Neuregulin/erythroblastic leukemia viral oncogene homolog 3 autocrine loop contributes to invasion and early recurrence of human hepatoma

Authors


  • Potential conflict of interest: Nothing to report.

Abstract

Intrahepatic metastasis is the primary cause of the high recurrence and poor prognosis of human hepatocellular carcinoma (HCC). However, neither its molecular mechanisms nor markers for its prediction before hepatectomy have been identified. We recently revealed up-regulation of erythroblastic leukemia viral oncogene homolog 3 (ERBB3) in human HCC. Here we examined the clinical and biological significance of ERBB3 in HCC. Up-regulation of ERBB3 in HCC was strongly associated with male gender (P< 0.001), chronic hepatitis B (P = 0.002), microscopic vascular invasion (P = 0.034), early recurrence (P = 0.003), and worse prognosis (P = 0.004). Phosphorylated ERBB3 and its ligands [neuregulins (NRGs)] were detected in both HCC tissues and cells. Phosphorylation of ERBB3 could be induced by conditioned media of HCC cells and abolished by the pretreatment of conditioned media with anti-NRG antibodies or by the silencing of the endogenous NRG expression of the donor HCC cells. Human epidermal growth factor receptor 2 was required for ERBB3 phosphorylation. The downstream phosphoinositide 3-kinase/v-akt murine thymoma viral oncogene homolog pathways were primarily elicited by NRG1/ERBB3 signaling, whereas the mitogen-activated protein kinase/extracellular signal-regulated kinase pathways were elicited by both epidermal growth factor/epidermal growth factor receptor and NRG1/ERBB3 signaling. The activation and silencing of ERBB3-dependent signaling had potent effects on both the migration and invasion of HCC cells, but neither had significant effects on the proliferation of HCC cells, tumor formation, or tumor growth in vitro and in vivo. Conclusion: The constitutive activation of ERBB3-dependent signaling via the NRG1/ERBB3 autocrine loop plays a crucial role in the regulation of cell motility and invasion, which contribute to intrahepatic metastasis and early recurrence of HCC. ERBB3 is a marker for the prediction of intrahepatic metastasis and early recurrence. ERBB3-dependent signaling is a candidate target for the treatment of microscopic vascular invasion and for the prevention of HCC recurrence. (HEPATOLOGY 2011;53:504-516) (This article firstt published online on January 18, 2011; the title has since changed; the correct version appears in print.

Frequent intrahepatic metastasis is a unique feature of hepatocellular carcinoma (HCC) and the primary cause of high rates of early recurrence after initial curative therapy. To improve the clinical outcomes of patients with HCC, there is a pressing need to elucidate the molecular mechanisms of intrahepatic metastasis and to identify markers for the detection of intrahepatic invasion and the prediction of early recurrence.

The human epidermal growth factor receptor (EGFR)/erythroblastic leukemia viral oncogene homolog (ERBB) family consists of four membrane-associated receptor tyrosine kinases: EGFR, human epidermal growth factor receptor 2 (HER2)/ERBB2, HER3/ERBB3, and HER4/ERBB4. Upon ligand binding to the extracellular domains, they form homodimers and heterodimers to one another, and this results in autophosphorylation or transphosphorylation and the initiation of downstream intracellular signaling cascades regulating cell proliferation, motility, and differentiation.1 HER2 does not bind any ligand and requires another ligand-bound EGFR/ERBB member for dimerization. ERBB3 has impaired kinase activity and also needs another EGFR/ERBB family member for dimerization to elicit downstream signals.2 Deregulation of EGFR/ERBB signaling is observed in most human cancers, and a wealth of data directly implicates EGFR/ERBB signals in cancer pathogenesis. Indeed, both EGFR and HER2 are validated targets for anticancer therapy.3 Recent studies have further disclosed the pivotal role of ERBB3 in oncogenic EGFR/ERBB signaling.4, 5 For example, both primary and acquired resistance to anti–tyrosine kinase therapies for lung cancers is attributable to persistent activation of ERBB3 signaling via either hepatocyte growth factor receptor c-MET gene amplification or v-akt murine thymoma viral oncogene homolog (Akt)–driven feedback signaling.6 Oncogenic HER2 signaling in breast cancer with HER2 up-regulation is dependent on ERBB3 activation,7 and resistance to HER2-targeted therapies results from escaped ERBB3 signaling.8-10 However, the roles of EGFR/ERBB signaling and ERBB3 in human HCC have rarely been addressed.

