p28gank overexpression accelerates hepatocellular carcinoma invasiveness and metastasis via...

p28GANK Overexpression Accelerates HepatocellularCarcinoma Invasiveness and Metastasis via

Phosphoinositol 3-Kinase/AKT/Hypoxia-InducibleFactor-1a Pathways

Jing Fu,1* Yao Chen,1* Jie Cao,1* Tao Luo,1 You-Wen Qian,1 Wen Yang,1 Yi-Bin Ren,1 Bo Su,1 Guang-Wen Cao,2

Yuan Yang,3 Yi-Qun Yan,3 Feng Shen,3 Meng-Chao Wu,3 Gen-Sheng Feng,1,4 and Hong-Yang Wang1,5

The overall survival of patients with hepatocellular carcinoma (HCC) remains poor, and themolecular mechanisms underlying HCC progression and aggressiveness are unclear. Here, wereport that increased expression of p28GANK (Gankyrin, PSMD10, or p28) in human HCCpredicts poor survival and disease recurrence after surgery. Patients with HCC who have largetumors, with vascular invasion and intrahepatic or distant metastasis, expressed high levels ofp28GANK. Invasive tumors overexpressing p28GANK were featured by active epithelial-mesen-chymal transition (EMT), and exhibited increased angiogenesis associated with vascular endo-thelial growth factor overexpression, whereas silencing p28GANK expression attenuated EMTand motility/invasion of tumor cells. The p28GANK activates phosphoinositide 3-kinase(PI3K)–V-akt Murine Thymoma Viral Oncogene Homolog (AKT)–hypoxia-inducible factor1a (HIF-1a) signaling to promote TWIST1, vascular endothelial growth factor, and metallo-proteinase 2 expression. Suppression of the PI3K–AKT–HIF-1a pathway interfered withp28GANK-mediated EMT and invasion. Consistently, we detected a significant correlationbetween p28GANK expression and p-AKT levels in a cohort of HCC biopsies, and the combi-nation of these two parameters is a more powerful predictor of poor prognosis. Conclusion:These results present novel mechanistic insight into a critical role of p28GANK in HCC pro-gression and metastasis. (HEPATOLOGY 2011;53:181-192)

Hepatocellular carcinoma (HCC) is the thirdleading cause of cancer death worldwide, andthe second in China.1,2 HCCs that grow rap-

idly with early vascular invasion are highly resistant tochemotherapy.3-5 The extremely poor prognosis ofpatients with HCC is largely due to the high frequencyof tumor recurrence or distant metastasis after surgicalresection.6 Extensive epidemiological studies have iden-tified major risk factors of HCC, and significant advan-ces have been made in understanding the pathogenesis.7

However, little is known about molecular mechanisms

underlying recurrence or metastasis. Therefore, a mostcritical issue is to search for molecular markers relatedto metastasis, which will provide new targets for inter-vention of metastatic recurrence of HCC.p28GANK (also known as gankyrin, PSMD10, or

p28) was identified as an oncoprotein that is frequentlyoverexpressed in human liver cancers.8,9 Up-regulationof p28GANK correlates well with cell cycle progression inhuman hepatocytes.10,11 Gankyrin overexpression con-fers tumorigenicity to NIH3T3 cells and inhibits apo-ptosis in cultured human tumor cells exposed to

Abbreviations: CRC, colorectal cancer; DFS, disease-free survival; EMT, epithelial mesenchymal transition; ESCC, esophageal squamous cell carcinoma; GFP,green fluorescent protein; HCC, hepatocellular carcinoma; MMP, matrix metalloproteinase; NF-jB, nuclear factor jB; OS, overall survival; PVTT, portal veintumor thrombus; qRT-PCR, quantitative reverse transcription polymerase chain reaction; SEM, standard error of the mean; siRNA, small interfering RNA; TMA,tissue microarray; TNM, tumor-node-metastasis; VEGF, vascular endothelial growth factor.From the 1International Cooperation Laboratory on Signal Transduction, Eastern Hepatobiliary Surgery Institute/Hospital, Shanghai, China; 2Department of

Epidemiology, Second Military Medical University, Shanghai, China; 3Department of Surgery, Eastern Hepatobiliary Surgery Hospital, Shanghai, China; 4Departmentof Pathology, School of Medicine, and Section of Molecular Biology, Division of Biological Sciences, University of California San Diego, La Jolla, CA; and 5State KeyLaboratory of Oncogenes and Related Genes, Cancer Institute of Shanghai Jiao Tong University, Shanghai, China.Received June 7, 2010; accepted September 20, 2010.Supported by grants for Creative Research Groups, National Natural Science Foundation of China (30921006, 81071778), the State Key Project for Infectious Diseases

(2008ZX10002, 2006AA02A310, and 2007CB914502), and Key Program of Basic Research of Shanghai, China (10JC1418500, 08ZR1405500, and 06Z059-08Z20).*These authors contributed equally to this work.

181

chemotherapeutic agents.8,12 The tumorigenic effect ofp28GANK might be associated with its antiapoptoticproperty,11–14 and down-regulation of p28GANK-induced apoptosis inhibits tumor growth.11,12,14 Theantiapoptotic activity is attributable, at least in part, toincreased degradation of p53, resulting in reducedexpression of p53-dependent proapoptotic genes.12

Additionally, p28GANK was shown to bind v-rel reticulo-endotheliosis viral oncogene homolog A (RelA)/nuclearfactor jB (NF-jB) and suppress NF-jB activity.15,16

Therefore, p28GANK may play a complex role in hepato-carcinogenesis, which is yet to be elucidated.In this study, we extensively investigated p28GANK

expression pattern and determined its contribution toHCC invasion and metastasis. We also dissected themolecular mechanisms by which p28GANK mediatestumor metastasis. Results presented here suggest thatp28GANK overexpression promotes HCC aggression viamodulation of phosphoinositide 3-kinase (PI3K)–v-aktmurine thymoma viral oncogene homolog 1 (AKT)–hypoxia-inducible factor-1a (HIF-1a) signaling. Wepropose that combination of p28GANK/p-AKT is anew powerful predictor for HCC recurrence and me-tastasis and a new potential target for adjuvant treat-ment of aggressive HCCs after surgical resection.

