Upregulation of Rac Gtpase-Activating Protein 1 Is Significantly Associated with the Early Recurrence of Human Hepatocellular Carcinoma

Published OnlineFirst August 8, 2011; DOI: 10.1158/1078-0432.CCR-11-0557

Clinical Cancer Imaging, Diagnosis, Prognosis Research

Upregulation of Rac GTPase-Activating Protein 1 Is Significantly Associated with the Early Recurrence of Human Hepatocellular Carcinoma

Suk Mei Wang1, London Lucien P.J. Ooi2, and Kam M. Hui1,3

Abstract Purpose: To assess the significance of Rac GTPase-activating protein 1 (RACGAP1) expression in identifying HBV-positive human hepatocellular carcinoma (HCC) patients who are at high risk for recurrent disease. Experimental Design: The prognostic significance of RACGAP1 was compared with clinicopathologic parameters available at diagnosis using multivariate and log-rank test. RACGAP1 expression and outcome in recurrence was compared between 35 patients with recurrence and 41 patients without recurrence using Kaplan–Meier analysis. RACGAP1-targeted molecules and pathways were identified and characterized by inhibition with siRNA duplexes. Results: Kaplan–Meier analysis showed that the level of RACGAP1 expression is sufficient to predict the early recurrence of HCC: high RACGAP1 expression correlates with high risk of postresection recurrent HCC (P < 0.0005). Silencing of RACGAP1 in Hep3B and MHCC97-H HCC cells with high endogenous RACGAP1 expression inhibited cell migration and invasion. Using Ingenuity Pathway Analysis, the target molecules silenced in the RACGAP1 interactome were mostly genes related to the mitotic roles of the polo- like kinases. These included PRC1, AURKB, CDC2, ECT2, KIF23, PAK1, and PPP2R5E. In providing clinical corroboration of these results, when expression of these transcripts was analyzed in an expression database that we have established previously for HBV-positive HCC patients, these genes was mostly upregulated in patients who exhibited early recurrent disease and hence provided important corroboration of these results. Conclusions: siRNA-silencing RACGAP1 mainly targeted genes in an interactome clinically relevant to early HCC recurrence. Besides being an independent informative prognostic biomarker, RACGAP1 could also be a potential molecular target for designing therapeutic strategies for HCC. Clin Cancer Res; 17(18); 6040–51. 2011 AACR.

Introduction of liver cirrhosis, including nonalcoholic steatohepatitis. The prevalence of HCC in Europe and the United States is Hepatocellular carcinoma (HCC) is the commonest pri- increasing and is currently the leading cause of death in mary cancer of the liver and is the third most frequent cause patients with cirrhosis, possibly resulting from the trans- of cancer-related deaths in the world, with more than mission of HCV by intravenous drug abuse and a rising 660,000 deaths per annum (1–5). The major etiologic prevalence of obesity and diabetes (6, 7). Surgery currently factors of HCC are hepatitis B virus (HBV) and hepatitis offers the only possibility of prolonged survival for HCC C virus infection (HCV), and various other nonviral-related patients. Unfortunately, recurrence occurs in more than causes such as aflatoxins, alcohol intake, and other causes two-thirds of these patients despite initial curative intent and converts the situation to a dismal prognosis (8, 9). It is presently a challenge to identify patients who are at Authors' Affiliations: 1Bek Chai Heah Laboratory of Cancer Genomics, high risk for early recurrence after undergoing potentially Division of Cellular and Molecular Research, Humphrey Oei Institute of Cancer Research, 2Department of Surgical Oncology, National Cancer curative treatment for HCC and various surrogate clinico- Centre; and 3Cancer and Stem Cell Biology Program, Duke-NUS Graduate pathologic features such as lymphovascular invasion, Medical School, Singapore capsular invasion, satellite lesions, and tumour numbers Corresponding Author: Kam M. Hui, Bek Chai Heah Laboratory of Cancer are often used with varying reliability. Such high-risk Genomics, Division of Cellular and Molecular Research, Humphrey Oei Institute of Cancer Research, National Cancer Centre, 11 Hospital Drive, patients could potentially benefit from closer surveillance Singapore 169610. Phone: 65-6436-8337; Fax: 65-6226-3843; E-mail: or receive adjuvant novel interventional measures, if they [email protected] could be accurately identified. DNA microarrays have been doi: 10.1158/1078-0432.CCR-11-0557 widely applied to the study of human cancer and compre- 2011 American Association for Cancer Research. hensive and systematic functional analyses of large number