Recently, we reported the up-regulation of ERBB3 in human HCC.11 Interestingly, ERBB3 plays important roles in liver development.12 Here we found that the up-regulation of ERBB3 in HCC was associated with aberrant activation of ERBB3 signaling, microscopic vascular invasion of HCC, early recurrence, and poor clinical outcomes. The constitutive activation of ERBB3 in hepatoma cells was mediated by a neuregulin 1b (NRG1)/ERBB3 autocrine loop that initiated the downstream phosphoinositide 3-kinase (PI3K)/Akt and mitogen-activated protein kinase (MAPK)/extracellular signal-regulated kinase (Erk) pathways to regulate cell motility and invasion activity rather than tumor formation and growth. Our findings suggest that ERBB3 plays a crucial role in the regulation of HCC invasion and metastasis rather than tumor initiation and growth. ERBB3-dependent pathways are candidate targets for the prevention and treatment of intrahepatic and extrahepatic metastases of HCC.

Abbreviations: Akt, v-akt murine thymoma viral oncogene homolog; CM, conditioned media; DMEM, Dulbecco's modified Eagle's medium; EGF, epidermal growth factor; EGFR, epidermal growth factor receptor; ERBB, erythroblastic leukemia viral oncogene homolog; ERK, extracellular signal-regulated kinase; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; HER, human epidermal growth factor receptor; JNK, c-Jun N-terminal kinase; L, largest superficial diameter; MAPK, mitogen-activated protein kinase; NRG, neuregulin; pAkt, phosphorylated v-akt murine thymoma viral oncogene homolog; pERBB, phosphorylated erythroblastic leukemia viral oncogene homolog; PI3K, phosphoinositide 3-kinase; sERBB3, secreted erythroblastic leukemia viral oncogene homolog 3; shRNA, short hairpin RNA; siRNA, small interfering RNA; W, smallest superficial diameter; XTT, sodium 3′-[1-(phenylaminocarbonyl)-3,4-tetrazolium]-bis(4-methoxy-6-nitro)benzene sulfonic acid hydrate.

Materials and Methods

Tissues and Patients.

The Internal Review Board for Medical Ethics of Chang Gung Memorial Hospital approved the specimen collection procedures, and informed consent was obtained from each subject or subject's family. Tumors and adjacent nontumor liver tissues were collected from 71 patients who underwent hepatectomy for HCC from 1999 to 2000 at our hospital. The diagnosis of HCC was made on the basis of histopathological criteria and clinical features. Follow-up continued until August 2008. The mean follow-up duration was 41 months (range = 2-106 months). Disease-free survival and overall survival were measured from the date of hepatectomy until tumor recurrence and death, respectively. Normal liver tissue was obtained from a patient with focal nodule hyperplasia who had undergone tumor resection.

Hepatoma Cell Lines, Reagents, Small Interfering RNAs (siRNAs), Short Hairpin RNAs (shRNAs), and Antibodies.

The sources of the human hepatoma cell lines (Tong, HepG2, Hep3B, Huh7, Sk-Hep1, and Mahlavu) have been described previously.13 Recombinant human NRG1 (R&D Systems, Inc.) and human epidermal growth factor (EGF; Gibco, Invitrogen) were purchased. Predesigned siRNAs targeting EGFR, HER2, ERBB3, and NRG1 plus siRNAs with scrambled sequences were also purchased (Applied Biosystems Business, Austin, TX). Lentivirus-based shRNA clones (TRC-Hs 1.0) were obtained from the National RNAi Core Facility (Academia Sinica, Taiwan). Antibodies against ERBB3, EGFR, HER2, NRG1, β-actin, ERBB4, and phosphorylated erythroblastic leukemia viral oncogene homolog 3 (pERBB3; Tyr1289) were commercially obtained (Abcam, Plc., and Cell Signaling, Inc.). Antibodies against Akt, phosphorylated v-akt murine thymoma viral oncogene homolog (pAkt; Ser473), p38, phosphorylated p38 (Thr180/Tyr182), Erk, phosphorylated Erk (Thr202/Tyr204), c-Jun N-terminal kinase (JNK), and phosphorylated JNK (Thr183/Tyr185) were purchased (Cell Signaling Technology, Inc.).

Induction or Blockade of ERBB3 Phosphorylation.

To demonstrate the induction of ERBB3 phosphorylation, Huh7 and HepG2 cells, which secreted relatively low amounts of bioactive NRG1 to activate their own ERBB3, were maintained in serum-free media for 24 hours, and this was followed by treatment with 1 or 10 ng/mL recombinant NRG1 or EGF for 15 to 30 minutes, respectively, or with conditioned media for 60 minutes. For the collection of conditioned media, 2 ×105 HCC cells (usually SK-Hep1 cells, whose conditioned medium contained the most activity for ERBB3 activation) were seeded with regular cultured media in 36-mm dishes overnight; after that, the media were removed from the dishes, washed three times with phosphate-buffered saline, and further cultured in serum-free media for 24 hours. Media were spun for the removal of any insoluble components for 15 minutes at 12,000g and then used to treat cells. For blockade assays, conditioned media were incubated with antibodies against NRGs (200 or 400 ng/mL) for 20 minutes at 37°C to neutralize the biological activity of NRGs and then were used to treat HCC cells.

