Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2010 Mar 12.
Published in final edited form as: J Med Chem. 2009 Mar 12;52(5):1247–1250. doi: 10.1021/jm801278g

Discovery of Novel Myc-Max Heterodimer Disruptors with a 3-Dimensional Pharmacophore Model

Gabriela Mustata 1,✉, Ariele Viacava Follis 2, Dalia I Hammoudeh 2, Steven J Metallo 2, Huabo Wang 3, Edward V Prochownik 3, John S Lazo 4, Ivet Bahar 1
PMCID: PMC2765518  NIHMSID: NIHMS95813  PMID: 19215087

Abstract

A 3-dimensional pharmacophore model was generated utilizing a set of known inhibitors of c-Myc-Max heterodimer formation. The model successfully identified a set of structurally diverse compounds with potential inhibitory activity against c-Myc. Nine compounds were tested in vitro, and four displayed affinities in the µM range and growth inhibitory activity against c-Myc-overexpressing cells. These studies demonstrate the applicability of pharmacophore modeling to the identification of novel and potentially more puissant inhibitors of the c-Myc oncoprotein.

De-regulation of the c-Myc oncogene is among the most frequent molecular abnormalities encountered in human cancer, and is often associated with aggressive tumors, of the breast, colon, cervix, lung, and hematopoietic organs.1,2 c-Myc is a member of the basic helix-loop-helix leucine zipper protein family (bHLH-ZIPa) whose dimerization with another bHLH-ZIP protein, Max, is necessary for various biological activities, including cellular transformation, apoptosis, and transcriptional activation.3–6 The fact that the oncogenic activity of c-Myc depends on its dimerization with Max makes the c-Myc-Max heterodimer not only an enticing target for drug design but also an important case study for the challenge of designing small molecule inhibitors of protein-protein interactions.7,8

Small molecules that specifically bind c-Myc and prevent c-Myc-Max heterodimerization have already been discovered.9–13 Recent results from our group have provided NMR-based models showing that structurally unrelated inhibitors bind to distinct regions of the intrinsically unordered c-Myc monomer and alter its conformation so as to render it incapable of interacting with Max.14

Given the availability of the activity data for several c-Myc-Max heterodimer disruptors,12,13 we decided to exploit this information to develop a molecule-derived pharmacophore model that would capture the primary chemical features common to these compounds. This is a powerful method for finding novel ligands, and has been used extensively in drug discovery research in both academia and pharmaceutical industry.15,16 Herein, we utilize GALAHAD (Genetic Algorithm with Linear Assignment for Hypermolecular Alignment of Datasets),17–19 implemented in SYBYL 8.020, a recently developed pharmacophore modeling program that allows for full ligand flexibility while taking strain energy and steric overlap into account.

The dataset used in these studies contains six c-Myc-Max heterodimer inhibitors identified previously by our group,12,13 (Figure 1a): the parental compound 10058-F4 (1)13 and five of its derivatives.12 All compounds bind to the same c-Myc region centered around residues Y402-K412,14 with affinities that were up to 6-8-fold greater than that of 1.12,13 Twenty models were produced in our study, which differed somewhat in the number and type of features, and in the conformations and overlay of the molecules. The pharmacophore model with the best overall score is displayed in Figure 2. It contains two hydrophobic features (yellow), one donor atom (blue), and two acceptor atoms (red).

Figure 1.

Figure 1

Open in a new tab

Molecules used in pharmacophore development: (a) active molecules used in the pharmacophore model generation with GALAHAD;18 (b) inactive molecules used in the model refinement stage with TUPLETS.20

Figure 2.

Figure 2

Open in a new tab

GALAHAD model obtained from six compounds in the biological data set includes two hydrophobes (yellow), one donor atom (blue), and two acceptor atoms (red). The sphere sizes indicate query tolerances.

