Evaluation of a chemical library of small-molecule Dishevelled antagonists that suppress tumor growth by down-regulating T-cell factor–mediated transcription

We describe the rational generation of small-molecule agents that suppress tumor cell growth by down-regulating canonical Wnt signaling. We first produced a chemical library of the derivatives of indole-2-ketones and carbinols; we then screened them by using scalable assays of biochemical antagonism of Dishevelled-1 PDZ domain interactions and cell-based assays of Dishevelled-1–driven T-cell factor–mediated transcription. Compounds showing parallel effects in these assays were tested for selective induction of apoptosis in cancer cells. A new compound (24) that met the criteria for high biochemical antagonism, T-cell factor–mediated transcription, and induction of tumor-selective apoptosis was found to significantly suppress the growth of tumor xenografts in mice. [Mol Cancer Ther 2008;7(6):1633–8]

Activation of Wnt signaling seems to play an important role in carcinogenesis (1, 2). Ectopic expression of Wnt-1 can result in tumorigenic conversion of primary human mammary epithelial cells (3). One important mechanism of Wnt-dependent tumorigenesis is autocrine up-regulation of extracellular Wnt growth factors (4), in which transcriptionally active β-catenin accumulates without mutation of adenomatous polyposis coli or β-catenin. Importantly, this mechanism induces cancer cell growth as robustly as the mutation-based mechanism (4). Therefore, overexpression of downstream components of the Wnt receptor complex can act synergistically with upstream Wnt autocrine up-regulation in tumorigenesis.

Dishevelled-1 (Dvl1; ref. 5), a PDZ domain-containing mediator of Wnt signaling, has been reported to be up-regulated in tumors with β-catenin activation (6). Overexpression of Dvl is a dominant event in malignant pleural mesothelioma and non–small cell lung cancer (7, 8). Amplification and abnormally high expression of Dvl1 has also been found in primary breast and cervical cancers, suggesting that Dvl1 may play a role in their pathogenesis (9, 10).

The Fz family of human proteins comprises Wnt receptors that maintain a conserved KTXXXW motif in the cytoplasmic domains that is essential in mediating Wnt signals (11). This motif of Fz7 has been shown to interact directly with the PDZ domain of Dvl1 (12). To date, Dvl is the only direct downstream signaling partner of Fz that has been identified. Dvl transmits the Wnt signal to at least three distinct pathways [β-catenin/T-cell factor (Tcf), c-Jun NH₂-terminal kinase, and planar cell polarity] and thereby organizes pathway-specific subcellular signaling (13).

Transcription mediated by Tcf is crucial for enhancement of cellular growth. In colon cancer, Tcf up-regulates transcription of well-characterized oncogenes (14, 15) and survivin (16), an inhibitor-of-apoptosis family member that prevents apoptosis of cancer cells. Wnt-1 inhibits the c-myc–induced release of cytochrome c (17) and prevents apoptosis through activation of Tcf-mediated transcription (18). The adenomatous polyposis coli tumor suppressor protein is an up-regulator of caspase-3 and caspase-7 (19), and its function is suppressed on Wnt stimulation. Accordingly, Wnt signaling consistently inhibits activation of caspase (17). Taken together, this information indicates that Wnt-mediated transcription of Tcf target genes in cancer cells not only facilitates tumor growth but also inhibits apoptosis. Further, we have shown that Wnt-inhibitory factor (20, 21) or Wnt small interfering RNAs (22, 23) induce apoptosis of cancer cells, suggesting another mechanism by which inhibition of Wnt signaling in cancer cells (by targeting another Wnt pathway component that inhibits adenomatous polyposis coli function) promotes their apoptosis.

We have made progress in the discovery and development of nonpeptide small-molecule antagonists of PDZ domain interactions that permeate cells and exert effects attributable to their targeted biochemical antagonism (24). As proof of principle, we developed a small-molecule compound (FJ9) that inhibits intracellular interactions of the Dvl PDZ domain, down-regulates Tcf-mediated transcription, and suppresses tumor growth in a mouse xenograft model (25). These findings suggest that a nonpeptide antagonist of the Dvl PDZ domain can provide novel antitumor therapy by reducing Tcf transcriptional activity. Here, we report an evaluation of chemical library to rationally generate small-molecule agents that suppress tumor cell growth by down-regulating canonical Wnt signaling.