Patients and Methods

Detailed description of Patients and Methods can befound in the online Supporting Information.

Results

Elevated Expression of P28GANK in HCC CellLines With Metastatic Potential and in InvasiveHCC Specimens. We first examined the p28GANK pro-tein amounts in several HCC cell lines, with varying met-astatic capability.17,18 The p28GANK protein levelsincreased progressively from normal liver cells (HL-7702and QSG-7701), low metastatic SMMC-7721 andMHCC-97L cells, to highly metastatic HepG2 andHCC-LM3 cells (Fig. 1A; Supporting Information Fig.1A). Quantitative reverse transcription polymerase chainreaction (qRT-PCR) analysis showed that p28GANK

expression was significantly higher in invasive HCC sam-ples than in normal liver tissue or noninvasive HCCtumors (28.4-fold and 9.3-fold, respectively; Supporting

Information Fig. 1B). Immunoblotting of protein extractsfrom the same set of patients’ samples confirmed the asso-ciation of p28GANK overexpression with features of tumormetastasis (Fig. 1B; Supporting Information Fig. 1C),suggesting p28GANK involvement in HCC aggressiveness.Portal vein tumor thrombus (PVTT) from HCC

notably deteriorates hepatic function and serves as a poorprognostic factor associated with frequent recurrence andintrahepatic metastasis.19 By analyzing an additional 10pairs of HCC samples, we found that p28GANK proteinlevels were low in normal liver tissue, relatively higher inprimary HCCs by 2.1-fold, and further increased inPVTT by 3.3-fold (Fig. 1C; Supporting Information Fig.1D), further suggesting its potential role not only at theorigin but also in invasive progression of HCCs.p28GANK Overexpression in HCCs Predicts a Poor

Prognosis. We next sought to determine whetherp28GANK expression in HCC is associated with diseaserecurrence and poor survival. Tissue microarrays from201 patients with HCC who underwent liver resection(Supporting Information Table 1) were examined by im-munostaining with p28GANK monoclonal antibody. Theaverage expression level of p28GANK protein was signifi-cantly higher in HCC tissues than in peritumoral tissues(Supporting Information Table 2). Intriguingly, thep28GANK expression levels were found to be significantlyhigher in HCCs with increased tumor size (P ¼ 0.002),vascular invasion (P ¼ 0.0366), and intrahepatic (P ¼0.0429) and distant metastasis (P < 0.0001; Table 1).Based on the immunohistochemical result, all 201patients with HCC were divided into two groups: thehigh-expression (n ¼ 100) and low-expression group (n¼ 101). Patients in the high expression group had eithershorter disease-free survival (DFS, median DFS timewere 11 and 32 months, respectively, difference ¼ 21month, P ¼ 0.002) or worse overall survival (OS, me-dian OS time were 28 and >49 months, respectively,difference >21 month, P ¼ 0.007) (Fig. 1D,E; Sup-porting Information Table 3) than the low-expressiongroup. Consistently, the 1-year, 3-year, and 5-year OSand DFS after surgery were much worse for p28GANK-high than p28GANK-low expression group (Table 2).Thus, p28GANK expression is a valuable predicting factorfor recurrence and survival in patients with HCC.p28GANK Enhances Invasive Potential of HCC

Cells In Vitro. To determine the significance of theabove clinical data, we generated lentiviral constructs

Address reprint requests to: Hong-Yang Wang, MD, Laboratory of Signal Transduction, EasternHepatobiliarySurgery Institute/Hospital, 225 Changhai Road,Shanghai 200438, China. E-mail: [email protected]; fax: 86 21 65566851.CopyrightVC 2010 by the American Association for the Study of Liver Diseases.View this article online at wileyonlinelibrary.com.DOI 10.1002/hep.24015Potential conflict of interest: Nothing to report.Additional Supporting Information may be found in the online version of this article.

182 FU ET AL. HEPATOLOGY, January 2011

expressing p28GANK (LV-p28GANK) to infect low-invading MHCC-97L cells. p28GANK overexpressionsignificantly enhanced their invasive capacity by 2.2-fold, as compared with lentiviral–green fluorescentprotein (LV-GFP)-treated cells (Fig. 2A), and enhancedadhesion to several cell matrix proteins (Supporting In-formation Fig. 2A). In contrast, silencing endogenousp28GANK with lentivirus-mediated microRNA (LV-mip28GANK) in HCC-LM3 cells significantly reducedcell invasion by 58% (Supporting Information Fig.2B), and resulted in inhibition of adhesion to cell ma-trix proteins (Supporting Information Fig. 2C). Evi-dently, p28GANK acts to promote the invasive propertyof HCC cells in vitro.p28GANK Promotes Localized Growth, Angiogene-

sis, and Metastasis of HCC In Vivo. We furtherexamined the effect of p28GANK on HCC growth andpulmonary metastasis by establishing an orthotopic liver

tumor model in nude mice. MHCC-97L cells havinglow metastatic potential were infected with LV-GFP orLV-p28GANK, and used for orthotopic model studies.Compared to mock control or LV-GFP groups,p28GANK overexpression resulted in significant increaseof tumor size, the number of pulmonary metastatic foci,increased average size of pulmonary metastatic lesions,and enlarged average microvessel density (Fig. 2B-D).Furthermore, the subcutaneous xenograft model basedon low-metastatic potential SMMC-7721 cells alsoshowed that p28GANK overexpression promoted tumorcell proliferation and lung metastasis, while suppressingretinoblastoma 1 (RB) expression (Supporting Informa-tion Fig. 3A-D), consistent with previous reportson p28GANK promotion of RB phosphorylation anddegradation.9,14 On the contrary, down-regulation ofp28GANK expression by LV-mip28GANK retardedgrowth of HepG2-derived or HCCLM3 cell–derived