The uniqueness of our approach is that the pharmacophore model generated by GALAHAD was further refined using two inactive analogs of 1 (Figure 1b). The refinement stage was performed using the Tuplets module in SYBYL 8.0,20 which allows for the decomposition of the full pharmacophoric pattern found for each inactive ligand into its constituent distance multiplets, that are encoded into a vector fingerprint. Compounds retrieved in this way are often of a different chemical class than those used to generate the database query, demonstrating lead-hopping capability.

For validation purposes, the Tuplets refined model was used to query a test set of 10 compounds containing 6 active and 4 inactive analogs of 1 (see Supporting Information) through a hierarchical clustering. The active compounds bind to the same c-Myc region centered around residues Y402-K412,13 with affinities that were up to 6-8-fold greater than that of 1.11,12 All molecules performed acceptably well, with all actives compounds, except one, clustered together in the dendrogram, whereas the inactive ones were distributed. One of the inactive compounds was clustered together with the active compounds.

The resulting Tuplets hypothesis was further used to screen the ZINC 7.0 database21 for drug-like molecules (~5 × 106 compounds) that were sufficiently similar to the selected hypothesis. Note that the search database was translated into multi-conformer Tuplets of the same type as the generated hypothesis. The hypothesis captured 15,822 hits (0.31% of database). The hits included a structurally diverse set of compounds as measured by their Tanimoto score of 0.5. This number progressively decreases with increasing Tanimoto similarity (e.g., 274 hits for 0.80 cutoff). Our choice of a Tanimoto cutoff of 0.5 is motivated by findings that similar biological activities may be shared by compounds that exhibit relatively weak structural similarities.22–24 Although, 2D similarity measures may overlook important structural/functional features, and 3D metrics may retrieve compounds with more diverse topology, at the initial screening stage, 2D metrics are conveniently used for a rapid way of finding new lead compound. The top 100 compounds, were filtered further for desirable ADME properties with ADME Boxes v4.0 software25,26 to rationally deconvolute the large number of compounds that resulted from the initial database screening process, with the understanding that those compounds could potentially serve as drugs. The top-ranking 30 compounds were selected as potential candidates for the design of novel c-Myc inhibitors.

Given the extensive metabolism and rapid clearance of the parental compound 1,27 we selected for experimental testing those compounds that were predicted to be less metabolically labile. Considering that cytochrome P450 isoform CYP3A4 is the major enzyme responsible for xenobiotic metabolism in human organism, and metabolizes >50% of drugs,28 we used ToxBoxes v2.929 to select compounds with the lowest predicted probability of being a CYP3A4 inhibitor at clinically relevant concentrations (Ki < 50 mM). Nine ZINC 7.0 compounds (Figure 3) were finally purchased from ChemBridge Corporation and tested in vitro for disruption of c-Myc-Max(S) heterodimer formation.

Figure 3.

Figure 3

Open in a new tab

Compounds selected for experimental testing as potential c-Myc inhibitors (ZINC 7.0 and ChemBridge database numbering is indicated, as well as the CAS number).

The experiments were performed as reported in our previous work12 and are also available as Supporting Information. The compounds were screened in a circular dichroism assay where the helical content of equimolar mixtures of c-Myc and Max (1.5 µM) was determined from ellipticity measurements at 222 nm. Compounds capable of disrupting the protein dimer cause a decrease in helical content, as the isolated monomers are disordered and flexible. Initial screening of the nine compounds at one single high concentration (200 µM) indicated a nearly complete disruption of c-Myc-Max dimers being induced by four molecules, and partial disruption by another three (Figure 4). Only two compounds proved to be entirely inactive in this assay. These results support the reliability of the computational model in predicting active inhibitors.

Figure 4.

Figure 4

Open in a new tab

Disruption of 1.5 µM c-Myc-Max dimer by 200 µM concentration of each tested inhibitor measured by circular dichroism.14 Data represent the average of three independent trials (error bars represent the standard error).

