Introduction
The identification of cell–cell communication signaling pathways that regulate embryogenesis and adult tissue homeostasis, and which are altered in hyperproliferative or degenerative diseases, has lead to the idea that specific pathway modulators may be efficacious therapeutic agents. For example, a variety of sporadic human cancers have been found to harbor hyperactive canonical WNT signaling, leading to the constitutive activation of β‐CATENIN, a key polyvalent protein that is modified by phosphorylation at multiple sites. In particular, early colon cancers commonly display loss of function of the tumor suppressor Adenomatous polyposis coli (APC), a key component of the β‐CATENIN destruction complex, or harbor N‐terminal gain of function mutations in β‐CATENIN that render it APC‐resistant. In both of these cases, this leads to canonical WNT pathway activation and the resulting regulation of target genes (reviewed in Kinzler & Vogelstein,
1996; MacDonald
et al,
2009; Valenta
et al,
2012; Varnat
et al,
2010). Other cancers also show an active canonical WNT pathway; these include carcinomas of the lung, stomach, cervix, endometrium, and lung as well as melanomas and gliomas (e.g., Nguyen
et al,
2009; Kandoth
et al,
2013;
http://cancer.sanger.ac.uk/cancergenome/projects/cosmic/).
In normal embryogenesis and homeostasis, the canonical WNT pathway is activated by secreted WNT ligands produced in highly controlled context‐dependent manners and in precise amounts. WNT activity is transduced in the cytoplasm, inactivates the APC destruction complex, and results in the translocation of activate β‐CATENIN to the nucleus, where it cooperates with DNA‐binding TCF/LEF factors to regulate WNT‐TCF targets and the ensuing genomic response (e.g., Kinzler & Vogelstein,
1996; Shitashige
et al,
2008; MacDonald
et al,
2009; Clevers & Nusse,
2012; Valenta
et al,
2012).
In the absence of WNT ligands, the APC destruction complex acts to phosphorylate cytoplasmic β‐CATENIN in its N‐terminal region (including at Ser33/Ser37 via CK1 and GSK3 β kinases) targeting it for proteasomal degradation. Without active nuclear β‐CATENIN, there is no activation of positive WNT‐TCF targets. In contrast, under active WNT signaling, β‐CATENIN escapes cytoplasmic degradation and is phosphorylated at its C‐terminus, for instance at positions Ser552 and Ser675, by various kinases including AKT, PAK, PKA, and possibly AMPK. Such C‐terminal phosphorylation has been found to be essential for the transcriptional function of β‐CATENIN with TCF factors (Hino
et al,
2005; Taurin
et al,
2006; Fang
et al,
2007; Zhao
et al,
2010; Zhu
et al,
2012a). These findings indicate that beyond the loss of activity of the APC destruction complex, for instance through
APC mutation, phosphorylation of β‐CATENIN at C‐terminal sites is required for the full activation of WNT‐TCF signaling and the ensuing WNT‐TCF responses in cancer.
Attempts to pharmacologically inhibit WNT signaling have lead to the identification of a number of small molecules that act at different levels in the core signaling cascade (e.g., Lepourcelet
et al,
2004; Curtin & Lorenzi,
2010; Anastas & Moon,
2013; see
http://www.stanford.edu/group/nusselab/cgi-bin/wnt/smallmolecules). These include inhibitors of Porcupine and Tankyrases, which promote normal WNT ligand secretion or cytoplasmic transduction, respectively (Huang
et al,
2009; Dodge
et al,
2012; Waaler
et al,
2012; Lau
et al,
2013). However, activation of WNT signaling below the level of ligand function or cytoplasmic transduction, as in the case of loss of APC for instance (see above), would appear to complicate the effects of pathway blockade at upstream levels. So far, there are no approved WNT signaling inhibitors in the clinics and clinical trials of new lead compounds are lengthy, very costly, and with less than 1/20 success rates.
