Overexpression of eukaryotic initiation factor 5A2 enhances cell motility and promotes tumor metastasis in hepatocellular carcinoma


  • Potential conflict of interest: Nothing to report.


A high incidence of tumor recurrence and metastasis has been reported in hepatocellular carcinoma (HCC) patients; however, the underlying molecular mechanisms are largely unknown. In the present study a novel metastasis-related gene, eukaryotic initiation factor 5A2 (EIF5A2), was characterized for its role in HCC metastasis and underlying molecular mechanisms. Overexpression of EIF5A2 messenger RNA (mRNA) was detected in 50/81 (61.7%) of HCCs, which was significantly higher than those in nontumorous liver tissues. Compared with matched primary HCC, higher expression of EIF5A2 protein was observed in 25/47 (53.2%) of metastatic tumors. Functional studies found that ectopic expression of EIF5A2 could enhance cancer cell migration and invasion in vitro and tumor metastasis in vivo in an experimental mouse model. Moreover, inhibition of EIF5A by small interfering RNA (siRNA) or deoxyhypusine synthase (DHPS) inhibitor GC7, which inhibits EIF5A2 maturation, could effectively decrease cell motility. Further study found that EIF5A2 was able to induce epithelial-mesenchymal transition (EMT), a key event in tumor invasion and metastasis, characterized by down-regulation of epithelial markers (E-cadherin and β-catenin) and up-regulation of mesenchymal markers (fibronectin, N-cadherin, α-SMA, and vimentin). In addition, EIF5A2 could also activate RhoA/Rac1 to stimulate the formation of stress fiber and lamellipodia. Conclusion: EIF5A2 plays an important role in HCC invasion and metastasis by inducing EMT, as well as stimulating cytoskeleton rearrangement through activation of RhoA and Rac1. (HEPATOLOGY 2010.)

A worldwide increase in mortality associated with hepatocellular carcinoma (HCC) has recently been reported.1, 2 Clinical treatment of HCC remains challenging due to a high incidence of tumor recurrence. The main cause of death in HCC patients is intrahepatic metastasis but the underlying mechanisms are still not fully understood. It is generally believed that to give rise to a metastatic tumor, cancer cells from a primary site must complete all of the following steps: invasion, intravasation, survival and arrest in the blood stream, extravasation, and colonization at a new site. The motility of cancer cells, driven by the actin cytoskeleton network, has been well documented to play crucial roles at multiple steps during the metastasis process.

A previous study found that 17 genes were associated with metastasis by comparing the gene expression profiles of human adenocarcinoma metastases of multiple tumor types to unmatched primary adenocarcinomas.3 Primary tumors that carry this 17-gene signature were found to be associated with tumor metastasis and poor prognosis. Moreover, an independent study revealed that out of the 17 human metastasis-associated genes, 16 murine orthologs showed the same differential expression pattern in an experimental mouse model of cancer metastasis.4 One of the 17 genes in the metastasis signature is deoxyhypusine synthase (DHPS), an enzyme that is required for posttranslational hypusination. During hypusination, a specific lysine residue is converted into the rare amino acid hypusine. Previously, eukaryotic translation initiation factor (EIF5A) was thought to be the only substrate of DHPS. It has been implicated that EIF5A and DHPS play essential roles in cell viability, cell growth, and proliferation.5–9 In our previous studies, we isolated EIF5A2 from 3q26,10, 11 a frequently amplified region in cancer, as another candidate target of DHPS.

Human EIF5A2 shares 83% amino acid identity with EIF5A, which includes the highly conserved domain required for hypusination and the lysine-50 residue, where the posttranslational modification occurs.12 Amplification of 3q26.2, the chromosomal locus of EIF5A2, has been frequently detected in many solid tumors including ovarian,13 lung,14 esophageal,15 prostate,16 breast,17 and nasopharyngeal carcinomas.18 Interestingly, gain of 3q has also been associated with the recurrence of HCC.19 Our previous studies have demonstrated that ectopic expression of EIF5A2 leads to tumor formation in nude mice11 and overexpression of EIF5A2 is associated with tumor metastasis in colorectal carcinoma and poor prognosis in colorectal,20 ovarian,21 and bladder cancers.22 However, the implication of EIF5A2 in HCC metastasis has not been investigated. Moreover, the molecular mechanism underlying the role of EIF5A2 in cancer metastasis is unknown.