The four compounds that exhibited the highest disruptive ability against c-Myc-Max(S) heterodimer formation at 200 µM concentration were further tested over a range of concentrations providing a full titration of the protein dimer disruption. Multiplication of the competition constant (Kcomp), employed to fit the experimental data by the independently determined dissociation constant of c-Myc-Max dimers provided an estimate of the inhibitors affinity for c-Myc monomers (assuming that, like the set of compounds employed to generate the pharmacophore model, they interact exclusively with this protein monomer). The compounds displayed affinities in the mid µM range, generally 2–10 times lower than that of 1 and structurally related compounds employed to generate the pharmacophore model (Table 1).

Table 1.

Micromolar affinities of the best tested inhibitors for c-Myc as estimated by disruption of c-Myc-Max dimers and competition against the parent inhibitor 1 for direct binding to monomeric c-Myc. Error ranges are indicated in parentheses. All the experiments were performed in triplicate.

Compound c-Myc-Max
Disruption
(µM)		 Competition
against 1
(µM)
230 10 (3) 2.5 (0.5)
331 25 (3) 40 (10)
4 19 (6) 18 (6)
5 45 (12) 70 (20)
Open in a new tab

The four compounds were tested further for direct competition to confirm that they interact directly with the c-Myc monomer in a mode similar to that displayed by 1. The compounds were tested by monitoring the fluorescence polarization of 1, which is inversely proportional to the compound’s tumbling rate in solution, and increases when it is bound to the relatively large c-Myc (353–437) monomer. Under these conditions, all four compounds were able to displace 1 from c-Myc. A titration competition was performed for each compound and data were fit in a similar way to the one described for the titrations of c-Myc-Max(S) disruption. Kcomp in this case was multiplied by the dissociation constant of the complex between 1 and c-Myc to provide an estimate of the affinity of the tested compounds for c-Myc (353–437) binding. The obtained values were in reasonable agreement with the estimates of binding affinity obtained from the titrations of c-Myc-Max(S) dimer disruption. The two best compounds were also tested in an electrophoretic mobility shifts assay (EMSA) for disruption of DNA binding by c-Myc-Max(S) dimers, showing an inhibitory efficacy comparable to that of 1 (Figure 5).12

Figure 5.

Figure 5

Open in a new tab

a. Disruption of E-Box DNA binding by c-Myc-Max dimer by the two newly identified inhibitors with the highest binding affinity to c-Myc − 2 and 4– at 200 µM concentration. b. Quantitative assessment of disruption of c-Myc-Max DNA binding for the parent compound 1 (white bars), 2 (black bars) and 4 (grey bars). Data represent the average of three independent trials (error bars represent standard error).

All nine compounds were tested in HL60 cells as described in our previous work,12 and also included as Supporting Information. As shown in Figure 6, compounds 5360134 (5) and 6370870 (6) proved to be significantly more active, with IC50s of 23 and 16.7µmol, as compared to 35 µmol for the parental compound 1. The lack of exact correlation between the growth inhibitory effects of these compounds and their abilities to interact with c-Myc and disrupt c-Myc-Max association likely reflects the more complex nature of the cell-based assay, which requires uptake and retention of the compounds, their transport to the nucleus, and sufficient intracellular stability over the several day time-span of the assay. Both compounds, 5 and 6 were tested with HL60 cells, with TGR1 (normal rat fibroblasts) along with TGR1 knockout cells with over-expressed HMGA1b (KO+HMG). These latter cells lacked c-Myc due to gene targeting; over-expression of the HMGAIb restored a normal growth rate in a c-Myc-independent manner.32 Our results demonstrated very good inhibition in HL60 cells with both ZINC compounds, and appeared to be somewhat selective in cells that expressed higher levels of c-Myc (HL60s) (see Supporting Information). They exerted the least effect on the KO+HMG cells, thus revealed a direct correlation between c-Myc levels and growth inhibition by these compounds. Further evidence for specificity came from the finding that compound 5 seemed to be more selective for HL60s than 6. From these studies, we concluded that the ability of both ZINC compounds to inhibit the growth of mammalian cells is c-Myc dependent. These compounds were well within the range of what was seen when we screened a large number of 1 analogs.12

Figure 6.

