresses HepG2 cell migration. Knockdown of DLC2 in HepG2 cells also reverses the suppressive effect of DLC2 in cell migration. The roles of DLC1 in metastasis have also been demonstrated in animal models. Overexpression of DLC1 in a metastatic breast cancer cell line reduced pulmonary metastases in an orthotopic nude mice model (16). Furthermore, stable HCC cell lines overexpressing DLC1 (7703K and Focus) reduced the dissemination of subcutaneous tumours to the lungs and livers of mice (16).
The anti-metastatic properties of DLC have been considerably explained by the detailed analyses of the cytoskeleton network. DLC exerts a drastic inhibitory impact on the cytoskeleton reorganization in various cell types (8). In HCC, DLC1 and DLC2 profoundly suppress actin polymerization as represented by perturbation of stress fibre formation (14, 15, 21, 24, 25). Also, DLC1 suppresses focal adhesion formation, which is important for the anchorage of stress fibres and the ECM (24, 25). Moreover, DLC1 suppresses myosin activation, which is crucial for actomyosin contractility (24). The inhibitory effects of DLC in the formation of cytoskeleton network are dependent on the RhoGAP regions of DLC (15, 18, 21, 26). All members of the DLC family induce severe cell shrinkage that can mainly be attributed to the disruption of the cytoskeleton network in cells.
All members of the DLC family possess RhoGAP activity specific for RhoA and Cdc42, but not Rac1 (9, 10, 23, 26). It has been reported that DLC1 RhoGAP is also specific for RhoB and RhoC (23). Different studies have shown that DLC1 exhibits the highest RhoGAP activity towards RhoA (9, 10, 23, 26). ROCK is one of the most well-characterized RhoA downstream effectors and our group has demonstrated previously that DLC1 antagonistically regulates the Rho/ROCK/MLC pathway in HCC (24). DLC1 abolishes ROCK-mediated events including the formation of stress fibres and focal adhesions, and the cortical phosphorylation of myosin light chain 2 in HCC cells (24). Also, dominant-active ROCK can partially reverse the morphological alterations induced by DLC1 in fibroblasts (24). Furthermore, as demonstrated by colony formation assay, knockdown of DLC1 could sensitize p53−/− and myc-overexpressing liver progenitor cells to ROCK inhibitors Y27632 and fasudil (14). All the above evidence leads to the conclusion that the DLC1/Rho/ROCK signalling axis is involved in regulating HCC cell proliferation and cell migration. Of note, overexpression of DLC1 does not affect the expression of ROCK (25) but suppresses ROCK-mediated myosin phosphatase and myosin light chain 2 phosphorylation (24). These findings are reasonable, as DLC1 inhibits RhoA activity, which influences only the activity but not the expression of ROCK. In addition, overexpression of DLC1 in HCC cell lines 7703K and Focus downregulates the expression of osteopontin and MMP9, both of which have been shown to be implicated in HCC metastasis (27–29). Moreover, overexpression of DLC1 in HCC cells, SNU-368, significantly represses the phosphorylation levels of focal adhesion kinase (FAK), paxillin and Cas proteins, which are all associated with focal adhesions dynamics. This finding again demonstrates the significance of DLC1 in the regulatory events of focal adhesions (18).
RhoGTPases in hepatocellular carcinoma
There are three members of Rho subfamily: RhoA, RhoB and RhoC. They share high homology and are all important for formation of stress fibres constituted by contractile actin and myosin filaments (30). Among all of the Rho family members, RhoA and RhoC are most well characterized in human cancers and will be discussed herein. RhoA was found to be significantly overexpressed in various cancers including HCC (31), breast (32), lung (33), colon (33), head and neck (34), testicular (35) and bladder cancers (36). In HCC, RhoA is commonly found to be overexpressed at both mRNA and protein levels. In two separate studies involving 128 HCC specimens and 64 HCC specimens, respectively, RhoA was found to be frequently overexpressed (37, 38) and the overexpression of the mRNA of RhoA correlated significantly with the overexpression of the protein of RhoA (37, 38). In addition, overexpression of RhoA was significantly associated with aggressive features of the tumours including cell differentiation, venous invasion, satellite lesions and advanced tumour stages (37, 38). Furthermore, overexpression of RhoA in HCC was found to be associated with a lower disease-free survival rate, indicative of a higher chance of tumour recurrence in another study (31).