To investigate whether overexpression of EIF5A2 is associated with HCC metastasis, the level of EIF5A2 expression was compared between primary and matched metastatic HCCs. The effect of EIF5A2 on cell motility and invasiveness was investigated using both in vitro and in vivo assays. Furthermore, the molecular mechanism of EIF5A2 in tumor metastasis was also addressed in this study.


DHPS, deoxyhypusine synthase; EIF5A2, eukaryotic initiation factor 5A2; EMT, epithelial-mesenchymal transition; GC7, N-guanyl-1,7-diaminoheptane; HCC, hepatocellular carcinoma.

Patients and Methods

HCC Patients, Specimens, and Cell Lines.

Total RNAs were extracted from 81 HCC specimens (tumor and adjacent nontumorous tissues) collected at the Cancer Center of Sun Yat-sen University (Guangzhou, China). Clinical information was available from 45/81 patients. In all, 10/45 patients further developed metastasis, including nine intrahepatic metastases and one lung metastasis.

A total of 47 pairs of primary and matched metastatic HCC specimens, as well as 10 cases of nontumor surrounding tissues, were collected from archives of paraffin-embedded tissues at the Department of Pathology, Sun Yat-sen University (Guangzhou, China). Metastatic tumors included 28 intrahepatic (26 portal vein and two cholangiotube) and 19 extrahepatic metastases (12 peritoneum, four lymph nodes, one kidney, one adrenal cortex, and one bone metastasis).

All samples were anonymously coded according to the local ethical guidelines (as outlined by the Declaration of Helsinki), and informed consent was obtained from all patients. Cell line information is described in the Supporting Materials and Methods.

Plasmid Constructs and Transfection.

EIF5A-ORF and EIF5A2-ORF were polymerase chain reaction (PCR)-amplified and cloned into expression vector pcDNA3.1(+) (Invitrogen, Carlsbad, CA) and stable EIF5A2-expressing clones in LO2 cells were selected as described.11

Tissue Microarray (TMA) Construction and Immunohistochemistry (IHC) Staining.

A detailed protocol of TMA construction and IHC staining is described in the Supporting Materials and Methods. The monoclonal anti-EIF5A2 antibody showed no crossreactivity with EIF5A (Supporting Fig. S1). To evaluate IHC staining of EIF5A2, expression of EIF5A2 was scored as negative (total absence of staining), weak (faint staining in <50%, or moderate staining in <25% of tumor cells), moderate (moderate staining in ≥25% to <75%, or strong staining in <25% of tumor cells), and strong (moderate staining in ≥75%, or strong staining in ≥25% of tumor cells). In this study we characterize negative/weak expression of EIF5A2 as “normal expression” and moderate/strong expression of EIF5A2 as “overexpression.” Both staining intensity and percentage of positive cells were scored by two experienced pathologists.

Total RNA Extraction, Semiquantitative Reverse-Transcription (RT)-PCR, and Real-Time Quantitative PCR (qPCR).

Refer to the Supporting Methods for a detailed description of experiment protocols.

Western Blot.

Western blot analyses were performed with the standard protocol. Antibody information is listed in the Supporting Methods.

Wound-Healing Assay.

The wound-healing assay is described in the Supporting Methods.

Cell Migration and Invasion Assay.

The transwell cell migration assay and invasion assay were performed in polyethylene terephthalate (PET)-based migration chambers and BD BioCoat Matrigel Invasion Chambers (Becton Dickinson Labware, Franklin Lakes, NJ) with 8 μm porosity according to the manufacturer's instructions for 48 hours.

Experimental Metastasis Model.

The mouse model of experimental metastasis is described in the Supporting Methods.

RNA Interference (RNAi).

The RNAi experiment protocol is described in the Supporting Methods.

Immunofluorescence (IF) Staining.

The detailed protocol of IF is described in the Supporting Methods.

Rho-GTPase Activation Assay.

PAK1 PBD-agarose (for isolating Rac1-GTP and Cdc42-GTP) and rhotekin-agarose (for isolating Rho-GTP) (Upstate Biotechnology, Lake Placid, NY) were used to pull down the GTP-bound form of Rho-GTPase according to the manufacturer's manual. The levels of active Rac1, Cdc42, and RhoA were detected by western blot using specific polyclonal anti-Rac1, Cdc42, and RhoA antibody (Cell Signaling Technology, Beverly, MA).

Statistical Analysis.

