Although endocannabinoid signaling is important for certain aspects of gastrointestinal homeostasis, the role of the cannabinoid receptors (CB) in colorectal cancer has not been defined. Here we show that CB1 expression was silenced in human colorectal cancer due to methylation of the CB1 promoter. Our genetic and pharmacologic studies reveal that loss or inhibition of CB1 accelerated intestinal adenoma growth in ApcMin/+ mice whereas activation of CB1 attenuated intestinal tumor growth by inducing cell death via down-regulation of the antiapoptotic factor survivin. This down-regulation of survivin by CB1 is mediated by a cyclic AMP–dependent protein kinase A signaling pathway. These results indicate that the endogenous cannabinoid system may represent a potential therapeutic target for prevention or treatment of colorectal cancer. [Cancer Res 2008;68(15):6468–76]
Cannabinoids are currently used to treat chemotherapy- or radiotherapy-induced nausea and vomiting as well as for pain relief, insomnia relief, mood elevation, and appetite stimulation (1, 2). Medicinal use of these compounds is also thought to be beneficial for digestive disorders such as diarrhea and Crohn's disease (3). Plant-derived cannabinoids and their derivatives exert a wide variety of biological effects by mimicking endogenous compounds (endocannabinoids), which primarily activate two cannabinoid-specific G-protein–coupled receptors, CB1 and CB2, encoded by the Cnr1 and Cnr2 genes, respectively. Studies have recently suggested that cannabinoids exert potential antitumor effects on a wide spectrum of human tumor cell lines in culture and xenograft studies (4, 5) and have anti-inflammatory properties (6). CB1 and CB2 can activate several different cellular pathways depending on the biological context being examined (4, 7). However, there is currently no in vivo genetic evidence showing the role of the endocannabinoid system in mouse models of cancer.
The gastrointestinal tracts of mice, rats, and humans produce two major endocannabinoids, anandamide and 2-arachidonoyl-glycerol (8–10). Interestingly, human colorectal cancers (CRC) and adenomas produce twice to thrice more anandamide and 2-arachidonoyl-glycerol than does the neighboring normal mucosa (10). In humans and mice, CB1 is expressed in normal colonic epithelium, smooth muscle, and the submucosal myenteric plexus (11, 12). CB2 is found mainly in subepithelial macrophages of normal colonic tissues (11). However, little is known about the expression of cannabinoid receptors in CRC.
Endocannabinoid signaling is important in regulating gastrointestinal motility, secretion, and neurotransmitter release (8, 13–15). For example, the endocannabinoid signaling has been implicated in an autoimmune intestinal disorder (16, 17); specifically, Cnr1−/− mice exhibited a stronger inflammatory response in the colon than did wild-type mice in response to treatment with proinflammatory agents (18), suggesting that CB1 provides intrinsic protection against colonic inflammation. Because chronic inflammation is a known risk factor for CRC, the aim of the present study was to determine the role of cannabinoid receptors in a mouse model of colon cancer and the mechanism of the receptor action.
Materials and Methods
Cell culture and reagents. Ten CRC cell lines were obtained from the American Type Culture Collection, and HCA-7 cells were a gift from Susan Kirkland (University of London, London, United Kingdom). All cells were maintained in McCoy's 5A medium with 10% fetal bovine serum (FBS). AM251 and R-1 methanandamide (R-1) were purchased from Cayman Chemical. 5-Aza-2-deoxycytidine (5-aza-dC) and 8-(6-aminohexyl)amino cyclic AMP (8-AHA-cAMP) were obtained from Sigma. PD98059, LY294002, and H-89 were obtained from Calbiochem. A negative control small interfering RNA (siRNAC) and a CB1-selective siRNA (siRNACB1) were purchased from Ambion.