It has been shown that dominant active RhoA enhances myosin phosphorylation and HCC cell invasion in vitro (39). Also, overexpression of constitutively active RhoA in rat hepatoma cells significantly enhances peritoneal invasion in a rat model (40). Notably, RhoA-mediated migratory and invasive ability in HCC cells can be suppressed by the ROCK inhibitor, Y27632 (40). In addition, the RhoA signalling pathway also plays an indispensable role in epithelial–mesenchymal transition (EMT). Overexpression of dominant-negative RhoA or ROCK mutants was sufficient to block the TGF-β-induced EMT phenotype in mammalian epithelial cell lines (41). The tumorigenic function of RhoA has recently been revealed in a system using p53−/− liver progenitor cells with myc and dominant-active RhoA expressed by retrovirus (14). In that study, it was demonstrated that activated RhoA cooperated with Myc and loss of p53 to enhance HCC formation and decrease the survival lengths of tumour-bearing mice. Consistently, knockdown of RhoA by the short hairpin approach in murine hepatoma lines suppressed tumour growth in a DLC1-dependent manner and that the RhoA expression level in these cell lines corresponded positively to the tumour size (14).
RhoC is implicated in cancer progression particularly metastasis in different cancers including HCC, melanoma, prostate, pancreatic and breast cancers (42–48). The metastatic properties of RhoC have been clearly revealed in a RhoC knockout mouse model with a polyomavirus middle T (PyV MT) background that developed a spontaneous palpable mammary gland tumour and secondary metastases (49). RhoC−/− PyV MT mice and RhoC+/− PyV MT mice showed no apparent differences in their primary tumours, but the number and the size of lung metastases were significantly reduced in RhoC−/− PyV MT mice (49). Furthermore, different groups have repeatedly reported that RhoC is frequently overexpressed in HCC and associated with metastatic and aggressive features of HCC including portal vein invasion, number of tumour nodules, metastatic lesions, poor cell differentiation, loss of tumour encapsulation and poor survival (42, 50–52). RhoC was found to be further overexpressed in metastatic lesions outside of the liver (52) and highly expressed in metastatic HCC cell lines, MHCC97 and HCCLM3, relative to the hepatoma cell line, HepG2 (47). Knockdown of RhoC in HCCLM3 cells suppressed cell migration and invasion but not cell proliferation and apoptosis in vitro (47). Knockdown of RhoC in HCCLM3 cells also suppressed the number of lung metastases formed and mildly attenuated the size of the primary HCC tumour in vivo (47).
Rac is responsible for the formation of lamellipodia and membrane ruffles, a flattened meshwork of actin filaments found at the cell periphery (53). There are four members in the Rac subfamily: Rac1, Rac2, Rac3 and RhoG (54). Among all the members of the Rac subfamily, Rac1 is the most extensively studied isoform, especially in cancer metastasis models (55, 56). Rac1 has been shown to enhance HCC cell migration, invasion and metastasis in a hypoxia/HIF1α/VEGF-dependent manner (57). The importance of Rac1 in HCC metastasis has been revealed by a pair of HCC cell lines derived from primary HCC (H2P) and its corresponding portal vein metastases (H2M) (58). Using the Rac-GTP pull-down assay, it has been shown that Rac1 has a significantly higher activity in H2M than H2P (58). From immunohistochemistry analysis on tissue microarray from 60 patients, Rac1 is significantly overexpressed in metastatic HCCs (including intrahepatic and extrahepatic metastases) as compared with their primary HCCs counterparts. Furthermore, the dominant-negative form of Rac1 inhibits cell invasion of the metastatic HCC cells, MHCC97H (57).