Statistical analysis was performed with SPSS for Windows v. 13.0 (Chicago, IL). The two-tailed chi-squared test was used to analyze the association of EIF5A2 overexpression with different clinicopathological characteristics. Independent Student's t test was employed to assess the statistical significance between two preselected groups. Statistical significance was declared if P < 0.05.


EIF5A2 Overexpression Is Frequently Detected in HCC.

Semiquantitative RT-PCR and real-time qPCR methods were used to compare EIF5A2 messenger RNA (mRNA) expression between 81 pairs of primary HCC tumor and nontumorous surrounding tissues. Overexpression of EIF5A2 was detected in 50/81 (61.7%, P < 0.0001, independent Student's t test) of HCCs as compared to their nontumorous counterparts (Fig. 1A,B), indicating that EIF5A2 was frequently overexpressed in HCC. Among these 81 HCCs, detailed clinicopathological information was available for 45 cases. The association study found that overexpression of EIF5A2 was positively associated with tumor metastasis (P = 0.036, chi-square test) and negatively associated with tumor encapsulation formation (P = 0.020, chi-square test, Table 1), suggesting that EIF5A2 may play a role in HCC metastasis.

Figure 1.

Expression of EIF5A2 in HCC samples. (A) EIF5A2 mRNA levels were measured with qPCR in 81 paired HCC samples and their nontumor counterparts Gene expression results were normalized to internal control 18S rRNA. ***P < 0.0001; independent Student's t test. (B) Representatives of EIF5A2 expression in HCC specimens detected by RT-PCR. EIF5A2 was frequently up-regulated in tumor tissues compared with their paired nontumorous tissues. 18S rRNA was used as an internal control. (C) Expressions of EIF5A2 and EIF5A proteins in three immortalized liver cell lines (MIHA, LO2, and Chang Liver) and nine HCC cell lines were detected by western blot analysis. β-Actin was used as a loading control. (D) Representatives of EIF5A2 expression in nontumor, primary, and metastatic tissues detected by IHC (magnification ×100). In some cases of primary tumor, high-level expression of EIF5A2 was shown by strong staining at the edge of tumor (E) and in the tumor cells invading to surrounding tissue (F, indicated by arrows, magnification ×100).

Table 1. Clinicopathological Correlation of EIF5A2 mRNA Expression in HCC (n = 45)
Clinicopathological FeaturesNumberEIF5A2 OverexpressionP
  • *

    Hepatitis B surface antigen.

  • Tumor size was measured by the length of the largest tumor nodule.

  • Partial data not available; statistics based on available data.

 Female33 (100%) 
 Male4226 (61.9%)0.183
 ≤603424 (70.6%) 
 >60115 (45.5%)0.130
 Negative85 (6.25%) 
 Positive3724 (64.9%)0.899
Serum AFP (ng/mL)   
 ≤4002916 (55.2%) 
 >4001613 (81.2%)0.080
Tumor size (cm)   
 ≤51310 (76.9%) 
 >53219 (59.4%)0.080
 Absent128 (66.7%) 
 Present2314 (60.9%)0.736
Tumor encapsulation    
 Absent1714 (82.4%) 
 Present188 (44.4%)0.020
 Absent2513 (52.0%)0.036
 Present109 (90.0%) 
Microsatellite formation    
 Absent2616 (61.5%) 
 Present74 (57.1%)0.833
Tumor stage(AJCC)    
 Stage I75 (71.4%) 
 Stage II3220 (62.5%) 
 Stage III64 (66.7%)0.898

Western blot analysis was applied to determine protein expression level of EIF5A2 in 12 cell lines including three immortalized liver cell lines (MIHA, LO2, and Chang Liver) and nine HCC cell lines (H2P, H2M, HepG2, Hep3B, Huh7, BEL7402, QSG7701, QGY7703, and PLC8024). EIF5A2 was undetectable in all three immortalized liver cell lines, whereas high-level expression of EIF5A2 was observed in 6/9 of HCC cell lines, including H2P, H2M, Hep3B, Huh7, BEL7402, and PLC8024 (Fig. 1C). The expression level of EIF5A in these 12 cell lines was also examined and a similar level of expression was observed in all tested cell lines, suggesting that EIF5A2, rather than EIF5A, plays an oncogenic role in HCC development and progression.

EIF5A2 Overexpression Is Associated with HCC Metastasis.

