Growth inhibition of colon cancer cells by polyisoprenylated benzophenones is associated with induction of the endoplasmic reticulum response

Authors

  • Petr Protiva,

    Corresponding author
    1. Division of Gastroenterology, University of Connecticut Health Center, Farmington, CT
    2. Strang Cancer Prevention Laboratory, Rockefeller University, NY
    3. Herbert Irving Comprehensive Cancer Center, Columbia University, NY
    • Division of Gastroenterology and Colon Cancer Prevention Program, University of Connecticut Health Center, 263 Farmington Avenue, Farmington, CT 06030-2817
    Search for more papers by this author
  • Michael E. Hopkins,

    1. Strang Cancer Prevention Laboratory, Rockefeller University, NY
    Search for more papers by this author
  • Scott Baggett,

    1. Department of Biological Sciences, Lehman College and the Graduate Center, City University of New York, Bronx, NY
    Current affiliation:
    1. Bionovo, Inc., 5858 Horton Street, Emeryville, CA 94608
    Search for more papers by this author
  • Hui Yang,

    1. Department of Biological Sciences, Lehman College and the Graduate Center, City University of New York, Bronx, NY
    Current affiliation:
    1. Bionovo, Inc., 5858 Horton Street, Emeryville, CA 94608
    Search for more papers by this author
  • Martin Lipkin,

    1. Strang Cancer Prevention Laboratory, Rockefeller University, NY
    Search for more papers by this author
  • Peter R. Holt,

    1. Strang Cancer Prevention Laboratory, Rockefeller University, NY
    Search for more papers by this author
  • Edward J. Kennelly,

    1. Department of Biological Sciences, Lehman College and the Graduate Center, City University of New York, Bronx, NY
    Search for more papers by this author
  • Weinstein I. Bernard

    1. Herbert Irving Comprehensive Cancer Center, Columbia University, NY
    Search for more papers by this author

Abstract

Polyisoprenylated benzophenones derived from Garcinia xanthochymus have cytotoxic activity in vitro and antitumor activity in rodent models, but the mechanism is unknown. The purpose of our study was to examine in parallel molecular pathways that are targeted by 3 Garcinia-derived benzophenones-xanthochymol (X), guttiferone E (GE) and guttiferone H (GH), in 3 human colon cancer cell lines, HCT116, HT29 and SW480. The IC50 concentrations were determined and the cells were then treated with X, GE or GH at their respective IC50 or IC50x2 concentrations. Effects on the cell cycle, mitochondrial membrane potential and apoptosis were assessed by flow cytometry and caspase activation. Changes in gene expression were assessed with Illumina 24 K gene arrays. We found that X, GE and GH induced loss of mitochondrial membrane potential and G1 arrest at their IC50 concentrations and induced caspase activation at IC50 × 2 concentrations. An analysis of the changes in gene expression revealed that with all 3 compounds and all 3 cell lines there was a marked increase in expression of several genes, including XBP1, ATF4 and DDIT3/CHOP, which are components of the endoplasmic reticulum stress response. The DDIT4/REDD1 gene, an inhibitor of the mTOR survival pathway, was also up-regulated. Therefore, X, GE and GH appear to inhibit the growth of human colon cancer cells, at least in part, by activating the endoplasmic reticulum stress response and inhibiting the mTOR cell survival pathway. These combined effects may contribute to the anticancer activity of these novel compounds. © 2008 Wiley-Liss, Inc.

Plant polyphenols are important candidates for cancer prevention and therapy. Thus some of these compounds are active against tumor cells in vitro and exhibit anti-tumor activity in animal models. However, there is only limited evidence for the efficacy of these compounds as anticancer agents in humans. The potential chemopreventive activity of green tea polyphenols is currently being investigated in clinical trials.1, 2 In addition, many of the available compounds have limited potency and their precise mechanisms of action are not known with certainty. Therefore, our research group has been engaged in identifying additional plant-derived compounds that display anticancer activity and analyzing their mechanisms of action.3–5

It was recently reported that fruits from the tropical plant Garcinia xanthochymus, also known as gamboge, contains 3 polyisoprenylated benzophenones-xanthochymol, guttiferone E and a novel compound guttiferone H, that inhibit the growth of human colon cancer cells in vitro at micromolar concentrations.3 Little is known about the use of gamboge in medicine. However, structurally similar polyisoprenylated benzophenones previously isolated from related plant species have been shown to exhibit important biological activities including growth inhibition and induction of apoptosis in cancer cells,6, 7 antibacterial activities8, 9 and chemopreventive activities in azoxymethane-induced aberrant crypt foci in the colon of rats10 and 4-nitroquinoline 1-oxide-induced tongue carcinogenesis in rats.11