Cdc42 is responsible for the formation of filopodia and the regulation of cell polarity in different cell types. Cell polarity is an importation step in directional movement. When a cell becomes polarized, its Golgi apparatus and microtubule organizing centre reposition to the front of the nucleus, facing towards the direction of movement to provide direct supply of proteins and polymerized actin required at the leading front of the cell (59). Cdc42 is responsible for regulating cell polarity through the partitioning PAR complex (PAR6–PAR3–aPKC) in epithelial cells (60–62). Filopodia are highly dynamic, actin-rich protrusions composed of parallel bundles of actin filaments responsible for sensing the extracellular environment (63). As demonstrated in different cell types, Cdc42 null cells exhibit loss of the actin cytoskeleton and reduce filopodia formations (63). A previous report on analysis on 20 human HCC samples has shown that Cdc42 exhibits a higher protein expression in HBV-related HCC as compared with pre-cancerous tissues and HBV-unrelated HCC (64). Cdc42 knockout mice are embryonically lethal (65); therefore, research groups have used a hepatocyte-specific Cdc42 knockout mouse model to elucidate the roles of Cdc42 in the liver. Liver-specific Cdc42 knockout mice do not show an increased mortality rate but exhibit chronic jaundice (66). These hepatocyte-specific Cdc42 knockout mice develop hepatomegaly after birth. These mice also show other multiple defects in the liver such as enlarged canaliculi between hepatocytes, absence of liver plates and cholestasis. Of interest, the hepatocyte-specific Cdc42 knockout mice spontaneously develop macroscopic dysplastic nodules in the livers at 6 months of age, then HCC at 8 months of age and finally lung metastases. It has been reported that Cdc42 plays important roles in the normal physiological functions of hepatocytes. Cdc42 becomes activated between 3 and 24 h after a partial hepatectomy (67). Using the hepatocyte-specific Cdc42 knockout mice, another group has demonstrated that Cdc42 plays an important role in liver regeneration. Cdc42 knockout mice exhibited delayed liver regeneration and decreased DNA synthesis in hepatocytes after a partial hepatectomy (67). Although the hepatocyte-specific Cdc42 knockout mice exhibited a slower recovery rate at time points 48–168 h post-hepatectomy, their livers finally restored to the original liver-to-body weight ratio 3 weeks after the partial hepatectomy. This indicates that Cdc42 knockout in the hepatocytes defers but does not completely block liver regeneration (67).
Rho-effectors in hepatocellular carcinoma
RhoGTPases exerted its cytoskeletal impact through activation of their specific effectors. In this section, we will review the effectors that have been shown to be involved in HCC metastasis, namely WAVE2, PAK1 and ROCK.
There are five members in the WASP family, which can be further divided into two different sub-groups: WASP (WASP, N-WASP) and WASP family verprolin-homologous protein (WAVE1, WAVE2 and WAVE3). The WASP family acts as downstream effectors of Cdc42 and Rac. WASP binds and activates the Arp2/3 complex, which in turn catalyses and nucleates actin polymerization (68). Among the five members of the WASP family, only WAVE2 has been reported to be associated with HCC and therefore will be further discussed herein. WAVE2 has been reported to be associated with metastasis in colorectal cancer, melanoma, as well as HCC (68). It has been demonstrated in melanoma cells that WAVE2 is the main downstream effector of Rac1 in melanoma invasion and metastasis (69). Knockdown of WAVE2 remarkably suppresses membrane ruffles, cell migration and invasion in vitro, and reduces pulmonary metastasis of melanoma cells in C57BL/6 mice (69). Furthermore, colocalization of Arp2 and WAVE2 is associated with colorectal metastasis to liver and lymphatic invasion (70). From a study on 112 HCC samples, immunohistochemistry demonstrated that 71 of 112 HCC cases had WAVE2 overexpression, and overexpression of WAVE2 was associated with aggressive HCC features including the presence of venous invasion and multiple tumour nodules, absence of tumour capsules, poor cellular differentiation and shortened survival (71).
