Synergistic role of sprouty2 inactivation and c-Met up-regulation in mouse and human hepatocarcinogenesis


  • Potential conflict of interest: Nothing to report.


Sprouty2 (Spry2), a negative feedback regulator of the Ras/mitogen-activated protein kinase (MAPK) pathway, is frequently down-regulated in human hepatocellular carcinoma (HCC). We tested the hypothesis that loss of Spry2 cooperates with unconstrained activation of the c-Met protooncogene to induce hepatocarcinogenesis via in vitro and in vivo approaches. We found coordinated down-regulation of Spry2 protein expression and activation of c-Met as well as its downstream effectors extracellular signal-regulated kinase (ERK) and v-akt murine thymoma viral oncogene homolog (AKT) in a subset of human HCC samples with poor outcome. Mechanistic studies revealed that Spry2 function is disrupted in human HCC via multiple mechanisms at both transcriptional and post-transcriptional level, including promoter hypermethylation, loss of heterozygosity, and proteosomal degradation by neural precursor cell expressed, developmentally down-regulated 4 (NEDD4). In HCC cell lines, Spry2 overexpression inhibits c-Met–induced cell proliferation as well as ERK and AKT activation, whereas loss of Spry2 potentiates c-Met signaling. Most importantly, we show that blocking Spry2 activity via a dominant negative form of Spry2 cooperates with c-Met to promote hepatocarcinogenesis in the mouse liver by sustaining proliferation and angiogenesis. The tumors exhibited high levels of activated ERK and AKT, recapitulating the subgroup of human HCC with a clinically aggressive phenotype. Conclusion: The occurrence of frequent genetic, epigenetic, and biochemical events leading to Spry2 inactivation provides solid evidence that Spry2 functions as a tumor suppressor gene in liver cancer. Coordinated deregulation of Spry2 and c-Met signaling may be a pivotal oncogenic mechanism responsible for unrestrained activation of ERK and AKT pathways in human hepatocarcinogenesis. (HEPATOLOGY 2010)

Human hepatocellular carcinoma (HCC) is a leading cause of cancer-related deaths worldwide, with limited treatment options and high mortality rate.1 Hepatocarcinogenesis is a multiphase process involving the deregulation of various signaling pathways.1 In particular, activation of the Ras/mitogen-activated protein kinase (MAPK) pathway is ubiquitous in human HCC.2 The importance of the Ras/MAPK cascade in hepatocarcinogenesis is underscored by the finding that treatment with Sorafenib, a Raf inhibitor, significantly prolongs the overall survival of HCC patients.3 Activated Ras induces the Raf-Mitogen-activate protein kinase kinase (MEK)-MAPK cascade, which regulates various cellular responses, including cell proliferation, survival, and differentiation.4 Mutations of either Ras or its downstream effector B-Raf are the most common targets for somatic gain-of-function mutations in human cancers.4

However, Ras and B-Raf are rarely mutated in human HCC.5, 6 Thus, it remains unclear how the Ras cascade is activated during hepatocarcinogenesis in the absence of Ras and B-Raf mutations. Overexpression of the c-Met protooncogene and loss of the MAPK inhibitor Sprouty2 (Spry2) have been implicated as possible mechanisms leading to unconstrained induction of the Ras/MAPK pathway in the absence of Ras mutations.1, 2

The c-Met gene encodes the receptor tyrosine kinase for hepatocyte growth factor (HGF),7, 8 and is involved in multiple cellular responses, including proliferation, survival, migration, tumorigenesis, and metastasis.9, 10 Activation of c-Met signaling has been implicated in various tumor types. In human liver cancer, c-Met is overexpressed, and its up-regulation is associated with patient's poor prognosis.11, 12 Genomic studies also identified a c-Met–regulated gene expression signature defining an aggressive biologic phenotype in human HCC.13 Furthermore, overexpression of human c-Met in mouse hepatocytes promotes malignant transformation.14

Sprouty (Spry)/Sprouty-related EVH1 domain containing (Spred) proteins are evolutionarily conserved inhibitors of receptor tyrosine kinases.15, 16 After induction by the Ras/MAPK pathway, Spry proteins function as negative inhibitors of Ras/MAPK activation by disrupting growth-factor-receptor bound-2–son of sevenless (GRB2-SOS) complex or inhibiting Raf.15, 16 Among the four mammalian homologs of Drosophila Spry (Spry1 to Spry4), Spry2 is frequently down-regulated in multiple tumor types, and its loss may contribute to abnormal activation of Ras/signaling in cancer.15, 16 In human HCC, previous reports showed frequent Spry2 down-regulation, presumably attributable to promoter hypermethylation and loss of DNA copy number at 13q31, and that its inactivation triggers elevated fibroblast growth factor (FGF)/MAPK signaling.17 In mouse models, it was recently demonstrated that inactivation of Spry2 using the dominant negative Spry2 (Spry2Y55F) construct cooperates with activated β-catenin to promote hepatocarcinogenesis in vivo, strengthening the hypothesis that Spry2 is a bona fide tumor suppressor gene in the liver.18

Most studies on the functional role of Spry2 have focused on Spry2 regulation over FGF and EGF signaling so far, whereas the functional interactions between c-Met activation and Spry2 down-regulation have been poorly characterized. To date, only one paper reported that overexpression of Spry2 inhibits HGF-mediated cell growth in SK-LMS-1 human leiomyosarcoma cells.19 Because of the frequent, concomitant induction of c-Met signaling and loss of Spry2 expression during hepatocarcinogenesis, we hypothesized that loss of Spry2 leads to unrestrained activation of c-Met signaling in HCC.