The mechanisms by which the polyisoprenylated benzophenones inhibit tumor cell growth are poorly understood. Recent studies suggest that these compounds cause a reduction in cell invasion and survival through inhibition of downstream signaling mediated by the protein kinase FAK.12 In addition, their anti-inflammatory and anticarcinogenic actions may result from their effects on arachidonic acid metabolism and nitric oxide synthesis.13 A recent study found that treatment of HeLa cells with garcinol results in posttranslational modifications of chromatin.14 Data from animal models suggest that these compounds may also induce phase II enzymes that are involved in the detoxification of many known carcinogens.10

In our study, we carried out mechanistic studies on 3 active polyisoprenylated benzophenones in human colon cancer cell lines. Our studies focused primarily on drug-induced changes in the cell cycle, apoptosis and mitochondrial membrane potential. To identify specific signaling pathways that are affected by these compounds, gene expression signatures were generated with the use of commercial 24 K microarrays, and the data were analyzed using Genespring GX software.

Material and methods

Reagents and cell culture

HT29, HCT116 and SW480 colon cancer cell lines were purchased from American Type Culture Collection and cultured in DMEM with 10% heat inactivated fetal bovine serum and penicillin 100 U/ml and streptomycin 100 mcg/ml (Invitrogen, Carlsbad, CA).

Polyisoprenylated benzophenones

Guttiferone H, guttiferone E, and xanthochymol were isolated and purified from Garcinia xanthochymus fruit as described previously.3 The fruits were collected in Fruit & Spice Park, Homestead, FL at peak ripeness. All 3 compounds were purified to ≥99% purity and verified by HPLC. These 3 benzophenones were dissolved in DMSO to form a stock solution which was added directly to the culture media.

Cell viability assay

Three thousand cells were plated in 96-well flat-bottom opaque white plates in 100 μl of medium. Cells were allowed to attach for 24 hr prior to drug exposure. The concentrations of benzophenones used to determine cell viability ranged between 0 and 50 μM. Cell viability was assessed 48 hr after benzophenone treatment using the ATP-based chemiluminescent Cell Titer Glo Assay (Promega, Madison, WI) carried out according to manufacturer's instructions. The relative luminescence was measured using a 96-well luminescence plate reader SpectraMax M5 (Molecular Devices Corporation, Sunnyvale, CA). Octuplicate wells were assayed for each condition, and IC50 values were calculated using a polynomial regression formula.

Cell cycle analysis

All flow cytometry analyses were performed on FACSCalibur system equipped with a 488-nm wavelength laser (Beckon Dickinson, San Jose, CA). Collections for cell cycle studies were performed as described previously.15 Briefly, floating and adherent cells were collected by trypsinization, washed with ice cold PBS and permeabilized with 70% ethanol chilled to −20°C. Cells were stored at −20°C in 70% ethanol. Before analysis, cells were washed with PBS and resuspended in 250 μl of staining solution, containing 0.05 mg/ml propidium iodide and 1 mg/ml RNAse A in PBS (Sigma Chemical.). The cells in suspension were then analyzed on a flow cytometer. At least 10,000 cells were gated for analysis. The analyses were performed using FlowJo software v.7.1.0 (Tree Star, Ashland, OR). Data were plotted as FL2A histograms and sub-diploid cells, G1, S and G2/M peaks were quantified.

Mitochondrial membrane potential analysis

Mitochondrial membrane potential was assessed by flow cytometry using the JC-1 cationic dye (Invitrogen, Carlsbad, CA) that indicates mitochondrial polarization by reversibly shifting its fluorescence emission from green to red.16 Cells with high membrane potential promote the formation of dye aggregates that fluoresce red; cells with low potential will contain monomeric JC-1 and fluoresce green. Therefore, the red/green ratio can help to identify a mitochondrial functional status. The JC-1 dye was dissolved in DMSO as ×1,000 concentrated stock solutions. Both floating and adherent cells were collected by trypsinization, washed with cell culture media, resuspended in media containing 5 μg/ml of JC-1 and incubated at 37 C for 20 min. Cells were subsequently washed once in dye free media and analyzed immediately by flow cytometry. The changes in mitochondrial potential were assessed as relative changes in FL1/FL2 fluorescence.

Caspase activation assay

The activities of the terminal effector caspases (caspase 3/7), as well as caspases 8 or 9, were assessed after treatment of the cells with each of the 3 Garcinia benzophenones at various time points at IC50 concentrations using chemiluminescent Caspase Glo assays by Promega. Assays were carried out according to manufacturer's instructions. Additionally, the activities of the terminal caspases were also assessed at IC50 × 2 concentrations at 8, 24 or 48 hr. The relative luminescence was measured on a 96-well luminescence plate reader SpectraMax M5. Octuplicate wells were assayed for each condition.