P21-activated protein kinase 1 is a downstream effector shared by Cdc42 and Rac. PAK1 is involved in the regulation of cell morphology and cell movement. PAK1 was found to be overexpressed in breast and colorectal cancers (72, 73). Stable expression of the kinase-active form of PAK1 enhances the mobility but not the proliferation of MCF7 cells, a breast cancer cell line (74). Reversely, stable expression of the kinase-inactive form of PAK1 suppresses breast cancer cell invasion by inhibiting the formation of focal adhesions and stress fibres (75). PAK1 is overexpressed in HCC at both mRNA and protein levels (76). Clinicopathologic correlation has revealed that overexpression of PAK1 mRNA is closely associated with features of aggressive HCC including the presence of venous invasion, poor cellular differentiation, advanced tumour stages and shortened disease-free survival (76). PAK1 is also highly expressed in H2M as compared with H2P (76). Stable overexpression of PAK1 in HepG2 hepatoma cells enhances cell migration and spreading through the regulation of stress fibres while knockdown of PAK1 in H2M cells suppresses HCC cell migration. Consistent with the findings in breast cancer, PAK1 has no major effect on cell proliferation in HCC (76).
Rho-kinase is a RhoA-GTP binding protein and a direct effector of RhoA (77–79). ROCK is a serine/threonine kinase that phosphorylates a wide repertoire of downstream targets involved in the cytoskeleton rearrangement. ROCK1 and ROCK2 are the two family members of the ROCK family (80). Overexpression of ROCK1 and ROCK2 was demonstrated in testicular and bladder cancers at the protein level (35, 36). In addition, ROCK inhibitor (Y27632) treatment was sufficient to abolish chemotactic migration in pancreatic cancer cell lines (81). Although ROCK1 and ROCK2 are highly homologous, sharing many common targets and cellular functions, only the ROCK2 protein is significantly overexpressed in HCC (82). Overexpression of ROCK2 is closely associated with the formation of a tumour microsatellite, which is an indicator of intrahepatic metastasis (82). Using constitutively active/inactive ROCK constructs, short hairpin RNA targeting ROCK and ROCK-specific inhibitors, the functional roles of ROCK in HCC invasion and metastasis have been elucidated by different research groups (24, 39, 40, 82, 83). Both ROCK1 and ROCK2 have been reported to participate in the activation of myosin and the polymerization of actin filaments in HCC cells. ROCK inhibitor, Y27632, suppresses actin reorganization, focal adhesion formation and myosin activity, and therefore suppresses cell migration and invasion in different HCC cell lines (24). Reversely, stable transfection of dominant-active ROCK1 or ROCK2 in rat hepatoma cells and human HCC cells significantly enhanced HCC cell motility and invasiveness (39, 40, 82). In a model using the injection of Li7 cells, a human HCC cell line, into the livers of SCID mice, Y27632 and stable expression of dominant-negative ROCK1 significantly reduced the incidence of macroscopic and microscopic intrahepatic metastasis and suppressed infiltrative tumour boundary (39, 83). In another model using the injection of MMP1 cells, a rat hepatoma cell line, into the peritoneal cavities of Donryu rats, Y27632 reduced RhoA-induced ascites and nodule formation and the frequency of tumour cell dissemination into the peritoneal cavity (40). In a model using the implantation of solid tumours into the livers of nude mice, our group showed that stable knockdown of ROCK2 remarkably suppressed different invasive features of the tumours, especially the incidence of tumour microsatellite formation, which is consistent with the clinical pathological correlation in human HCC (82). Intriguingly, almost all of the findings reported by different groups indicate that both ROCK1 and ROCK2 are more involved in the metastatic aspect instead of the tumour-proliferative aspect (24, 39, 40, 82, 83). The ROCK inhibitor shows no significant impact on HCC cell proliferation in vitro (24). In ROCK inhibitor-treated animals, HCC cells overexpressed with dominant-negative ROCK1, and HCC cells with ROCK2 knocked down all, resulting in a slight decrease of primary liver tumour size; however, these differences were mild and statistically insignificant (39, 82, 83).
