Male BALB/c nu/nu mice (4–6 weeks old) weighing approximately 20 g (Shanghai Institute of Materia Medica, Chinese Academy of Sciences, Shanghai) were housed in laminar-flow cabinets under specific pathogen-free conditions. The experimental protocol was approved by the Shanghai Medical Experimental Animal Care Committee. The model of human HCC in nude mice (LCI-D20) established in our institute 12 34 via orthotopic implantation of a histologically intact metastatic tumor tissue (at an average volume of 1 mm³) was used in this study.

To observe the effect of IFN-α treatment, 12 HCC-bearing mice were randomized to receive normal saline (group A, subcutaneous injection of 0.2 ml normal saline, n=6) or IFN-α (group B, subcutaneous injection of 0.2 ml at a dose of 1.5×10⁷ U/kg/day, n=6; recombinant human interferon-α-1b, Kexing Bioproducts Co., Shenzhen, China) 34 . Treatment was started on day 7 after implantation and continued for 20 days (Figure 1, Additional file 1: Figure S1).

To examine tumor growth after discontinuation of IFN-α treatment, 12 HCC-bearing mice were first treated with normal saline (group C) or IFN-α (group D) for 20 days, after which treatment was discontinued and the mice observed for another 20 days (both groups, n=6). To further investigate the effect of re-initiated IFN-α treatment, another 12 HCC-bearing mice were first treated with IFN-α for 20 days at the same dosage as group B, then observed for another 20 days after treatment was discontinued. The mice were then randomized to receive either normal saline (group E, n=6) or IFN-α (group F, n=6, at the group B dosage) for 20 days.

To study the effect of IFN-α treatment on a relatively larger tumor, tumors were allowed to grow for 35 days after implantation in 12 HCC-bearing mice (tumor weight was approximately 1.5 g, the same size as the tumor at the beginning of the re-initiated IFN-α treatment). Mice were then randomized to receive either normal saline (group G, n=6) or IFN-α at a dose of 1.5 × 10⁷ U/kg/day (group H, n=6) for 20 days.

The serum VEGF concentration was analyzed by enzyme-linked immunosorbent assay (ELISA) using the Quantikine VEGF ELISA kit (R&D Systems, Minneapolis, MN). All analyses were conducted in duplicate. The concentrations of VEGF in unknown samples were determined by comparing the optical density of the samples to the standard curve.

The immunohistochemical study was performed as detailed previously 35 . The primary antibody was rat anti-mouse CD31 (1:40, BD Biosciences, San Jose, CA) to measure MVD. The components of Envision-plus detection system (EnVision+/HRP/Mo; Dako, Carpinteria, CA) were applied. Reaction products were visualized by incubation with 3,3’-diaminobenzidine. Negative controls were treated identically but with the primary antibody omitted. To measure MVD, any brown-staining endothelial cell cluster distinct from adjacent microvessels, tumor cells, or other stromal cells was considered a single countable microvessel. The entire section was observed under low magnification, and three highly vascularized areas per tumor were then evaluated at high magnification (200×).

Frozen tissues were sectioned into 8-μm-thick slices, which were stained with rat anti-mouse CD31 (1:40, BD Biosciences) overnight. Bound antibodies were detected by incubation with biotin-coupled rabbit anti-rat antibody (1:200, Southern Biotech, Birmingham, AL), sections were then incubated for 30 min at room temperature with DyLight 488 streptavidin (1:500, Vector, Burlingame, CA). Vascular pericyte staining with anti-α-SMA (1:100, sigma, St. Louis, MO) was performed. The biotin-conjugated secondary antibody and DyLight 549 streptavidin (1:500, Vector, Burlingame, CA) were used to visualize pericytes simultaneously. Terminal deoxynucleotidyl transferase-mediated dUTP-biotin nick-end labeling was used to detect apoptosis according to the manufacturer’s instructions. The apoptotic tumor cells within areas of a viable tumor were counted (more than 2000 cells in each sample) at 400× magnification and proliferative indexes were calculated.