6040 Clin Cancer Res; 17(18) September 15, 2011

RACGAP1 and HCC Recurrence

Translational Relevance Cancer Centre Singapore and conducted in accordance with the policies of its Ethics Committee. Informed consent Hepatocellular carcinoma (HCC) is the third com- was obtained from all participating patients, and all clinical monest cause of cancer-related deaths in the world. and histopathologic data provided to the researchers were Surgery currently offers the only possibility of long-term rendered anonymous. Cancerous and some of the corre- survival for these patients. Unfortunately, recurrences sponding distant noncancerous liver tissues were obtained occur in more than two-thirds of these patients and from patients who underwent surgical resection as curative confer a dismal prognosis. In this study, we have sys- treatment for HCC. All tumor tissues were divided into tematically presented molecular evidence and provided 2 portions and immediately snap frozen in liquid nitrogen. clinical corroboration of these data to show that, inde- Half of the sample was stored in liquid nitrogen until use, pendently from clinical risk factors, aggressive early whereas the other portion was employed for hematoxylin recurrent HCC tumors have their Rac GTPase-activating and eosin staining and evaluated by an independent protein 1 (RACGAP1) expression significantly upregu- pathologist. All the cancerous tissues studied were made lated. For the first time, our data provide clinical support up of at least 70% of cancer cells. for possible drug developments targeting the various An early recurrence was defined as a recurrence within important oncogenic signaling molecules in an inter- 2 years after a curative resection. To assess recurrence, all actome clinically relevant to early HCC recurrence. Our treated HCC patients were monitored by routine clinical results also suggest the importance of RACGAP1 as a follow-up once every 3 months. The level of serum alpha- stratification factor for the design of future comparative fetoprotein (AFP) and liver function tests were determined therapeutic trials, which is especially important for early every 3 months and ultrasound scans of the liver were done recurrent HCC tumors. every 6 months. Computerized tomography (CT) or MRI scans of the liver were done when the serum levels of AFP showed a rising trend or when the ultrasound results indicated the presence of possible recurrent disease. A total of genetic and epigenetic alterations provide unbiased of 76 HCC liver biopsies with 24 histologically normal analytical approaches to decipher the molecular heteroge- tissues were collected and studied: Thirty-five samples were neity of cancer (10, 11). Through these strategies, distinct from patients who had early recurrent disease over the subclasses of HCC patients based on their differing gene 24-month observation period, whereas 41 did not have expression patterns, which were also associated with early recurrent disease. In addition, samples of histologi- patient survival, were identified, indicating the presence cally normal liver tissues of 10 colorectal cancer patients of distinct molecular subtypes of HCC (12–14). However, who had liver metastases resected were used as reference the identification of the early recurrence of HCC remains a normal liver tissues. major challenge and the development of new prognostic markers are urgently needed to identify HCC patients who Oligonucleotide gene chips microarray analysis are at higher risk of having recurrence. Global gene profiling experiments of the clinical Previous studies by our group and others have success- samplesweredoneusingtheHumanGenomeU133 fully shown that specific gene expression signatures can Set (HG-U133A and HG-U133B) from Affymetrix be established from frozen and formalin-fixed cancerous (Affymetrix Inc.) as previously described (18). Gene pro- tissues to accurately predict early recurrent disease follow- filing analyses following the inhibition of RACGAP1 ing curative hepatic surgery (15–19). In this context, we expression with siRNA duplexes in the MHCC97-H hu- have reported a 57-member gene signature earlier for man HCC cell line were done with Affymetrix Human selecting HCC patients who are at higher risk of having Genome U133 Plus 2.0 Arrays. The microarray data have recurrent disease (18). Although a number of options for been deposited in the European Bioinformatics Institutes gene set analysis exist, we have chosen to investigate the of the European Molecular Biology Laboratory database prognostic significance of individual members of the gene (http://www.ebi.ac.uk/arrayexpress/) and are accessible set using multivariate and log-rank test and observed that through ArrayExpress public database with accession Rac GTPase-activating protein 1 (RACGAP1), a member of numbers E-MEXP-84 and E-TABM-292. Signal intensities this 57-member gene signature, gave the best ability to were transformed to log2 baseandimportedtoPartek predict early recurrent disease and survival outcome. In this Genomics Suite software (Partek Inc.) to conduct statis- study, we report the identification and molecular charac- tical analyses. The Affymetrix probeset Id for RACGAP1 is terization of RACGAP1 as a clinically relevant prognostic 222077_s_at. predictor for recurrent HCC disease. Immunohistochemistry Materials and Methods Sections of 6 mm were mounted onto Superfrost Plus microscope glass slides (Thermo Fisher scientific) and Patient samples stored at 80C until use. All frozen sections were fixed All tissue samples employed in this study were approved with chilled 100% acetone at 20C for 10 minutes prior and provided by the Tissue Repository of the National to incubation with the mouse RACGAP1 MoAb (M01)

www.aacrjournals.org Clin Cancer Res; 17(18) September 15, 2011 6041

Wang et al.

clone 1G6 (Abnova) followed by adding dextran carrying Culture Collection. Cells were maintained in Dulbecco’s anti-mouse IgG conjugated to horseradish peroxidase modified Eagle’s medium (DMEM) supplemented with (Chemicon) and positive staining was developed using 10% heat-inactivated FBS (Hyclone). the Dako REAL EnVision detection system. Isotype- matched mouse IgG2b (Dako) was used as a negative siRNA treatment control. Images of stained sections were imported into Two pooled RACGAP1 siRNA sequences, R1 and R2, Image-Pro Plus, Version 7.0 (Media Cybernetics) for quan- were employed in this study. The sequences for RACGAP1 tifying RACGAP1-stained cells. siRNA duplexes R1 consisted of a mix of 3 different sequences: duplex 1 minus-CAC ACU GUC UGU CUC Cell cultures AGU UCU UGG C, plus-GCC AAG AAC UGA GAC AGA The human HCC cell line MHCC97-H was a generous CAG UGU G; duplex 2 minus-UUU ACU GUG CGG UCA gift from Prof. Tang Zhao-You and Dr. Liu Bin Bin (Liver CAG CCA GAG A, plus-UCU CUG GCU GUG ACC GCA Cancer Institute and Zhongshan Hospital, Fudan Univer- CAG UAA A, and duplex 3 minus-UUG CCU UGU CGU sity, Shanghai, PR China; ref. 20). Hep3B, HepGG2, and CCU AGG UUA GUG G, plus-CCA CUA ACC UAG GAC PLC/PRF5 cells were obtained from the American Type GAC AAG GCA. R2 also consisted of a mix of 3 sequences:

P A BCP < 0.0001 = 0.02 FC = 2.31 30 FC = 2.03 < 0.0001 8 P 25 P 8 < 0.0001 16 = 0.025 FC = 2.50 FC = 2.12 14 6 6 12 P = 0.2752 0.92 FC = 0.86 10 4 4 8 6 P = 0.052 2 2 4 FC = 0.93 2 Relative expression signal expression Relative 0 0 0 GC-RMA signal (Log base 2) NN ST HCC GC-RMA signal (Log base 2) NN ST NR R NN ST NR R (N = 10) (N = 24) (N = 76) (N = 10) (N = 24) (N = 41) (N = 35) (N = 7) (N = 7) (N = 18) (N = 18)

D Surrounding tumor Tumor (L104) Recurrent HCC (L98) Nonrecurrent HCC

Figure 1. Overexpression of RACGAP1 in primary HCC tumors. A, expression of RACGAP1 in primary HCC tumors compared with NN (histologically normal liver tissues from patients with colorectal metastases) and ST (histologically normal liver tissues of HCC patients) using Affymetrix gene chips. B, expression of RACGAP1 in tumor tissues of HCC with recurrent (R) and nonrecurrent (NR) disease within 2 years. C, expression of RACGAP1 as detected by quantitative real-time PCR analysis. P < 0.05 is statistically significant; FC, fold change. D, representative images of RACGAP1 expression in tumor and nontumorous tissues of NR and R HCC patients following IHC analysis. Using IHC, cells that stained brown were scored as positive and RACGAP1 staining was mainly localized in the nuclei. Insert in lower left corner shows image obtained under high magnification.