Animals. CB1- and CB2-deficient mice on a C57BL/6J genetic background were generated as previously described (19, 20). CB1-deficient ApcMin/+ mice (Cnr1−/−/ApcMin/+), CB2-deficient ApcMin/+ mice (Cnr2−/−/ApcMin/+), and their controls (Cnr1+/+/Cnr2+/+/ApcMin/+) were derived from same littermates by breeding Cnr1−/− or Cnr2−/− mice with ApcMin/+ mice on a C57BL/6J genetic background (The Jackson Laboratory). Male mice were killed at the age of 13 wk. For the AM251 treatment, 6-wk-old male Cnr1+/+/ApcMin/+ control mice (n = 28) with a genetic background identical to that of Cnr1−/−/ApcMin/+ mice were randomly assigned to one of two groups treated with 150 μL of 0.5% carboxymethylcellulose with or without AM251 (10 mg/kg body weight) by daily gavage for 7 wk. For the R-1 treatment, 32 male Cnr1+/+/ApcMin/+ control mice with a genetic background identical to that of Cnr1−/−/ApcMin/+ mice and 20 male Cnr1−/−/ApcMin/+ mice at the age of 6 wk were assigned to one of two groups and injected i.p. with 150 μL of PBS with or without R-1 (10 mg/kg body weight) daily for 8 wk. At the end of the experiments, mice were killed and the number and size of polyps were measured as previously described (21). After the tumors were counted, intestinal tissues were embedded in paraffin. For histologic analysis, 5-μm-thick sections from all groups were stained with H&E to examine polyp morphology.
Quantitative real-time PCR. Cnr1 and Cnr2 mRNAs were quantified by quantitative real-time PCR with an iCycler and iQ SYBR Green Supermix (both from Bio-Rad) as previously described (22). Primers for the Cnr1, Cnr2, and Actin genes were chosen by using the Beacon Designer 4 program (Premier BioSoft International). The sequences of the specific PCR primers were as follows (5′-3′): hCnr1, AAGACCCTGGTCCTGATCCT (forward) and CGCAGGTCCTTACTCCTCAG (reverse); hCnr2, ATCATGTGGGTCCTCTCAGC (forward) and GATTCCGGAAAAGAGGAAGG (reverse); hβ-actin, AGAAAATCTGGCACCACACC (forward) and AGAGGCGTACAGGGATAGCA (reverse).
5-Aza-dC treatment. The cells (1 × 106 per 100-mm plate) were cultured in McCoy's 5A medium containing 10% FBS and 5 μmol/L 5-aza-dC for 3 d and subjected to the assays described below.
Bisulfite genomic sequencing PCR. Sodium bisulfite modification was carried out with the CpGenome DNA Modification Kit (Chemicon) according to the manufacturer's instructions. The bisulfite sequencing PCR primers were designed with the Web software program MethPrimer.7
Transfection and retroviral transduction. SW-480 cells (1.8 × 105) were transfected with either a negative control siRNA (siRNAC) or a Cnr1-selective siRNA (siRNACB1) at 100 nmol/L by using LipofectAMINE 2000 reagent according to the manufacturer's instructions (Life Technologies, Inc.). Transfection efficiency was monitored with 100 nmol/L of a control fluorescence-labeled RNA (Invitrogen) at 24 h after transfection. Transfection efficiency was 90% with this system. Retrovirus transduction of human survivin in MIEG3 vector or vector alone into SW-480 cells was carried out as described (23). SW-480 cells stably expressing survivin were selected by fluorescence-activated cell sorting.
Apoptosis assays. The cells (9 × 105 per 100-mm plate) were incubated in serum-free medium for 1 d and then treated with either vehicle, R-1, or H-89 at the indicated concentrations in serum-free medium for 1 d. For the 8-AHA-cAMP treatment, the cells were pretreated with 8-AHA-cAMP for 1 h and then treated with R-1 in serum-free medium for 1 d after serum starvation for 24 h. For the 5-aza-dC treatment, the cells were treated with 5-aza-dC for 3 d as noted above, followed by incubation with R-1 in medium containing 1% FBS for 1 d. For the siRNA assays, the transfected cells were treated with R-1 in serum-free medium for 1 d. The numbers of apoptotic cells were determined by flow cytometry with the TACS Annexin V-FITC Apoptosis Detection Kit according to the manufacturer's instructions (R&D Systems). Data are expressed as the mean ± SE of the percentages of apoptotic cells from three separate experiments.
Membrane protein preparation. Plasma membranes were prepared according to the procedure described previously (24). CB1 protein levels in the plasma membrane were measured by Western blotting as described below.