Dichlorofluorescein assay for reactive oxygen intermediates

The detection of the presence of intracellular ROI is based on the ability of cells to oxidize the fluorogenic dye, CM-H2DCFDA (Molecular Probes, Eugene, OR), to its corresponding fluorescent analogue, dichlorofluorescein, that can be detected by fluorometric methods. We used a 96-well fluorometer for detection of ROI. Cells were seeded in white opaque 96-well plates at 6,000 cells/well and allowed to attach for 24 hr. Cells were then incubated with 10 μM CM-H2DCFDA for 30 min, washed thrice with dye free media and then incubated with diamide or the Garcinia benzophenones, with or without catalase (100 U/ml) for 45 min. The ROI levels were measured by using a fluorescence plate reader at an excitation wavelength of 485 nm and emission wavelength of 530 nm.

Gene arrays

For microarray analyses, 0.5 μg of total RNA was used to make biotin-labeled cRNA using the Ambion Illumina cRNA amplification and labeling kit according to manufacturers' instructions (Ambion, Austin, TX). The cRNA quality was verified using the Agilent 2100 Bioanalyzer (Palo Alto, CA) prior to hybridization. Biotin-labeled cRNA was labeled with the fluorescent dye in the Rockefeller University Gene Array Facility and hybridized onto a Sentrix HumanRef-8 24 K Expression Array Bead Chip (Illumina, San Diego, CA). The arrays were then scanned by the Illumina Bead Station laser scanning imaging system; an average of 30 beads/gene transcript was used to generate the expression data. The data were analyzed using Genespring software from Agilent after normalization.

Gene array analyses

The hierarchical clustering algorithm used to generate the dendrogram is based on the complete-linkage method.17 The distance between 2 individual samples is calculated by Pearson distance with the normalized expression values. To determine if treatment by Garcinia benzophenones has a significant effect on gene expression behavior across any of the groups under study, we used one-way analysis of variance (ANOVA) tests using biological duplicate samples or the Cross-Gene Error Model as a way to estimate measurement precision by combining variability of gene expression data. We applied conservative Bonferroni multiple testing correction with Tukey post-hoc testing to detect the differences in gene expression between specific treatment groups. The list of genes that were differentially expressed following the treatments by Garcinia benzophenones was subjected to gene ontology analysis using GO/GO SLIMS built into the Genespring software package v 7.3 (www.geneontology.org). The gene expression correlations were computed using Pearson correlation test. The statistical significance of overlap between gene groups was calculated using the standard Fisher's exact test with the p-value adjusted with Bonferroni multiple testing correction or hypergeometric probability. p-values less than 0.05 were considered statistically significant.

Statistical analysis

Results are expressed as mean ± SD unless stated otherwise. Differences between group means were tested by a 2-sided t test or nonparametric Wilcoxon's Rank Sum Test. Correlations were tested by linear regression analysis. Statistical significance is defined as a p-value <0.05.

Results

Effects of polyisoprenylated benzophenones on growth inhibition and cell cycle progression

HCT116, HT29 and SW480 cells were treated with each of the 3 Garcinia benzophenones (0–50 μM), and cell viability was assessed at 48 hr. Viability was plotted against concentration and the IC50 was calculated from the linear regression curve. The IC50 concentrations with individual confidence curve intervals found for the 3 benzophenones are shown in Figure 1. HCT116, HT29 and SW480 cells were then treated with benzophenones for 24 or 48 hr at the IC50 concentrations, and cells were collected and analyzed by flow cytometry for DNA content. Figure 2 shows the results of the cell cycle analyses. The cells treated with benzophenones showed an increase in percent of cells in the G1 phase accompanied by a decrease of cells in the S and G2 phases. There was no increase in the sub-G1 population at the IC50 concentration suggesting that the reduction in viability seen is primarily because of growth arrest, rather than apoptosis.

Figure 1.

Top part: chemical structures of Garcinia benzophenones. Middle part: IC50 values with intervals corresponding to individual confidence curves for Garcinia benzophenones in 3 colon cancer cell lines studied. Individual confidence curve interval displays the upper and lower 95% confidence limits for an individual predicted value. The confidence limits reflect variation in the error and variation in the parameter estimates. Cell viability was assessed 48 hr after benzophenone treatment using the ATP-based chemiluminescent assay. Bottom part: Cell viability curves with individual confidence curve intervals are shown for Guttiferone H in 3 colon cancer cell lines. Y-axis corresponds to % viability in untreated controls.

Figure 2.

