The uses of the Chinese medicine formula, BYHWD, and its disassembled prescriptions, Yiqi decoction (YQD) and Huoxue decoction (HXD), in the present study were authorized according to the medical publications; Formulas of Chinese medicine (20) and the Chinese Pharmacopoeia (21). BYHWD was composed of seven medicinal components: 120 g milkvetch root, 6 g Chinese angelica, 5 g red peony root, 3 g earth worm, 3 g Szechwan lovage rhizome, 3 g peach seed and 3 g safflower. YQD consisted of only 120 g milkvetch root, and HXD consisted of the remaining medicinal components: 6 g Chinese angelica, 5 g red peony root, 3 g earth worm, 3 g Szechwan lovage rhizome, 3 g peach seed and 3 g safflower. All the herbal components were purchased from Shanghai Pharmacy (Shanghai, China), and authenticated by experts at the Department of Pharmacology, Shanghai University of Chinese Medicine (Shanghai, China). The three medicinal herb decoctions were prepared as follows. The mixture of the crude drugs were soaked in 500 ml distilled water and decocted by boiling for 35 min. The resulting decoction was then filtered through a 0.22 µm polytetrafluoroethylene filter (Whatman; GE Healthcare Life Sciences, Chalfont, UK) and collected. The remnants were added to 350 ml distilled water, and decocted by boiling was continued for 20 min, followed by filtering. The filtered decoctions were combined and condensed to a specific dose. The doses were calculated according to the formula: dB = dA ^* RB/RA^* (WA/WB)¹/³ (22), where dA represents the human dose and dB represents the mouse dose; WA is the average human weight (60 kg) and WB is the average mouse weight (0.2 kg). R is the build coefficient (RA=1.00; RB=0.59). The combined filtrates of BYHWD, YQD and HXD were condensed to 1.03, 0.56 and 0.166 g/ml, respectively. All the combined filtrates were stored at 4°C until use.

Male and female athymic BALB/c nu/nu mice (15–50 g; 4–6 weeks old) were purchased from Shanghai Slack Experimental Animals Co., Ltd. (Shanghai, China) and maintained under specific pathogen-free conditions. All mice were handled according to the Guidance Suggestions for the National Institutes for the Care and Use of Laboratory Animals (23). The study was approved by the ethics committee of Shanghai University of Traditional Chinese Medicine.

The HCCLM3 HCC cell line, which was cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (Gibco; Thermo Fisher Scientific, Inc., Waltham, MA, USA), was established at the Liver Cancer Institute of Fudan University (Fudan, China) (24). The human HCC tumor models were established using HCCLM3 in nude mice via the orthotopic implantation of intact metastatic tumor tissue, as described in previous reports (25,26). Briefly, following the acquirement of HCCLM3, 5×10⁶ (0.2 ml) cells were injected subcutaneously into four nude mice (male or female). When the subcutaneous tumor had reached ~1.5 cm in diameter, the mice were sacrificed by cervical dislocation. The tumor tissue was removed, cut into small sections (~1 mm³) and implanted into the liver of separate recipient mice, which were kept in standard facilities. This animal model showed 100% spread into the liver and metastasis to the lungs. Abnormal serum α-fetoprotein was excreted, and hepatitis B surface antigens were also identified in this model.

A total of 96 nude mice bearing orthotopic xenografts were randomly divided into four groups: Model control group (LM); BYHWD-treated group (LB), YQD-treated group (LY) and HXD-treated group (LH). Each of these groups was randomized into three subgroups, which were treated consecutively for 21 days (LM21, LB21, LY21 and LH21), 28 days (LM28, LB28, LY28 and LH28), and 35 days (LM35, LB35, LY35, and LH35), respectively. Each subgroup contained eight mice. Treatment began the day following that on which xenograft surgery was performed. The groups of animals (n=8) were orally gavaged with 0.2 ml of the respective treatment solution twice daily, with the exception of the animals in the model group, which received 0.2 ml of 0.9% sodium chloride solution twice daily. Daily general observations and weekly mice body weights were all recorded.

The mice were sacrificed after 21, 28 and 35 days, respectively. The tumors were removed, images were captured, and the tumor tissues were weighed and processed for histology. Following autopsy, the longest (a) and the smallest (b) diameters of the tumors were measured using a slide gauge (Control Co., Friendswood, TX, USA) under an operating microscope (OPMI Pentero; Carl Zeiss, Oberkochem, Germany). The tumor volume was calculated as follows: Tumor volume =ab²/2. Both lungs of each mouse were removed, and paraffin blocks of 10% buffered formalin (Tissue Prep-II; Thermo Fisher Scientific, Inc.)-fixed samples of the lungs were prepared. A total of five coronal sections were selected from each paraffinized lung sample, with each coronal section cut consecutively into five slices. Serial sections were cut at 4 µm, and 20 slices in the group were randomly selected and stained with hematoxylin and eosin (Beijing Solarbio Science & Technology Co., Ltd., Beijing, China) to determine the presence of lung metastases under a light microscope (Leica DMIRBl; Leica Microsystems GmbH, Wetlzar, Germany). If at least one in these 20 slices was found to exhibit lung metastasis, the mice were confirmed to have lung metastasis. The degree of lung metastasis was graded by the number (N) of tumor cells counted in the maximum section of a solitary pulmonary metastatic nodule: Grade I, N<20; grade II, N=20–50; grade III, N=50–100; grade IV, N>100.