6042 Clin Cancer Res; 17(18) September 15, 2011 Clinical Cancer Research

RACGAP1 and HCC Recurrence

duplex 4 minus-UAU ACA GGC CUG UCU CAG UCA GAC (Clontech), which served as a reconstituted basement C, plus-GGU CUG ACU GAG ACA GGC CUG UAU A; membrane in vitro. Hep3B and MHCC97-H cells were duplex 5 minus-UUC UGC UGC UUC CAU AAA GGC transfected with RACGAP1 R1 and R2 siRNA sequences UCU G, plus-CAG AGC CUU UAU GGA AGC AGC AGA A, for 2, 4, or 6 days. A total of 1.5 105 transfected cells and duplex 6 minus-UUG AGA AGC UGA UGU UCA GGA were seeded and plated in 500 mL of 0.1% bovine serum GUG G, plus-CCA CUC CUG AAC AUC AGC UUC UCA A. albumin–DMEM to the upper chamber. The lower The MHCC97-H cells were transfected separately with chamber was filled with 750 mL of 10% FBS–DMEM as 100 nmol/L each of the 2 pooled RACGAP1 R1 and R2 the chemoattractant. After being cultured for 48 hours, siRNA sequences or with the stealth RNAi negative control noninvaded cells in the inserts were removed by using medium GC duplex (Invitrogen) using Lipofectamine 2000 cotton-tipped swabs. The cells that had invaded to the (Invitrogen). Using Western blot and cell lysate of HCC membrane undersurface were enumerated by microscopy cells, the RACGAP1 MoAb (M01) clone 1G6 gave an following fixation by 10% formaldehyde for 10 minutes, expected protein band of 72 kDa. permeabilized with 0.2% Triton X-100 and mounted in Vectashield Mounting Media with 40,6-diamidino-2-pheny- Migration and invasion assay lindole (DAPI; Vector Laboratories). At least 3 random Migration assays were carried out in 24-well plates fields per insert were counted and a representative field Boyden chambers with an 8-mm pore size PET membrane of each experiment was photographed. Results are expressed (Falcon). Invasion assays were done in a similar way except as means SE of 3 independent experiments. Statistical that each membrane had a thin layer of GFR Matrigel analysis was done using t test.

Table 1. Univariate and multivariate analyses showing that RACGAP1 expression could serve as an independent prognostic factor for recurrent HCC

Univariate analysis Multivariate analysis

Variable RR (95% CI) P RR (95% CI) P

Gender 1.56 (0.61–4.02) 0.358 n.s. Male (n ¼ 61) vs. female (n ¼ 15) Age 1.07 (0.55–2.08) 0.839 n.s. >60 y (n ¼ 41) vs. 60 y (n ¼ 35) Hepatitis 0.585 n.s. HBV (n ¼ 59) vs. non-B/C (n ¼ 14) 0.76 (0.33–1.76) 0.525 HCV (n ¼ 3) vs. non-B/C (n ¼ 14) 1.47 (0.31–7.09) 0.632 Encapsulation 0.590 n.s. Partial (n ¼ 19) vs. no (n ¼ 40) 0.81 (0.36–1.82) 0.604 Complete (n ¼ 17) vs. no (n ¼ 40) 0.63 (0.26–1.57) 0.323 Tumor size 1.62 (0.84–3.16) 0.153 n.s. >5cm(n ¼ 35) vs. 5cm(n ¼ 41) AFP 0.328 n.s. 10–300 ng/mL (n ¼ 27) vs. 10 ng/mL (n ¼ 31) 1.66 (0.72–3.78) 0.232 >300 ng/mL (n ¼ 21) vs. <10 ng/mL (n ¼ 31) 1.85 (0.78–4.35) 0.161 Lesion 1.18 (0.46–3.04) 0.732 n.s. Multiple (n ¼ 10) vs. single (n=66) Differentiation 0.626 n.s.

G2 (n ¼ 42) vs. G1 (n ¼ 10) 1.58 (0.47–5.38) 0.460

G3 (n ¼ 18) vs. G1 (n ¼ 10) 2.23 (0.62–8.01) 0.218

G4 (n ¼ 5) vs. G1 (n=10) 1.61 (0.27–9.65) 0.601 Cirrhosis 1.91 (0.95–3.85) 0.048* 2.55 (1.20–5.40) 0.0148* Yes (n ¼ 41) vs. no (n ¼ 35) Vascular invasion 3.17 (1.62–6.21) 0.011** 4.06 (1.93–8.53) 0.0002*** Yes (n ¼ 29) vs. no (n ¼ 47) RACGAP1 3.42 (1.64–7.16) 0.001** 2.71 (1.27–5.74) 0.0096* High (n ¼ 39) vs. low (n ¼ 37)

NOTE: Recurrence is defined as recurrent disease occurred within a 2-year time point. *, P < 0.05; **, P < 0.005; ***, P < 0.001; n.s. ¼ not significant.

www.aacrjournals.org Clin Cancer Res; 17(18) September 15, 2011 6043

Wang et al.

Table 2. Correlation between RACGAP1 ex- domain (PBD) of human Pak1 (GST-human Pak1-PBD) pression and clinicopathologic features of the that specifically binds active (GTP bound) Cdc42 and Rac1. For Rho pull-down assay, the same amount of protein 76 HCC patients in this study using Fisher exact was incubated with GST-fusion protein containing Rho- test binding domain (GST-Rhotekin-RBD). Inactive (GDP bound) Cdc42, Rac1, and Rho were washed 3 times in RACGAP RACGAP P m high low lysis buffer and bound proteins eluted in 30- L SDS-PAGE (n ¼ 39) (n ¼ 37) sample buffer. Normalized amounts of lysates were loaded on the gel and assessed for the presence of active small Gender 0.78 GTPases by Western blot using Cdc42, Rac1, and Rho Ab. Female 7 8 Pull-down assay was quantified by calculating the fold ratio Male 32 29 of the pull-down active GTPase (GTP-Cdc42, GTP-Rac1, Age, y 1 and GTP-Rho A) after normalization to the corresponding 60 18 17 total protein. The fold ratio obtained was then compared >60 21 20 Hepatitis (HBV or HCV) 0.24 Positive 34 28 Negative 5 9 A Tumor encapsulationa 0.06 1.0 Partial or complete 14 22 RACGAP1 low None 23 14 0.8 (R = 4, NR = 34) Tumor size (cm) 0.5 51921 >52016 0.6 AFP (ng/mL)b 0.0002 20 9 25 0.4 >20 29 12 RACGAP1 high (R = 17, NR = 21)

Lesions 0.19 Recurrence-free rate Multiple 3 7 0.2 Single 36 30 b Differentiation (staging) 0.62 P = 0.00005 0.0 G1–G2 25 27