Western blot analysis. Whole-cell extracts were prepared from cells (1.5 × 106) treated with the indicated concentration of R-1 or H-89 in serum-free medium for 1 d after serum starvation for 24 h. For the 8-AHA-cAMP treatment, SW-480 cells were pretreated with 8-AHA-cAMP for 1 h and then treated with R-1 for 1 d after serum starvation for 24 h. Western blotting was done as described elsewhere (25). Antibodies to survivin, phospho-survivin (Thr34), Bcl-2, PTEN, and cdc2 (Santa Cruz Biotechnology); cleaved caspase-3, caspase-3, and cleaved poly(ADP-ribose) polymerase (Cell Signaling Technology); and CB1 (Affinity Bioreagents) were used in 1:500 dilutions. The blots were stripped and then reprobed with β-actin antibody (Sigma).
cAMP-dependent protein kinase assay. cAMP-dependent protein kinase (PKA) kinase activity in cultured cells was measured by using the SignaTECT PKA Assay system (Promega) according to the manufacturer's instructions. Briefly, SW-480 cells (3 × 106) were cultured in serum-free medium for 24 h and then were treated with the indicated concentration of R-1 in serum-free medium for 24 h. The cell extracts were subjected to in vitro PKA kinase assays. Data are represented as the mean ± SE of the kinase activity from three independent experiments with duplicate.
Statistical analysis. A post hoc ANOVA was used to calculate P values for the experiments testing the effect of CB1 deletion or treatment with a CB1 agonist or antagonist on intestinal polyp formation.
Loss of CB1 expression in colorectal carcinomas. To understand the role of endocannabinoid signaling in CRC, we examined the expression pattern of CB1 and CB2 in human grade 2 to 3 colon carcinomas and human CRC cell lines. Quantitative real-time PCR analysis revealed greatly reduced expression of CB1 in 18 of 19 cancer specimens as compared with that in adjacent normal mucosa (Fig. 1A) and in 9 of 10 cell lines (Fig. 1B). In contrast, no recognizable pattern of CB2 expression was found in tumor tissues (see Supplementary Fig. S1A) or in the CRC cell lines examined (Supplementary Fig. S1B). Similarly, CB1 protein levels were lost in 15 of 16 cancer specimens in paired samples as measured by Western blotting. CB1 protein levels were barely detected in samples 17 to 19, consistent with very low levels of Cnr1 mRNA (Fig. 1A and C). Overall, CB1 mRNA and protein levels correlate well in these human biopsies. These results led us to consider that loss of CB1 expression could be associated with CRC progression.
Inactivation of tumor suppressor genes in cancer results from epigenetic silencing as frequently as that due to genetic mutations (26). Therefore, we first examined whether epigenetic silencing (DNA methylation and histone modifications) of Cnr1 contributes to loss of its transcription. Treatment with the demethylating agent 5-aza-dC restored Cnr1 mRNA expression in seven of eight CRC cell lines tested (Fig. 1D,, left) and CB1 protein expression in three cell lines (Fig. 1D , right), whereas treatment with the histone deacetylase inhibitor sodium butyrate did not significantly affect Cnr1 mRNA expression (data not shown). These results suggest that aberrant methylation of CpG islands within the promoter results in transcriptional silencing of Cnr1.
We next determined the methylation status of 39 CpG sites flanking the transcription start site of Cnr1 (−212 to +140) using bisulfite sequencing in four CRC cell lines and in the first 13 sets of human paired samples with high levels of CB1 mRNA in normal tissues. The CpG sites of this region were either fully or partially methylated in these tumor samples (Fig. 2). When the threshold for methylation was set at 30% (the upper limit of normal), the hypermethylation of this region was found in three low Cnr1 expressing cell lines (HCT-116, HT-29, and LS-174T) and in 8 of 13 human tumor tissues (T3, T5, T7, T8, T9, T10, T12, and T13; 62%). Multiple CpG sites were methylated in 10 of 13 (77%) tumor tissues but not in the matched normal tissues. None of the above 39 CpG sites assessed were methylated in the high-Cnr1-expressing cell line SW-480 or in the normal tissues (N1, N5, N8, and N13) showing high Cnr1 expression (Fig. 1). Overall, methylation of Cnr1 correlated well with loss of its transcription in these samples (Figs. 1 and 2). The most methylated CpG sites are located in the transcription factor binding sites and the transcription start site (Supplementary Table S1). Moreover, after 5-aza-dC treatment, the Cnr1 promoter changed from fully methylated to unmethylated at any site in HCT-116 cells and from fully to partially methylated in LS-174T and HT-29 cells (Supplementary Fig. S2), providing evidence that silencing of Cnr1 in human CRCs is due to methylation of its promoter. These findings uncover a new and previously unrecognized regulatory mechanism responsible for CB1 expression in CRC and help to explain one mechanism for the loss of CB1 in CRC.