We investigated the molecular mechanisms of down-regulation of Spry2 in hepatocarcinogenesis and characterized the biochemical and genetic interactions between Spry2 and c-Met using human HCC samples, and in vitro and in vivo models. We provide robust evidence that disruption of Spry2 and c-Met balance might be a dominant oncogenic event in HCC leading to the activation of extracellular signal-regulated kinase (ERK) and v-akt murine thymoma viral oncogene homolog (AKT) pathways and uncontrolled tumor growth.


EGFR, epithelial growth factor receptor; ERK, extracellular signal-regulated kinase; FGF, fibroblast growth factor; HCA, hepatocellular adenoma; HCC, hepatocellular carcinoma; HCCB, hepatocellular carcinoma with better outcome; HCCP, hepatocellular carcinoma with poorer outcome; HE, hematoxylin-eosin; HGF, hepatocyte growth factor; LOH, loss of heterozygosity; MAPK, mitogen-activated protein kinase; mRNA, messenger RNA; mTOR, mammalian target of rapamycin; NEDD4, neural precursor cell expressed, developmentally down-regulated 4; PCR, polymerase chain reaction; siRNA, small interfering RNA; SD, standard deviation; SL, surrounding nontumor livers; Spry2, Sprouty2; VEGF, vascular endothelial growth factor.

Patients and Methods

Human Tissue Samples.

Ten normal livers, 82 surgically resected HCCs, and corresponding surrounding nontumor livers (SL) were used. Patients' clinicopathological features are shown in Supporting Table 1. HCCs were divided into two groups based on patient's survival length: HCC with poorer outcome (HCCP; n = 44), and HCC with better outcome (HCCB; n = 38), which were characterized by a shorter (<3 years) or longer (>3 years) survival after liver partial resection, respectively.20 Tissues were kindly provided by Dr. Snorri S. Thorgeirsson (National Cancer Institute, Bethesda, MD). Institutional Review Board approval was obtained at participating hospitals and the National Institutes of Health.

Cell Lines and Treatments.

Human HCC cell lines were subjected to either small interfering RNA (siRNA) or demethylating treatments as reported in Supporting Materials. Transient transfection experiments with either Spry2 wild-type complementary DNA in pCMV6-XL vector (OriGene Technologies, Rockville, MD) or a dominant negative form of Spry2 (Spry2Y55F)21 in a pCS2 vector were performed on HCC cell lines. Cell viability, apoptosis, and vascular endothelial growth factor-α (VEGF-α) secretion were determined by WST-1 Cell Proliferation Reagent, Cell Death Detection Elisa Plus kit (Roche Molecular Biochemicals, Indianapolis, IN), and VEGF-α enzyme-linked immunosorbent assay kit (R&D Systems, Minneapolis, MN), respectively, following the manufacturers' instructions. Experiments were repeated at least three times in triplicate.

Hydrodynamic Injection and Mice Monitoring.

Ink4A/Arf−/− mice were obtained from Mouse Model of Human Cancer Consortium (National Cancer Institute), and wild-type FVB/N mice were purchased from the Charles River Laboratory (Wilmington, MA). Hydrodynamic injection was performed as described.22 All mice were housed, fed, and treated in accordance with protocols approved by the committee for animal research at UCSF.

Statistical Analysis.

Student t and Tukey-Kramer tests were used to evaluate statistical significance. Values of P < 0.05 were considered significant. Data are expressed as means ± standard deviation (SD).

See Supporting Information for more detailed descriptions of Materials and Methods.


Frequent Down-Regulation of Spry2 in Aggressive Human HCC.

