Human pancreatic cancer cell lines PANC-1, AsPC-1, HPAF-II, BxPC-3 and MiaPaCa-2 were obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA). Immortalized human lung epithelial cells MRC-5 were provided by Dr Sittampalam at the University of Kansas Medical Center, and were used as a comparison to cancer cells. All the cells were cultured at 37°C in 5% CO₂/95% air in recommended growth media containing 10% fetal calf serum. Pao extract was provided by Natural Source International (New York, NY, USA). Pao and Gem (Sigma, St. Louis, MO, USA) were prepared in sterile water and stock at −20°C.

Cells were assessed for viability by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay at 72 h of treatment. Cells in exponential growth phase were exposed to serial dilutions of Pao, Gem, or the combination of the two, for 72 h. Then, cells were changed into fresh media containing MTT and were incubated for 4 h. The colorimetric MTT assay assessed relative proliferation, based on the ability of living, but not dead cells, to reduce MTT to formazan (15,16). Cells did not reach plateau phase during the incubation period. Fifty percent inhibitory concentration (IC₅₀) was defined as the concentration of drug that inhibited cell growth by 50% relative to the untreated control. Pilot experiments for each cell line were performed to optimize cell density and assay duration and to center drug dilution series approximately on the IC₅₀.

Anchorage-independent colony formation assay in soft agar was utilized to determine survival of tumorigenic cancer cells following the treatments. In 6-well plates, PANC-1 cells (5,000 cells/well) were seeded in the upper layer containing 0.5% agar, DMEM medium, and 10% fetal bovine serum (FBS), with or without 400 μg/ml Pao. The solid agar base (lower layer) contained 0.75% agar, DMEM medium, 10% FBS with or without 400 μg/ml Pao, respectively. The number of colonies was counted after 20 days.

Cells were exposed to various concentrations of Pao for 48 h. Cells were washed in PBS, resuspended in binding buffer, and subjected to FITC-conjugated Annexin V and propidium iodide (PI) staining according to the manufacturer’s protocol (BD Biosciences, San Jose, CA, USA). Cells were analyzed by flow cytometry. Annexin V positive cells and Annexin V-PI double positive cells were identified as early and late stage apoptotic cells, respectively. PI single-positive cells were identified as necrotic cells.

A total of 40 μg protein was loaded for SDS-polyacrylamide gel electrophoresis. Primary and secondary antibodies were from Cell Signaling Technology Inc. (Danvers, MA, USA): rabbit anti-poly-(ADP-ribose)-polymerase (PARP) (1:2,000), rabbit anti-caspase-3 (1:1,000), rabbit anti-capase-8 (1:1,000), mouse anti-β-actin (1:1,000), and goat anti-rabbit or anti-mouse IgG (1:5,000). Blots were developed using immobilon chemiluminescent substrate (Thermo Fisher Scientific, Waltham, MA, USA).

Animal experiments were conducted following a protocol (#2012–2035) approved by the Institutional Animal Care and Use Committee. Through an intra-pancreas surgical procedure, the human pancreatic ductal adenocarcinoma cells PANC-1 were orthotopically implanted into the pancreas of nude mice (3.2×10⁵/mice). Ten days after tumor cells were implanted, treatment began with intraperitoneal (i.p.) injection of Gem (20 mg/kg, every 4 days), Pao (20 or 50 mg/kg daily), the respective combination of Gem and Pao and saline as control. To allow in vivo imaging, PANC-1 cells were transfected with luciferase gene. Luciferin at 150 mg/kg was administered i.p. each time prior to imaging for the luminance of tumor cells. Mice were kept anesthetized with ~2.5% isoflurane. An IVIS in vivo imaging system (Caliper Life Sciences, Hopkinton, MA, USA) was used to scan the mice. The Living Image 4.1 software (Caliper Life Sciences) was used for analysis of the images. After 70 days of treatment, mice were euthanized. All tumor lesions in the peritoneal cavity were collected and weighed. Major organs such as liver, kidney and spleen were subjected to histological analysis for any damage due to potential drug toxicity.