Gene Silencing via siRNAs: Silencing Efficiency, Specificity, and Off-Target Effects.

To knock down the expression of EGFR, HER2, ERBB3, or NRG1, cells were transfected with siRNAs or transduced with lentivirus-based shRNAs targeting EGFR, HER2, ERBB3, or NRG1. siRNAs with randomly scrambled sequences were used as the controls. To guarantee the specificity and to avoid off-target effects, we used two clones of siRNAs or shRNAs for each gene and separately examined the silencing efficiency with respect to their target genes and their effects on the related biological results. For example, we used two clones of siRNA targeting HER2. Silencing of HER2 expression via both siRNA clones efficiently suppressed the phosphorylation of ERBB3 and its downstream Akt (Supporting Information Fig. 1A). Also, two clones of siRNAs targeting ERBB3 were used, and the specificity for the silencing of ERBB3 expression and for the suppression of phosphorylation of downstream Akt was examined (Supporting Information Fig. 1B). In addition, consistent effects of both siRNA clones targeting ERBB3 and both siRNA clones targeting HER2 on cell proliferation (Supporting Information Fig. 1C) and tumor sphere formation for HepG2, Huh7, and SK-Hep1 cells were observed (data not shown). Basically, 2 × 105 cells were seeded onto six-well plates and transfected with 5 nM siRNA with Lipofectamine as the transfectant reagent according to the manufacturer's protocols (Lipofectamine RNAiMAX, Invitrogen). Forty-eight hours after transfection, the cells were harvested or subjected to further assays.

Figure 1.

Up-regulation of ERBB3 in HCC. (A) Immunoblotting of ERBB3 expression in HCC cells. Lane 1 represents the results derived from normal liver tissue used as the control; lanes 2 to 7 represent the results derived from HCC cells (Huh7, HepG2, Mahlavu, SK-Hep1, Tong, and Hep3B, respectively). (B) Representative immunoblotting of ERBB3 in eight pairs of HCC (T) and matched nontumor liver (N) tissues (top). The detection of β-actin was used as the loading control for each sample (bottom). (C,D) Immunohistochemistry of ERBB3 expression in normal liver and HCC tissues, respectively. The left panels are shown at ×100 magnification; their close-ups in the right panels are shown at ×400 magnification. ERBB3 was barely detectable in normal liver tissues but was highly expressed in HCC tissues.

Quantitative Real-Time Polymerase Chain Reaction, Immunoblotting Analysis, and Immunohistochemical Staining.

RNA extraction, reverse transcription, and quantitative real-time polymerase chain reaction were performed as previously described. Immunoblotting analysis and immunohistochemistry assays were conducted as previously described13, 14 (see the Supporting Information).

Migration, Invasion, and Cell Proliferation Assays.

The invading activities of HCC cells were analyzed with Boyden chambers (8-μm pore size; Corning, Inc.), cell motility was assayed with wound healing assays, and cell proliferation was determined via colorimetric sodium 3′-[1-(phenylaminocarbonyl)-3,4-tetrazolium]-bis(4-methoxy-6-nitro)benzene sulfonic acid hydrate (XTT) assays (see the Supporting Information).

Tumor Formation Assays.

The care of animals and the use and treatment of mice were in strict compliance with the guidelines of the Animal Care and Use Committee of Chang Gung Memorial Hospital (Taiwan) and Guide for the Care and Use of Laboratory Animals (National Academy of Sciences, 1985). Equal numbers (5 × 106/0.2 mL of phosphate-buffered saline) of Huh7 or SK-Hep1 cells transduced with lentivirus vectors bearing shRNAs targeting either the ERBB3 or luciferase gene were injected subcutaneously into the dorsal flanks of athymic nude mice (6- to 8-week-old BALB/c-nu mice), and tumor growth was observed for up to 8 weeks after inoculation. Tumor growth was followed every week with electronic caliper measurements. Each tumor volume was calculated with the following formula:

equation image

where W is the smallest superficial diameter and L is the largest superficial diameter. The Student t test was used to analyze the in vivo growth parameter. Each group contained six mice, and the experiments were conducted at least twice for each HCC cell line.

Statistical Analysis.

The χ2 test or Student t test were used for comparisons between variables. Kaplan-Meier analysis and the log-rank test were used to illustrate differences between each potential risk factor in probabilities of recurrence-free and overall survival after patients underwent primary curative hepatectomy. In our analysis of the probability that patients would remain free of hepatoma recurrence, we defined recurrence as the first event in treatment failure; data for all other patients were censored at the date of the last follow-up visit, death from causes other than hepatoma, and any subsequent recurrence of hepatoma. Data for patients were analyzed from the date of surgery to the time of the first event or to the date on which data were censored (according to the Kaplan-Meier method), and the curves were compared with the log-rank test.

Results

Up-Regulation of ERBB3 in Human HCC.