A recent study suggests that suppression of Spry2 triggers uncontrolled activation of the c-Met mitogenic and survival cascades, namely, the MAPK and the AKT pathways, in leiomyosarcoma cells.19 To determine whether this applies to human liver cancer, we investigated the levels of Spry2 and c-Met in a collection of normal livers, HCC and corresponding non-neoplastic surrounding livers (Fig. 1; Supporting Fig. 1). Progressive up-regulation of Spry2 protein expression was detected from normal livers to HCC with better outcome (HCCB). In the latter subclass, only six HCCs (15.8%) showed down-regulation of Spry2 at protein levels. In striking contrast, a significant down-regulation of Spry2 protein characterized most of HCC with poorer outcome (HCCP) (36/44, 81.8%; P = 1.54E-13). A similar trend of Spry2 expression at messenger RNA (mRNA) levels was detected (Fig. 1B). Intriguingly, 2 of 38 HCCB and 12 of 44 HCCPs with low levels of Spry2 protein displayed an increase of Spry2 mRNA expression (Fig. 1C), suggesting the existence of posttranscriptional mechanisms strongly reducing Spry2 activity in some human HCC. Expression of c-Met (total and activated) and its downstream effectors (activated ERK and AKT) were ubiquitously high in HCCP (Fig. 1A,C). Next, to rule out the possibility that elevated ERK and AKT was due to Ras mutations in HCC, we determined the frequency of mutations in Ha-Ras, Ki-Ras, N-Ras, A-Raf, B-Raf, and Raf-1 genes by sequencing analysis. No mutations were detected in the members of the Ras cascade (not shown). These data indicate that suppression of Spry2 function and up-regulation of c-Met and its putative targets characterize tumor progression in human HCC in a context of wild-type Ras.

Figure 1.

Inactivation of Spry2 in human HCC samples with poor outcome. (A) Whole cell lysates were prepared from normal livers (NL), surrounding livers (SL), and HCC with better (HCCB) or poorer (HCCP) outcome and immunoblotted with indicated antibodies. Representative western blots are shown. β-Actin was used as loading control. (B) Spry2 mRNA levels were determined by quantitative reverse transcription PCR in normal livers (NL), nonneoplastic surrounding livers (SL), and HCCs with either better (HCCB) or poorer (HCCP) outcome. N-Target (NT) = 2 −ΔCt; ΔCt = RNR18-Ct target gene. Each bar represents mean ± SD. Tukey-Kramer Multiple Comparison Test: P < 0.0001 a, versus NL; b versus SL; c, versus HCCB. (C) Scheme representing the frequency of Spry2 promoter hypermethylation (Spry2 HYP), loss of heterozygosity (LOH) at Spry2 locus, Spry2 mRNA expression, as well as protein levels of Spry2 and its interactors as assessed by western blot analysis.

Multiple Mechanisms Mediate Spry2 Inactivation in HCC.

We investigated whether epigenetic events (promoter hypermethylation) or somatic alterations (point mutations or genomic deletions) were responsible for Spry2 down-regulation in HCC. Frequency of Spry2 promoter methylation was investigated using methylation-specific polymerase chain reaction (PCR). No hypermethylation at Spry2 promoter was found in normal and surrounding nontumorous liver samples (data not shown). Spry2 promoter hypermethylation was detected in both HCC prognostic subclasses, but at a significantly higher frequency in HCCP (13/44, 29.5%) than in HCCB (2/38, 5.3%; P < 0.003; Fig. 1C, Supporting Fig. 2A). No somatic mutations in the Spry2 gene were detected in the whole sample collection. The genomic status of Spry2 was further investigated via locus of heterozygosity (LOH) analysis. Again, LOH at Spry2 locus was more frequent in HCCP (18/44, 38.6%) than in HCCB (4/38, 10.5%; P < 0.002; Fig. 1C, Supporting Fig. 2B). Importantly, all HCCs exhibiting promoter hypermethylation or LOH of Spry2 gene showed down-regulation of Spry2 (Fig. 1C), indicating these molecular mechanisms as the causative events for Spry2 inactivation in a HCC subset.

The role of methylation on Spry2 expression was further studied in vitro. We screened nine HCC cell lines for Spry2 promoter methylation. The latter was detected in Alexander, SNU-387, and SK-Hep1 cell lines (Fig. 2A). Treatment with the demethylating agent Zebularine caused a dose-dependent up-regulation of Spry2 in Alexander and SNU-387 cells, but not in HepG2 cells, which have unmethylated Spry2 promoter (Fig. 2B). Up-regulation of Spry2 by Zebularine in Alexander and SNU-387 cells was paralleled by reversed methylation of the Spry2 promoter, as assessed by combined bisulfite restriction analysis (Fig. 2C).

Figure 2.

Regulation of Spry2 expression by promoter methylation and NEDD4-mediated proteosomal degradation. (A and B) Role of promoter methylation on Spry2 silencing in human HCC cell lines. (A) Summary of Spry2 promoter methylation status in human HCC cell lines (n = 9). (B) Treatment for 24 and 48 hours with the demethylating agent Zebularine restores Spry2 mRNA levels in a dose-dependent manner in Alexander and Snu-387 cells (with Spry2 methylated promoter), but not in HepG2 cells (with Spry2 unmethylated promoter). Levels of Spry2 were determined by quantitative real-time reverse transcription PCR. N-Target (NT) = 2-ΔCt; ΔCt = RNR18-Ct target gene. Each bar represents mean ± SD. Values represent the mean and SD from three independent, triplicate experiments. (C) Treatment of Alexander and Snu-387 cells with 100 or 200 μmol/L Zebularine for 48 hours reduces promoter methylation of Spry2, as assessed by combined bisulfite restriction analysis (COBRA). Primers amplifying the Spry2 promoter did not contain CpG dinucleotides so that the amplification step did not discriminate between templates according to their original methylation status. PCR products were then digested with 100 units of the methylation-sensitive restriction enzyme, BstUI. The presence of cleaved fragments attributable to BstUI digestion is markedly reduced after Zebularine treatment, implying a significant reduction of Spry2 promoter methylation in Alexander and Snu-387 cell lines. Abbreviations: C, untreated cells; ZEB, Zebularine-treated. (D) Silencing of NEDD4 in HuH7 cells via siRNA increases Spry2 levels and decreases ubiquitinated levels of Spry2. The position of full-length Spry2 protein in the poli-ubiquitinated immunoprecipitate is indicated by arrows. (E) Silencing of MNK2 via siRNA induces Spry2 expression in HuH7 cells. Experiments at 24 and 48 hours after siRNA transfection are shown.