Figure 6

Open in a new tab

Dose-response profiles of compounds 1, 5 and 6 on HL60 cell growth. IC50s were calculated based on dose-response profiles on day 5 following the addition of each compound.

We recently identified the binding site and provided a model of the interaction between the parental compound, 1, and c-Myc.14 The c-Myc-Max disruption assays along with the competition assays clearly show that the active compounds described here bind in the same region as 1, residues Y402-K412 of c-Myc. These compounds disrupt the formation of the highly ordered c-Myc-Max dimer by binding and stabilizing the intrinsically disordered monomer of c-Myc. NMR based studies of 1 binding to c-Myc demonstrated clear NOE signals with the binding site but the overall flexibility of the disordered target resulted in insufficient NOE data to generate a standard structural model.14 Disordered regions are over represented in disease related protein interactions; the ligand-based pharmacophore approach may be of especial importance in the search for inhibitors of these proteins.33

This is the first report of a pharmacophore model that provides a hypothetical picture of the main chemical features responsible for the activity of c-Myc-Max heterodimer disruptors that may prove to be useful for the future development of more potent analogs based on rational design. The newly identified lead compounds exhibit novel chemical scaffolds, and will be further optimized to enhance their inhibitory activity.

Supplementary Material

1_si_001. Supporting Information Available.

Details of Pharmacophore model generation, refinement and validation; Summary of HPLC purity and NMR data for the tested compounds; Expression and Purification of Recombinant c-Myc-353–437 and Max; Screening of c-Myc-Max dimer disruption; Competition assay against 1 for c-Myc353–437 binding; Electrophoretic Mobility Shifts Assays (EMSA); Dose response experiments; Cell-based assay. This material is available free of charge via the Internet at http://pubs.acs.org.

Acknowledgment

Support by NIH grant 1U54MK074411 is gratefully acknowledged by JSL and IB. GM is grateful to Ahmet Bakan and Dr. Gunther Stahl from Tripos International for valuable assistance and fruitful discussions.

Footnotes

a

Abbreviations: bHLH-ZIP, basic helix-loop-helix leucine zipper; GALAHAD, Genetic Algorithm with Linear Assignment for Hypermolecular Alignment of Datasets; EMSA, Electrophoretic Mobility Shifts Assay

