SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Patients and Methods
  4. Results
  5. Discussion
  6. References
  7. Supporting Information

Hepatocellular carcinoma (HCC) is one of the most common cancers and the third leading cause of death from cancer worldwide. HCC has a very poor prognosis because of tumor invasiveness, frequent intrahepatic spread, and extrahepatic metastasis. The molecular mechanism of HCC invasiveness and metastasis is poorly understood. The homeobox protein PROX1 is required for hepatocyte migration during mouse embryonic liver development. In this study, we show that high PROX1 protein expression in primary HCC tissues is associated with significantly worse survival and early tumor recurrence in postoperative HCC patients. Knockdown of PROX1 expression in HCC cells inhibited cell migration and invasiveness in vitro and HCC metastasis in nude mice while overexpression of PROX1 in HCC cells promoted these processes. PROX1's pro-metastasis activity is most likely attributed to its up-regulation of hypoxia-inducible factor 1α (HIF-1α) transcription and stabilization of HIF-1α protein by recruiting histone deacetylase 1 (HDAC1) to prevent the acetylation of HIF-1α, which subsequently induces an epithelial-mesenchymal transition response in HCC cells. We further demonstrated the prognostic value of using the combination of PROX1 and HDAC1 levels to predict postoperative survival and early recurrence of HCC. Conclusion: PROX1 is a critical factor that promotes HCC metastasis. (Hepatology 2013;58:692-705)

Abbreviations
ChIP

chromatin immunoprecipitation

Co-IP

coimmunoprecipitation

EMT

epithelial-mesenchymal transition

HCC

hepatocellular carcinoma

HDAC1

histone deacetylase 1

HIF-1

hypoxia-inducible factor 1

IHC

immunohistochemistry

IP/MS

immunoprecipitation/mass spectrometry

LNM

lymph node metastasis

mAb

monoclonal antibody

OS

overall survival

qrtPCR

quantitative real-time polymerase chain reaction

TMA

tissue microarray

TTR

time to tumor recurrence

Hepatocellular carcinoma (HCC) is one of the most common cancers and the third leading cause of death from cancer worldwide.[1] Although successful curative hepatectomy has significantly improved survival, the prognosis of HCC remains poor owing to tumor invasiveness, frequent intrahepatic spread, and extrahepatic metastasis. The molecular mechanism of HCC invasiveness and metastasis is ill-defined and its elucidation is fundamental to the improvement of HCC prognosis and treatment.

Epithelial-mesenchymal transition (EMT) is a process in which epithelial cells lose polarity and cell–cell adhesion, and are converted to a mesenchymal phenotype. The molecular hallmarks during EMT include down-regulation of epithelial markers (e.g., E-cadherin) and up-regulation of mesenchymal markers (e.g., vimentin).[2] EMT has a crucial role in the progression and metastasis of multiple cancers including HCC.[3, 4] EMT is triggered and controlled by signals cancer cells receive from their microenvironment. One of the major EMT triggers in cancers is the signaling through hypoxia-inducible factor 1 (HIF-1), activated via hypoxia-dependent or hypoxia-independent pathways.[5, 6] Enhanced HIF-1 activities have been reported to promote angiogenesis and invasiveness in HCC.[7, 8]

HIF-1 is composed of a hypoxia-inducible α subunit (HIF-1α) and a constitutively expressing β subunit (HIF-1β). HIF-1α is rapidly degraded under normoxic conditions.[5] During this process, HIF-1α is hydroxylated by prolyl hydroxylase domain proteins (PHDs) at two proline residues (P402 and P564) and subsequently interacts with the E3 ubiquitin ligase von Hippel-Lindau protein (VHL). Acetylation at K532 by ARD1 favors the interaction of HIF-1α with VHL and is coordinated with prolyl hydroxylation and ubiquitination,[9] leading to proteasomal degradation of HIF-1α. Under hypoxia conditions, the activities of PHDs are inhibited and HIF-1α acetylation can be prevented by histone deacetylase 1 (HDAC1).[10] Consequently, HIF-1α is stabilized, translocates to the nucleus, heterodimerizes with HIF-1β, and activates the expression of a broad range of genes including essential regulators for EMT.[11, 12]

The homeobox protein PROX1 is crucial for the development of multiple organs and tissues.[13] Gene knockout analysis in mice indicates that PROX1 is required for hepatocyte migration during embryonic liver development.[14] The role of PROX1 in cancer development has been studied in several cancers. A positive correlation is present between PROX1 protein expression and the malignancy grades of gliomas.[15] High PROX1 protein expression is also associated with poor clinical outcomes of colon cancer.[16] PROX1 is thought not to be responsible for the initiation of colon cancer but rather promotes cancer progression from benign to highly dysplastic phenotype.[17] The connection between PROX1 and HCC is rather obscure. Shimoda et al.[18] showed that the PROX1 messenger RNA (mRNA) level was down-regulated in HCC, and patients with high PROX1-expressing tumors had favorable prognosis compared with those with low PROX1 expression, but there was no statistically significant difference in the disease-free survival rates between these two PROX1 expression groups. However, in a later report, variable levels of PROX1 mRNA were observed in HCC tissues.[19] PROX1 protein expression in HCC samples, nevertheless, has not been investigated systematically. Whether PROX1 plays an important role in HCC invasiveness and metastasis remains unclear.

PROX1's activities in promoting hepatocyte migration during liver development and in colon cancer malignant transformation led us to hypothesize that PROX1 might be intimately involved in HCC invasiveness and metastasis. In this study, we discovered that high PROX1 protein expression in primary HCC tissues was associated with significantly worse clinical outcomes. PROX1 promoted HCC cell migration and invasiveness in vitro and HCC metastasis in nude mice. Mechanism studies revealed that PROX1 induced EMT response in HCC cells via up-regulating HIF-1α transcription and HIF-1α protein stability. We have thus identified PROX1 for the first time as a crucial factor that promotes HCC invasiveness and metastasis.

Patients and Methods

  1. Top of page
  2. Abstract
  3. Patients and Methods
  4. Results
  5. Discussion
  6. References
  7. Supporting Information
Patients and Clinical Samples

Primary HCC samples were obtained from cohort 1 (n = 227, collected between February 2005 to November 2006), cohort 2 (n = 125, collected between February 1999 to December 2003), and cohort 3 (n = 93) patients who had undergone curative hepatectomy at Zhongshan Hospital. Cohort 3 contained 43 patients with lymph node metastasis (LNM) and 50 patients without LNM randomly picked by computer. The study was approved by the Zhongshan Hospital Research Ethics Committee. Follow-up procedures were described previously.[20] Each patient was followed until March 2010, with the longest follow-up up to 72 months in cohort 1 and 126 months in cohort 2. The clinical characteristics of the HCC patients are presented in Supporting Table S1.