To investigate the role of EIF5A2 in HCC invasion and metastasis, EIF5A2 expression was compared between primary and metastatic HCCs by immunohistochemistry using an HCC tissue microarray containing 47 pairs of HCC specimens. Each pair consisted of primary and metastatic lesions derived from the same patient. In all, 25 pairs of HCCs (53.2%) showed higher expression of EIF5A2 in metastatic lesions compared with their individually matched primary tumor samples. In a subset of primary tumors, EIF5A2 protein expression was already elevated (18/47, 38.3%). IHC staining of EIF5A2 protein in representative samples of nontumor, primary, and metastatic lesions are shown in Fig. 1D. Moreover, in some metastatic lesions we observed that the expression level of EIF5A2 was obviously higher at the edge of tumor (Fig. 1E) and in tumor cells invading the surrounding tissue (Fig. 1F, indicated by arrows).

Overexpression of EIF5A2 Promotes Cell Motility.

We have described LO2-EIF5A2, a stable liver cell line overexpressing EIF5A2.11 Overexpression of EIF5A2 in LO2-EIF5A2 cells was determined by RT-PCR and western blot (Fig. 2A). Because cell motility is an important factor regulating cancer invasion and metastasis, the effect of EIF5A2 on cell motility was characterized by wound-healing, transwell migration, and Matrigel invasion assays. The wound-healing assay showed that cell migration at the edge of exposed regions was remarkably faster in LO2-EIF5A2 cells than in vector-transfected cells (LO2-Vec) cells (Fig. 2B). Moreover, the expression level of EIF5A2 appeared to be higher at the edge of the wound in LO2-EIF5A2 cells (Supporting Fig. S3); however, it was less obvious than that observed in tumor samples (Fig. 1E,F). The transwell migration assay showed that overexpression of EIF5A2 led to a marked increase in cell motility, as more cells were observed migrating through the 8-μm pores in LO2-EIF5A2 compared with control LO2-Vec (P < 0.05, Fig. 2C). Similarly, the invasion assay showed that LO2-EIF5A2 cells obtained a significantly higher rate of cell invasion than that of control cells (P < 0.01, Fig. 2D). These data demonstrate that overexpression of EIF5A2 in LO2 cells enhanced cell motility.

Figure 2.

EIF5A2 enhances LO2 cell migration and invasion in vitro. (A) RT-PCR, western blot, and immunofluorescence staining (green: EIF5A2; blue: nuclear DAPI staining, magnification ×400) were used to detect EIF5A2 expression in EIF5A2-transfected LO2 cell lines (LO2-EIF5A2) and empty vector-transfected cells (LO2-Vec). 18S and β-actin were used as loading controls. (B) LO2-EIF5A2 cells showed higher motility in a wound-healing assay, 48 hours posttreatment. (C) Effect of EIF5A2 overexpression on cell migration in a transwell assay. Examples of cells migrated through the PET-membrane (pore size: 8 μm) are shown in the left panel. Columns: mean of triplicate experiments; *P < 0.05; independent Student's t test. (E) Effect of EIF5A2 overexpression on cell invasion through Matrigel. Examples of cells migrated through Matrigel-coated transwell are shown in the left panel (magnification ×100). Columns: mean of triplicate experiments; **P < 0.01; independent Student's t test.

Overexpression of EIF5A2 Leads to Cancer Metastasis in an Experimental Mouse Model.

To test whether EIF5A2 overexpression is causative in an experimental metastasis model, we injected LO2-EIF5A2 cells into the tail vain of SCID mice; LO2-Vec were used as control (five mice per group). Mice were sacrificed 6 weeks after cell injection and metastatic tumor nodules formed in the lung and in the liver were examined. No tumor nodules were detected in the lung in any mice examined. However, overexpression of EIF5A2 increased liver metastasis by 2-fold, as shown in Fig. 3A. Interestingly, higher-level expression of EIF5A2 was also observed in cancer cells invading the surrounding tissue as described before (Fig. 3B, indicated by arrows).

Figure 3.

EIF5A2 overexpression leads to increased live metastasis in an experimental mouse model. (A) Metastatic nodules on the surface of liver are shown in the left panel (indicated by arrows). Number of nodules formed in the liver of SCID mice were compared 6 weeks after tail vein injection of LO2-EIF5A2 or LO2-Vec cells (five mice per group; *P < 0.05; independent Student's t test). (B) Detection of EIF5A2 by IHC staining in two liver nodule samples originated from LO2-EIF5A2-injected mice. Higher expression level of EIF5A2 was observed at the edge of tumor and cells invading into surrounding tissue (indicated by arrows, magnification ×100).