All commercial chemicals were purchased from Aldrich or Fisher Scientific. Human cell lines (HEK293T, MDA-MB231, HepG2, BJ, and NCI-H460) were obtained from the American Type Culture Collection. These cells were cultured in Eagle's MEM supplemented with 10% fetal bovine serum without antibiotics (HEK293T, HepG2, and BJ), L-15 supplemented with 10% fetal bovine serum without antibiotics (MDA-MB231), or RPMI 1640 supplemented with 10% fetal bovine serum, penicillin (100 IU/mL), and streptomycin (100 mg/mL; NCI-H460). Glutathione S-transferase–tagged Dvl1 PDZ domain protein was expressed as reported previously (25).

The chemical library was synthesized as shown in Fig. 1. The detailed procedure is described in Supplementary Data.³

All compounds (except for aliquots tested in mice) were stored as 10 mmol/L solutions in DMSO.

The potency with which each compound inhibited interaction between the Dvl1 PDZ domain and the Fz7 PDZ-binding region was measured by using the AlphaScreen Glutathione S-Transferase Detection kit (Perkin-Elmer). The probe (biotin-IWSGKTLNSWRKF, 10 μmol/L) and a glutathione S-transferase–tagged PDZ domain of human Dvl1 (100 nmol/L) were mixed into an assay solution (pH 7.4, 25 mmol/L HEPES, 100 nmol/L sodium chloride, and 0.1% bovine serum albumin). Each compound was serially diluted into the probe solution and incubated for 30 min at room temperature. Each sample solution (15 μL) was then added in triplicate to a 384-well plate, and 5 μL of anti–glutathione S-transferase acceptor beads were immediately added. After incubation at room temperature for 30 min, 5 μL of streptavidin donor beads were added and the mixtures were incubated for an additional 45 min. The AlphaScreen signal was measured with an EnVision 2103 Reader (Perkin-Elmer). The binding curves were generated by using GraphPad Prism software to calculate the IC₅₀ values.

HEK293T cells cultured in a 6-cm dish were transfected when 90% confluent with the superTOPflash reporter plasmid (3 μg) and mouse Dvl1-pCS2+ plasmid (3 μg) by using Lipofectamine LTX (Invitrogen) according to the manufacturer's recommendations. After incubation for 5 h, cells were resuspended in fresh medium, distributed into 384-well plates (15,000 per well), and incubated for 19 h. The test compounds (100 μmol/L final concentration) were added in triplicate and the cells were then incubated for 1 d. The same amount of DMSO was used as negative control. AlamarBlue reagent (1 μL; BioSource) was added, and the fluorescence signal was recorded according to the manufacturer's recommendations. Steady-Glo reagent (20 μL; Promega) was then added, and the luminescence signal was read in an EnVision 2103 Reader (Perkin-Elmer) according to the manufacturer's recommendations. Tcf reporter activity was calculated for each well as the luminescence signal was normalized by the fluorescence signal, and the activity of each compound was evaluated as the mean percent inhibition of Tcf activity compared with that of DMSO control (assigned a value of 0% inhibition).

hTERT-immortalized BJ human skin normal fibroblast cells, MDA-MB231 breast cancer cells, and HepG2 hepatocellular carcinoma cells were distributed into 384-well plates (15,000 per well), the test compounds (100 μmol/L final concentration) were added in triplicate, and the plates were incubated for 2 d. The same amount of DMSO was used as a negative control. Caspase-Glo 3/7 reagent (20 μL; Promega) was added, and the luminescence signal was read as described above. Caspase activity was calculated as the mean luminescence signal, and the mean caspase activity for each compound was expressed as a proportion of the caspase activity obtained for DMSO control (assigned as 0% activation).

H460 non–small cell lung cancer cells in 24-well plates were incubated for 4 d with culture medium (500 μL/well) containing 10 μmol/L of each test compound and then treated with trypsin, stained with Annexin V-FITC (Apoptosis Detection kit, Oncogene) according to the manufacturer's protocol, and immediately analyzed by flow cytometry (FACScan, Becton Dickinson) as described previously (23). Cytometry values represented Annexin V-FITC staining (X axis) and propidium iodide staining (Y axis). Percent apoptosis was calculated as the percentage of FITC-positive cells (sum of upper-right and lower-right areas of the graph). The percent increase in the apoptotic population was calculated by subtracting the value for DMSO control.