Here, we have thus focused on the end part of the WNT‐TCF pathway, using a well‐established transcriptional reporter assay (Barker & Clevers,
2006) for TCF activity driven by activated, APC‐insensitive N‐terminal mutant β‐CATENIN (N'∆45 β‐CATENIN). However, with the aim to find blockers of WNT‐TCF responses in cancer, we implemented two key modifications: i‐ a reliable dynamic range determined by activated β‐CATENIN as the activating pole and activated β‐CATENIN plus dominant‐negative TCF (dnTCF) as the repressing pole, and ii‐ sine qua non mimicry of genetic blockade by dnTCF activity for any hit.
We find that macrocyclic lactones of the Avermectin family have specific anti‐WNT‐TCF response activity in human cancer cells and that the clinically approved compound Ivermectin (EMEA‐ and FDA‐approved) is a specific WNT‐TCF response blocker at low micromolar concentrations.
Ivermectin and other macrocyclic lactones are well‐tolerated agents that have been used to treat millions of people for river blindness (Thylefors,
2008; Traore
et al,
2012) and other parasite infections (Campbell
et al,
1983; González
et al,
2012; Nolan & Lok,
2012). These actions of Ivermectin have been ascribed to its blockade of parasite ligand‐gated chloride channels (Hibbs & Gouaux,
2011; Lynagh & Lynch,
2012). The WNT‐TCF response blockade that we describe for low doses of Ivermectin suggests an action independent to the deregulation of chloride channels. Its low concentrations effects are rescued epistatically by direct constitutive activation of TCF transcriptional activity and involve the repression of the levels of C‐terminally phosphorylated β‐CATENIN forms and of CYCLIN D1, a critical target that is an oncogene and positive cell cycle regulator. Moreover, the Avermectin single‐molecule derivative Selamectin, a drug widely used in veterinarian medicine (Nolan & Lok,
2012), is ten times more potent acting in the nanomolar range. Finally, Ivermectin has
in vivo efficacy against human colon cancer xenografts sensitive to TCF inhibition with no discernable side effects. The ability of systemic Ivermectin to also block lung cancer growth
in vivo supports the possible use of Ivermectin in particular, and other macrocyclic lactones such as Selamectin in general, as WNT‐TCF pathway response blockers to combat WNT‐TCF‐dependent human diseases including cancers of the intestine, breast, skin, and lung.
Discussion
Here we report that Ivermectin (Campbell
et al,
1983), an off‐patent drug approved for human use, and related macrocyclic lactones, have WNT‐TCF pathway response blocking and anti‐cancer activities. Whereas the exact molecular target for Ivermectin and Selamectin that affects WNT‐TCF responses remains to be identified, the present findings show that these drugs block WNT‐TCF pathway responses, likely acting at the level of β
‐CATENIN/TCF function, affecting β
‐CATENIN phosphorylation status.
The similar anti‐proliferative activities of Abamectin, Doramectin, and Moxidectin with those of Ivermectin and Selamectin suggest that macrocyclic lactones of the Avermectin and Milbemycin families share common properties and structural features (e.g., Awasthi
et al,
2012) that may be the basis of the anti‐WNT‐TCF activities of Ivermectin and Selamectin. Moreover, the potent WNT‐TCF response inhibitory activity of Selamectin strongly argues that the activity of macrocyclic lactone preparations made through fermentation, such as Ivermectin, is not due to secondary or residual components since Selamectin is a semi‐synthetic single compound drug. The finding that Selamectin is tenfold more potent begs its clinical testing but also the further exploration of the macrocyclic lactone chemical space.