To examine the expression of ERBB3 in human HCC, we assayed the relative messenger RNA levels of ERBB3 in 2 normal liver tissues and 71 pairs of HCC and matched para-HCC liver tissues by quantitative real-time polymerase chain reaction. In comparison with the expression levels of the corresponding nontumor liver tissues, up-regulation of ERBB3 in HCC (2-fold or higher) was found in 50 cases (70.4%; see Supporting Information Table 1). Moreover, ERBB3 proteins were detected in all six HCC cell lines (Fig. 1A) and most of the HCC tissues (Fig. 1B). In contrast, ERBB3 proteins were barely detectable in normal liver tissues (Fig. 1A,B). Up-regulation of ERBB3 in HCC was further confirmed in liver tissue sections by immunohistochemistry (Fig. 1C,D).

Table 1. Correlation of ErbB3 Up-Regulation to Clinical Presentation
FactorErbB3 Up-Regulation*P ValueMethod
No (n = 21)Yes (n = 50)
  • Abbreviation: HBV, hepatitis B virus.

  • *

    Twofold or greater up-regulation of ErbB3 messenger RNA in HCC tissues versus the corresponding nontumor liver tissues.

  • P< 0.05 was regarded as statistically significant.

  • Microscopically positive for tumor invasion into local blood vessels or bile ducts.

  • §

    Clinical evidence of recurrence within 24 months after the initial tumor resection.

SexMale7420.000Two-sided Fisher's exact test
Female148
Age<55 years4220.061
≧55 years1728
CauseHBV8370.002
Non-HBV1310
CirrhosisNo4210.101
Yes1729
Alpha-fetoprotein<400 ng/mL16280.046
≧400 ng/mL113
Histology grade1 + 212230.443
3 + 4927
Tumor size<5 cm8250.443
≧5 cm1325
Microscopic vascular invasionNo15160.034
Yes633
Tumor stageT111190.040Kruskal-Wallis
T242
T309
T4720
Recurrence   0.000Log rank
Early recurrence§   0.003
Overall survival   0.004

Up-Regulation of ERBB3 Was Associated With Intrahepatic Microscopic Invasion, Early Recurrence, and Poor Prognosis.

To clarify the clinical significance of ERBB3 up-regulation, we correlated the expression of ERBB3 to clinical presentations in 71 patients with HCC (Table 1). Up-regulation of ERBB3 was strongly associated with male gender (P< 0.001), chronic hepatitis B (P = 0.002), higher serum alpha-fetoprotein levels (P = 0.046), higher tumor recurrence rates (P< 0.001, log-rank test), and lower overall survival (P = 0.004, log-rank test). The association of ERBB3 up-regulation with higher tumor recurrence and lower overall survival was further demonstrated via Kaplan-Meier analyses (Fig. 2A,B).

Figure 2.

Kaplan-Meier analysis of the probabilities of recurrence-free survival and overall survival for patients. The analyses were performed according to the up-regulation of ERBB3. (A) Probability of HCC recurrence after primary tumor resection. (B) Probability of overall survival. (C) Probability of HCC recurrence within 24 months after the initial tumor resection. The P values were calculated with the log-rank test.

Because intrahepatic metastasis contributes to the unusually high frequency of early recurrence of HCC and its poor prognosis, we further examined the correlation of ERBB3 up-regulation with microscopic vascular invasion and early recurrence of HCC (within 24 months after the initial curative therapy). Up-regulation of ERBB3 was strongly associated with microscopic vascular invasion of HCC (P = 0.034; Table 1) and early recurrence (P = 003; Fig. 2C).

Constitutive Activation of ERBB3 via an NRG1/ERBB3 Autocrine Mechanism.

We next asked whether up-regulation of ERBB3 is associated with constitutive activation of ERBB3. We assayed the coexpression of other ERBB members and NRGs, the ligands of ERBB3. EGFR and HER2 were expressed in most of the HCC cells, whereas ERBB3 and NRG1 were expressed in all of the HCC cells (Fig. 3A). ERBB4 was not detected in any of the HCC cells (Fig. 3A). In addition, both NRG1 and pERBB3 were also detected in all of the tested HCC tissues (Fig. 3B), and this suggested constitutive activation of ERBB3 in HCC, very likely via an NRG1/ERBB3 autocrine mechanism.

Figure 3.

Expression of the ERBB family proteins and NRG1 in HCC cells and tissues. (A) Immunoblotting of the expression of EGFR, HER2, ERBB3, ERBB4, and NRG1 in HCC cells. MCF7 (a breast cancer cell line) and HeLa (a cervical cancer cell line) were used as positive controls for HER2 and EGFR, respectively. (B) Immunoblotting of the expression of ERBB3, pERBB3, and NRG1 in HCC tissues. Lane 9 represents the results derived from normal liver tissue used as the control.