CB1 signaling and intestinal tumor growth in ApcMin/+ mice. To further explore the in vivo significance of endocannabinoid signaling in CRC progression, we examined the consequences of either silencing or enhancing cannabinoid signaling in ApcMin/+ mice by deletion of cannabinoid receptor genes or by treatment with a selective CB antagonist or agonist. ApcMin/+ mice have widely been used to study CRC progression because they contain a germ-line mutation in the APC gene and, like humans, spontaneously develop multiple polyps in the intestine. Moreover, the reduction of CB1 expression was not observed in intestinal adenomas as compared with the adjacent normal tissues in Cnr1+/+/ApcMin/+ mice (Supplementary Fig. S1C). As a first step, we investigated the effect of the loss of Cnr1 or Cnr2 in ApcMin/+ mice. Thirteen-week-old male CB1-deficient ApcMin/+ mice exhibited 2.5- to 3.8-fold increases in small intestinal (Fig. 3A) and colonic (Fig. 3B) polyp burden relative to littermate control mice (Cnr1+/+/ApcMin/+). Disruption of Cnr1 in particular resulted in a >10-fold increase in the number of large polyps (> 2 mm) in the small intestine. In contrast, deletion of Cnr2 had no effect on intestinal polyp burden (Fig. 3A and B). Histologic analysis revealed that large, medium, and small polyps from mice of different genotypes were all adenomas (Fig. 3C). These results provide the first genetic evidence that CB1 is involved in regulating polyp growth in vivo.
To determine whether a CB1 antagonist would mimic the phenotypic changes found in Cnr1−/−/ApcMin/+ mice, male Cnr1+/+/ApcMin/+ mice with a genetic background identical to that of Cnr1−/−/ApcMin/+ mice were treated with carboxymethylcellulose with or without a CB1-selective antagonist (AM251) for 7 weeks. The mice treated with AM251 exhibited a 2- to 6-fold increase in small intestinal and colonic tumor burden relative to controls (Fig. 4A and B). Notably, AM251 treatment mainly increased the number of polyps >1 mm in the small intestine, whereas AM251 stimulated the growth of colon polyps of all sizes. Interestingly, ApcMin/+ mice treated with carboxymethylcellulose itself had fewer and smaller polyps than the untreated mice (Figs. 3A and 4A), suggesting that the carboxymethylcellulose has some inhibitory influence on tumor growth in this model. However, this effect was not observed in the large intestine. We further explored the role of CB1 in regulating tumor growth by evaluating the effect of a CB1 agonist in ApcMin/+ mice. Treating Cnr1+/+/ApcMin/+ mice with a CB1 agonist, R-1, in PBS for 8 weeks resulted in half to one sixth as many tumors in the small intestine and colon compared with control mice (Fig. 4C). Again, this effect was consistently observed mostly in terms of fewer large polyps. This finding is interesting because larger polyps are known to have a higher risk of progressing to carcinomas. Similarly, the higher average polyp number of the control group in Fig. 4C than that in Fig. 4A is due to the effect of carboxymethylcellulose and 1-week difference of treatment between two experiments. Unlike the results seen in Cnr1+/+/ApcMin/+ mice, administration of R-1 failed to affect small and large intestinal polyp burden in male Cnr1−/−/ApcMin/+ mice, showing that CB1 is required for the tumor-inhibiting effects of R-1 (Fig. 4D). Collectively, these genetic and pharmacologic results show that silencing of endocannabinoid signaling via CB1 accelerates intestinal adenoma growth and that activation of CB1 attenuates intestinal tumor growth, suggesting that CB1 has tumor suppressor effects.