G3–G4 13 10 1 23 45 6 Cirrhosis 0.25 Months Yes 24 17 B No 15 20 1.0 Vascular invasion 0.35 RACGAP1 low Yes 17 12 (R = 8, NR = 30) No 22 25 0.8 Recurrence 0.00004 Yes 28 7 0.6 No 11 30

a 3 patients not reported. 0.4 b1 patient not reported. Recurrence-free rate RACGAP1 high 0.2 (R = 27, NR = 11) Pull-down assays for Cdc42, Rac1, and Rho P = 0.000017 The activation of Rho family small GTPases was detected 0.0 using EZ-Detect Cdc42 Activation Kit, Rac1 Activation Kit 02 4681012141618202224 (both from Pierce Biotechnology) and Rho Activation Kit Months (Upstate/Millipore). Cell lysate was prepared with ice-cold lysis buffer provided in the kit supplemented with protease inhibitors (Roche Diagnostics). Lysates were clarified by Figure 2. RACGAP1 overexpression could serve as an independent centrifugation at 13,000 g for 10 minutes and aliquots prognostic factor for recurrent HCC. A, recurrent HCC disease 6 months. B, recurrent HCC disease 24 months. Kaplan–Meier plots stored frozen (80C) until use. For Cdc42 and Rac1 pull- showed that high RACGAP1 expression was significantly associated down assay, 1 mg of total protein was incubated overnight with early HCC recurrence. The P value was generated using log-rank at 4 C with GST-fusion protein containing p21-binding test between R and NR.

6044 Clin Cancer Res; 17(18) September 15, 2011 Clinical Cancer Research

RACGAP1 and HCC Recurrence

with the fold ratio of the corresponding untreated null importantly, RACGAP1 expression was significantly sample which was normalized to 1. upregulated in primary HCC biopsies from patients who had recurrent disease within 2 years (N ¼ 35) DNA fragmentation (TUNEL) assay compared with patients who did not have recurrence The ApoAlert DNA Fragmentation Assay Kit (Clontech) (NR, N ¼ 41; fold change ¼ 2.31, P < 0.0001, false was employed. Triplicate cultures of Hep3B and MHCC97- discovery rate (FDR) ¼ 0.043%), Figure 1B. Consistent H cells were transfected with the RACGAP1 R1 and R2 with results obtained with the microarray analysis, quan- siRNA sequences and harvested at day 2, 4, and 6 following titative real-time PCR studies showed that samples from transfection. Harvested cells were fixed with 1% formal- patients with high risk of recurrent disease had RACGAP1 dehyde for 20 minutes at 4C. The fixed cells were then expressions significantly upregulated compared with incubated with a mixture of terminal deoxynucleotidyl samples from patients with low risk of recurrent disease transferase and fluorescein-labeled nucleotide mix for (fold change ¼ 2.03, P < 0.02, Fig. 1C). Immunostaining 1 hour at 37C and stained with propidium iodide in experiments using paired R and NR samples further the presence of 0.5 mg/mL DNase-free RNase. validated the observation that RACGAP1 was upregulated in HCC biopsies from patients having early re- Statistical analysis currence. The frequency of RACGAP1 positively stained Correlation between clinicopathologic features and re- cells was significantly higher in samples from patients currence was done using the statistical package for the social having early recurrent disease than in NR samples sciences (SPSS) for Windows (version 15.0). A P value of [median positive signal for recurrent HCC samples ¼ less than 0.05 was taken as statistically significant. 7% (N ¼ 9); median positive signal for NR biopsies ¼ Tukey Biweight is an M-estimator that has the ability to 0.6% (N ¼ 9); P ¼ 0.014, Fig. 1D]. RACGAP1 staining down-weight data points that are far from the data center in was mainly localized in the nuclei. In addition, signi- the calculation of the mean. The Tukey Biweight is derived ficant difference in RACGAP1 expression could be using the following calculations: Assuming ‘x’ is the num- detected between R and paired surrounding nontumor- ber of gene expression values to be studied, m ¼ median ous tissues (P ¼ 0.011), whereas no significant difference of x; s ¼ median absolute deviation ¼ median (absolute (x) in RACGAP1 expression was observed between NR and m); c ¼ cut-off value determined by software based on paired surrounding nontumorous tissues (P ¼ 0.149). s; epsilon ¼ internal constant ¼ 0.0001; w ¼ weight of This observed difference in immunohistochemistry each data point in x based on Tukey curve; u ¼ (x m)/((c (IHC) staining could provide a potential avenue for times s) þ epsilon); w ¼ (1 u2)2 for each data point; application in clinical screening. Tukey Biweight ¼ sum (w times x)/sum (w). RACGAP1 overexpression could serve as an Results independent prognostic factor for recurrent HCC The Tukey Biweight mean of RACGAP1 expression was Overexpression of RACGAP1 correlates with early 4.7876 for all the HCC biopsies studied. We have arbi- recurrence of HCC trarily considered samples with RACGAP1 expression Expression of RACGAP1 was significantly upregulat- above 4.78 as high RACGAP1 expressers and samples ed in the HCC biopsies compared with available paired with RACGAP1 expression less than 4.78 were considered adjacent matched histologically normal liver tissues as low expressers. We carried out univariate and multi- (N ¼ 24) as well as histologically normal liver tissues variate Cox regression analysis to determine whether (N ¼ 10) from patients who underwent surgery for RACGAP1 overexpression could serve as an independent metastatic colorectal cancer in the liver (Fig. 1A). More adverse survival prognostic factor for HCC. In univariate

Real-time PCR Migration 200 220 180 200 160 180 160 Figure 3. High levels of RACGAP1 140 MHCC97-HHepG2 PLC/PRF5Hep3B 140 120 correlated with high migration 120 100 rates of human HCC cells in vitro RACGAP1 100 80 as shown by Northern blot 80 60 analysis, real-time PCR assays, 60 β 40 40

-Actin of cells migrated No. and the Boyden migration Relative expression (%) expression Relative 20 20 in vitro chambers . 0 0

Hep3B HepG2 Hep3B HepG2

PLC/PRF5 PLC/PRF5 MHCC97-H MHCC97-H

www.aacrjournals.org Clin Cancer Res; 17(18) September 15, 2011 6045

Wang et al.