Tissue Microarray (TMA) and Immunohistochemistry (IHC)

Preparation of TMA and IHC procedures were performed as described.[20] The antibodies used in IHC are listed in Supporting Table S2. All IHC staining was independently assessed by two experienced pathologists. The staining intensity was graded from 0 to 2 (0, no staining; 1, weak; 2, strong) (Supporting Fig. S1). The staining extent was graded from 0 to 4 based on the percentage of immunoreactive tumor cells (0%, 1%-5%, 6%-25%, 26%-75%, 76%-100%) (Supporting Fig. S2). A score ranging from 0 to 8 was calculated by multiplying the staining extent score with the staining intensity score, resulting in a low (0-4) level or a high (6-8) level for each sample.

Plasmids and Cell Lines

The human HCC cell lines BEL-7402, Huh7, HepG2, QGY7701, QGY7703, SMCC7721, and embryonic kidney cell line HEK293T were obtained from the Cell Bank of Shanghai Institutes of Biological Sciences, Chinese Academy of Sciences. HCC cell line MHCC-97H was established at the Liver Cancer Institute.[21]

PROX1, HIF-1α, HIF-1b, and HDAC1 cDNAs were cloned downstream of Flag or HA tag in pcDNA3 (Invitrogen, Shanghai, China). Lentiviral vectors pLKO.1 TRC (Addgene 10879) and pWPI.1 (Addgene 12254) were used for constructing recombinant lentiviruses. Oligonucleotides encoding hairpin precursors for si259 (5′-TTTCCAGGAGCAACCA TAATT-3′, corresponding to nt259-279 of PROX1 ORF) and si1646 (5′-GGCTCTCCTTGTCGCTCA TAA-3′, corresponding to nt1646-1666 of PROX1 ORF) were used for generating short interference RNA (siRNA) constructs. A scrambled sequence (Scr) was used as a control. Synonymous mutations were introduced into the target sequence of si1646 (5′-GGCTCTCATTATCACTCATAA-3′, mutations underlined) in PROX1 ORF to generate si1646-resistant PROX1. pEGFP-Prox1 was generated by inserting PROX1 cDNA into pEGFP-C2 (ClonTech, Mountain View, CA). pGL2-HIF-1α contained the promoter (−572/+284) of HIF-1α in pGL2-Basic (Promega, Madison, WI). Cell transfection and luciferase reporter assays were performed as described.[22]

Immunoprecipitation and Mass Spectrometry (IP/MS)

The procedures are detailed in the Supporting Methods.

Coimmunoprecipitation (Co-IP) and GST Pulldown

Co-IP, GST pulldown, and western blot were performed as described.[22] Antibodies used in western blot experiments are listed in Supporting Table S2.

RNA Isolation and Quantitative Real-Time Polymerase Chain Reaction (qrtPCR)

The procedures are detailed in the Supporting Methods.

Chromatin Immunoprecipitation (ChIP)

Anti-PROX1 monoclonal antibody (mAb) (Cell Signaling Technology, Danvers, MA) was used to immunoprecipitate sonicated chromatins prepared from Huh7 or MHCC-97H cells. Preimmune IgG was used for specificity control. Immunoprecipitated DNA was quantitated for HIF-1α promoter (Ct[IP]) using qrtPCR (forward, 5′-GGTGAGGCGGGCTT GCGGGAG-3′; reverse, 5′-GAGGAGCTGAGGCAG CGTCAGGG-3′). DNA from 5% input was quantitated for HIF-1α promoter in parallel (Ct[Input]). The relative occupancy was calculated using the equation: 2^(Ct[Input]-Ct[IP]) *100%.

Cell Migration and Invasiveness Assay and 3D Cell Culture

The procedures are detailed in the Supporting Methods.

Nude Mice and Metastasis Assay

Animal experimental protocols were approved by the Animal Ethics Committee of Shanghai Medical College, Fudan University. Eight-week-old nude mice (BALB/c) were randomly divided into groups (six mice/group) before inoculation or injection. Cells were inoculated into the liver parenchyma of nude mice (BEL-7402 derived cells, 2.0 × 106 cells/mouse) or subcutaneously injected into nude mice (MHCC-97H derived cells, 1.0 × 107 cells/mouse). The mice were sacrificed after 8 weeks and the number of metastatic tumors was assessed by double-blinded evaluation.

Statistical Analysis

Statistical analysis was performed with SPSS v. 13.0. Kaplan-Meier analysis was used for survival analysis and the log-rank test was chosen to compare the difference. Pearson χ2 test or Fisher's exact test were employed to compare qualitative variables, while Student t test was used for quantitative variables. A Cox proportional hazards model was adopted for multivariate analysis. Receiver operating characteristic (ROC) curve analysis was applied to assess the predictive values of variables. P < 0.05 was considered statistically significant for all tests.

Results

  1. Top of page
  2. Abstract
  3. Patients and Methods
  4. Results
  5. Discussion
  6. References
  7. Supporting Information
High PROX1 Protein Expression in Primary HCC Is Associated With Worse Clinical Outcomes

IHC on tissue microarray was performed to measure PROX1 protein levels in primary HCC samples from 227 postoperative HCC patients (cohort 1) whose 5-year follow-up data were available. In all, 196 patients (86.3%) had a history of hepatitis B and 82 (36.1%) patients were HBV e antigen (HBeAg)-positive. Additionally, 185 patients (81.5%) showed liver cirrhosis (Supporting Table S1). Except one sample damaged during array preparation, the rest of the 226 samples were analyzed. PROX1 was observed mainly in the nuclei of tumor cells and absent in most stroma cells (Supporting Fig. S1). All the samples could be stratified into high PROX1 level (PROX1_hi) and low PROX1 level (PROX1_lo) according to IHC staining scores. Patients with a high serum α-fetoprotein (AFP) level, microvascular invasion, and advanced TNM stage appeared to possess high PROX1 levels in primary HCC tissues (Supporting Table S3).