To confirm the involvement of an NRG1/ERBB3 autocrine loop in ERBB3 activation in HCC, we determined whether HCC cells secrete bioactive NRG1 to activate ERBB3 of HCC cells. Phosphorylation of ERBB3 of starved Huh7 and HepG2 cells was induced (presumably activated) by treatment with the recombinant NRG1 (Fig. 4A) or the conditioned media of most of the HCC cells (Fig. 4B). To verify the presence of bioactive NRG1 in the conditioned media, we used antibodies against NRG1 to block its interaction with ERBB3. As shown in Fig. 4C, phosphorylation of ERBB3 was abolished because the conditioned media had been treated with antibodies against NRG1 before its administration to HCC cells. Pretreatment of the conditioned media with antibodies against the extracellular domain of ERBB3 was used as the positive control (Fig. 4C). In parallel, HeLa cells, which did not express ERBB3, were treated with the conditioned media of HCC cells to rule out the possibility of contaminants of pERBB3 in the conditioned media due to lysis of the donor HCC cells (Fig. 4D). Treatment of HeLa cells with recombinant NRG1 was used as the control.

Figure 4.

Activation of ERBB3 via an autocrine mechanism in HCC cells. (A) Induction of ERBB3 phosphorylation by treatment with recombinant NRG1b. (B) CM collected from Tong (T), HepG2 (G2), Hep3B (3B), Huh7 (H7), Mahlavu (ML), SK-Hep1 (SK), human fibroblast (F), and HeLa cells were used to treat G2 and H7 cells and examine the phosphorylation of ERBB3 of H7 and G2 cells. CM of G2 and H7 cells had relatively low activity for phosphorylating ERBB3. (C) The neutralizing ERBB3 phosphorylation activity of the CM of SK cells by pretreatment with antibodies against NRG1 (anti-NRG) or ERBB3 (anti-ERBB3) was demonstrated. A dose-dependent suppression effect on the phosphorylation of ERBB3 of H7 and G2 cells was noted for both anti-NRG and anti-ERBB33. The total amount of ERBB3 in each experiment is shown as the loading control. (D) CM collected from SK cells were used to treat G2, H7, T, and HeLa cells. HeLa cells, which did not express ERBB3, were used as a negative control. (E) Silencing of NRG1 expression in SK and T cells suppressed the phosphorylation of their own ERBB3. siRNAs with scrambled sequences (siNS) or targeting endogenous NRG1 (siNRG) were used to transfect SK and T cells. Recombinant NRG1 was added to restore the phosphorylation of ERBB3. (F) The phosphorylation of ERBB3 of H7 and G2 cells by the CM of SK cells was abolished via the silencing of the endogenous NRG1 expression of SK cells. Abbreviation: CM, conditioned media.

If ERBB3 of the HCC cells was activated via an autocrine loop, we expected that silencing of the expression of endogenous NRG1 would suppress phosphorylation of their own ERBB3 and abolish the bioactivity of the conditioned media to phosphorylate ERBB3 of the target cells. As shown in Fig. 4E, silencing of the expression of endogenous NRG1 by RNA interference in HCC cells suppressed their own ERBB3 phosphorylation (Fig. 4E) and eliminated the activity of their conditioned media to phosphorylate ERBB3 of the target cells (Fig. 4F). Altogether, we conclude that the constitutive activation of ERBB3 in HCC cells was achieved via an autocrine mechanism by the synthesis/secretion of bioactive NRG1 from HCC cells to activate their own ERBB3.

HER2 Was Required for ERBB3 Phosphorylation/Activation.

To identify the partners for the dimerization and activation of ERBB3 upon NRG1 binding, we investigated whether EGFR or HER2 was required for ERBB3 activation in SK-Hep1, Huh7, and HepG2 cells. Silencing the expression of HER2 efficiently suppressed ERBB3 phosphorylation, whereas silencing EGFR only minimally affected ERBB3 phosphorylation of HCC cells (Fig. 5A). In addition, simultaneously silencing EGFR and HER2 expression had only minimal synergistic effects on ERBB3 phosphorylation. These findings suggest that the dimerization and activation of ERBB3-dependent signaling in HCC cells are primarily dependent on HER2.

Figure 5.

ERBB3-dependent signaling. (A) Roles of EGFR and HER2 in ERBB3 phosphorylation. After transfection with siRNAs targeting HER2, EGFR, or both, cells were treated with NRG1 for 15 minutes and then harvested for the detection of pERBB3. Lanes 1 to 4 represent the results derived from the HCC cells transfected with siRNAs containing scrambled sequences targeting HER2, EGFR, and both HER2 and EGFR. The detection of β-actin was used as the loading control. (B,C) Differential roles of EGR/EGFR and NRG1/ERBB3 in the activation of the Akt and MAPK pathways. (B) After starvation for 24 hours, Huh7 cells were treated with NRG1 or EGF for 15 minutes, and then the activation of Akt, p38, Erk1/2, and JNK was assayed. (C) To further confirm the differential roles of EGF and NRG1 in the activation of Akt, the expression of EGFR, HER2, and ERBB3 was silenced, and then the cells were treated with either NRG1 (N) or EGF (E). (-) indicates no treatment; siNR indicates transfection with siRNA containing scrambled sequences; and siB1, siB2, and siB3 indicate transfection with siRNAs targeting EGFR, HER2, and ERBB3, respectively.