Fig. 1. Increased levels of p28GANK indicate worsening prognosis and recurrence/metastasis of HCC. (A) p28GANK expression was evaluated inthe indicated cell lines. (B) Protein levels of p28GANK were determined in 10 normal liver (Norm), 12 noninvasive HCC (Non-inv, without vascularinvasion), and 15 invasive HCC (Inv, with vascular invasion) samples. Box plot graph shows the statistical analysis of p28GANK expression in allsamples. The expression of p28GANK was normalized against b-actin. (C) Protein levels of p28GANK in 10 pairs of HCC tissue samples (Normal,normal liver; Primary, primary HCC; PVTT, portal vein tumor thrombus). Box plot graph shows the quantitative results of p28GANK expression. (D,E)HCC samples in a tissue microarray were immunostained with a monoclonal anti-gankyrin antibody. Representative low-p28GANK (L: below themedian value) and high-p28GANK expression (H: above the median value) samples are shown (�200). The disease-free and overall survival ratesof 201 patients with HCC were compared between the low-p28GANK and high-p28GANK groups. [Color figure can be viewed in the online issue,which is available at wileyonlinelibrary.com.]

HEPATOLOGY, Vol. 53, No. 1, 2011 FU ET AL. 183

tumors and reduced pulmonary metastases dramatically(Supporting Information Fig. 3E,F). Together, theseresults reveal functional significance of increasedp28GANK expression in metastatic HCC and in aggres-sive tumors with high propensity to develop recurrenceafter surgery.

p28GANK Promotes Tumor Metastasis by Enhanc-ing EMT. Given that p28GANK promotes HCC metasta-sis, we investigated the effect of p28GANK on EMT, a criti-cal event in tumor invasion. p28GANK overexpression byLV-p28GANK resulted in morphologic changes fromtightly packed colonies to scattered growth structure (Fig.3A), suggesting induction of EMT. We performed qRT-PCR for molecular markers of EMT. The epithelialmarkers such as E-cadherin, cytokeratin-8, cytokeratin-17,cytokeratin-18, claudin-1, and claudin-8 were lower inthe LV-p28GANK group than that in the LV-GFP group,whereas the mesenchymal markers such as N-cadherin,vimentin, HEY1 (hairy/enhancer-of-split related withYRPW motif 1), HEY2, Jagged 1, Jagged 2, Goosecoid,and the EMT major regulator TWIST1 were increased(Fig. 3B). Immunoblotting also detected lower expressionof E-cadherin in MHCC-97L-LV-p28GANK cells but anincrease in HCC-LM3-LV-mip28GANK cells. In contrast,the expression of N-cadherin, vimentin, and TWIST1increased in MHCC-97L-LV-p28GANK cells but decreasedin HCC-LM3-LV-mip28GANK cells (Fig. 3C). However,no alteration in other EMT inducers, such as Snail, Slug,and SIP1 (survival of motor neuron protein interactingprotein 1), was observed in MHCC-97L-LV-p28GANK orHCC-LM3-LV-mip28GANK cells (Supporting Informa-tion Fig. 4A). We next investigated the occurrence ofEMT in vivo. LV-p28GANK tumors exhibited the typicalEMT phenotype, including focal loss of the epithelialmarker E-cadherin, translocation of b-catenin (dissocia-tion of membranous b-catenin and translocation into thenucleus), and concurrent gain of the mesenchymal markervimentin and N-cadherin (Fig. 3D). Thus, p28GANK over-expression induced oncogenic EMT in HCC in vivo.We asked whether TWIST1 is involved in p28GANK-

induced E-cadherin down-regulation. E-cadherin expres-sion could be rescued by silencing TWIST1 in SMMC-7721-LV-p28GANK or MHCC-97L-LV-p28GANK cells

Table 1. Relationship Between Intratumoral p28GANK

Expression and Clinicopathologic Features

Variable

p28GANK Density

P ValueLow(n 5 101) High(n 5 100)

Sex NS

Male 92 87

Female 9 13

Age(years) NS

� 50 50 54

> 50 51 46

Tumor size(cm) 0.002

� 5 57 30

> 5 44 70

AFP NS

� 20 U/L 25 28

>20 U/L 76 72

� 1000 U/L 66 57

>1000 U/L 35 43

HBsAg NS

Positive 83 89

Negative 18 11

Anti-HCV NS*

Positive 5 2

Negative 96 98

Liver cirrhosis NS

Yes 63 51

No 38 49

Vascular invasion 0.0366

Yes 52 66

No 49 34

Intrahepatic metastasis 0.0429

Yes 18 30

No 83 70

Capsular formation NS

Yes 70 59

No 31 41

Capsular invasion NS

Yes 25 24

No 76 76

Distant metastasis <0.0001

Yes 15 41

No 86 59

TNM stage NS

I-II 81 72

III-IV 20 28

Edmondson NS

I-II 11 8

III-IV 90 92

Patients with HCC were divided into p28GANK ‘‘High’’ group (whose final den-

sity was higher than the median) and ‘‘Low’’ group (whose final density was

lower than the median). The patient and disease profiles in each group were

compared.

*Fisher’s exact test. AFP, serum alpha-fetoprotein; anti-HCV, anti-hepatitis C

virus antibody; HBsAg, hepatitis B surface antigen; NS, not significant between

any groups; TNM, tumor-node-metastasis.

Table 2. Relationship Between Intratumoral p28GANK

Expression and Survival Rate

Survival Measurement

p28GANK Density

P Value*Low High

1-year overall survival (%) 83.2 6 3.7 70.0 6 4.6 0.007

3-year overall survival (%) 65.2 6 4.8 44.8 6 5.0

5-year overall survival (%) 53.7 6 5.7 39.2 6 5.6

1-year disease-free survival (%) 69.3 6 4.6 46.0 6 5.0 0.002

3-year disease-free survival (%) 45.8 6 5.0 27.8 6 4.5

5-year disease-free survival (%) 39.0 6 6.1 24.6 6 4.6

Disease-free survival is defined as a period without recurrence, the diagnosis

of which is based on the typical feature presented in a computed tomography/

magnetic resonance imaging scan and an elevated serum alpha-fetoprotein

level.

*Log-rank test.


