The 95% ethanol extract of BU was provided by Dr. Arunporn Itharat from the Faculty of Medicine of Thammasat University (9). The BU formula consisted of aloe vera (2.48%), red physic nut (9.92%), kaffir lime (33.06%), asafoetida (2.48%), gamboge resin (4.96%), Javanese long pepper (2.48%), black pepper (2.48%), ginger (2.48%) and epsom salt (39.66%). Dried plant material (300 g) in a hot air oven at 50˚C was macerated in 95% ethanol for 3 days, filtered through a Whatman No. 1 filter paper and concentrated using an evaporator.

Liquid chromatography-mass spectrometry (LC-MS) was used to determine the compounds in the BU extract. The analysis was performed on an Agilent HPLC 1260 series consisting of a vacuum degasser, a binary pump, an autosampler and a column thermostat equipped with QTOF 6540 UHD accurate mass (Agilent Technologies GmbH). MassHunter Software B06.0 (Agilent Technologies GmbH) was used to control the LC-MS.

The separation of the sample solution was performed on a Luna C18, 150x4.6 mm, 5 µm column (Phenomenex, Torrance). A 10 µl sample of each filtrated extract at a concentration of 20 mg/ml was injected into the LC system with a solvent flow rate of 500 µl/min. The mobile phase consisted of a gradient elution between water (solvent A) and acetonitrile (solvent B), each containing 0.1% v/v formic acid. The linear gradient elution was 5-95% for solvent B starting at 0-35 min with holding for 5 min and post-run for 5 min. The column temperature was controlled at 35˚C. The mass analysis was performed using a QTOF 6540 UHD accurate mass spectrometer. The conditions for the negative electrospray ionization source were drying gas (N₂) at a flow rate of 10 l/min, a drying gas temperature of 350˚C, nebulizer 30 psi, fragmentor 100 V, capillary voltage 3,500 V and scan spectra from m/z 100-1,000 amu. The auto MS/MS for the fragmentation was set with collision energies of 10, 20 and 40 V. The positive mode was also set up with the same MS conditions as the negative mode.

All animal experiments were performed according to the Thai guidelines for the care and use of experimental animals, subsequent to being approved by the Animal Ethics Committee of the Faculty of Medicine, Srinakharinwirot University (Bangkok, Thailand; approval no. 3/2558). A total of 42 male Wistar rats (6-7 weeks old) weighing 200-250 g were obtained from the National Laboratory Animal Center of Mahidol University (Bangkok, Thailand). The animals were acclimatized for one week. All animals were maintained under standardized hygienic conditions throughout the experimental period, including a temperature of 21-22˚C, humidity at 55±5%, a standard 12 h light-dark regime and ad libitum access to standard diet and tap water.

The experimental protocol for HCC induction was based on El-Ashmawy et al (29). For the induction of HCC, 200 mg/kg diethylnitrosamine (DEN; Sigma-Aldrich; Merck KGaA) was injected intraperitoneally (i.p.) in a single dose. Following 14 days, the rats were subjected to i.p. injections of 300 mg/kg thioacetamide (TAA) (Sigma-Aldrich; Merck KGaA) 3 times weekly for 4 weeks. Then the rats were left for 2 further weeks without any treatment. At the end of the induction period (8 weeks), HCC rats were weighed and randomly divided into 6 groups: i) No treatment; ii) treatment with propylene glycol: Tween 80: deionized water (4:1:4), a solvent of BU; iii) treatment with 30 mg/kg Sorafenib (30-34); or treatment with iv) 1 mg/kg, v) 10 mg/kg or vi) 50 mg/kg BU. Doses of BU used in the present study were based on those previously used in vitro (9) and demonstrated to be safe in a toxicity test in rats (Intharit et al, preliminary study). During the time course of the experimental tumor study (16 weeks), a set of criteria was developed to follow the rats' condition. These included their external physical appearance, appearance of any visible lesions, changes in body weight and behavioral responses to external stimuli (including light or noise and so on), which reflect pain and distress in the animals. The humane endpoint in the present study was based on a weight loss exceeding 20% of the body weight of the rats in the control group. Subsequent to 16 weeks, experimental rats (n=7 per group) were anesthetized with an i.p. injection of 45 mg/kg pentobarbital sodium prior to sacrifice by decapitation with a rodent guillotine. Liver tissues and blood samples were collected for histological and immunohistochemical analyses, liver function tests, and reverse transcription-quantitative PCR (RT-qPCR) and western blot analyses.

At the end of the treatment period, the body weight of all animals along with their respective livers were measured in order to determine the liver-to-body-weight ratio in each group.