Changes in cell cycle distribution following the treatment of HCT116, HT29 or SW480 colon cancer cells with Garcinia benzophenones for 24 or 48 hr at IC50 concentrations. Data is presented as mean % of cells in G1, S or G2/M phase ± SD. Note that all 3 Garcinia benzophenones induced the accumulation of cells in the G1 phase accompanied by a decrease of cells in S phase following the treatment by benzophenones. The lower part of the figure shows the actual changes in cell cycle histogram following the treatment of SW480 colon cancer cells with the 3 Garcinia benzophenones at IC50 concentration for 48 hr (GH, guttiferone H; GE, guttiferone E; X, xanthochymol).

Effects of Garcinia benzophenones on mitochondrial membrane potential

Loss of mitochondrial potential is an early marker of cellular stress and subsequent cell death. To investigate whether the Garcinia benzophenones lead to loss of mitochondrial membrane potential, HCT116, HT29 or SW480 cells were treated with each of the 3 polyisoprenylated benzophenones at IC50 or IC50 × 2 for 24 hr. Mitochondrial potential was assessed by flow cytometry using the dye JC-1. As shown in Figure 3, there was a dose-dependent loss of mitochondrial membrane potential in HCT116 cells treated with Garcinia benzophenones at 24 hr. Identical results were observed in HT29 and SW480 cells (data not shown). Interestingly, some loss of mitochondrial potential was also clearly apparent at an IC50 concentration, which was not associated with increased sub-G1 phase population (Fig. 2) of cells or caspase activation (Fig. 4, upper part).

Figure 3.

Changes in mitochondrial membrane potential in HCT116 colon cancer cells following treatment with 3 Garcinia benzophenones at IC50 or IC50 × 2 concentrations for 24 hr. The mitochondrial potential was measured by flow cytometry using the JC-1 dye. The loss of mitochondrial potential is detected as a relative loss of red (FL2) fluorescence and gain in green fluorescence (FL1) compared to controls (upward and leftward shift in fluorescence in the figures). Note that there is a significant and dose-dependent loss of mitochondrial potential following the treatment with each of the 3 benzophenones.

Figure 4.

Upper part: Caspase 3 and 7 activation measured by a highly sensitive chemiluminescent assay in HCT116 or HT29 colon cancer cells after treatment with each of the 3 Garcinia benzophenones at IC50 or IC50 × 2 concentrations for 8 or 24 hr. The data are presented as percent of control activity ± coefficient of variation. Significant increases in caspase 3/7 activation were observed in HCT116 cells after treatment with benzophenones at IC50 × 2 concentrations for 24 hr. Treatment with benzophenones at IC50 concentrations was not associated with caspase 3/7 activation. In HT29 cells, treatment with xanthochymol was not associated with caspase 3/7 activation at IC50 × 2. Lower part: Effects of Garcinia benzophenones on reactive oxygen intermediates in HCT116 colon cancer cells after treatment with IC50 × 2 or IC50 × 4 concentrations. An increase in fluorescence indicates an increase in formation of reactive oxygen intermediates. As expected, there was an increase in fluorescence following treatment with the positive control compound diamide. There was no increase in fluorescence following a 45-min treatment with each of the 3 Garcinia benzophenones. Moreover, the addition of catalase (100 U/ml) 5 min prior to incubation with benzophenones did not result in significant changes in fluorescence when compared to benzophenone treatment alone. GE, guttiferone E; GH, guttiferone H; X, xanthochymol.

Effect of Garcinia benzophenones on caspase 3/7 activation

To determine whether the caspase-dependent apoptotic pathway is associated with the loss of mitochondrial membrane potential detected by flow cytometry, terminal caspase activation was assessed (caspases 3 and 7) in HCT116 and HT29 cells at 2 concentrations (IC50 and IC50 × 2) at 8 and 24 hr after treatment. There was no significant activation of caspases 3 or 7 following treatment with either xanthochymol, guttiferone E or guttiferone H at the IC50 concentration. However, at a concentration of IC50 × 2, significant activation of caspases 3 and 7 occurred at 24 hr but not at 8 hr (Fig. 4, upper part). Since all 3 compounds induced loss of mitochondrial membrane potential at the IC50 concentrations, the activation of caspases 8 and 9 was assessed at 3, 6, 12 and 24 hr. No significant activation of caspases 8 or 9 by Garcinia benzophenones at this concentration occurred in HCT116 or HT29 cells (data not shown).