Crucial Role of NRG1/ERBB3 in the Activation of the PI3K/Akt Pathways.

We then examined whether EGF/EGFR signaling and NRG1/ERBB3 signaling play redundant or different roles in the transmission of transmembrane oncogenic signals in HCC cells. As shown in Fig. 5B, the induction of phosphorylation of Akt and JNK was observed when HCC cells had been treated with NRG1 to activate ERBB3 but not when they had been treated with EGF to activate EGFR. The induction of Erk1/2 phosphorylation was observed when HCC cells had been treated with EGF as well as NRG1. On the other hand, the phosphorylation of p38 was not changed by treatment with either NRG1 or EGFR. Because the PI3K/Akt pathways are generally regarded as key to oncogenic signaling, we further examined the differential roles of NRG1/ERBB3 and EGF/EGFR in the activation of Akt in Huh7 cells (Fig. 5C). Again, Akt phosphorylation was primarily induced by the treatment of HCC cells with NRG1 but not EGFR. In addition, silencing of the expression of HER2 or ERBB3 (but not EGFR) suppressed Akt phosphorylation by NRG1 (Fig. 5C). Apparently, EGF/EGFR and NRG1/HER2/ERBB3 play different roles in transmembrane cellular signals. NRG1/HER2/ERBB3 rather than EGF/EGFR plays a pivotal role in the activation of the PI3K/Akt pathways in HCC cells.

The finding of differential roles of EGFR- and HER2/ERBB3-dependent signaling in eliciting downstream pathways was further validated by the observation that the proliferation and viability of HCC cells were much more sensitive to lapatinib, an EGFR- and HER2-specific inhibitor, than to gefitinib, an EGFR-specific inhibitor. The median inhibitory concentrations of lapatinib (17-50 nM) for the six HCC cell lines were much lower than those of gefitinib (29 to >150 μM; Supporting Information Fig. 2).

ERBB3-Dependent Signaling Played the Crucial Role in the Migration and Invasion of HCC Cells.

Because the up-regulation of ERBB3 was strongly associated with microscopic vascular invasion and early recurrence of HCC (Fig. 2C and Table 1), we speculated that ERBB3-dependent signaling regulates tumor cell motility and invasion. We used wound migration and Transwell invasion assays to examine this hypothesis. Activation of ERBB3 signaling by treatment with recombinant NRG1 significantly enhanced the motility and invasion activity in SK-Hep1, Huh7, and HepG2 cells in a dose-dependent manner (Fig. 6A,B and Supporting Information Fig. 3). On the other hand, the silencing of ERBB3, HER2, or both ERBB3 and HER2 expression efficiently suppressed the invasion activity of HCC cells (Fig. 6C,D). These findings were consistent with those found clinically: (1) up-regulation of ERBB3 was strongly associated with both microscopic vascular invasion and intrahepatic invasion (Table 1 and Supporting Information Table 1), and (2) among the four patients with early spreading of HCC to the lungs and/or brain (<8 M), there was up-regulation of ERBB3 in the tumors (Supporting Information Table 1). ERBB3-dependent signaling apparently plays the crucial role in regulating the motility and invasion of HCC cells.

Figure 6.

Effects of ERBB3-dependent signaling on tumor cell invasion. (A,B) Transwell (Boyden chamber) assays showed the invasion activity of Huh7 and SK-Hep1 (SK) cells in the presence of 0, 1, or 10 ng/mL recombinant NRG1 in the culture medium (DMEM) of the lower chambers. (C,D) Silencing of HER2 (siHER2), ERBB3 (siERBB3), or both HER2 and ERBB3 (siHER2+siERBB3) significantly suppressed the invasion activity of Huh7 and SK cells. In comparison with the control sets, P values less than 0.001 (t test) were regarded as statistically significant (indicated by asterisks). The results were based on the averages of six high-power fields of two experiments, and the Student t test was used to evaluate the statistical significance. siNS indicates siRNAs with scrambled sequences. Abbreviation: DMEM, Dulbecco's modified Eagle's medium.

ERBB3-Dependent Signaling did not Significantly Affect the Proliferation, Tumorigeicity of HCC Cells.

To examine the roles of EBBB3-dependent signaling in the tumor development and growth of HCC, we examined the biological consequences of aberrant activation or suppression of ERBB3-dependent signaling in hepatocellular carcinogenesis. Although the activation of ERBB3 by treatment with recombinant NRG1 modestly enhanced cell proliferation (Fig. 7A,B), the silencing of ERBB3, HER2, or both did not suppress the proliferation of HCC cells (Fig. 7C,D).

Figure 7.