The PROX1_hi group displayed significantly worse overall survival (OS) (median OS: 38.9 months versus >55 months; log-rank = 9.689, P = 0.002) and shortened time to tumor recurrence (TTR) (median TTR: 27.0 months versus >55 months; log-rank = 6.837, P = 0.009) compared to the PROX1_lo group (Fig. 1A). During the 5-year follow-up period, there were 43 deaths out of 80 patients (53.8%) of the PROX1_hi group compared with 52 deaths out of 146 patients (35.6%) of the PROX1_lo group. These observations were further validated in another cohort comprised of 125 postoperative HCC patients (cohort 2) with about 10-year follow-up data (Supporting Table S1). The second analysis confirmed that high PROX1 protein expression in primary HCC tissues was associated with significantly worse OS (P < 0.001) and shortened TTR (P < 0.001) (Fig. 1B).

image

Figure 1. High PROX1 protein expression in HCC is associated with worse clinical outcomes. Significant differences in OS (a,d), TTR (b,e), and early recurrence (c,f) between postoperative HCC patients with high and low PROX1 levels in primary HCC tissues. (A) cohort 1, n = 226; (B) cohort 2, n = 125. Log-rank P values are shown.

Download figure to PowerPoint

Two biologically different forms of HCC recurrence have been proposed. Early recurrence, which occurs within 2 years after treatment, mainly results from dissemination of metastatic HCC cells, while late recurrence is usually a result of a multicentric new tumor in liver.[23] Using 2 years as a cutoff value, the PROX1_hi group was shown to display a significantly higher early recurrence rate compared with the PROX1_lo group (P = 0.026 for cohort 1, P < 0.001 for cohort 2) (Fig. 1A,B). No significant difference was observed for late recurrence between the two groups (P = 0.275 for cohort 1, P = 0.093 for cohort 2) (Supporting Fig. S3).

HBeAg positivity, high AFP level, large tumor size, microvascular invasion, multiple tumors, and advanced TNM stage were found associated with worse OS and shortened TTR in univariate analysis (Table 1). To assess the correlation between high PROX1 level and other risk factors, a Cox proportional hazards analysis was performed, which indicated that high PROX1 level is an independent risk factor for worse OS (hazard ratio = 1.931, P = 0.002) and shortened TTR (hazard ratio = 1.602, P = 0.019) (Table 1).

Table 1. Univariate and Multivariate Analysis of Factors Associated with Survival and Recurrence of Cohort 1 HCC Patients
 Overall SurvivalTime to Tumor Recurrence
  Multivariate Multivariate
FeaturesUnivariate P ValueHazard Ratio95% CIP ValueUnivariate P ValueHazard Ratio95% CIP Value
  1. Abbreviations: AFP, alpha fetoprotein; ALT, alanine aminotransferase; NA, not adopted; NS, not significant.

Age (≤52 vs. >52 years)0.594  NA0.879  NA
Gender (female vs. male)0.386  NA0.333  NA
Hepatitis B history (yes vs. no)0.580  NA0.710  NA
Hepatitis B e antigen (positive vs. negative)0.038  NS0.0091.6271.110–2.3840.013
Liver cirrhosis: yes vs. no0.946  NA0.825  NA
AFP (<400 vs. ≥400 ng/mL)0.024  NS0.005  NS
Preoperative ALT (≤75 vs. >75 U/L)0.636  NA0.507  NA
Tumor size (≤5 vs. >5 cm)<0.0013.2332.020–5.176<0.001<0.0012.2641.491–3.438<0.001
Tumor encapsulation (complete vs. none)0.469  NA0.216  NA
Microvascular invasion (yes vs. no)<0.001  NS0.001  NS
Intrahepatic metastasis (yes vs. no)0.004  NS0.008  NS
Tumor differentiation0.027  NS0.235  NA
TNM stage (IIIa vs. II vs. I)<0.0011.6561.245–2.2020.001<0.0011.4891.173–2.0240.004
PROX1 level (low vs. high)0.0021.9311.272–2.9310.0020.0091.6021.082–2.3740.019
PROX1 Promotes HCC Cell Migration and Invasiveness

Association of high PROX1 expression in primary HCC samples with early recurrence suggests that PROX1 might play an important role in HCC invasiveness and metastasis. We then examined PROX1 expression in HCC cell lines possessing varying metastatic potentials.[24] PROX1 was readily detected in HepG2, Huh7, and MHCC-97H cell lines which are highly metastatic in nude mice.[24] However, the absence or a very low level of PROX1 was observed in HCC cell lines with a low metastatic potential (BEL-7402, QGY7701, QGY7703, SMCC7721) (Fig. 2A). Huh7, MHCC-97H, and BEL-7402 were used in the following investigations.

image

Figure 2. PROX1 promotes HCC cell migration and invasiveness in vitro. (A) Western blot analysis of PROX1 expression in HCC cell lines which have varying metastasis potentials. (B) Left, knockdown and rescue of PROX1 expression in MHCC-97H cells by si1646 and si1646-resistant PROX1 (Res). Scr, scrambled siRNA. Right, overexpression of PROX1 in BEL-7402 cells. (C) The migration (left) and invasiveness (right) of MHCC-97H cells were inhibited by si1646, which could be restored by expression of si1646-resistant PROX1. (D) The migration (left) and invasiveness (right) of BEL-7402 cells were promoted by overexpression of PROX1. Cell migration assays were repeated three times for each cell line and transwell invasion assays were performed in triplicate wells and repeated twice for each cell line. Representative images are shown. Scale bars, 100 μm.

Download figure to PowerPoint

Lentivirus-mediated knockdown and rescue of PROX1 expression in the highly invasive MHCC-97H cells[21] was performed to assess the functional involvement of PROX1 in HCC cell invasiveness in vitro. MHCC-97H-si1646 was established by infection of MHCC-97H cells with a lentivirus expressing si1646 precursor that achieved an 80% reduction in PROX1 expression (Fig. 2B). The control cells were infected with a lentivirus expressing a scrambled siRNA sequence. Compared with the control cells, MHCC-97H-si1646 cells displayed sharp declines in both cell migration and invasiveness, which could be prevented by the expression of si1646-resistant PROX1 (Fig. 2C). On the other hand, exogenous expression of PROX1 in BEL-7402 cells (Fig. 2B) enhanced cell migration and invasiveness (Fig. 2D). These results indicate that PROX1 promotes HCC cell migration and invasiveness in vitro.