Effect of Garcinia benzophenones on formation of reactive oxygen intermediates

To investigate whether Garcinia benzophenones exert a pro-oxidative effect, the presence of ROI in HCT116 or HT29 colon cancer cells was measured after exposure of the cells to each of the 3 benzophenones. In this study, the cells were pretreated with the compound CM-H2DCFDA that produces a fluorogenic product upon oxidation. Diamide, a compound known to increase the production of ROI in cell culture was included as a positive control. As shown in Figure 4 (lower part), there was an increase in fluorescence in HCT116 cells following treatment with diamide. However, there was no increase in fluorescence following a 45-min treatment of HCT116 cells with the Garcinia benzophenones at either 2 × IC50 or 4 × IC50 concentrations. Moreover, the addition of catalase 5 min prior to incubation with benzophenones did not result in significant changes in fluorescence when compared to benzophenone treatment alone. Identical results were obtained with HT29 cells (data not shown). Therefore, cytotoxicity of these compounds does not appear to be due to induction of ROI.

Effects of Garcinia benzophenones on gene expression in colon cancer cells

Garcinia benzophenones activate expression of a set of genes involved in the endoplasmic reticulum stress response

To further investigate the mechanisms underlying growth inhibitory effects of the Garcinia benzophenones on colon cancer cells HCT116 or HT29 cells were treated with each of the 3 Garcinia benzophenones at an IC50 concentration for 18, 24 or 48 hr. Duplicate samples for each compound and time point, and their respective controls, were collected for gene array analyses using the Illumina 24,000 RefSeq genome-wide array platform.

First we performed an unsupervised clustering analysis of the HCT116 and HT29 samples from the 18, 24 and 48 hr experiments using a distance similarity measure and complete linkage. Figure 5 shows that the array samples clustered into 3 main groups in both the HCT116 and HT29 cells. Surprisingly, the 3 main groups separated according to treatment times, suggesting that Garcinia compounds induced only small changes in overall gene expression. The individual compounds did not cluster in any particular order within the 3 groups, suggesting that these structurally similar compounds exhibit similar effects on gene expression.

Figure 5.

Hierarchical gene tree clustering using Pearson correlation. The conditions that exhibit similar gene expression profile cluster together. The distance (D) between the branches of the gene tree is proportional to the strength of correlation between gene expression profiles; a lower the distance indicates a higher correlation. HCT116 or HT29 colon cancer cells were treated with xanthochymol (X), guttiferone E (GE) or guttiferone H (GH) for 18, 24 or 48 hr. Treated samples were collected with their respective controls. True biological duplicates for each condition were used to generate gene array data. Note that both HCT116 cells (left part of the figure) and HT29 cells (right part of the figure) cluster in 3 distinct groups following the benzophenone treatment. Each of these 3 groups corresponds to treatment time and benzophenone-treated samples cluster with their respective controls. The actual compounds do not cluster in any particular order within the 3 groups suggesting that these structurally similar compounds exhibit similar effects on gene expression.

To examine whether exposure to Garcinia benzophenones significantly altered gene expression across the treatment groups, we used ANOVA with Bonferroni correction for mutiple observations. To detect significant differences between treatment groups and their respective controls we used Tukey post hoc testing. In HCT116 cells, there were a total 346 genes with statistically significant differences in levels of expression for at least 1 treatment group when compared to controls, and in HT29 cells, there were 552 genes with statistically significant differences in at least 1 treatment group compared to controls. Gene Ontology analysis of the genes that displayed significant changes in their expression after treatment with Garcinia benzophenones revealed that these genes are often involved in cell cycle arrest, response to stress, the unfolded protein response, DNA damage, and/or metabolism. Because at the IC50 concentrations the benzophenones exhibited the same effects on cell cycle, apoptosis and mitochondrial potential (Figs. 2, 3 and 4) we analyzed an intersection of significant genes from post hoc testing. In both HT29 and HCT116 cells there were 58 genes whose expression was significantly altered after treatment with Garcinia benzophenones. The GO analysis revealed that these genes play important roles in protein folding, regulation of progression through cell cycle and cell cycle arrest. The most significantly enriched GO category was protein folding (p < 0.001). Interestingly, a mammalian unfolded protein induced transcription factor XBP1 was significantly upregulated in both HCT116 and HT29 cells, suggesting that the unfolded protein response and/or the endoplasmic stress response pathway play important roles in the growth inhibitory effect of benzophenones in colon cancer cells.

There were 18 genes whose expressions were significantly changed at 18, 24 and 48 hr in HCT116 or HT29 colon cancer cells after treatment with each of the 3 Garcinia benzophenones (Supplemental Table A). Several of these genes are directly involved in the endoplasmic reticulum response or nutrient deprivation stress, and one of the most highly induced genes, DDIT4/REDD1, is involved in the cellular stress response. Thus, 2 principle pathways emerged as possible targets for these 3 compounds: (i) the endoplasmic reticulum stress response and (ii) the TSC/mTOR pathway, since DDIT4/REDD1, a critical transducer of cellular stress response, inhibits the TSC/mTOR pathway.18, 19 It is known that modulation of these pathways can lead to G1 arrest and apoptosis. We observed all of these changes following the treatment of both HCT116 and HT29 colon cancer cells with each of the 3 Garcinia benzophenones.