Biological effects of ERBB3-dependent signaling on the proliferation and tumor growth of HCC cells. (A,B) XTT assays revealed the effects of the activation of ERBB3 signaling on the proliferation of HepG2 and Huh7 cells, respectively, by treatment with recombinant NRG1 (0, 1, or 5 ng/mL). (C,D) XTT assays showed the effects of the silencing of the expression of HER2 (siB2) or ERBB3 (siB3) on the proliferation of HepG2 and Huh7 cells. The results were generated via experiments conducted twice in triplicate. siB3#1 and siB3#2 indicate two different clones of siRNA targeting ERBB3.siNS indicates siRNAs with scrambled sequences. (E,F) Xenograft assays for tumorigenicity in vivo. SK-Hep1 and Huh7 cells were transduced with shRNA targeting ERBB3 (siB3) or shRNA targeting the luciferase gene (control). At each site, 5 × 106 cells were injected subcutaneously into nude mice. Each group contained six mice. The experiments were performed at least twice. Tumor volumes were measured weekly for up to 5 or 7 weeks after injection. The averages with standard deviations are shown.

We then examined the biological effects of ERBB3-dependent signaling on the tumor formation and growth of HCC. Neither the activation of ERBB3-dependent signaling via treatment with recombinant NRG1 nor the silencing of the expression of ERBB3, HER2, or both in Huh7 and SK-Hep1 cells significantly suppressed tumor sphere formation in soft agar assays (data not shown). The silencing of EGFR expression did not significantly inhibit tumor sphere formation either (data not shown). On the other hand, xenograft tumor growth of Huh7 and SK-Hep1 cells in nude mice was not significantly suppressed whether the expression of ERBB3 or HER2 had been silenced or not via transduction with lentivirus-based RNA interference (Fig. 7E,F). The silencing of ERBB3 expression in the xenografts was confirmed with immunoblotting assays (Supporting Information Fig. 4). ERBB3-dependent pathways apparently did not have significant effects on the development and growth of HCC in in vitro or in vivo assays.

Discussion

Microscopic vascular invasion of HCC cells is frequently found in surgically removed HCC tumors. Such microscopic vascular invasion is strongly associated with high rates of intrahepatic metastasis and early recurrence of HCC.15-17 However, the mechanisms behind the high frequency of microscopic vascular invasion have not been well elucidated. In this study, we provide the first direct evidence that ERBB3-dependent signaling contributes to microscopic vascular invasion and intrahepatic metastasis of HCC. First, we found that up-regulation of ERBB3 in HCC was strongly associated with microscopic vascular invasion, early recurrence, and poor prognosis for patients with HCC. Second, we demonstrated that aberrant activation of ERBB3-dependent signaling enhanced the migration and invasion of HCC cells, whereas the suppression of ERBB3-dependent signaling significantly inhibited the migration and invasion of HCC cells. ERBB3-dependent signaling apparently plays a crucial role in the regulation of the invasion and metastasis of HCC. Our findings also suggest that ERBB3 is a marker for the prediction of microscopic vascular invasion, intrahepatic metastasis, and early recurrence. Indeed, in an independent study designed to discover novel HCC markers for diagnosis and prognosis, we recently identified secreted erythroblastic leukemia viral oncogene homolog 3 (sERBB3) isoforms in the sera of patients with HCC. We also found that the levels of serum sERBB3 isoforms were strongly associated with portal vein invasion and metastasis of HCC; this suggests that serum sERBB3 isoforms are markers for detecting tumor invasion and metastasis and for predicting early recurrence and poor prognosis (S.-Y.H., unpublished data).

We still need to determine why EGFR-targeted and HER2-targeted therapies had only modest effects on advanced HCC in clinical trials, although EGFR and HER2 are validated therapeutic targets in many human cancers.18, 19 In this study, we found that ERBB3-dependent signaling regulates tumor cell motility and invasion rather than tumor formation and growth. Silencing of the expression of ERBB3, HER2, or EGFR in HCC cells could not effectively suppress tumor formation and growth in both in vitro and in vivo assays; this indicates that the aberrant growth signaling required for tumor initiation and growth is elicited from tyrosine kinase receptors other than EGFR, HER2, and ERBB3. Alternatively, combined signaling elicited from more than one kind of tyrosine kinase receptor orchestrates tumor initiation and tumor growth for HCC. For example, growth signaling from other receptor tyrosine kinases such as insulin-like growth factors/insulin-like growth factor receptors and hepatocyte growth factorhepatocyte growth factor receptor (c-MET) has been shown to elicit the PI3K/Akt and MAPK/Erk pathways in HCC cells.20-22 To further improve the efficacy of anti-HCC therapy, a systematic search for the receptor tyrosine kinases implicated in the tumor initiation and growth of HCC is required. Our findings suggest that ERBB3-dependent signaling is a potential therapeutic target or cotarget for the prevention and treatment of HCC recurrence and metastasis instead of the treatment of advanced HCC.