CB1 activation and tumor cell apoptosis via down-regulation of survivin. Cannabinoids were previously shown to induce apoptotic cell death and to inhibit proliferation in cancer cell lines (4). To determine how CB1 signaling inhibits tumor growth, we examined the ability of the CB1 agonist R-1 to induce apoptosis in cultured CRC cells. As expected, SW-480 cells expressing high CB1 levels were sensitive to R-1–induced apoptosis, whereas LS-174T cells with low CB1 were more resistant to R-1 (Fig. 5A). Importantly, knockdown of Cnr1 mRNA with a selective siRNA (Fig. 5B,, left) prevented CB1 agonist–induced apoptosis of SW-480 cells (Fig. 5B,, right), confirming that CB1 mediates the proapoptotic effect of a CB1 agonist (R-1). Because treatment with 5-aza-dC led to increased CB1 expression in HCT-116 cells (Fig. 1D), we next examined the proapoptotic effect of R-1 on HCT-116 cells treated with 5-aza-dC. Similar to the results seen in SW-480 and LS-174T cells, 5-aza-dC–treated HCT-116 cells were more sensitive to R-1–induced apoptosis than controls (Supplementary Fig. S3A), showing that restoration of CB1 expression can reprogram CB1 ligand–resistant cells to become responsive to CB1 agonist. These results show that CB1 mediates the proapoptotic effect of R-1 and that the antitumor effect of CB1 relies, at least in part, on the ability of CB1 to induce tumor cell death.
To investigate the molecular mechanism(s) by which activation of CB1 induces tumor cell apoptosis, we tested whether activation of CB1 regulates genes known to control cell death. Bcl-2 and inhibitor of apoptosis protein (IAP) represent members of a large family of genes that regulate different aspects of apoptosis. IAP genes encode proteins that inhibit cellular apoptosis by binding to caspases and inhibiting their activity (27). Survivin is unique among the IAP gene family in that it is overexpressed in almost every human tumor studied but is barely detectable in most normal adult tissues (28). Overexpression of survivin is associated with a poor clinical outcome and reduced tumor cell apoptosis in patients with CRC (29, 30). Therefore, we examined the ability of a CB1 agonist to regulate these apoptotic genes and found that treatment of SW-480 cells with R-1 decreased survivin expression but had no effect on Bcl-2 or PTEN expression (Fig. 5C). R-1 treatment also led to increased caspase-3 activity, which in turn resulted in cleavage of its target, poly(ADP-ribose) polymerase, in SW-480 cells (Fig. 5C). In contrast, inhibition of CB1 expression with siRNA prevented the R-1–induced down-regulation of survivin in SW-480 cells (Fig. 5C), showing that CB1 mediates the down-regulation of survivin. Consistent with its effects on apoptosis, R-1 failed to down-regulate survivin expression in LS-174T cells, which lack the CB1 receptor (Supplementary Fig. S3B). Similarly, treatment with 5-aza-dC, which increased CB1 expression, restored the ability of R-1 to reduce survivin expression in HCT-116 cells (Supplementary Fig. S3C). Overexpression of survivin in SW-480 cells inhibited R-1–induced apoptosis (Fig. 5D), showing that survivin mediates the effects of CB1 agonist on inducing apoptosis. Because the phosphorylation of survivin at Thr34 by cdc2 increases protein stability (31), we also examined whether treatment of R-1 would diminish survivin phosphorylation at Thr34 and reduce cdc2 expression. Indeed, reductions in both survivin phosphorylation and cdc2 expression were observed in response to R-1 in SW-480 cells (Supplementary Fig. S3D), indicating that activation of CB1 enhances survivin degradation by dephosphorylation via down-regulation of cdc2. Taken together, these results provide the first evidence that survivin is a target of CB1 to regulate programmed cell death in CRC cells.
Because activation of CB1 regulates multiple cellular pathways, including stimulation of mitogen-activated protein kinase (MAPK) and PI3K-Akt signaling and inhibition of cAMP-dependent PKA activity, we examined whether these pathways are involved in CB1 down-regulation of survivin. As shown in Fig. 6, treatment of SW-480 cells with R-1 decreased PKA kinase activity (Fig. 6A) and a PKA inhibitor (H-89) mimicked the effects of CB1 on down-regulation of survivin and promotion of cell death (Fig. 6B). In contrast, a PKA activator (8-AHA-cAMP) inhibited the effects of CB1 (Fig. 6C). These results suggest that the effects of CB1 are mediated, at least in part, via a cAMP-dependent PKA pathway because inhibition of the MAPK and PI3K pathways failed to affect the CB1 regulation of survivin (Fig. 6D).
Study of the potential application of cannabinoids as antitumor drugs is an exciting consideration. Several studies have already suggested that cannabinoids exert potential antitumor effects, but they were done in cultured cell lines or in xenograft models. However, one recent study showed that increased endocannabinoid levels by a fatty acid amide hydrolase inhibitor reduce the development of precancerous lesions in an azoxymethane-treated mouse model (32). Our results further show that the CB1 receptor mediates the antitumor effects of endocannabinoids in vivo.