Verification of endoplasmic reticulum stress activation by Garcinia benzophenones in SW480 colon cancer cell line

The mammalian endoplasmic reticulum stress response consists of 3 primary sensors/transducers; the IRE1, PERK and ATF6, that upon activation result in reduction in protein translational rate and the induction of 2 transcription factors XBP1 and ATF4.20 These factors, in turn, activate a set of genes involved in cell cycle arrest, apoptosis and protein folding and degradation.20 As mentioned earlier, DDIT4/REDD1 is a stress response gene that was identified as an essential regulator of mTOR activity through the TSC1/2 complex, in both Drosophila and mammalian cells. To investigate if these pathways are indeed involved in the effects of Garcinia benzophenones on colon cancer cells, we performed a validation experiment investigating the effects of Garcinia benzophenones on gene expression in SW480 colon cancer cells at 4 and 12 hr. The results were evaluated by One-Way ANOVA or t test using the Cross-gene error model.

We further tested whether 2 known transducers of the endoplasmic reticulum stress response, ATF4 and XBP1, and their downstream targets were induced significantly by benzophenone treatment. Overall, there were 43 genes whose expression was significantly increased upon treatment of SW480 cells with each of the 3 benzophenones for either 4 or 12 hr (Supplemental Table B). GO analysis revealed that these genes are primarily involved in cell cycle arrest, cell death, cell proliferation, response to stimulus, response to stress and apoptosis. Indeed, XBP1 was significantly induced; ATF4 was also increased, but this increase did not reach statistical significance. To confirm that there was activation of the endoplasmic reticulum stress pathway in SW480 cells, we found 16 genes that were described in the literature as downstream targets of the XBP1 or ATF4 transcription factors (Supplemental Table C), and we then examined if some of these genes were induced in SW480 cells by benzophenones. Indeed, out of 43 genes significantly induced in SW480 cells 5 genes overlapped with the downstream targets of XBP1/ATF4, a highly significant result (p < 0.001). Finally, we performed a Pearson correlation of XBP1 transcription factor expression in the study with the SW480 colon cancer cells. There were 50 genes with correlation coefficient ≥0.95. As expected, this gene list showed highly significant overlap with XBP1/ATF4 target gene list (p < 0.001).

The role of oxidative stress in Garcinia benzophenone-induced stress

The DDIT4/REDD1 stress response gene that was consistently upregulated in all of our target identification experiments with HCT116 and HT29 cells was also significantly upregulated in the target validation experiment with SW480 colon cancer cells. A well known function of this gene is its response to cellular oxidative and other stress factors. However, as shown in Figure 4, treatment with benzophenones did not induce reactive oxygen intermediates in the culture media. To further address the issue of oxidative stress, we carried out a simple gene array experiment in which HCT116 cells were treated with guttiferone H, in the presence or absence of catalase (100 U/ml, added to the cell culture media), for 4 hr. RNA was isolated and hybridized to Illumina arrays. We then tested if DDIT4/REDD1 gene upregulation was diminished in the presence of catalase. DDIT4 was upregulated 2.7 times in the absence of catalase and 3.3 times in the presence of catalase. Thus the addition of catalase to the culture media did not prevent the upregulation of DDIT4/REDD1. It is therefore unlikely that the effects of Garcinia benzophenones on gene expression are simply a consequence a pro-oxidative effect of Garcinia benzophenones in the culture media.

Garcinia benzophenones activate expression of a set of genes involved in the endoplasmic reticulum stress response in a dose-dependent fashion

We found that Garcinia benzophenones induced endoplasmic reticulum stress response and mitochondrial membrane potential loss at their IC50 concentrations in HCT116, HT29 and SW480 colon cancer cells. However, terminal caspase activation was not associated with the IC50 concentrations and was observed only with IC50 × 2 concentrations (Fig. 4). To investigate the effect of benzophenone concentration on gene expression we treated HCT116 cells with IC50/2, IC50 or IC50 × 2 for 4 hr, extracted RNA and performed gene arrays experiments. Figure 6 shows the results of this experiment. There was a dose-dependent upregulation of both the ATF4 and XBP1 transcription factor mRNAs accompanied by a dose-dependent upregulation of their respective downstream target mRNAs. Notably, expression of the DDIT3/CHOP gene exhibited a striking 5.1 times upregulation at IC50 × 2.

Figure 6.