Next, we investigated the mechanism responsible for down-regulation of Spry2 protein in HCC samples where Spry2 mRNA levels were elevated. Recent reports indicate that Spry2 can be proteolytically degraded by Casitas B-lineage lymphoma proto-oncogene (c-Cbl),23 seven in absentia homolog-2 (SIAH2),24 or neural precursor cell expressed developmentally down-regulated 4 (NEDD4).25 Thus, we determined the protein expression of c-Cbl, SIAH2, and NEDD4, and their relationship with Spry2 expression in human HCC (Fig. 1A,C; Supporting Fig. 1). C-Cbl was heterogeneously expressed in normal livers, HCCs, and corresponding nonneoplastic livers, with no correlation with Spry2 levels. Seven in absentia homolog-2 was similarly expressed in normal livers and nonneoplastic surrounding tissues, whereas it was frequently down-regulated in HCCs, with no differences between the two HCC subclasses. In contrast, NEDD4 was up-regulated only in a subset of HCCs (13/82, 15.8%). Of note, all HCCs exhibiting NEDD4 up-regulation displayed low Spry2 protein levels, suggesting a possible role of NEDD4 in Spry2 degradation. Accordingly, an increase of NEDD4-Spry2 complexes was detected in the same HCCs displaying high NEDD4 and low Spry2 levels (Fig. 1A,C; Supporting Fig. 1). The role of NEDD4 in Spry2 down-regulation function was further investigated in HuH7 cells with high NEDD4 expression and low Spry2 expression in the absence of Spry2 promoter hypermethylation. Indeed, suppression of NEDD4 via specific siRNA led to reactivation of Spry2 and decrease of Spry2 poly-ubiquitinated levels (Fig. 2D). Because Spry2 needs to be phosphorylated by MAPK-interacting kinase 2 (MNK2) before NEDD4-dependent degradation,25 we assessed the effect of MNK2 silencing in HuH7 cells. As expected, suppression of MNK2 via specific siRNA triggered Spry2 up-regulation (Fig. 2E). The current data indicate that multiple events concur to impair Spry2 function during human HCC pathogenesis.

Spry2 Modulates c-Met Signaling in Human HCC Cell Lines.

The importance of Spry2 in the control of c-Met–driven cell signaling and cell growth was investigated in human HCC cell lines. Among the latter, we chose the 7703 cell line for induction experiments, because it shows appreciable but not elevated levels of Spry2. Also, we used the HepG2 and Focus cell lines for transfection/overexpression experiments because of their very low levels of Spry2, whereas Hep3B and HuH6 cells, exhibiting a very high expression of Spry2, were selected for silencing experiments.

In 7703 cells, a rise in Spry2 protein expression was detectable as early as 10 minutes after HGF administration (not shown), peaking at 4 hours after treatment with a kinetic similar to that of phosphorylation (activation) of c-Met, AKT, and ERK (Fig. 3A). This observation suggests that up-regulation of Spry2 is a compensatory mechanism leading to modulation of HGF signals. In accordance with this hypothesis, induction of activated ERK and AKT proteins driven by HGF administration was inhibited when Spry2 was transfected into HepG2 and Focus cells, whereas Spry2 overexpression did not influence c-Met levels (Fig. 3B). Inhibition of c-Met–induced ERK and AKT signals resulted in a significant growth restraint of the two cell lines, increase in apoptosis, and decline of VEGF-α secretion in the medium (Supporting Fig. 3). Conversely, when treatment of Hep3B cells with HGF was associated with transfection of the Spry2 dominant negative form, Spry2Y55F, further amplification of activated ERK and AKT occurred (Fig. 3C), as well as an additional cell growth increase, decline in apoptosis, and rise of VEGF-α secretion (Supporting Fig. 4). Equivalent results were obtained with HuH6 cells (data not shown). Intriguingly, suppression of Spry2 in untreated Hep3B cells either via transfection of Spry2Y55F (Fig. 3D) or siRNA against Spry2 (not shown) did not trigger activation of c-Met, ERK, and AKT proteins, suggesting that loss of Spry2 expression alone does not sufficiently activate MAPK or AKT cascades. Equivalent results were obtained in HuH6 cells (data not shown). Altogether, our data indicate that Spry2 functions as a feedback inhibitor and regulates c-Met–induced cell growth via modulating ERK and AKT signaling cascade.

