All clinical samples were obtained with informed consent from the HCC surgery undertaken in Eastern Hepatobiliary Surgery Hospital, Shanghai, China. From 2007 to 2010, 212 paired HCC and non-tumor samples from HBV-related HCC patients were examined, who were HBV-positive (HBsAg, HBV DNA or HBeAg serum-positive) and HCV-negative (anti-HCV serum-negative) (Table I). Thirty-one non-tumor tissues served as healthy liver controls, and were obtained from hepatic hemangioma patients, without HBV infection (serum HBsAg and anti-HBc-negative). The present study was approved by the Ethics Committee of Eastern Hepatobiliary Surgery Hospital.

Cells: HL-7702 and SMMC-7721 cells were maintained in RPMI-1640 medium (HyClone, Logan, UT, USA) supplemented with 10% fetal bovine serum (FBS; Gibco, Carlsbad, CA, USA). Huh7 and 293T cells were maintained in Dulbecco's modified Eagle's medium (DMEM) (HyClone) supplemented with 10% FBS. The cells were purchased from the Cell Bank of Institutes for Biological Sciences, Chinese Academy of Sciences, Shanghai. Primary human umbilical vein endothelial cells (HUVECs) were purchased from Lonza and were maintained in EGM2 medium (Lonza, Walkersville, MD, USA). Plasmids: the pre-miR-210 fragment was cloned from human hepatocyte genome into the pCDH-CMV-MCS-EF1-puro vector (SBI), for miR-210 lentivirus production and infection. Lentivirus package plasmids, CMV-VSVG and δ-8.9, were purchased from SBI. Lentivirus production, infection and establishment of stable transduced cell lines were performed according to the manufacturer's protocol. Reagents: TRIzol agent (Invitrogen, Carlsbad, CA, USA), puromycin, deferoxamine (DFX) (both from Sigma, St. Louis, MO, USA), Matrigel™ Matrix (BD Biosciences, Bedford, MA, USA), recombinant mouse FGF R5/FGFRL1 Fc Chimera (R&D Systems, Minneapolis, MN, USA). miR-210 mimic, mimic control, miR-210 inhibitor and inhibitor control (RiboBio, Guangzhou, China). BCA protein assay kit (Beyotime, Jiangsu, China) and Complete Protease Inhibitor Cocktail Tablets (Roche). Antibodies: anti-HIF1α, anti-FGFRL1 and anti-GAPDH (Sigma), anti-human and mouse CD34 (Abcam, Cambridge, MA, USA).

Total RNA was isolated from clinical samples or cultured cells by TRIzol according to the manufacturer's protocol. The quality of the RNA was assessed by NanoDrop 2000 (NanoDrop Technologies, Wilmington, DE, USA). miRNA qRT-PCR was performed using the TaqMan Human miRNA Assay kit (Applied Biosystems, Foster City, CA, USA) according to the manufacturer's instructions. The expression of miR-210 was normalized relative to the expression of small nuclear RNA U6, as an internal control. Data were analyzed using the comparative Ct method (2^−ΔΔCt).

Secreted FGFRL1 protein in the culture medium was purified by heparin-sepharose as previously described (40). Briefly, heparin-sepharose was washed with phosphate-buffered saline (PBS) twice, and then added to PBS 1:1 diluted culture medium, the culture medium was filtered using a 0.45 μm filter. Heparin-sepharose and the medium mixture were rotated at 4°C for 4 h. Heparin-sepharose was pelleted using a centrifuge and washed 5 times with PBS. Heparin binding proteins were eluted with Laemmli buffer at 95°C, and then subjected to SDS-PAGE for WB assay. For WB assay, the cells were lysed in lysis buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 1% Nonidet P-40 and complete protease inhibitor mixture), for 30 min at 4°C. The cell lysates were centrifuged at 12,000 × g and 4°C for 10 min, and the supernatant was used in the next step. The cell lysates were quantified using the BCA kit and adjusted to the same level in each experiment. The protein samples were then subjected to SDS-PAGE and transferred to nitrocellulose membranes for WB assay.