analysis, besides vascular invasion and cirrhosis, RACGAP1 expression was significantly associated with A 2 d 4 d 6 d the recurrence of HCC (Table 1). RACGAP1 expression gave the relative risk (RR) of 3.42 and P ¼ 0.001 in its Null C R1 R2 Null C R1 R2 Null C R1 R2 association with early recurrent disease following Racgap1 hepatic resection. Multivariate survival analysis using the Cox’s regression model also showed that RACGAP1 Hep3B Actin overexpression, vascular invasion, and cirrhosis were the only independent statistically significant risk factors for HCC recurrence (Table 1). Multivariate analysis sug- Racgap1 gested that the risk of developing early recurrent disease Actin was increased 2.7-fold for patients with high RACGAP1 MHCC97-H expression. Migration Additional correlation between RACGAP1 expression B and clinicopathologic features in the 76 HCC patients were 250 Hep3B 300 MHCC97H done using Fisher exact test, similar to that reported re- 250 cently by Li and colleagues (21). Table 2 shows that high 200 200 RACGAP1 expression significantly correlates with the early 150 150 recurrence of HCC (P ¼ 0.00004). P 100 ** = 0.002 100 Kaplan–Meier analysis on the HCC patients studied * P = 0.027 (N ¼ 76) over a period of 6 and 24 months was done 50 50 No. of migrated cells of migrated No. cells of migrated No. in relationship to their RACGAP1 expression. For both 0 0 analyses, it was determined that high RACGAP1 expression C R2 C R2 gave a significantly shorter recurrence-free duration com- C Invasion pared with low RACGAP1 expression (P ¼ 0.00005 and 25 Hep3B 40 MHCC97H P ¼ 0.000017, respectively, Fig. 2A and B). 35 20 30 25 Human HCC cells expressing high levels of RACGAP1 15 * P = 0.03 20 in vitro * P = 0.02 correlated with high migration rates 10 15 Four human HCC cell lines were employed to study the 10 5 effect of expression of RACGAP1 and their ability to cells of invaded No. cells of invaded No. 5 migrate in vitro. It was determined by Northern blot 0 0 analysis and real-time PCR assays that MHCC97-H and C R2 C R2 Hep3B cells expressed high levels of RACGAP1, whereas D HepG2 and PLC/PRF5 cells expressed negligible amount Null C siR2 of intrinsic RACGAP1 (Fig. 3). When the ability of these 4 cell lines to migrate were studied in vitro by the Boyden chambers, it was shown that MHCC97-H and Hep3B cells

gave much higher migration rates compared with HepG2 Migration and PLC/PRF5 cells (Fig. 3).

Blocking RACGAP1 expression by siRNA reduced cell migration and invasion activities in the RACGAP1-

positive HCC cell lines Hep3B and MHCC97-H Invasion The HCC cell lines Hep3B and MHCC97-H express high endogenous RACGAP1. The effect of silencing RACGAP1 expression in Hep3B and MHCC97-H cells was studied Figure 4. Silencing of RACGAP1 modulated the cell migration and invasive with the siRNA duplexes (R1 and R2) designed for properties of Hep3B and MHCC97-H cells with high endogenous RACGAP1 RACGAP1 (see Materials and Methods). Suppression of expression.A, time course tostudy siRNA-mediated silencing of RACGAP1in endogenous RACGAP1 expression for both Hep3B and Hep3B and MHCC97-H cells using Western blot. Null, untreated controls; MHCC97-H cells was most apparent at day 4 (Fig. 4A). C, RNAi universal negative control-treated cells; R1 and R2, pool 1 and pool 2, respectively, of a total of 3 sequences of siRNA duplexes designed to However, the suppression of RACGAP1 expression in knockdown RACGAP1 expression (see Materials and Methods). Expression MHCC97-H cells was not complete at day 2 with R2. This of actin acted as the internal loading control for mRNA. B, quantitation of cell is likely due to the relatively larger amount of RACGAP1 in migration in Boyden chambers. Data were obtained from 3 independent the MHCC97-H cells. For subsequent experiments, the experiments. P < 0.05 is statistically significant. C, quantitation of cell invasion in vitro P < siRNA duplex R2 was employed. . Data were obtained from 3 independent experiments. 0.05 is statistically significant. D, representative images showing the effect of siRNA The migration and invasiveness of Hep3B and MHCC97- treatment on the migration and invasion of MHCC97-H cells in vitro.DAPI- H cells were tested. At day 4 following transfection with R2, stained nuclei in blue, 2 days after seeding in insert chamber were shown.

6046 Clin Cancer Res; 17(18) September 15, 2011 Clinical Cancer Research

RACGAP1 and HCC Recurrence

A 10 MHCC97H Null Hep3B C Fold = 2.1 siR2 P = 0.001** 8 Null C siR2

6 RACGAP1 Fold = 3.0 P = 0.001** 4

(TUNEL positive) Actin 2 % of DNA fragmentation % of DNA fragmentation

0 Day 2 Day 4 Day 6 GTP-Cdc42 B Fold = 5.7 30 MHCC97H P = 0.001** 1 1.12 0.43 Null C Total Cdc42 25 siR2 Fold = 4.6 20 P = 0.001** GTP-Rac1 15 1 1.07 0.40 10 (TUNEL positive) Total Rac1

% of DNA fragmentation % of DNA fragmentation 5

0 Day 2 Day 4 Day 6 GTP-RhoA Null C siR2 1 1.07 1.10 Total RhoA

Figure 6. Pull-down and immune-blot assays showing that silencing C Hep3B MHCC97H of endogenous RACGAP1 by R2 in MHCC97-H cells impaired the Null C siR2 Null C siR2 GTPase activity of Cdc42 and Rac1 but not RhoA. The relative signal of each band in the GTP-bound form of the pull-down experiments was normalized to the total amount detected in the whole-cell lysates and RACGAP1 followed by normalization to the untreated control of the same cell lysate. Cleaved caspase-9 Null, untreated cells; C, RNAi negative control–transfected cells; siR2, RACGAP1 siRNA duplexes–transfected cells. Cleaved caspase-7 Cleaved PARP