(Fig. 3E), suggesting that TWIST1 is required forp28GANK-driven EMT.HIF-1a Is Essential for p28GANK-Mediated

Vascular Endothelial Growth Factor and MatrixMetalloproteinase 2 Expression. Given that orthotopicintrahepatic implantation of MHCC97L-LV-p28GANK

cells generated aggressive and highly vascularized tumors,we examined the expression levels of vascular endothelialgrowth factor (VEGF) and matrix metalloproteinases

(MMPs). Compared with vector control, the proteinamounts of both VEGF and MMP2, but not MMP9,were significantly increased in LV-p28GANK groups butdecreased in LV-mip28GANK group (Fig. 4A, upper). Gel-atin zymography assay showed that MMP2 activityincreased in MHCC-97L-LV-p28GANK cells but decreasedin HCC-LM3-LV-mip28GANK cells (Fig. 4A, lower).HIF-1a plays a pivotal role in promoting angiogen-

esis, through regulation of target genes including

Fig. 2. Overexpression of p28GANK in HCC cells promotes invasiveness in vitro and accelerates proliferation and metastasis in vivo. (A) The numberof invasive MHCC-97L cells infected with lentiviral-delivering GFP (LV-GFP) or p28GANK (LV-p28GANK) was calculated with crystal violet staining asdescribed in Patients and Methods, and represented the average count of six random microscopic fields (magnification, �100). Bar graphs showmean 6 standard error of the mean (SEM) performed in triplicate compared with mock cells. p28GANK expression was analyzed by immunoblot. (B)Photomicrographs were taken for liver tumors and lung metastasis in nude mice, 4 weeks after orthotopic xenograft transplantation of mock, LV-GFP–infected or LV-p28GANK–infected MHCC-97L cells. Representative images of a mouse in each group were presented. Arrowheads indicate the metastaticnodes. Tumor volumes from each group were measured as described in Patients and Methods. (C) (Upper) Representative lung tissue sections fromeach group were shown (hematoxylin and eosin stain; magnification, �100). Black arrows indicate lung metastatic tumors. (Lower) The number oflung metastatic foci in each group was calculated. Classification of metastatic foci was based on the cell number in the metastatic lesion. (D) Immu-nostaining of p28GANK and CD31 was performed, and the microvessel density (MVD) was calculated in tumors from each group. Results display themean 6 SEM from triplicate experiments. *P < 0.05, **P < 0.001 versus mock and control. [Color figure can be viewed in the online issue, whichis available at wileyonlinelibrary.com.]


VEGF and MMPs.20–22 Thus, we asked whetherp28GANK regulates HIF-1a activity in HCC cells. Asshown in Fig. 4B, overexpression of p28GANK inMHCC-97L cells resulted in up-regulation of HIF-1a–responsive luciferase reporter which contains hy-poxia response element–binding sites (9-fold), whereasdown-regulation of p28GANK in HCC-LM3 cells ledto a decrease in the HIF-1a reporter level (79.5%).Consistently, HIF-1a protein level was higher in

MHCC-97L-LV-p28GANK cells than in LV-GFP cells,whereas silencing p28GANK suppressed HIF-1a expres-sion in HCC-LM3 cells (Fig. 4C). Increased level ofHIF-1a accompanied with VEGF and MMP2enhancement was also observed in p28GANK-overex-pressing tumors (Fig. 4D).HIF-1a small interfering RNA (siRNA) significantly

down-regulated HIF-1a protein levels but notp28GANK, whereas p28GANK miRNA inhibited both

Fig. 3. Overexpression of p28GANK enhances EMT in HCC cells. (A) Morphology of SMMC-7721 and MHCC-97L infected with either LV-GFP or LV-p28GANK is shown by phase contrast microscopy (magnification, �200). (B) qRT-PCR was performed to assess messenger RNA levels of epithelialmarkers E-cadherin, cytokeratin-8, cytokeratin-17, cytokeratin-18, claudin-1, and claudin-8 and mesenchymal markers N-cadherin, vimentin, HEY1,HEY2, Jagged1, Jagged2, Goosecoid, and an EMT regulator TWIST1 in LV-p28GANK–infected versus LV-GFP–infected MHCC-97L cells. Results, normal-ized against b-actin, are presented as mean 6 SEM from three independent experiments. (C) Expression of indicated molecules in MHCC-97L cellsinfected with LV-GFP or LV-p28GANK and in HCC-LM3 cells infected with LV-miNon or LV-mip28GANK were detected by western blot. (D) Tumors fromMHCC-97L mock, LV-GFP, or LV-p28GANK groups were immunostained for indicated molecules. The yellow arrows indicate membranous expression ofE-cadherin, N-cadherin, or b-catenin; blue arrows indicate the cytoplasmic translocation of b-catenin; and red arrows indicate the nuclear expressionof b-catenin. Pictures are representative of three independent experiments. (E) Immunoblots were performed to detect expression of indicated mole-cules in SMCC-7721 LV-p28GANK or MHCC-97L LV-p28GANK cells transfected with siRNA targeting TWIST1 (TWIST1 siRNA) or scrambled siRNA (ConsiRNA). *P < 0.05, **P < 0.001. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]


p28GANK and HIF-1a expression, indicating that HIF-1a is downstream of p28GANK (Fig. 4C,E). Moreover,p28GANK-mediated VEGF and MMP2 production wascounteracted by silencing HIF-1a in SMMC-7721 orMHCC-97L cells (Fig. 4E). In addition, HIF-1a sup-pression reduced p28GANK-induced TWIST1 butrestored p28GANK-reduced E-cadherin (Fig. 4E). Theseresults suggest that p28GANK may promote EMTresponse and tumor cell invasion through HIF-1a.