PROX1 Interacts With HIF-1α

To explore the mechanism underlying PROX1's promotion of HCC cell invasiveness, IP/MS was conducted to identify key factors associated with PROX1. Flag-PROX1 produced in HEK293T cells was immunoprecipitated by anti-Flag mAb and coprecipitated proteins were visualized by silver staining after electrophoresis and identified by MS. One coprecipitated factor turned out to be HIF-1β (Fig. 3A). The association of PROX1 with HIF-1β was verified in HEK293T and Huh7 cells (Fig. 3B). As HIF-1β is able to heterodimerize with HIF-1α, it can be deduced that PROX1 may be associated with HIF-1α as well. This assumption was confirmed in HEK293T transfected with the expression plasmid for Flag-Prox1 and in Huh7 cells with endogenous proteins (Fig. 3C). Moreover, coexpressed EGFP-PROX1 and HA-HIF-1α were found to colocalize in Huh7 cells (Fig. 3D). Finally, GST-HIF-1α was expressed in Escherichia coli and the full-length PROX1 was produced via in vitro translation. The results of GST pulldown assays indicate that PROX1 directly interacts with HIF-1α (Fig. 3E).

image

Figure 3. PROX1 interacts with HIF-1α. (A) Identification of PROX1-associated factors using IP/MS. HEK293T cells were transfected with pFlag-Prox1 or empty vector. Identified HIF-1β peptides are shown. (B) Association of PROX1 with HIF-1β. HEK293T and Huh7 cells were transfected with pFlag-Prox1 or empty vector (control) and subjected to immunoprecipitation using anti-Flag mAb. Coimmunoprecipitated HIF-1β was detected using anti-HIF-1β antibody. NS: nonspecific signals. (C) Association of PROX1 with HIF-1α. HEK293T cells transfected with pFlag-Prox1 or empty vector (control) were subjected to immunoprecipitation using anti-Flag mAb. Endogenous PROX1 in Huh7 cells was immunoprecipitated using anti-PROX1 polyclonal antibody with rabbit IgG as nonspecific control. Coimmunoprecipitated HIF-1α was detected using anti-HIF-1α antibody. HIF-1α level was elevated in HEK293T cells transfected with pFlag-Prox1. (D) Colocalization of PROX1 and HIF-1α in HCC cells. Huh7 cells cotransfected with pEGFP-Prox1 and pHA-HIF-1α were stained using anti-HA mAb and visualized under fluorescent microscopy. Nuclei were stained with DAPI. (E) Mapping of the region(s) of HIF-1α that interacts with PROX1. Upper, the domain organization of HIF-1α and the deletion constructs. bHLH, basic helix-loop-helix; PAS, PER-ARNT-SIM; NTAD, N-terminal transactivation domain; CTAD, C-terminal transactivation domain. Lower, Input, Coomassie blue staining of one-tenth of the amount of each GST-HIF-1α fragment incubated with in vitro translated PROX1. PROX1 was detected by anti-PROX1 mAb after pulldown. (F) Upper, the domain organization of PROX1 and the deletion constructs. HD, homeodomain; PD1, prospero domain 1; PD2, prospero domain 2. Lower, Input, western blot of one-tenth of the amount of each Flag-PROX1 fragment used in GST pulldown assays. Flag-PROX1 fragments were detected by anti-Flag mAb after pulldown. IB, immunoblot; IP, immunoprecipitation.

Download figure to PowerPoint

To determine the regions mediating the interaction, several GST-fused HIF-1α fragments were produced (Fig. 3E) and each was tested for interaction with PROX1. The region spanning the residues 1-344 of HIF-1α appeared responsible for the interaction between HIF-1α and PROX1, and the amino-terminal 84 residues containing the basic helix-loop-helix domain might be the main interactive motif (Fig. 3E). Reciprocally, Flag-tagged PROX1 fragments were produced in HEK293T cells (Fig. 3F) and tested for interaction with GST-HIF-1α. The results of GST pulldown assays suggest that the amino-terminal two-thirds (amino acids 1-570) but not the putative DNA binding domain of PROX1 was responsible for the interaction with HIF-1α (Fig. 3F).

PROX1 Up-Regulates HIF-1α Expression and Induces EMT Response in HCC Cells

Next, we wondered whether PROX1 might regulate HIF-1α expression in HCC cells. Western blot analysis showed that HIF-1α expression was reduced by knockdown of PROX1 expression in Huh7 and MHCC-97H cells (Fig. 4A) but increased by overexpression of PROX1 in BEL-7402 and Huh7 cells (Fig. 4B). Meanwhile, the expression of E-cadherin was up-regulated in PROX1-knockdown cells and down-regulated in PROX1-overexpressing cells, which was reversely correlated with the change in the expression of vimentin (Fig. 4A,B). Importantly, exogenous expression of HIF-1α could counteract the effects of knockdown of PROX1 expression in Huh7 cells (Fig. 4C). These results indicate that PROX1 can induce EMT response in HCC cells and HIF-1α is most likely involved in this process.

image

Figure 4. PROX1 up-regulates HIF-1α expression and promotes EMT response in HCC cells. (A) Knockdown of PROX1 expression in MHCC-97H and Huh7 cells increased E-cadherin expression and reduced vimentin and HIF-1α expression. (B) Overexpression of PROX1 in BEL-7402 and Huh7 cells reduced E-cadherin expression and increased vimentin and HIF-1α expression. (C) HIF-1α counteracted knockdown of PROX1 expression. Huh7 cells were coinfected with the lentiviruses expressing PROX1 siRNA (si259 or si1646) precursor and Flag-HIF-1α. E-cadherin and vimentin were detected by respective antibodies. (D) MHCC-97H cells were infected with the lentivirus expressing either PROX1 siRNA (si259 or si1646) precursor or scrambled siRNA precursor (Scr). Morphological changes of MHCC-97H cells were observed under phase-contrast microscope. Scale bars, 100 μm.

Download figure to PowerPoint

We further investigated whether knockdown of PROX1 expression in the highly invasive MHCC-97H cells might cause any change in cell morphology and behavior. PROX1-knockdown cells (MHCC-97H-si259, MHCC-97H-si1646) appeared to have more of an epithelial-like morphology than the mesenchymal spindle-like morphology characteristic of MHCC-97H and the control MHCC-97H-Scr cells (Fig. 4D), implying that PROX1 is required for maintenance of the mesenchymal morphology of MHCC-97H cells.