The protocol for endothelial cell tube formation assay was based on the procedure outlined by Corning (Corning, NY, USA). Briefly, for pre-treatment, HUVEC cells were serum-starved 12 h before the tube formation assay, and the Matrigel™ Matrix was thawed on ice overnight. Thawed Matrigel Matrix (50 μl) was added to each well of a pre-chilled 96-well sterile plate. The plate was incubated for 30 min at 37°C to allow the Matrigel Matrix to form a gel. HUVEC cells were harvested and re-suspended in the indicated medium at 1.5×10⁵ cells/ml. Cell suspension (100 μl) (1.5×10⁴ cells) was added to each well containing solidified Matrigel Matrix. The assay plate was incubated at 37°C for 4 h. The endothelial tubes were pictured for calculation using the angiogenesis-analyzer tool plugin for ImageJ.

Immunohistochemical staining for CD34 in both human and mouse liver sections was performed in the Immunohistochemistry Department of Eastern Hepatobiliary Surgery Hospital. The MVD CD34 value was assessed using the method developed by Weidner et al (41). Briefly, the CD34 immunohistochemically stained tissue sections were examined in the 200× field, 5 hotspots were pictured for MVD assessment. Any endothelial cell or endothelial-cell cluster stained brown which was clearly separate from adjacent microvessels, tumor cells and other connective tissue elements was considered a single countable microvessel. The average of the 5 hotspots in each sample was considered the MVD CD34 value.

miR-210 stably transduced SMMC-7721 and control cells were trypsinized and re-suspended in the medium. Male BALB/c nude mice (6-weeks-old) were subcutaneously inoculated in the dorsal area with 2.0×10⁶ cells in 0.1 ml. Six weeks after inoculation, the mice were sacrificed and the tumors were dissected and fixed for CD34 immunohistochemical staining.

Data are expressed as the mean ± SEM. Differences between groups were assessed using the t-test or Mann-Whitney U test. Time-to-recurrence was calculated according to the Kaplan-Meier method and compared using the log-rank test. All statistical analyses were performed using Prism GraphPad software (GraphPad Software, La Jolla, CA, USA). A P-value <0.05 denoted the presence of a statistically significant difference.

In order to examine the expression of hypoxic miR-210 in HCC, miR-210 expression was measured by qRT-PCR in 212 paired HCC tumor and non-tumor samples obtained from surgery (Fig. 1A). miR-210 expression in HCC tissues was significantly higher than that in non-tumor tissues. In addition, to determine whether hepatitis stress induced miR-210 expression, 31 healthy liver samples obtained from HBV-negative hepatic hemangioma patients were examined, and we found that miR-210 expression level in healthy liver was significantly lower than that in the HCC non-tumor tissues (Fig. 1A). Thus, miR-210 expression was increased progressively from normal liver and adjacent non-tumor tissues, to HCC tissues.

We also determined whether miR-210 expression was correlated with the progression of liver disease. According to the pathological reports, we divided non-tumor samples into the hepatitis group (100 patients) and cirrhosis group (112 patients), the HCC samples were divided into the BCLC I+II group (33 patients) and the BCLC III+IV group (179 patients). The hepatitis group had a higher miR-210 expression level than that in the cirrhosis group (P=0.0093), and the BCLC III+IV group had a higher miR-210 expression level than that in the BCLC I+II group (P=0.029) (Fig. 1B). The decrease of miR-210 in the cirrhosis group suggests that miR-210 expression may be correlated with liver damage (42), and the increase of miR-210 in the BCLC III+IV group suggests that miR-210 induction is correlated with HCC progression.