Paraffin-embedded tumor tissues were cut into 4-µm-thick sections, dewaxed in xylene (Sinopharm Chemical Reagent Co., Ltd., Shanghai, China) and dehydrated in ethanol. Two-step methods (EnVision™ system; DakoCytomation, Glostrup, Denmark) were used for CD105 staining. Briefly, the endogenous peroxidase activity was quenched with 3% H₂O₂ (Sinopharm Chemical Reagent Co., Ltd.) in methanol for 10 min. The CD105 antigen was unmasked by microwave oven (Panasonic 1380W; Panasonic Corporation, Osaka, Japan) pre-treatment in pH 9.0 ethylene diamine tetraacetic acid buffer (Invitrogen; Thermo Fisher Scientific, Inc.), and then cooled to room temperature. The sections were then incubated with the following primary monoclonal antibodies: mouse anti-endoglin anti-CD105 (SN6 h; DAKO; 1:50 dilution) for 2 h at 37°C, followed by incubation with rabbit anti-mouse horseradish peroxidase (HRP)-conjugated streptavidin (P0397; DAKO; 1:50 dilution) for 45 min at 37°C. Antibody binding was visualized with 3,3-diaminobenzidine (DAB; Sigma-Aldrich, St. Louis, MO, USA), and the sections were counterstained with Mayer's hematoxylin (Sgma-Aldrich). Prior to each subsequent step, the sections were washed extensively with phosphate-buffered saline (PBS) twice for 5 min. Tissues incubated with PBS as the primary antibody served as negative controls.

The quantification of MVD was performed in accordance with a previously described method (27). Briefly, the three most vascularized areas of the tumor tissues were initially identified under a low-power field (×40). The numbers of microvessels were then counted in each of these areas under a high-power field (×400) using a LEICA DMLB light microscope (Leica Microsystems GmbH). Any brown-stained single endothelial cells, or cell clusters with or without a discernible lumen and separated from adjacent microvessels and other connective tissue elements, were considered to be an individual, countable microvessel. The mean value of three ×400 field (0.40 mm²) counts was recorded as the MVD of the section. All microvessel counts were performed independently by two investigators, who were blinded to the clinicopathological data. The rate of disagreement between the two investigators' analyses was <10%.

Immunohistochemistry was performed using an EnVision™ method using the reagents supplied within the kit. Briefly, 4 µm-thick sections were cut consecutively from paraffin-embedded tumor tissue. Following deparaffinization and dehydration, the tissue sections were repaired for 40 min with 10% ethylenediamine tetraacetic acid (Invitrogen; Thermo Fisher Scientific, Inc.). The slides were then incubated with either primary VEGF monoclonal mouse antibody (clone VG1; M7273; DAKO; 1:100 dilution) for 60 min at 37°C, primary rabbit ant-mouse RGS-5 polyclonal antibody (SAB1411523; Sigma-Aldrich; 1:80 dilution) overnight at 4°C, or primary rabbit anti-mouse HIF-1α polyclonal antibody (sc-10790; Santa Cruz Biotechnology, Inc., Dallas, TX, USA; 1:200 dilution) overnight at 4°C. Following three rinses in PBS, the slides were incubated with the secondary antibodies (goat anti-mouse/rabbit unbiotinylated antibody-HRP; EnVision™ System; K5007; DAKO, 1:100 dilution) for 30 min at 37°C. The tissue staining was visualized with DAB substrate (DAKO), and the sections were counterstained with Mayer's hematoxylin. Antibody binding was visualized with DAB. Prior to each subsequent-step, the sections were washed extensively with 0.01 M (pH 7.4) triethanolamine-buffered saline (TBS; Wuhan Boster Biological Technology, Ltd., Wuhan, China). Tissues incubated with TBS as the primary antibody served as negative controls.

Positive staining was located in the cytoplasm for VEGF and RGS-5. Positive HIF-1α expression was located as brown staining, predominantly in the nucleus. A total of five areas were counted under ×400 field. The levels of expression were graded, as follows: Negative (–), <5% of the cancerous cells were positively stained. Weak positive (+), 5–25% of the cancerous cells were positively stained and the cytoplasm was light brown. Positive (++), 25–50% of the cancer cells were positively stained, and positive cells were light brown with a granule-like appearance in the cytoplasm. Strongly positive (+++), 50% of the cancer cells were positively stained and the cytoplasm was dark brown. The level of expression was represented by the grade: ≥5%=positive, <50%=low level expression, ≥50% high level expression. The percentage of positive tumor cells and the mean optical density (OD) values were calculated.