PROX1 Activates HIF-1α Transcription

An increase in HIF-1α expression might result from activated HIF-1α transcription and/or enhanced HIF-1α protein stability. Indeed, HIF-1α mRNA levels were increased in PROX1-overexpressed BEL-7402 and Huh7 cells (Fig. 5A) but reduced in PROX1-knockdown Huh7 and MHCC-97H cells (Fig. 5B). Moreover, luciferase reporter assays indicated that PROX1 can activate HIF-1α promoter (Fig. 5C). Finally, ChIP-qrtPCR assays suggest that endogenous PROX1 is associated with HIF-1α promoter in Huh7 and MHCC-97H cells (Fig. 5D). Together, these results clearly indicate that PROX1 can activate HIF-1α transcription.

image

Figure 5. PROX1 activates HIF-1α transcription. (A) Overexpression of PROX1 increased HIF-1α mRNA expression. BEL-7402 and Huh7 cells were infected with the PROX1-expressing lentivirus or the vector lentivirus. HIF-1α mRNA was measured using qrtPCR. Means ± standard deviation (SD) from three independent experiments are presented as relative ratio to the control, whose value was taken as 1.0. (B) Knockdown of PROX1 expression repressed HIF-1α mRNA expression. Huh7 and MHCC-97H cells were infected with the lentivirus either expressing PROX1 siRNA precursor (si259 and si1646) or scrambled siRNA precursor (Scr). Means ± SD from three independent experiments are presented as relative ratio to the control, whose value was taken as 1.0. Inset, western blot verification of PROX1 expression. (C) PROX1 activated HIF-1α promoter. HEK293T cells were cotransfected with indicated plasmids. Means ± SD of normalized luciferase activity from three independent experiments are presented. (D) PROX1 associated with HIF-1α promoter. ChIP-qrtPCR was performed with sonicated chromatins immunoprecipitated from Huh7 and MHCC-97H cells by anti-PROX1 mAb or preimmune IgG (control). HIF-1α promoter segments were quantified using qrtPCR against 5% input. Means ± SD from three independent experiments are presented.

Download figure to PowerPoint

PROX1 Inhibits HIF-1α Acetylation

HDAC1 recruited by HIF-1α-associated factors can prevent acetylation of HIF-1α to stabilize HIF-1α.[10] Whether PROX1 employed this strategy was investigated. First, PROX1 was shown to interact with HDAC1 in HEK293T and Huh7 cells (Fig. 6A). Second, coexpressed EGFP-PROX1 and HA-HDAC1 colocalized in Huh7 cells (Fig. 6B). Next, HEK293T cells without endogenous PROX1 expression were cotransfected with PROX1 and Flag-HIF-1α expression constructs. Flag-HIF-1α was pulled down by anti-Flag mAb, followed by detection of HDAC1 and acetylated Flag-HIF-1α. With PROX1 coexpression, much more HDAC1 was coprecipitated with Flag-HIF-1α, while the amount of acetylated Flag-HIF-1α as detected by pan-Acetyl antibody was markedly decreased (Fig. 6C). Furthermore, HDAC1 expression was reduced by knockdown of PROX1 expression in Huh7 and MHCC-97H cells but increased by overexpression of PROX1 in BEL-7402 and Huh7 cells (Fig. 6D). Collectively, these results suggest that PROX1 inhibits HIF-1α acetylation by recruiting HDAC1 and up-regulating HDAC1 expression. However, the possibility that HDAC1 might be recruited by other HIF-1α-associated factors cannot be excluded.

image

Figure 6. PROX1 inhibits HIF-1α acetylation and the prognostic value of combining PROX1 and HDAC1 levels. (A) Association of PROX1 with HDAC1 in HEK293T and Huh7 cells. HEK293T cells transfected with pFlag-Prox1 or empty vector (control) were subjected to immunoprecipitation using anti-Flag mAb. Endogenous PROX1 in Huh7 cells was immunoprecipitated using anti-PROX1 polyclonal antibody with rabbit IgG as nonspecific control. HDAC1 was detected by anti-HDAC1 antibody. (B) Colocalization of PROX1 and HDAC1 in Huh7 cells. Cells cotransfected with pEGFP-Prox1 and pHA-HDAC1 were visualized using immunofluorescence microscopy. Nuclei were stained with DAPI. (C) PROX1 repressed HIF-1α acetylation. HEK293T cells were cotransfected with expression plasmids for PROX1 and Flag-HIF-1α. Flag-HIF-1α was immunoprecipitated by anti-Flag mAb and used as loading control. Acetylated Flag-HIF-1α was detected by pan-Acetyl antibody. PROX1 and HDAC1 were detected by anti-PROX1 mAb and anti-HDAC1 antibody, respectively. (D) PROX1 up-regulates HDAC1 expression. Overexpression of PROX1 in HCC cells increased HDAC1 expression. BEL-7402 and Huh7 cells were infected with the Flag-PROX1-expressing lentivirus or the vector lentivirus. Knockdown of PROX1 expression in HCC cells reduced HDAC1 expression. Huh7 and MHCC-97H cells were infected with the lentivirus either expressing PROX1 siRNA (si259 or si1646) precursor or scrambled siRNA precursor (Scr). (E) PROX1 and HDAC1 expression are positively correlated in HCC tissues. Representative IHC staining images of PROX1 and HDAC1 are shown. Scale bars, 100 μm. (F) Variables showing significance in the multivariate analysis (Table 1) and the combination of PROX1 and HDAC1 levels were adopted in ROC curve analysis. Larger areas under curve indicate more predictive power of the variable.

Download figure to PowerPoint

IHC on tissue microarray of primary HCC samples from cohort 1 patients was then performed to assess whether PROX1 and HDAC1 expression was correlated in HCC. A strong positive correlation between PROX1 and HDAC1 levels was observed (r = 0.419, P < 0.001) (Fig. 6E). The patients whose HCC samples showed above-medium levels (scoring >4) of both PROX1 and HDAC1 suffered exacerbated adverse clinical outcomes than the patients with below-medium levels (scoring ≤4) of both PROX1 and HDAC1 (OS P < 0.001, TTR P = 0.003, Supporting Fig. S4). Finally, the combination of PROX1 and HDAC1 levels appeared to have a better prognostic value for OS and early recurrence than PROX1 alone according to the ROC curve analysis (Fig. 6F; Supporting Table S4).