To analyze the correlation between miR-210 expression and HCC prognosis, we divided the patients into two groups according to miR-210 expression in HCC using the median value as the cut-off. The clinicopathological and demographic characteristics of the patients in these two groups are listed in Table I. Compared to the miR-210 low group, miR-210 high group had significantly less tumor encapsulation (P=0.020), larger tumor size (7.82±4.07 vs. 6.29±3.89 cm; P=0.035), more vascular invasion (P=0.008) and a higher proportion of undifferentiated tumors (P=0.014). Furthermore, patients with high miR-210 expression in HCC had significantly shorter tumor-free survival (P=0.036; Fig. 1C) and overall survival (P=0.012; Fig. 1C), as compared to that of the miR-210 low patients. Moreover, multivariate analysis showed that HCC miR-210 expression (P=0.03; Table II) and vascular invasion (P=0.013; Table II) were significant independent factors for tumor-free survival, and HCC miR-210 expression (P=0.012; Table III) and tumor diameter (P=0.001; Table III) were significant independent factors for overall survival. Thus, miR-210 expression is significantly correlated with the prognosis of HCC patients.

To further assess the possible prognostic value of miR-210, we evaluated miR-210 expression in non-tumor liver tissues. The expression of miR-210 in non-tumor tissues was not correlated with tumor-free survival (P=0.479) or overall survival (P=0.290) of HCC patients (data not shown). However, the distribution of relative tumor/non-tumor (T/nT) miR-210 expression suggested the complicated regulation of miR-210 expression in the tumor microenvironment (Fig. 1D). According to the T/nT miR-210 expression fold-change, we used one fold as the cut-off value (35) and further divided all patients into 3 groups, T/nT >2.0 (91 patients), T/nT <0.5 (35 patients), and T/nT 0.5–2.0 (86 patients). The Kaplan-Meier analysis of both tumor-free and overall survival showed that the 3 groups of patients were clearly distinguished (P=0.0001; Fig. 1E), suggesting that miR-210 expression T/nT <0.5 may serve as a favorable prognostic factor for HBV-related HCC. Thus, miR-210 expression in HCC, particularly the T/nT value, is a new prognosis predictor for HCC patients.

As the role of miR-210 in HCC angiogenesis remains unclear, we then investigated the correlation between miR-210 expression in HCC and intratumor MVD. The results of CD34 immunohistochemistry showed that HCC tissues with high miR-210 expression had high CD34 staining (Fig. 2A). In addition, miR-210 expression in HCC was positively correlated with the MVD CD34 value (R2=0.1925; P=0.0018; Fig. 2B), in a randomly selected group of 48 HCC patients. To determine whether miR-210 is an independent angiogenesis inducer, we established miR-210 stably overexpressed HCC cell lines SMMC-7721 and HL-7702 and the corresponding control cell lines for in vivo and in vitro assays. The expression level of miR-210 was confirmed by qRT-PCR in each cell line (Fig. 2C and E), and we inoculated both control and miR-210 stably overexpressed SMMC-7721 cells into nude mice. CD34 staining of the xenograft showed that miR-210 stably overexpressed SMMC-7721 cells had higher CD34 staining as compared to that of control cells (Fig. 2D), suggesting that HCC cells with increased miR-210 may send a signal to recruit more endothelial cells for vascular formation. Moreover, we used the medium supernatants from control and miR-210 stably overexpressed HL-7702 cells for the tube formation assay, and the result showed that the supernatants from HL-7702 of increased miR-210 expression induced greater tube formation (Fig. 2F). Thus, we conclude that increased miR-210 expression in HCC promotes angiogenesis, which may be dependent on the secreted cytokines.

We next investigated the target genes of miR-210, which are responsible for miR-210 mediated promotion of HCC angiogenesis. As the target genes of miR-210 have been studied in several genomic approaches (8,43), we only screened for the experimentally proved target genes, which may regulate angiogenesis with extracellular secretion capability and which were relatively conserved between human and mouse. One of the target genes of miR-210, FGFRL1, was found to fulfill all the criteria listed above. FGFRL1, also known as FGFR5, and belongs to the fibroblast growth factor receptor (FGFR) family. FGFRs consist of an extracellular region of 3 immunoglobulin-like domains, a single hydrophobic membrane-spanning segment, and a cytoplasmic tyrosine kinase domain for cascading downstream signals. The extracellular portion of FGFRs binds to fibroblast growth factors (FGFs). However, as a decoy FGF receptor (44), FGFRL1 lacks the cytoplasmic tyrosine kinase domain (Fig. 3A).