Ivermectin has a well‐known anti‐parasitic activity mediated via the deregulation of chloride channels, leading to paralysis and death (Hibbs & Gouaux,
2011; Lynagh & Lynch,
2012). The same mode of action has been suggested to underlie the toxicity of Ivermectin for liquid tumor cells and the potentiation or sensitization effect of Avermectin B1 on classical chemotherapeutics (Drinyaev
et al,
2004; Sharmeen
et al,
2010). In contrast, the specificity of the blockade of WNT‐TCF responses we document, at low micromolar doses for Ivermectin and low nanomolar doses for Selamectin, indicate that the blockade of WNT‐TCF responses and chloride channel deregulation are distinct modes of action. In support of this, WNT‐TCF response blocking activity is detected at up to tenfold lower concentrations than those reported for chloride ion deregulation (this work; Drinyaev
et al,
2004). The finding that Moxidectin is more potent than Ivermectin in controlling intestinal nematodes (Fatima
et al,
2007; Cringoli
et al,
2009) but similarly or less active on human cancer cells (this work), further supports different modes of action of macrocyclic lactones on cancer cells vs. parasites. The specificity and selectivity of Ivermectin and Selamectin we describe are also inconsistent with ubiquitous effects of these macrocyclic lactones through alterations of anion‐selective Cys‐loop channels (Hibbs & Gouaux,
2011; Lynagh & Lynch,
2012) or Farnesyl X receptors (Jin
et al,
2013). The idea that Ivermectin can target different molecules is further supported by its inhibition of flaviviral helicase activity present only during viral replication in infected cells (Mastrangelo
et al,
2012;). What is key then is to find a dose and a context where the use of Ivermectin has beneficial effects in patients, paralleling our results with xenografts in mice.
Cell toxicity appears at doses greater (> 10 μM for 12 h or longer or > 5 μM for 48 h or longer for Ivermectin) than those required to block TCF responses and induce apoptosis. General toxicity related to chloride channel deregulation has been suggested to underlie the high micromolar toxicity of Ivermectin for liquid tumor cells
in vitro (Drinyaev
et al,
2004; Sharmeen
et al,
2010) and might complicate the treatment of WNT‐TCF‐dependent brain diseases since Ivermectin can affect glutamate‐gated and other Cys‐loop ion chloride channels (Kokoz
et al,
1999; Hibbs & Gouaux,
2011; Lynagh & Lynch,
2012). Ligand‐gated chlorine channels are found in the mammalian central nervous system and are normally protected from systemic Ivermectin treatments by the blood–brain barrier (Schinkel
et al,
1994), but this is often broken in brain tumors.
Our data point to a repression of WNT‐β‐CATENIN/TCF transcriptional responses by Ivermectin, Selamectin and related macrocylic lactones. This conclusion is based on (i) The ability of Avermectin B1 to inhibit the activation of WNT‐TCF reporter activity by N‐terminal mutant (APC‐insensitive) β‐CATENIN as detected in our screen; (ii) The ability of Avermectin B1, Ivermectin, Doramectin, Moxidectin and Selamectin to parallel the modulation of WNT‐TCF targets by dnTCF; (iii) The finding that the specific WNT‐TCF response blockade by low doses of Ivermectin and Selamectin is reversed by constitutively active TCF; (iv) The repression of key C‐terminal phospho‐isoforms of β‐CATENIN resulting in the repression of the TCF target and positive cell cycle regulator CYCLIN D1 by Ivermectin and Selamectin; (v) The specific inhibition of in‐vivo‐TCF‐dependent, but not in‐vivo‐TCF‐independent cancer cells by Ivermectin in xenografts.
Analyses of phospho‐isoforms of β
‐CATENIN after treatment with Ivermectin or Selamectin under PP2A/PP1 protein phosphatase‐blocked conditions suggest that these drugs may act by enhancing, directly or indirectly, phosphatase activity involved in dephosphorylating P‐Ser552/P‐Ser675. This effect can help explain the phenotype of Ivermectin‐treated cells since P‐Ser552‐ and P‐Ser675‐β‐CATENIN show enhanced transcriptional activity in cooperation with TCF factors and are essential for WNT signaling in colon cancer cells (Hino
et al,
2005; Taurin
et al,
2006; Fang
et al,
2007; Zhu
et al,
2012a). Support for an involvement of PP2A also derives from the finding that its Bα (PR55α) subunit is required to downregulate the levels of P‐Ser552 and P‐Ser675 C‐terminal phosphoforms of β
‐CATENIN in colon cancer cells (Zhang
et al,
2009). The role of PP2A and its multiple subunits is thus likely to be complex since it can also act positively on WNT signaling with its B56 subunit through the inhibition of N‐terminal β
‐CATENIN phosphorylation and thus inhibition of β
‐CATENIN destruction (Seeling
et al,
1999). Nevertheless, in cancer cells lacking APC destruction complex function (e.g. lacking APC), the overall effect of enhancing PP2A function via Ivermectin treatment is predicted to be pathway silencing (Fig
6H).