Figure 3.

Spry2 is feedback inhibitor of HGF/c-Met signaling in human HCC cell lines. (A) HGF stimulation induces expression of Spry2 in 7703 human HCC cell line. Cells were maintained as monolayer cultures in Dulbecco's modified essential medium supplemented with 10% fetal bovine serum, serum-starved for 24 hours, and then treated with HGF (50 ng/mL) for 1, 4, and 24 hours. Expression levels of Spry2, total and phosphorylated (activated) c-Met, AKT, and ERK1/2 were analyzed by western blotting. β-Actin was used as the loading control. (B) Overexpression of Spry2 inhibits HGF-induced ERK and AKT cascades without affecting activated c-Met levels in HepG2 human HCC cell line. SPRY2 complementary DNA was transiently transfected and cells were serum-starved for 24 hours and treated with HGF (50 ng/mL) for 1 and 4 hours. Activation of c-Met, ERK1/2 and AKT was analyzed by western blotting. β-Actin was used as loading control. (C) Transient transfection of a nonphosphorylated mutant form of Spry2 (Spry2Y55F) acting as dominant negative amplifies AKT and ERK signals in the Hep3B cell line, with no effect on c-Met levels on treatment with HGF. β-Actin was used as loading control. Experiments at 4 and 24 hours after HGF stimulation are shown. (D) Overexpression of Spry2Y55F does not trigger per se activation of c-Met, AKT, and ERK pathways in untreated Hep3B cell line. β-Actin was used as loading control. All cell lines experiments were repeated at least three times in triplicate.

Suppression of Spry2 and Overexpression of c-Met Cooperate to Promote Hepatocarcinogenesis in Ink4A/Arf−/− Mice.

Our clinical and in vitro data suggest that loss of Spry2 activity and activation of c-Met play a synergistic role during hepatocarcinogenesis. Thus, we developed a mouse model to examine whether the combination of these two genetic alterations promotes hepatocarcinogenesis in vivo. However, we envisaged the possibility that activation of c-Met and loss of Spry2 may not be able per se to induce liver tumor formation, because we and others have demonstrated that activation of the Ras/MAPK signaling alone is not sufficient for hepatocarcinogenesis using mouse models.18, 26 Therefore, we added another genetic alteration in our model, namely, the loss of the Ink4A/Arf locus, which is frequently disrupted in human HCC.27 Using an in vivo transfection method that combines hydrodynamic injection and sleeping beauty mediate somatic integration, we stably expressed c-Met (V5-tagged) or a dominant negative mutant form of Spry2, Spry2Y55F (HA-tagged) into the hepatocytes of Ink4A/Arf−/− mice and monitored for liver tumor development. Expression of Spry2Y55F alone did not induce histological abnormalities in the mouse liver (not shown), whereas overexpression of c-Met alone resulted in the formation of clear-cell foci of altered hepatocytes (FAH), proven to be preneoplastic in various rodent models of hepatocarcinogenesis (Supporting Fig. 5).28, 29 These lesions were often located in zone 3 of the liver acinus (Supporting Fig. 5B-E) and showed an excess in glycogen storage (Supporting Fig. 5D), resulting in enlargement and clear-cell phenotype of hepatocytes in hematoxylin-eosin (HE) staining (Supporting Fig. 5B,C). Moreover, these lesions were proliferating, as indicated by the expression of the proliferation-associated marker proliferating cell nuclear antigen (Supporting Fig. 5E) and the detection of occasional mitotic figures. However, no HCC or hepatocellular adenomas (HCA) were observed in these mice. In striking contrast, 54% (7/13) of the Ink4A/Arf−/− mice cotransfected with c-Met and Spry2Y55F developed numerous liver tumors between 14 and 20 weeks after injection (Fig. 4A-C). Tumors varied in size and histopathologic features, and were classified as HCA or HCC based on the criteria by Frith et al. (Fig. 4C,D).28 Tumors were characterized by the presence of a trabecular or pseudoglandular pattern. Small tumors usually exhibited a clear-cell phenotype (Fig. 4C), thus retaining the morphology of preneoplastic lesions developed in the model with exclusive overexpression of c-Met (Supporting Fig. 5). However, with increasing tumor size, particularly in large HCCs, some tumor cells lost their glycogen content and transformed into mitotically more active glycogen-poor, basophilic hepatocytes (Fig. 4D), recapitulating the usual sequence of morphological progression in the clear-cell type of rodent hepatocarcinogenesis.28, 29 Corresponding to their hepatocellular origin, tumors showed typical high RNA expression levels of α-fetoprotein (Fig. 4E). The two cotransfected genes, c-Met and Spry2Y55F, were detected in the tumors by immunohistochemistry and immunofluorescence with antibodies against their respective epitope tags (Fig. 5). Sporadic expression of the injected genes was observed also in the surrounding nontumor liver (Fig. 5A). Altogether, our observations indicate that coexpression of Spry2Y55F and c-Met promotes hepatocarcinogenesis in Ink4A/Arf−/− mice.