Silencing EIF5A2 Expression by Small Interfering RNA (siRNA) Inhibits Cell Motility.

We next studied whether endogenous EIF5A2 is important for cancer cell motility. High-level EIF5A2 expression was detected in several liver cancer cell lines including H2M (Fig. 1C), a metastatic liver cancer cell line established from metastatic lesion of a liver cancer patient.23 We evaluated the effect of EIF5A2 silencing by RNAi on H2M cell migration. Compared with scrambled siRNA (siSCR), treatment with specific siRNA against EIF5A2 (siEIF5A2) resulted in about 80% silencing of EIF5A2 in H2M cells at both mRNA and protein levels, whereas EIF5A remained unaffected (Fig. 4A,B). Further study showed that EIF5A2 knockdown could significantly inhibit cell migration in H2M cells (Fig. 4C, P < 0.05).

Figure 4.

Silencing of EIF5A2 expression suppresses H2M cell migration. EIF5A2 was efficiently silenced by the treatment of siEIF5A2, as determined by RT-PCR (A) and western blot (B), whereas EIF5A was not affected. Corresponding siSCR was used as negative control for siRNA treatment; 18S and β-actin were used as loading controls. (C) Cell migration assay was used to compare the frequency of migratory cells between H2M cells treated with siSCR and siEIF5A2. (D) Combined treatment of GC7 (200 μM) and siEIF5A2 (20 nM) abolished H2M cells migration. Columns: mean of triplicate experiments; *P < 0.05; **P < 0.01; independent Student's t test.

Combined Treatment of DHPS Inhibitor GC7 and siEIF5A2 Abrogates Cell Motility in a Synergistic Manner.

Posttranslational hypusination, which is mediated by DHPS, is required for EIF5A to function properly.5, 8 We speculated that this would also be an essential maturation step for EIF5A2 due to their high level of sequence homology, especially at the region of hypusine modification.12 It is therefore expected that inhibiting the maturation of EIF5A2 by DHPS inhibitor N1-guanyl-1,7-diaminoheptane (GC7) could inhibit the effect of EIF5A2 on cell motility. Indeed, a reduction in cell motility was observed in H2M cells treated with 200 μM GC7 for 16 hours (Fig. 4D); however, the effect was not as profound as that seen in cells treated with siEIF5A2. Although GC7 could potentially pose its effect on both EIF5A and EIF5A2, under the dose and treatment duration mentioned above, EIF5A in H2M cells was not affected (data not shown). Notably, in cells treated with both GC7 and siEIF5A2 cell migration was almost completely abrogated (Fig. 4D), suggesting that hypusination is essential for EIF5A2 to exert its role in cell motility.

EIF5A2 Induces Epithelial-Mesenchymal Transition (EMT).

Morphological changes were observed in EIF5A2 overexpressing LO2 cells, suggesting that these cells may undergo EMT. LO2-Vec cells maintained highly organized cell-cell adhesion; however, when plated at the same cell density, LO2-EIF5A2 cells exhibited a cell scattering phenomenon and loss of cell-cell contact, accompanied by the spindle-shaped, fibroblastic morphology (Fig. 5A). To further demonstrate this phenotype in EIF5A2 overexpressing cells, western blot and IF analysis were carried out. In LO2-EIF5A2 cells, expression of E-cadherin and β-catenin decreased, whereas the expression of α-catenin remained unaltered; on the other hand, all mesenchymal markers tested, including fibronectin, N-cadherin, vimentin, and α-smooth muscle actin were elevated in LO2-EIF5A2 cells. These data reinforced that EIF5A2 overexpression may induce EMT (Fig. 5B,C). The western blotting results were confirmed by IF analysis (Fig. 5D). In particular, vimentin intermediate filaments localized in a concentrated and polarized pattern in LO2-Vec cells; however, in LO2-EIF5A2 cells a network of vimentin intermediate filaments was clearly visible. In addition, the mesenchymal marker fibronectin was completely undetectable in LO2-Vec cells, whereas LO2-EIF5A2 cells showed positive staining of fibronectin in the cytoplasm. Taken together, these results strongly suggested that EMT may be activated by EIF5A2 overexpression.

Figure 5.