Caspase-9 the migratory and invasive ability of both Hep3B and MHCC97-H cells were significantly reduced (Fig. 4B and Caspase-7 C). Representative images showing the effect of siRNA PARP treatment on the migration and invasion of MHCC97-H in vitro Actin cells was shown in Figure 4D. In addition, transfection with R2 induced significant DNA fragmentation in Hep3B and MHCC97-H cells at Figure 5. Silencing of RACGAP1 expression induced apoptosis in Hep3B day 4 and 6. At day 6, there was a 2.1-fold increase in and MHCC97-H cells. A, quantitation of DNA fragmentation (TUNEL assay) R2-transfected Hep3B cells that were stained positively for in vitro at d4 following siRNA treatment on Hep3B cells . Data were obtained the terminal deoxynucleotidyl transferase–mediated dUTP from 3 independent experiments. P < 0.05 is statistically significant. B, quantitation of DNA fragmentation (TUNEL assay) at d4 following siRNA nick end labeling (TUNEL) assay compared with both treatment on MHCC97-H cells in vitro.Datawereobtainedfrom3 null and control cells (P ¼ 0.001, Fig. 5A). The observed independent experiments. P < 0.05 is statistically significant. The bottom increase was more significant with the MHCC97-H panels show representative immunofluorescent images obtained after cells (5.7-fold, Fig. 5B). Corroborating with these observa- TUNEL assays. The nuclei of apoptotic cells were stained with green fluorescence. C, Western blot showing silencing of RACGAP1 expression tions, Western blot further showed that molecules activated caspases 9, 7, and PARP. Null, untreated cells; C, RNAi negative relating to apoptosis, including cleaved caspase 9, cleaved control–transfectedcells; siR2, RACGAP1 siRNA duplexes–transfected cells. caspase 7, and PARP were all activated in Hep3B and

www.aacrjournals.org Clin Cancer Res; 17(18) September 15, 2011 6047

Wang et al.

A CP: Cell cycle: G2–M DNA damage checkpoint regulation CP: ERK/MAPK signaling

CP: Rac signaling

Legend

Source (S) gene drives target (T) gene ST Source and target drive each other ST Source and target regulate each other ST Source gene inhibits target gene ST

Indirect relationship between source and target ST

CP: Mitotic roles of polo-like kinase CP: PI3K/AKT signaling CP: Wnt/β-catenin signaling

B A. MHCC97-H (siRNA2 vs. CON) B. HCC patient gene expression dataset [R (<24 mo) vs. NR]

Activates Increased expression Indirectly activates Indirectly acts on Decreased expression Phosphorylates/ Protein-protein binding dephosphorylates

Figure 7. A, IPA of global gene expression profiling of siRNA-treated MHCC97-H cells. Interactome showing the interactions between RACGAP1 and the most significant perturbed canonical pathways detected using IPA software for pathway analysis. The molecules highlighted in green were downregulated following siRNA treatment. B, comparison of the interactomes obtained following IPA analysis of differentially downregulated genes of siRNA-treated MHCC97-H cells and differentially upregulated genes in primary tumor samples of HCC patients with recurrent disease within 24 months. Molecules highlighted in green and red indicated differentially decreased and increased expression respectively.

6048 Clin Cancer Res; 17(18) September 15, 2011 Clinical Cancer Research

RACGAP1 and HCC Recurrence

C = = FDR ≤ 0.00001 FDR 0.008 FDR 0.004

11.8 11.5 11.0

11.1 10.74 10.08

10.4 9.98 9.16

9.7 9.22 Continued 8.24

Figure 7. ( )C, expression 2 expression expression 2 comparison of the signal intensity 2 9.0 8.46 for RACGAP1, ECT2, and PRC1 7.32 between early recurrent (R) and nonrecurrent (NR) HCC patients. 8.3 7.7 FDR (or Q value) of 0.05 implies 6.4 that 5% of significant tests will result in false positives. 7.6 6.94 5.48

6.9 6.18 4.56 ECT2 (219787_s_at) Log PRC1 (218009_s_at) Log Racgap1 (222077_s_at) Log 6.2 5.42 3.64

5.5 4.66 2.72

4.8 3.9 NR R NR R 1.8 NR R

MHCC97-H cells following silencing with the siRNA R2 Pathway Analysis (IPA) software tool (Ingenuity Systems, duplexes (Fig. 5C). This is likely to be associated with the Inc.). Functional and gene network analysis with dif- differentiation of RACGAP1 depleted cells as suggested by ferentially expressed genes identified for R2-treated and O’Brien and colleagues (22). universal negative control siRNA-treated cells revealed canonical pathways of mitotic roles of polo-like kinase, RACGAP1 regulates the GTPase activity of Cdc42 and PI3K/AKT signaling, Wnt/b-catenin signaling, cell-cycle Rac1 but not RhoA G2-M DNA damage checkpoint regulation, ERK/MAPK Silencing of endogenous RACGAP1 by R2 in MHCC97-H signaling, and Rac signaling were among the most signif- cells further impaired the GTPase activity of Cdc42 and icant canonical pathways altered. Transcripts that were Rac1 but inactive toward RhoA as shown by pull-down significantly silenced included RACGAP1, PRC1, SFN, assays (Fig. 6). Reduction in GTP-Cdc42 and GTP-Rac1 CDC2 (CDK1), AURKB, PRSS23, ECT2, SH3RF1, PAK1, activities was prominent in RACGAP1-depleted TGFB1, KIF23, and PPP2R5E (Fig. 7A). Importantly, when MHCC97-H cells (fold ratio ¼ 0.43 and 0.4, respectively) expression of these transcripts were analysed in a data- when compared with control cells, whereas the activity of base that we have established previously for HBV-posi- GTP-RhoA was not significantly affected (fold ratio ¼ 1.1, tive HCC patients, it was shown that all these transcripts Fig. 6), suggesting that the GAP domain of RACGAP1 could were differentially upregulated in patients that exhibited strongly stimulate Rac1 and Cdc42 GTPase activity but early recurrent disease (Fig. 7B). In addition, ECT2 and inactive toward RhoA. PRC1, two downstream genes of the RACGAP1 signaling pathway, were both significantly upregulated in early Functional pathway analysis revealed that transcripts recurrent HCC patients (Fig. 7C). These results strongly that were specifically silenced by R2 in MHCC97-H support the hypothesis that siRNA against RACGAP1 cells were differentially upregulated in HCC patients targeted genes in an interactome clinically relevant to that had early recurrent disease early HCC recurrence. To elucidate the signaling pathways that were significantly altered following the silencing of RACGAP1 Discussion expression by R2, we conducted pathway analysis with the differentially expressed genes identified in MHCC97-H Recurrent HCC disease is the major obstacle in achieving cells at day 4 post-R2 transfection using the Ingenuity long-term survival outcomes for the treatment of HCC via

www.aacrjournals.org Clin Cancer Res; 17(18) September 15, 2011 6049

Wang et al.