Blood samples, collected by cardiac puncture, were assayed in a standard clinical lab for serum ALT and albumin. The reagent kits for ALT (cat. no. 7D56-21) and albumin (cat. no. 7D53-23) were used (Abbott Pharmaceutical Co. Ltd.). The signals were detected by ARCHITEC model Ci16200 (Abbott Pharmaceutical Co. Ltd.).

Liver tissues were fixed overnight at 4˚C in 4% (v/v) formaldehyde solution, dehydrated in an ascending series of ethanol (50, 70, 80, 90, 95 and 100%), cleared in xylene and embedded in paraffin. Specimens were sliced into sections that were 5 µm thick. The slides were stained with hematoxylin for 6 min and eosin for 1 min at room temperature and scanned with a panoramic digital slide scanner (3DHISTECH Ltd.). For each image, an area of 4,000x2,500 µm (10 mm²) was randomly selected to locate and calculate the cancer area characterized histopathologically by the presence of thick-cell cords (35) using the CaseViewer software (v1.3.0.41885; https://www.3dhistech.com/caseviewer). A total of 21 images of each animal group were sampled, in which three specialists in liver histopathology identified the thick-cell cords and located the cancer areas. The mean cancer area in each group was then calculated.

The deparaffinized tissue sections were treated with xylene and rehydrated in a descending series of ethanol (100, 95, 90, 80, 70 and 50%), followed by antigen retrieval in 10 mM sodium citrate buffer (pH 6.0) for 10 min at 120˚C using an autoclave. The blocking solution TENG-T (10 mM Tris, 5 mM EDTA, 150 mM NaCl, 0.25% gelatin and 0.05% Tween 20; pH 8.0) containing 10% goat serum was applied for 30 min at room temperature to the slides to block any non-specific binding, and the slides were incubated with mouse anti-VEGF immunoglobulin G (IgG; cat. no. sc-53462; Santa Cruz Biotechnology, Inc., Dallas; 1:50) at 4˚C overnight. The slides were then incubated for 2 h at room temperature with alkaline phosphatase-conjugated goat anti-mouse IgG (cat. no. 11569520, Sigma-Aldrich; Merck KGaA; 1:100). Subsequent to washing in phosphate buffered saline (PBS), the sections were incubated in substrate containing nitroblue tetrazolium chloride/5-bromo-4-chloro-3-indolyl phosphate (toluidine salt; Dako; Agilent Technologies GmbH) diluted in 100 mM Tris (pH 9.5), 100 mM NaCl and 50 mM MgCl₂ at room temperature for 30-120 min to visualize the immunopositive areas in the tissues. The staining reaction was stopped by washing with distilled water. The sections were dehydrated with ethanol, cleared in xylene and covered with Permount^® prior to being examined and photographed under a light microscope (Olympus Corporation).

The livers were homogenized with a sonicator (Sonics & Materials, Inc.) followed by total RNA isolation using TRIzol reagent (Invitrogen; Thermo Fisher Scientific, Inc.). Total RNA with 2 µg of each RNA sample was reverse transcribed with the High Capacity cDNA Reverse Transcriptase kit (Applied Biosystems; Thermo Fisher Scientific, Inc.) according to the manufacturer's protocol (25˚C for 10 min, 37˚C for 120 min, 85˚C for 5 min and hold at 4˚C). RT-qPCR was set up using the SsoAdvanced^™ SYBR^® Green Supermix (Bio-Rad Laboratories, Inc.), along with cDNA and commercial PrimePCR^™ primers (Bio-Rad Laboratories, Inc.). The primer used was rat Vegfa (unique assay ID: qRnoCED0002159). The differences in sample RNA content were normalized to rat β-actin (Actb) expression (unique assay ID: qRnoCID0056984). The conditions of the reactions were as follows: 95˚C for 2 min of polymerase activation, 40 cycles of 95˚C for 5 sec for denaturation and 60˚C for 30 sec for primer annealing and extension. Each sample's mRNA expression was measured in triplicate to ensure the fidelity and accuracy of the results using a CFX96 Real-Time PCR Detection System and Bio-Rad manager^™ software version 1.3.1 (Bio-Rad Laboratories, Inc.). The quantification of relative mRNA expression was calculated using the 2^-∆∆Cq method (36).

