Despite in the last decade the improvement of colon cancer management increased patient’s life expectancy, the occurrence of high malignancy forms and the progression toward metastatic and drug resistant phenotypes, remain a still open challenge to contrast disease relapses. Thus, discovery and development of new drugs aimed to improve therapeutic protocol for advanced malignancy are urgently needed.

It has become clear that CRC is a heterogeneous disease, both at intratumoral and intertumoral level (Punt et al., 2016). This increases difficulties in its management and, therefore, development of personalized medicine protocols based on molecular subtype could be the more advisable solution for an ideal therapy. Tumor heterogeneity implies that only a small subset of tumor cells retains tumorigenic ability, while a largest part differentiate in tumor cells that lost the ability to initiate cancer. The so-called CSCs, are certainly associated to tumor recurrence given their high chemoresistance to a large part of chemotherapeutics (Colak and Medema, 2014; Zhao, 2016).

Recently, Guinney et al. (2015), proposed a novel gene expression–based subtyping classification system for CRC. They identified four Consensus Molecular Subtypes (CMSs), with different signatures. CMS2 (canonical) is the largest subgroup (about 37%) presenting Myc and Wnt pathway hyperactivation. This result confirmed that Wnt/β-Catenin pathway plays a pivotal role in the onset and progression of CRC and the study of its intricate regulation is now emerging as a promising field to identify potential target of intervention for CRC treatment.

Wnt and c-Myc hyperactivity are undisputable markers of colon CSCs (Zeuner et al., 2014). Understanding of plastic and dynamic regulation surrounding cancer stem cells intra-clonal heterogeneity, raised the need to develop in vitro models that faithfully reflect patient’s tumors, aimed to improve personalized medicine through the knowledge of CSCs biology.

For decades, despite some unresolved disputes, endocannabinoid system modulators have gained increasing attention not only for their effects against cancer-related symptoms (such as anti-inflammatory, pain relief or anti-emetic action), but also as anti-cancer agents in many tumor types, including colon cancer (Maida and Daeninck, 2016; Velasco et al., 2016). Many reports indicate that cannabinoid compounds are able to reduce progression of CRC in vivo in the azoxymethane (AOM) induced ACF model in mice and to improve the efficacy of chemotherapeutic drugs used in the clinical practice (Izzo et al., 2008; Gustafsson et al., 2009; Santoro et al., 2009; Gazzerro et al., 2010). Moreover, in CRC cells, high levels of cannabinoids induced up-regulation of CB1-receptor through co-localization of PPARγ and RXRα at its promoting region, suggesting a potential epigenetic modulation in cannabinoid-mediated anti-tumor effects (Proto et al., 2012).

DeMorrow et al. (2008) reported that anandamide (AEA), a CB1 agonist, exerts anti-tumor effects in cholangiocarcinoma model, through activation of the non-canonical Wnt pathway. Moreover, in human breast cancer cells, a stable analogue of endogenous anandamide (Met-F-AEA) inhibits β-Catenin transcriptional activity on TCF/LEF responsive elements (Laezza et al., 2012). More interesting, some reports demonstrated the ability of cannabinoids to inhibit gliomagenesis, targeting glioma stem cells (GSCs) (Aguado et al., 2007; Nabissi et al., 2015), and spheroid formation in prostate cancer stem cells (Sharma et al., 2014).

Previous works from our group reported that Rimonabant, an antagonist/inverse agonist of CB1 receptor, is able to exert anti-tumor action in different cancer models (Malfitano et al., 2012; Ciaglia et al., 2015). Of note, in DLD1 CRC cells, Rimonabant produced mitotic catastrophe and modulated the expression of Cyclin B1, PARP-1, Aurora B and phosphorylated p38/MAPK and Chk1, improving the efficacy of Oxaliplatin (Santoro et al., 2009; Gazzerro et al., 2010).