Rho-kinase effectors in hepatocellular carcinoma
The major characterized functional roles of ROCK are related to its kinase activity in phosphorylating downstream substrates. ROCK is a serine/threonine kinase that recognizes the consensus phosphorylation sequence RXXS/T or RXS/T (80). Basic amino acids such as arginine (R) are required to be adjacent to the phosphorylation site of ROCK. Because of the high homology of ROCK1 and ROCK2, they phosphorylate common targets. ROCK phosphorylates a number of substrates important for cytoskeletal-related events and are implicated in HCC (Table 1). In this section, the most-characterized substrates of ROCK and their implications in HCC will be further discussed.
|ROCK substrates||Phosphorylation sites||Functions||Information in HCC|
|MLC2||Thr 18, Ser 19 (84)||Actomyosin contraction||Phosphorylated in HCC cells and is sensitive to ROCK inhibitor treatment (24)|
|MYPT1||Thr 697, Ser 853 (85)||Actomyosin contraction||Phosphorylated in HCC cells and is sensitive to ROCK inhibitor treatment (24)|
|GFAP||Thr 7, Ser 13, Ser 38 (119, 120)||Cytokinetic segregation of glial filaments||GFAP-positive stromal cells were highly expressed in ICC than in scirrhous HCC (121)|
|Vimentin||Ser 38, Ser 71 (122)||In vitro filament formation||Upregulated in metastatic HCC cell lines (123, 124)|
|Adducin||Thr 445, Thr 480 (89)||Formation of membrane ruffles and cell movement||Not available|
|Calponin||Thr 170, Thr 184 (125)||Actin filaments binding||Positively stained on the walls of blood vessels in the parenchyma of the HCC tissues (126)|
|Tau||Ser 262 (89)||Microtubules assembly and disassembly||Not available|
|LIM kinase 1||Thr 508 (91)||Actin polymerization||Not available|
|LIM kinase 2||Thr 505 (88)||Actin polymerization||Not available|
|FAK||Ser 732 (106)||Focal adhesion formation||Upregulated in HCC (108)|
|Cofilin||Ser 3 (89)||Actin depolymerization||Downregulated in highly metastatic HCC cell line (96)|
|Ezrin||Thr 567 (97, 98)||Cell membrane and cytoskeleton anchorage;|
Focal adhesion formation
|Upregulated in metastatic HCCs (101)|
Rho-kinase activates myosin and mediates muscle contraction, neurite retraction and other mechanisms driven by actomyosin contraction such as cancer cell movement, via phosphorylating two major substrates: myosin light chain 2 (MLC2) and myosin phosphatase 1 (MYPT1). ROCK phosphorylates and activates MLC2 at Thr 18 and Ser 19, while MYPT1 dephosphorylates and inactivates MLC2 at these sites (Fig. 5) (84). MYPT1 and ROCK counteract each other in MLC2 phosphorylation status, but ROCK can inactivate MYPT1 by phosphorylating MYPT1 at Thr 697 and Ser 853 (Fig. 5) (85). Myosin is an important component of stress fibres. Activated myosin connects the actin filaments to form stress fibres that generate actomyosin force to allow cell movement. So far, to our knowledge, there is no clinical report showing aberrant expression and activation of myosin in HCC. This may be because of the abundance of myosin in normal liver cells. Nonetheless, the phosphorylation statuses of MLC2 and MYPT1 in cancer cells have been examined. MLC2 phosphorylation can be suppressed by ROCK inhibitor using different cancer cell models such as HCC and breast cancer cells (24, 86). Previously, our group has reported that that MLC2 phosphorylation (Ser 18) at the cell cortex is important for HCC cell migration and was suppressed by ROCK inhibitor (24). DLC1 also suppressed MLC2 phosphorylation by inhibiting the Rho/ROCK pathway in HCC cells (24). Also, we reported that loss of ROCK2 expression led to decreased MYPT1 phosphorylation at Thr 853 in HCC cells (82).