The PI3K/AKT Pathway Plays a Critical Role inMediating p28GANK Function. Signaling pathwaysactivated by p28GANK were analyzed by expression ofphosphorylated forms of AKT, p38 mitogen-activatedprotein kinase, extracellular signal-regulated kinase(ERK), and JNK by immunoblot. Only the p-AKTsignal was observed to be significantly higher inMHCC-97L-p28GANK cells, whereas there was adecrease in HCC-LM3-LV-mip28GANK cells (Fig. 5A;

Fig. 4. p28GANK activates VEGF and MMP2 correlated with HIF-1a. (A) (Upper) Protein contents of VEGF and p28GANK were examined by immuno-blot in LV-p28GANK–infected versus LV-GFP–infected MHCC-97L cells, and LV-mip28GANK–infected versus LV-miCon–infected HCC-LM3 cells. (Lower)The above-mentioned cells were serum-starved for 18 hours. The supernatants were immunoblotted by MMP2 or MMP9 antibodies, and MMP2 activitywas analyzed by zymography. Equal amounts of total proteins were loaded in each lane. (B) Luciferase activity was measured in the indicated cells intriplicate. Firefly luciferase activity was normalized by Renilla luciferase activity, and the value of the empty pGL3-basic vector was used as control. Val-ues are given as the mean6 SEM of three independent experiments. **P < 0.001. (C) The effect of p28GANK on HIF-1a expression was analyzed byimmunoblot in indicated cells. (D) Representative immunohistochemistry of VEGF, MMP2, and HIF-1a in MHCC-97L mock, MHCC-97L LV-GFP, orMHCC-97L LV-p28GANK induced tumors. The yellow arrows indicate mesenchymal expression of MMP2, blue arrows indicate the cytoplasmic transloca-tion of MMP2, and red arrows indicate nuclear expression of HIF-1a. Pictures are representative of three independent experiments. (E) Expression ofindicated molecules in SMCC-7721 LV-p28GANK or MHCC-97L LV-p28GANK cells transfected with HIF-1a siRNA or scrambled siRNA (Con siRNA). MMP2expression in cell supernatants was detected as above. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]


Supporting Information Fig. 4B). Silencing AKTexpression by siRNA suppressed both proliferation ofLV-GFP and LV-p28GANK cells. Interestingly, com-pared with LV-GFP, LV-p28GANK still significantlypromoted proliferation of MHCC-97L cells evenwhen AKT was down-regulated (Supporting Informa-tion Fig. 4C), suggesting that AKT does not play akey role in p28GANK-mediated cell proliferation. How-ever, suppression of AKT profoundly blockedp28GANK-induced matrigel invasion in MHCC-97Lcells, and ectopic expression of AKT restored the inva-siveness of LV-mip28GANK–treated HCC-LM3 cells(Fig. 5B). LV-p28GANK–enhanced adhesion to cell ma-trix proteins was reduced in the absence of AKT (Sup-porting Information Fig. 4D). Consistently, PI3K-AKT inhibitors (LY294002 and rapamycin), ratherthan Ras-ERK (PD98059) or p38–mitogen-activatedprotein kinase (SB203580), could markedly blockp28GANK-induced invasion (Supporting InformationFig. 4E). Taken together, these data show requirementfor the PI3K/AKT pathway in p28GANK-mediatedinvasion and adhesion, but not proliferation.Either knockdown or small AKT inhibitors (LY294002

and rapamycin), rather than ERK inhibitor (PD98059)

or p38 inhibitor (SB203580), blocked p28GANK-activatedHIF-1a activation, whereas mip28GANK-diminishedHIF-1a reporter was reversed by ectopic expression ofAKT, indicating PI3K-AKT involvement in induction ofHIF-1a by p28GANK (Supporting Information Fig.5A,B). Moreover, AKT knockdown repressed p-AKT sig-nal and HIF-1a expression in LV-p28GANK groups (Fig.5C). More importantly, inhibition of PI3K-AKT pathwayin vivo by rapamycin dramatically impeded the pulmo-nary metastasis of p28GANK-overexpressing cells (Fig.5D). Collectively, these findings support a critical role ofAKT during p28GANK-induced invasiveness and metasta-sis in cancer cells.We did not observe any change in expression of AKT

signaling regulators (PI3K, pyruvate dehydrogenase ki-nase 1 [PDK1], phosphatase and tensin homolog[PTEN], SH-2 containing inositd phosphatase-1(SHIP-1), and SH-2 containing inositd phosphatase-2(SHIP-2)), nor any change in PI3K, PDK1 kinase activ-ity by p28GANK. Moreover, no association of p28GANK

with those regulators was observed by immunoprecipita-tion in MHCC-97L cells (data not shown). Throughcoimmunoprecipitation, we detected a complex consist-ing of p28GANK, RhoGDIa (RhoGDIa (Rho GDP

Fig. 5. PI3K-AKT pathway is involved in mediating p28GANK action. (A) p-AKT was evaluated by immunoblot analysis with AKT protein and b-actinas controls. (B) Matrigel invasion assay was done for the indicated cells following treatment with AKT siRNA or ectopic AKT expression. (C) SMCC-7721LV-p28GANK or MHCC-97L LV-p28GANK cells were transfected with AKT siRNA or Con siRNA, and were subjected to immunoblot with the indicated anti-bodies. (D) The MHCC-97L vector control or p28GANK overexpressing cells were injected into right flanks of 10 nude mice. Fourteen days after injection,mice were randomly assigned into two groups and set for the rapamycin sensitivity assay. Pulmonary metastasis lesions were analyzed at 30 days later.(E) Immunoblot analysis of the interaction between p28GANK, RhoGDIa, and RhoA in lysates of MHCC-97L LV-GFP or MHCC-97L LV-p28GANK cells afterimmunoprecipitation with anti-p28GANK and whole-cell lysates (WCL). IgL, immunoglobin light chain.


dissociation inhibitor (GDI) a), and RhoA (ras homo-log gene family, member A) proteins (Fig. 5E), whichresults in repression of ROCK2 (Rho-associated, coiled-coil containing protein kinase 2) activity (SupportingInformation Fig. 6). This result suggests a mechanismin which p28GANK activates p-AKT signaling throughcontrol of the RhoGDIa/RhoA/ROCK2 pathway. Con-sistent with this, Man et al. reported recently thatp28GANK promoted RhoGDIa interaction with RhoA,leading to inhibition of RhoA/ROCK2 activity, andprolonged AKTactivation in NIH3T3 cells, human em-bryonic kidney 293 cells, and lung cancer cells.23