PROX1 Promotes HCC Metastasis in Nude Mice

Metastasis assays by inoculation or injection of HCC cells into nude mice were conducted to investigate whether PROX1 promotes HCC metastasis. First, BEL-7402-Prox1 was established by infection of the low metastatic BEL-7402 with the PROX1-expressing lentivirus. The control BEL-7402-Mock was generated by infection of BEL-7402 with the vector lentivirus. These cells were inoculated into the liver parenchyma of nude mice to create orthotopic xenograft models. After 8 weeks, the BEL-7402-Prox1 group displayed significantly more metastases in mesenteric lymph nodes (mean = 46.0 ± 35.9 per mouse) than the control group (mean = 2.3 ± 4.1 per mouse) (P = 0.014) (Fig. 7A). The number of lung metastatic nodules revealed by hematoxylin and eosin staining was also higher in the BEL-7402-Prox1 group, although the difference did not reach statistical significance (P = 0.135). The BEL-7402-Prox1 group also had larger tumors in liver (P = 0.012) (Fig. 7B). None of the mice in the BEL-7402-Prox1 group survived at day 65 postinoculation, while 50% of the mice in the control group did (P = 0.007) (Fig. 7C). Compared with the tumors of BEL-7402-Mock origin, a reduction in E-cadherin expression and increase in vimentin and HIF-1α expression were observed in the tumors derived from BEL-7402-Prox1 cells, suggesting that PROX1-induced EMT plays a key role in HCC metastasis in this model (Fig. 7D).

image

Figure 7. PROX1 promotes HCC metastasis in nude mice. (A) Overexpression of PROX1 in BEL-7402 cells promoted spontaneous mesenteric lymph node metastasis. BEL-7402-Mock and BEL-7402-PROX1 were inoculated into liver parenchyma in nude mice, respectively (2.0 × 106 cells/mouse, n = 6). Mesenteric lymph node metastatic nodules were counted manually and the number of metastases per mouse was presented as mean ± SD, and P value was determined by Student t test. Representative images of mesenteric lymph node metastases at day 56 postinoculation are shown with arrowheads indicating primary HCC and arrows pointing to mesenteric lymph nodes metastases. Comparison of tumor weight (B) and survival (C) between mice inoculated with BEL-7402-Mock and BEL-7402-Prox1. Tumor weight is presented as mean ± SD and P value was determined by Student t test. For survival, P value was calculated by two-sided log-rank test. (D) Representative images of IHC staining of PROX1, E-cadherin, vimentin, and HIF-1α in HCC of mice inoculated with BEL-7402-Mock or BEL-7402-Prox1. Scale bars, 20 μm. (E) A significantly higher proportion of patients whose HCC samples showed high PROX1 level (scoring >4) were found among HCC patients with LNM (n = 43) than those without (n = 50). P value was calculated by Fisher's exact test. (F) Knockdown of PROX1 expression in MHCC-97H cells reduced spontaneous lung metastasis. MHCC-97H-Scr and MHCC-97H-si1646 were injected subcutaneously into nude mice (1.0 × 107 cells/mouse, n = 6). Number of metastatic nodules per lung is presented as mean ± SD and P value was determined by Student t test. Representative images of H&E-stained lung tissues are shown. Scale bars, 200 μm.

Download figure to PowerPoint

Lymph node metastasis (LNM) was reported in about 7% of Chinese HCC cases, correlated with advanced disease and extremely poor survival.[25] We compared PROX1 expression in primary HCC samples between 43 patients with LNM and 50 randomly picked patients without LNM. Thirty (69.8%) out of 43 patients with LNM showed high PROX1 expression in primary HCC tissues in comparison to 22 (44%) out of 50 patients without LNM (P = 0.001) (Fig. 7E).

Lung metastasis frequently occurs in HCC. To further investigate the role of PROX1 in HCC lung metastasis, we used the MHCC-97H cell line, which has a high rate of spontaneous lung metastasis when injected subcutaneously.[21] MHCC-97H-si1646 and MHCC-97H-Scr were respectively established by infection of MHCC-97H cells with the lentivirus expressing si1646 precursor and the lentivirus expressing scrambled siRNA precursor. These cells were subcutaneously injected into nude mice. All the mice in both groups developed tumors. After 8 weeks, five out of six mice in the control group were found to have lung metastases (mean of number of metastatic nodules per lung = 4.7 ± 3.3), while none in the MHCC-97H-si1646 group developed lung metastases (P = 0.015) (Fig. 7F).

Discussion

  1. Top of page
  2. Abstract
  3. Patients and Methods
  4. Results
  5. Discussion
  6. References
  7. Supporting Information

Although several genes associated with HCC metastasis have been identified in the past few years,[24, 26-31] the molecular mechanism of HCC invasiveness and metastasis is still not well understood. Much evidence has been presented to indicate that EMT is crucial for cancer invasiveness and metastasis.[3, 4, 32] Nevertheless, a more in-depth understanding of the factors promoting EMT in HCC and HCC metastasis is urgently needed in order to identify specific biomarkers for improving HCC prognosis and treatment. Because of its essential function in promoting hepatocyte migration during mouse embryonic liver development, PROX1's role in HCC invasiveness and metastasis is intriguing but unsolved. In this study, we demonstrated that PROX1 promoted HCC cell migration and invasiveness in vitro and HCC metastasis to lymph nodes and lung in nude mice. The molecular mechanism underlying PROX1's Pro-metastasis activity is most likely attributed to its up-regulation of HIF-1α transcription and HIF-1α protein stability, which consequently induces an EMT response in HCC cells. Accordingly, high PROX1 expression in primary HCC tissues is associated with significantly worse postoperative survival and early tumor recurrence. Collectively, we pinpointed PROX1 for the first time as a critical factor that promotes HCC metastasis.