EIF5A2 promotes metastasis by inducing EMT. (A) Cell morphology of LO2-EIF5A2 and LO2-Vec cells. Expressions of epithelial markers E-cadherin, α-catenin, and β-catenin (B) and mesenchymal markers fibronectin, N-cadherin, vimentin, and α-smooth muscle actin (C) were compared by western blot analysis between LO2-EIF5A2 and LO2-Vec cells. β-Actin was used as a loading control. (D) IF was used to compare expression level/pattern of epithelial markers and mesenchymal markers between LO2-EIF5A2 and LO2-Vec cells. Epithelial markers α-catenin (red cytoplastic signal) remained unchanged, whereas β-catenin (red membrane signal at cell border) was down-regulated in LO2-EIF5A2 cells; mesenchymal markers fibronectin (green cytoplastic signal), vimentin (green cytoplastic signal), and N-cadherin (red cytoplastic signal) were up-regulated in LO2-EIF5A2 cells (magnification ×400).

EIF5A2 Overexpression Activates Rho/Rac GTPases.

During tumor invasion and metastasis, changes in Rho-GTPase activity will lead to subsequent reorganization of actin cytoskeleton, which in turn causes disruption of adherent junctions. Because LO2-EIF5A2 cells displayed morphological alterations, we further investigated whether EIF5A2 could modulate cytoskeleton rearrangement and activation of Rho-GTPase. F-actin staining revealed that stress fiber and lamellipodia were observed in LO2-EIF5A2 cells but not in control LO2-Vec cells (Fig. 6A). Conversely, EIF5A2 knockdown by siRNA resulted in loss of stress fiber in H2M cells (Fig. 6B). These findings suggested that the migratory phenotype induced by EIF5A2 may be associated with activation of Rho-GTPases. To test this possibility, pull-down assays were used to quantify the amount of the GTP-bound active form of RhoA, Rac1, and Cdc42. Despite a higher level of RhoA in control cells, the active form of RhoA could only be detected in LO2-EIF5A2 cells (Fig. 6C). Similarly, a higher level of active Rac1 was observed in LO2-EIF5A2 cells compared with LO2-Vec cells. Cdc42 was barely detectable in either LO2-Vec or LO2-EIF5A2 cells; no active form was observed (Fig. 6C). These results indicated that EIF5A2 could activate small GTPase Rac1 and RhoA; however, the precise mechanism underlying EIF5A2-mediated Rho-GTPase activation requires further investigation.

Figure 6.

EIF5A2 induces the formation of stress fiber and lamellipodia through activation of RhoA and Rac1. (A) Stress fiber and lamellipodia (indicated by arrows) were formed in LO2-EIF5A2 cells, but not in LO2-Vec cells. (B) Stress fiber formation was inhibited in H2M cells treated with siEIF5A2, but unaffected in cells treated with control siSCR. (C) Total and active forms of Rho-GTPases, including RhoA, Rac1, and Cdc42, were compared between LO2-EIF5A2 and LO2-Vec cells by western blot analysis. GTP-bound (active) forms of RhoA, Rac1, and Cdc42 were pulled down and detected by western blot analysis using corresponding antibodies. Active forms of RhoA and Rac1 were much higher in LO2-EIF5A2 than that in LO2-Vec cells (magnification ×400).


The vast majority of cancer deaths result from cancer metastasis, rather than the influence of the primary tumors. In patients with HCC, the early stages of the disease are usually asymptomatic; in addition, the incidence of tumor recurrence is high. As a result, the 5-year survival rate for HCC patients is poor and most patients die of intrahepatic metastasis. A better understanding of the molecular events governing the pathogenesis of cancer metastasis in HCC is highly desirable for improvement of clinical management. Recently, overexpression of EIF5A2 have been associated with metastasis in multiple cancer types, including colon,20 ovarian,21 and bladder cancer.22 In this study, we first demonstrated that EIF5A2 was frequently overexpressed in HCC, which was associated with the metastatic state of cancer progression. Interestingly, the invasive border between tumor and nontumorous tissues showed a higher level of EIF5A2 expression, indicating that this oncoprotein may contribute to a more malignant and invasive phenotype of the cancer cells.

A series of in vitro and in vivo assays were carried out to characterize the role of EIF5A2 in regulating liver cancer cell motility and invasiveness, and the results showed that overexpression of EIF5A2 could significantly enhance cell motility and invasiveness. In the tail-vein-injection mouse model of cancer metastasis, overexpression of EIF5A2 led to a significant increase in the number of lesions of liver metastasis. Again, we observed a higher level of EIF5A2 on the tumor margin with an aggressive front. In addition, gene silencing of EIF5A2 by siRNA or disruption of posttranslational hypusination inhibited its effect on cell migration. All these findings strongly supported that overexpression of EIF5A2 played an important role in HCC invasion and metastasis.