Figure 4.

Overexpression of c-Met and dominant-negative form of Spry2 (Spry2Y55F) cooperates with the loss of Ink4A/Arf to induce liver carcinogenesis in vivo. (A) Cumulative hazard curve, which represents the relative probability of tumor development in mice injected with c-Met, Spry2Y55F, or both. (B) Gross images of normal liver and liver with numerous tumor nodules (indicated with white arrows) from c-Met/Spry2Y55F injected Ink4A/Arf−/− mice. (C) Histology of liver tumors induced by c-Met/Spry2Y55F. Left, HE staining overview over the liver at low magnification, showing numerous tumors of different size, corresponding to several HCAs and HCCs; Right: HE staining of part of an HCC lesion of c-Met/Spry2Y55F tumors. (D) Periodic acid-Schiff staining of this HCC shows a subpopulation of PAS-negative basophilic (glycogen-poor) tumor cells—marked by arrowheads—that transformed from the surrounding PAS-positive glycogen-rich hepatocytes. These basophilic cells are generally more mitotically active (Inset with mitotic figure). (E) Real-time PCR analysis of AFP expression in normal wild-type liver (n = 5), Ink4A/Arf−/− liver (n = 5), and liver tumors (n = 5). Stars on the top represent P values comparing tumor samples versus wild-type liver (left) and tumor samples versus Ink4A/Arf−/− liver (right) using the Tukey-Kramer Multiple Comparisons Test. ***P < 0.001.

Figure 5.

Co-expression of c-Met and Spry2Y55F in mouse liver tumors. (A) Immunohistochemical staining of c-Met and Spry2Y55F in nontumorous (NT) and tumorous (T) liver; magnification, 200×. (B) Localization of c-Met (red) or Spry2Y55F (green) in liver tumors developed in c-Met/Spry2Y55F mice by immunofluorescence; magnification, 200×. Insets are close-ups of the images.

Next, we determined how cellular processes were affected during c-Met/Spry2Y55F-driven hepatocarcinogenesis. c-Met/Spry2Y55F liver tumors were characterized by an increase in proliferation, as shown by positive staining for the proliferation markers, proliferating cell nuclear antigen (Fig. 6A) and Ki67 (not shown). Accordingly, mRNA levels of cell cycle–positive regulators, cyclin B1, E1, and CDC20, were up-regulated in tumors (Fig. 6D). In HCC, apoptosis was also induced, as indicated by terminal deoxynucleotidyl transferase-mediated 2′-deoxyuridine 5′-triphosphate nick-end labeling staining (Fig. 6B). However, the mean apoptotic index was remarkably lower than the proliferation index in c-Met/Spry2Y55F tumors (4.4 ± 2.2 versus 20.8 ± 3.8, respectively; n = 18), indicating the prevalence of growth over death stimuli. Tumors samples were then assayed for angiogenesis by immunohistochemistry for the liver tumor endothelial marker PODXL1.30 Positive PODXL1 immunolabeling was detected only in neoplastic liver lesions from c-Met/Spry2Y55F mice, implying the presence of neovasculature in these lesions (Fig. 6C). In addition, c-Met/Spry2Y55F tumors displayed increased mRNA levels of angiogenic markers, angiogenin 1 and 2, and VEGF receptor-1 (Fig. 6D). In summary, the current data indicate that c-Met/Spry2Y55F coexpression promotes hepatocarcinogenesis by inducing cell proliferation and angiogenesis.

Figure 6.

Overview of cell proliferation, apoptosis, and angiogenesis in c-Met/Spry2Y55F tumors. (A) Proliferating cell nuclear antigen immunostaining showing strong proliferation in a HCA when compared with nonneoplastic surrounding tissue. (B) Apoptotic bodies (indicated by arrows) in a well-differentiated HCC. (C) Immunohistochemical staining of tumor vessel marker, PODXL1 in nontumorous (NT) and tumorous (T) liver. (D) Real-time PCR analysis of the expression levels of proliferation markers (cyclin B1, cyclin E1, and CDC20) and angiogenic factors (Ang1, Ang2, and VEGFR1) in normal wild-type liver (n = 5), Ink4A/Arf null liver (n = 5), and c-Met/Spry2Y55F liver tumors (n = 5). Stars on the top represent P values comparing tumor samples versus wild-type liver (left) and tumor samples versus Ink4A/Arf null liver (right) using the Tukey-Kramer Multiple Comparisons Test. ***P < 0.001; **P < 0.01; and *P < 0.05.

Up-regulation of MAPK and AKT Signaling in c-Met/Spry2Y55F Tumors.