The role of PROX1 in HCC development was unsolved because previous reports had not investigated the role of PROX1 in HCC metastasis. Dudas et al.[19] analyzed PROX1 mRNA levels in small numbers of normal, cirrhotic, and HCC liver samples using quantitative RT-PCR and northern blot, but no clear correlation with disease was observed. In another earlier work, however, Shimoda et al.[18] also used quantitative RT-PCR to evaluate PROX1 mRNA levels in 52 HCC samples and identified a positive correlation between PROX1 mRNA level and better prognosis. The discrepancy between the latter results and our data reported in this work might be attributed to several factors. First, PROX1 mRNA expression rather than protein expression was examined in the previous report. High PROX1 mRNA expression itself does not necessarily result in high PROX1 protein expression. Second, the differences in HCC patients' background, especially a history of infection with HCC-inducing pathogens, may have a profound influence on HCC development. For example, only one-fourth of the HCC patients in Shimoda et al.'s study had a history of HBV infection. In contrast, more than 70% of Chinese HCC patients are infected with HBV. In our case, the rate of HBV infection is 86.3% for cohort 1 and 84% for cohort 2. There were 31 non-HBV patients in cohort 1, and among them 11 and 20 belong to the high PROX1 level group and low PROX1 level group, respectively. Interestingly, there is no statistical significance for the differences in OS and TTR between these two groups (Supporting Fig. S5). Nevertheless, it is premature to reach a conclusion because the number of non-HBV patients in cohort 1 is rather small. Third, a relatively small number of HCC patients (n = 52) were included in the previous study, although statistical significance was strong (P = 0.014).[18]

PROX1-mediated up-regulation of HIF-1α expression in HCC cells occurs at two levels: the activation of HIF-1α transcription and the stabilization of HIF-1α protein through prevention of HIF-1α acetylating. Although it contains a DNA binding domain located at the carboxyl terminal one-third, PROX1 does not usually regulate transcription by directly binding to target gene promoter DNA. Instead, PROX1 often serves as a coregulator for transcription factors.[22, 33] HIF-1α transcription is activated by nuclear factor kappa B (NF-κB),[34] which is, to date, not known to partner with PROX1. Therefore, how PROX1 activates HIF-1α transcription remains an interesting topic for future study. On the other hand, recruitment of HDAC1 to stabilize HIF-1α has been reported to be employed by metastasis-associated protein 1 (MTA1) in breast cancer cells.[10] Interestingly, MTA1 belongs to a family of proteins associated with tumor metastasis. The similarity between PROX1 and MTA1 in HDAC1 recruitment to stabilize HIF-1α suggests a possible common strategy cancer cells may utilize to achieve metastasis.

Given the extremely poor prognosis of HCC, biomarkers for improving HCC prognosis and intervention are urgently needed. According to our ROC curve analysis, tumor size and TNM stage are the most effective predictors of survival and early recurrence among postoperative HCC patients. The combination of PROX1 and HDAC1 levels appears a potentially useful predictor for survival and early recurrence. Since a high PROX1 level is an independent risk factor for poor OS and shortened TTR (Table 1), we speculate that combining PROX1/HDAC1 levels with tumor size and TNM stage may increase the predictive power. This hypothesis should be tested in a prospective study with postoperative HCC patients. From a therapeutic viewpoint, the concurrent increase in HDAC1 protein expression in HCCs with high PROX1 level suggests that inhibitors of HDAC1 might be potent drugs against HCC metastasis. On the other hand, elucidation of the molecular mechanism leading to high PROX1 expression in certain HCC patients and discovery of the means to suppress PROX1 activity may provide novel therapies for preventing HCC metastasis.