Effect of increasing concentrations of guttiferone H on changes in mRNA expression of genes involved in endoplasmic reticulum stress response after 4 hr of treatment in HCT116 cells. Upper part of the figure shows the changes in mRNA of ATF4 and XBP1, 2 transcription factors associated with endoplasmic reticulum stress response. Lower part of the figure shows the changes in the downstream target mRNAs of these 2 transcription factors. Please note the dose-response effect on gene expression and marked upregulation of DDIT3/CHOP at IC50x2 concentration. Y-axis indicates fold upregulation compared to untreated controls.

Discussion

In previous studies, we isolated and purified 3 polyisoprenylated benzophenones from the edible tropical fruit Garcinia xanthochymus that exhibited growth inhibitory activity against colon cancer cells in vitro.3 All 3 benzophenones are structurally related; xanthochymol and guttiferone E were described previously3 and guttiferone H was first discovered by our group. Structurally similar benzophenones displayed chemopreventive activity in animal models of colorectal and tongue carcinogenesis10, 11 but the underlying mechanism of action of this group of compounds had not been previously elucidated.

In our study, we found that guttiferone H, guttiferone E and xanthochymol exhibited similar effects on growth inhibition, cell cycle progression and caspase activation in human colon cancer cells. Treatment with each compound arrested the colon cancer cell lines in the G1 phase and, at a high concentration, the 3 compounds caused caspase activation (Fig. 4). At a low concentration, the G1 arrest occurred with virtually no detectable caspase activation or increase in the sub-G1 fraction. However, increasing the concentration of the Garcinia benzophenones to IC50 × 2 values led to terminal caspase activation in HCT116 cells. In HT29 cells, treatment with GE or GH at a concentration of IC50x2 also significantly increased caspase activation, whereas treatment with X did not (Fig. 4). We also found that all 3 compounds interfered with the mitochondrial membrane potential in colon cancer cells at concentrations that do not cause apoptosis or activation of caspases 3, 7, 8 or 9 (Fig. 3). This observation implicated mitochondrial energy pathways as potential critical targets for these compounds. Several plant-derived polyphenolic compounds have antioxidant activity, but under certain circumstances such as the high oxygen tension present in cell culture systems, they can also induce oxidative stress.21 However, employing an assay for intracellular reactive oxygen species we found that these 3 benzophenones did not cause a detectable increase in oxidative stress (Fig. 4, lower section). Therefore it is unlikely that our compounds exhibit significant prooxidative properties in the tissue culture media. However, we could not exclude endogenous production of reactive oxygen species as a result of cellular energy depletion or nutritional deprivations.

Subsequently, a genome wide array analysis was used to analyze changes in gene expression that might explain the earlier-described biological effects of these compounds in colon cancer cells. The genes that were consistently differentially expressed after treatment of the cells with each of the 3 Garcinia benzophenones are involved in several pathways including cell cycle arrest, responses to cellular stress, the unfolded protein response and apoptosis. These changes in gene expression are consistent with the biological effects of the Garcinia benzophenones we observed in the colon cancer cell lines. Two principal stress pathways were activated by all 3 compounds in 3 colon cancer cell lines: (i) the endoplasmic reticulum stress pathway which was associated with induction of the transcription factors XBP1 and ATF4 and activation of downstream genes involved in cell cycle arrest and apoptosis, such as p21 and DDIT3/CHOP; and (ii) the cellular energy stress pathway, associated with induction of DDIT4/REDD1, which can cause growth arrest via downregulation of the mTOR pathway.19

These findings are consistent with the fact that the DDIT3/CHOP gene is downstream of ATF4 and possibly XBP1.22 In addition, about a third of the well known mammalian genes that are downstream of XBP1/ATF4 were also significantly upregulated by the benzophenones (Supplemental Table C). The DDIT3/CHOP pathway is an apoptotic component of the endoplasmic reticulum stress response and is known to be involved, at least in part, in caspase-dependent apoptosis. DDIT3/CHOP participates in growth arrest and activates apoptosis when cellular stress defenses are overwhelmed.22 The precise mechanism by which the effector caspases are activated in the DDIT3/CHOP apoptotic pathway are still not known but probably involve an apoptosome-dependent mechanism with functional APAF and caspase 9.23 The gene expression studies obtained at low and high doses of guttiferone H are consistent with the fact that we could only demonstrate a significant increase in apoptosis and effector caspases activation at a IC50 × 2 concentration of the benzophenones (Figs. 3 and 4). Indeed, the decrease in cell viability at IC50 was virtually entirely due to growth arrest and not apoptosis. This would explain why we did not observe a significant early activation of caspase 9 at the IC50 concentrations. These results are consistent with the activation of endoplasmic reticulum stress response genes that are downstream of XBP1/ATF4 transcription factors, most notably sharp dose-dependent induction of DDIT3/CHOP pro-apoptotic gene at IC50 × 2 concentration (Fig. 6). Others have also suggested that the magnitude of cellular stress correlates with DDIT3/CHOP expression and ultimately distinguishes adaptive stress response accompanied by translational attenuation and growth arrest from terminal apoptotic pathway activation,24 and our data support this finding. Many types of malignancies rely on intact unfolded protein response to survive unfavorable conditions such as lack of oxygen or nutrients. Treatment with Garcinia benzophenones may result in balance shift from adaptive response to apoptotic activation within these tumors.