Rho-kinase phosphorylates and activates LIM kinase (LIMK) (87, 88). LIMK phosphorylates cofilin at Ser 3 and blocks the activity of cofilin in depolymerizing actin filaments (89); hence, activation of LIMK is important for actin cytoskeletal organization. The LIMK/cofilin pathway is regulated by different members of the RhoGTPases and their effectors. Rac activates LIMK1 while RhoA and Cdc42 activate LIMK2 through their effectors (89, 90). PAK and ROCK are able to phosphorylate LIMK1 at Thr 508 and LIMK2 at Thr 505, thereby enhancing the activity of LIMK (88, 91–93). Phosphorylated and activated LIMK then enhances phosphorylation of cofilin at Ser 3, which inactivates cofilin and inhibits cofilin-induced F-actin depolymerization (89). Cofilin therefore is often referred to as the terminal effector of the cell signalling cascades that regulate the cytoskeleton rearrangement. LIMK and cofilin are generally overexpressed in cancer cells especially in invasive and metastatic cell lines (94). LIMK was overexpressed in rat and mouse mammary tumour, especially in the invasive subpopulation (94, 95). Cofilin is frequently overexpressed in human cancers including breast, renal, ovarian and oral cancers (94). Depletion of cofilin expression by siRNA suppressed breast cancer cell invasion (86). However, a contradictory result was found in HCC cell lines. A proteomic study shows that cofilin was downregulated in a highly metastatic HCC cell line, MHCC97H, relative to the less metastatic HCC cell line, MHCC97L (96). Because it is difficult to draw a conclusion based on just two cell lines and there are no clinical data on the expression and phosphorylation status of LIMK and cofilin in HCC thus far, a more detailed investigation on the implications of the LIMK/cofilin pathway in HCC is awaited.
Rho-kinase also regulates a group of proteins closely linked to the cytoskeleton, the ezrin/radixin/moesin proteins, responsible for anchoring actin filaments to the plasma membrane. ROCK phosphorylates ezrin at Thr 567 (97, 98) and ROCK-mediated ezrin Thr 567 phosphorylation is involved in focal adhesion formation (97). Because of the importance of this phosphorylation site of ezrin, the non-phosphorylated mutant of ezrin (ezrin T567A) was used universally as dominant-negative ezrin (97, 99, 100). Ezrin has been clearly demonstrated to be an important gene that confers metastatic capability by microarray analysis comparing highly and poorly metastatic murine rhabdomyosarcoma (RMS) cell lines (100). Ezrin was found to be significantly overexpressed in highly metastatic murine RMS and osteosarcoma cell lines as compared with the poorly metastatic counterparts (99, 100). Knockdown of ezrin expression or overexpression of dominant-negative ezrin T567 suppressed lung metastases formed by an intravenous injection in the mice (99, 100). Conversely, overexpression of wild-type ezrin enhanced the metastatic properties of non-metastatic RMS cells (100). Interestingly, ezrin did not affect primary tumour growth as well as in vitro cell proliferation in the osteosarcoma model (100). In a study on 49 HCC patients, ezrin mRNA was frequently overexpressed in HCC relative to the non-tumorous counterparts (101). Ezrin was found to be highly expressed in HCC cells with a high metastatic capability including SF7721, MHCC-1 and MHCC97-H relative to those with a low metastatic capability including SMMC7721, Hep3B and HepG2 (102). It was reported that suppression of ezrin reduced the in vitro migration, invasion and cell proliferation in HCC cell lines (102, 103).