References

  1. Top of page
  2. Abstract
  3. Patients and Methods
  4. Results
  5. Discussion
  6. References
  7. Supporting Information
  • 1
    Ferlay J, Shin HR, Bray F, Forman D, Mathers C, Parkin DM. Estimates of worldwide burden of cancer in 2008: Globocan 2008. Int J Cancer 2010;127:2893-2917.
  • 2
    Zeisberg M, Neilson EG. Biomarkers for epithelial-mesenchymal transitions. J Clin Invest 2009;119:1429-1437.
  • 3
    Van Zijl F, Zulehner G, Petz M, Schneller D, Kornauth C, Hau M, et al. Epithelial-mesenchymal transition in hepatocellular carcinoma. Future Oncol 2009;5:1169-1179.
  • 4
    Yang J, Weinberg RA. Epithelial-mesenchymal transition: at the crossroads of development and tumor metastasis. Dev Cell 2008;14:818-829.
  • 5
    Majmundar AJ, Wong WJ, Simon MC. Hypoxia-inducible factors and the response to hypoxic stress. Mol Cell 2010;40:294-309.
  • 6
    Semenza GL. Hypoxia-inducible factors: mediators of cancer progression and targets for cancer therapy. Trends Pharmacol Sci 2012;33:207-214.
  • 7
    Maxwell PH, Dachs GU, Gleadle JM, Nicholls LG, Harris AL, Stratford IJ, et al. Hypoxia-inducible factor-1 modulates gene expression in solid tumors and influences both angiogenesis and tumor growth. Proc Natl Acad Sci U S A 1997;94:8104-8109.
  • 8
    Rosmorduc O, Housset C. Hypoxia: a link between fibrogenesis, angiogenesis, and carcinogenesis in liver disease. Semin Liver Dis 2010;30:258-270.
  • 9
    Jeong JW, Bae MK, Ahn MY, Kim SH, Sohn TK, Bae MH, et al. Regulation and destabilization of HIF-1alpha by ARD1-mediated acetylation. Cell 2002;111:709-720.
  • 10
    Yoo YG, Kong G, Lee MO. Metastasis-associated protein 1 enhances stability of hypoxia-inducible factor-1alpha protein by recruiting histone deacetylase 1. EMBO J 2006;25:1231-1241.
  • 11
    Mak P, Leav I, Pursell B, Bae D, Yang X, Taglienti CA, et al. ERbeta impedes prostate cancer EMT by destabilizing HIF-1alpha and inhibiting VEGF-mediated Snail nuclear localization: implications for gleason grading. Cancer Cell 2010;17:319-332.
  • 12
    Yang MH, Wu MZ, Chiou SH, Chen PM, Chang SY, Liu CJ, et al. Direct regulation of twist by HIF-1alpha promotes metastasis. Nat Cell Biol 2008;10:295-305.
  • 13
    Elsir T, Smits A, Lindström MS, Nistér M. Transcription factor PROX1: its role in development and cancer. Cancer Metastasis Rev 2012;31:793-805.
  • 14
    Sosa-Pineda B, Wigle JT, Oliver G. Hepatocyte migration during liver development requires Prox1. Nat Genet 2000;25:254-255.
  • 15
    Elsir T, Eriksson A, Orrego A, Lindstrom MS, Nister M. Expression of Prox1 is a common feature of high-grade malignant astrocytic gliomas. J Neuropathol Exp Neurol 2010;69:129-138.
  • 16
    Skog M, Bono P, Lundin M, Lundin J, Louhimo J, Linder N, et al. Expression and prognostic value of transcription factor PROX1 in colorectal cancer. Br J Cancer 2011;105:1346-1351.
  • 17
    Petrova TV, Nykanen A, Norrmen C, Ivanov KI, Andersson LC, Haglund C, et al. Transcription factor Prox1 induces colon cancer progression by promoting the transition from benign to highly dysplastic phenotype. Cancer Cell 2008;13:407-419.
  • 18
    Shimoda M, Takahashi M, Yoshimoto T, Kono T, Ikai I, Kubo H. A homeobox protein, Prox1, is involved in the differentiation, proliferation, and prognosis in hepatocellular carcinoma. Clin Cancer Res 2006;12:6005-6011.
  • 19
    Dudas J, Mansuroglu T, Moriconi F, Haller F, Wilting J, Lorf T, et al. Altered regulation of Prox1-gene-expression in liver tumors. BMC Cancer 2008;8:92.
  • 20
    Zhu XD, Zhang JB, Zhuang PY, Zhu HG, Zhang W, Xiong YQ, et al. High expression of macrophage colony-stimulating factor in peritumoral liver tissue is associated with poor survival after curative resection of hepatocellular carcinoma. J Clin Oncol 2008;26:2707-2716.
  • 21
    Ye QH, Qin LX, Forgues M, He P, Kim JW, Peng AC, et al. Predicting hepatitis B virus-positive metastatic hepatocellular carcinomas using gene expression profiling and supervised machine learning. Nat Med 2003;9:416-423.
  • 22
    Qin J, Gao DM, Jiang QF, Zhou Q, Kong YY, Wang Y, et al. Prospero-related homeobox (Prox1) is a corepressor of human liver receptor homolog-1 and suppresses the transcription of the cholesterol 7-alpha-hydroxylase gene. Mol Endocrinol 2004;18:2424-2439.
  • 23
    Portolani N, Coniglio A, Ghidoni S, Giovanelli M, Benetti A, Tiberio GA, et al. Early and late recurrence after liver resection for hepatocellular carcinoma: prognostic and therapeutic implications. Ann Surg 2006;243:229-235.
  • 24
    Fu J, Chen Y, Cao J, Luo T, Qian YW, Yang W, et al. P28gank overexpression accelerates hepatocellular carcinoma invasiveness and metastasis via phosphoinositol 3-kinase/akt/hypoxia-inducible factor-1alpha pathways. Hepatology 2011;53:181-192.
  • 25
    Xiaohong S, Huikai L, Feng W, Ti Z, Yunlong C, Qiang L. Clinical significance of lymph node metastasis in patients undergoing partial hepatectomy for hepatocellular carcinoma. World J Surg 2010;34:1028-1033.
  • 26
    Chen L, Chan TH, Yuan YF, Hu L, Huang J, Ma S, et al. CHD1L promotes hepatocellular carcinoma progression and metastasis in mice and is associated with these processes in human patients. J Clin Invest 2010;120:1178-1191.
  • 27
    Chuma M, Sakamoto M, Yasuda J, Fujii G, Nakanishi K, Tsuchiya A, et al. Overexpression of cortactin is involved in motility and metastasis of hepatocellular carcinoma. J Hepatol 2004;41:629-636.
  • 28
    Lau SH, Sham JS, Xie D, Tzang CH, Tang D, Ma N, et al. Clusterin plays an important role in hepatocellular carcinoma metastasis. Oncogene 2006;25:1242-1250.
  • 29
    Lee TK, Poon RT, Yuen AP, Ling MT, Kwok WK, Wang XH, et al. Twist overexpression correlates with hepatocellular carcinoma metastasis through induction of epithelial-mesenchymal transition. Clin Cancer Res 2006;12:5369-5376.
  • 30
    Sun CK, Ng KT, Sun BS, Ho JW, Lee TK, Ng I, et al. The significance of proline-rich tyrosine kinase2 (pyk2) on hepatocellular carcinoma progression and recurrence. Br J Cancer 2007;97:50-57.
  • 31
    Wu X, Jia HL, Wang YF, Ren N, Ye QH, Sun HC, et al. HTPAP gene on chromosome 8p is a candidate metastasis suppressor for human hepatocellular carcinoma. Oncogene 2006;25:1832-1840.
  • 32
    Yang MH, Chen CL, Chau GY, Chiou SH, Su CW, Chou TY, et al. Comprehensive analysis of the independent effect of Twist and Snail in promoting metastasis of hepatocellular carcinoma. Hepatology 2009;50:1464-1474.
  • 33
    Song KH, Li T, Chiang JY. A prospero-related homeodomain protein is a novel co-regulator of hepatocyte nuclear factor 4alpha that regulates the cholesterol 7alpha-hydroxylase gene. J Biol Chem 2006;281:10081-10088.
  • 34
    Rius J, Guma M, Schachtrup C, Akassoglou K, Zinkernagel AS, Nizet V, et al. NF-kappaB links innate immunity to the hypoxic response through transcriptional regulation of hif-1alpha. Nature 2008;453:807-811.

Supporting Information

  1. Top of page
  2. Abstract
  3. Patients and Methods
  4. Results
  5. Discussion
  6. References
  7. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
hep26398-sup-0001-suppfig1.tif5562KSupporting Information Figure 1
hep26398-sup-0002-suppfig2.tif3731KSupporting Information Figure 2
hep26398-sup-0003-suppfig3.tif1200KSupporting Information Figure 3
hep26398-sup-0004-suppfig4.tif1652KSupporting Information Figure 4
hep26398-sup-0005-suppfig5.tif1219KSupporting Information Figure 5
hep26398-sup-0006-supptab1.doc34KTable S1. Clinicopathologic features of patients from two cohorts
hep26398-sup-0007-supptab2.doc35KTable S2. Primary antibodies for western blot, immunohistochemistry and co-immunoprecipitation
hep26398-sup-0008-supptab3.doc80KTable S3. Comparison of clinicopathologic profiles between HCC patients of high PROX1 level and low PROX1 level in cohort 1
hep26398-sup-0009-supptab4.doc39KTable S4. Prognostic values of variables for overall survival, recurrence and early recurrence
hep26398-sup-0010-suppinfo.doc42KSupporting Information

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.