It has been reported that FGFRL1 can be shed from the cell membrane and secreted in myoblasts and HEK293 cells (40), but it is unclear whether hepatocytes can secrete FGFRL1. We established an FGFRL1 stably transduced SMMC-7721 cell line to evaluate the shedding of FGFRL1 from hepatocytes. As previously mentioned, FGFRL1 binds to heparin (40), therefore, the cell culture supernatant was subjected to heparin purification and then was analyzed by WB, while the whole cell lysate was served as a control (Fig. 3B). We obtained a WB signal for shed FGFRL1 in the cell culture supernatant, with a lower molecular weight due to loss of transmembrane and intracellular domains. This finding showed that FGFRL1 is released from the cell membrane in a soluble form, suggesting that FGFRL1 acts as a secreted regulator of FGF signaling.

Next, we determined whether hypoxia treatment changed the expression of FGFRL1. SMMC-7721 cells were treated with DFX for different time periods. HIF1α expression was induced while the expression of FGFRL1 declined (Fig. 3C). To assess whether the decrease of FGFRL1 following hypoxia was miR-210-dependent, we transiently transfected SMMC-7721 cells with miR-210 mimic, miR-210 inhibitor and their respective controls. The results showed that compared with mimic-control transfected cells, miR-210 mimic transfected cells had lower expression of FGFRL1, which is similar to the decrease seen in DFX-treated cells (Fig. 3D). miR-210 inhibitor transfected cells showed no decrease of FGFRL1 when treated with DFX (Fig. 3D). These results indicated that hypoxia-induced downregulation of FGFRL1 was mainly through miR-210. We further assessed whether expression of miR-210 downregulated FGFRL1 secretion. The heparin extract of the supernatant and whole cell lysate of miR-210 stably transfected and control cells were examined by WB. The results showed that miR-210 expression downregulated FGFRL1 both inside and outside the cells (Fig. 3E). Hence, we conclude that hypoxia-induced miR-210 targets FGFRL1 expression and secretion in HCC.

Although there are no reports on the angiogenesis functions of FGFRL1, we propose that FGFRL1 may inhibit angiogenesis by blocking FGF2 (45). To prove our hypothesis, we used a commercially available recombinant FGFRL1 protein in an in vitro angiogenesis assay. An in vitro HUVEC tube formation assay was performed to evaluate the effect of FGFRL1 on the anti-angiogenic response. HUVEC cells were seeded on Matrigel support using different stimulations: complete EGM2 medium without FGF2 (EGM2 - FGF2), complete EGM2 medium (EGM2), complete EGM2 medium with FGFRL1 (EGM2 + FGFRL1), and complete EGM2 medium with FGFRL1 plus FGF2 (EGM2 + FGFRL1 + FGF2). As shown in Fig. 4A, EGM2 + FGFRL1 significantly inhibited in vitro tube formation as compared to EGM2, while additional FGF2 rescued inhibition shown in the EGM2 + FGFRL1 + FGF2 group. Thus, FGFRL1 is a negative regulator of HCC angiogenesis, and hypoxia-induced miR-210 promotes HCC angiogenesis by targeting FGFRL1 expression and secretion.

Based on our findings, we propose the following working model to explain how induced miR-210 expression promotes HCC angiogenesis. Hypoxia induced the transcription factor HIF1α expression and activation in HCC cells. HIF1α then promotes the secretion of FGFs and VEGF to enhance HCC angiogenesis. Alternatively, HIF1α induced the expression of hypoxic miR-210 to target FGFRL1 expression and secretion. As FGFRL1 is a decoy FGF receptor and an inhibitor of HCC angiogenesis, hypoxia-induced miR-210 promotes HCC angiogenesis by targeting FGFRL1 expression (Fig. 4B).