Combination of p28GANK and p-AKT Levels HasBetter Prognostic Value for HCC. We further ana-lyzed the expression levels of p28GANK, p-AKT, E-cad-herin, TWIST1, HIF-1a, RB, and p53 in clinicalHCC samples. Tissue microarray analysis of 130patient specimens revealed a strong correlation ofp28GANK expression with p-AKT levels (r ¼ 0.2505,P ¼ 0.0004) (Fig. 6A). Moreover, patients whosetumors expressed above-average levels of p28GANK orp-AKT exhibited significantly decreased trend in anyof the prognostic indicators, including time to DFSand OS due to HCC-related death (Supporting Infor-mation Fig. 7A,B). For patients whose tumors hadabove-average levels of both p28GANK and p-AKT,adverse outcomes were exacerbated (Fig. 6B). Usingthe combination of these two parameters increased theprognostic value, as compared to p28GANK or p-AKToverexpression alone (Fig. 6B; Supporting InformationFig. 7A,B). In conclusion, evaluation of both p28GANK

expression and p-AKT signal is a powerful predictor ofpoor prognosis, further supporting a model ofp28GANK activation of PI3K–AKT–HIF-1a signaling,resulting in EMT occurrence, VEGF and MMP2 pro-duction, and thus metastases of HCC cells (Fig. 6C).

Discussion

Efforts to elucidate the molecular mechanism under-lying HCC tumorigenicity, invasion, and metastasis ofHCC are warranted in order to identify biomarkersfor prediction and intervention. In this study, wedetermined the significance and underlying mechanismfor p28GANK overexpression in HCC progression andmetastasis. The p28GANK content was low in normalhepatocytes, increased in noninvasive and primaryHCC cells, and reached the highest level in invasiveHCC and PVTT cells. This progressively increasedexpression profile paralleled with deterioration of thedisease, suggesting a role of p28GANK in progression ofHCC.

Analyzing the association of p28GANK expressionwith pathological characteristics in 201 patients withHCC by tissue microarray revealed a significant corre-lation of p28GANK expression with tumor size, veininvasion, intrahepatic metastasis, and distant metasta-sis, which are all hallmarks for poor prognosis ofHCC.24,25 Indeed, the Kaplan-Meier analysis showsthat patients with HCC who had high p28GANK

expression in general had worse prognosis than thosewith low expression. We believe that p28GANK is anattractive candidate gene for risk prognostication andtherapy of HCC. However, our data is apparently atodds with a recent report suggesting that the cumula-tive survival rate of patients with gankyrin-positiveHCC was significantly higher than those patients withgankyrin-negative HCC.26 The discrepancy may bedue to different backgrounds of specimens used,including the proportion of hepatitis C virus and hep-atitis B virus infection, sex of patients, and the classifi-cation/criteria of tumor-node-metastasis (TNM) stag-ing. Recently, Ortiz and Tang reported that gankyrinmessenger RNA and protein increased in humanesophageal squamous cell carcinoma (ESCC) or colo-rectal cancer (CRC), and its overexpression is poorprognosis of ESCC or CRC due to its significant cor-relation with TNM stages and metastasis of thesetumors, respectively.27,28 Therefore, p28GANK overex-pression may be involved in development of humandigestive malignancies such as HCC, ESCC, andCRC.The effect of p28GANK on tumor invasion and me-

tastasis was directly demonstrated in our in vitro andin vivo studies. In both subcutaneous and orthotopicxenografts, overexpression of p28GANK generated largerprimary tumors and more lung metastasis foci, andhigher levels of vascularization and angiogenesis, indi-cating their more aggressive and metastatic properties.Moreover, down-regulation of p28GANK led to severesuppression of tumor growth and lung metastasis ofHCC in mice. To our knowledge, this is the firstreport that p28GANK expression is critical for HCCmetastasis, in addition to tumor proliferation andgrowth.In this study, we found that TWIST1 is indeed

involved in p28GANK-driven EMT. Moreover,p28GANK modulated HIF-1a hyperactivation andexpression correlated with TWIST up-regulation andE-cadherin down-regulation. Thus, our data suggest arequirement for HIF-1a in p28GANK-driven EMT. Wealso observed a role of HIF-1a in p28GANK-regulatedVEGF and MMP2 expression, consistent with previousreports that HIF-1a up-regulates VEGF, promoting


angiogenesis and invasion of HCC.29–31 Taken to-gether, this study clearly demonstrates a crucial role forp28GANK in induction of EMT and angiogenesisthrough regulation of HIF-1a, VEGF, and MMP2expression.An increase in AKT signal is a key tumor survival

mechanism, and promotes tumor metastatic processesincluding EMT, resistance to apoptosis, and angiogen-esis.32–34 Previous studies have demonstrated that acti-vated AKT plays a critical role in hematogenous intra-hepatic metastasis in an orthotopic implantationmodel of HCC.35 Our group previously showed a pro-tective role of p28GANK in HCC cells against endo-plasmic reticulum stress-induced apoptosis, partially

through enhancing AKT phosphorylation.14 Our cur-rent in vitro and in vivo studies suggest that p-AKT isresponsible for p28GANK-mediated invasion/metastasis.Furthermore, we found that p28GANK interacted withRhoGDIa and RhoA, resulting in inhibition ofROCK2 activity in HCC cells, an observation sup-ported by a recent report that p28GANK negatively regu-lates the RhoA/ROCK2/PTEN pathway for activationof AKT.23 Given complex p-AKT pathways, whetherother upstream regulators are involved in p28GANK-pro-moting p-AKT signal remains to be further determined.Remarkably, the predictive range of p28GANK expressionlevels combined with p-AKT signal was more sensitivethan that of p28GANK alone for OS and cumulative