In our recent work, we demonstrated that Rimonabant inhibits Wnt pathway activation and β-Catenin nuclear translocation both in vitro and in vivo models of CRC. The Rimonabant-mediated inhibition of TCF/LEF transcription factors allow to reduction of COX-2, Cyclin D1 and c-Myc expression. Moreover, through computational studies, we identified potential protein targets of Rimonabant and, among others, p300/Kat3b. We then demonstrated that Rimonabant was able to inhibits p300-HAT activity, suggesting that Wnt inhibition could be ascribable to downstream epigenetic regulation (Proto et al., 2017). In this work we tested the hypothesis that Rimonabant could control colon cancer stemness. In line with previous results, we found that Rimonabant reduces CD133+/CD44+ population in DLD1 and HCT116 cells, where it also downregulates Lgr5 expression. Moreover, in HCT116 cells, Rimonabant was able to produce a strong synergistic interaction with 5-Fluorouracil, to date one of the widely used drugs in CRC therapies. In DLD1 cells, as well as in GTG7 cells, we did not found downregulation of Lgr5 protein expression. Beyond its role as CSCs marker, Lgr5 receptor is strongly involved in negative feedback control of Wnt receptors turnover. Among R-Spondins (RSPO) ligands family, RSPO2 binding to Lgr5 regulates receptor turnover through stabilization of tumor suppressors ZNFR3 and RNF43 membrane E3 ligases, that mediate ubiquitination and consequent degradation of LRP6/5 receptors (Hao et al., 2016). As widely reported, the response to almost all chemodrugs is highly influenced by genotypes. Throughout RSPO2-Lgr5 signaling in CRC cells, several gene mutations were found. Both HCT116 and DLD1 cells, express low levels of RSPO2, due to its promoter methylation (Wu et al., 2014). Moreover, in HCT116 cells homozygous RNF43 mutation has been reported. Interestingly, the growth of RNF43 mutant tumors, such as HCT116, seem to be dependent on paracrine Wnt signal and, unlike to APC mutant such as DLD1, those tumors are sensitive to inhibitors of Wnt ligand secretion (Koo et al., 2012; Hao et al., 2016). We previously found a Rimonabant-mediated direct inhibition of Wnt/β-Catenin pathway in RNF43 and β-Catenin HCT116 mutant cells. On the other hand, in APC mutant DLD1 cells, we reported an indirect inhibition of Wnt signaling, substantially ascribable to p300-HAT inhibition (Proto et al., 2017). Finally, the Lgr5 splicing variants reported in CRCs, seem to influence also CSCs chemoresistance (Osawa et al., 2016).

Vermeulen et al. (2010) developed a 3D in vitro model of primary CSCs in which it is possible to take into account the tumor heterogeneity, monitoring heterogeneous Wnt activity. Here, in this model we found that Rimonabant was able to control cancer stemness, through activation of caspase 3, and then induction of cell death, in both differentiated tumor cells (characterized by Wnt-low activity) and in CSCs (defined as Wnt-high active cells). Together with the shorter survival of CSCs observed in long-term cultures and the reduction of β-Catenin expression and CD133+/CD44+ population, this result strongly supported Rimonabant ability to control colon cancer stem cells and their plasticity. Unfortunately, in primary CSCs Rimonabant did not ameliorate nor Oxaliplatin neither 5-Fluorouracil effects. Highest chemoresistance of CSCs seems to be ascribable to several mechanisms such as the high activity of drug transporters (e.g., MRP5) and to slowed cell cycling (Dean et al., 2005). Interestingly, Todaro et al. (2007) demonstrated that IL-4 inhibition enhances colon CSCs response to Oxaliplatin and 5FU. It was previously reported that, compared to differentiated cells, CSCs display a higher chemoresistance because of decreased mitochondrial priming and that BCL2, BCLXL or BCLW overexpression, increase chemotherapy resistance. However, in the same models we used in this work, BCL2-specific inhibitors failed to sensitize CSCs to chemodrugs such as Oxaliplatin (see Colak et al., 2014, where GTG7 are labeled as Co100 cells). In HCT116 cells, we previously found that Rimonabant induces apoptosis through reduction of BCL2 expression (Proto et al., 2017). Thus, this observation could explain the different results from combined treatments, observed in HCT116 and GTG7, where Rimonbant failed to sensitize the cells to Oxaliplatin and 5FU.