MTT data were normalized to their corresponding untreated controls for each condition (drug and cell type) and were expressed as percentage viability. Dose-reduction index (DRI) values for Gem were calculated by the equation DRI_ICx=(D_Gem/D_Gem+Pao), where D_Gem is the dose of Gem alone required to produce an ICx level of cytotoxicity, and the divisor D_Gem+Pao is the dose of Gem needed to produce the same ICx level of cytotoxicity when it is combined with Pao (at a given molar ratio). DRI_Gem is defined with respect to Gem. Combination index (CI) values were calculated by the equation CI_ICx = (D_{GemCombo ICx}/D_{Gem ICx}) +(D_{PaoCombo ICx}/D_{Pao ICx})+α[(D_GemComboICx)(D_PaoComboICx)/(D_{Gem ICx})(D_{Pao ICx})], where D is the dose of Gem and Pao either alone or in combination at a given constant ratio required to produce an ICx level of cytotoxicity (17–19). The more conservative assumption of mutual exclusivity was adopted (α=0). SPSS15.0 was used for additional statistical analysis.

Human pancreatic cancer cell lines (PANC-1, AsPC-1, HPAF-II, MiaPaCa-2 and BxPC-3) were compared to an immortalized epithelial non-tumorigenic cell line (MRC-5) for sensitivity to Pao. All the cancer cells were susceptible to Pao treatment with IC₅₀ ranging from 75 to 215 μg/ml. MRC-5 cells were less sensitive to the same treatment with an IC₅₀ of 547 μg/ml (Fig. 1A). At the concentration of 400 μg/ml Pao selectively killed 85–100% of cancer cells, and only decreased viability in MRC-5 cells for 25%. Colony formation in soft agar was used to assess the survival of tumorigenic cancer cells, which has been positively correlated to in vivo tumorigenicity of the cancer cells in animal models (20,21). Untreated PANC-1 cells formed colonies in soft agar at a rate of 12%. Pao at 400 μg/ml completely inhibited formation of colonies of PANC-1 cells in soft agar (Fig. 1B), indicating no survival of tumorigenic cancer cells with this treatment.

To assess Pao-induced death pathway, Annexin V/PI staining was performed to detect apoptosis vs. necrosis in PANC-1 cells treated with Pao. Flow cytometry demonstrated that the percentage of cells positive with Annexin V/PI staining increased from 4.3% in untreated cells to the highest, 96%, in 400 μg/ml Pao-treated cells (Fig. 2A). The induction of apoptosis was dependent on concentrations of Pao. Apoptosis contributed to the majority of cell death, while necrosis contributed to only 2–16.5% of cell death at all tested concentrations. As a consistent finding, cleavage of caspase-8, caspase-3 and PARP was also detected in Pao-treated PANC-1 cells. The cleavage was induced dependent on the concentration of Pao and time of treatment (Fig. 2B).

After determining the dose-response relationships to Pao in the cancer cells (Fig. 1A), the dose-response relationships to Gem were also established in PANC-1, AsPC-1, HPAF-II and BxPC-3 cells (Fig. 3A, dotted lines). A constant ratio design was used to systematically examine the combination effect of Gem and Pao (Gem+Pao). Ratio of Gem:Pao was chosen as IC_50Gem:IC_50Pao. Cell viability was examined after 72 h of treatment to obtain the optimum Gem effects. Results clearly showed that when Pao was added to Gem, drops in cell viability were markedly enhanced in all tested cells, compared to Gem as a single agent (Fig. 3A). The enhancement of the cytotoxic effect was particularly obvious in AsPC-1 cells which were more resistant to Gem treatment.