Fig. 6. Combination of elevated p28GANK expression and p-AKT signal is a powerful predictor of poor clinical outcome in HCCs. (A) Correlationbetween p28GANK expression and p-AKT level was examined in tumor tissues derived from 130 patients, r ¼ 0.2505, P ¼ 0.0004. Representa-tive immunostaining for p28GANK and p-AKT is shown for three patient samples. (B) Combination of p28GANK and p-AKT enhanced correlation toclinical parameters and the significance for poor prognosis. The median value for p28GANK and p-AKT expression in HCCs was used to divide thepatients into high (above median) and low (below median) p28GANK and p-AKT groups. (C) A model for p28GANK action in HCC progression andmetastasis is shown, with p28GANK acting as a critical driving factor in metastatic carcinoma cells counteracting with various barriers. [Color figurecan be viewed in the online issue, which is available at wileyonlinelibrary.com.]


recurrence, strongly suggesting that the concerted activ-ities of p28GANK and p-AKT detected in our experi-ments are recapitulated in clinical patients with HCC.Identification of tumor p28GANK alone or combinedevaluation of p28GANK/p-AKT levels as a new prognos-tic marker in patients with HCC is important becausethey provide not only a new criterion for prognosis, butalso a potential therapeutic target.The most interesting part of the results shown here is

the remarkable function of p28GANK in transformingnoninvasive HCC cells into highly aggressive cells thatgenerate tumors similar to those in patient-derived sam-ples (Fig. 6C). p28GANK is a cytoplasmic protein thatcontains seven ankyrin repeats to mediate protein-pro-tein interactions, and acts as a chaperone for the assem-bly of the 19S structure of the 26S proteasome.36 How-ever, neither 26S proteasome activity nor the overalllevels of polyubiquitinated proteins were changed inp28GANK-overexpressed or knockdown cells (Support-ing Information Fig. 8A,B), indicating that the protea-some system is not involved in p28GANK-mediated inva-sion/metastasis. Previous studies showed that p28GANK

plays its oncogenic role by controlling the activities ofpRb and p53.8,12 Intriguingly, even in both Rb andp53-deficient Hep3B cells, p28GANK overexpression stillpromoted their invasion (Supporting Information Fig.8C). Combined with no evident correlation of pRb orp53 with p28GANK in clinical HCC samples (data notshown), it is likely that p28GANK-induced invasion/me-tastasis is independent of Rb and/or p53 status.In conclusion, we have identified p28GANK as a key

regulator that controls multiple facets essential for HCCdevelopment and metastasis. In particular, the data hasled us to propose that p28GANK or combination ofp28GANK with p-AKT is a novel marker in the progno-sis of HCC and a potential therapeutic target. Becausep28GANK is also overexpressed in other types of cancers,including lung,23 esophageal,27 colon,28 gastric carci-noma, rectal, bladder, breast, ovary, and uterus endome-trium cancers (Fu and Chen, unpublished observations),we believe that this oncoprotein may be widely involvedin tumorigenesis in human cancers.Acknowledgment: We thank Dangsheng Li (Shanghai

Institutes for Biological Sciences, China) for critical reviewof the manuscript and helpful suggestions. We also thankGuoqiang Chen (Shanghai Jiao Tong University, Shang-hai, China) for the gift of pGL3–HIF-1a plasmid.

References

1. Parkin DM, Bray F, Ferlay J, Pisani P. Global cancer statistics, 2002.CA Cancer J Clin 2005;55:74-108.

2. He J, Gu D, Wu X, Reynolds K, Duan X, Yao C, et al. Major causesof death among men and women in China. N Engl J Med 2005;353:1124-1134.

3. Yeo W, Mok TS, Zee B, Leung TW, Lai PB, Lau WY, et al. Arandomized phase III study of doxorubicin versus cisplatin/interferonalpha-2b/doxorubicin/fluorouracil (PIAF) combination chemotherapyfor unresectable hepatocellular carcinoma. J Natl Cancer Inst 2005;97:1532-1538.

4. Pang RW, Joh JW, Johnson PJ, Monden M, Pawlik TM, Poon RT. Biol-ogy of hepatocellular carcinoma. Ann Surg Oncol 2008;15:962-971.

5. Lopez PM, Villanueva A, Llovet JM. Systematic review: evidence-basedmanagement of hepatocellular carcinoma--an updated analysis ofrandomized controlled trials. Aliment Pharmacol Ther 2006;23:1535-1547.

6. Portolani N, Coniglio A, Ghidoni S, Giovanelli M, Benetti A, TiberioGA, et al. Early and late recurrence after liver resection for hepatocellu-lar carcinoma: prognostic and therapeutic implications. Ann Surg 2006;243:229-235.

7. Aravalli RN, Steer CJ, Cressman EN. Molecular mechanisms of hepa-tocellular carcinoma. HEPATOLOGY 2008;48:2047-2063.

8. Higashitsuji H, Itoh K, Nagao T, Dawson S, Nonoguchi K, Kido T,et al. Reduced stability of retinoblastoma protein by gankyrin, an onco-genic ankyrin-repeat protein overexpressed in hepatomas. Nat Med2000;6:96-99.

9. Fu XY, Wang HY, Tan L, Liu SQ, Cao HF, Wu MC. Overexpressionof p28/gankyrin in human hepatocellular carcinoma and its clinical sig-nificance. World J Gastroenterol 2002;8:638-643.

10. Iwai A, Marusawa H, Kiuchi T, Higashitsuji H, Tanaka K, Fujita J,et al. Role of a novel oncogenic protein, gankyrin, in hepatocyte prolif-eration. J Gastroenterol 2003;38:751-758.

11. Li H, Fu X, Chen Y, Hong Y, Tan Y, Cao H, et al. Use of adenovirus-delivered siRNA to target oncoprotein p28GANK in hepatocellular carci-noma. Gastroenterology 2005;128:2029-2041.

12. Higashitsuji H, Higashitsuji H, Itoh K, Sakurai T, Nagao T, SumitomoY, et al. The oncoprotein gankyrin binds to MDM2/HDM2, enhanc-ing ubiquitylation and degradation of p53. Cancer Cell 2005;8:75-87.

13. Wang X, Li H, Chen Y, Fu J, Ren Y, Dong L, et al. p28GANK knock-down-derived reactive oxygen species induces apoptosis through mito-chondrial dysfunction mediated by p38 in HepG2 cells. Int J Oncol2008;33:743-750.