Groups of 6-wk-old female nude mice (strain: NSWNU-M; Taconic) were injected s.c. in the dorsum with 4 million H460 tumor cells in 100 μL PBS. After 7 d, animals in which tumors had developed essentially uniformly were selected and groups of nine began receiving daily i.p. injections of compound (50 mg/kg in 80 μL PBS-20% DMSO). Groups of 11 control mice received 80 μL PBS-20% DMSO. Injections were administered on days 8 to 21 after cell injection. Tumor volume was calculated as x²y/2, where x = width, y = length, and x < y, and was reported as the mean and SD.

A one-tailed Student's t test was used to compare the effect of treatment with that of no treatment.

We designed an indole-2-carbinol scaffold to allow generation of a chemical library via diversity of substituents at its 2- and 3-positions, and we developed synthetic methods to achieve this diversity (Fig. 1). Triethylsilylated alkynes (compound 3) were easily prepared from terminal alkynes (2) and, when coupled with an iodoaniline (1) by a palladium catalyst, yielded indoles with a triethylsilyl group at the 2-position and diverse structures at the 3-position (4). Compound 4 was next treated with acid chlorides to replace the 2-silyl group with diverse acyl groups to afford diverse indole-2-ketones (5–25) and indole-2-carbinols (26–47). The chemical structures of these compounds are shown in Supplementary Table S1.³

We used an AlphaScreen protocol to compare the biochemical inhibition (IC₅₀) of the binding of the compounds of Dvl1 PDZ domain and peptide probes designed based on native ligand Fz7. Several compounds showed greater biochemical activity than that of FJ9 (26), our previous lead compound (Supplementary Table S1;³ ref. 25). To determine whether the biochemical antagonists of Dvl1 also down-regulated Tcf transcription, we used a scalable cell-based reporter assay. HEK293T cells are commonly used in Wnt pathway analyses because they have an intact but not enhanced Wnt pathway system (no mutation/amplification in Wnt signaling molecules) and therefore have very low endogenous Tcf transcriptional activity. We have confirmed that exogenous Dvl1 strongly activates Tcf activity in these cells as measured by superTOPflash reporter (26) but not as measured by superFOPflash, the negative control mutant promoter reporter. We measured Tcf transcriptional activity in Dvl1-transfected HEK293T cells after treatment with the library compounds.

We next wished to identify compounds that both biochemically antagonized Dvl1 PDZ domain and actively down-regulated Tcf-mediated transcription. Several compounds met these criteria (Fig. 2), including 8, 9, 24, 25, 30, 33, and 47. In contrast, all compounds with polar functional groups in the R1/R2 moieties (e.g., 17, 18, 31, 40, and 41) showed weak down-regulation of Tcf activity regardless of the potency of their biochemical Dvl1 antagonism, presumably because of less ability to penetrate the cell membrane. No compounds from this chemical library showed any down-regulation of reporter activity of activator protein-1, an unrelated transcription factor, in the same cells and under the same conditions (data not shown). These results suggest that the down-regulation of Tcf-mediated transcription by these compounds is not nonspecific.

To gain an overview of the selectivity of our indole compounds for cancer cells, we compared their induction of apoptosis in cancer cells and noncancer cells by using a caspase activation assay. We used hTERT-immortalized BJ human skin normal fibroblast cells, which are proven to be noncancerous (27) but are sensitive to nonspecific antitumor agents (28). For cancer cells, we chose MDA-MB231 breast cancer cells, whose Wnt signaling is up-regulated in an autocrine manner (4), and HepG2 hepatocellular carcinoma cells, which have high intrinsic Tcf-mediated transcription activity (29). All compounds tested except compound 28 showed little or no caspase up-regulation in the BJ cells, modest caspase up-regulation in the HepG2 cells, and significant up-regulation in the MDA-MB231 cells (Fig. 3), suggesting that the apoptotic effect of the chemical library generated from this scaffold is generally selective for cancer cells. HepG2 cells were more resistant to apoptosis than MDA-MB231 presumably because the high Tcf transcription activity in HepG2 is believed to be at least partially due to mutations of β-catenin, which is downstream of Dvl (30), whereas the Tcf-mediated transcription in MDA-MB231 reflects autocrine Wnt up-regulation that is upstream of, and therefore dependent on, Dvl.