The elucidation of the effects of Ivermectin discussed above benefited from the fact that most human colon cancer cells tested
in vitro in monolayers are TCF‐dependent (e.g., van de Wetering
et al,
2002; Varnat
et al,
2010), thus allowing the
in vitro screen. Interestingly, the capacity of cancer cells to form 3D spheroids in culture, as well as the growth of these, is also WNT‐TCF‐dependent (Kanwar
et al,
2010) and they were also affected by Ivermectin treatment. These results together with the reduction of the expression of the colon cancer stem cell markers
ASCL2 and
LGR5 (e.g., Hirsch
et al,
2013; Ziskin
et al,
2013) raise the possibility of an inhibitory effect of Ivermectin, Selamectin and related macrocyclic lactones on TCF‐dependent cancer stem cells.
In vivo, our previous work has shown that most colon cancer tested in xenografts are TCF‐independent. For example, Ls174T and primary CC14 cells are TCF‐dependent
in vitro but become TCF‐independent in xenografts
in vivo (and vice versa) (Varnat
et al,
2010). In contrast, DLD1 remain TCF‐dependent
in vitro and
in vivo (Varnat
et al,
2010). The basis for these changes remains unclear although DNA methylation might be involved (de Sousa
et al,
2011). Notwithstanding the mechanism, these differences afforded a key test for the specificity of Ivermectin
in vivo. If Ivermectin is specific, it should only block TCF‐dependent tumor growth. Indeed, the sensitivity and insensitivity of DLD1 and CC14 xenografts to Ivermectin treatment, respectively, together with the desensitization to Ivermectin action
in vivo by constitutively active TCF provide evidence of the specificity of this drug to block an activated WNT‐TCF pathway in human cancer.
Ivermectin has a good safety profile since only
in‐vivo‐dnTCF‐sensitive cancer xenografts are responsive to Ivermectin treatment, and we have not detected side effects in Ivermectin‐treated mice at the doses used. Whereas it remains likely that higher doses may begin to show non‐specific toxicity, previous work has shown that side effects from systemic treatments with clinically relevant doses in humans are rare (Yang,
2012), that birth defects were not observed after exposure of pregnant mothers (Pacqué
et al,
1990) and that this drug does not cross the blood–brain barrier (Kokoz
et al,
1999). Similarly, only dogs with mutant
ABCB1 (MDR1) alleles leading to a broken blood–brain barrier show Ivermectin neurotoxicity (Mealey
et al,
2001; Orzechowski
et al,
2012).
Oral Ivermectin is already used by millions of people to combat multiple parasite infections, notably through the Mectizan donation program against river blindness (Thylefors,
2008). Given our present data, this drug could therefore be additionally used as a WNT‐TCF blocker against different diseases, including multiple WNT‐TCF‐dependent human cancer types. In this case, this will likely involve a combinatorial approach with standard chemotherapeutics as debulking agents. Indications may include treatment for incurable β‐CATENIN/TCF‐dependent advanced and metastatic human tumors of the lung, colon, endometrium, and other organs. Ivermectin, Selamectin, or related macrocyclic lactones could also serve as topical agents for WNT‐TCF‐dependent skin lesions and tumors such as basal cell carcinomas. Moreover, they might also be useful as routine prophylactic agents, for instance against nascent TCF‐dependent intestinal tumors in patients with familial polyposis and against nascent sporadic colon tumors in the general aging population. Formulations of Ivermectin and other hydrophobic macrocyclic lactones with agents that enhance tissue penetration may improve their efficacy.
