Heregulin-1ß and HER3 in hepatocellular carcinoma: status and regulation by insulin

Eighty-five HCC (T) and paired adjacent non-tumour (NT) liver tissues were collected from patients undergoing curative liver resection for HCC at the Saint-Antoine hospital (Paris, France). Clinicopathological characteristics are summarized in Table 1. Part of this collection was used in our previous study where it was designated as collection #2 [21]. All patients gave informed consent to the study, which was conducted in accordance with the French laws and regulations (CNIL n° 1913901 v 0).

HepG2, Hep3B, and Huh7 cells were obtained from the American Type Culture Collection (ATCC). PLC/PRF5 cells were provided by Dr Christine Perret (Institut Cochin, France). Cell line authentication was routinely performed by using a panel of nine ATCC short tandem repeats. Cell lines were cultured as previously reported [22] and routinely controlled for mycoplasma contamination. Primary cultures of human hepatocytes were established as previously described [23]. In some experiments, serum-deprived cells were treated with insulin, IGF-II (Sigma-Aldrich), heregulin-1ß (R&D Systems), cycloheximide, bafilomycin A1 (Sigma-Aldrich) and/or MG-132 (Cell Signaling Technology).

Protein electrophoresis and transfert to nitrocellulose were performed according to standard procedures. The primary antibodies are summarized in Additional file 1: Table S1. Blot revelations and quantifications were performed using ChemiDoc™ Touch Imaging System (BIORAD). Total amounts of HER3 and heregulin-1β were quantified in human liver tissue extracts by ELISA according to manufacturer’s instructions (R&D Systems).

Total RNA was extracted from cell cultures using RNeasy Mini kit (Qiagen). For liver tissues, a preliminary RNA extraction step was performed using TRIzol Reagent (Life Technologies). Quantitative measurements of transcripts were performed by real-time PCR on a LightCycler 480 instrument (Roche) using SYBR Green chemistry and specific primers (Additional file 1: Table S2). For each sample, gene expression was normalized to that of hypoxanthine guanine phosphoribosyltransferase mRNA content and was expressed relatively to the same calibrator. The relative quantity of each target gene was determined from replicate samples using the formula 2^-∆∆Ct.

Paraffin-embedded 4-μm sections were dewaxed in xylene and rehydrated in graded alcohol series and microwave antigen retrieval was performed in 10 mM citrate buffer pH 6 during 15 min. Primary antibody detection was performed using Novolink Polymer Detection System (Novocastra, Leica Biosystems) according to the manufacturer’s protocol on an automated staining system (Dakocytomation®). Aminoethyl carbazole was used to reveal the peroxidase activity. The sections were counterstained with haematoxylin. Four different HER3 monoclonal antibodies were tested (Additional file 1: Table S1) but only the clone RTJ1 (Thermo Scientific, 1:30 dilution, overnight at 4 °C) gave signals. The clone RTJ1 was validated by siRNA-mediated knockdown and Western blotting (Additional file 2: Figure S1).

Cells seeded on glass coverslips were fixed with 4 % paraformaldehyde, blocked with 1 % BSA and 10 % goat serum in PBS, followed by overnight incubation with a 1:250 dilution of the primary antibody in PBS at 4 °C (Additional file 1: Table S1). Cells were then incubated with a 1:200 dilution of conjugated secondary antibody (Alexa Fluor® 488 or 546 dye) in PBS for 1 h at room temperature. The slides were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) for nuclei detection. Fluorescence was visualized using an immunofluorescence microscope (Leica Microsystems) with a DFC300 FX digital camera.

Migration was performed in 6.5 mm Transwell® with 8-μm pore polycarbonate membrane insert (Corning). Cells (1 × 10⁵) in medium without serum were plated in the upper chamber and medium containing 10 % fetal bovine serum (FBS) was added in the lower chamber as a chemoattractant. After 24 h, cells were treated with or without insulin or heregulin-1β for 24 h. Cells that had not migrated were removed from the upper surface of inserts by using cotton-tipped swabs and migrated cells that were attached to the lower surface were enumerated by microscopy following fixation by 4 % paraformaldehyde for 15 min and nucleic acid staining with DAPI. Four random fields were counted per insert. For the wound-healing assay, cells (5 × 10⁴) were seeded in a 24-well dish, incubated for 24 h in complete medium and serum-harvested overnight. The cell monolayers were pretreated with mitomycin C (1 μg/ml) for 2 h to inhibit cell proliferation, scraped with a p200 pipet tip and washed. Cells were then treated with or without insulin or heregulin-1β. Photographs were taken at 0 and 48 h with a phase-contrast microscope and at least ten fields were recorded for each treatment.