Because both Spry2 and c-Met are important regulators of the Ras pathway, we investigated whether simultaneous overexpression of c-Met and Spry2Y55F results in up-regulation of Ras effectors, namely, the MAPK and AKT cascades, during hepatocarcinogenesis. Western blotting showed that preneoplastic lesions (not shown) and tumors from c-Met/Spry2Y55F mice exhibited high levels of activated ERK and AKT (Fig. 7). Activation of ERK and its downstream effector, ELK1, was elevated in tumors from c-Met/Spry2Y55F mice, lower in c-Met–injected livers, and absent in livers from Spry2Y555-injected and uninjected Ink4A/Arf−/− mice (Fig. 7). A similar pattern was found for AKT and its downstream effectors, including activated mammalian target of rapamycin (mTOR) (Fig. 7). Because tumor suppressor gene PTEN (phosphatase and tensin homolog) is the key regulator of AKT activity, we assessed the samples for total and phospho-PTEN levels. Western blotting showed a consistent expression of PTEN but an increased phospho-PTEN on c-Met overexpressing livers and tumor samples (Fig. 7). Because phospho-PTEN is known to be associated with inactivation of PTEN,31 this observation may represent an important mechanism of AKT signaling activation by c-Met in hepatocarcinogenesis. Levels of other regulators of the AKT pathway, including phospho-PDK-1 and phospho-Raf, were similar among the whole samples collection (not shown). Because other mitogenic cascades, including the FGF receptor and the epidermal growth factor receptor (EGFR) pathways, may induce activation of the ERK and AKT pathways, we determined the levels of activation of EGFR and FGF receptor in the mouse collection. Neither consistent induction of EGFR phosphorylation nor increase of fibroblast growth factor receptor substrate 2/growth-factor-receptor bound-2 (FRS2/GRB2) complexes (sign of FGF receptor activation) was detected in the mouse collection (Fig. 7). Similarly, no consistent up-regulation of other members of the EGFR family, including ERBB2 and ERBB3, occurred in c-Met/Spry2Y55F lesions, suggesting that c-Met might represent the major inducer of ERK and AKT in c-Met/Spry2Y55F mice (Fig. 7). In summary, our data indicate that overexpression of c-Met and Spry2Y55F leads to unconstrained activation of MAPK and AKT signaling during hepatocarcinogenesis.

Figure 7.

Activation of ERK and AKT signaling in c-Met/Spry2Y55F lesions. Western blot analysis of c-Met, activated c-Met, and its downstream effectors in uninjected Ink4A/Arf−/− livers, Spry2Y55-injected livers, c-Met-injected livers, and c-Met/Spry2Y55F–injected liver tumors. Three to five samples per each group were used for the analysis, and representative images are shown. β-Actin was used as loading control.

Loss of Spry2 Activity and Overexpression of c-Met Promote Hepatocarcinogenesis in Wild-Type Mice.

Finally, to rule out the possibility that hepatocarcinogenesis was induced by disruption of the Ink4A/Arf locus, mice with an intact Ink4A/Arf locus were hydrodynamically transfected with c-Met and Spry2Y55F genes. Our previous studies demonstrated that injection of either Spry2Y55F18 or c-Met22 alone into wild-type mice does not lead to tumor development. After co-expression of c-Met and Spry2Y55F into wild-type mice, liver tumors developed at similar frequency (46%, 6/13) as in the Ink4A/Arf null background. However, hepatocarcinogenesis required longer latency, and a lower number of tumors developed (Supporting Fig. 6). Morphologically, HCCs developed in wild-type mice were similar to the malignant lesions from c-Met/Spry2Y55F mice with a disrupted Ink4A/Arf locus (Supporting Fig. 6). Furthermore, the apoptotic index was equivalent in c-Met/Spry2Y55F tumors with intact or disrupted Ink4A/Arf locus (4.8 ± 2 versus 4.4 ± 2.2, respectively; n = 12 and 18, respectively), whereas the proliferation index was significantly higher in HCCs from mice in which the Ink4A/Arf locus was disrupted than in wild-type mice (20.8 ± 3.8 versus 15.2 ± 3; P = 0.0004). No significant differences were detected in the levels of activated ERK and AKT in c-Met/Spry2Y55F tumors with intact or disrupted Ink4A/Arf locus, indicating that the different proliferation rate in the two genetic backgrounds is independent of the latter cascades (Supporting Fig. 7). Interestingly, mRNA and protein levels of p16INK4A and p19ARF were higher in c-Met/Spry2Y55F–induced HCC samples when compared with normal livers from wild-type mice (Supporting Fig. 7), suggesting that up-regulation of p16INK4A and p19ARK may partly counteract cell proliferation in this tumor model. In accordance with this hypothesis, we found the highest expression of Cyclin D1/CDK4 complexes, whose levels are negatively regulated by p16INK4A and mark the G1-S progression of the cell cycle,27 in tumors from c-Met/Spry2Y55F with the Ink4A/Arf null alleles. Altogether, these data indicate that overexpression of c-Met and Spry2Y55 is sufficient to induce HCC development in mice, and that disruption of Ink4A/Arf locus accelerates this process by further increasing cell proliferation.