14. Dai RY, Chen Y, Fu J, Dong LW, Ren YB, Yang GZ, et al. p28GANK

inhibits endoplasmic reticulum stress-induced cell death via enhance-ment of the endoplasmic reticulum adaptive capacity. Cell Res 2009;19:1243-1257.

15. Higashitsuji H, Higashitsuji H, Liu Y, Masuda T, Fujita T, Abdel-AzizHI, et al. The oncoprotein gankyrin interacts with RelA and suppressesNF-kappaB activity. Biochem Biophys Res Commun 2007;363:879-884.

16. Chen Y, Li HH, Fu J, Wang XF, Ren YB, Dong LW, et al. Oncopro-tein p28 GANK binds to RelA and retains NF-kappaB in the cyto-plasm through nuclear export. Cell Res 2007;17:1020-1029.

17. Tang ZY, Ye SL, Liu YK, Qin LX, Sun HC, Ye QH, et al. A decade’sstudies on metastasis of hepatocellular carcinoma. J Cancer Res ClinOncol 2004;130:187-196.

18. Li Y, Tian B, Yang J, Zhao L, Wu X, Ye SL, et al. Stepwise metastatichuman hepatocellular carcinoma cell model system with multiple meta-static potentials established through consecutive in vivo selection andstudies on metastatic characteristics. J Cancer Res Clin Oncol 2004;130:460-468.

19. Giannelli G, Pierri F, Trerotoli P, Marinosci F, Serio G, Schiraldi O,et al. Occurrence of portal vein tumor thrombus in hepatocellular car-cinoma affects prognosis and survival. A retrospective clinical study of150 cases. Hepatol Res 2002;24:50.

20. Semenza GL. HIF-1 and tumor progression: pathophysiology and ther-apeutics. Trends Mol Med 2002;8:S62-S67.

21. Ema M, Taya S, Yokotani N, Sogawa K, Matsuda Y, Fujii-Kuriyama Y.A novel bHLH-PAS factor with close sequence similarity to hypoxia-


inducible factor 1alpha regulates the VEGF expression and is poten-tially involved in lung and vascular development. Proc Natl Acad Sci US A 1997;94:4273-4278.

22. Gupta GP, Massague J. Cancer metastasis: building a framework. Cell2006;127:679-695.

23. Man JH, Liang B, Gu YX, Zhou T, Li AL, Li T, et al. Gankyrin playsan essential role in Ras-induced tumorigenesis through regulation ofthe RhoA/ROCK pathway in mammalian cells. J Clin Invest 2010;120:2829-2841.

24. Poon RT, Fan ST, Lo CM, Liu CL, Wong J. Difference in tumor inva-siveness in cirrhotic patients with hepatocellular carcinoma fulfilling theMilan criteria treated by resection and transplantation: impact on long-term survival. Ann Surg 2007;245:51-58.

25. Velazquez RF, Rodriguez M, Navascues CA, Linares A, Perez R, Sotor-rios NG, et al. Prospective analysis of risk factors for hepatocellular car-cinoma in patients with liver cirrhosis. HEPATOLOGY 2003;37:520-527.

26. Umemura A, Itoh Y, Itoh K, Yamaguchi K, Nakajima T, HigashitsujiH, et al. Association of gankyrin protein expression with early clinicalstages and insulin-like growth factor-binding protein 5 expression inhuman hepatocellular carcinoma. HEPATOLOGY 2008;47:493-502.

27. Ortiz CM, Ito T, Tanaka E, Tsunoda S, Nagayama S, Sakai Y, et al.Gankyrin oncoprotein overexpression as a critical factor for tumorgrowth in human esophageal squamous cell carcinoma and its clinicalsignificance. Int J Cancer 2008;122:325-332.

28. Tang S, Yang G, Meng Y, Du R, Li X, Fan R, et al. Overexpression ofa novel gene gankyrin correlates with the malignant phenotype of colo-rectal cancer. Cancer Biol Ther 2010;9:88-95.

29. Huang GW, Yang LY, Lu WQ. Expression of hypoxia-inducible factor1alpha and vascular endothelial growth factor in hepatocellular carci-noma: Impact on neovascularization and survival. World J Gastroen-terol 2005;11:1705-1708.

30. Maxwell PH, Dachs GU, Gleadle JM, Nicholls LG, Harris AL, Strat-ford IJ, et al. Hypoxia-inducible factor-1 modulates gene expression insolid tumors and influences both angiogenesis and tumor growth. ProcNatl Acad Sci U S A 1997;94:8104-8109.

31. Carmeliet P, Dor Y, Herbert JM, Fukumura D, Brusselmans K, Dewer-chin M, et al. Role of HIF-1alpha in hypoxia-mediated apoptosis, cellproliferation and tumour angiogenesis. Nature 1998;394:485-490.

32. Larue L,Bellacosa A. Epithelial-mesenchymal transition in developmentand cancer: role of phosphatidylinositol 30 kinase/AKT pathways.Oncogene 2005;24:7443-7454.

33. Sheng S, Qiao M, Pardee AB. Metastasis and AKT activation. J CellPhysiol 2009;218:451-454.

34. Franke TF, Yang SI, Chan TO, Datta K, Kazlauskas A, Morrison DK,et al. The protein kinase encoded by the Akt proto-oncogene is a targetof the PDGF-activated phosphatidylinositol 3-kinase. Cell 1995;81:727-736.

35. Nakanishi K, Sakamoto M, Yasuda J, Takamura M, Fujita N, TsuruoT, et al. Critical involvement of the phosphatidylinositol 3-kinase/Aktpathway in anchorage-independent growth and hematogeneous intrahe-patic metastasis of liver cancer. Cancer Res 2002;62:2971-2975.

36. Saeki Y, Toh EA, Kudo T, Kawamura H, Tanaka K. Multiple protea-some-interacting proteins assist the assembly of the yeast 19S regulatoryparticle. Cell 2009;137:900-913.


p28gank overexpression accelerates hepatocellular carcinoma invasiveness and metastasis via...

Documents