We analysed the expression of HER3 mRNA in a French collection of 85 resected HCC. We observed that HER3 mRNA was increased in 52 % of HCC (T) versus paired non-tumour (NT) liver tissue with a median fold-induction of 1.94 (Fig. 1a). High HER3 T/NT ratios were associated with HBV, while lower values were associated with HCV (Table 2). Except for tumour size, the HER3 T/NT ratio was not associated with histological and biological markers reminiscent of tumours with poor outcome such as CK19 expression, satellite nodules, multiplicity, microvascular invasion and high serum levels of AFP (Table 2). The HER3 T/NT ratio was higher in well/moderately differentiated tumours than in poorly differentiated tumours. Finally, there was no difference in overall survival (OS) or recurrence-free survival (RFS) in patients with high or low HER3 T/NT ratio (Fig. 1b).

Similar analyses were conducted for heregulin-1ß ligand. Heregulin-1ß mRNA levels were lower in 82 % of HCC compared with surrounding liver tissue (median fold-ratio of 0.09) (Fig. 1c). The heregulin-1ß expression ratio (T/NT) correlated neither with tumour size, multiplicity, microvascular invasion, OS nor RFS (Table 2 and Additional file 3: Figure S2); it was higher in CK19-negative and well/moderately differentiated tumours.

We then attempted to investigate HER3 protein expression by immunohistochemistry. Out of four primary antibodies tested, only the RTJ1 clone yielded staining. This staining was exclusively cytoplasmic. This finding raises questions regarding the validity of these signals since HER3 is supposed to be also localized at cell membrane. Moreover, HER3 staining turned out to be nonreproducible in terms of intensity as illustrated in Fig. 2a and no reliable association studies could be performed from these analyses.

Thirty-two paired T/NT tissue samples previously analysed for HER3 mRNA expression were also measured for HER3 protein levels by ELISA. We observed no statistical difference between HER3 protein contents in tumours vs nontumour liver tissues (Fig. 2b, left). Spearman correlation analysis indicated no correlation between HER3 mRNA levels (evaluated by real-time PCR) and HER3 protein levels (evaluated by ELISA) (Spearman r = −0.230; p = 0.071; Fig. 2b, right). In contrast, heregulin-1ß protein levels were decreased in tumours compared with nontumour liver tissues (Fig. 2c, left) and heregulin-1ß mRNA and protein levels tended to be positively correlated (Spearman r = 0.289; p = 0.023; Fig. 2c, right). The lack of correlation between HER3 transcript and protein levels suggested that HER3 is subjected to post-transcriptional and/or post-translational regulation in human liver samples.