In the present study we found that overexpression of EIF5A2 had a significant impact on EMT, as shown by increased expression of mesenchymal markers (fibronectin, N-cadherin, vimentin, and α-SMA) and decreased expression of epithelial markers (E-cadherin and β-catenin). EMT is a key event in tumor invasion and metastasis in which epithelial cells lose epithelial adherence and tight junction proteins, lose polarity and cell-cell contacts, and undergo a remarkable remodeling of the cytoskeleton to facilitate cell motility and invasion.24–26 Thus, HCC cells overexpressing EIF5A2 probably undergo EMT to achieve higher motility and invasiveness.

The role of Rho small GTP binding proteins in the regulation of actin cytoskeleton organization and cell migration has been well documented.27–29 Actin filaments are essential for the maintenance of cytoskeleton networks that determine cell shape and cell motility. Our study, for the first time, provided evidence supporting the role of EIF5A2 in the regulation of cytoskeleton through a Rho-GTPase signaling pathway. Our results demonstrate that ectopic expression of EIF5A2 in LO2 cells stimulated the formation of stress fiber (through the activation of RhoA) and lamellipodia (through the activation of Rac1) without affecting actin expression level in the cells (Supporting Fig. S2). Furthermore, in metastatic H2M cells that express a high level of EIF5A2 endogenously, treatment of specific siRNA against EIF5A2 suppressed stress fiber formation. The level of total RhoA and Rac1 remained the same between LO2-EIF5A2 and LO2-Vec cells, indicating that EIF5A2 did not affect the expression of Rho/Rac GTPases. Rho-GTPase activity is determined by the amount of the GTP-bound form, which is regulated by RhoGAP (Rho-GTPase activating protein), RhoGEF (Rho-GTPase guanine exchange factors), as well as RhoGDI (Rho GDP dissociation inhibitor). Thus, EIF5A2 may target members of GEF, GAP, or GDI to increase GTP-bound Rho in EIF5A2 overexpressing cells. Further study is needed to identify which Rho regulator is targeted by EIF5A2 that leads to Rho-GTPase activation.

A previous study has implicated DHPS as one of the metastasis signature genes.3 To date, EIF5A and EIF5A2 are the only known proteins that require posttranslational modification by DHPS. It is therefore a logical hypothesis that one or both of the proteins may be involved in the pathogenesis of cancer metastasis. EIF5A and EIF5A2 share 83% amino acid identity. Although the lethal effect of EIF5A disruption could be partially rescued by EIF5A2, evidence is accumulating that they are not functionally redundant.30 EIF5A is ubiquitously expressed at a high level in all tissues, whereas EIF5A2 is normally undetectable except in testis and brain.12 Moreover, amplification and overexpression of EIF5A2 was frequently detected in various malignancies,11, 20–22 indicating its unique role in cancer development and progression. In our HCC sample set analyzed for EIF5A2 mRNA expression, 43 cases (23 of which with overexpression of EIF5A2) were also analyzed for potential EIF5A2 gene copy number change. We did not detect any significant copy number change in these HCC samples using semiquantitative genomic PCR (data not shown), suggesting that EIF5A2 gene amplification may not be the major reason for its overexpression found in the current study. However, we could not rule out the possibility that small copy number changes may exist that were not detected by the assay used. Although DHPS was suggested to be a metastatic signature gene,3 overexpression of DHPS was not reported in HCC. In this study we screened the DHPS mRNA expression status in 45 HCCs by qPCR. The result showed that overexpression of DHPS (fold change >2) was detected in 6/45 (13.3%) of HCCs, indicating that EIF5A2 overexpression frequently detected in HCC could not be explained by the DHPS expression level alone. Other factors yet to be defined may also be involved in the regulation of EIF5A2 expression at the level of transcription, translation, and posttranslational modification.

Taken together, our data demonstrate that EIF5A2 plays a pivotal role in HCC development and metastasis by way of enhancing cell motility and invasiveness, regulating cytoskeleton, and activating EMT. These findings will not only support EIF5A2 as an important biomarker for cancer diagnosis, but also provide insights for the development of novel anticancer therapies.