surgical resections (8, 9). Several studies have addressed the scripts detected on silencing RACGAP1 expression in clinical value of gene expression profiling in predicting MHCC97-H cells were KIF23, PRC1, PPP2RE, PPP2RC, the early recurrence of human HCC. RACGAP1, insofar as and PPP2RB that interacts directly with RACGAP1, whereas we know, has not been previously implicated in HCC. Cdc42 and Rac were also interconnected through AURKB Recently, we have shown the feasibility of conducting (Fig. 7A). Silencing of RACGAP1 expression in Hep3B and genome-wide expression analysis to derive gene signatures MHCC97-H cells with high endogenous RACGAP1 expres- to identify HCC patients who are at higher risk for early sion and high metastatic potential resulted in reduced cell recurrence and who could potentially benefit from more migration and invasion, reduction of activated Cdc42 intense surveillance and, possibly, adjuvant disease man- and Rac1, and strongly augmented DNA fragmentation agement. RACGAP1 is one of the genes identified in a that led to cell death in vitro. In this study, we were able signature for predicting risk of early recurrence (18). In this to correlate results obtained from siRNA-mediated silenc- study, we further showed, by Kaplan–Meier analysis, the ing of gene expression in MHCC97-H cells with gene potential that high RACGAP1 expression can be an inde- profiling results of primary HCC clinical samples. Tran- pendent adverse prognosticator for early HCC recurrence scripts that were targeted by siRNA against RACGAP1 genes after curative resection (P < 0.0005). The upregulation of namely KIF23, PRC1, PPP2RE, PPP2RC, PPP2RB, Cdc42, RACGAP1 and its association with early HCC recurrence and Rac were all upregulated in primary HCC biopsies were consistent with reports suggesting that RACGAP1 is from patients with early recurrent disease. RACGAP1 could linked to more frequent aggressive tumor phenotypes of regulate the activation of Rac1 and Cdc42 to trigger cyto- epithelial ovarian cancer (23), invasive cervical cancer (24), skeletal reorganization and, consequently, influence cell and high-grade breast cancer in the transition from pre- morphology, cell migration, chemotaxis, and the establish- invasive to invasive disease (25). ment of cell polarity that may lead to tumor metastases In vertebrate cells, RACGAP1 interacts with KIF23 to (31, 32). In concordance with this hypothesis, expression form the central spindlin complex which plays an essential of PRC1, the main downstream effector of Rac1 and Cdc42 role in cytokinesis (26, 27). RACGAP1 has been reported to (33, 34), was also found to be upregulated in samples of be involved in controlling the initiation of cytokinesis by early recurrent HCC patients, and its expression was regulating ECT2, which in turn induces the assembly of the repressed following R2-mediated silencing of RACGAP1 contractile ring and triggers the ingression of the cleavage in MHCC97-H cells (Fig. 7B). Therefore, it is likely that furrow to complete cytokinesis via interactions with Rac1, RACGAP1 could contribute to cancer progression and Cdc42, and RhoA (26, 28). RACGAP1 together with ANLN, metastasis through PRC1 to modulate cytoskeletal and ECT2, AURKB, PRC1, and KIF23 (MKLP1) are cytokinesis- transcription pathways that enhance cell motility, prolif- related cluster genes (29). In this study, analysis of the eration, and survival. differentially expressed transcripts obtained from R2- and control siRNA-treated MHCC97-H cells with the IPA soft- Disclosure of Potential Conflicts of Interest ware tool revealed that the most perturbed canonical No potential conflicts of interest were disclosed. pathway identified was the mitotic roles of polo-like kinase (Fig. 6A), the major mechanism involved in cell division Acknowledgments (30). The differentially downregulated transcripts obtained by comparing the gene expression profiles of R2- and We thank the NCC Tissue Repository for providing the tissue specimens control siRNA-treated MHCC97-H cells included KIF23, for this study and Professors Tang Zhao-You and Liu Bin Bin for providing the MHCC97-H cell line. PRC1, PPP2RE, PPP2RC, and PPP2RB that interacts between RACGAP1 and the mitotic roles of polo-like kinase– Grant Support mediated mitosis pathway. Further interactome mapping suggested that Cdc42, Rac, and actin cytoskeleton signaling This study was supported by grants from the National Medical Research were also interconnected through the ERK pathway via Council of Singapore, Biomedical Research Council of Singapore, and Singapore Millennium Foundation. AURKB (Fig. 7A). The costs of publication of this article were defrayed in part by the payment Although RACGAP1 plays key roles in controlling of page charges. This article must therefore be hereby marked advertisement in cell growth and differentiation, the mechanism by which accordance with 18 U.S.C. Section 1734 solely to indicate this fact. RACGAP1 contributes to HCC recurrence, however, Received March 8, 2011; revised July 19, 2011; accepted July 25, 2011; remains unclear. The differentially downregulated tran- published OnlineFirst August 8, 2011.

References 1. Llovet JM, Burroughs A, Bruix J. Hepatocellular carcinoma. Lancet 4. Poon D, Anderson BO, Chen LT, Tanaka K, Lau WY, Van Cutsem E, et al. 2003;362:1907–17. Management of hepatocellular carcinoma in Asia: consensus statement 2. Farazi PA, DePinho RA. Hepatocellular carcinoma pathogenesis: from from the Asian Oncology Summit 2009. Lancet Oncol 2009;10:1111–8. genes to environment. Nature Rev Cancer 2006;6:674–87. 5. Bruix J, Sherman MPractice Guidelines Committee, American Asso- 3. El-Serag HB, Rudolph KL. Hepatocellular carcinoma: epidemiology ciation for the Study of Liver Diseases. Management of hepatocellular and molecular carcinogenesis. Gastroenterology 2007;132:2557–76. carcinoma. Hepatology 2006;43:373.