A total of 50 mg frozen liver was homogenized in 500 µl radioimmunoprecipitation lysis buffer (Santa Cruz Biotechnology, Inc.,). The lysate was centrifuged at 4˚C and 12,000 x g for 15 min to collect the supernatant fraction. Protein concentration was determined using a Bradford protein assay (Bio-Rad Laboratories, Inc.). Total protein samples (40 µg) were separated on 12% SDS-polyacrylamide gels (Bio-Rad Laboratories, Inc.) and transferred onto a 0.2 µm polyvinylidene difluoride membrane. The membrane was blocked at room temperature for 1 h with 5% non-fat milk in 1X PBS-0.1%Tween 20 (PBS-T), then washed with 1X PBS-T, and incubated overnight at 4˚C with a primary mouse anti-VEGF antibody (cat. no. sc-53462; Santa Cruz Biotechnology; diluted 1:500). Subsequent to washing thrice in 1X PBS-T for 5 min, the membrane was incubated at room temperature for 90 min with horseradish peroxidase-conjugated goat anti-mouse secondary antibody (cat. no. 7076S lot 32; Santa Cruz Biotechnology; diluted 1:10,000) followed by 3x10 min washes with 1X PBS-T. Protein bands were developed using enhanced chemiluminescence (Bio-Rad Laboratories, Inc.) and quantified using densitometry with Scion Image software version Beta 4.0.3 (Meyer Instruments, Inc.). β-actin was used as a loading control.

Data were analyzed using a one-way analysis of variance with a Tukey's post-hoc test (PSPP 0.10.4; http://gnu.org). All results were represented as the mean ± standard deviation. P<0.05 was considered to indicate a statistically significant difference.

In the negative mode, 12 compounds were tentatively identified based on their molecular mass and fragmentation pattern (Fig. 1 and Table I). Aloin, aloesin and their derivatives together with dihydrogambogic acid and gambogic acid were indicated to be present in the extract. These compounds were identified in aloe vera and gamboge resin. In addition, five alkaloids from black pepper were identified in the positive mode (Fig. 2 and Table II). The chromatograms, mass spectra data and compounds proposal were presented in Figs. 1 and 2 and Tables I and II. This analysis confirms the previously published results for the original extract (9).

Fig. 3A presents the changes in the body weight in the respective groups of rats. Apart from the anticipated adverse effects of DEN administration at the beginning of the experiment and of thioacetamide during weeks 2-6, the experimental animals gained weight at the same rate as the control animals. To identify the potentially harmful general effect of the treatments on the liver, the liver/body weight ratio was determined (Fig. 3B). Rats treated with Sorafenib had a significantly reduced liver/body weight ratio compared with all other groups (P<0.05). There were no significant effects in the BU-treated groups when compared with the non-treated and vehicle-treated groups (Fig. 3).

Figs. 4 and 5 revealed the effects of BU on liver function tests (serum ALT activity and albumin content) in the experimental animals. The serum ALT level was significantly increased in the Sorafenib-treated group compared with all other groups (P<0.05; Fig. 4). On the other hand, the serum albumin concentration was significantly lower in the Sorafenib-treated group compared with all other groups (P<0.05; Fig. 5). The results demonstrate that Sorafenib causes liver injury.

Fig. 6 revealed that Sorafenib (Fig. 6C) and BU (Fig. 6D-F) suppressed nodule growth compared with untreated and vehicle-treated rats with HCC (Fig. 6A and B). Sections revealed the characteristic histopathological thick-cell cord changes in the HCC nodules of non-treated (Figs. 7A and 8, 22±9%) and vehicle-treated rats (Figs. 7B and 8, 18±4%). The percentage of the cancer area was lowest in the Sorafenib-treated group compared with all other groups (Figs. 7C and 8, 1.5±0.7%). The percentage of the cancer area in the BU-treated group decreased dose-dependently from 2.7±1.6% (Figs. 7D and 8; 1 mg), to 2.6±1.4% (Figs. 7E and 8; 10 mg) to 2.1±0.8% (Figs. 7F and 8; 50 mg), and was significantly reduced when compared with the vehicle-treated group (P<0.05). These data reveal that BU reduces cancer growth in vivo.

Next, the mechanisms by which BU exerts its antitumor effect in vivo were investigated. Immunohistochemical analysis revealed that the cytoplasmic VEGF concentration was markedly increased in cancerous areas (Fig. 9A; arrowheads) and that BU solvent alone did not change that result (Fig. 9B). In contrast, Sorafenib (Fig. 9C) and BU treatment prevented the formation of VEGF-positive cancer areas (Fig. 9D-F) in rats with HCC. In agreement with these results, VEGF mRNA expression was revealed to be significantly downregulated by Sorafenib compared with the control group (P<0.05) and an even stronger and dose-dependent downregulation by increasing doses of BU compared with the control groups (P<0.05; Fig. 10). Similarly, western blot analysis of the liver revealed that VEGF protein content was, compared with untreated and vehicle-treated rats with HCC, decreased significantly by Sorafenib treatment and treatment with the two highest doses (10 and 50 mg) of BU (P<0.05; Fig. 11). These results suggest that the anticancer activity of BU is mediated at least in part by the inhibition of VEGF expression in rats with HCC.