Focal adhesion kinase is a non-receptor tyrosine kinase localized at the cytoplasm and integrin-rich focal adhesions. It is responsible for the turnover of the focal adhesion complex, which is essential for cell movement. FAK autophosphorylates at tyrosine residue 397 and binds to Src kinase (104). Src kinase further phosphorylates FAK to activate FAK and enhances the binding of FAK with a number of signalling proteins and focal adhesion proteins such as Cas130 and paxillin (104). Tyrosine phosphorylation plays an important role in the regulation of FAK activity (104). It has been first reported that inhibition of ROCK by Y27632 or by overexpressing the dominant-negative forms of RhoA or ROCK suppressed VEGF-induced Tyr 407 phosphorylation of FAK and suppressed the recruitment of paxillin and vinculin to focal adhesions in human umbilical vein endothelial cells (105). This phenomenon was further explained in the next report using the same cellular system showing VEGF-induced ROCK phosphorylation of FAK at Ser 732, which in turn affected Tyr 407 phosphorylation of FAK (106). A non-phosphorylatable FAK mutant (S732A) can suppress Tyr 407 phosphorylation of FAK and inhibit the recruitment of vinculin to the ventral focal adhesions (106). This FAK mutant also suppresses VEGF-induced cell migration (106). Different research groups have reported that FAK is frequently overexpressed in various cancers including thyroid, prostate colon, rectum, ovary, cervix and oral epithelial cancers (107). In HCC, FAK is also frequently overexpressed (108). Like many other cancers, FAK overexpression contributes to metastasis as well as a poor patient prognosis in HCC (108). Overexpression of FAK is closely associated with portal venous invasion, a lower level of serum albumin, larger tumour size and poor survival (108).
Therapeutic insight of targeting Rho/Rho-kinase
Despite the significant contribution of metastasis towards HCC mortality, the only currently available therapy targeting HCC is the application of Sorafenib, which has been shown to improve the survival rate of advanced HCC patients for only 3 months (109). As highlighted in this review, the Rho/ROCK pathway confers HCC cells with a metastatic potential and hence acts as an attractive therapeutic target for metastatic HCC and recurrent HCC. Because ROCK is a key molecule that orchestrates the phosphorylation of a wide repertoire of cytoskeletal molecules, ROCK serves as an attractive therapeutic target to repress HCC metastasis. In addition to cancer metastasis, aberrant activation of ROCK may contribute to disastrous physiological consequences such as blood vessels constriction and extensive retraction of neuritis (110, 111). Therefore, the pharmaceutical industry has been making immense efforts in the search for ROCK-specific inhibitors in treating human diseases. The two most common groups of ROCK inhibitors developed are the pyridine and the isoquinoline derivatives. They are small molecule inhibitors competing for the ATP-binding sites of ROCK (110). Fasudil (HA1077) is the most widely used isoquinoline derivative. Fasudil was found to be a potent protein kinase inhibitor acting as a vessel dilator in treating cerebral vasospasm in 1993 and was later found to exert an inhibitory effect on ROCK activity (112). Fasudil has been marketed clinically in Japan since 1995. It is used in the clinical treatment of cerebral vasospasm, prolonged blood vessel constriction in the nervous system and other cardiovascular diseases. Recently, fasudil has been found to be effective in improving the prognosis and memory of Alzheimer's patients (113). Currently, the efficacies of fasudil in the treatments of atherosclerosis and hyperlipidaemia as well as Raynaud's phenomenon are being examined in clinical trials (ClinicalTrials.gov Identifier: NCT 00120718 and NCT 00498615). Y27632 is the most widely used pyridine derivative and was later discovered to show a more selective inhibitory effect towards ROCK as compared with fasudil (114); however, Y27632 is less studied in clinical trials. Y27632 profoundly suppresses metastatic tumour formation in liver cancer cells in rodent models and tumour recurrence in rats after liver transplantation (40, 83, 115). Importantly, fasudil and Y27632 are well tolerated and safe, without exerting a severe negative effect in patients and animals (40, 83, 115, 116). Until now, ROCK inhibitors have not been used in cancer treatment clinically and intensive investigation is warranted to validate the efficacy of ROCK inhibitors. As mentioned earlier, Y27632 profoundly decreased the occurrence of macroscopic and microscopic intrahepatic metastasis, repressed infiltrative tumour boundary and was well tolerated in animals; it is a good drug candidate for targeting metastatic and recurrent HCC. However, improvements should be made to devise more specific ROCK inhibitors to further prevent side effects from the drug upon prolonged intake. All current ROCK inhibitors do not specifically discriminate ROCK isoforms. The lack of specificity may be due to the lack of knowledge about the distinguished functions and structures of ROCK isoforms. Detailed studies on the crystal structures of ROCK1 and ROCK2 will be very helpful towards the understanding of the drug–target interaction and the intervention of more specific drugs.