The GEArray Q Series Angiogenesis Gene Array kit (SuperArray Bioscience, Bethesda, MD) was used to characterize the gene expression profiles of differently treated mice. Hybridization was performed according to the manufacturer’s instructions, as described previously 34 36 . Experiments were performed in duplicate, and the reproduction rate was >98% between experiments.

Total RNA was extracted by the Trizol method (Life Technologies Inc., Grand Island, NY). RNA was pretreated with DNase I (Invitrogen, Carlsbad, CA), and 3 μg of RNA was subjected to reverse transcription at 42°C for at least 1 h in a 20-μl reaction mixture. The PCR products were analyzed by electrophoresis in a 1.5% agarose gel. Details on the VEGF 37 and GAPDH primers and the reaction parameters are provided in Additional file 1: Table S1.

The ABI Prism Primer Express Software (Applied Biosystems, Foster City, CA) was used to design probes, forward primers, and reverse primers of VEGF165 38 , PDGF-A, and GAPDH (Additional file 1: Table S2, Sangon Biotech, Shanghai, China). An ABI Prism 7700 Sequence Detection System (Applied Biosystems) was used to analyze RT-PCR, as described previously 38 . The relative amounts of VEGF165 and PDGF-A mRNA were calculated as detailed previously 39 .

Protein expression was assessed only in the non-necrotic part of the tumor. Western blotting was performed for rabbit anti-VEGF, rabbit anti-PDGF-A (Santa Cruz), rabbit anti-phosphor-VEGF receptor-2 tyrosine 951, rabbit anti-phosphor-PDGF receptor α tyrosine 754, rabbit anti-AKT, rabbit anti-phosphor-AKT, rabbit anti-ERK, rabbit anti-phosphor-ERK, and mouse anti-β-actin (Cell Signaling Technology, Beverly, MA) after SDS-PAGE and electrophoretic transfer to polyvinylidene fluoride membranes. Then incubation was performed with horseradish peroxidase–conjugated goat anti-rabbit or goat anti-mouse antibodies (Amersham Life Technologies, Arlington Heights, IL) for 1 h.

Continuous data were expressed as mean and standard deviation or mean and 95% confidence intervals (CIs). Groups were compared with the Mann–Whitney U-test using the software package SPSS 10.0 (SPSS Inc., Chicago, IL). All tests were two-tailed, and data were considered to be statistically significant at P < 0.05.

Tumors in IFN-α treatment group B were smaller than those in control group A (0.27 ± 0.19 g vs. 0.68 ± 0.24 g, P = 0.009; Figure 2A). The serum VEGF levels were also lower in the IFN-α–treated mice than in the control mice (24.5 ± 8.7 pg vs. 41.6 ± 12.0 pg/ml, P = 0.018, Figure 2B), and MVD was also markedly decreased (Figure 2C), with a mean MVD of 22/HP (95% CI = 15–29/HP) in group B versus 46/HP (95% CI = 32–60/HP, P = 0.004) in group A. Furthermore, a higher apoptotic rate (Figure 2D) was found in IFN-α–treated tumors compared with control tumors (8.60% ± 2.42% vs. 3.47% ± 1.13%, P = 0.001).

However, there was no significant difference in the tumor weight and MVD between groups C and D when IFN-α treatment was discontinued for 20 days (1.93 ± 0.58 g vs. 1.54 ± 0.49 g, P = 0.233; 59, 95% CI = 43–73 vs. 51, 95% CI = 38–64, P = 0.369, respectively). Serum VEGF levels in group D were not statistically different from those in control group C (54.3 ± 13.8 pg/ml vs. 63.7 ± 17.5 pg/ml, P = 0.327) and were much higher than those in group B immediately after the first treatment course (Figure 2), suggesting a significant relapse of tumor growth and angiogenesis happened after IFN-α treatment was stopped.

No significant inhibitory effect on tumor size (3.81 ± 0.92 g vs. 3.22 ± 0.62 g, P = 0.217), MVD (70, 95% CI = 59–81 vs. 61, 95% CI = 48–74, P = 0.198) and apoptotic index (4.84% ± 1.16% vs. 5.56% ± 1.41%, P = 0.360) (Figure 2) was observed between group E (normal saline control) and group F (re-initiated IFN-α treatment). However, the serum VEGF concentration in group F was decreased again when IFN-α treatment was re-initiated as compared with control mice (92.8 ± 24.3 pg/ml vs. 58.0 ± 13.7 pg/ml, P = 0.012, Figure 2B).