The combination index (CI) for each cell type was calculated based on the isobologram principle to examine whether drug combinations may be synergistic (CI<1), additive (CI=1), or antagonistic (CI>1) (17–19). In addition, DRI for Gem was calculated to indicate whether Gem concentration can be reduced when Pao was combined to produce the same cytotoxic effect. DRI values of >1 indicate a favorable combination. DRI<1 is interpreted as an antagonistic combination. In all cell lines tested, CI values were ≤1 and DRIs were >1 (Fig. 3B) across the desired levels of effect (fraction affected, f_a), indicating additive to synergistic effect when Pao was combined with Gem. These data also showed that the concentration of Gem is decreased to produce an equitoxic effect on pancreatic cancer cells when Pao is combined.

An orthotopic pancreatic cancer mouse model was used to evaluate the effect of Pao and Gem plus Pao (Gem+Pao) treatment. Compared to a subcutaneous xenograft model, this model better mimics clinical conditions of human pancreatic cancer, as the local environment for pancreatic cancer development was represented.

Treatment was carried out as described in Materials and methods. Tumor progress was monitored through longitudinal live animal imaging. Representative images in 5 mice from each group are shown in Fig. 4A. Gem at an early stage of the treatment showed tumor inhibitory effects (day 20, Fig. 4A), but it failed to inhibit the tumor at a later stage (day 69, Fig. 4A). At the end of experiment, Gem-treated mice had only insignificant reduction of tumor burden. By contrast, Pao at either 50 mg/kg or a lower dose of 20 mg/kg strongly inhibited tumor growth throughout the treatment period, and resulted in significantly reduced tumor burden at the end of experiment. Of note, the combination treatments mimicked the effect of Pao treatment alone, without showing additional effects with added Gem. These observations were confirmed by quantification of bioluminescence intensity in all images to show total tumor burden (Fig. 4B). Gem did not provide significant inhibition of the tumor at the end of the treatment. By contrast, a significant inhibition was shown between the control group and the Pao-presenting treatment groups. There was no difference whether Gem was present or absent in the Pao-treated groups. Both low (20 mg/kg) and high dose (50 mg/kg) of Pao provided similar effects. A dose-dependent effect to Pao was absent at the tested doses.

At necropsy, all tumors were weighed and metastatic lesions were counted. The results clearly confirmed data from live animal imaging (Fig. 5). Whereas Gem had minimal reduction on total tumor weight (P=0.25), Pao alone decreased tumor weight by 72% at 20 mg/kg/day and by 70% at 50 mg/kg/day (P<0.05 for both groups vs. control) (Fig. 5A). The Pao-treated groups had significantly lighter tumor burden compared to the Gem-treated group (P<0.05). By combining Pao and Gem, tumor weight decreased by 82% (Gem + 20 mg/kg Pao) and by 78% (Gem + 50 mg/kg Pao) relative to control. Notably, the enhancement was not significant compared to Pao single-drug treatment, but was highly significant compared to Gem single-drug treatment. Again, there was no dose-dependence to Pao, consistent with imaging.

Regarding metastasis, the effect mirrored what was described above. Gem did not inhibit the formation of metastatic lesions in mice. Pao at both tested doses markedly inhibited the number of metastatic lesions (Fig. 5B). Pao treatment also decreased the number of mice that developed metastasis disease (Fig. 5C), while Gem failed to do so. The combinations of Gem+Pao had better effect compared to Gem in reducing either number of mice having metastasis, or the number of metastatic lesions in the mice, but had no difference compared to Pao treatment alone. There was no difference between the different doses of Pao.

Proteins were isolated from tumor samples of the treated and control mice. Western blot analysis showed cleavage of caspase-3 and PARP in Pao and Gem+Pao treatment groups at all doses (Fig. 5D). These results confirmed the in vitro data that Pao induced apoptosis in tumor cells.

Pao treatment possessed low toxicity as no toxic effect was observed associated with the treatments. At the end of the experiments, major organs (kidney, liver and spleen) were subjected to haematoxylin and eosin (H&E) staining and histological analysis. No tissue damage was detected in any of the groups (Fig. 5E).