REFERENCES

  • 1.Nesbit CE, Grove LE, Yin X, Prochownik EV. Differential apoptotic behaviors of c-Myc, N-Myc, and L-Myc oncoproteins. Cell Growth Differ. 1998;9:731–741. [PubMed] [Google Scholar]
  • 2.Soucek L, Whitfield J, Martins CP, Finch AJ, Murphy DJ, Sodir NM, Karnezis AN, Swigart LB, Nasi S, Evan GI. Modelling Myc inhibition as a cancer therapy. Nature. 2008;455:679–683. doi: 10.1038/nature07260. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Amati B, Dalton S, Brooks MW, Littlewood TD, Evan GI, Land H. Transcriptional activation by the human c-Myc oncoprotein in yeast requires interaction with Max. Nature. 1992;359:423–426. doi: 10.1038/359423a0. [DOI] [PubMed] [Google Scholar]
  • 4.Amati B, Littlewood TD, Evan GI, Land H. The c-Myc protein induces cell cycle progression and apoptosis through dimerization with Max. EMBO J. 1993;12:5083–5087. doi: 10.1002/j.1460-2075.1993.tb06202.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Amati B, Brooks MW, Levy N, Littlewood TD, Evan GI, Land H. Oncogenic activity of the c-Myc protein requires dimerization with Max. Cell. 1993;72:233–245. doi: 10.1016/0092-8674(93)90663-b. [DOI] [PubMed] [Google Scholar]
  • 6.Amati B. Integrating Myc and TGF-beta signalling in cell-cycle control. Nat. Cell Biol. 2001;3:E112–E113. doi: 10.1038/35074634. [DOI] [PubMed] [Google Scholar]
  • 7.Wells JA, McClendon CL. Reaching for high-hanging fruit in drug discovery at protein-protein interfaces. Nature. 2007;450:1001–1009. doi: 10.1038/nature06526. [DOI] [PubMed] [Google Scholar]
  • 8.Xu Y, Lu H, Kennedy JP, Yan X, McAllister LA, Yamamoto N, Moss JA, Boldt GE, Jiang S, Janda KD. Evaluation of “credit card” libraries for inhibition of HIV-1 gp41 fusogenic core formation. J. Comb. Chem. 2006;8:531–539. doi: 10.1021/cc0600167. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Berg T, Cohen SB, Desharnais J, Sonderegger C, Maslyar DJ, Goldberg J, Boger DL, Vogt PK. Small-molecule antagonists of Myc/Max dimerization inhibit Myc-induced transformation of chicken embryo fibroblasts. Proc. Natl. Acad. Sci. U. S. A. 2002;99:3830–3835. doi: 10.1073/pnas.062036999. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Kiessling A, Sperl B, Hollis A, Eick D, Berg T. Selective inhibition of c-Myc/Max dimerization and DNA binding by small molecules. Chem Biol. 2006;13:745–751. doi: 10.1016/j.chembiol.2006.05.011. [DOI] [PubMed] [Google Scholar]
  • 11.Prochownik EV. c-Myc as a therapeutic target in cancer. Expert. Rev. Anticancer Ther. 2004;4:289–302. doi: 10.1586/14737140.4.2.289. [DOI] [PubMed] [Google Scholar]
  • 12.Wang H, Hammoudeh DI, Follis AV, Reese BE, Lazo JS, Metallo SJ, Prochownik EV. Improved low molecular weight Myc-Max inhibitors. Mol. Cancer Ther. 2007;6:2399–2408. doi: 10.1158/1535-7163.MCT-07-0005. [DOI] [PubMed] [Google Scholar]
  • 13.Yin X, Giap C, Lazo JS, Prochownik EV. Low molecular weight inhibitors of Myc-Max interaction and function. Oncogene. 2003;22:6151–6159. doi: 10.1038/sj.onc.1206641. [DOI] [PubMed] [Google Scholar]
  • 14.Follis AV, Hammoudeh DI, Wang H, Prochownik EV, Metallo SJ. Binding of small-molecule inhibitors to local sequence sites on the intrinsically disordered c-Myc protein. Chem Biol. 2008 In Press. [Google Scholar]
  • 15.La Jolla: International University Line Biotechnology Series; Pharmacophore Perception, Development, and Use in Drug Design. 2000:1–531.
  • 16.Pharmacophores and Pharmacophore Searches. Weinheim Germany: Wiley-VCH Verlag GmbH & Co. KGaA; 2008. pp. 1–365. [Google Scholar]
  • 17.Clark RD, Abrahamian E. Using a staged multi-objective optimization approach to find selective pharmacophore models. J. Comput. Aided Mol. Des. 2008 doi: 10.1007/s10822-008-9227-2. [DOI] [PubMed] [Google Scholar]
  • 18.Richmond NJ, Abrams CA, Wolohan PR, Abrahamian E, Willett P, Clark RD. GALAHAD: 1. pharmacophore identification by hypermolecular alignment of ligands in 3D. J. Comput. Aided Mol. Des. 2006;20:567–587. doi: 10.1007/s10822-006-9082-y. [DOI] [PubMed] [Google Scholar]