6050 Clin Cancer Res; 17(18) September 15, 2011 Clinical Cancer Research

RACGAP1 and HCC Recurrence

6. El-Serag HB. Epidemiology of hepatocellular carcinoma in USA. cell line HCCLM3 with high metastatic potential: comparative geno- Hepatology Res 2007;37:S88–94. mic hybridization and multiplex fluorescence in situ hybridization. 7. Altekruse SF, McGlynn KA, Reichman ME. Hepatocellular carcinoma Cancer Genet Cytogenet 2005;158:180–3. incidence, mortality, and survival trends in the United States from 21. Li JC, Yang XR, Sun HX, Xu Y, Zhou J, Qiu SJ, et al. Up-regulation of 1875 to 2005. J Clin Oncol 2009;27:1485–91. Kruppel-like€ factor 8 promotes tumor invasion and indicates poor 8. Poon TP, Fan ST, Wong J. Risk factors, prevention, and management prognosis for hepatocellular carcinoma. Gastroenterology 2010;139: of postoperative recurrence after resection of hepatocellular carcino- 2146–57. ma. Ann Surg 2000;232:10–24. 22. O’Brien RN, Shen Z, Tachikawa K, Lee PA, Briggs SP. Quantitative 9. Portolani N, Coniglio A, Ghidoni S, Giovanelli M, Benetti A, Tiberio GA, proteome analysis of pluripotent cells by iTRAQ mass tagging reveals et al. Early and late recurrence after liver resection for hepatocellular post-transcriptional regulation of proteins required for ES cell self- carcinoma: prognostic and therapeutic implications. Ann Surg renewal. Mol Cell Proteomics 2010;9:2238–51. 2006;243:229–35. 23. Lu KH PA, Wang L, Marquez RT, Atkinson EN, Baggerly KA, Ramoth 10. Cowin PA, Anglesio M, Etemadmoghadam D, Bowtell DD. Profiling the LR, et al. Selection of potential markers for epithelial ovarian cancer cancer genome. Annu Rev Genomics Hum Genet 2010;11:133–59. with gene expression arrays and recursive descent partition analysis. 11. McDermott U, Downing JR, Statton MR. Genomics and the continuum Clin Cancer Res 2004;10:3291–300. of cancer care. N Engl J Med 2011;364:340–50. 24. Rosty C SM, Tsafrir D, Stransky N, Tsafrir I, Peter M, de Cremoux P, 12. Villanueva A, Hoshida Y, Toffanin S, Lachenmayer A, Alsinet C, Savic et al. Identification of a proliferation gene cluster associated with HPV R, et al. New strategies in hepatocellular carcinoma: genomic prog- E6/E7 expression level and viral DNA load in invasive cervical carci- nostic markers. Clin Cancer Res 2010;16:4688–94. noma. Oncogene 2005;24:7094–104. 13. Katoh H, Ojima H, Kokubu A, Saito S, Kondo T, Kosuge T, et al. 25. Ma XJ SR, Tuggle JT, Gaudet J, Enright E, McQuary P, Payette T, et al. Genetically distinct and clinically relevant classification of hepatocel- Gene expression profiles of human breast cancer progression. Proc lular carcinoma: putative therapeutic targets. Gastroenterology Natl Acad Sci U S A 2003;100:5974–9. 2007;133:1475–86. 26. Mishima M, Pavicic V, Gruneberg U, Nigg EA, Glotzer M. Cell cycle 14. Aravalli RN, Steer CJ, Cressman EN. Molecular mechanisms of regulation of central spindle assembly. Nature 2004;430:908–13. hepatocellular carcinoma. Hepatology 2008;48:2047–63. 27. Zhao WM, Fang G. MgcRacGAP controls the assembly of the con- 15. Iizuka N, Oka M, Yamada-Okabe H, Nishida M, Maeda Y, Mori N, et al. tractile ring and the initiation of cytokinesis. Proc Natl Acad Sci U S A Oligonucleotide microarray for prediction of early intrahepatic recur- 2005;102:13158–63. rence of hepatocellular carcinoma after curative resection. Lancet 28. Yuce O, Piekny A, Glotzer M. An ECT2-centralspindlin complex 2003;361:923–9. regulates the localization and function of RhoA. J Cell Biol 2005;170: 16. Kurokawa Y, Matoba R, Takemasa I, Nagano H, Dono K, Nakamori S, 571–82. et al. Molecular-based prediction of early recurrence in hepatocellular 29. Miller AL, Bement WM. Regulation of cytokinesis by Rho GTPase flux. carcinoma. J Hepatol 2004;41:284–91. Nat Cell Biol 2009;11:71–7. 17. Lee JS, Chu IS, Heo J, Calvisi DF, Sun Z, Roskams T, et al. Classi- 30. Lens SMA, Voest EE, Medema RH. Shared and separate functions of fication and prediction of survival in hepatocellular carcinoma by gene polo-like kinases and aurora kinases in cancer. Nature Rev Cancer expression profiling. Hepatology 2004;40:667–76. 2010;10:825–41. 18. Wang SM, Ooi LL, Hui KM. Identification and validation of a novel gene 31. Raftopoulou M HA. Cell migration: Rho GTPase lead the way. Dev Biol signature associated with the recurrence of human hepatocellular 2004;265:23–32. carcinoma. Clin Cancer Res 2007;13:6275–83. 32. Grewal S, Carver JG, Ridley AJ, Mardon HJ. Implantation of the 19. Hoshida Y, Villanueva A, Kobayashi M, Peix J, Chiang Derek Y, human embryo requires Rac1-dependent endometrial stromal cell Camargo A, et al. Gene expression in fixed tissues and out- migration. Proc Natl Acad Sci U S A 2008;105:16189–94. come in hepatocellular carcinoma. N Engl J Med 2008;359: 33. Bokoch GM. Biology of the p21-activated kinases. Annu Rev Biochem 1995–2004. 2003;72:743–81. 20. Yang J, Qin LX, Li Y, Ye SL, Liu YK, Gao DM, et al. Molecular 34. Molli PR, Li DQ, Murray BW, Rayala SK, Kumar R. PAK signaling in cytogenetic characteristics of the human hepatocellular carcinoma oncogenesis. Oncogene 2009;28:2545–55.

www.aacrjournals.org Clin Cancer Res; 17(18) September 15, 2011 6051

Upregulation of Rac GTPase-Activating Protein 1 Is Significantly Associated with the Early Recurrence of Human Hepatocellular Carcinoma

Suk Mei Wang, London Lucien P.J. Ooi and Kam M. Hui

Clin Cancer Res 2011;17:6040-6051. Published OnlineFirst August 8, 2011.

Updated version Access the most recent version of this article at: doi:10.1158/1078-0432.CCR-11-0557

Supplementary Access the most recent supplemental material at: Material http://clincancerres.aacrjournals.org/content/suppl/2015/12/11/1078-0432.CCR-11-0557.DC1

Cited articles This article cites 34 articles, 9 of which you can access for free at: http://clincancerres.aacrjournals.org/content/17/18/6040.full#ref-list-1

Citing articles This article has been cited by 11 HighWire-hosted articles. Access the articles at: http://clincancerres.aacrjournals.org/content/17/18/6040.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at Subscriptions [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/17/18/6040. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.