Materials and Methods
Cell culture
Human colon cancer CC14 and CC36 primary cells and Ls174T and HT29 cell lines were cultured as previously described (Varnat
et al,
2009,
2010). Colon cancer DLD1 and lung non‐small cell bronchioalveolar carcinoma H358 cells were cultured as attached 2D layers in DMEM‐F12 or RPMI 1640 with 10% FBS. 293T cells were cultured in DMEM with 10% FBS. For drug treatments, cells were plated in 2% FBS and treated the next day for 12–48 h. GBM primary cells were cultured as described in Zbinden
et al (
2010). U251 glioma and SKMel2 melanoma cell lines were cultured as described by ATCC.
Primary screening
A luciferase reporter assay was performed in 293T cells to identify compounds able to decrease TCF‐driven gene expression using a TCF‐binding site fused to a firefly luciferase reporter construct. Activation was driven by transfected N'∆β‐CATENIN acting on endogenous TCF factors. Internal controls derived from Renilla luciferase activity resulting from a HSV‐TK expression plasmid co‐transfected in all cases. Transfection of 293T cells was preformed with calcium phosphate and plated at a density of 5,000/well into 96‐well plates. Each transfection combined a constitutively active β‐CATENIN plasmid as effector, a TCF‐binding site firefly luciferase plasmid as reporter, and a HSV‐TK Renilla luciferase plasmid as control. Cells were harvested 12 h after compounds treatment. Secondary reporter assays were repeated at least three independent times.
Drugs
Abamectin (31732, Fluka), Ivermectin (Fagron Iberica, and Sigma #I8898), Doramectin (33993, Fluka), Selamectin (32476, Sigma), Stromectol™ (Merck) and Moxidectin (33746, Sigma) were diluted in 100% DMSO or ethanol and used at 0.01–10 μM for in vitro assays. Okadaic acid (OA, Enzo Life Science) was used at 1–15 nM.
BrdU incorporation
Colon primary cell culture and cell lines were plated in medium containing 2% FBS and treated with different drug concentrations for 48 h. After a 15′ pulse with BrdU (10 mg/ml), cells were fixed in 4% PFA, incubated in 2N HCl for 15′, neutralized with 0.1 M Boric acid for 10′, blocked in PBS‐10% HINGS for 10′, and labeled with anti‐BrdU antibodies (1:5,000, University of Iowa Hybridoma Bank). Signal was revealed with a rhodamine‐coupled anti‐mouse secondary antibody (1:500, Invitrogen Molecular Probes) and counterstained with DAPI (Sigma). At least 10 fields were quantified for each condition.
Apoptosis assay
Colon cancer cells were plated at 30% confluency and treated with different drug concentrations for 48 h. After treatment, cells were fixed in 4% PFA for 2 min and blocked in 10% HINGS for 1 h. Apoptosis was evaluated by immunofluorescence using a rabbit anti‐cleaved Caspase3 antibody (1:200, overnight at 4°C; Cell Signaling) and Cy3‐labeled secondary antibody (1:500, 1 h at RT; Jackson ImmunoResearch). Cells were counterstained with DAPI (1:10,000, Sigma). At least 10 fields were quantified per plate for each condition.
Quantitative reverse transcription PCR
RNA extraction, reverse transcription and qPCR were performed as previously described (Varnat
et al,
2009).
qPCR primer sequences were as follows (5′‐>3′):
In vitro clonogenic assays and CD133 sorting
Adherent colon cancer cells were treated with Ivermectin or Selamectin at different concentrations as described, washed, lifted, and plated at 1 cell/well in 96‐well non‐adherent plates in colon spheroid medium (DMEM‐F12 supplemented with B27 and 10 ng/ml EGF) without drugs and grown for 14 days. Spheres were then counted visually using an inverted Zeiss optical microscope. For secondary clonogenic assays, primary spheres were dissociated and replated at 1 cell/well in 96‐well plates and cultured as described above. Each experiment was repeated at least thrice. Alternatively, untreated cells were plated and treated in 96‐well plates with drugs and subsequently scored for spheroid formation. CD133 MACS (Miltenyi Biotech) was performed as described previously (Varnat
et al,
2009,
2010) using either limited Trypsin or StemProAccutase (Gibco LifeTech) for cell dissociation.