Recently, there has been a surge of interest in a group of microRNAs (miRNAs) that downregulate gene expression through interaction with the 3′ untranslated region of their gene targets. miRNAs are found to control a diverse variety of biological mechanisms such as cell cycle control, cell proliferation and cell movement. It has been demonstrated recently that systemic delivery of anti-proliferative miRNA (miR-26a) via an adeno-associated viral vector (AAV) by a tail-vein injection suppressed myc-induced hepatocarcinogenesis in mice without any toxic effects to the mice (117). This study sheds light on the application of specific miRNA delivery as an attractive strategy for treating HCC. As the experimental advances in miRNAs study expanded, different groups have identified miRNAs targeting the RhoGTPase pathways in various cancer cell models, yielding a new layer of knowledge about the underlying mechanisms regulating the RhoGTPases pathway. Recently, an elegant study demonstrated that miR-31 inhibited breast cancer metastasis through suppression of RhoA (118). Whether miR-31 could act as an anti-metastatic miRNA in HCC and the consequences of anti-metastatic miRNAs systemic delivery in targeting advanced HCC are unknown and need to be examined. Nevertheless, the safety of an AAV systemic delivery platform in humans still remains the most important concern.
One of the most dangerous traits of cancer cells is their capability to metastasize. Unfortunately, the precise underlying mechanisms involved in this lethal feature of HCC remains to be elusive. So far, there is no promising curative therapy to target metastatic HCC. In this review, we have highlighted the molecular and clinical importance of RhoGTPases regulators, RhoGTPases, Rho-effectors and Rho-effector targets in HCC progression and metastasis. Although it is clear that the RhoGTPases/Rho-effector pathway is important to HCC metastasis, several uncertainties still exist about this pathway in HCC. For instance, the complexity of the cross talks between different RhoGTPases themselves and other signalling pathways, the activities of RhoGTPases and Rho-effectors in HCC tissues and the efficacy of drugs targeting this pathway in human HCC metastasis are typical unanswered questions. Furthermore, it will be intriguing to examine whether the expression and activities of RhoGTPases and Rho-effectors can act as indicators of recurrent and metastatic HCC. The recent burgeoning advances in animal imaging technology will remarkably facilitate the in vivo study of this signalling pathway and broaden our knowledge about HCC metastasis. Finally, although the review shows that current findings on RhoGTPases/Rho-effectors in HCC are not scarce, more effort should be made towards further characterization. More thorough understanding of the RhoGTPases/Rho-effectors pathway in the metastatic process is very much needed so that better therapeutic strategies can be devised to treat advanced HCC.
Financial support: The work was funded in part by Hong Kong Research Grants Council Collaborative Research Fund (HKU 7/CRF/09). I. O. L. Ng is Loke Yew Professor in Pathology.