Mechanistic insights about Rimonabant effects and Endocannabinoid System regulation in GTG7, CSCs expressing high levels of CB1 receptor (see Supplementary Figure S2), need to be improved. Lgr5 is involved in metastasis dissemination and CSCs plasticity, more than in tumor maintenance (Melo et al., 2017). Its expression was not affected by Rimonabant treatments but, on the other hand, its localization seems to be modified. Lgr5 is constitutively internalized through a clathrin-dependent mechanism (Snyder et al., 2013). The anti-proliferative effect of Rimonabant in breast cancer cells occurs through a CB1-lipid raft/caveolae-mediated mechanism (Sarnataro et al., 2006). This evidence, together with observation of complete abolishment of LRP6 expression, allows us to speculate that an unknown mechanism occurs through plasma membrane. Moreover, the different regulation of Lgr5 receptor, observed in HCT116, DLD1 and GTG7 cells carrying different genotypes, suggests that Rimonabant could affect the intricate feedback regulation of Wnt ligands and receptors also through plasma membrane. Therefore, further studies are required to dissect the precise events underling Rimonabant-mediated effects.

Since Wnt/β-Catenin pathway regulates both normal and cancer tissues homeostasis, safely eradication of CSCs and, generally, of cancer cells, need to be improved by discovery of selective drugs. In recent years, development of ex vivo cultures “organoids,” in which material from patients can be cultured in vitro, retaining both architecture and heterogeneity of original tissue, pioneered the better understanding of both tumor biology and normal tissue homeostasis. These models are currently used for a wide spectrum of application, such as personalized medicine, living biobanks, genetic analysis, but also drug discovery (Leushacke and Barker, 2014; Ohta and Sato, 2014). In our previous work, we reported the anti-proliferative effect of Rimonabant in glioma cell line and primary cells from patients, without significant toxicity in Normal Human Astrocyte (NHA), thus demonstrating moderate selectivity toward glioma cells, in vitro (Ciaglia et al., 2015). Here, for the first time, we demonstrated that in normal colon human organoids (wild type organoids that represent a high fidelity surrogate of normal colon epithelium) Rimonabant did not shows toxicity or clonogenicity reduction in normal cells, also at higher doses. To date, many reports tried to elucidate distribution of receptors, ligands and enzymes of endocannabinoid system in the gut. While CB2 receptor shows relevant role in mucosal immunity, being expressed prevalently in intestinal macrophages, CB1 receptor seems to be expressed in smooth muscle, submucosal myenteric plexus and normal human colon (Wright et al., 2005). The intricate regulation of this system seems to be contest- and cell-specific and may totally differ depending on patho-physiological conditions, but the exact regulation must still be clarified to optimize clinical use of specific cannabinoid compounds. Some authors reported that acute administration of Rimonabant (3 and 5.6 mg/kg dose), increases gastrointestinal motility, diarrhea and nausea, in a dose-dependent manner. However, Carai et al. (2004) demonstrated that in mice, after initial induction of intestinal peristalsis, tolerance mechanisms occurred (Carai et al., 2004; Izzo and Sharkey, 2010). Of course, since normal colon organoids contain all cell type of intestinal crypt and villus compartments (e.g., Lgr5+ niche stem cells, goblet cells, Paneth cells etc.) (Leushacke and Barker, 2014), with these preliminary results we cannot establish if Rimonabant influences differentiation or, generally, homeostasis of normal epithelium, in cell-specific manner. Thus, additional experiments would be needed to deepen this aspect.

Substantially, these promising results could candidate Rimonabant as a novel lead compound for CRC treatments, and could help its optimization as chemotherapic agent, also taking advantage of information obtained in previous clinical studies as anti-obesity drug.