There are three possible factors contributing to the constitutive activation of EGFR/ERBB signaling: up-regulation of ERBB3 per se, activation mutations, and autocrine loops. Because the silencing of endogenous NRG1 expression suppresses the phosphorylation of ERBB3, NRG1 is required for the activation of ERBB3-dependent signaling. Therefore, the possibilities of up-regulation per se and activation mutations of ERBB3 in HCC cells are less likely. Additionally, we detected bioactive NRG1 in the conditioned media of the HCC cells, and this suggests an NRG1/ERBB3 autocrine loop driving the aberrant activation of ERBB3-dependent signaling in HCC cells. This speculation was further verified by the finding that the activity to phosphorylate ERBB3 of the conditioned media of HCC cells was eliminated by pretreatment of the conditioned media with anti-NRG1 antibodies and by silencing of the NRG1 expression of the donor HCC cells. The involvement of an NRG1/ERBB3 autocrine loop in the constitutive activation of ERBB3 signaling was also found in human ovarian cancer cells very recently.23 Also, an amphiregulin/EGFR autocrine loop was found in some lung cancers, and the detection of amphiregulin could predict the sensitivity to EGFR-targeted therapies of lung cancer with wild-type EGFR.24 We speculate that the NRG1/ERBB3 autocrine loop in HCC is a marker for the sensitivity to HER2-targeted and ERBB3-targeted therapies of HCC. Further studies are required.

Recently, Schoeberl et al.25 used a systems biology approach to identify ERBB3 as a key node in the EGFR/ERBB signaling network. Sheng et al.23 further demonstrated an activated ERBB3/NRG1 autocrine loop supporting in vivo proliferation of ovarian cancer cells. In addition, a fully human anti-ERBB3 monoclonal antibody, MM-121, binding with high affinity to ERBB3 was identified with a phage library screen.23 This antibody not only suppresses xenograft tumors with ligand-dependent activation of ERBB3 but also, when concomitantly used with cetuximab, blocks ERBB3 activity and the ensuing development of resistance to EGFR-targeted therapies in lung cancer cells with the EGFR mutation.23, 26 However, because we did not find any significant effects of ERBB3- or EGFR-dependent signaling on tumor growth in vitro and in vivo, we speculate that targeting ERBB3-, EGFR-, and HER2-dependent signaling will not be sufficient to suppress HCC tumor growth in patients with HCC; this is consistent with clinical observations. In contrast, our findings suggest that ERBB3-targeted therapies may be effective in the prevention and/or treatment of HCC invasion and metastasis. Therefore, instead of being used for advanced HCC, NRG1/ERBB3-targeted therapies should be used to treat microscopic vascular invasion in the early stages of HCC and to prevent the early recurrence of HCC, particularly for those patients who have high ERBB3 expression or are positive for the NRG1/ERBB3 autocrine loop.

It is intriguing to ask why the novel NRG1-ERBB3/HER2-Akt pathways of HCC cells identified in this study dictate HCC cell migration/invasion and not proliferation or tumor growth. It has been hypothesized that a low level of ERBB3-dependent signaling is sufficient for tumorigenesis, whereas a moderate to high level of HER2- and ERBB3-dependent signaling enhances the invasion of metastasis in human breast cancer cells.27 Alternatively, activation of a specific isoform of Akt, such as Akt2, enhances motility and invasion but not proliferation and tumor growth by the NRG1/ERBB3 autocrine loop of HCC cells. Indeed, recent studies have provided evidence for distinct functions of the three mammalian Akt isoforms.28 In breast cancer cells, Akt1 promotes cell survival and limits cell invasion, whereas Akt2 functions downstream of Twist to promote cancer cell migration and invasion.29 Further studies of the roles of ERBB3-dependent signaling and signaling elicited from other receptor tyrosine kinases in the activation of distinct isoforms of Akt in the regulation of cell transformation, proliferation, migration, and invasion in human HCC are currently ongoing in our laboratory.

In conclusion, constitutive activation of ERBB3-dependent signaling driven by an NRG1/ERBB3 autocrine mechanism is strongly associated with microscopic vascular invasion, early recurrence, and poor prognosis of HCC. ERBB3-dependent signaling plays a crucial role in the regulation of tumor invasion and metastasis of HCC rather than tumor initiation and growth. ERBB3 is a marker indicating microscopic vascular invasion and a predictor for the early recurrence of HCC. ERBB3-dependent signaling is a candidate target for the treatment of microscopic intrahepatic invasion and for the prevention of HCC recurrence.

Acknowledgements

The authors are grateful to Professor Yun-Fan Liaw for his comments on this study and to Miss Shao-Jung Lo for her technique assistance. They also thank the Taiwan Liver Cancer Network for providing some of the clinical samples for this study and the National RNAi Core of Taiwan for providing the lentivirus-based shRNA clones.

Ancillary