Western blots
Cells were plated in media containing 2% FBS and treated with Ivermectin or Selamectin at different concentrations for 12 h. Protein lysates were in RIPA buffer. 10–20 μg of protein were loaded onto 12% polyacrylamide gels and the proteins transferred onto nitrocellulose membranes (Hybond), which were probed with primary and secondary antibodies as noted. Signal was revealed using the ECL system (GE Healthcare) and film. ImageJ was used for quantification. Membranes were sequentially reprobed after washing in TBS‐1% Tween‐20 and controlling for loss of signal. The following primary antibodies were used: anti‐GAPDH (1:6,000 for 1 h at RT, 2118 Cell Signaling), anti‐cyclinD1 (1:100 for 1 h at RT, Ab‐3 Oncogene research), anti‐Pser552 β‐CATENIN (1:1,000 for 1 h at RT, 9566 Cell Signaling), anti‐Pser675 β‐CATENIN (1:1,000 for 1 h at RT, 4176 Cell Signaling), and anti‐total β‐CATENIN (1:1,000 for 1 h at RT, 8480 Cell Signaling). As secondary antibodies, HRP‐coupled anti‐rabbit or anti‐mouse (1:6,000, Promega) were incubated for 1 h at RT.
Mouse xenografts
5 × 105 control or lentivector‐infected cancer cells were injected subcutaneously into the flanks of 6‐ to 8‐week‐old female NMRI Nude mice. After tumor establishment, mice were treated with cyclodextrin carrier alone or Ivermectin (or Stromectol™) conjugated with cyclodextrin (45%) via daily intraperitoneal injections at 10 mg/kg. Tumor volumes were measured every 2–3 days.
Sequencing APC and β‐CATENIN commonly mutated regions
One‐hundred nanograms of genomic DNA from primary colon cancer cells (CC14 and CC36) or cell lines (DLD1 and Ls174T) was used for PCR using Phusion HF DNA Polymerase (BioLabs). Specific primers were designed to amplify commonly mutated regions in APC and CTNNB1 (β‐CATENIN) genes. The Mutation Cluster Region (MCR) of APC was amplified with two pairs of primers: (1) MCR1FWD 5′‐GATACTCCAATATGTTTTTC‐3′ and MCR1REV 5′‐GGAAGATCACTGGGGCTTAT‐3′, (2) MCR2FWD 5′‐GTGAACCATGCAGTGGAATG‐3′ and MCR2REV 5′‐TCTGAATCATCTAATAGGTC‐3′. The primers for CTNNB1 were as follows: CTNNB1FWD 5′‐CAATGGGTCATATCACAGATTCTT‐3′ and CTNNB1REV 5′‐TCTCTTTTCTTCACCACAACATTT‐3′.
β‐CATENIN immunohistochemistry
Colon cancer cells were plated at 40% confluency and treated with 5 μM Ivermectin or 2% DMSO for 6 h. After treatment, cells were fixed in fresh 4% PFA for 2 min. Antigen retrieval was conducted by boiling for 7 min in 10 mM citrate buffer (10 mM citric acid, 0.05% Tween‐20, pH 6.0). Cells were then incubated at RT for 30 min, washed with PBS and permeabilized in PBS with 1% Tween‐20 for 15 min at RT. Anti‐β‐CATENIN antibodies (1:400, Cell Signaling) were applied overnight at 4°C after blocking with 3% BSA in PBS. Signal was revealed with a rhodamine‐coupled anti‐rabbit secondary antibody (1:1,000) and counterstained with DAPI (1:10,000). Samples were imaged with Zeiss LSM 510META confocal microscope. Five independent slides were analyzed per each condition.