Activation of the Ras pathway is a key oncogenic event in carcinogenesis. In many tumor types, such as pancreatic cancer, colorectal cancer, and melanoma, oncogenic activating mutations of Ras or B-Raf are frequently observed. These mutations are rare in other tumor types, such as human HCC.2, 4-6 Previous studies showed that Ras pathway is up-regulated by multiple factors in cancer, including down-regulation of the Ras inhibitor RASSF1A2 or loss of the ERK inhibitor DUSP1 (dual specificity phosphatase 1).32 Here, we showed the concomitant activation of c-Met and the loss of Spry2 expression in a subset of human HCCs with aggressive clinical behavior. Furthermore, using in vitro assays and in vivo mouse models, we demonstrated that Spry2 is an important regulator of c-Met signaling, and loss of Spry2 propagates the activation of ERK and AKT signaling initiated by c-Met, leading to HCC development in vivo. Thus, the current study provides evidence that the coordinated deregulation of Spry2 and c-Met signaling is oncogenic in the liver and may represent an important mechanism for the uncontrolled activation of Ras pathway in the absence of Ras or B-Raf mutations.

Sprouty/Spred gene family members are feedback inhibitors of the receptor tyrosine kinase signaling.15, 16 Mounting evidence indicates that loss of Sprouty/Spred leads to abnormal activation of the Ras pathway. For example, genetic studies demonstrated that germline loss-of-function of Spred1 results in a neurofibromatosis 1–like syndrome, which is phenotypically similar to other genetic disorders caused by mutations of members of the Ras pathway.33 Loss of expression of Sprouty/Spred gene family members has been reported in multiple tumor types.17 Intriguingly, the tumor types showing frequent down-regulation of Spry expression (breast cancer, prostate cancer, and HCC) are all characterized by a relative low frequency of mutations of Ras or B-Raf genes. Conversely, our analysis of microarray studies performed in tumor types with frequent Ras or B-Raf mutations, including pancreatic and colorectal cancer, revealed no significant down-regulation of Spry/Spred genes (X. Chen and S.Y. Leung unpublished observation). Thus, this body of evidence suggests that down-regulation of members of the Spry family may be a key and alternative mechanism leading to the propagation of the Ras signaling in a context of wild-type Ras and B-Raf genes.

Previous studies in humans and rodents have envisaged the oncogenic role of c-Met and the oncosuppressor role of Spry2, respectively, in hepatocarcinogenesis.11, 12, 14, 18 However, the functional interaction between c-Met and Spry2 during tumorigenesis has never been examined in vivo. Here, we demonstrated that co-expression of c-Met and Spry2Y55F promotes hepatocarcinogenesis in mice, providing strong genetic evidence that deregulation of c-Met and Spry2 activation may have a pivotal role in HCC. Interestingly, overexpression of Spry2Y55F alone in mice does not induce neither alterations of liver morphology or activation of ERK and AKT cascades. These findings indicate that other genetic or epigenetic alterations are required for HCC development in addition to the loss of Spry2. However, hepatic preneoplastic lesions developed after overexpression of c-Met alone. Similar to our data, c-Met overexpression in FVB/N mouse liver resulted in the appearance of dysplastic, but not neoplastic, lesions.22 In many rodent models, hepatocarcinogenesis is defined by the emergence of glycogen-rich preneoplastic lesions, followed by progression through mixed-cell to predominantly glycogen-poor (basophilic) cell foci.28, 29 In accordance with these models, our current findings suggest that c-Met over-expression is sufficient for the appearance of glycogen-rich preneoplastic lesions in the mouse liver, whereas Spry2 disruption by Spry2Y55F is necessary for full malignant transformation of the liver.

Our mouse model demonstrated that coexpression of Spry2Y55F and c-Met leads to activation of both ERK and AKT/mTOR pathways, a signature shared by human HCCs with aggressive phenotype. Although the role of the MAPK pathway has been clearly demonstrated in HCC pathogenesis, the critical functions of AKT/mTOR pathway have only been recently elucidated.34 Clinical studies with mTOR inhibitors, such as RAD001, are currently in progress, with promising preliminary outcomes for HCC treatment.35 However, it seems unlikely that inhibition of AKT/mTOR pathway alone is sufficient to inhibit HCC growth. Because of the concomitant activation of ERK and AKT/mTOR pathways in a human HCC subset, it seems likely that better clinical outcomes can be achieved via combinatory inhibition of ERK and AKT/mTOR pathways. Indeed, recent studies with HCC cell lines suggest an enhanced antitumor activity when combining Sorafenib (a Raf inhibitor) with Rapamycin (mTOR inhibitor).36 However, the efficacy of such combinatorial treatment needs to be further validated in preclinical settings, especially mouse models with genetic changes that resemble human HCC pathogenesis. Liver tumors induced by Spry2Y55F and c-Met overexpression, which are characterized by elevated ERK and AKT/mTOR activity, may represent a promising preclinical model to determine the potential synergistic effect of suppressing concomitantly the Ras/MAPK and AKT/mTOR pathways in HCC treatment.


We thank Sandra Huling of the UCSF Liver Center Morphology Core for histology support and Dr. Eisuke Nishida of Kyoto University for Spry2Y55F/pCS2 construct.