  • 19.Shepphird JK, Clark RD. A marriage made in torsional space: using GALAHAD models to drive pharmacophore multiplet searches. J. Comput. Aided Mol. Des. 2006;20:763–771. doi: 10.1007/s10822-006-9070-2. [DOI] [PubMed] [Google Scholar]
  • 20.SYBYL 8.0. www.tripos.com.
  • 21.Irwin JJ, Shoichet BK. ZINC--a free database of commercially available compounds for virtual screening. J. Chem. Inf. Model. 2005;45:177–182. doi: 10.1021/ci049714. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Martin YC, Kofron JL, Traphagen LM. Do structurally similar molecules have similar biological activity? J. Med. Chem. 2002;45:4350–4358. doi: 10.1021/jm020155c. [DOI] [PubMed] [Google Scholar]
  • 23.Wallqvist A, Huang R, Thanki N, Covell DG. Evaluating chemical structure similarity as an indicator of cellular growth inhibition. J. Chem Inf. Model. 2006;46:430–437. doi: 10.1021/ci0501544. [DOI] [PubMed] [Google Scholar]
  • 24.Covell DG, Huang R, Wallqvist A. Anticancer medicines in development: assessment of bioactivity profiles within the National Cancer Institute anticancer screening data. Mol. Cancer. Ther. 2007;6:2261–2269. doi: 10.1158/1535-7163.MCT-06-0787. [DOI] [PubMed] [Google Scholar]
  • 25.ADME Boxes v4.0. http://pharma-algorithms.com/
  • 26.Japertas P, Didziapetris R, Petrauskas A. Fragmental methods in the analysis of biological activities of diverse compound sets. Mini. Rev. Med. Chem. 2003;3:797–808. doi: 10.2174/1389557033487601. [DOI] [PubMed] [Google Scholar]
  • 27.Guo J, Parise RA, Joseph E, Egorin MJ, Lazo JS, Prochownik EV, Eiseman JL. Efficacy, pharmacokinetics, tisssue distribution, and metabolism of the Myc-Max disruptor, 10058-F4 [Z,E]-5-[4-ethylbenzylidine]-2-thioxothiazolidin-4-one, in mice. Cancer Chemother. Pharmacol. 2008 doi: 10.1007/s00280-008-0774-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Wrighton SA, Schuetz EG, Thummel KE, Shen DD, Korzekwa KR, Watkins PB. The human CYP3A subfamily: practical considerations. Drug Metab Rev. 2000;32:339–361. doi: 10.1081/dmr-100102338. [DOI] [PubMed] [Google Scholar]
  • 29.ToxBoxes v2.0. http://pharma-algorithms.com/
  • 30.Schwyzer R, Feurer M, Iselin B. Activated esters. III. Reactions of activated esters of amino acid and peptide derivatives with amines and amino acid esters. Helvetica Chimica Acta. 1955;38:83–91. [Google Scholar]
  • 31.Watanabe S, Ogawa K, Ohno T, Yano S, Yamada H, Shirasaka T. Synthesis of 4-[1-(substituted phenyl)-2-oxo-pyrrolidin-4-yl]methyloxybenzoic acids and related compounds, and their inhibitory capacities toward fatty-acid and sterol biosynthesis. European Journal of Medicinal Chemistry. 1994;29(9):675–686. [Google Scholar]
  • 32.Rothermund K, Rogulski K, Fernandes E, Whiting A, Sedivy J, Pu L, Prochownik EV. C-Myc-independent restoration of multiple phenotypes by two C-Myc target genes with overlapping functions. Cancer Res. 2005;65(6):2097–2107. doi: 10.1158/0008-5472.CAN-04-2928. [DOI] [PubMed] [Google Scholar]
  • 33.Uversky VN, Oldfield CJ, Dunker AK. Intrinsically disordered proteins in human diseases: introducing the D2 concept. Annu. Rev. Biophys. 2008;37:215–246. doi: 10.1146/annurev.biophys.37.032807.125924. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1_si_001. Supporting Information Available.

Details of Pharmacophore model generation, refinement and validation; Summary of HPLC purity and NMR data for the tested compounds; Expression and Purification of Recombinant c-Myc-353–437 and Max; Screening of c-Myc-Max dimer disruption; Competition assay against 1 for c-Myc353–437 binding; Electrophoretic Mobility Shifts Assays (EMSA); Dose response experiments; Cell-based assay. This material is available free of charge via the Internet at http://pubs.acs.org.

NIHMS95813-supplement-1_si_001.pdf (556.3KB, pdf)

RESOURCES