The data are presented as the mean ± standard deviation, and were analyzed using SPSS 17.0 software (SPSS, Inc., Chicago, IL, USA). One-way analysis of variance and Student's t-test were used for comparisons between groups. P<0.05 was considered to indicate a statistically significant difference.

At the point at which the mice were sacrificed, lumps were grossly visible. The tumors removed from the mice in the treatment groups and model control group at day 35 are shown in Fig. 1. The tumor weights and volumes in each group are presented in Table I. The results revealed that the changes in tumor weight (g) and tumor volume (mm³) in the LB35 group were smaller, compared with those in the LM35 group, but without statistical significance (P>0.05; Table I). No animals experienced weight loss of >10%.

Visible metastases were observed in all groups, as shown in Fig. 2A, the number of which were recorded. At 35 days post-treatment, the rate of metastasis to the lungs was 100% in all groups, however, the number of metastases in the lungs of the treatment groups were significantly lower, compared with the number in the LM group (P<0.05). In the LB group, the number of lung metastases was the lowest (P<0.05). The majority of the lung sections in the LM35 group were observed to contain scattered hemorrhagic spots. The degree of lung metastasis ranged between grade I and grade III, with 10% of metastases in the LB35 group being grade II and the others in the group being grade I. In the LH35 group, 40% were grade I and 60% were grade III. In the LY35 group, 20% were grade III and the others in the group were grade I.

A significant reduction in the number of stained regions were revealed following herbal medicine treatment. The expression levels of CD105 in the herbal medicine treatment groups were weak or even negative, whereas the expression of CD105 was more marked in the control group (Fig. 2B). Only the results at 35 day post-treatment have been presented, as the results on days 21 and 28 were similar to those at day 35. The newborn endothelial cells were stained brown or yellow, and were sinusoidally distributed in the capillary walls of the fiber interval and portal area of the liver tissues (Fig. 2B). In the 35 day groups, microvessel counting revealed that the MVD counts in the LB, LY, LH and LM groups at a high-power field (×400) were 5.60±3.02, 6.08±1.42, 8.07±2.65 and 12.00±4.45, respectively. The MVD in the LM group was higher, compared with those in the medicine treatment groups, and the lowest MVD was found in the LB35 group (P<0.05). These results showed that BYHWD treatment caused a marginal decrease in MVD.

To further determine whether BYHWD treatment inhibits the angiogenesis of HCC by inhibiting VEGF or other factors, the present study measured the expression levels of VEGF, RGS-5 and HIF-1α. The expression values of VEGF (positive cell rate × OD) are shown in Fig. 3B. The expression values in the LB35, LY35, LH35 and LM35 groups were 285.16±75.80, 228.12±81.45, 231.12±81.84 and 200.00±80.96, respectively. The results showed that the expression of VEGF was higher in the LB35 group, compared with LM35 group (P<0.05), and the IHC results revealed a significant increase in the expression levels of VEGF in the herbal medicine treatment groups, compared with LM group (Fig. 3A). The expression levels of VEGF were also significantly increased in the LB28 group, compared with the LM28 group (P<0.05; data not shown).

The IHC results and the expression values of RGS-5 are presented in Fig. 4 and Table II. Statistical analysis indicated that the expression levels of RGS-5 in herbal medicine treatment groups were significantly different from that in the model control group. The results showed that the expression value of RGS-5 on day 28 following treatment in the LB, LY, LH and LM groups were 262.64±36.33, 280.75±32.46, 281.00±31.64 and 312.87±39.36, respectively. The expression values of RGS-5 in the LB35, LY35, LH35 and LM35 groups were 294.16±63.70, 303.66±30.02, 310.20±51.23 and 349.25±48.27, respectively. These results showed that the herbal medicine treatment groups had lower expression levels of RGS-5 on days 28 and 35, compared with the corresponding time points in the LM group (P<0.05). BYHWD treatment exerted a more significant inhibitory effect, compared with the other herbal medicines. However, no significant difference was found on day 21 in the medicine treatment groups, compared with the model group (P>0.05; Fig. 4; Table II).

The present study also examined the change in expression of HIF-1α following herbal medicine treatment, compared with the model control group (Fig. 5). The result showed that the expression of HIF-1α in the tumor tissue was downregulated by BYHWD and HXD treatment, compared with the LM group, at day 35 post-treatment (P<0.05). The expression values (positive cell rate × OD) are shown in Table III. The expression values in the LB35 and LH35 groups were 250.95±32.08 and 256.29±24.08, respectively. The expression levels of HIF-1α were lower in the medicine-treatment groups, compared with the model control groups at the corresponding time points. However, no significant differences were found, compared with the control on days 21 and 28 post-treatment (P>0.05). In addition, no significant difference between the LY group and LM group were identified.