Epigenetic silencing plays a role in carcinogenesis and most CRCs show some epigenetic abnormalities (33). Aberrant DNA methylation occurs in normal mucosa at very early preneoplastic stages of cancer development as well as during the later stages of cancer progression, but seems to be a defining event in about half of all sporadic colon tumors (34). Hypermethylation of CpG islands is one mechanism for silencing tumor suppressor and DNA repair genes such as the cyclin-dependent kinase inhibitor P16/INK4α, P14/ARF, APC, hMLH1, and MGMT (33). However, these genes are methylated in relatively few cases (10–30%) of CRC. We found the frequency of Cnr1 methylation to be quite high (77%), indicating that loss of CB1 could be involved in many, if not most, CRCs. These results also suggest that increased levels of endogenous cannabinoids in humans may not affect tumor growth because loss of CB1 in most CRCs would make cells resistant to these ligands. Our findings that CB1 down-regulation is due to methylation of the CB1 promoter in CRC may represent a general mechanism in other cancer types and may provide a new therapeutic strategy for treatment and/or prevention of CRC. For example, an initial treatment with a demethylating agent to boost CB1 levels followed by administration of a CB1 agonist to achieve optimal stimulation of programmed cell death might be effective. Moreover, down-regulation of the CB1 is also correlated with a number of neurodegenerative diseases including Huntington's disease, Alzheimer's disease, and multiple sclerosis (35). Further investigation is required for examining whether CB1 expression is reduced in other cancer types as well. Because CB1 antagonists are known to inhibit appetite and induce weight loss (36), a CB1 antagonist, rimonabant (SR141716), is approved in Europe and Latin American for the treatment of obesity and associated dyslipidemias. However, our findings that treatment with a CB1 antagonist accelerated intestinal adenoma growth raise concerns about developing CB1 antagonists for human use, especially in people who are at high risk of developing CRC.
Dysregulation of apoptosis, proliferation, and angiogenesis all participate in cancer progression. In addition to stimulating apoptosis, cannabinoids have been shown to have antiproliferative effects on the CRC cell lines Caco-2 and DLD-1 (10) and antiangiogenic effects on gliomas (37). However, whether the inhibitory effect of cannabinoids on proliferation and angiogenesis depends on the presence of CB receptors is not clear because Caco-2 and DLD-1 cells express very low levels of CB1 (Fig. 1B) and some biological effects of cannabinoids are known to be independent of CB receptor binding (38). Further investigation is needed to determine whether cannabinoids have antiangiogenic effects in CRC and whether CB receptors mediate these effects. In addition, we examined several potential signaling pathways of CB1, including the MAPK, PI3K, and cAMP-dependent PKA pathways. We found that, primarily, a cAMP-dependent PKA intracellular signaling mediates the effects of activated CB1 on down-regulation of survivin and induction of programmed cell death. It has been established that PKA could regulate cdc2 activity and expression via two cdc2 kinase regulators, cdc25 phosphatase and Wee 1 kinase, in oocytes (39). Our results showed that activation of CB1 down-regulated cdc2 expression. Consistent with our results, a recent study showed that treatment of breast cancer cells with Δ9-tetrahydrocannabinol resulted in a decrease in cdc2 and survivin protein expression with enhanced Wee 1 and reduced cdc25C protein expression (40), supporting the hypothesis that PKA regulates survivin expression via the Wee 1/cdc25C-cdc2 cascade.
In conclusion, our studies reveal the molecular mechanism by which cannabinoids inhibit tumor growth. We found that aberrant methylation of CB1 represents a clear mechanism for loss of expression in CRC. Using multiple approaches, we provide in vivo evidence showing that endocannabinoid signaling via CB1 plays a key role in regulating intestinal tumor growth. Importantly, we elucidated the signaling pathway that mediates the proapoptotic effects of CB1. Moreover, our results may provide a rationale for the development of CB1 agonists that do not cross the blood-brain barrier for cancer prevention or treatment in combination with a demethylating agent.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
Grant support: NIH grants R01-DK-62112, P01-CA-77839, R37-DK-47297, and P30-DK-58404 (R.N. DuBois); the National Colorectal Cancer Research Alliance (R.N. DuBois); and NIH grants R37-DA06668 and R37-HD12304 (S.K. Dey).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.