Limited data are available regarding the post-transcriptional and post-translational mechanisms controlling HER3 expression in hepatocytes. A few years ago, Carver and colleagues reported that in cultured rat hepatocytes insulin impaired heregulin-1ß signalling by promoting HER3 downregulation [24]. We examined whether such a mechanism was relevant to human normal and transformed hepatocytes, that expressed both HER3 and IR receptors [21, 22]. As shown in Fig. 3a, insulin down-regulated HER3 protein expression in primary cultures of normal human hepatocytes. A similar effect was observed in the presence of IGF-II, another ligand to the insulin receptor (IR). Insulin and IGF-II also decreased HER3 protein levels in HepG2 and PLC/PRF5 cancer cells but not in Hep3B and Huh7 cells (Fig. 3b). There was no decrease in HER3 mRNA levels after insulin treatment in HepG2 and PLC/PRF5 cells (Additional file 4: Figure S3A), showing that the effect of insulin was posttranscriptional. To determine whether the decreased levels of HER3 in insulin-stimulated HCC cells were due to increased degradation, we blocked the de novo protein synthesis with cycloheximide and analyzed the rates of HER3 decay. The half-life of HER3 in insulin-stimulated cells was shorter (2–3 h) than in unstimulated cells (>6 h) (Fig. 3c) indicating that insulin destabilizes HER3 protein. We then examined how insulin promoted HER3 degradation. First, we observed that treatment of HepG2 and PLC/PRF5 cells with the specific proteasome inhibitor MG132 for 24 h reversed the effect of insulin on HER3 downregulation while bafilomycin A, a selective inhibitor of lysosomal v-ATPase, was ineffective (Fig. 3d). This indicated that the negative regulation of HER3 protein levels by insulin was a proteasome-mediated process. Additionally, it seemed that the effect of insulin did not require the involvement of NRDP1, an E3 ubiquitin ligase, which can control the steady-state levels of HER3 [25]. Indeed, insulin did not modify NRDP1 expression and more importantly siRNA-induced downregulation of NRDP1 expression did not impact insulin effect on HER3 expression (Additional file 4: Figure S3 B and C).

We then examined whether insulin impacted heregulin-1ß/HER3-dependent biological effects in Hep3B and Huh7 cell lines in which the hormone did not down-regulate HER3 expression. As shown in Fig. 4a, heregulin-1ß promoted cell migration as evaluated by Transwell migration assays in these cell lines while it was ineffective to stimulate proliferation and viability (Additional file 5: Figure S4). Insulin was a potent mitogen in these cell lines [21] but had no effect on cell migration (Fig. 4a). When insulin was combined to heregulin-1ß, we observed that the promigratory effect of heregulin-1ß was reduced in Hep3B and Huh7 cell lines (Fig. 4a). The ability of insulin to counteract heregulin-1ß migratory effect was confirmed in Huh7 cells using a wound healing assay (Fig. 4b).

As a first step to investigate the mechanisms underlying the functional interaction between heregulin-1ß- and insulin-dependent pathways, we examined the cellular localization of HER3 and IR by immunofluorescence (Fig. 5a). In HCC cell lines, the two receptors showed a partially overlapping pattern of distribution. A potential interaction between IR and HER3 in HCC cells was next assessed by immunoprecipitation. HER3 was immunoprecipitated from whole-cell lysates, and the resulting precipitates were subjected to immunoblot analysis with an antibody to IR. In all cell lines, we observed that HER3 coimmunoprecipitated with IR (Fig. 5b).

We then examined the impact of insulin on the heregulin-1ß/HER3 signalling axis in HCC cells. The effect of insulin on HER3 phosphorylation was evaluated by Western blot by analysing two major tyrosine phosphorylation sites, Y1289 and Y1197. Strikingly, the hormone induced an increase of HER3 Y1289 phosphorylation in Hep3B cells while it was ineffective to promote HER3 Y1197 phosphorylation (Fig. 6a, left). As a comparison, both tyrosine residues were potently phosphorylated by heregulin-1ß (Fig. 6a, right). Insulin-induced phosphorylation of HER3 Y1289 was reduced when IR expression was downregulated with siRNA (Fig. 6b). The stimulatory effect of insulin on HER3 Y1289 phosphorylation was observed in the three other cell lines but to a lesser extent (Additional file 6: Figure S5). As insulin also promoted EGFR phosphorylation in HCC cells (Fig. 6c), we wondered whether insulin induction of Y1289 HER3 phosphorylation required EGFR activity. The effect of insulin on HER3 was maintained in the presence of gefitinib, an EGFR TKI, or after EGFR downregulation with siRNA suggesting that it occurred independently of EGFR activation (Fig. 6c). Finally, we examined the consequence of IR downregulation with siRNA on HER3 and AKT phosphorylation in the presence of increasing doses of heregulin-1ß. As shown in Fig. 6d, HER3 phosphorylation tended to be reduced while AKT phosphorylation was enhanced and detectable at lower doses of ligand after IR depletion. Altogether, these data indicate that HER3 is in a close proximity to IR in HCC cells and that IR exerts a negative constraint on heregulin-1ßlHER3 that could involve IR-mediated HER3 Y1289 phosphorylation.