Finally, we selected compounds for testing in vivo. Compounds active in our Tcf down-regulation assay are believed to directly target Dvl1 in cancer cells because our assay measured Tcf-mediated transcriptional activity stimulated by exogenous Dvl1. Therefore, these compounds are expected to be active against tumors whose Tcf-mediated transcription is up-regulated by Wnt up-regulation (4) or by overexpression of Dvl1 (7, 8). We selected the compounds that down-regulated Tcf-mediated transcription (Fig. 2) and evaluated their induction of apoptosis in Dvl-overexpressing cancer cells. Previously, we found that Dvl1⁴

and Dvl3 (23) proteins are overexpressed in H460 non–small cell lung cancer cells but not in normal lung or skin (7) and that apoptosis of H460 cells is induced by inhibition of Wnt activity (23). H460 cells were incubated with the compounds, and their apoptotic population was then analyzed by Annexin V staining. We anticipated that apoptotic population after longer exposure to a lower drug concentration (10 μmol/L) would reflect in vivo efficacy more accurately than activation of caspases after a shorter exposure to a high concentration because in vivo drug concentration in tumors is not expected to be high. Figure 4 shows induction of apoptosis in relation to down-regulation of Tcf transcriptional activity. Compounds 9, 11, and 24 showed high activity in both H460 apoptosis induction and Tcf down-regulation. Compound 28 showed high activity in induction of apoptosis but not in Tcf down-regulation, suggesting that its apoptosis induction is unrelated to Tcf down-regulation.

Finally, we tested the in vivo efficacy of two compounds (24 and 28) that most potently induced apoptosis of the H460 cells (Fig. 4, arrows). H460 cells were transplanted into nude mice and allowed to grow for 7 days. Compounds 24 and 28 were then administered daily for 14 days. The mice showed no significant weight loss (<5%; data not shown). Compound 24 exerted a statistically significant growth-suppressive effect (P = 0.002; Fig. 5) that was greater than the effects of compound 28 (P = 0.082) and FJ9 (P = 0.02; ref. 25).

In this study, we showed a strategy useful for the rational generation of novel agents to down-regulate Dvl1-driven Tcf transcriptional activation. The initial stage was to produce chemical libraries and screen them for biochemical antagonism of Dvl PDZ interactions and for Tcf transcriptional activation. As observed in this study, there is often some discrepancy between biochemical potency and the cellular pharmacologic effects of chemical agents because the latter is affected by other factors, such as solubility, stability in cultured cells, and cell membrane permeability. Drug discovery programs have recently tended to collect all of these data for all biochemical hits, a practice that is enormously costly. To streamline this process, our strategy incorporated a second stage designed to straightforwardly identify bona fide lead compounds that show both biochemical potency and cellular pharmacologic effects. Specifically, we selected agents that inhibited both Tcf-mediated transcription and the biochemical interactions of Dvl and that selectively induced apoptosis in cancer cells.

We tested two lead compounds, 24 and 28, in a mouse xenograft model. Compound 24 was highly active in both apoptosis induction and down-regulation of Tcf transcription, whereas 28 was highly active only in apoptosis induction (Fig. 4). The effects of compound 28 were also less specific to cancer cells, as it activated caspases in BJ cells (Fig. 3). These observations suggest that 24 may act as an antitumor agent by selectively down-regulating Tcf-mediated transcription, whereas the effects of 28 may not be due solely to Tcf down-regulation. Therefore, if tumor growth is Dvl1 dependent, then Tcf-specific compound 24 is likely to suppress that growth more efficiently. Another possibility is that the indole-2-carbinol 28 is less metabolically stable in mice than the indole-2-ketone 24. In general, 2-carbinol compounds tended to show weaker potency in inducing apoptosis than in down-regulating Tcf-mediated transcription (Fig. 4). This effect may be due to greater metabolic instability of the 2-carbinol moiety in the cancer cell cultures during the 4-day apoptosis assay than during the 1-day Tcf assay.

Our compounds are designed to selectively antagonize the Dvl1-Tcf pathway in cells. New mechanism-based antitumor drugs targeting the Dvl-Tcf signaling pathway should be effective against cancers in which Tcf signaling is up-regulated and should exert dramatic synergism with existing cytotoxic chemotherapy agents. We will use the method described in this study to generate the next lead compounds, which will possess further improved potency, efficacy, and selectivity. An inhibitor of the canonical Wnt pathway was also recently reported to be useful in promoting adipogenesis (31). Therefore, our approach will improve the development not only of new antitumor agents but also of new antidiabetic agents. Efforts in both directions are under way.

No potential conflicts of interest were disclosed.

We thank Randall T. Moon (University of Washington) for the generous gift of superTOPflash and superFOPflash plasmids, Michele Connelly for maintaining cell cultures and technical support, and Sharon Naron for editorial advice.