The DDIT3/CHOP apoptotic pathway was recently identified as a target in lung cancer cells treated with a celecoxib/EGCG combination that synergistically induces apoptosis while neither of these compounds alone was able to activate this pathway.25 Additionally, the celecoxib/EGCG cotreatment did not induce the expression of other apoptosis related genes such as p21 or GADD45. It is of note that each of these benzophenones alone was able to activate numerous apoptotic and cell growth pathway genes, including DDIT3/CHOP, p21 and GADD45.

Another gene that was consistently up-regulated after Garcinia benzophenone treatment, in all 3 colon cancer cell lines, is DDIT4/REDD1. This gene is known to be a critical transducer of the cellular response to energy depletion through the TSC-mTOR pathway.19 This pathway plays a central role in growth and proliferation by integrating information about nutrient availability and growth factor stimuli.26 DDIT4/REDD1 acts downstream of AKT and upstream of TSC2 to inhibit the function of mTOR.18 This gene is also regulated by hypoxia or reactive oxygen species.27 However, it is unlikely that the Garcinia benzophenones induce this stress response by a direct prooxidant effect because catalase did not abolish the ability of guttiferone H to induce DDIT4/REDD1. Nevertheless, we did observe a perturbation of mitochondrial membrane potential by benzophenones which occurred at a relatively low concentration of these compounds i.e., their IC50 values (Fig. 3). The resulting cellular stress may explain the consistent up-regulation of DDIT4/REDD1 and the upregulation of endoplasmic stress genes seen in our study. The impressive G1 arrest may be due to the synergistic effect of our compounds on these 2 critical pathways that control cellular growth and stress responses.

We believe that the changes in gene expression produced by the Garcinia compounds are highly significant for several reasons. The Illumina gene array technology uses an average of 30 replicates per transcript within a single gene array and we used biological duplicates. Furthermore, identical results were generated in 3 different cell lines at 5 different experimental time points. Using our gene array results in conjunction with gene ontology analysis and analyses of overlaps with target gene lists generated by careful review of the literature we obtained identical results for each of our experiments in all 3 colon cancer cell lines. Recent advances and discoveries in the area of endoplasmic reticulum stress pathway are congruent with our gene array results.24

A recent study found that when HeLa cells were exposed to garcinol, a structurally related benzophenone, there was inhibition of histone acetylation.14 However, the concentration was quite high, 100 μM. We found that guttiferone E also inhibits histone acetylation in SW480 cells but only at concentrations above IC50 × 3. None of our 3 compounds was able to inhibit histone acetylation at concentrations below IC50 × 2 (unpublished studies). In our study, we obtained marked biological effects at much lower concentrations, so we believe that the reported effects on histone acetylation are probably not relevant to the growth inhibitory effects of these compounds.

Our finding that all 3 Garcinia polyisoprenylated benzophenones activated similar pathways in all 3 colon cancer cell lines, despite differences in the spectrum of mutations in these 3 cell lines suggests that these compounds may be useful for colon cancer prevention and therapy. Furthermore, the chemopreventive activity of benzophenones has already been demonstrated in rodent models.10 Future studies will determine which of these cellular targets of benzophenones is essential in chemoprevention or therapy of cancer and whether Garcinia benzophenones may act by modulating posttranslational modifications of proteins including interference with protein isoprenylation.

In summary, in our study we identified 3 structurally related polyisoprenylated benzophenones originally isolated from the edible tropical fruit Garcinia xanthochymus, xanthochymol, guttiferone E and guttiferone H, that induce G1 arrest of the cell cycle and terminal caspase activation in 3 human colon cancer cells lines. Our detailed analysis of changes in gene expression profiles induced by these compounds in all 3 colon cancer cell lines suggest that these biological effects are associated with interference with mitochondrial membrane potential and subsequent activation of the endoplasmic reticulum stress/apoptotic pathway via the XBP1, ATF4 and DDIT3/CHOP, and the cellular energy stress response gene DDIT4/REDD1 which can negatively regulate the activity of TSC/mTOR. These effects may explain similar cytotoxic effect of these 3 compounds.

Acknowledgements

This work was supported by The Irwing Weinstein Foundation, Inc. (to PP) and by the New Investigator Award of the Clinical Nutrition Research Unit, NIH (to PP).

Ancillary