It is possible that poorer tumor response to re-initiated IFN-α treatment may result from a larger tumor size prior to treatment. To test this hypothesis, tumors were allowed to grow for 35 days until the tumor weight was about 1.5 g (the same size as the tumor at the beginning of the re-initiated IFN-α treatment). IFN-α treatment for 20 days (group H) resulted in a significant reduction of tumor weight compared with control group G (2.28 ± 0.63 vs. 3.90 ± 0.80 g, P = 0.003), which suggested that the reduced inhibitory effect of re-initiated IFN treatment was not influenced by the pre-treatment tumor size.

Double immunofluorescence staining of apoptotic cells and endothelial cells was used to investigate whether IFN-α treatment resulted in endothelial cell apoptosis. IFN-α treatment induced a marked increase of endothelial cell apoptosis and decreased MVD in group B compared with control group A. In the mice treated by re-initiated IFN-α, very few apoptotic endothelial cells could be detected in groups E and F, both of which had a higher MVD (Figure 3A).

Two independent cDNA microarray analyses showed that IFN-α treatment down-regulated several pro-angiogenesis factors, including VEGF, angiogenin, basic fibroblast growth factor (bFGF), PDGF-A, transforming growth factor (TGF), tumor necrosis factor (TNF), granulocyte colony stimulating factor (G-CSF), epidermal growth factor (EGF), insulin-like growth factor (IGF), and interleukin (IL)-8 in the first treatment course. All the pro-angiogenesis factors except PDGF-A were again down-regulated in group F when compared with group E (Table 1).

PCR showed two transcripts whose base-pair length corresponded to those of VEGF165 and VEGF121, as well as a small transcript corresponding to VEGF189 (Figure 3B). IFN-α treatment decreased VEGF165 expression in the first IFN-α treatment group (group B) and the re-initiated IFN-α treatment group (group F) compared with their control groups A and E, respectively. PDGF-A expression was decreased in group B compared with control group A, but increased in the re-initiated IFN-α treatment group F compared with control group E (Figure 3C, 3D). These results were consistent with the cDNA microarray analyses.

Western blotting for VEGF165 of HCC tissues under reducing conditions showed two bands of 23 and 21 kDa, respectively (Figure 4A), consistent with two glycosylation variants of VEGF165. No other isoform of VEGF was detected. Tumors from group B expressed a lower level of VEGF165 than those of group A. Similarly, the VEGF165 concentration of group F was lower than that of group E. In addition, VEGF receptor-2 phosphorylation of Y951, which plays a critical role in pathological angiogenesis, was higher in control groups A and E but present at a very low level in the IFN-α–treated groups B and F, which suggests IFN-α retained its ability to inhibit VEGF signaling in the first and re-initiated treatment courses. PDGF-A in group B was down-regulated significantly compared with control group A. However, in the second IFN-α treatment course, group F had a higher PDGF-A level than its control group E (Figure 4).

Combination treatment with imatinib and IFN-α resulted in the decrease of PDGF receptor phosphorylation level and a lower AKT and ERK phosphorylation level, compared with IFN treatment alone or the control group (Figure 5A). Meanwhile, tumor growth was inhibited significantly (2.23 ± 0.43 g) in the combination treatment group in contrast with IFN-α treatment alone (3.22 ± 0.62 g, P = 0.01), imatinib treatment alone (3.10 ± 0.62 g), and the normal saline group (3.81 ± 0.92 g, P = 0.003). The difference in tumor size between IFN-α or imatinib treatment alone and the control group was not significant(P = 0.751). Furthermore, the decrease in angiogenesis was demonstrated by lower MVD in IFN-α combined with imatinib treatment group (42, 95% CI = 32–52) compared with IFN-α treatment alone (61, 95% CI = 48–74, P = 0.015). The results of immunofluorescence double staining showed that, compared with the groups treated with IFN-α or normal saline, the tumor vessels in the combined treatment group had fewer signals detected by anti-α-SMA antibody (Figure 5).